Quick Links

• Where we started

• Goals and Objectives

  • Needs and wants

  • Power station metrics

  • Strategies and options

• Finding a power station

  • Anker Solix

  • Ecoflow

  • Bluetti

  • A point about standalone batteries

  • Picking a power station

• Now, theorycrafting

• About ROI...

• My current setup

• Milestone

Where we started

When I rebuilt my network cart a few years ago, I had originally planned for having two UPS units. I had hoped to deploy a maintenance-bypass ($$) or automatic-transfer switch ($$$) so I could actually do staggered testing of the units without having to first power everything down. Powering down everything beforehand is a necessity - should the UPS fail the test, it would actually drop everything it was meant to power up.

A dual-UPS setup (specifically with an ATS) has the additional benefit of [nearly] doubling the runtime. According to the manufacturer, with my network cart pulling 300W, I should expect somewhere in the range of 20-25 minutes however, real-world outages are slightly less optimistic — definitely less than 20 minutes. Considering the fact that it it takes a non-insignificant (in reality, a substantial) amount of time for my NAS to actually power down once the determination has been made, I realistically plan for having a measly 7 minutes before things get exciting.

Even hypothetically doubling that runtime doesn’t exactly give me a ton of breathing room.

Currently, I do have an EGO Power+ Nexus Portable Power Station and four 5Ah batteries: a cumulative 1120Wh of power. This can (and has!) easily kept the network cart running for over two hours.

So, when the power goes out for more than a blip, I need to do a mental evaluation of whether this will end up being an extended outage or not. Once the determination has been made to transition over to this separate power station, the process is:

1. Grab the EGO power station from the garage and bring them to the office

2. Grab four batteries, bring them to the office and plug it in

3. Power up the power station

4. Move the network cart over to the EGO power station

Luckily I have been home for every major power outage, so I just handle this — it’s definitely doable within 7 minutes, but there certainly isn’t a lot of wiggle room.

Last fall, I got around to deploying a small proof of concept for using solar/batteries to peak-shave a work laptop. Even with the reduced solar generation over the winter months, for such a small draw (mostly averaging 30W), the power station has easily been able to peak- (and mid-) shave the laptop for my entire work day without any intervention. It’s been great.

Thinking about how easily the EGO power station handles the load, I started thinking about deploying a dedicated power station for the office…

At a high level, I wanted some kind of battery system that I could use to carry some (or all) of the office electrical load for some (to be determined) amount of time. I wanted to address both backup-power and peak-shaving. There was a bit of wiggle-room in the final requirements to account for cost, complexity, scalability, ease of use etc.

There are three broad solution categories:

1. An off the shelf, ready-to-go solution. You pay a premium for the convenience but you have to fit your needs inside the manufacturer’s product stack (and there is the risk of being locked in an ecosystem)

2. A fully DIY solution where you mix and match inverters, batteries, charge controllers etc. In the long-run, this ends up being a much more cost effective solution, but getting to the first (and possibly each) checkpoint costs more (absolutely) and brings with it a lot of complexity

3. A professionally installed, home-level system. This is something we’d like to consider down the line but definitely not at this time and definitely outside of the scope of what I want to do

Although the DIY route is definitely appealing and I think I will eventually get there, this isn’t my only project — having a power station is just a checkpoint on a much more complicated journey. For the time being, I need the simplicity of ‘open the box and go’.

[Back to Top]

Goals and Objectives

Needs and wants

Whatever system (or systems) I landed on, I absolutely wanted

• To have a “substantial” amount of runtime to the network cart. While I didn’t initially have a specific runtime I wanted to hit, after seeing that my EGO power station could run the network cart for 2 hours, I set that as my starting target (as this would be largely influenced by cost)

• To be able to peak-shave as much as a I can (and to do it robustly, without minimal fiddling on automations)

Dependant on pricing, system complexity etc., I wanted to extend both of the previous goals to the computers in my office

Looking at the on-peak windows for the above pie charts:

• Winter: I need to be able to carry the office for 4 hours in the morning (0700-1100). There’s a 6-hour window mid-day where I can either solar-charge (winter solar will be a challenge) or AC-charge (at a lower rate). My experience in the winter will largely be determined by the battery capacity.

• Summer: here, I need to be able to carry the office for a continuous 6-hours (1100-1700) but summer solar generation is more promising; my experience in the summer will largely be determined by solar input.

[Back to Top]

Power station metrics

1. Inverter size. This is measured in watts and the metric sets the limit on what you can do. For example, something like a hair dryer might need 1500W to operate — if you have a 200W inverter, you’ll never be able to run it.

2. Battery capacity. Now that you can run whatever it is you’re trying to run, this determines how long you can run it for. Measured in watt-hours, and continuing our example,
   

   • If your battery has a capacity of 500Wh, then you could run your 1500W hair dryer for 1/3rd of an hour (20-minutes)

   • If your battery capacity was 1500Wh, then you could run it for 1 hour

   • If your battery capacity was 3000Wh, then you could run it for 2 hours

3. Solar input capacity. Measured in watts, this expresses how fast you can recharge your battery capacity (even while actively draining it). By focusing on solar input, we are looking at how fast (or how much) we can recharge the battery for free. Indirectly, this value affects how long you can run things for. Assuming we have a 500Wh battery and continuing our example:

   • If we have 500W of solar-input, then our 1500W hair dryer effectively only pulls 1000W, extending our runtime from 1/3 hour to 1/2 hour

   • If we have 1000W of solar-input, then our hair dryer effectively only pulls 500W, extending our runtime from 1/3 hour to 1 hour

   • If we have 1500W of solar-input, then our hair dryer effectively pulls 0W, meaning we can run indefinitely (while the sun is shining of course)

As I alluded to previously, there is a bit of a ‘soft-limit’ on battery size determined by:

(1440W - Continuous Draw) x 12-hours (of off-peak)

or

(1440W - Continuous Draw) x 18-hours (of off-peak + mid-peak)

With zero draw, this gives us a (silly) hypothetical peak of 17.3-25.9kWh for battery capacity - any larger and we wouldn’t (really) be able to recharge it within a 24-hour period on a typical 15A residential circuit. Note that for safety purposes, even though a 15A circuit can pull 1800W, we shouldn’t be doing so continuously.

More importantly though — my draw is never going to be zero. But can come back to this later when we go down the theory-crafting rabbit hole.

[Back to Top]

Strategies and options

Option 1

I would get a smaller power station exclusively for the network cart; I can integrate it into the cart directly and this would be the end of the project. This gives me a self-contained starting point that fulfills my core requirement of giving the cart extended runtime and peak-shaving.
For the main computers in the office, I could follow up with a more substantial setup (either DIY or a larger off-the-shelf solution). The downside with this option is that, down the road, there will be two separate systems to manage and I would want to try and find a way to charge both systems with a single solar array.

Option 2

Here, I get substantially larger power station — with enough oomph to drive everything I need. In addition to the higher initial cost, I would need to accept that initially, my runtime may be sub-optimal — since I’d be loading the everything in the office on battery power.

I would need to limit myself to models that support expansion batteries (either officially or DIY) and I would have to prioritize power-stations that more usable solar input (as there is more draw to try and offset).

For both strategies, I wanted to tie this into Home Assistant - with or without official support (but obviously, platforms with official integration support would be prioritized). Ultimately, despite the jump in price, it made a bit more sense to go with Option 2 — this way I only have one system to worry about maintaining, integrating and automating.

[Back to Top]

Finding a power station

I definitely wanted a ‘open the box and go’ solution (and recognized that I would be paying a premium for that). One advantage of going this route is having access to operating modes out of the box, without having to further do custom automation work. Specifically, I was looking to do some level of basic time-of-use/peak-shaving natively with whatever app the power station came with.

Anker Solix

Since I already have an Anker Solix C300, I naturally started my search with Anker products. When I initially looked into this, there was an unofficial integration to get Solix power stations into Home Assistant. Since then, Anker now has an official integration. At this time, there’s only support for a single product, so the official route is something to keep an eye on.

• [$] C1000 Gen 2

  This is the smallest, most cost-effective power station of the bunch. It technically has enough oomph to handle the entire office but only just.

  • Inverter: 2000W. My nominal load for the office is 1100W so this can carry everything

  • Battery capacity: 1024Wh. Without considering solar, I have less than 1-hour of battery-based runtime — not enough to meaningfully do any peak-shaving. Since the Gen 2 model doesn’t support expansion batteries, I’m at a dead end here.

  • Solar Input: 600W. Even with maximum solar input, with my 1100W load, I’m draining my battery at a rate of 500W. This puts my hypothetical best case runtime at roughly 2 hours.

  • Price point: ~$600

  Even if I were to use a standalone battery to act as faux-server, the built-in battery will continue to drain — and once the built-in battery drains down to empty, the inverter will shut off.

• [$$] C2000 Gen 2

  Bigger in every way compared to the C1000G2, this unit realistically represents the smallest offering I would consider.

  • Inverter: 2400W. With a [slightly] bigger inverter, my hope is that, because my nominal load represents a lower utilization (of the inverter), there’s less noise in the office (fans, inverter whine etc.). Unless the noise is extreme, this is just a nice perk.

  • Battery Capacity: 2048Wh. Doubling the battery capacity is only half the equation: this unit support expansion batteries as well. Out of the box this gives me just under 2-hours of runtime

  • Solar input: 800W. Although it’s a small bump in the solar rating, it does have a meaningful impact on my runtime. With a nominal 1100W load, the (peak) 800W solar input makes the effective drain only 300W. This gives me a runtime of roughly 6 hours.

  • Price point: ~$1000

  To actually hit 800W is a different challenge: I would probably do this via standalone battery (at additional cost) acting as faux-solar. But at least it’s possible. Like the C1000G2 though, with a 1100W load and only 800W solar input, I will eventually run the battery down and then the inverter cuts off.

• [$$$] F3000

  This is the biggest/most-expensive power station I wanted to consider. This does ‘all of the above’ and more.

  • Inverter: 3600W. For the office, I definitely don’t need the extra oomph here, but my hope is that the fans would be minimal

  • Battery capacity: 3072Wh. Out of the box I’m looking at roughly 3-hours of run time (for the entire office, no less). With the tons of options for battery expansion this power station has a lot of points for scalability (if you have the money to spend)

  • Solar input: 2400W. Here we finally start getting into some seriously big numbers. Here I can (theoretically) run the office entirely on solar.

  • Price point: ~$2000

  The F3000 brings a bit more to the table too (at additional cost).

  • Multiple units can be paired together for 240V power

  • There’s more support for directly tying into your breaker-box to act as an automatic peak-shaving transfer-switch

I was really torn between between the C2000G2 and the F3000. For what I wanted to do, the C2000G2 was a near-perfect fit — the only limiting factor was the solar input. The F3000 is clearly a much more capable unit - but to unlock that extra solar, I was paying for a lot of [awesome] features I wasn’t going to use; also, some of the awesome features (like 240V/home-transfer switch) required buying an additional F3000 to capitalize on…

[Back to Top]

Ecoflow

Ecoflow was one of the two brands I wanted to look into. They seem to be positioned as a premium, tier-1 power station option. The downside? Their entire 3rd Gen product stack is a disaster. I did some spot-checks for some models but there was no way I was going to deep-dive into their product stack (and there was no way I was going to spend a ton of money without doing a deep dive). Just for this post, I did a very quick blast through the three attributes I care about and loosely grouped things (without considering price) by solar input:

• Minimal solar

  • Delta 3 1000 Air: 500W, 960Wh, 500W solar

  • Delta 3 Classic: 1800W, 1024Wh, 500W solar

  • Delta 3 1500: 1800W, 1536Wh (expandable), 500W solar

  • Delta 3 Max: 2400W, 2048Wh, 500W solar

• Medium Solar

  • Delta 3 Plus: 1800W, 1024Wh (expandable), 1000W solar

  • Delta 3 Ultra: 3600W, 3072Wh (expandable), 800W solar

  • Delta 3 Max Plus: 3000W, 2048Wh (expandable), 1000W solar

• Plenty

  • Delta 3 Ultra Plus: 3600W, 3072Wh (expandable), 1600W solar

  • Delta Pro 3: 4000W, 4096Wh (expandable), 2600W solar

• Crazy

  • Delta Pro Ultra: 7200W, 6144Wh (expandable), 5600W solar

  • Delta Pro Ultra X: 12000W, 12288Wh (expandable), 10000W solar

As someone who was struggled to make sense of QNAP’s NAS offerings, I find the entire 3rd Gen EcoFlow product stack to be triggering.

From an automation perspective, there’s an unofficial Home Assistant integration.

[Back to Top]

Bluetti

After the disappointment that was the Ecoflow product stack, I was mostly resigned to going back to debating the C2000G2 vs F3000 but for completeness sake I decided to look at Bluetti. Until this point, I hadn’t considered Bluetti as my impression at the time was solidly Tier-2 vibes: “everything works but the polish isn’t there”. After looking at their product stack, I was refreshingly surprised…

One really compelling point in Bluetti’s favor is official support for Home Assistant. To make my life easier on the automation front, I specifically wanted to focus on models with Wi-Fi, which rules out the more affordable AC-line of power stations and mostly restrict myself to Elite (or more upmarket) models.

• [$] Elite 100 V2

  On paper, this appears to be similar to the C1000 Gen 2 but with a much more substantial 1000W solar input

  • Inverter: 1800W. While this is the smallest inverter, it can carry all my needs. The only concern is how hard the fans might run

  • Battery Capacity: 1024Wh. By itself, the battery capacity only affords me ~1hr so my only hope is too leverage the substantial solar to increase this.

  • Solar Input: 1000W. This is a massive amount of solar for such a tiny unit. At peak solar input, my runtime increases to ~10 hours.

  • Price Point: <$600

  Looking at just these core metrics, this is an easy win over the C1000G2. My only concern whether or not my inverter utilization% will cause any fans to run and how loud they might be.

  Since there is no way to expand the battery here, the most practical strategy would be to use a standalone battery and hook it up as a 1000W faux-solar. Doing things this way would give me nearly 10 hours theoretical run time for the whole office. Very compelling.

• [$$] Elite 200 V2

  Double the battery, and a bit bigger of an inverter, on paper this is better ’value’ than the 100V2. Compared to the C2000G2, has a 200W solar input advantage but loses the ability to expand the battery.

  • Inverter: 2600W

  • Battery Capacity: 2074Wh

  • Solar Input: 1000W

  • Price Point: $1000

  Again, using a standalone battery as a 1000W faux-solar setup, this can give my office a theoretical peak run time of 20 hours. Despite this being an absolutely bonkers runtime, this doesn’t seem like a compelling option for me (as 10-hours from the 100V2 is already plenty sufficient).

• [$$$] Elite 300

  Despite the existence of the 200V2, this power station feels like it would be a better option.

  • Inverter: 2400W

  • Battery Capacity: 3014Wh

  • Solar Input: 1200W

  • Price Point: $1700

  With this power station, I have the option (using a standalone battery configured as faux-solar) to run my office ‘indefinitely’ from solar. The asking price for this is a lot though.

• [$$$$] Apex 300

  W solar input

  • Inverter: 3840W

  • Battery Capacity: 2765Wh

  • Solar Input: 2400W / 6400W

  • Price Point: $1900

  Now this is an interesting option. The inverter on this power station is a tank and can do a ton more than I need at this time (but all within the wish-list of what I might want down the road).

  The super attractive point is the dual high-wattage solar charge controllers for a very respectable 2400W solar (out of the box). With a unique option to add an additional 4000W of solar input via an accessory, a power station like this could grow with me over time.

  Compared to the Elite 300, this is clearly the better strategic option to go for.

[Back to Top]

A point about standalone batteries

So far I’ve considered the idea of using a standalone-battery as faux-solar: the idea here is to connect the battery to the solar-input port on the power station. Under the hood, this is DC-DC power transfer — the power station distinguish (or care) that the power is coming from a standalone battery versus a solar panel. When doing this, the power station happily charges its internal battery using the ‘solar’, subject to solar-input limits. This is why I’ve focused so much on the solar-input of the power stations.

Something to note though: if the power station is net-draining (i.e., Load > Solar Input), then eventually the battery on the power station will hit zero — when this happens, the inverter on the power station will shut off, regardless of how much power is remaining on the standalone-battery. Since the standalone battery is providing ‘solar’, the power station will continue to charge and eventually it will turn back on. Assuming your load is consistently bigger than the solar input, you will enter a loop where the power station runs down, shuts off, charges back up, turns on, runs down, turns off…

Using the C1000 Gen 2 just as an example, it has

• 1024Wh battery

• 600W solar-input

Given my nominal 1100W draw, this means that I would have a constant drain of 500W on the battery which would yield a theoretical runtime of just over 2-hours.

Without any way to expand the battery, this is a hard limit no matter how big our standalone battery is.

There is one more thing we could do: double-dip the standalone battery with a separate inverter. The C1000 Gen 2 has a limit of 1200W AC charging, so a small 2000W inverter is plenty for our experiment. With 600W faux-solar and 1200W AC charging (or really, with just the AC charging alone), we can run our loads entirely off battery power (of course, until the standalone battery runs out).

Between the solar and inverter, we would only be pulling 55A — well within the 100-200A operational limit of most standalone batteries. There is one minor downside: the power station has no way to directly query the standalone battery (remaining life, etc.). While many standalone batteries do have their own [Bluetooth] app, this becomes an additional complication to consider from an automation perspective.

While double-dipping technically works, it’s really inefficient: we would be taking DC power, converting it to AC via the inverter, just to have the power station convert it back to DC as part of the charging process.

Another thing to consider is the potential for noise: both the inverter and the dedicated charger (for the standalone battery) have fans that will likely be running in addition to the fans on the power station. For my first go at this, I wanted to keep things a bit simpler, so I’m not considering double-dipping.

[Back to Top]

Picking a power station

Putting all of the models under consideration in a spreadsheet, we get can narrow things down a bit:

• The Elite 100V2 beats out the C1000G2 based on the substantial jump in solar input

• Between the Elite 200V2 and the C2000G2, it’s much closer. Entirely by itself, I think the Elite 200V2 takes the win here with the not-insignificant advantage in solar-input however, the option to get an expansion battery on the C2000G2 should not be ruled out (despite OEM expansion batteries being silly expensive).

• In the context of the available options, the Elite 300 is a bit of an odd duck and only really makes sense if you were considering the C2000G2 with the expansion battery (~2200CAD). If we scale the price of C2000G2+Expansion down to match the battery capacity of the Elite 300, it is roughly 1620CAD. In that context, the Elite 300’s substantial advantage in solar input makes it a compelling alternative.

• The F3000 and Apex 300 are fairly evenly matched: the F3000 has more battery capacity out of the box vs the Apex 300 being able to do 240V natively. Having the option to get an additional 4000W of solar input into the Apex 300 is really unique.

My final short list was:

1. Apex 300 — the most expensive option of the three, but it’s a ‘do everything’ option

2. Elite 100V2 — going with this option means that I would definitely have to get a standalone battery (the sooner the better)

3. C2000G2 — this unit is ‘medium’ at everything; I think I would value the potential for an official ($$) expansion battery versus not having the option at all

[Back to Top]

Now, theorycrafting

Now that I have a power station in mind, we can start to do the fun theory-crafting. I went down a bit of a rabbit hole to try and model power-usage in a spreadsheet. My scenario is a bit unique in that I’m modelling power-draw that is relatively stable - my computer is always doing something.

View fullsize

View fullsize

View fullsize

View fullsize

For each scenario, I tackle them in the same way:

1. See how much of an on-peak period we can offset using the battery

2. If there’s remaining battery capacity, try and offset mid-peak where applicable. Otherwise, allow the battery to recharge during mid-peak

3. During off-peak periods, switch the power station to AC charging

The two variables at play are: the amount of starting energy available (i.e., whether we have an expansion battery or not) and the amount of solar generation available. The following pictures provide a bit of context in the thinking behind each strategy

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

When we start going down the strategy rabbit-hole, there are few details we need to remember:

1. Inverters consume power. Even if there’s nothing plugged in, just the fact that the inverter is enabled and standing by consumes power; how much power is ‘wasted’ depends on the model. In our spreadsheet-model, we can account for this a bit by reducing the ‘efficiency’ number for our power station.

2. Solar generation can be very fickle. In all but the most intensely sunny days, solar production will not be consistent - a single cloud rolling by, branches swaying in the wind and casting shadows or even bird poop can impact solar generation. We can tackle this two ways: either over-panelling ($$) or reducing the numbers around solar input in the spreadsheet.

3. We don’t (and probably shouldn’t) run the power station down to absolutely empty. For my testing, I set my operating mode to bottom out at 15%; this means that if the battery drops below 15%, the power station will cut-over to grid-pass through to keep the battery from going lower. This ensures that there is always a minimum amount of power available in case there is an actual power outage. We can tackle this by either using additional expansion batteries or by reducing the efficiency metric in the spreadsheet even further to give us some buffer.

All this to say that the spreadsheet is very much a dynamic, work in progress that also may be very uniquely dialed into my specific use case.

[Back to Top]

About ROI...

After all the scenario work, we can pull all the aggregate data together and try and draw some conclusions. We can see from the summary below that theoretically, the best options are ‘power station with lots of solar input’ and the distinctly second-best options are “power station with an expansion battery but also lots of solar input”. For my electrical needs, it seems that two expansion batteries are definitely not the way to go.

Here, we composite all of the daily-operating costs of my office across all of the different scenarios, accounting for both winter- and summer- billing schedules.

This spreadsheet is a bit of a work in progress and I’ll need to do some fine-tuning over time both to tailor the baseline settings (assumptions), change strategies and/or correct formula errors. Despite the raw numbers suggesting otherwise, my gut feeling is that going for an expansion battery is the better option for three reasons:

1. The expansion battery gives me additional runtime outside of any of this cost-optimization stuff

2. It’s possible that the factors I’ve built into my theory-crafting around the sun are unrealistic/optimistic

3. Even if my assumptions around sunlight are reasonable, any additional capacity allows me to offset load without the additional variability of depending on the sun

That being said, price (and availability) obviously plays a massive part in this equation.

[Back to Top]

My current setup

So I went with the Bluetti Apex 300 and currently, I’ve only got 100W of solar feeding it. I’ve defined the following in my configuration of the power station:

• Custom Mode [SoC 15-100%, with time-of-use and schedule defined]

  • The 15% means that no matter what, if the power station goes below 15%, then AC pass-through will be enabled. This won’t charge the battery, it just keeps the battery from draining any lower

  • The 100% is the point where, if the battery level goes higher than this, it stops the grid input with the idea being that the grid will top up the battery to the specified percentage and then solar power will fill up the rest of the way: if you have some decent solar available, you might set this to 80% (or lower!). In my case, I’ve got such an insignificant amount of solar input, it doesn’t really make a difference. I’ll revisit this later on.

  • The schedule allows me to define two types of time blocks: [a] charge the battery and [b] discharge/use the battery. Times not explicitly falling in that window will use AC-pass through (where the grid powers the load but doesn’t charge the battery). Note that even when the battery is in discharge/use mode, it still have the 15% safety net

• Backup Mode

  • I use this mode as a “make it charge now”

For my locale, in the winter-months, I have two on-peak time windows (0700-1100 and 1700-1900) and as such, I need to find a way to get the battery topped up before the second on-peak window. The behavior (of ‘Custom Mode’) is only to provide pass-through power during mid-peak periods.

Ideally, I would like to charge the power station during the mid-peak period but specifically avoiding charging more than necessary to handle that second on-peak window. This way, I can leave the majority of the charging to be done during the off-peak period. One way to work around the limitation of the app is to define some of that mid-peak period as off-peak. While this would work, it’s not a clean way of doing it. The second problem I have with this app is that it doesn’t support seasonal TOU scheduling so in the long run, this means I’ll need to handle the automations myself.

To get started, I know that if I have 75% or more remaining, I can definitely handle that second on-peak period regardless of what loads might be on the power station. To do what I want, I setup a few automations in Home Assistant:

• 1100h - If the battery is less than 65%, then I switch to Backup Mode allowing the power station to charge. I’m always running on empty (treading 15%) at 1100 so this always happens.

• 1330h and every half-hour after - If the battery is at least 75%, then I switch back to Custom Mode. This effectively stops charging the battery and just does AC pass through

• 1700h - Since the power station is already in Custom Mode and 1700h is defined as on-peak, it now uses the battery to power everything

[Back to Top]

Milestone

For the time being, the Apex 300 has been great - with the quick Home Assistant automations, the day to day operations are entirely hands-off.

Based on my theory-crafting, I knew that, even in the most optimistic scenarios, the power station doesn’t have enough juice to carry the office through the initial [winter] 4-hour on-peak period. Empirically, I’ve found that I can carry my office for 2.5-4 hours (most of the time, it lasts to 3.5 hours) — and this is running down to 15% reserve. The deviation comes from tweaking user-behaviour: I can explicitly chose to scale down intensive operations during the morning peak window.

That being said, with the summer ToU scheduling coming up, I know that I absolutely can’t offset the entire 6-hour continuous block.

I’d like to get a DIY expansion battery ($$) but I’m actually waiting for pricing and availability to go down that road. As the theory-crafting has already identified, a single 5kWh battery would allow me to offset the entire work day.

A much cheaper option would be to get started with one (or two!) 500W-class solar panels. My office is already ‘plumbed’ to run a panel in — it’s just a matter of identifying a specific panel to get and then the harder task of finding a vendor to order it from. That will be a the focus of another post ;)

[Back to Top]

Product links may be affiliate links: MinMaxGeek may earn a commission on any purchases made via said links without any additional cost to you.

• Data collection — what is using power? how much is it using? when is it using it? and

• Automation — what can I do about it? (without having to change my behavior)

And this is the perfect way to start talking more about home automation. This first post is intended to set the historical context and layout the groundwork for future posts on this.

Quick Links

• Phase 1: Smart:Life

• Phase 2: KASA

• Phase 3: Home Assistant

• Actual Zigbee deployment

• Milestone

Phase 1: SmartLife

About ten years ago, I got my first ‘smart device’ — a Wi-Fi controlled RGB strip. At the time I was just looking for a cheap RGB strip that I could stick onto a bed-frame that I had hastily built in a weekend. It wasn’t long before I moved onto ‘smart plugs’ and (compared to old-school mechanical timers), the advantages were obvious:

• Can remotely trigger them on/off

• Can set multiple schedules!

• Don’t t have to worry about daylight savings or power outages throwing schedules out of whack

• I wasn’t limited to a 30-minute scheduling resolution

When I learned that some models could do power monitoring, I was sold: this was a [somewhat] practical way to get insight into the instantaneous power draw for arbitrary devices around my home. It was non-invasive: I could just drop a smart plug in between an existing device and the wall. Over the next few years, I deployed almost 50 smart plugs — enough to start running into issues with my networking setup.

I started with Tuya devices and used the SmartLife app platform to manage them (I’ll use Tuya/SmartLife somewhat interchangeably to refer to these IoT devices). I explored a few other categories of products (contact sensors, RGB bulbs, etc.) but the overwhelming majority of this phase was deploying smart plugs everywhere.

Looking back, there was one glaring problem I had noticed but didn’t fully understand (and overlooked): every batch of devices seemed to come with it’s own (dead-end) firmware! I would faithfully initialize each device and it would report that it already had the latest available firmware, but you never knew what features you’d get until you actually went and used it. Obviously the most basic functionality (turn on/off) was consistent but beyond that, it seemed like everything was fair game:

• On some, you could only do double-bound scheduling (i.e., power-on from 07:00 to 07:30) whereas others allowed you to do ‘one-way’ scheduling (i.e., turn-on at 07:00, turn-off at 07:30)

• In addition to ‘on’ and ‘off’, some firmware released also have a ‘toggle’ action

• ‘Away mode’ where the switch randomly toggles on and off to try and simulate someone being home

• ‘Auto off timers’ which turn a switch off after a predetermined time, no matter what

• Timers: some allowed you to set a countdown timer (turn off in 5 minutes), some allowed you to set count-up timers (turn on in 5 minutes)

• Some smart plugs exposed ‘current power draw’ as an actionable data point — you could react to plug A drawing more than 50W by turning on plug B

The frustrating part is that these feature differences were all from the ‘same product’ — where I would just go to an Amazon page and reorder the same smart plugs.

The biggest pain point across all the hardware was that there was no elegant (or consistent) way to do anything with the power monitoring data. It was largely designed to do spot-checking of power consumption rather than long term monitoring. Even with that in mind, there was no way to get an Android desktop widget to be able to display this data at a glance, so you truly had to go into the app, connect to each smart plug and check them one at a time.

Compared to mechanical timers, I accepted that I was trading a bit of that bulletproof reliability for fancy IoT features and I was okay with that. I was however, a bit surprised at how many smart plugs failed over time (with no readily identifiable failure patterns). Of the almost 50 smart plugs I had in rotation, 25% of them were substantially defective enough that I ended up recycling them.

Anecdotally, only two seemed to have a failure in the the electrical relay whereas the rest seemed to have issues just on the software side - the thought that the companies were soft-killing their switches to ‘encourage’ additional sales of new products did cross my mind. Sometimes, you could rescue a smart device by doing a hard reset and re-initialize it, but as the number of misbehaving devices tallied up, I started to make plans for my second go at this.

[Back to Top]

Phase 2: KASA

For my second go around (~2023) , I had a better picture of what I wanted from my IoT devices:

• Hardware

  • I was done with surprise-feature sets - I no longer wanted to deal with differing features for the same devices

  • I wanted to be able to reliably buy replacement/additional parts as needed

• App/UX

  • I wanted a much more modern UX for the app with high scores for usability (subjective)

  • I wanted an app/ecosystem that was big enough that I was no longer bouncing between multiple apps to manage my devices

  • Ultimately, I wanted to make a move to Home Assistant as my home-automation platform

• Power Monitoring

  • I wanted a convenient way to export my data for analysis if needed

  • In addition to instantaneous power draw (W), I wanted to be able to tally power consumption over time (kWh)

  • Bonus points if there was a [basic] way for me to build pricing into the monitoring

  • Bonus points if there was a way I could have an Android widget to have quick monitoring

• Forward compatibility

  • Where available, the devices must support/implement Matter in some way and

Despite knowing that my long-term plan was to transition to Home Assistant, I still put a lot of emphasis on the native app-experience because it was going to be some time before I even started my Home Assistant deployment let alone get my system anywhere close to ‘final form’ and until that time, my primary means of interacting with the IoT devices was going to be via the native app. Even after getting a Home Assistant deployment “finished”, I suspect for “one off” interactions with smart plugs etc. will be quicker to do with the native app.

I ultimately settled on TP-Link’s KASA as my platform of choice, specifically starting with the KP125M Smart Plug.

Compared to the SmartLife experience, KASA was a breath of fresh air:

• Device initialization and setup was quite painless; you could even do soft-resets easily and without losing any customizations you already setup

• The app experience was snappy and modern

Even if I didn’t want to go down the Home Assistant route, the KASA UX is something I could definitely live with — and I did, for a few years. So, every time the KASA plugs went on sale, I bought a bunch and slowly replaced my SmartLife plugs. At the tail end of 2025, I had finally decommissioned all my remaining SmartLife plugs and only had a small handful of ancillary SmartLife devices left to migrate.

I had just a handful of devices that required the SmartLife (or other) app:

• SmartLife

  • Door contact sensor (super reliable)

  • Plug-in leak-sensor (super unreliable)

  • A switch for an outdoor outlet

  • The RGB strips that I started with all those years ago - still connected to the bed!

  • I had a couple other RGB strips that I stuck on my Leveled Floor to reduce toe-stubbing at night

• Other

  • An outdoor Govee Hygrometer

  • An Inkbird thermostat controller for my indoor Grow Station

There were some other miscellaneous RGB strips that I had configured as static colors and have no intention of interacting with again, so I’ve left them off this. My migration away from SmartLife was largely complete.

[Back to Top]

Phase 3: Home Assistant

Over this year end break, I finally bit the bullet and deployed a Zigbee network; after doing some initial validation, I was able to move some of my KASA plugs around for better usability targeting. From a power-monitoring and Home Assistant perspective, both my KASA and Zigbee smart plugs were mostly equivalent, but the direct usability of the KASA app lends itself better to some situations where I might need to ‘quickly set a schedule’ or ‘manually turn on/off a plug’.

The overarching ethos was to reserve the KASA plugs for when scenarios where I might want to directly interact with a plug and leverage Zigbee for when I just wanted quiet data collection or automation that could be driven by Home Assistant. As an hypothetical example, consider something like this where I have a combination of KASA and Zigbee devices nested together.

One limitation of KASA (and traditional WiFi based devices) is the lack of sensors. The reason is largely because WiFi just chews up so much power that it is untenable to apply for devices that need to constantly report a state. Being a low-power protocol, Zigbee is well suited to filling that gap and now that I’m finally getting around to using my Home Assistant setup, I can start to introduce additional sensors into my house.

One valid question is why I’m bothering with Zigbee smart plugs when I’m quite happy with KASA ones. This is for three reasons:

1. Zigbee is a mesh protocol and non-battery devices (like bulbs and smart plugs) contribute to the resiliency of that mesh network - so the more of them I deploy, the better my Zigbee connectivity will be for battery devices

2. Cost effectiveness. KASA Matter plugs, after taxes, come out to just under $16 each while my Zigbee plugs come out to $12.50 each.

3. As mentioned before, the KASA UX is pretty great, so having some Zigbee plugs, where it might (currently) be more of a burden to directly interact with, allows me to redeploy my KASA plugs to devices where I might want to directly interact with more

[Back to Top]

Actual Zigbee deployment

At this point in time, my Home Assistant is run off a dedicated Raspberry Pi 5; I have future plans to move this to x86 for the horsepower jump, but I’m still playing around for now. With that in mind, I specifically wanted an IP-based Zigbee coordinator to avoid having to deal with migrating a USB-based one down the road. I settled on the SMLite SLZB-06MR series and specifically got the SLZB-06MR3

Fun Story

I actually ordered the SLZB-06MR2 but I received the SLZB-06MR3 (this was where I got mine). I’m not complaining. :)

I got lucky in that we have a water line for the fridge — and there is a super convenient hole in the wall for me to access the stud bay and drill down into the basement where I can run my ethernet down to the network cart.

This gives me direct access into the basement drop-ceiling

Easy access into the basement

A quick command hook keeps the ethernet cable out of the way and deflects the slack so the cable is tidy

There’s plenty of space up there to place the coordinator

I left enough slack in the line that I could bring the coordinator down for troubleshooting/replacement purposes if needed.

Just a couple pieces of scrap and a hook

The wire holding jig manages the tension and spring force in the ethernet wire.

I can push it further back so you don’t see it at all, but I like being able to see it just barely as a reminder that it’s there.

Since I already have my own MQTT server, it was a no-brainer to connect my coordinator via Zigbee2MQTT. Everything just worked the first time.

[Back to Top]

Milestone

To me, home automation has always been a game of interconnected building blocks: to do <something> when <some condition> has been met etc. By leveraging Home Assistant and manufacturer-agnostic (and sometimes platform-agnostic) systems like Matter and Zigbee[2MQTT] I can start to finally get creative with my automations. Without the core infrastructure to connect the basic building blocks and test things out, I find it really hard to come up with things to automate, so I’m grateful that getting all of this spun up was super painless.

Before I close off, I think it’s worth addressing a really important question: why?

• Why go through so much effort to try and spin up automations? This is a fun question because the real question being asked is “what kind of automations are you making? [that you can’t use something easy like Google Home]”. The answer is fun because ultimately it requires thinking outside of the box [provided by things like Google Home] and when you go outside of that box, I find that the limiting factor is what inputs you have access to — it’s very difficult to imagine reacting to something if you don’t know what or how you can react to something. For some tangible examples:

  • Turn on the lights, speakers and accessories at my desk while I’m sitting there and to turn them off when I’m gone

  • When someone comes to the door, flash status lights (in various places around the house) to [silently] indicate that someone is at the door (handy if you’re on a remote call)

  • When my HTPC spins up, detect the power spike and turn on my powered subwoofer, power on the receiver, switch to the corresponding input and send an IR signal to the TV to turn it on (because, for some reason HDMI CEC doesn’t work for my combination in that scenario); do the reverse when the HTPC powers down

  • I have two heaters in a room (on a single circuit): when one heater powers up, the other powers down and vice versa

  • If it’s dark outside and the curtains are open, close the curtains

  • Water my plants, but only if they need it

  It’s also important to consider that platforms like Google Home require cloud connectivity — they won’t work during Internet/power disruptions and without getting into the privacy angle, they are reliant on those providers continuing to allow access to their systems without changes to the terms of service, addition of paywalls etc.

• The actions you are automating are so simple, does it really kill you to do it manually? Why does it need to be automated? The simple answer is they don’t need to be automated and I won’t die if I have to do things manually but that’s not my bar — if the task truly is easy and mundane, why I am I doing it in the first place?

• All of this seems really complicated, why does it have to be so hard? Can’t we do something similar, but without the complexity! I talked about leveraging super reliable motion triggers in a previous post — and for my setup, I intend to continue using them

Yes, a lot of this can be construed as a solution in search of a problem, but I didn’t call it a rabbit hole for nothing…

[Back to Top]

Product links may be affiliate links: MinMaxGeek may earn a commission on any purchases made via said links without any additional cost to you.

Networking Series

This post is part of a series, check out the other posts!

Quick Links

• Starting point

• Deciding on a switch

• New switch!

• Re-racking

• Additions to the rack

• Leveraging my managed switch

• Looking ahead

Starting point

Early on in my network journey, I made a very deliberate choice not to use a POE switch and instead, explicitly went with POE injectors to handle my access points. The huge selling point was that it was dead simple to forcibly reboot an access point if it was acting up — you’d just have to flip a power switch. This actually came in handy when my original outdoor AP was acting up and Nicole was able to just power cycle it - no having to fiddle around with management portals, the entire process was under 5 seconds: locate the appropriate switch and hit it.

Since then, some things have changed:

• My replacement AP has been quite stable; this removes my requirement for quick-and-easy power-cycling

• I wanted to start scheduling some of my APs to only be active during certain hours: I probably don’t need full-house Wi-Fi saturation at 0500h…

• I wanted to migrate my two remaining unmanaged switches to managed ones

As something that has been bugging me for a few years now, I would use this opportunity to swap out all of the front-to-back cables (primarily, ethernet but HDMI and USB as well) from 2ft to 3ft. The extra slack would better allow me to do live-servicing of the cart.

I had a few other punch list items I wanted to do (or have done and I’m just getting around to writing about it)…

[Back to Top]

Deciding on a switch

In my network stack, I had two unmanaged switches that I was hoping to consolidate. I knew that no matter what, I needed to get a 1G POE switch to handle the IP cams and APs so that was supposed to be the easy part. What I was trying to avoid, was sinking money into a big, expensive, 10G switch…

Luckily, I had just enough ports on my SX3008F that I can kick that can down the road a little bit longer: right now I only need three 10G ports (1xNAS, 2xPC) but I imagine in the future, that will expand to a second 10G-capable NAS. After that, I’ll need to either pony up, or get creative. But that is a problem for future me :)

In searching for a 1G POE switch, I had a passing thought that perhaps I could find a switch where I could just do an in-place swap of my existing 24-port 1G switch. This would have the benefit of not ‘using up’ my precious few 10G SFP+ ports. After setting my criteria, the options were a bit depressing:

The 52-port switches were easier to rule out: the SG3452XP was more money than I wanted to spend and for the SG2452LP, I really didn’t want to drop down to 1G. This left the two 28-port switches: was it better to have fan-less or an additional 8xPOE ports? It wasn’t clear if the ‘full-fat’ switch had a well-defined power-threshold where the fans would cut-off or not - that would have swayed my decision process. With the uncertainty, I went to the truly fan-less option - one less thing to fail over time.

If I had a time machine…

While I don’t regret my original 1G switch purchase, in hindsight, I probably should have went with the SG3452XP right off the hop as I ultimately spent ~$750 between my original 1G switch and now this 1G POE switch. I’m sure it’s not an uncommon reaction, but I really didn’t think I would come close to exhausting the number of available ports!

[Back to Top]

New switch!

With the decision made, I can consolidate my two unmanaged switches and this still leaves me with a scant three remaining POE ports. For the time being, I can’t think of anything I would want to power with POE - famous last words! I had previously considered racking up Raspberry Pi’s and running them via POE, but I don’t think I’ll really go down that road. Time will tell if I come back in a few years with a follow up on this ha.

[Back to Top]

Re-racking

Popping in a new switch is pretty straightforward: I did a quick test and firmware update when switch arrived but I needed to wait for the winter time off to perform the actual install as there were some additional punch-list tasks I wanted to do:

• Swap the 2ft ethernet/HDMI/USB cables running front-to-back with 3ft ones — the extra reach allows me to completely pull out the patch panel from the rack

• For the 8 devices I was now powering via POE, I wanted to switch to full-thickness ethernet

• Since I was eliminating the POE injectors from the equation, I no longer needed one of the PDUs

• Install emergency lighting

• Install an exhaust fan

Why on earth do I need to be able to pull the patch panel out?

Not all of the slots on the patch panel are ethernet — some are USB and HDMI and allow me to connect to the Raspberry Pi units and the USB ports on my NAS through the front of the rack. While swapping out ethernet keystones for USB/HDMI can be done without pulling the panel out entirely, it’s so much easier when I can directly access the entire back side of the port.

I’ve also run into a few scenarios where the keystone itself was popping loose (falling backwards, into the network rack) so being able to easily get to the back side to replace/fix this is so much nicer than trying to work through the panel.

Every cable is labelled to indicate which port on the patch panel it belongs to, so it’s easy to just leave the cable plugged into the device and disconnect the patch-panel side.

Setting things aside while I focus on the rack first

A better angle of the labels indicating where exactly the cable needs to be plugged into

Now we’re down to the 2ft front-to-back connections.

For the non-POE connections, I stayed with thin cables as they are nicer to bundle up. For the [active] POE connectors, I went with regular ethernet wires.

Here I was testing out having all of the network devices stacked in a series. It works, but it’s not as neat as I’d like it.

It was neater to have the new POE switch separated by the patch panel.

This is pretty easy - every cable is explicitly marked with where i needs to go.

Good enough for now

[Back to Top]

Additions to the rack

Emergency lighting

Previously, I talked about configuring my network for Endurance Mode — to be able to provide some degree of uptime during extended power outages. Currently, I have a manual process where I need to move the UPS from the wall to my EGO Power Station - a process that is made annoying by the fact that during a power outage, I’m working in the dark. Flashlights are no stranger in my house but it would be so much neater if I could just have an emergency light…

So one gotcha is that the entire network cart is powered by a UPS which means the emergency light won’t come on until the UPS dies — which is too late. I have a few options:

1. Directly connect the emergency light to the wall. This would work, but for sheer simplicity (during a power outage) I specifically only wanted a single power cable going to the wall

2. I could connect the the light to a different outlet on the UPS which isn’t supported by the battery. This also would work, but [for now] I have two UPS units and it’s one more cable to have [to remember] to transfer over between UPS units

3. I could put the light on a switch and during a power outage, I can just turn the switch off

Network exhaust

I don’t think my network cart runs hot, but it’s a problem I’d rather not find out if I can help it. I’m a huge fan of the Noctua 200mm USB fan for these types of applications. I opted to put a fan grill on each side of the fan to minimize the chances of accidentally chopping a wire or whatnot. In order to quickly disconnect the fan, I use an inline kill switch.

[Back to Top]

Leveraging my managed switch

So now that I went ahead with all that work to transition to an entirely managed network stack, I spent some time tinkering with it. Obviously, everyone’s setup (and their needs) will be a bit different, but I finally got around to getting

• Firmware update on my Endurance Mode travel router

• Getting my VPN working again

• Configured my POE switch to schedule downtime for some of my access points

• Consolidate a couple IoT SSIDs, I used this as an opportunity

By far the most tedious (and rewarding) task was to update my documentation on every single device on the network. Previously, I had poked around with using something like phpIPAM but I found it to be a bit too cumbersome for my needs. I ended up just going with a simple spreadsheet

I specifically itemize the MAC address using dashes and colons because it makes doing a lookup of the MAC address a bit easier when I’m just given an arbitrary value to search for. Using the formula =SUBSTITUTE(C1,"-",":") makes it so you don’t need to enter the value twice. If you have more VLANs, simply clone the set of columns for each VLAN. A great benefit is that all of your devices are shown on a single view, without having to bounce back and forth between tabs.

[Back to Top]

Looking ahead

I recently started poking around with solar power and toying with the idea of trying to peak-shave part (or all!) of my office electrical draw. Tying into Endurance Mode and from a resiliency angle, I think I’m going to go down the road of building a power station into the network cart.

This would allow me to handle most extended power outages and/or peak-shave the network cart. Depending on the size of the battery reserve there is a corner case where an extended power outage begins at the tail end of peak-shaving (i.e., the battery may be nearing depletion). Being able to set a reserve battery amount and/or supplementing with solar can be a way to reduce this risk window, but this is an adventure for #future-me. :)

[Back to Top]

Product links may be affiliate links: MinMaxGeek may earn a commission on any purchases made via said links without any additional cost to you.

Quick Links

• Where we started

• Goals and Objectives

• Theorycrafting

  • What is peak-shaving?

  • Some easy examples

  • What about a bigger load?

  • A realistic example

• Step 1 - Using the C300 to offset on-peak

• Step 2 - Bypassing the inverter

• Step 3- Getting solar into the office

• Step 4 - Thinking ahead

• Step 5 - ROI

• Step 6 - Window shopping

• Step 7 -  Fine tuning

• Milestone

• Appendix - Parts list

Where we started

A few months ago, I was wandering through a liquidation store when we came across the Ryobi 18V Hybrid Cooler. Growing up, we had a 12V car-cooler that came in handy when we went to far-away grocery stores and meat markets. While some 12V coolers come with an AC wall option (and I guess I can always buy a dedicated 12V adapter), the Ryobi cooler gives me the option to totally disconnect, relying on batteries entirely.

Looking at this from a different perspective, the cooler comes with it’s own built-in UPS. With a regular price of $700CAD (!!), there was never a chance in hell I would have considered this but the liquidation price of $300? This puts it in the same ballpark as many 12V coolers. Early in the [camping] season, a pair of 9ah batteries (~162Wh each) was more than enough to last the weekend but as the weather really heated up, I needed to find a better way to power the cooler without packing a whole bunch of batteries…

As luck would have it, I stumbled upon the Anker Solix C300 shortly after. From a camping perspective, this provides me with 288Wh of power: between the Solix, the two batteries installed in the cooler and a couple of extras, I’ve got 936Wh of power. Enough to reliably go off-grid during a weekend camping trip regardless of the temperature. But enough about camping — how I can I really use this?

You know, for computers.

View fullsize

View fullsize

[Back to Top]

Goals and Objectives

It’s no surprise that, the computers represent a significant portion of my electrical bill. I’ve been toying with the idea of using power-stations to try and offset some of the load when electricity is the most expensive throughout the day. Obviously, getting a beefy enough power-station that can comfortably drive my computers is not a cheap undertaking, so I wanted to use this small, camping-oriented power-station as a sandbox test-bed.

At a glance, these are the relevant specs:

• 288Wh battery capacity

• LiFePO4 batteries rated for 3000-cycles (meaning, after 3000 charge-discharge cycles, it should retain 80% original capacity - 230Wh or so)

• 300W inverter with 600W max draw

• Rechargeable via: AC (100W min, 330W max), USB-PD (140W max) or solar (11-28V, 100W max)

• Idle power consumption of 11-12W

Since I already have the C300, I wanted to a way to leverage it’s battery.

[Back to Top]

Theorycrafting

What is peak shaving?

Before diving too deep into things, here’s a sample time-of-use cost of electricity chart.

So there’s this idea of peak shaving: if we can find a way to reduce electrical draw from red and orange periods — we shave the cost of electricity during the peak periods of the day.

[Back to Top]

Some easy examples

It can be easier to visualize and explore how we might leverage this with an example. For a simple example, let’s suppose I have a 100W load that needs to run 24x7; let’s make some assumptions to make things a bit simpler:

• That there are no efficiency losses from the inverter

• There’s no idle-power consumption

• The power-station has a max recharge rate of 300W

• We are working with an 1800W circuit or so (a standard 15A outlet)

• Once the battery is depleted, we will immediately switch to grid-power and start recharging the battery - I don’t know of any system that allows you to switch to grid-power without recharging the battery. What this actually means is that our load (currently 100W) and our recharging (currently 300W) must fit within this 1800W allowance.

  • You can easily see how we can run into issues as our load and charging rates go up

  • Some power-stations allow you to adjust the recharge rate so you can optimize things - more on this in a bit

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

[Back to Top]

What about a bigger load?

Things get a but more complicated as we start to introduce bigger loads: since the circuit’s capacity (1800W) is fixed, this means that we will need to reduce the charging rate so that we’re tripping the breakers.

In the case of an 1100W load, this means that we would have to limit our charging rate to a maximum 700W or so. Since our charging rate is less than the load, there is an upper limit to how much we can recoup during off-peak hours. This ends up being a rough ceiling on the battery size that makes sense.

For this specific scenario, we have a soft-limit of roughly 8400Wh on the battery since that is the most we can replenish during off-peak hours. Adding more battery capacity doesn’t change this and just kicks the can down the road so to speak. At some point we will run out of battery and when that time comes, we will only be able to put 8400Wh back into the battery - the only way to stretch the runtime is to either increase the rate of charge or reduce the load.

So far I’ve left this off, but there are two efficiency losses to consider:

• Inverter efficiency losses: if you’re using the inverter, do note that the is power used just for the inverter to function. This could be 10-watts, or 100-watts — this all depends on the inverter

• Battery degradation: that 100Wh battery we bought years ago might only have 85Wh of usable charge left in it now

[Back to Top]

A realistic example

To bring things back down to reality a bit, consider a much more realistic example. This might be my computer, with three power bands:

• 2hours @ High: 850W (i.e., gaming)

• 11 hours @ Medium: 500W (monitors on, computer still churning at 100%)

• 11 hours @ Baseline: 250W (monitors off, computer churning at 100%)

When we plot this out we can begin to see where we might try and offset costs.

This is all very exciting, but I’m not ready to tackle a real computer yet — I need to find a baby-steps version of this. By coincidence, my work laptop is a perfect candidate to test this all out:

• Not a significant power draw

• Fairly predictable [work] schedule

• If I somehow run out of battery, it’s not the end of the world — the laptop has a built in battery

[Back to Top]

Step 1 - Using the C300 to offset on-peak

Since the C300 was intended to be used in a camping capacity and totally not in a home power-offsetting capacity, a lot of the fancier time-of-use settings aren’t available. Especially for the way I intended use it, I had to make the following setting changes:

• Disable ‘Smart AC Output Mode’ — this forces the inverter to run continuously, even when the laptop isn’t running. I could leave this setting enabled but then I would have to power on the inverter (a single button press) every time I started my work day. That’s one more hurdle I don’t really want to deal with although another way to look at things: this is an option I can keep in my back pocket if I need to squeeze more runtime out of this setup

• Disable ‘Device Timeout’ — this also keeps the C300 from going to sleep in general

To work around the limitation of not having the built-in ability to control power scheduling, I a pair of power-monitoring smart-outlets. This would allow me to control (and measure) power coming into the C300 as well as measure the power going to the laptop.

By placing a power monitoring outlet upstream of the C300, I can monitor the idle power consumption when I know everything is off. The C300 has between 11-12W of idle consumption.

The loads are:

• My work laptop uses roughly 25W

• The C300 itself has roughly 11-12W idle power consumption

• Each of the smart outlets consumes 0.5W (idle) or 1.0W (active)

The ideal goal is to shut-off power recharging the C300 at 0700h (when off-peak rates end) and restore power after 1900h (when off-peak rates resumes). Would the 288Wh capacity be sufficient to carry everything for 12-hours?

Not quite, but there is an important nuance to consider: that I only needed to offset the laptop’s power draw during my work day and more specifically, I only needed to offset that power draw if it happened to fall during mid- and on-peak hours. What this means is that if I can shift some of my working hours into the off-peak period, that consumption would be out of scope of any optimization.

After testing for a few weeks, I found that I was able to offset the laptop’s power draw almost all of the time. To plan for this a bit better, I scheduled Switch A to turn on at 1000h for 10 minutes. With a 330W charging speed, this 10-minute window allows me to recharge roughly 55Wh (19%). This quick top-off ensures that I can make it all the way to off-peak hours no matter what my work day looks like, and this top-off is done at mid-tier rates.

[Back to Top]

Step 2 - Bypassing the inverter

So far I’ve been powering the laptop using a fairly standard AC laptop power brick but using the AC outlets a power station is inefficient because the inverters themselves require power to operate. In the case of my C300, it’s 11-12W — enough to drain the entire power station in a day with just idle consumption alone! If we can find a way to power the laptop using DC power, we could leave the AC outlets turned off, boosting overall efficiency. There are three ways to accomplish this:

1. Using the USB-C outlet

2. Via the 12V car socket directly using a 12V to USB-C charger

3. Via the 12V car socket indirectly using your own inverter; in my case, I happened to have an old Ridgid 100W inverter kicking around. By using your own inverter, you’re gambling that your inverter has less idle power consumption than the one built into the power-station

Just out of curiosity, I did an informal test:

• The laptop would be powered via the USB hub

• The laptop would not be at 100% power (since there can be some weird on-off cycling when the battery is that full) - I targeted 90% charge

• The laptop would be powered up, all applications closed and it would idle at the desktop, one VNC session would be open. I also did another test where I left the laptop running a stress-test for the duration of the test

• I would watch the app for a 5-minute period and track the reported power output values (it updates every few seconds)

• In between tests, I would top up the C300

• The upstream smart outlet would be powered off for each test

I performed each test (roughly x100 data-points) three times (x3), for each power source (x3), at idle and full load (x2). A ton of data. We already ‘know’ that powering the laptop via USB will be the most efficient, but it is kinda cool to see the numbers come in and validate this understanding.

It was also cool to validate that my standalone inverter was less efficient than the built-in one. The only test I wasn’t able to do was to power the laptop using a 12V socket to USB-C (since I don’t own an adapter like that). Since it’s all DC to DC power, I can’t imagine it being substantially more efficient than connecting directly via USB.

View fullsize

As a result of this test, I discovered two neat things:

1. Having a network card connected to the hub results in a 0-2W draw. If I was desperate to shave this off, I could resort to wireless (never) or consider something like a physical network kill switch (questionable reliability based on the reviews).

2. When powering the laptop via USB: if the laptop battery is not at 100% and we hibernate/power down, the power-station will show a 7-8W draw (this amount includes the 0-2W from the network connection). Once the battery is charged to 100% (either by manually charging it before shutting down, or the trickle charge happens to ding 100%), the draw disappears an the power station returns to ‘just’ the 0-2W.
   
   This oddity does not appear when using the AC ports (but may still be the case) — the power-station is not able to reliably report AC loads under 20W. This may just be an oddity of my specific laptop, or could be a quirk of Thunderbolt, or USB-PD, I’m not entirely sure.

Sadly, I won’t be able to power my actual computer using USB so when I eventually apply this to my computer, I’ll have to go with the inverter route.

[Back to Top]

Step 3 - Getting solar into the office

The next logical step in the experiment was to leverage solar charging to top up the battery during the day. I picked up a Renogy 100W panel and got this plumbed into the basement. I did this by mounting a junction box to the exterior of the house and fishing my MC4 cables through the wall and into the box. By doing it this way, I keep all the connections in one place: if I have to move/service anything, I can handle physical connection/disconnects without having to run back and forth into the house.

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

Looking online, I saw a lot of guides showing people attaching MC4 connections directly to the sides of the junction box. I specifically opted to run my connections through the front/lid of the box because:

1. This was my first time doing this, I didn’t want to gamble on screwing up crimping MC4

2. Replacing a lid is so much easier than unbolting and replacing a box

3. Servicing/upgrading connections through the lid is also much easier: I can remove the lid, take it over to my workbench and work on it rather than hunching over by the junction box

I specifically went with an RV-style gland with cables because I wasn’t sure how much depth I would need for the MC4 connections (if I were to pass them through the front) versus how deep the junction box would be. The extra length of cabling from the gland is a bit unwieldy but it does give me a bit of operating flexibility when servicing.

I went with a bigger junction box for the possibility to pass two sets of cables through the front but the extra space in the box was super helpful when coiling up the extra length of MC4.

I used a 15-ft MC4 cable inside the house be able to go from the junction box over near by C300 (in hindsight, I probably should have went with 20-ft inside the house as well). I used a spring-locked grommet to keep things tidy. The C300 requires an XT60 connector for solar input, so I used an adapter cable to run the last mile.

I’ve only had this installation for a few weeks so I only have limited (summer) experience but overall, it’s been great.

Obviously, my summertime experience represents the most ideal conditions and as the seasons change, it will be more challenging to keep the battery topped off. That being said, I did my initial testing with no solar recharging, so even if I get a small amount of solar over the winter months, I should come out ahead.

Worst case, I re-enable the schedule to add top-off charging during mid-peak.

View fullsize

[Back to Top]

Step 4 - Thinking ahead

I briefly considered over-panelling: running two 100W solar panels (in parallel) to feed into the 100W-max solar input of my C300. By over-panelling I can

• Expand my charging-window to cover more hours of the day

• Increase the charging output during cloudy periods

With parallel connections, voltage is constant and the current is added up; doing this allows me to keep the incoming voltage within the (low) voltage tolerance range of the C300 and I would be relying on the built in charge controller to clip the current, limiting itself to the 100W it can handle.

I ultimately chose to pass on this idea — for now. I want to get a bit more experience with my setup the way it is, before making substantial changes to it and also because it’s very likely that I’ll upgrade to a bigger power-station in the medium-term.

A bigger solution would allow for a much higher solar input and I’d be better served by transitioning to larger (i.e., 400W) panels.

To preemptively handle cloudy days (and winter), I opted to schedule Switch A to power-on for 5-minutes at 1705h (mid-peak); this would allow me to feed 27.5Wh (9.5%) back into the power station. On sunny/ideal days, these 5-minutes wouldn’t matter (since the battery would already be near 100%), but the quick-top off on a cloudy-day would allow me to stretch the battery to off-peak time.

[Back to Top]

Step 5 - ROI

Inevitably, when thinking about solar, it’s just a matter of time before the idea of ROI (return on investment) creeps up for most people because getting this stuff (at least at scale), is not cheap. Everyone’s situation will be different as it depends on how much power you use, when you use it and the difference between the on- and off-peak rates (which can also change, seasonally). I have an alternative set of rates available to me and it’s neat to see how the theoretical numbers may work out.

I’m not getting too excited about the 75% reduction in daily-operating cost since I’m totally not ready for it. This sample scenario is just for one computer, and a rate-change would apply to all the other things in the house. At the house-level, there are several substantial loads that I don’t necessarily have the leeway to reschedule (i.e., stove, air-conditioner).

For the time being, I’ll focus on the savings that I can get using Rate 1, just with offsetting mid- and on-peak consumption.

[Back to Top]

Step 6 - Window shopping

Without scaling up to whole-house solar, having the MC4 pass-through into the basement does give me some more consumer-oriented options: with a beefy enough power-station, I can start to move towards leveraging the savings while still remaining on Rate 1. Consider some of the larger power stations like the Solix F3000 or Delta 3 Ultra Plus, we are starting to get to the point where there is enough battery capacity and solar-recharge capacity that I can make a meaningful dent into my power consumption.

With either of these power stations (+ single expansion battery), I have enough capacity that I can definitely cover my on-peak usage (approximately 3500Wh) entirely, without relying on any solar recharging. I specifically selected these two power-stations for their above average solar-input numbers: even with just a few hours of usable solar, I can put a substantial amount of charge back into the batteries - likely enough to get me to the off-peak period. There’s also a more than likely chance I can start to add multiple computers, or my network rack to this power optimization as well.

Some exciting stuff - all the more reason I need to play with my C300 in it’s ‘basic state’ to get a better picture of how (and when) I use my power. Obviously, there are other products out there and there’s and entire world of DIY/custom to consider as well. One tempting option would be to get a high-capacity battery and connect that to the C300 (or other power station). The power-station just sees this battery as a ‘solar source’ and I can worry about charging the standalone battery separately.

In addition to getting more real-world experience with this baby-steps C300, it would be beneficial to get per-circuit continuous power monitoring so I can get a better understanding of how and when I use power.

Super long-term, I think the right play is to go with whole-house integrated solar - with enough capacity/generation that I can take a serious look at switching to the alternative electrical rate.

[Back to Top]

Step 7 - Fine tuning

The following

• Every day: 0655 OFF. Turning off while we are still in off-peak pricing

• Weekdays: 1705 ON → 1710 OFF. For weekdays, this represents mid-tier rates and gives me a very quick top-off (if needed) in case it’s a cloudy day and I’ve been pulling more power than typical. Assuming the solar panels haven’t contributed anything, this will put 27.5Wh back into the battery.

• Weekends: 1400 ON → 1415 OFF. For weekends, this is all off-peak. Assuming the solar panels haven’t contributed anything, this will put 82.5Wh back into the battery.*

• Every day: 2020 ON. This is well within off-peak hours and the grid can power the idle-power consumption.

[*] I won’t need this if I’m using just the USB ports on the C300, but I’m playing around with the idea of migrating additional devices to the power-station (like my pedestal fan)

[Back to Top]

Milestone

For the time being, I’ve got my work laptop situation fully handled by the power-station and I’m just looking for small things that I can migrate from grid-power to my power-station. As winter is coming, it’ll be neat to see what changes I need to make (if any) deal with the fewer hours of usable sun.

In the spring, I’ll figure out a way to get the panel’s installed on some tiny angle-stands — for the time being, they are just flopped against the house/fence. I’ll also be looking at what options are available for upgrading to a larger power-station to begin to migrate more things off peak-power.

I’ve got some research to do and a ton of data I need to collect (power usage patterns etc.). This will be an interesting rabbit hole to explore!

[Back to Top]

Appendix - Parts list

• Power station - Anker Solix C300

• Solar panel - Renogy 100W, N-Type, 16BB

• MC4 Cables - Renogy 10AWG (20-ft outside, 15-ft inside)

• RV-style Solar entry gland (3ft 10AWG)

• Cheapo MC4 disconnect tool

[Back to Top]

Product links may be affiliate links: MinMaxGeek may earn a commission on any purchases made via said links without any additional cost to you.

Self-hosted Series

This post is part of a series, check out the other posts!

Quick Links

• Starting point

• Building out the Rules

  • Amazon

  • Reddit

  • YouTube

• Creating your own lists

• My Workflow

• Bonus Shenanigans

• Summary

Starting point

Over the last year or two, Google has been making some moves that ultimately have/will lead to the end of one of the most effective browser based ad-blockers. As I run network-wide DNS based ad-blocking, this may affect me less than others but DNS based blocking doesn’t catch everything and you really do need browser-based blocking to handle the last-mile.

I don’t blame Google for making these moves — they are an ad company which naturally puts them at odds with ad-blocking, but with the writing on the wall for uBlock Origin (in Chrome), I cut over to Firefox. As I run multiple machines where I want to maintain a consistent rule-set, this was a good opportunity to spend a bit of time on refining my ‘last-mile’ ad-block process. uBlock Origin supports the idea of cloud storage to handle synchronizing rules between machines, but I wanted to take this a bit further.

The one thing that you can do with browser-based shenanigans is to restructure a website.

• Your favourite website changed the colour of a button that you don’t like? Override that

• Don’t care about sports/arts/entertainment/business/weather on your preferred news site? Hide it

• A website you visit often has a fixed divider that just isn’t big enough (Azure DevOps I’m looking at you) - resize it

• A site has a button placed way out in left field? Move it

And then the rabbit hole got deeper:

• Could we hide entire products/brands/categories on Amazon?

• What about blocking out specific channels on YouTube? (AI generated puke, I’m looking at you)

• Can we filter out entire subreddits on Reddit?

[Back to Top]

Step 1 - Building out the Rules

We’ll jump right into it, the super star that makes all the magic happen is…

:has-text(/your-text-here/i)

Find the element archetype you want to clean up, append that magic block and go on a censoring spree. It’s a slow, iterative and cumulative process as you build out your desired filters, but it gets better over time.

Note: the optional element in red allows the filtering to be case-insensitive.

Amazon

Here, I'm interested in cleaning up the Deals page: my thought process is that I like to browse through the Deals page, perhaps looking for random impulse/super-sale items, but there are definitely things that I'm not interested in:

• Brands (i.e., Apple)

• Product categories (i.e., sweaters -- this isn't something I'm going to impulse buy)

• Specific products (i.e., a specific coffee maker -- I already have one, so I won't be buying a second one)

• Alphabet-soup products and companies -- you know the type, where the company name is a random jumble of letters and the product is just copy-paste-generic-crap

For each thing you want to block, I've found you need two entries:

amazon.ca##li.dcl-carousel-element.a-carousel-card:has-text(/thing-you-dont-care-about/i)
amazon.ca##div.GridItem-module__container_PW2gdkwTj1GQzdwJjejN:has-text(/thing-you-dont-care-about/i)

You do need to be a bit cautious (sometimes being more-vague with your filters, sometimes more-specific) otherwise you might accidentally block items/categories you are interested in. For example, let's say you do want to see phone accessories, chargers, cables etc. but you don't want to see anything Apple-specific. If you were to add a block for Apple, you would definitely be inadvertently blocking some/many things that you might be interested in as many of them will use Apple in the description. So you might build out a few blocks that are more granular:

• for Apple

• for iPhone

• for iPad

• etc.

Will you still [inadvertently] block stuff? Yes, but it’s a balancing act. The ‘clean-up’ process is pretty fun to watch in real-time — the product blocks get lazy-loaded and then hidden by the ad-blocking.

Ad-block on Amazon purging items/brands/categories you don’t care about

Sometimes, you can throw a :not in there to simplify your life. For example, on the main Amazon page, the only block I might be interested in will be ‘Keep Shopping For’, so I have the following rules to allow it, but filter the rest

[Back to Top]

Reddit

I admit, I didn’t look too hard, but from what I can tell, there isn’t an option in Reddit to say “please don’t show me this subreddit ever again, I’m not interested”. Enter ad-block. Again, I’ve found that you need two rules to hide things fully (one to hide from the top carousel, and the other to hide from the main listing):

[Back to Top]

YouTube

Again, you can use this to filter out channels you never want to see again because you find them boring or… you know the channels: the ones that are purely text-to-speech product-puke. Like before, I found that you need a couple rules to hide the channels (one hides it from the general listing, the other hides it from search results):

Bonus: if you want to just rid yourself of YouTube shorts…

[Back to Top]

Creating your own lists

So the wonderful thing about uBlock (and other ad-blockers) is that you can leverage lists of things to block. Most of the time, we leverage lists curated by others but nothing stops us from making our own list. At a high level, we just need a text file to be accessible by the browser - where and how you get that text file is up to you.

One easy (and free) way to do it is to use GitHub.

1. Create your lists of rules, optionally organizing them into subfolders and multiple lists (so you can easily troubleshoot by turning off an individual list)

2. Create an account and somehow, get your list into GitHub

3. In uBlock, import your list

The signature for your list will be something like:

[Back to Top]

My Workflow

For myself, I run my own self-hosted Gitea instance at home, where I have a repository for holding all the ad-blocking rules (both for PiHole and for uBlock). Since I batch out a whole bunch of rules at a time, I threw together a quick little desktop app to build out rules for a bunch of websites I regularly visit.

Once I have a bunch of rules done up, I do a pull-request which then triggers a sync from my local Gitea instance to GitHub. In my uBlock configuration, I import all of the GitHub-hosted rules. Since I have a ton of different files, I also keep an index.txt file that contains a list of all the GitHub files in a ready to copy-paste format so I can readily deploy to a new machine if needed.

Using Amazon for example, I will have my app open while I do a quick scroll through the deals page and add blocks as I go — it doesn’t help me this time, but after the rules get integrated, I’ll never see those annoyances again.

At the time of this post, I have 62-lists and just under 9500 rules defined.

Bonus: you can use uBlock (with these lists) on Firefox for Android as well

View fullsize

View fullsize

[Back to Top]

Bonus shenanigans

Above, I alluded to the idea of if you don’t like the way a website is laid out - you can change it (to a degree). For that, I use a tool called Stylebot. As real-world example, for work, we use a tool called Azure DevOps and for the pipeline details page, the individual job steps are in a fixed-sized column that just isn’t wide enough to show the full text some/most of the time.

Since I am dialed in for 4K/8K and have resolution for days, it’s all kinds of frustrating to then have a (web!) UI that feels like it’s fixed to 1280x960 or something silly — with the double-whammy of the obsession of (web!) UI to have gobs and gobs of white space for the sake of white space.

Rant aside. For Stylebot, the corrective override is

Set the styling override to your AzDo portal’s wildcard URL (so, https://your-azdo-url.com/*) and you’re set. Depending on your resolution (and size needs), you can play around with the size there

[Back to Top]

Summary

A lot of the websites that I visit tend to have a lot if crap on there, with a smattering of things that might be a bit interesting mixed in. Being able to filter out the crap helps to keep the experience more enjoyable and ultimately save me time — I’m never going to be interested in X,Y,Z so having to scroll through it again and again is no good for anyone.

Sometimes, websites will have “hide this” functionality and while that is a definite improvement over not having that option, this only hide’s that specific instance — another similar post can be made tomorrow and you’ll just have to hide it all over again. Being able to permanently block things categorically means, you only have to be annoyed by any given thing once (or certainly only a few times).

[Back to Top]

Product links may be affiliate links: MinMaxGeek may earn a commission on any purchases made via said links without any additional cost to you.

Self-hosted Series

This post is part of a series, check out the other posts!

Quick Links

• Starting point

• Automating Private DNS

• Novel use case: parental controls

• Summary

Starting point

A few years ago, I switched to using Pi-Hole as a network-wide ad/malware sinkhole and generally speaking it’s been great - I even deployed a second Pi to act as a secondary DNS for my network.

Phones in particular, can be tricky to blanket protect as your protection mechanism needs to work at home (Wi-Fi) and on the go (data). At the time, I mentioned that I was hoping to go with a two-prong solution: Pi-Hole whilst at-home and NextDNS while out and about.

For Pi-Hole, there’s nothing to do — it’s configured as the DNS server on the router, so in general, everything on the entire network is protected. For NextDNS, there are a few ways to get things hooked up; thankfully, in Android, I can manually specify Private DNS settings (and skip having to install any VPN-esque apps or do any per-network configuration tinkering). There’s only one catch: Android uses this private DNS setting regardless of whether I’m out and about (using data), or if I’m at home on Wi-Fi (under the coverage of Pi-Hole).

The bottom line is that, yes, everything will still work but there are some nuances that made me reconsider just setting this blanket setting. As a simplified example, consider making a request to example.com:

In both cases, we ultimately get the IP address that we’re looking for (1.2.3.4) and for the purposes of this example, we’re going to disregard any extra (likely negligible) latency required for performing the initial lookup to NextDNS. Everything works. If this is all you’re after: then you can likely just set the Private DNS setting on your phone and be done with it.

So what’s the big deal? Things are a bit different when we look at the Pi-Hole logs:

For my use case, this is a bit of an maintenance annoyance: because Android is now obfuscating my actual request (to example.com) via a request to NextDNS,

• None of the log-entries in Pi-Hole tell me anything (this would be pretty similar to running a higher privacy level in Pi-Hole)

• None of the blocking rules provided by Pi-Hole ever kick in (because NextDNS isn’t on a block list)

• If I’m running all of my queries through NextDNS (and especially with multiple devices), I’ll hit the 300K/month limit for the free-tier

The big deal though, is having different blocked/not-blocked resolution depending on the device, while at home, on Wi-Fi. This is because my Pi-Hole and NextDNS have different blocking rules in place (broadly speaking - I’m sure there are some overlaps).

Turning off the Private DNS setting would solve the problem at home but this removes all of the filtering and protection while on data and honestly, if I have to toggle something (even if it’s a simple toggle) every time I connect/disconnect from Wi-Fi, that’s just too much friction — it needs to be automated in some way.

[Back to Top]

Automating Private DNS

In an ideal world, we would have something like this:

• If connected to home Wi-Fi, disable Android’s Private DNS — filtering is done by Pi-Hole

• Otherwise, enable Private DNS and let NextDNS handle it

We can leverage Tasker to do the automation, but before we get started, we need a few things:

1. Obviously, setup a NextDNS account, define your rules etc.

2. Your phone needs to be rooted. You may be able to do this without root, but I don’t know, it’ll certainly be more involved.

3. You need to set the Private DNS endpoint in Android in advance - Tasker will just toggle between ‘off’ and ‘Specified DNS’

Create the Tasks

View fullsize

View fullsize

View fullsize

View fullsize

Create the Profiles

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

This is a super simplified starting point to get you started. There are a couple easy ways to spruce things up:

• You can add [persistent] notifications to the each of the Tasks to add a visual cue to indicate whether you are in ‘Pi-Hole mode’ or ‘NextDNS mode’ (as I have in a few of my screenshots). If you go down this route, make sure you also add a ‘Notify Cancel’ action (to clear the previous persistent notification)

• You can react to users clicking on the notification using ‘Notification Click’, by triggering a Browse URL action (a useful NextDNS URL would be https://my.nextdns.io/YOUR_NEXTDNS_ID/analytics)

[Back to Top]

Novel use case: parental controls

Just thinking about this for a moment, there’s definitely a novel use case specific for parents here. At home, you might go down the route of adding in protections, filtering and parental controls but what if your kid goes to a friend’s house and connects to Wi-Fi there? Now with this automated Private DNS, regardless of what filtering may or may not be in place at the friend’s house, there everything will still go through NextDNS.

As a bit of a corner case if you happen to use a generic SSID (i.e., ‘Guest’, or ‘Kids’): you may want to consider specifying the MAC address of your router/AP instead of just the SSID. This way, when your kid goes to a friend’s house and connects to the ‘Kids’ network, there will still be filtering in place.

Of course, permanently setting the Private DNS setting will always provide you with the protection and you only need to look for logs in one location (in NextDNS) but, depending on the number of kids/devices, you will be more likely to hit the 300K/month query limit/

One more neat thing: you can add some identifiers to the DNS logging. For example, if your NextDNS identifier is A1B2C3, then you might set your Private DNS for multiple kids/devices:

• john-phone-A1B2C3.dns.nextdns.io

• john-tablet-A1B2C3.dns.nextdns.io

• susy-phone-A1B2C3.dns.nextdns.io

• shared-device-A1B2C3.dns.nextdns.io

Now, when you go to your NextDNS analytics, you can see which device/user is going where.

[Back to Top]

Summary

As I mentioned above, everything still works if you permanently set the Private DNS for all devices — just that we bypass the value of having Pi-Hole protecting our home network. Yes, by having a two-pronged approach, we will still need to maintain two sets of rules (I’m a bit more lax with my NextDNS settings) but there is consistency in having “all home devices are covered by Pi-Hole and nothing else”.

[Back to Top]

Product links may be affiliate links: MinMaxGeek may earn a commission on any purchases made via said links without any additional cost to you.

Prior to starting down the self-hosted journey, I made the jump to a slightly more structured networking setup, so this was the next logical progression.

Self-hosted Series

This post is part of a series, check out the other posts!

Quick Links

• In the beginning…

• Getting started

• My self-hosting setup

  • At a glance

  • Flow

  • Hooking up the reverse-proxy

• Services spotlight: what kinds of stuff do I host?

  • Vaultwarden

  • Actual Budget

  • Vikunja

• Looking ahead

In the beginning...

A long time ago, I was in the market for a password manager and the big player in that space at the time was LastPass. I was no stranger to password complexity, but I was looking for a usability improvement (as compared to a local passwords file). To me, LastPass scored very highly on usability and not long after, I migrated to their Premium plan and then a little time later, their Family plan.

Generally speaking (and despite a few high-profile security breaches), I was fairly happy with the service. What I wasn’t too impressed with, were the crazy high prices.

But that’s just how these online services work right? They wow you with a fantastic onboarding experience but to get out? Well, that’s going to be a bit more involved, now that you’re invested into their ecosystem. Looking at this list now, it’s kind of wild that LastPass represents one of the ‘more affordable options’.

Imagine my surprise when Synology announced a password service, C2 Password in 2021 and I saw their pricing:

I had to do a double-take on the numbers. For just a tiny bit more than the monthly cost of LastPass, I would get a year’s worth of service. Sure, naysayers have/had some very valid points to consider:

• This might be introductory pricing and they will jack the prices up down the road

• They might decide this service isn’t worth it for them and pull the plug

• Can you trust them?

At the end of the day, for that kind of a price drop, it was worth at least trying. It took me about three days fully cut over to C2 and never look back. For $5 a year, it’s pretty unbeatable, no complaints.

But I had a nagging thought:

And it was this goal of saving $5 that set me off down the rabbit hole of self-hosted services…

[Back to Top]

Getting started

Talk about self-hosting usually brings with it a few predictable questions

• What is self-hosting?

  As the name suggests, it refers to running your own services, on your own hardware, on your own terms. While this isn’t an exhaustive, definitive definition, it’s pretty close, I think. A few easy examples:

  • A business might have an ‘on premise’ email server

  • Someone might have their own multiplayer server for a game

  These are pretty clear cut: you can run whatever version of the software you want, with whatever configuration you want, any way you want.

  A few grey-area examples:

  • You pay Amazon for an EC2 instance to run your own services

  • You rent a dedicated game server

  These are a bit grey area because the amount of ultimate control you have may be limited: Amazon may discontinue your server instance, or the game server must run a specific version or configuration of the game.

  Self-hosted can mean different things to different people, but as more and more restrictions are placed on when, how, where etc. you run your software, the more it moves away from the idea of being ‘self hosted’

• Why would you self-host?

  Sometimes phrased as ‘what’s wrong with using <existing service>’,

  • You control all aspects of the service. This can mean different things for different people:

    • You control what version of the service you run. Lots of us have encountered a scenario where a company has rolled out an update we don’t necessarily agree with: either they change some functionality or worse, they remove it. When you self-host, you have some options: you can simply not update, you can skip/disable/bypass the change or you can implement fixes/workarounds independently

    • You control your data: if you want to take a quick backup right now? You can. If you want to purge all traces everything and start from scratch? You can. And if you do (or don’t) want your data to feed AI training models? You’re in control. More on this in a bit.

    • Generally speaking, you control when your service has downtime (if applicable). For example, gamers love it when game companies roll out big patches on Friday night…

  • The cost angle. Services may be affordable now, but [a] they add up and [b] we all know that as soon as they can, the subscription cost will go up. On the note of free services (Google I’m looking at you), not only are we at the mercy of features being paywalled down the road, with ‘free services’, we are the commodity. We are being profiled for ad targeting.

  • Self investment. Suppose there’s a service you’re going to use anyways — something like ChatGPT for example. Without getting into tinfoil hat territory, the more you use their services, the more it benefits them (rightfully so) - but what if you could run your own and thus, benefit your own training pool? While you’re not likely going to outcompete with their resources, this does go a bit towards minimizing your risks when they inevitably change the services (either the functionality or the price, or both)

  At a high level, just looking at the streaming industry over the last decade should give a pretty clear (and dire) picture of how the next decade will roll.

  1. First prices will go up and product segmentation will try and carve up the consumer base to make new ‘norms’.

  2. Next, companies will reduce service, coverage, quality, etc. (and spin it as an effort to be equal to everyone or how it’s ‘no longer viable to support’ — woe is them)

  3. Repeat for more profit.

• But what about...?

  • Self-hosting isn’t all rainbows either: sometimes features can be hidden, removed, changed etc. and while it may (and likely will) be possible to mitigate these changes, not everyone has the time, resources or care to go through the damn hassle

  • Truthfully, if you’re not living and breathing this space, it can be really intimidating and frustrating. I cannot begin to count the number of times I’ve yelled “JUST #&^$%! WORK” while figuring out some nuance about a service

  • Yes, you’re 100% on the hook for your own uptime, data security, etc. If you’re self-hosting for a business, this can mean added costs in the way of fail-over, redundancy etc. There can be a lot of upfront costs that may sting in comparison to a “small monthly fee”

  • You’re 100% on the hook for your own backups too. Obviously this would be covered by the previous bullet, but given how ‘lax a lot of people are with backups, I thought it’d be worth it’s own bullet

  • Worth noting: not everyone’s spouse, family etc., is as open-minded and forgiving as mine. I’m pretty lucky in that respect…

• What can I self-host?

  For a lot of people, thing that run through a web-browser makes for a great place to start (it’s wild the breadth and depth of stuff that can run through a browser). I get a lot of inspiration from lists like

  • Awesome-Selfhosted

  • Awwesome

  • ASH UI

[Back to Top]

My self-hosting setup

At a glance

I started self-hosting via Container Manager (essentially, an implementation of Docker) on my Synology NAS. The first container I ran was Portainer which I use as a wonderful front-end for managing all of my actual containers. There are a ton of Synology-centric guides, but I used a combination of guides from Wundertech and MariusHosting when I got started. After I got Portainer up and running, I went headfirst down the rabbit hole, voraciously trying out new tools left, right and center.

For awhile now, I’ve held steady with around 100 or so containers spread across two Docker hosts. For my second Docker host, I went with Raspberry Pi 5 for a couple reasons:

1. Mostly, so I could test out Frigate using a Coral TPU and,

2. There are more than a few services I can’t run on off Synology NAS due to OS using an old kernel version. For my NAS, the specific error to look out for is ‘failed to get urandom’ or something to related to random-numbers or cryptographic seeding. Thankfully, most of these services that I couldn’t run on the NAS, I was able to run using the Pi (note the architecture difference: x86 vs ARM)

Longer term, I would like to get a dedicated (x86) mini PC to sit above my current desktop to be my primary Docker host. By moving the bulk (or ideally, all) of my containers from the NAS to this higher-performance Docker host, I can realize a few benefits:

1. From time to time, I need to reboot my NAS (say, for patching) — the extra load of firing up a ton of Docker containers on reboot is brutal (it can add 30 minutes to the boot time, ha)

2. I prefer to use my NAS purely as a storage-only device and to let computing happen independently of that

3. A standalone Docker host will offer a ton more performance and, given that I have dedicated connectivity (USB, power, networking, etc.) for this machine directly built into the upper cubby, it’ll be easy to interact with

4. In theory, I should be able to be free from the limitations of old-Linux-kernel

   [Back to Top]

Flow

Looking at the flow and connectivity of my services, I’m setup something like this:

The two big pieces of the puzzle (at least for me) were getting a reverse-proxy and local-DNS in place.

1. Reverse-proxy: think of a reverse-proxy as a switchboard operator for network traffic.

   • This is the system that routes traffic from a domain to the appropriate server (or in this case, the Docker container):

     • serviceA.domain.com → docker-container-A

     • serviceB.domain.com → docker-container-B

     • serviceC.domain.com → docker-container-C

   • You can have multiple incoming URLs point to the same server/container. This can be handy if you haven’t figured out how you want to call your services, or if you want to make a quick test mapping without affecting live mappings

     • serviceA.domain.com → docker-container-A

     • service-A.domain.com → docker-container-A

     • a.domain.com → docker-container-A

     • a-service.domain.com → docker-container-A

   • The reverse proxy can also handle wrapping everything with SSL certificates (handy for getting mitigating the pesky “this site isn’t secure” warnings that browsers throw at you) and can be integrated with authentication services like Authelia. Some/many (but not all) services have integrated user authentication but:

     • You can use a standalone authentication service like Authelia to cover the services that don’t and

     • You could even use Authelia to cover the services that do have their own integrated authentication, giving you the option of disabling the built-in authentication. This would turn Authelia into something of a sorta-single-sign-on.

2. Local DNS: we’re familiar with what DNS is — it’s the magic that converts an IP address like 1.2.3.4 into a more memorable thing like domain.com. Here, we use DNS to make ‘custom shortcuts’ that point to the same domain.

   • serviceA.domain.com → domain.com

   • serviceB.domain.com → domain.com

   • serviceC.domain.com → domain.com

View fullsize

View fullsize

[Back to Top]

Hooking up the reverse-proxy

Over the two services, NGINX Proxy Manager (NPM) running on the NAS (specifically), was the [much] harder of the two — due to Synology quirkiness: Synology claims port 80 and 443 for itself. You can use the built-in Synology Reverse Proxy (which I tried out - but it’s clunky and I didn’t like it). It’s a bit hazy, but at a super high level:

1. Create a MACVLAN that binds to a physical network port (thankfully, my NAS has several)

2. Create a bridge network

3. On the NPM Docker compose, assign both networks

• Creating the MACVLAN

  For me this was the more complicated task to do (this was my starting point).

  At a high level

  1. Enable remote access to your NAS

  2. SSH into your NAS

  3. List your network cards (ifconfig) and pick one (i.e., ovs_eth1). Now,

     • Identify the gateway IP for that NIC (i.e., 192.168.10.1) This will be listed in the ifconfig call.

     • Set an IP range for this network we are creating. For me, I only plan to have NPM on this special network, so I only needed one IP (i.e., 192.168.10.15/32)

     • Pick a name for your network (i.e., mymacvlan)

  4. Create the network:

     sudo docker network create -d macvlan -o parent=ovs_eth1 --gateway=192.168.10.1 --ip-range=192.168.10.15/32mymacvlan

Now, within NPM, you can create, a proxy host for your example service:

• The domain names here line up with whatever CNAMES we’ve created in our local DNS

• The forward IP just needs to match the gateway for our mybridgenetwork

• The port will need to match whatever the port number is for your Docker container

I am, by no means, an expert on this (not even close) and there may be better ways to do things, but this was what ended up working for me after trying a few options and more importantly, a ton of cursing. Your mileage may vary haha.

[Back to Top]

Services spotlight: what kinds of stuff do I host?

Vaultwarden

My original objective had been to self-host a password manager, so this was one of my earlier projects. I had originally set out to try and host Bitwarden but I found that to be a bit more stressful to deploy than I liked so I gave Vaultwarden a spin and it was a dream to get running. Vaultwarden is feature compatible with extensions/apps made for Bitwarden so this was a win all around.

Migrating from C2 was a different story though: there was no easy way to export/import so I had to do things manually. There is a script that promises to help alleviate this pain but I found it wasn’t too bad to do things manually once I got underway - doing things manually was a good opportunity to rotate passwords, increase complexity and where appropriate, close out accounts. All in all it took me a couple days to fully migrate over.

Having the local-desktop-app makes managing your credentials en-mass much more pleasant - your pace is limited only by your keyboard and mouse and not any webserver-refresh/fetch.

Deploying Vaultwarden was a great intro project and allowed me to meet my goal of saving that $5/yr.

• Notes

  • For most users, you’ll want to allow this to be exposed to the outside in some way (with all of the security gotchas that go along with that). A Cloudflare Tunnel setup can be a fairly low-effort and low-stress way to expose this service.

  • When connecting the Bitwarden extension to your server, note that you’ll need to switch the mode drop-down to self-hosted

Actual Budget

For a few years, I’ve used YNAB as a way to track expenses and do budgeting. YNAB wasn’t cheap and for me (at least for a time), there was some value in having transactions auto-import. At some point though, something changed: it could be the banks themselves, the middle-ware responsible for handling imports, or YNAB, frankly, I don’t know but the process of importing because a colossal pain in the butt (mostly stuck on minutes-long loading screens). Between this frustration and YNAB increasing it’s prices, this was a kick in the pants for me to take a look at something — anything else.

I settled on Actual Budget and it was a breath of fresh air - they even have documentation for migrating from YNAB! Having this service run locally where everything is stupid-fast and responsive makes staying on top of transactions very easy (once you get over the initial hump of transactions to import). I made the cut-over right at the end of the year so it was all kinds of perfect to start my Actual Budget journey on January 1st.

Deploying Actual Budget saved me an eye-opening $110/yr (!!) compared to YNAB.

• Notes

  • Actual Budget doesn’t support multiple logins directly yet. For couples/families, you can have a single server/login, but multiple files. Each person sticks to their own file (there’s no connectivity between files).

  • If you absolutely need your files to be private, then you will need to spin up a separate Actual Budget instance for each person (also probably worth looking at Encryption settings)

  • I don’t see a reason to have this exposed to the outside, so for me, this is only hosted internally

Vikunja

Over the years I’ve struggled to find a tool to do task-tracking/to-do and nothing quite had the features I wanted. Google Tasks was kind of serviceable until the forced integration with Calendar and then killing the standalone web-app). The main thing I wanted from my tasks was to be rich: bullets, embedded files, links, sub-tasks, related tasks, categories, historical tracking etc. The overwhelming majority of apps in this space are mobile-first and I explicitly wanted a desktop-first solution — I don’t want to be creating extensive recurring tasks, embedding files etc. using a phone.

For me, Vikunja hit the spot for most things - I suspect for some people, the lack of ‘see all my tasks on a calendar view’ will be a bummer (— although, it’s coming!).

• Notes

  • Something like Vikunja probably benefits from being accessible from outside your network (again, a Cloudflare Tunnel is probably a half-decent way to do this)

  • Vikunja has a mobile app in early development, however, the experience using the PWA is pretty decent (I primarily use via desktop though anyways)

[Back to Top]

Looking ahead

It’s funny, that what started me down this path was a chuckle thought along the lines of ‘heh, I wonder if I can save $5’ and here I am, a year later, with a bazillionty number of containers running locally…

When I initially started, I spun up all of my services using the web view within Portainer. Since then, I’ve deployed source control infrastructure (using Docker no less), so I’ve begun the process of pulling the docker.compose content out from Portainer into source control and transitioning my stacks to pull from git.

In addition to more services to talk about in ‘Services Spotlight’, there are some tangential things I’d like to cover in a future post:

• A closer look at a few ways of getting external access working

• Juggling internal and external access to the same service

• Getting SMTP working

All with a focus on trying to keep costs down where viable.

This post has been a long time in the making — I’ve mostly been hit by writers block (and churn!). Trying to keep this post (and series) from becoming too technical, but at the same time, offering some value in the technical bits… is hard!

[Back to Top]

Product links may be affiliate links: MinMaxGeek may earn a commission on any purchases made via said links without any additional cost to you.

Digging into this a bit further, I found a whole bunch of posts online specifically about the Stream Deck and how using anything but the original cable (implying USB-A) resulted in bad connectivity. Thankfully, USB-C does work, there’s just a small caveat to keep in mind.

Quick Links

• Testing connectivity

• Conclusion

Testing connectivity

This initial test is just there to verify that everything works as expected: I’m using the original cable, the way the manufacturer ‘intended’ for me to connect the device. No surprises here, the Stream Deck is picked up by the computer shortly after plugging in.

Test 1: connecting via USB-A

In my specific configuration at this time, I had a single USB-A port remaining (without resorting to using a hub), so I had hoped to just convert the Stream Deck to USB-C and make use of an extra USB-C port that I had handy. This is where the excitement starts: since the Stream Deck only gets recognized when the cable gets plugged in a specific orientation.

Test 2: converting to USB-C

It would be as bit of a stretch (but not entirely unheard of) where the adapter converting USB-A to USB-C is [partially] defective and as such, only works in a given orientation. Just to rule this out for peace of mind, I tested this again using a straight USB-C to USB-C cable. Sadly, I can verify that again, it only works when plugged in using a specific orientation.

Test 3: Using a straight USB-C cable

[Back to Top]

Conclusion

Despite hearing that USB-C is the best thing ever and works any way you connect it, the fact is, there are definitely some examples where it doesn’t ‘just work’. If you came here because you couldn’t figure out why your Stream Deck isn’t working or is ‘working intermittently’ (because you haven’t been paying super close attention to the orientation of your cable — rightfully so), just try flipping the cable.

Also bonus oddity: you have to flip the connection at the PC-end. Flipping the connection at the Stream Deck end has no impact.

Why Elgato (now Corsair) have done things this way? I have no clue but at least if you’re stuck with a Stream Deck that 'only works some of the time’, at least now you know you’re not alone and how to work around the problem.

[Back to Top]

Product links may be affiliate links: MinMaxGeek may earn a commission on any purchases made via said links without any additional cost to you.

Quick Links

• Starting Point

• Goals and Objectives

• Rough Design

• Step 1 - Revisiting the wall component

• Step 2 - Remaking the bottom panel

• Step 3 - Remaking the middle panel

• Step 4 - Tweaking the top panel

• Step 5 - RGB

• Step 6 - Side panels

• Step 7 - First power up

• Step 8 - Revisiting the drawers

• Step 9 - Speaker mounts

• Step 10 - Magnet all the things

• Step 11 - Misc touches

• Milestone

• Mixups, surprises and lessons learned

• Appendix - Parts List

Starting Point

The original Wall-Desk is still one of my most ambitious projects ever, encompassing a few different hobby sectors and by far and large, the desk is still, very effective. As such, I don’t really want to reinvent the desk so to speak.

The first catalyst for revisiting the wall-desk was actually when I built a pre-levelled platform to compensate for the severe sloping in my office: the unforeseen five-inch platform raised the altitude of my left/right cubbies and by proxy, raised the desk. While I got lucky with the connections at the floor (none of those were blocked by the platform), the new altitude of the desk sides meant that the desk top would now interfere with the wall-mounted keystones and speaker mounts that were attached to the wall (since the wall was not included in the five-inch platform).

The second catalyst came in the form of a new monitor purchase or rather the power brick that the monitor uses — there was no way I was going to be able to neatly mount that behind the monitors, so I needed to find a way to put that somewhere.

As a bare minimum, I probably could have gotten away with rebuilding the side cabinets to lower the altitude of the desk, but this was a good opportunity to revisit some enhancements.

View fullsize

View fullsize

View fullsize

View fullsize

Something cool to see is how my cheap RGB strip fared over the years; with some minor exceptions, I ran the RGB strip at 100% brightness on a single color (blue) for five years. That’s over 40,000 hours if you’re counting. The fact that my strip still turns on and provides something is kind of impressive; but it’s definitely earned it’s well-deserved retirement (or at least to be deployed with a different color). The strip is the Govee H6163 which I’m pretty sure if all kinds of discontinued now.

View fullsize

View fullsize

View fullsize

View fullsize

[Back to Top]

Goals and Objectives

I booked vacation time to work on this, so everything needed to be back up and running ‘enough’ by the time my vacation ran out. I had roughly two weeks to tackle this; I also didn’t want to ‘reinvent the desk’ so to speak: I wanted to reuse as much as I could from the existing desk.

As for the more specific considerations:

• I wanted to be able to level my triple monitors more easily: when you look at the back side of a monitor to see where the VESA mounts are, some (most?) monitors have the mounting point in the dead center of the monitor. With a lot of cheaper monitors (like my side displays at the time), the mounting point sometimes shifts towards the bottom. This shift makes aligning monitors trickier (if you’re using fixed wall mounts). You can use a VESA slider to give yourself some vertical wiggle room, but this is tedious and annoying

• The original wall supported 4x USB-A and 1x USB-C port integrated into it; I wanted to add a whole bunch more USB ports into my wall (without resorting to using hubs); this would allow my to support future PC upgrades

• Based on how I used the wall over the last few years, I wanted to move where some of the connection points were on the desk. In particular

  • I wanted to run the networking down through the top of the faux wall; this would allow me to run the cables from my network rack up into the drop ceiling

  • I wanted to move the rear-speaker connections to the side of the wall, and higher up (to reduce the chaos in the corners)

  • The XLR connection for my microphone could be moved out of sight and routed to the side of the wall (where the boom arm is mounted)

  • I wanted to revisit how I was mounting the speakers to the wall as well

• I wanted more outlets, both on the face and the right-side of the desk. This would also allow me to connect everything to the wall, and then connect the wall to the UPS with a single cord

• On a related note, I wanted to reduce the amount of crap on the desk. I accept that I probably can’t get away from having to account for keyboard, mouse and mousepad but I wanted to find a way to reduce some desk clutter:

  • Capture card

  • SATA dock for backups

  • Google Nest Mini, Amazon Echo Show 5

  • Clock

  • Speaker control pod

  • A label printer

• I wanted to replace the RGB strips with something that I could have local control with (and I wanted to take repair, replacement and upgrading into consideration)

And a big one: I wanted to support an additional PC on this wall (a real computer, not an anemic work laptop). In short, I wanted to add a whole bunch of connectivity to the wall, including supporting a whole new computer and simultaneously reducing the amount of stuff that will live on the desk.

A note about ‘minimalist’ trends

“Reducing desktop clutter” and “minimalism” are trendy buzzwords that influencers like to push around. My drive to reduce the amount of [tech] stuff on the desk has nothing to do with minimalism - I want to reduce the number of devices I need to unplug/relocate if I need to take the entire desk apart for whatever reason.

[Back to Top]

Rough Design

At a high level, I was going to do three things:

1. Change two of the three monitor arms from fixed monitor mounts to gas-shock arms. In theory, I can leave the center monitor on a fixed mount and having adjust-ability on the left-right monitors would allow me to get everything aligned

2. Add a bunch more outlets to the wall

3. Add a boatload more keystone connections

This allows me to keep most of the faux-wall elements — I would mostly be changing out the cutouts on the ‘face’ and the sides of the wall. This should keep the cost of materials down.

It was quite chaotic to look at all of this on a single diagram, so I broke things out by category.

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

[Back to Top]

Step 1 - Revisitng the wall component

When we initially got the office setup, we knew the floor was sloped but we were in a crunch so we shimmed the faux-wall up as best we could and screwed it to the actual wall. When I recently levelled out the floor in my sitting area, I swapped out the shim for a more ‘proper’ levelling foot. Now that I’m going through the effort of rebuilding the wall, this is a good time to just fix it more properly by pre-levelling out the base.

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

This was a good time to cut/drill all the access through-holes for wires I specifically cut everything (in the 2x4s) oversized so I wouldn’t have to think about alignment — when I cut the openings in the plywood ‘skin’, I’ll need to be much more careful.

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

Regarding the screws that we used to initially attach the faux-wall to the actual wall, I didn’t want to make new holes and I didn’t want to wear out the screw holes so I needed to find a way to make this faux-wall (and any additional unit) “compatible” with the existing screw holes as well.

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

[Back to Top]

Step 2 - Remaking the bottom panel

The bottom panel definitely needs to be re-cut as there were substantial changes to the number and placement of keystone panels. I ended up doing it a super hard way by cutting everything out using an oscillating multi-tool; I know you can drill four holes and connect them using a jigsaw, but I simply didn’t trust (myself, the tools, nor the process) to be able to make the cuts super precisely (since this is a finish surface). The overlap of the trim panels I had on hand was something tiny (3/8? 1/4? I’m not sure) and I’ve seen drill bits skate more than that...

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

Spoiler: in my original plan, I was going to run the speaker connections on the right hand side of the wall but at some point, I decided I didn’t want to do that and I would come out the face of the wall as I did previously. So I’m missing two keystone panel cutouts that I would have to make later.

[Back to Top]

Step 3 - Remaking the middle panel

Although most of the middle panel would be ‘hidden’ due to having another panel on top of it, the placement of the keystones required me to make the panel anew. I probably could have figured a way to reuse the existing panel, but I wasn’t going to expend the brainpower to deal with it at the time. Same as previously, I cut everything much more generously because there’s no reason to fight for stupid-exact tolerances on cutouts that won’t be seen in any way.

One change I did have to consider was switching two of the fixed monitor mounts for gas-spring mounts which had a different bolt pattern.

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

Once the woodworking phase was complete, I continued with the wiring phase — the middle plate is the most complicated since there’s a ton of stuff coming in/through it (from below and now, from above too). Thankfully, a Rubbermaid bin was a good height to bring it to a close enough that I could work on it.

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

[Back to Top]

Step 4 - Tweaking the top panel

I was able to reuse the top panel and just make some cutouts to it which saves me a bit of material and time. For some reason (it made sense in my head), I opted not to fully secure the middle (nor top) panels down and just leaned/propped them up. It’s a bit complicated to explain, but I did have to go back and forth with taking the middle and top panel off to fish wires around. As I got closer to the finish line, I did run some temporary screws to hold the middle and top plates in place.

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

Although I could have used the cubby as it was, I opted to knock out the divider and have a large compartment up top. This would afford me plenty of space to put all the tech stuff I wanted to have up top. While I had the cubby on the worktable, I reinforced the microphone mount area as I noticed it sometimes had a bit of flex.

View fullsize

View fullsize

View fullsize

View fullsize

[Back to Top]

Step 5 - RGB

Coming off the LED Slat Wall project, I knew I was so totally done with chopping up strips of RGB up; the size of the faux-wall (roughly 48” across) was a perfect use case for using two 1m RGB strips. Since I was buying individual strips anyways, I opted for a 144 led/m strip. The only gotcha with these strips is whether they will fit inside the aluminum channel I use. Thankfully they do — you do have to carefully push them in there though, there’s probably 0.5mm wiggle room.

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

[Back to Top]

Step 6 - Side panels

The side panels are the last bit of the desk that needed to be done to finish the ‘woodworking’ component. Due to the way I made things with the Mk I, the 48” span of the panels does not run flush with the frame. In the Mk I design, I just screwed it down regardless, but I like to think I’m a bit more refined now. With the Mk II, I routed out a gap at the front so that if there is a bit of overhang on the panels, they wont interfere with the side panels and I can still screw the side panels flat to the frame.

View fullsize

View fullsize

View fullsize

View fullsize

[Back to Top]

Step 7 - First power up

The downside about this wall project is that you can’t really test connectivity until everything is already in place; or at least, you can’t do so easily. This initial power up is just to verify the core bits work before I look at rebuilding vs modifying the desk portion.

I ended up getting really close with the placement of the keystone panels on the middle plate. With the cabinets the way they were now, there was maybe 1/4” interference between the plate and the desk — totally serviceable, but I was planning on redoing the cabinets anyways (or at the least, revisiting them).

View fullsize

View fullsize

View fullsize

View fullsize

[Back to Top]

Step 8 - Revisiting the drawers

I debated rebuilding the drawers from scratch, but truthfully, I didn’t feel like it yet — I wanted to use the setup for a bit (especially get through a PC upgrade) before revisiting a total rebuild. I opted to take the cabinet apart, shorten it a bunch and put it back together. How hard could it be?

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

The drawer aspect was extra difficult because I was cramming the same amount of drawers into (substantially) less space. I was able to get things to fit but just barely.

[Back to Top]

Step 9 - Speaker mounts

My previous DIY speaker mounts were okay but they had a few problems with them

• They weren’t beefy enough so they were splitting at the connection to the wall (this could have very well been low quality plywood I was using at the time)

• While I was able to adjust left-right, it was a bit cumbersome

• I wasn’t able to adjust up-down at all

I ended up just buying a few cheap speaker mounts and I also had a similar set on hand that I picked up from a surplus store. For this phase, I only needed to mount the Front Left, Center and Front Right speakers to the wall.

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

For the rear speakers, I specifically didn’t want to have speaker stands that I would inevitably knock over or bump into; previously I used a drawer slide to have a way to retract the rear speakers but I wanted to try something different this time around. I tried using a pair of low-profile microphone boom arms: these boom arms came with a 3/8 stud on them — luckily I happened to have some 3/8 T-nuts so I was able to adapt my previous speaker mounts to them. The boom arm is rated for 3lbs — my rear speakers are a bit under 2lbs so, on paper, it should hold. Giving it a bit of a ‘push test’, the arms seem to be okay for now, but time will tell.

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

[Back to Top]

Step 10 - Magnet all the things

I’ve been using magnetic charging cables for our for a long time now: it’s been wonderful to not have to think about whether something is Micro-USB or USB-C — it’s the same magnetic charging connector for both. A special bonus is that the magnetic cables handle power and data.

I ended up getting a Logitech G915 (linear) - more to try it than anything else - I’m still very much in love with my G510 (but I do recognize that my stockpile of backup G510s will eventually run out one day). I do like having a wireless keyboard — not because I’m opposed to wires, but the wire on my keyboard is so long (and it’s not detachable); the distance between the end of the keyboard and my keystone is 18” so having a 6-ft cord is just an exercise in frustration.

The way I have things setup, I connect the keyboard via wireless and I keep a magnetic adapter plugged into the back. When the battery runs low, I just connect it to a magnetic charge cable I keep around my desk anyways. Since the charge cable is power-only, there’s no transition between wired and wireless — the keyboard continues to work using the wireless dongle while charging.

View fullsize

View fullsize

I also opted to put a magnetic adapter on my charging mouse-pad. Since the mouse-pad itself can act as a mouse receiver, I have the option to either

• Connect the mouse-pad to the computer using a data link; the mouse connects to the mouse-pad or

• Connect the mouse-pad purely to power; the mouse then connect to the computer using it’s original nano-receiver

For the most part, there’s no functional difference (connectivity wise) between these two options. The change for me is that I have a few more options on how to connect the mouse-pad to power source (and reduce the number of cables draping across the desk). Being a magnetic cable, I can just take the mouse and mouse-pad in one go.

View fullsize

View fullsize

For my Stream Deck, it’s a bit more of nuanced implementation; in my specific case, I wanted to connect the Stream Deck using USB-C (the cable that ships with it has USB-A on the computer-side). I paired a USB-A to USB-C adapter along with the magnetic adapters to make the whole thing quickly detachable. It was a bit of work to stuff the long cord into the included mount, but I got it in there.

[Back to Top]

Step 11 - Misc touches

One of the goals of this rebuild was to reduce the amount of stuff on the desk that I would have to carefully unplug/move if I ever needed to remove the top to access the bottom/back of the desk. For the time being, I opted to shelve the Amazon Echo Show 5 so that still leaves the Google Nest Hub. From some scrap plywood and some magnets, I just attached this to the network cart, just at the edge of my periphery (I suppose I could also make a magnetic mount for the Show 5 if I ever decided to revisit that).

View fullsize

View fullsize

View fullsize

The desk features a big access panel with an internal shelf, at this time, I keep just a monitor brick, my WLED controller and a USB capture dongle. On one side, the capture dongle connects to the USB feeder ports in the bottom left and up at the desk level, it transitions to an HDMI port. This way I can directly connect a camera to the wall.

View fullsize

View fullsize

I also added a remote [physical] power switch to my desk; at this time it’s linked to a fan I have pointed at me, but in the future I might consider getting an additional one to remotely trigger a light or something. Depending on how often I may bump into this with my armrest, I may adjust the altitude of the switch, but for now it’s working out great.

View fullsize

View fullsize

View fullsize

I normally keep some kleenex in one of the segments of the upper cubby but since I removed a divider, I lost the allocated space for that. I made a quick tissue box holder from some scraps and mounted it up top.

The speaker control pod was the last thing to be mounted up: I debated making a quick shelf for it off to the side, but I do actually need to have reasonable access to it to switch the inputs and possibly to turn it on/off. I ended up using some steel strapping and some VHB tape to keep it secured right overhead.

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

[Back to Top]

Milestone

All in all, I think this desk rebuild turned out well and has breathed new life into my office. I do have to give credit to my wife though, for putting up with the absolute disaster zone that was the office for a week while this was under construction. In the next month or two, I’ll be revisiting her wall-desk and levelling her floor area in one go. Thankfully her wall-desk is a bit less involved cable-wise, so it should go by much easier.

No KVM or USB switcher

I specifically did not want to include any kind of KVM or USB switcher to toggle between the three machines. From PC-A to PC-B, I only need to migrate the two nano-recievers for my mouse and keyboard and the work laptop can connect via Bluetooth (keyboard) and VNC. Remoting is also an option to pass keyboard/mouse through as well. For PC-A and PC-B, the monitors are directly connected.

Reducing Clutter

Reducing the number of ‘permanent residents’ on my desk was one of the motivators for this rebuild: I started out with nine or so devices that were parked on my desk regularly:

• Keyboard

• Mouse

• Mouse pad

• Stream Deck

• Google Nest

• Amazon Echo Show 5

• Label printer

• SATA/USB Dock

• Speaker control pod

After the Mk II rebuild, I’ve got this number down to to just four: keyboard, mouse, mouse pad and Stream Deck. Of these four devices, only mouse pad and Stream Deck require ‘unplugging’ before moving them from the desk but given that they are connected by magnetic cables, it hardly registers.

What’s in store for the Mk III?

• Looking ahead, I will almost certainly be rebuilding the side drawers once I settle on the next system build. I will certainly be swapping out the stupid soft-close drawers for traditional, undampened drawer slides.

• Expanding on the idea of connecting the keyboard and mouse pad with power-only connection, I’d like to explore alternative ways of getting USB-power to these devices without having cables strewn across the desk (perhaps a strategically placed USB grommet).

• I would also like to have more ‘hidden USB keystones’; by hidden I mean, that they can be in the upper cubby, be placed behind the monitors or even below the desk — dongles for my keyboard, mouse and wireless headset don’t actually need to be visible

• You might have noticed that the monitors aren’t fully aligned with each other — this is because the VESA mount on the right monitor is located near the bottom of the monitor. This seems to be a fairly common thing with ultra-cheap monitors. In comparison, the VESA mount location for my two other monitors is located dead-center. I thought that the gas strut monitor mount would allow me to compensate for this — and it does — but I’ve maxed out the declination and it’s still out of alignment. I’ve been low-key planning another monitor upgrade this year so maybe that problem will sort itself out.

[Back to Top]

Mixups, surprises and lessons learned

This was a surprisingly uneventful project with only a few minor setbacks. The first was having a WLED control brick itself during an update (thankfully, I hadn’t gone down the rabbit hole of configuring things yet - this was just during initial setup).

The second was a bit more annoying: ultimately, I discovered that a fairly new network cable had a short in it. In hindsight, I definitely noticed something was off months ago, just never thought to check the cable itself (since it was relatively new and not abused). Thankfully, the network cable in question was outside of the wall (it was one of the feeder lines providing connectivity to the wall) so it was an easy thing to replace.

This brings up an interesting point: it is really challenging to ‘test as you go’ with a build like this. For a lot of it, I have to be nearly completed before I can really do any testing. I suppose I could bust out a laptop to do some spot checking of keystones as I go but really, a keystones generally ‘just work’.

Testing some connections, like the capture dongle require testing multiple jumps too and given the chaotic nature of the build, it’s not entirely viable.

Thankfully, I actually had no issues with any of the connections within the wall.

[Back to Top]

Appendix - Parts list

To make things even more complicated, the cables noted below are self-contained within the wall-desk — each keystone endpoint will require an additional cable in order to connect that port to the corresponding port on a computer to ‘feed’ it. Think of the wall as a docking station for the computer.

Networking

• 8x CAT6 keystones — I happened to buy a few pre-loaded wall-plates that came with them

• 4x runs of CAT6 inside the wall

Audio

• 8x 3.5mm stereo keystones (6x for both ends of 5.1 signal, 2x for both ends of a stereo signal)

• 1x 5-speaker binding post plate (to connect the surround speakers)

• 2x RCA keystones (you don’t need different colors, but it helps)

• 1x Female Stereo to 2-RCA + 1x 3.5mm stereo cable with one right-angle (this allows us to connect the stereo signal to the second input on the Z906)

• 3x 3.5mm stereo cable with two right-angle (this allows us to connect a 5.1 signal to the primary input on the Z906)

• 1x Right-angle XLR adapter

• 2x XLR wall plates (one male, one female) — you can get something like this and separate them

• 1x XLR cable with one right-angle on the male or female end (the specifics depend on what gear you’re trying to connect together

• 1x cheapo microphone boom arm

• 20x Banana plugs (need them for each speaker as well as for the connection to the wall)

• 10x Banana plugs for spring loaded inputs (used on the speaker end)

• 5x Round 2-way speaker box terminal

• 3x Cheap speaker mounts

• 2x Low-profile microphone arms

• 14AWG speaker wire

USB

• An ungodly number of keystones, 30x USB-A and 10x USB-C if I‘m counting correctly

• 15x USB A-to-A cables

• 1x USB C-to-C 20Gbs/DisplayPort-AltMode (this is used as a video cable)

• 4x USB C-to-C cables (you could get five of the previous but these are cheaper since we don’t need the enhanced video spec)

• USB capture card (can go with a cheap unit or something fancier)

Video

• 3x HDMI (2.1 or 2.0) keystones

• 3x HDMI (2.1 or 2.0) cables

• 3x DisplayPort 1.4 cables

• 1x Fixed monitor mount

• 2x Gas strut monitor arm

• VESA vertical adjustment adapter (if needed)

• Right angle DisplayPort adapter (if needed)

RGB

• 2 WS2812B RGB Strips (144 led/m)

• 1x WLED controller

• 1x 10A power brick

• 18AWG RGB extension wire (if needed)

• 2x Aluminum channel with diffuser

Power

• 1x Inline extension switch cord

• 2x Desk power bar with USB

Misc

• Large wall mounted USB clock

• 6-port wall plates

• Blank keystone insert

• 1x Access Panel (8x8)

• 3x Cable pass through wall plate

• 10+ Low Voltage boxes

Two Special Call Outs

There are two things that definitely made this project so much easier to tackle: the Milwaukee M12 Installation Drill/Driver and CordIDs from Dotz. The former is pretty straightforward: an ultra compact and versatile drill/driver that excels in ‘close quarters’ actions. The latter provides a means of labelling the tires but due to their clip-on/clip-off nature, provides you with three game changing benefits:

1. You can easily flip the clip so all your labels are oriented the same way, you can also rotate the clip to show either side

2. You can clip them onto cords after you’ve already wired them if needed (so you can focus on one thing at a time)

3. The smooth plastic surface means that labels both stay on there if you want them to but can be peeled off if you need to update it

I used probably close to a full bulk pack (100) of these CordIDs on this wall; at the very least, each wire in the wall gets two labels (one at each end) with some additional ones getting two more (in the case of video cables, they are labelled on both ends, inside and outside of the wall. It’s so nice to not have to guess what a given wire is — it’s labelled.

[Back to Top]

Product links may be affiliate links: MinMaxGeek may earn a commission on any purchases made via said links without any additional cost to you.

Networking Series

This post is part of a series, check out the other posts!

Quick Links

• Starting point

• New router - TP-Link ER8411

• New access points

• Dual DNS

  • Redundancy

  • Syncing and Caching

• Looking ahead

Starting point

In my previous update, the focus was on establishing a secondary, low-power network that I could run during an extended power outage (say, an 8 hour work day). With a sub-30W footprint, this isn’t a hard target to hit at all. I actually have some a process in place to be able to calmly transition the network rack from utility-power to use battery-based ‘generator’ to support longer outages (because, priorities right?).

Over the last year, I’ve made a few hardware upgrades and there are a few specific problems that I’d like to tackle. In no particular order:

• I ended upgrading to a 10G router. I don’t need it yet (ISP capped at 1G symmetric), but when the times comes for faster-than-gigabit, I’d like to have everything already in place to go. There is a benefit that I can realize now — the vastly improved VPN performance

• My outdoor AP sort of died so I used the opportunity to swap it out for a WiFi6 model

• During the summer, there was a one day sale on a high-density AP so I bought that to act as my new ‘main’ AP and transitioned my current one to be dedicated to my low power network

• The batteries in one of my UPS units hit their end of life which got me thinking (more on that later)

[Back to Top]

New router - TP-Link ER8411

I’ve wanted to upgrade my router for a while now. Originally it was to get ahead of a potential service upgrade from our ISP but more recently, I’ve wanted make better use of my current connection. In particular, I wanted to make better use of my available upload speed through more modern VPNs (OpenVPN, SSL, WireGuard, etc). We can see a monumental performance uplift when looking at the 10G router.

In a vacuum, the ER707-M2 is actually an interesting choice: it gets me 2.5G WAN/LAN capability and a respectable performance numbers all around (in particular, OpenVPN performance) without breaking the bank too badly. However, since I’m coming from the ER7206, the performance jump is less impressive and so, the 10G ER8411, even with the massive price tag, starts to make a bit more sense as a meaningful upgrade.

This upgrade has one additional benefit: the ER7206 is not naively rackable: I do have a 3d-printed rack-cage for the ER7206. One catch is that the specific 3d-printed mount I ordered was bigger than 1U, which threw my spacing out of whack (and this has been low-key bothering me since I got it racked up). It’s an OCD thing.

During the upgrade, I took the opportunity to try and eliminate as many SFP+ to RJ45 adapters I had in use as I could, swapping them out for DAC cables. Why? Because the DAC cables don’t need to perform any signal conversion. This has the benefit of having the least amount of latency but also the least amount of power consumption and heat. Before I started racking up networking gear, I didn’t think that it would make a meaningful difference, but seeing the temperature difference with DAC cables is eye-opening:

On the ER8411, the 10G WAN port is SFP+, so if my ISP ever provides faster-than-gigabit, I will likely need to run a transceiver; but for now, I opted to use one of the 1G WAN ports and connect to the modem via CAT6. While a single module isn’t going to make a huge dent on the room temperature, if I can avoid extra heat, I might as well.

To make the [future] transition easier, I included the SFP+ when making router-rules; so in the future, I should just have to worry about switching the physical connection and not having to update any of my ACLs.

Why not something like OPNSense?

This is something I’d like to revisit down the road but at this time, truthfully, it’s not in my current skill set and comfort zone. From a hardware management perspective, I wanted all of this to be accessible from a single portal/app. Perhaps, down the road, when I outgrow this, I may look at something for serious.

[Back to Top]

New access points

My outdoor AP started to develop some weird behavior where it would broadcast SSIDs and allow devices to connect but would absolutely refuse to provide a connection to the internet. Indirectly, this failure ended up being a selling point for having everything on physical switches — Nicole was able to power down the AP without fuss (and thus, connect to any of the other APs that were working correctly). I did not want to deal with calling into support to troubleshoot the issue, so I just opted to replace this with my first WiFi6 AP, the EAP650 Outdoor.

Over the summer, there was a sale on the EAP 620HD and even though I was really pining to get the EAP 690e HD, I just couldn’t say no to the price. The model I received was a V2 (rather than the latest, V3) so I missed out on two significant improvements:

• Approximately 60% smaller footprint (290 sq-in vs 125 sq-in)

• The V3 can take passive-POE and includes an injector; the V2 can only take active-POE and doesn’t include one

Thankfully for my scenario, neither of these downsides really affects me: I have the AP mounted in the furnace room, out of sight and for POE, I picked up a cheap TRENDnet POE Injector. With this new ‘main’ AP, I was able to transition the previous AP to my low-power Endurance network. I did this by ‘forgetting’ the previous AP from the control panel and then configuring it using standalone mode. In standalone mode, I configured the same SSID/password combination so things should mostly work as expected when cutting over.

This leaves me with three remaining WiFi-5 APs (all EAP-235Wall units) and barring a fantastical sale of some sort, I don’t see myself upgrading any time soon.

[Back to Top]

Dual DNS

Redundancy

A few updates ago, I set up an ad-blocking DNS server on my network and it’s been pretty great but I’m always super wary about the robustness of the Raspberry Pi (or maybe it’s Linux, lol). Everything works great — until it inexplicably doesn’t and it’s not a quick fix because you can’t even remote into the machine because it’s got a kernel panic. In fairness (and definitely hindsight), I suspect that the kernel panic was caused by ‘not enough power being delivered’:

1. There isn’t really a nice and easy way to verify that now and

2. Frankly, I don’t really want to deal with it

3. Outages are never at a convenient time!

To mitigate this, I bought a second Raspberry Pi — I went with an identical RPi4 unit to what already had. The plan was to just clone my existing system image, tweak a few things to change ‘Pi1’ to ‘Pi2’ and that would be the end of it. For once, I am super happy to say the process went exactly that smoothly.

When I rebuilt the network rack, I made to sure add USB ports (connected to each Pi) to the front side of the rack. This way I can do fairly easy system backups by just connecting a card reader or SATA-USB dock - I keep two full backups on standby. In the event of a device failure, it’s a matter of just unplugging the existing boot drive and plugging in the new one.

View fullsize

View fullsize

View fullsize

[Back to Top]

Syncing and Caching

Once both Pi units are up and running, it was simply a matter of having the primary and secondary DNS servers for my whole network point at them. You could stop here and be done: it’s totally possible to manage both Pi-Hole instances individually but it would be much nicer to be able to have [config] changes automatically sync between them. This is where Gravity Sync comes into play.

Gravity Sync was super easy to setup and verify; it sync’s blocking rules, exceptions, local DNS settings - everything I would want it to do. I’m still responsible for doing updates on the Pi units but one nice thing about having the extra Pi, is I can stagger my updates: I can update Pi1 more often compared to Pi2 with the safety net that if an update goes wrong, I have the Pi2 to carry the network while I restore a backup.

One thing I did learn was that ‘primary and secondary’ DNS servers are automatically load-balanced (to a degree) by the router. I had previously thought that the secondary was really a hot-spare and only really came into play when something went wrong with the primary. The load-balancing definitely favors the primary, but the secondary certainly doesn’t sit idly by.

While I was in the head-space to be working with this, I opted to setup a recursive DNS on one of my Pi units - I figured that if something went wrong with the recursive DNS lookup, at least the other Pi can perform ‘normal’ look-ups. The motivation here was less "'privacy concerns’ but more ‘self-reliance’ and the possibility that the local-recursive DNS server would be faster (with the caveat that the first unique inquiry would be slower).

I don’t know if there any conclusions I can draw from any of this — it’s just neat to see! Obviously, Pi-Hole does have some localized caching but there is a non-insignificant number of queries that it has to ‘reach out to Cloudflare’ to resolve.

[Back to Top]

Looking ahead

I’m in a holding pattern, waiting for some hardware to come in which should sort out the last bit of my network stack. Aside from that, I need to spend a bunch of time handling the configuration aspect of everything:

• IP address management, documentation and diagramming

• review of access rules, permissions, etc.

• review of backup strategy

On a related tangent, I’ve started exploring the rabbit hole that is self-hosting services…

[Back to Top]

Product links may be affiliate links: MinMaxGeek may earn a commission on any purchases made via said links without any additional cost to you.

Quick Links

• Setting the stage

• Power outage!

• Blinking status!

Setting the stage

I have two UPS units — the long term plan is to ‘combine’ them into an Automatic Transfer Switch (ATS) so that I can fail over to a secondary UPS in case one fails and also for ~double the runtime in the event that there is an outage. I’m still waiting for the ATS to come in. When I previously rebuilt the network cart, I consolidated the wiring for my NAS units — both units plug into a single compact power bar and that power bar plugs into the UPS. Something like this:

In normal every operation, everything works the way you would expect it to:

• Both NAS units obviously receive power

• The USB connection allows the DS1817+ to get a reading of how much battery life remains and the NAS can self-power down before the battery on the UPS drains

• The DS1817+ acts as a network UPS server for the DS920+ allowing the DS920+ to also power down before we fully drain the battery on the UPS

[Back to Top]

Power outage!

We recently had a two-hour power outage — just as Nicole was getting started on an important call (of course). When it was obvious that this was going to be a longer-than-usual power outage, there were a few things I could do to extend my runtime:

• Power down NAS units

• Power down extra networking gear (POE equipment, extra APs, switches that aren’t being used etc.)

With the easy things powered down, my UPS was reporting about 1-hr runtime. Thankfully there was a quick break in the call that allowed me to fully transition over to Endurance-Mode. Running on the Endurance network stack, we were good for roughly 10-hours of uptime.

On the power/UPS front, I moved the power bar from UPS A to UPS B, so things looked something like this (if you’re clever, you can see the problem already).

[Back to Top]

Blinking status!

After the utility power was restored, I decided it might be a good idea to continue using UPS B to power the network stack: this way I get a bit more even ‘wear’ on the circuitry (both UPS models are identical so they can handle the load). I powered down the Endurance stack and turned the main network stack on.

Once the network came up, I powered on both NAS units and this was when things got strange:

• I could see both NAS units get their statically assigned IPs

• I could ping both NAS units

• But I could not connect to the management portal either by hostname nor by IP

• Using the ‘find NAS’ methods got me nothing.

A clue from earlier…

• At the start of the power outage, I powered down the DS1817+ and then the DS920+

• The DS1817+ powered down without a fuss, but the DS920+ seemed to not want to power down. I tried to remote into management interface, but it wouldn’t connect

• At the time, I wasn’t in a network-connectivity-troubleshooting mode, so I didn’t think about it much — but the symptoms were the same as I was seeing now: I could ping the NAS IP, but for some reason, just could not connect to the portal by any means (I ended up performing a hard-shutdown of the DS920+)

Looking at the NAS units though, I noticed that the status light was blinking. On both of them.

Synology’s help page (as of November 21, 2022) tells us that STATUS: Blinking Green means one of three things:

1. No drive installed, or

2. DSM not installed, or

3. ‘Configuration lost’

All three possibilities are quite alarming but it turns out that the actual root cause in my scenario is much more comical.

The root cause was forgetting to move the USB cable over! When I properly moved the USB cable form UPS A to UPS B, both NAS units came online and everything was running smoothly as one would expect.

When I initially moved everything from UPS A to UPS B, the original plan was to move everything back after the power was restored and had I done that, none of this excitement would have happened. Even still, this is still a bit perplexing:

• The USB cable was plugged into UPS A — which was powered down

  • How could the NAS possibly know that the other end of the USB cable was a UPS?

  • I would have expected the DS1817+ to boot up, thinking that there was no UPS connected

  • Similarly, I would have expected the DS920+ (using the DS1817+ as a UPS server) to also boot up, thinking there was no UPS connected

• In the event that neither of them boot up, I would not expect them to respond to network pings

What about that clue from earlier?

It looks like when I powered down the DS1817+ first, it took the UPS server down with it. This leaves the DS920+ in a state where it’s expecting to have a UPS server (that is no longer there) and the NAS doesn’t know what to do and then gets stuck in a limbo state.

Lesson learned: power down all of the dependant NAS units first!

[Back to Top]

Product links may be affiliate links: MinMaxGeek may earn a commission on any purchases made via said links without any additional cost to you.

Quick Links

• Starting Point

• Rough Design

• Step 1 - Switching out the base

• Step 2 - UPS enclosure

• Step 3 - Handle bar

• Step 4 - The rear patch panel

• Step 5 - Making my own rack mount

• Step 6 - POE Injector mount

• Step 7 - NAS shelf, finishing touches

• Milestone

• Mixups, Surprises and Lessons Learned

• Appendix - Parts and equipment

Starting point

Before embarking on an upgrade/rebuild, I wanted to highlight my top three favorite features of the Mk II rack

• All of the connections were self-documenting: by using CordIDs, I can unplug any and all of my cables and know at glance which device and which port (on the patch panel) it should be plugged into.

• Almost all of the components have individual (analog) power switches associated with them. Being able to manually power down elements of your network is incredibly handy if you’re dealing with a power outage scenario (and need to cut the power footprint to maximize UPS runtime) and it’s also way faster to just walk up to a device and toggle the power switch versus logging into a management interface and telling the device to reboot

• The steel side panels allowed me to magnetically attach everything to the rack (both on the inside and outside surfaces). The top-plate is also magnetically attached, so it’s super easy to pop the top off and I can then snake around inside the rack

Of course it’s not without some downsides.

• I had some doubts about the stability of using a washing-machine mobile base to support everything on a very sloped floor. Specifically, I was concerned about flexing on the base plate

• I really, really hate the second PDU I added to the rack. On a day-to-day basis, it bothered the hell out of me seeing the 1.25U PDU which then reminded me of the dumb design where rack-mounting was clearly an afterthought.

• So all of my cables are connected to the patch panel through the back side (like they would be in a traditional rack). But as I use this more as a cart, this is problematic: [1] anytime I want to swap something out, I have to snake my hands around on the inside to disconnect the device and [2] if I want to disconnect the entire cart, I have to take everything apart for the same reason

[Back to Top]

Rough design

I had three broad considerations I wanted to keep in mind for the new cart:

• Redo the base to remove any concern about sagging or flexing in the base plate

• Try to keep the overall height of the network rack down if I can - I park the network cart in front of a cabinet with the breaker panel and it would be fantastic to be able to access the breaker panel without having to move the cart

• Make room for two UPS units and where possible, consider any room for a future NAS unit

View fullsize

View fullsize

View fullsize

I really liked the idea of Option 2 but it would definitely make connecting the network elements together tricky — even if I punched a hole through the two sides connecting them. I ultimately went with Option 3 because I wasn’t sold on being able to fit everything I wanted now (and in the future) into a small rack.

The big change would be adding rear-facing patch panels and permanently connecting them to the front-facing patch panels. This way, I don’t have to snake around inside the rack whenever I want to remove a device. It turns the back side of the network rack into a dock.

[Back to Top]

Step 1 - Switching out the base

The previous base was a panel of 1/2” plywood that I ‘kind of’ clamped onto an adjustable base and while I don’t have any overt reason not to trust it, over time, I just ‘had a feeling’ that due to the uncentered weight was causing the wood to deflect and eventually result in tears.

To replace this, I had a universal tool base that I’ve had kicking around for a long time (and truthfully, have been tripping over). To try and get ahead of deflection, I opted to make give the base a bit of a torsion-box. Being designed as a mobile base for heavy workshop tools, it was easy to bolt this platform to the base.

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

For actually bolting the platform to the mobile base, I used 2” machine screws and star knobs. At some point down the road, I may swap them out for a pair of nuts but for now, the star knobs are great for being able to quickly get the board on and off.

[Back to Top]

Step 2 - UPS enclosure

I laminated some 3/4 ply together to make some side panels; the idea is that the side panels of the rack will rest on these, transferring weight to the base. I connected a couple quick stretchers to keep the two panes co-planar. During test fitting, I realized that one of the braces blocked the power cord for the UPS so I drilled some access holes to be able to feed the plug through it.

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

[Back to Top]

Step 3 - Handle bar

One of tricky things of the old base is there was no easy spot to grab anything in order to move the cart. I wanted something I use to manoeuvre the cart around. I was going with a pair of wall-mount racks and wanted a solid, physical way to transfer the weight load to the ground — I am not going solely rely on anchoring the racks to my vertical back posts — I would prefer to use that more for ‘alignment’.

I raided the scrap bin to get some plywood to make a ‘bushing’ to separate the two racks - yet provide a continuous, solid contact surface to transfer weight down.

View fullsize

View fullsize

View fullsize

View fullsize

As you may have noticed on the previous step (in the final picture), the rack has a support brace on the top side — which meant that I had to cut a small relief on the side strips so that everything would lay flush. After connecting up the handle bar, I ran a round-over bit make the handle bar aspect of this glorified spacer, nicer to touch.

I didn’t get any photos of it but using 3/4 OSB, I laminated together a frame (like in my previous rack build). Even though the weight of the rack is directly transferred downward, the frame gives me something to bolt everything to, if only to keep the rack from sliding. I had to wait until I had the handle bar together so I could stack the second rack — in order to get the final height needed for the back support frame.

[Back to Top]

Step 4 - The rear patch panel

I bought a small 2U wall-mounted rack in order to mount the back-side patch panel. I had originally had a vision of recessing this into the frame (somehow!). By recessing this into the frame, the front and the back patch panels would be closer - meaning I could use 2-ft cables to connect them. Logistically, I couldn’t think of a clever way to recess the panel (without adding more height), so now, the back patch panel was 4-inches further away than I had originally intended, which was a bit of a problem since I had already bought 2-ft cables…

[Back to Top]

Step 5 - Making my own rack mount

Since it was just a matter of a couple inches of clearance (and I had nothing to lose), I took a side quest to try and make my own rack-mount. Using scrap plywood and the rack-mount I already had handy, I carefully marked out the hole locations and then drilled out the holes using a drill press. I debated using threaded inserts to then hold the bolts captive (turning them into studs), but I didn’t want to risk tear out with the holes so close and honestly, for a 2U rack like this, it’s not a big deal.

The custom rack route worked great, allowing me to get the back side of the rear patch-panel close enough that I could use my 2-ft cables . Of course, had I anticipated this, I would have considered 3-ft cables and avoided all of this drama to begin with.

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

[Back to Top]

Step 6 - POE Injector mount

Previously, I had mounted the POE injectors one at a time, independently because, for my main AP, I was just using a POE injector I keep around for testing purposes, not the injector I wanted to use long term. This was a good opportunity to get around to getting a dedicated injector and mounting everything on a common plate. I kept the injector mount for the low-power AP separate because [1] ran out of room on the mounting plate and [2] thematically, it’s kind of its own thing anyways. I did opt to directly screw the low-power AP-injector into the back frame though.

[Back to Top]

Step 7 - NAS shelf, finishing touches

For the NAS shelf, I opted to dismantle my previous two-tier shelf and reuse it here, rotated sideways. I also opted to route a power bar for the NAS units so that I wouldn’t have to snake the around inside the network rack to access the power. The extra few inches to get to power doesn’t sound like much, but I have an upcoming change with the way power is handled in the rack so I wanted to get ahead of it this way.

While I was here, I took the time to finally mount everything down to the base plate! Until now, everything was just relying on mass/friction to not really move. I had taken the time to drill pocket holes to secure things down but due to a comedy of errors, every single one of the pocket holes had been covered up by something else and I was just too lazy to undo it — so I went with some angle brackets I had lying around.

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

View fullsize

[Back to Top]

Milestone

This was a fairly quick project as I was able to leverage existing pieces (the NAS shelf) and the fact that all of the cables are individually labeled (for both ends) made it easy to be able to switch between “powering everything down to work on it” and “okay we need the network up because the weekend is over”.

I have to thank my wife for putting up with zero-network-at-all for hours at a time while I tinkered with this. Oh, and the massive puke of cables and tech that sprawled in the basement as I took everything apart ha-ha.

In the near future, I might buy a pair of Noctua 200mm USB fans to act as top-exhaust. For the time being (and for the last few years) I’ve not had any temperature issues with the rack mostly-closed up. If there ever is a heat issue, I can simply remove the top plate as it is just magnetized down to the rack.

At the moment, I have nine empty bays which sounds like a ton but I imagine once I take a closer look at what I can rack mount, they will fill up fast.

[Back to Top]

Mixups, Surprises and Lessons Learned

There were three notable screw-ups

1. I intentionally drilled some pocket holes so that I could secure the UPS enclosure down to the platform. I also drilled pocket holes on the back support from (which is secured to the UPS enclosure) to secure that to the platform. I managed to face the pocket holes for both into each other and thus, had to resort to using angle brackets to actually tack the network stack down

2. For the UPS enclosure, I really should have opened up the cable pass through holes with a bandsaw/jigsaw to reduce future headache. As it stands, I can get power cords in and out of the enclosure but it’s a bit annoying.

3. The last careless mistake was not drilling the holes for the magnets before assembling the shelf for holding the POE injectors. Another careless oopsie.

The only other hurdle was the issue of the rear rack-mount using up too much of the limited 2-ft length span I had planned for the wiring, something that could have been avoided with 3-ft wiring or just DIYing the rear rack.

For the mobile tool base, I’m simultaneously really happy and really annoyed with it: it’s quite sturdy, well built and I love the fact that it has adjustable sizing but I hate, hate, hate the fast that the back wheels are fixed. I have this same hatred for rolling tool chests as well. Having fixed real wheels isn’t an issue in a large automotive shop but in any smaller/home scenario, I want to be able to shuffle and snake the tool chest (or network cart) in small tight spaces. Instead of just scooting the cart over, I have to do the game of snaking it back and forth incrementally. Honestly, if I see the mobile base go on super sale again, I might legit consider buying another one just to cannibalize the front wheels to mount on the back of this network cart. Totally infuriating.

Regarding the rack-mount cable concealer, I’m kind of torn. On the one hand, I really do like that it gets the random jumble of cables tidied up but at the same time, I really miss being able to directly see the CordIDs for each cable. Thankfully it comes with a quick-snap panel, so if I needed to see the label, I could do so quickly enough so for now, this is just an irk for me.

[Back to Top]

Appendix - Parts and equipment

• Racks

  • StarTech 12U Wall-mount Rack. Since I already had a 12U, it made sense for me to go with another one as one of my candidate designs had them stacked side by side (thus, requiring the same rack). The only thing to note is that the brace bar on the top can interfere with all but the most shallow devices (read: small PDUs). In my case, I was planning on leaving the top as an air-gap anyways, but something to be aware of. Otherwise, fantastic rack mount, no complaints.

  • StarTech 2U (Hinged) Wall-mount Rack. Even under the hypothetical assumption that I had bought 3-ft cables and could use this, I’m kind of torn on this product. Like it’s fine, it does the job but the hinge is not confidence inspiring - just having the two patch panels hanging off the hinge has some serious deflection. In hindsight (and unique to my scenario), considering how easy it was to make my own 2U rack, I think I might give something like this a pass.

  • ETS 24-port loaded patch panel. This is one of the more cost effective units and comes with everything ready to go out of the box

  • Rackstud Series II Duo. An alternative to cage nuts, this is totally worth every penny. The ‘Duo’ only works for 1U equipment (which by coincidence, all of my gear is) but it’s a massive time and frustration saver. Having studs (instead of nuts) allows your to temporarily hang your equipment off the stud while you reach for nuts, instead of playing the balancing act of trying to support equipment whilst lining up holes.

  • Monoprice SlimRun CAT6a cables. Available in a variety of colors and lengths, I went with these to try and keep the cable mess under control. Note that for POE devices, anything ‘downstream’ of the power-injection source (whether it be an injector or a POE switch), you should try and use non-slim cables — for better heat management.

• Organizing

  • CordID tags. I use these extensively to label every single cable, knowing exactly where to plug (both!) ends of a random cable is pretty awesome

  • 1U network cable organizer. As I mentioned above, I’m not entirely sold on this for my specific use case (where I want to see the CordIDs) but for most people, I imagine this will be great. I might suggest a 2U organizer as it allows for a bit more flexibility (and maybe even the possibility of using CordIDs) — but for my specific application, I only had 1U available. The front plate can swivel from the top or the bottom and can be removed so it’s actually quite a pleasure to work with.

  • Bulk Velcro. I bought a huge roll of bulk Velcro to keep things my cables all grouped together and organized where possible

• Magnets

  • 130lb rare earth magnets. I use these magnets to make magnetic mounting plates - the solid steel surface of the wall rack gives me plenty of area to mount everything

  • 25lb magnet hooks, I use these to gently route cables around on the rack and take up slack. These generally don’t need to have a ton of holding power as you’re just gently routing things, so finding magnet hooks at your local surplus store will be fine too

• Power

  • ADJ Pow-R Rack. This rack has it all — switched power outlets, two always-on power outlets (front + back) and front-USB

  • Pyle Pro PDBC70. This is a more basic PDU that just works and provides nine switched rear-outlets

  • Splitter extension cord. You can get them in 2-split and 3-split and these are great for when you have just a few items you want to power on/off together. In my case, my main modem, fiber-switch (allowing me to toggle between main-router and low-power router) will always be powered on together, so I can use this splitter to link them to the same port on the PDU

  • Short power cords: 1ft ‘outlet saver’ extension cords, C13 ‘PC power’ cords and C7 ‘figure-8’ unpolarized power cords. One way to reduce the amount of cable management and wrapping needed is to literally reduce the physical amount of cable required to hook things up. Shorter cables allow you to get near exact-length power runs.

• Mastercraft Universal Tool base. I happened to use the Mastercraft model as I had bought one years ago on super sale but there are any number of adjustable mobile tool bases that might be suitable alternatives. I did see that the Portamate PM-3550 comes with all-swivel wheels but that price tag is something else — you could totally buy two bases and cannibalize parts for less. You can also easily put one together using just four swivel locking castors as well.

• A small 4V screwdriver: I have an older version of this 4V Black & Decker — makes working with rack nuts/screws so much less annoying as otherwise, I’m just sitting there spinning screws forever. Even since switching to rackstuds, the low power of this screwdriver means I can just cinch the nuts down on top of the RackStuds without concern that anything is going to break or strip.

[Back to Top]

---

 

 

Product links may be affiliate links: MinMaxGeek may earn a commission on any purchases made via said links without any additional cost to you.

]]>Build Log: Mini Network Rack Mk IIIAVTechDPMon, 06 Mar 2023 12:00:00 +0000https://www.minmaxgeek.com/blog/skype-station-update-2023-16020ac0b21a9b86f0577e29e:60535320d9a09304764a163f:64050bee33a42373b63389a6I have some plans for upgrading Skype Station this year — just waiting on some parts and time to work on it. I plan to split the current setup into two, much more specialized variants. For one of those specialized variants, I wanted to build it around a very intriguing camera setup; luckily I was able to test drive this camera and see where it falls in the line-up of gear I have on hand.

---

Quick Links

• Specialized variants

• The contenders

• Results

  • Sony A7 III

  • Sony ZV-E10

  • Sony RX10

  • OnePlus 9 Pro

  • Logitech C920

  • Insta360 Link

• Conclusions?

---

Specialized variants

Objectively speaking, Skype Station is pretty slick: it’s easy to use, fairly reliable, performant and delivers a great experience all around. Over time, I want to improve on what Skype Station can do — better image quality, easier to use etc. One thing Skype Station isn’t good at, is being portable.

We have a fairly nice backyard and I think my wife would like to work outside, amidst all the gross fresh air and nature. I prototyped some ideas for getting decent audio in the backyard — but I’m waiting on the cost of Raspberry Pi’s to come down in order to work on making it a bit more robust.

But what about video? I had previously performed some ad-hoc tests using a bridge camera (funny enough, it was the camera I originally built Skype Station around) but the ease of setup and use was a weak point. Using a mobile phone or a webcam addresses the ease of setup but now I have to deal with battery life and abysmal image quality.

Last year, Insta360 announced a very exciting webcam that threw a ton of buzzwords at us: 4K, gimbal, AI, etc. I was definitely interested — but I’ve also been ‘burned’ by Insta360 before…

Memory lane with Insta360

I own the Insta360 ONE camera which is neat. A single device that lets you take a spherical photo? That’s awesome (and totally handy for doing house-walkthroughs). My problem was with the software - or at least the Android experience. You couldn’t really do anything meaningful with the camera without physically tethering it it. Essentially, it turns your phone into a glorified Bluetooth shutter button for the otherwise fairly awesome gear.

Wirelessly tethering to the camera was a mess too - think olden days of wondering if the damn thing will pair or not.

So I wasn’t keen on being a guinea-pig for this camera. By sheer luck, a colleague of mine was excited enough to go for it and I let me try it out.

The contenders

I have a fairly decent camera line-up to test ranging from full-frame all the way down to the ubiquitous webcam of the 2000s.

 

 

For the tests to be meaningful and allow each camera to perform at its best:

• Each camera is tested with the best available video settings including hand-tuned color-grading for each camera

• The camera is mounted into the same teleprompter setup with minor positioning adjustments as needed to keep the framing as consistent as possible (to account for different focal lengths)

• A mannequin head mounted on a tripod was used as a subject to ensure minimize/eliminate any influence subject-positioning might have

• Where possible/feasible, I locked-off as many settings as possible (i.e., white-balance)

• Lighting was controlled for the environment and consistent from camera to camera. For some of the cameras, I needed to move the lighting ever so slightly further away in order to keep the lights out of frame

Note

One thing to note though: although I did tune each camera to try and give the best results, the ZV-E10 has a not-insignificant advantage since it’s dedicated to Skype Station and has had the most time spend fine-tuning.

[Back to Top]

Results

Sony A7 III

For full-frame cameras, I have both the A7iii and the A7R iv available; they are roughly equivalent in the headline feature: both will provide 4K with 8-bit color. I opted to actually test the A7 iii here because [a] the price point is lower, making it a more useful data point and [b] the A7 iii has an AA filter, which will give it a slight advantage on video. Will it matter? Likely not for scenarios like this.

Shot specifics:

• The video stream is 4K

• I used the Tamron 28-75 f/2.8 at 22mm

• Aperture was locked off at f/2.8

• Shutter speed was locked off at 1/50s

• White balance was locked off at 5500K

In practice, the A7Riv may have a usability advantage as it has realtime eye-autofocus in video but for the purposes of this test, the camera was pre-focused as best as reasonably available.

[Back to Top]

Sony ZV-E10

This camera is an APS-C, vlogger-centric camera that I bought as an upgrade camera to the original Skype Station; as much as I would have wanted 10-bit, the cheapest Sony option at the time would have been the A7S iii which was prohibitively expensive to use as a webcam lol. Even now, the cheapest option would be the A7 iv which is still pushing the limits on what I would consider reasonable as a webcam.

I have this camera and lens permanently allocated for Skype Station - having a prime lens on here allows me to have an even lower aperture than achievable with my full-frame camera (with the lenses available to me at this time).

Shot specifics:

• The video stream is 4K

• I used the Sigma 30mm f/1.8. Other than having a faster aperture, one nice advantage of prime-lenses is reliability: I never have to worry about whether the lens has shifted focal length ever

• Aperture was locked off at f/1.8

• Shutter speed was locked off at 1/50s

• White balance was locked off at 5500K

Even though this setup has a smaller sensor than the A7iii, I’m able to leverage the amazing Sigma prime to really drive the shot here. Since my full-frame glass is ‘only’ f2/.8, technically speaking, this setup has a lower aperture (f/2.1).

[Back to Top]

Sony RX10

This was the camera that started it all — the original camera I used for Skype Station. This is an old camera (2013!) but covers all the basics: constant f/2.8 aperture, a relatively large 1.0” sensor, eye-AF and even supports modern Sony hot-shoe accessories. After seeing the modern cameras, you can totally tell this is a substantial downgrade but if you take into account it’s age, this is stellar. This camera has a maximum video resolution of 1080, so keep that in mind.

Shot specifics:

• The video stream is 1080

• This camera has a built in lens with a constant aperture of f/2.8 I shot this at 18mm (about 48mm equivalent)

• Aperture was locked off at f/2.8

• Shutter speed was locked off at 1/50s

• White balance was locked off at 5500K

Initially this image looks noisier than the previous but the specific choice of using ‘bricks’ as a backdrop helps a lot and lets some of the noise be ‘explained away’. If you remember that this camera is limited to 1080, this is actually very workable quality.

[Back to Top]

OnePlus 9 Pro

I got new phones a little while back so this was a good opportunity to see just how far smartphones have come. Using a plugin called DroidCam, we can feed a camera stream via the network into OBS. This is actually really cool because this is a super easy way to get into multi-cam streaming/recording. From what I can tell, the stream is limited to 1080 (coming into OBS) but the camera on the phone is capable of 4K so I suspect that through the process of super-sampling, noise is reduced substantially.

Shot specifics:

• The video stream is super-sampled 1080

• The phone has multiple lenses, from what I can tell, this is shot using the primary lens which is 1/1.43”, 23mm (equivalent)

• Aperture is f/1.8 (but with crop factor, this comes out to f/16.7)

• There’s no option to lock off shutter speed

• There is an option to lock off white-balance but there’s no numbers — just a scale that you drag, so I opted to leave this on ‘auto’

So right off the bat, one huge takeaway here is that smartphones are awesome in terms the image quality they can put out - everything is crisp and sharp and there’s virtually no noise. This comes at the cost of

1. Limited adjustability. Maybe this is a OnePlus limitation, maybe it’s an app limitation etc.

2. Smartphone video tends to trend towards overdoing the sharpness filters though; in this specific context, I think it’s fine however, smartphones also overdo the saturation levels: this is the most saturated result by a huge margin

3. Battery life, heat etc.,

For short bursts, phones make an excellent supplemental camera if you can deal with the extra saturation. Specific to Skype Station, one thing to note is that a lot of phones have their cameras in the corners of the phone — which makes it tricky to physically place the phone behind a small teleprompter as it’s really easy to hit the sides of the teleprompter.

[Back to Top]

Logitech C920

What might be the original premium webcam, this was the webcam to get in the early 2010s and sadly, that’s about when Logitech’s drive to innovate and push out better products went out to lunch. They launched the Brio 4K a little while back but that feels like a case of too little, too late. Now that we’ve had a taste for what is possible, the C920 is categorically bad.

Shot specifics:

• The video stream is 1080

• Information on specifics of this webcam are exceedingly difficult to find

• From what I can guess, this has a 1/3” sensor and uses a 28mm equivalent focal length

• There’s almost nothing we can set, everything is ‘auto’

A decade ago, this webcam was hot stuff but I don’t think it has aged well. I think Logitech got comfortable with ‘well, it’s better than integrated webcam, what else are you going to do’ and when COVID and Elgato’s Camlink carved up a whole new market segment.

Logitech has released a couple minor follow-up webcams recently, but I can’t imagine they’ve substantially improved anything - certainly not for the ~$100 they are asking.

• C920

  • C920S: (lol) Adds a privacy cover (something you could just add anyways)

  • C920E: (lol) Microphone is disabled by default, also adds a privacy cover

  • C920X: (lol) Adds a 3-month license for Xplit

• C922: Adds a nearly pointless 60fps recording — only at 720P

  • C922X: (lol) Adds a 3-month license for Xplit

  • C922E: (lol) Microphone is disabled by default, also adds a privacy cover

• C930: widens the field of view to 90° (up from 78°); whether this is useful is dependent on your specific setup

[Back to Top]

Insta360 Link 4K

And finally we get to the camera I set out to test, the Insta360 Link. While it’s not truly unique — there is a similar, slightly cheaper OBSBOT Tiny PTZ 4K but I would argue the Link is more compelling because of the [ever so slightly] bigger sensor and smaller footprint - the Insta360 Link is positively tiny, even when compared to the OBSBOT which is already quite small. Of course, what I’m here to investigate is the image quality.

Shot specifics:

• With HDR, the stream is limited to 1080. Without HDR, the stream is 4K

• This camera has a built in lens of 26mm (equivalent)

• Aperture was locked off at f/1.8 (f/9.7 equivalent)

• There’s no control for the shutter speed

• White balance was locked off at 5500K

View fullsize
With HDR (1080) View fullsize
Without HDR (4K)

It’s a bit unfortunate that HDR comes at the cost of a resolution drop (you also lose high-framerate recording but that’s less of an issue in this application). If all I knew was the 1080-footage-with-HDR, I think I would be satisfied but now knowing how sharp it can be without HDR? In my opinion, you can claw back a bit of the colors with a bit of grading and looking at them back to back, the HDR footage starts to look bad.

 

---

 

One small gotcha…

When run through OBS, the video feed introduces a noticeable amount of lag. Oddly enough, I noticed that there was substantially more lag if the video feed was configured as H264 rather than MJPEG — and here I thought encoding tech had gone past the days when this was even something to consider. Pardon the following grainy captures — there’s only so much I can do with animated gifs ;)

The process of doing the stream capture and then converting to an animated gif does tend to exaggerate the motion a bit (which is helpful in this case when I’m trying to identify lag). In the real world, (where frames are not being decimated to create animations), there is virtually no lag when configured as MJPEG. A bit disappointing that this limitation cropped up given that I thought almost anything with more computational horsepower than a glass of water could handle H264 now. So for best results with the Insta360 Link, it appears we should:

• Set HDR off

• Use 4K resolution

• Bring colors back with a quick grade

• Configure the feed to use MJPEG

[Back to Top]

Conclusions?

The image quality from the Insta360 Link (non-HDR) looks great. Compared back to back with the ZV-E10, for sharpness, it definitely punches well above its weight. It’s only when you start to look for differences:

• The lamp in the background is sharper (because the ZV-E10 has a much shallower depth of field)

• Looking back and forth between the photos, you might notice bit of the mannequins face might have some noise (the sensor on the ZV-E10 is much bigger so there will be less noise)

Categorically, the ZV-E10 is better — but the fact that the Link passes the quick-glance test is incredible. The Link isn’t exactly cheap ($400) but it’s in a great position where anything cheaper is substantially worse and anything meaningfully better will cost a lot more. Worth noting is the possibility of using a smartphone as a portable-camera although this depends on your specific phone (as you start looking at older and older phones, the capabilities may fall off dramatically) and you’ll need to figure a way to handle the heat and drain on the battery.

I initially set out to see about finding a camera option for a more portable Skype station — and the Insta360 looks like the camera that fulfills that niche.

[Back to Top]

---

Product links may be affiliate links: MinMaxGeek may earn a commission on any purchases made via said links without any additional cost to you.

]]>Interim Skype Station UpdateTechDPFri, 03 Mar 2023 13:00:00 +0000https://www.minmaxgeek.com/blog/new-work-laptop-20236020ac0b21a9b86f0577e29e:60535320d9a09304764a163f:63ff5caa9f640902c843ba1fOur work laptops were up for a refresh - and not a moment too soon: my previous laptop has some serious issues with the communication with the battery (a, uh, somewhat important component of a mobile-focused device). This laptop also is my first actual interaction with Windows 11, so that’s gonna be fun given the bad press I’ve heard about it.

This post is after about a week of transitioning to the new laptop and I wanted to capture the initial thoughts and document some of the fixes I had to apply to make it less stupid.

---

Quick Links

• New machine

• Actual performance observations

  • SSD Performance

  • High resolution performance

  • Network performance

• Other fixes

  • Additional drives to organize your files

  • Sticky mouse edges in Windows 11

  • Changing wallpaper/lockscreen

  • ‘Improving’ the resolution

• Early thoughts on this laptop

---

New machine

So I got a new work laptop recently and sadly, it’s mostly sad news: there are some neat things about it but on the whole, it’s almost certainly downgrade for two user groups:

• Those looking for high-performance (even high-performance by laptop standards) and

• Anyone who uses this laptop-as-a-desktop. You know: almost perma-docked, hooked up to monitors, external keyboard, mouse etc. — so ‘most people’

 

---

 

The Good

Since this is mostly bad news, let’s get the positives out of the way:

• The new laptop is thinner, lighter, sleeker, feels nicer to the touch

• They added another USB-C port!

• Reduced the [required] charging brick down to 65W. This is great because it has a trickle-down benefit of allowing a much wider variety of chargers to power the laptop

• The laptop keyboard didn’t get any worse (not a very high bar I know, but I’ll take it)

• Out of the box it supported my weird work monitor setup that was based around adding two eVGA UVPlus+ adapters

• Some meaningful performance gains!

• Unique to me: the old laptop battery cells would constantly fail. When the battery failed, you would know really fast because, not only would it fail to charge, the processor would be locked at 800MHz until the battery was removed/replaced.

The Bad

• Windows 11. Full stop. More on this in a bit.

• The force-power-down time has been increased from 12-seconds to 14-seconds (!!). I can’t fathom why they would need to do this, perhaps, their customer base was accidentally holding the power button down for 12-seconds at a time and unintentionally powering down their laptops?

• They removed the ubiquitous Dell DC barrel port (likely in an effort to keep the laptop thin). This is super unfortunate because those power bricks were much more reliable and (perhaps unique to me), I have a bunch of those handy. There is an official adapter that I might pick up to continue using my tried and true power bricks

• Went from a quad-core processor to a ‘dual core’ processor and all the mess that comes with that

Seriously? Dual-core?

Thankfully not, I’m just being playful about it and poking a bit of fun at how Intel uses P-cores (performance) and E-cores (efficiency) in their processor architecture. The high-performance purist in me only really counts P-cores — especially when I’m using the laptop primarily as a faux-desktop configuration anyways.

For reference, the old laptop had an i7-8665U and the replacement now has an i7-1265U. High performance jokes aside, the new processor is faster but my problem with both the old and new processors is the abysmal 15W TDP rating.

In other words: it’s a laptop processor, so it’s always going to be terrible. Since I use it like a desktop, I’m going to compare to to a desktop.

• Hopefully unique to me: in an ironic twist, within a week, the primary Thunderbolt/USB-C port partially failed. It still accepts power, but the data connectivity no longer works (or at least, works intermittently). Thankfully, this laptop has two such ports. I don’t blame Dell for this per-se, this is just Thunderbolt being dumb.

[Back to Top]

Actual performance observations

SSD performance

The old laptop shipped with a 512GB Samsung PM991 NVME drive. On paper, this is fairly performant — it should be about on par with the more recognizable Samsung 960 EVO. The real world performance is eye opening. It’s not perfect, but just as an approximation, we can look at CrystalDiskMark to benchmark the drives.

 

 

I don’t think it’s fair to think the drives will benchmark at their theoretical numbers but the relative performance here is atrocious. I don’t know how much blame to associate with a bottlenecked processor versus ‘enterprise bloatware’.

But there are some performance improvements! The new laptop comes with a 1TB SK Hynix PC801 NVME drive and on paper, this is a total monster of a drive.

 

 

Compared to the theoretical ceiling, the improvement in write-throughput here is monumental, but IOPs still seem to be held back. If we compare the performance from old-laptop to new-laptop, the difference is eye opening:

 

 

To shed some light, let’s compare against an actual desktop PC (happens to be running WD SN850)

 

 

I did run the numbers multiple times on all the machines to verify that there is at least some sense of consistency. Knowing that both laptops have a similar amount of enterprise bloat, the improvement from old- to new-laptop comes from a combination of the new drive being inherently faster but even more so, what must be reduced bottle-necking of the PCIE controller (in the processor).

That being said, even a newer processor architecture leaves a ton of IOPs performance on the table. When we compare to a full-fat desktop setup — which doesn’t have nearly as fast of an NVME drive, we can see that the full desktop environment has enough horsepower to fill out the numbers!

Take away re: slow drive performance:

Your solid state drive is slow because your processor is weak.

[Back to Top]

High-resolution performance

I imagine this won’t apply to most people, but there are definitely some scenarios where connecting a virtual monitor can introduce quite a bit of lag. Definitely made me step back and try to think of the last time I ever saw this happen — other than on the previous 15W laptop.

 

 

[Back to Top]

Network performance

Moving a bunch of data across a network takes computational horsepower (Captain Obvious strikes again) but just how much is needed isn’t that obvious. On my network, I can test somewhere in the 7500/6000 Mbit/s with the hardware I have on hand. Using OpenSpeedTest isn’t perfect, but certainly gives me some general performance pictures.

One thing I wanted to check was if the new laptop would choke out with network traffic. Using a basic USB 2.5GbE dongle, I did some tests on a freshly rebooted machine at the desktop.

 

 

The synergy between the network performance and processor load is definitely interesting: remember this test is done on a fresh reboot, with nothing running, at the desktop. And with those ideal conditions, we are still seeing CPU saturation -- imagine actually using the computer!

As we scale to newer and/or more powerful machines, we can see the required CPU load go way down -- leaving room for you to actually use the computer while downloads are happening

Take away re: slow network performance

Your network is slow because the processor is weak.

[Back to Top]

Windows 11

For the first five minutes of using the laptop, I thought Windows 11 was fine — great even. You get a refreshed aesthetic and some nicer touches on things etc. and my overall opinion of it remains about the same: it’s great if you’re only ‘stopping by to use it quickly’. As soon as you sit down to actually do anything? It’s a trainwreck. For me, almost all of my issues revolve around changes to Explorer and the Start Menu.

• No Drag and Drop

  This one is a doozy. As of 21H2, drag and drop functionality just does not work the way you would think it should. You can’t drag a file into an open Notepad instance, you can’t drag a file onto a Notepad icon, you can’t drag a file into Notepad on the start bar.

  You can however, do a bunch of bs to indirectly get drag and drop to work (like drag ⟶ alt+tab,⟶ drop) which I’m never ever going to do.

  To directly and specifically fix this issue, navigate to HKLM\SOFTWARE\Microsoft\Windows\CurrentVersion\Shell\Update\Packages and add a new DWORD32 UndockingDisabled and set it to 1.

  Restart the explorer process and you can drag and drop again! Except there’s a catch…

• Start Button doesn't work

  Once you’ve fixed drag and drop you’ll notice the start button no longer functions (clicking on it, ctrl+esc, windows key - nothin).

  It took me awhile to notice this because I primarily use Win+R to interact with things and shutdown -f -r -t 0 is less frustrating than clicking restart, walking away and coming back to a dialog saying “are you sure you want to restart, but these apps are still open yo”

  Funny enough, if you undo the drag & drop registry setting hack, it will restore the start menu functionality until the next reboot. So: apply the registry fix, restart explorer, undo the registry fix.

  If you’re lazy you can probably automated this into a startup batch file. I just kept a pair of exported registry keys on my desktop.

• Taskbar is just plain dumb

  This is really an extension of the above two related issues. In its ‘broken state’ you can’t do anything useful with the taskbar — you’re stuck with “combine always”, never having labels, not being able to resize it or move it. You can, however, change the size of the icons in the taskbar (whoopie!). There’s got to be a better way to fix this.

• Show all notification icons

  I’ve always wanted to see all the notification icons — without having to click on a stupid thing to expand it. In Windows 10 this was readily accessible via a ‘always show all icons and notifications in the taskbar’ setting. Of course this option is now gone.

  If you run the following shortcut, it restores the Windows 10 option dialog allowing you to set it.

  C:\Windows\explorer.exe shell:::{05d7b0f4-2121-4eff-bf6b-ed3f69b894d9}

 

---

 

Please, fix it all

There are more annoyances (like the context menu or the Explorer ribbon menu) but at least for me, those pale in comparison. Thankfully, there is an easy fix that knocks out most of the issues

1. Grab Explorer Patcher which is also free. Run this, it’ll blank our your desktop for a moment and then you’re done, now your start bar behaves ‘correctly’

2. Grab WinAero Tweaker, which thankfully, is freeware

Within WinAero Tweaker, have a look through the settings and adjust your system to be less dumb

[Back to Top]

Other fixes

Additional drives to organize your files

I like to keep my files organized into separate partitions/drives. The work laptop comes with Bitlocker enabled so there’s a bit of a dance that needs to be done:

1. Suspend Bitlocker

   • Open the BitLocker settings window: Win+R, explorer.exe shell:::{D9EF8727-CAC2-4e60-809E-86F80A666C91}

   • Locate the drive you want to suspend BitLocker for and click Suspend protection

2. Do your partition shuffling

   • Open the disk management snap-in: Win+R, diskmgmt.msc

   • To make room for new partitions, first I shrink C

   • Then create new partitions, don’t format during the partition creation wizard

   • If a BitLocker popup is shown, go ahead and encrypt the drive. If no such popup is shown, switch to the Bitlocker settings window, locate your drive and encrypt it

   • Once all the new partitions are encrypted, go ahead and resume Bitlocker protection on C. Format any outstanding partitions if needed and configure the drive letters as desired

3. Done, now you have a bit better file organization

If you don’t feel comfortable (or can’t) do the partition and Bitlocker dance, there is a fallback method to do this. To do a cheater way of doing this, navigate to your root drive and create some folders to organize you data. For example:

• C:\MyFiles

  • C:\MyFiles\Logs

  • C:\MyFiles\Archive

  • C:\MyFiles\Repository

Now from Explorer, utilize the Map a Network Drive functionality and map your folder using the pattern localhost\C$\MyFiles\Logs to your desired drive letter. For most scenarios, this will behave the way you’d expect with a negligible penalty.

[Back to Top]

Sticky mouse edges in Windows 11

This was definitely an odd one: in Windows 10 introduced the idea of "sticky edges" for multi-monitor setups. In theory, it's supposed to make dragging windows between multiple monitors a bit easier but some people (i.e., me) don't like it -- I use hotkeys to bounce applications between monitors. You'll know sticky edges is at work when your mouse cursor gets to the edge of a monitor, and if you move the mouse cursor slowly, it will be 'stuck' in that monitor. In order to get the cursor to the next monitor, you'll need to move the mouse cursor faster -- to 'jump over the edge' so to speak.

In Windows 11, settings for this have effectively disappeared (and/or the functionality officially seemed to be removed) but I was still encountering a sticky mouse situation. Comically, the root cause was actually the UX for handling multiple monitors. In Windows 10, when you rearrange multiple monitors around, they snap up next to each other. In Windows 11, this doesn't appear to be the case and you have to be extra careful to make sure the monitors are really butting up against each other. This problem seems to be magnified with monitors arranged up-down rather than left-right.

Consider the these two scenarios, specifically looking at monitors 1, 2, 6 and 5. In both scenarios below, there is no issue moving the cursor left-right between screens but definitely vertically.

View fullsize
Not working as desired View fullsize
Now, working as desired

[Back to Top]

Changing wallpaper/lockscreen

For these laptops, the wallpaper and lockscreen are intrinsically linked and will display the same picture. To change your wallpaper:

1. C:\Windows\Web\Wallpaper\Windows

2. Replace img1.jpg with whatever image you would like to use. You can either delete/overwrite the original file or rename it so you have a copy to roll back to

[Back to Top]

'Improving' the resolution

The laptop screen is 1920x1080 which we can sanity check by visiting something like screenresolutiontest. There are two ways to ‘force’ a higher resolution but they come limitations.

 

 

Method 1

• Open the laptop (so the screen activates)

• Connect a higher resolution display to the laptop

• Select the option to ‘Duplicate these displays’

Now, you’ll notice that your laptop’s built in screen is displaying the higher resolution and if you scroll down you can even verify that it isn’t ‘cheating’ by using a 25% scaling factor. Simply leave the monitor physically connected and switch the input on the monitor to something else. The downside here is that you don’t have the benefit of multiple physical displays but your laptop will behave as if it was 4K natively. Not sure if this will work on all laptops, but it certainly does on this work one.

View fullsize
Method 1

Method 2:

You’ll need to have a VNC (or maybe remote desktop) setup configured. I made a more in-depth write-up here using my previous laptop

• Buy a virtual display dongle and plug it in

• Select the option to ‘Show only on 2’

• Now you can close the laptop lid if you want

Now, when you remote into this machine, you’ll have access to the higher resolution. Again, you’ll have some limitations based on the number of dongles you can plug in etc.

View fullsize
Method 2

[Back to Top]

Early thoughts on this laptop

Out of the box, the new laptop is a mixed bag: the device itself is definitely superior (even with the ‘dual-core’, low-TDP processor) than the predecessor. The underwhelming processor hamstrings what would otherwise be a stellar NVME, but even with that handicap, it’s still very good.

Low-performance perspective

For a long time, I’ve thought that maybe I was being a bit too harsh on mobile processors for simply existing, but getting a this new machine was a great opportunity to get some cold hard numbers. Seeing the processor redline at 100% with downloads and seeing the write-performance get absolutely tanked; so much of what we do with these machines boils down to ‘internet with the network’ and ‘write some stuff on the local machine’, this provides an explanation of why the machine feels sluggish.

Hardware limitations aside though, Windows 11 as it renders the entire machine dead on arrival - or at least until the annoyances are mitigated. I try not to hate on things because they are new and different and objectively, there are a lot of neat UX improvements with Windows 11, but it just came at the expense of dumbing down the user experience when you actually sit down and try and use it. I’ve heard that a future update has fixed some of the annoyances, but that’s not currently available to us so, I’ll have to revisit that in the future.

Is the old laptop that bad?

Truthfully, I don’t think so — I think a large part of the perceived (and measurable) pain with the old laptop lies in what I’ll call ‘enterprise bloatware’ — software, settings and policies that [rightfully] should be on a work machine. If there’s an opportunity to sterilize the machine from scratch, the performance would be substantially improved (the issue of what appears to be a faulty battery connection notwithstanding).

[Back to Top]

---

Product links may be affiliate links: MinMaxGeek may earn a commission on any purchases made via said links without any additional cost to you.

]]>New work laptop…TechDPFri, 06 Jan 2023 07:05:00 +0000https://www.minmaxgeek.com/blog/security-audit6020ac0b21a9b86f0577e29e:60535320d9a09304764a163f:63b5d5dbe19d3600e6fa934bOver the year-end break, I moved my password manager away from LastPass. I didn’t move for security reasons but more so for usability issues and the security issues were purely an unfortunate coincidence. I used this transition as an opportunity to do a full audit of the credentials and I stumbled across some fun (read: annoying) discoveries.

---

Quick Links

• Why was I with LastPass originally? Why change?

• Thoughts on the credential audit

  • Stealth wiping

  • Messy wipes

  • Automatic lockdown

  • Have I been pwnd?

• Revisiting 2FA with Authy

---

Why was I with LastPass originally? Why change?

When I originally transitioned to using a password manager forever ago, I chose LastPass for the same reasons ‘everyone else’ did — it was the popular option that was well marketed. It was approachable, did well in terms of usability and value-add. Over the years, the security issues haven’t bothered me nearly as much as the usability quirks.

A few things really got to me over the years:

1. It takes four clicks to get a username/password into your clipboard. So potentially eight mouse clicks to log into something. The count is even higher on phones where you may need to switch back and forth between processes (and sometimes having to re-authenticate in between). Thankfully with phones, you can have the option of pinning copy-notifications to make it a bit less frustrating

1. By default, the extension will clear the clipboard after 60 seconds. 99% of the time this isn’t an issue (especially if you’re aware of this) but there’s a rare occasion when the “would you like to save this site” prompt doesn’t trigger (say, due to the way the site’s account-creation is set up) and if it takes >60s to get the activation email etc., you’re in for a surprise when your password is cleared.

2. Seemingly random reasons to re-authenticate. I understand the monthly re-authentication, that’s fine and I can even understand getting signed out randomly here and there, but I’ve had a few cases where I’m authenticating (with 2FA too, for extra fun) five or six times in a 30 second period

3. The extension will populate a little icon in text-boxes that allows you to invoke LastPass (normally, to trigger a “search for logins for this site” functionality). To things bother me about this feature: [1] many times it interferes with the ‘eye’ (show me what I typed) and [2] it’s rare, but sometimes the button just doesn’t work.

4. There wasn’t the ability to (easily and neatly) add ‘sub-entries’. For example: logging into your bank will include a login and password but what if there are challenge questions? PIN numbers? PIN number specific for calling into phone-banking vs support? There’s no way to consolidate this into a single object if you wanted to.

On the big scale, these are relatively small issues by themselves, but LastPass presents itself as a premium player in this space (and they certainly charge a premium price tag). At this level (and price!) I’m much less tolerant of these issues.

[Back to Top]

Thoughts on the credential audit

When I started, I had about 500 credentials. After everything was migrated, consolidated and cleaned-up, I now have just over 265. That’s a huge drop looking at consolidating multiple related credentials (for example., combining PIN and challenge questions into the ‘main’ entry) only accounted for about 60 or so entries. Maybe 30 accounts were for websites that were no longer kicking around.

So what about the almost 150 remaining entries? I’ve got some stories.

 

---

 

Stealth wiping

Let me tell you a story about product registration.

Oftentimes when you buy stuff, the product manufacturer wants you to register it (ostensibly to get your contact information so they can sell you more stuff but that’s beside the point). So let’s say you create an account and register your product. Some time passes and a few months/years later, you buy another product and go to dutifully register it. Depending on how long it’s been, there is a more than good chance your account doesn’t exist anymore and all the things you registered? Probably gone.

I can understand how some services might want to occasionally purge inactive accounts for product registration, that’s a bit silly. Oddly, there are some notable web-services that purge accounts (I should note, there’s no notification of the purge — if there was an email sent saying “your inactive account is going to get purged” that’s totally fine)

• Web services

  • BlazeMeter

  • Fongo

  • MyPrepaidCenter

  • OpenDNS

  • Webex

• E-Retailer

  • CDW

  • DataPro

• Product registration

  • Aukey

  • Anker

  • Black & Decker

  • Bosch (power tools)*

  • Flir

  • LG (appliances)

  • NEEWER

  • Porter-Cable

  • RAVPower

  • Worx

  • Zotac

Bosch is a bit of an interesting one. My account (with product registration) got deleted, so I created a new account (using the same email). When the ‘new’ account was created, my previously registered tools showed up. While this was convenient for me, I’m not sure if I like the idea that a ‘deleted’ account (the notification said the account was deleted) still has associated information floating around.

Looking just at product registrations, this now has me looking at purchases in a different light. Specifically looking at power tools for a second:

• Accounts still in working order despite not logging in for years:

  • Milwaukee

  • Ridgid power tools (orange) and Vacs (red)

  • Ryobi

  • Kreg Tools

  • Bosch (my account got wiped but was able to be restored, so I’ll log it here for now)

  • EGO (but my account is still quite new so I wouldn’t expect it to be wiped)

• Accounts that got wiped without notice

  • Black & Decker

  • Porter-Cable

  • Worx

Having my accounts (and associated registered items) quietly wiped leaves me with an odd feeling about the vendors. In fairness, while I’ve not had any issue with the tools … what if I did? It costs next to nothing to store the minuscule amount of data associated with my account & registered products, so what gives?

All things being equal (or perhaps, close), I’d certainly shy away from brands where my data gets stealth wiped.

[Back to Top]

Unclean wipes

Accounts getting stealth-wiped seems to be the signature move for power-tool companies; for small electronics companies they took it another step: not only is your account is stealth wiped but they got rid of the entire concept of ‘creating an account’ and/or ‘registering your product’ — so forget even making a new account to register a recent purchase. Fun times.

Sometimes, you get some companies that just drop the ball even when it comes to wiping, like UPS. My account was stealth wiped; I know this because the credentials don’t work.

• When I trigger “I forgot my username”, I get an email from them telling me “this is the username and email associated with the login”

• If I try the “Please reset my password”, that process fails

• If I try and create a new account, I can’t because the account is still in use

As someone who understands a bit of how the software works, all i can do is shake my head. Thankfully, I don’t actually really need this account for anything.

[Back to Top]

Automatic lockdown

I had a fun time with Marriott — if you change your email and phone number, you can’t log in for at least a day. Something about “you recently changed your contact info, you can’t log in at this time”.

I can understand why they might do this — if I was going to steal an account, the first thing I would do would be to change the contact information of the account! But if they are going to do this lock-down, maybe put a warning that “hey we noticed you just changed your email, if you change your phone, you won’t be able to log in”. I don’t travel enough (nor care for Marriott) so this was just an amusing nuisance of having to make a note to come back in a day or two to verify the credential migration was successful.

[Back to Top]

Have I been pwnd?

I’m a huge fan of the site haveibeenpwnd where you can punch in your email (or phone number) to see if it’s shown up on a leak somewhere in the past. This can be a bit of an warning beacon to go and do a comprehensive password update. There’s a FAQ to cover the obvious questions that arise from a service like this.

[Back to Top]

Revisiting 2FA with Authy

After I got my accounts migrated over, this was a pretty good time to revisit 2FA as well. I regularly check a directory of sites that support 2FA to see if I can add 2FA to any of my existing accounts (and in some cases, move away from SMS to time-generated codes).

Until now, my code-generator app of choice has been Google Authenticator and it while it works great, I did start to have problems with it a few updates back (it’s been bugging me for about a year now). Making an educated guess, when I launch Google Authenticator, it goes through all of the entries and starts generating the one-time keys and then it finishes loading the app. You’ll notice that when the app launches, all of the entries already have keys being generated.

At time of writing, I have about 70 accounts protected by 2FA and it regularly takes 3+ seconds to launch the app. Add my gripe of it being a general pain in the butt to manage the entries (relabelling and sorting), I wanted to see if there was a better option.

I settled on trying Authy and it’s been great. I don’t care for the backup & multi-device sync features but the app launches fast. The one-time keys are only generated when I click on the entry I want which I think is why the application launches so faster and is generally more responsive. I can very easily relabel entries and sort everything more easily. It’s also a nice touch that I can add icons to the entries and that I can also perform search.

What about the other [dis]advantages between Google Authenticator vs Authy?

Most of the time, people focus on the ability to sync across multiple devices and/or do backups. For me, this is a non-issue since I run rooted, so snapshotting the entire application’s data provides me with the ability to clone my authenticator to another device if needed.

There are still a ton of sites that use SMS as means of 2FA and one thing that has helped me keep my sanity is to create a single contact (i.e., ‘2FA’) that has all of the phone numbers (and/or email addresses) so that when the 2FA codes come in, they come in to a singular entry in your messages-view rather than spread across a billion different entries. This is a long-game optimization hack as you’ll need to add the SMS contacts as they come in, over time. But it does get better — and you can migrate it with you to new phones.

[Back to Top]

---

Product links may be affiliate links: MinMaxGeek may earn a commission on any purchases made via said links without any additional cost to you.

]]>Security Audit (2023.1)TechDPTue, 03 Jan 2023 21:18:08 +0000https://www.minmaxgeek.com/blog/fix-synology-ssl6020ac0b21a9b86f0577e29e:60535320d9a09304764a163f:63b49b908222a052ac40b96cWith HTTPS is enabled, whenever I access the management portal for my NAS, I get a very well-intentioned warning about how the certificate isn’t trusted and how all the baddies could be trying to bust into my NAS. I have dutifully clicked the Advanced ➞ Proceed anyways option but this has been bugging me for a while now and I finally got around to fixing this nuisance.

---

Quick Links

• Starting point

• The easy options

• Step 1 - Prerequisites

• Step 2 - Become a CA

  • [A] Generate a private key

  • [B] Create our root certificate

  • [C] Install the Root Certificate

  • Recap

• Step 3 - Create the signed certificate

  • [A] Get the CSR from DSM

  • [B] Create our own CSR

  • [C] Special sauce

  • [D] Sign the NAS certificate

  • Recap

• Step 4 - Add signed certificate to DSM

• TLDR Steps

---

Starting Point

Every time I navigate to my NAS’s management portal, Chrome makes sure to tell me how my connection isn’t private and that attackers are lurking around the corner and while it’s not a huge deal to just carry on, it’s a nuisance.

My NAS isn’t open to the internet, I don’t use the QuickConnect functionality and have no desire to go down that road - if that’s something you want, Synology has a KB article on it. For users that don’t want this, it’s initially an exercise in frustration because Google search results will yield two types of results:

1. How to setup LetsEncrypt to handle SSL certificates for your NAS or

2. How to setup a self-signed certificate for your NAS

The latter option sounds great until you realize that this functionality was only available in DSM6. Surely, we can’t let this stop us!

[Back to Top]

The easy options

There are a few easy options

1. It’s not a big deal, just do the Advanced ➞ Proceed dance every time

2. If you’re only ever accessing the management portal from your local network, maybe turn off HTTPS? Is it really necessary after all?

3. Utilize site-settings to allow insecure content

Since I use HTTPS Everywhere, turning off HTTPS doesn’t really work for me so realistically, I’m left with the last option. This workaround doesn’t make the problem go away but can reduce the nuisance-factor. By leveraging Site Settings, we can override the behavior for the management portal site specifically. Just scroll down and set Insecure content to Allow and that’s it!

After setting this to Allow, I noticed that I was prompted once about the super insecure site and then I wasn’t bothered again for a long time. You still get the ‘Not Secure’ flag in the address bar so this still bugs the hell out of me.

 

---

 

To fix this, we’ll need to “make the certificate valid”. At a high level, this has three stages:

1. Become a Certificate Authority (CA)

2. Create a certificate for the NAS

3. Install the certificate for the NAS

Seems easy enough.

Of course, to make things more challenging, I’m opting to do this purely in Windows since I just don’t want to spin up and maintain a virtual Linux environment just for this. If you’re more versed with Linux, this may flow a bit more naturally for you.

[Back to Top]

Step 1 - Pre-requisites

Working within Windows (and not counting WSL or anything), before we can do anything, we have to get the openssl tools. Pick whichever pre-compiled binary you feel comfortable with (specifically, I went with the ‘Light installer’ option).

During installation, I opted to copy the binaries to the \bin folder, rather than the \Windows directory.

We will be working entirely out of the \bin folder, so go ahead and open a Command Prompt window and navigate to our \bin folder

[Back to Top]

Step 2 - Become a CA

At a high level, we want to create a private key (protected by a password) and then, using that private key, go ahead and create a [public] certificate.

[A] Generate a private key

We don’t need to become a full-fledged CA, we just need the bits and pieces absolutely necessary just so we can create some certs. In our working folder, go ahead and create a file ‘pass.txt’. This file should contain a strong password as this is what secures our CA. By using a separate file to store the password, we can just refer to this file instead of having to type the password out again and again,

OpenSSL has two commands that will create a private key for us: genpkey and the older (but still valid) genrsa. In the following sample, I’ve opted for 4096-bit keys (because, heck, why not).

openssl genpkey -algorithm RSA -pass file:pass.txt -out ca-key.pem -pkeyopt rsa_keygen_bits:4096 -aes256

openssl genrsa -aes256 -passout file:pass.txt -out ca-key.pem 4096

Running this command will create a ca-key.pem file in our working folder. We can verify that the private key is good by using the following command:

openssl rsa -in ca-key.pem -check -noout -passin file:pass.txt

If all goes well, the output should be: RSA key ok

[Back to Top]

[B] Create our root certificate

We can use the ca-key.pem private key we just created, to create our root certificate. Here, I opted for 10-year expiry on the root certificate.

openssl req -new -x509 -sha256 -days 3650 -passin file:pass.txt -key ca-key.pem -out ca.pem

When we run this command, we will be prompted to provide some information. I opted to specify the country and the common-name, leaving everything else as blank

[Back to Top]

[C] Install the Root Certificate

By installing the root certificate on our machine(s), any certificates signed by this root-certificate are automatically trusted. On Windows it’s probably easiest to first convert the certificate:

openssl x509 -outform der -in ca.pem -out ca.der

This creates a more familiar certificate file in our working folder - install this into the Trusted Root Certificates store. You can do this per-user or per-machine (I opted to do per-machine).

[Back to Top]

Recap

At this point we have a few files in our working folder:

• pass.txt - This file contains the password that we secure our private key with.
• ca-key.pem - This is the private key for the CA
• ca.pem - This is the public facing certificate
• ca.der - This is the public facing certificate, in a different format. For Windows users, this will be the file we install on each machine

With these files, we are now a certificate authority and we can sign downstream certificates! Make sure we put these files in a safe spot somewhere.

[Back to Top]

Step 3 - Create the signed certificate

[A] Get the CSR from DSM

So the first thing to do is get the Certificate Signing Request (CSR) from our NAS Management portal. Instructions may vary for different versions of DSM but for 7.1x, the steps are:

1. Log into the NAS

2. Open Control Panel

3. Open Security

4. Select Certificate

5. Click Settings

6. Click Advanced

7. Click Create certificate signing request (CSR)

View fullsize
View fullsize

Creating the CSR requires you to fill out some information — I opted to set the private key length to be the same as everything else so far (4096), and for the common name, I set it as one of the hostnames for my NAS.

DSM will prompt you to download a zip with two files in it:

• server.key — this is the key that we need to sign using our CA

• server.csr — this contains the settings we just filled out when creating the CSR

The server.csr file, as exported, is intended for us to send to a third party (paid) certificate authority to then sign (and they can’t help us with local IPs anyways) so we won’t actually need that file, we’ll make our own csr in a moment.

[Back to Top]

[B] Create our own CSR

From the archive.zip we just downloaded, extract the server.key file to our working directory. Before we totally forget what this file is, I would strongly suggest renaming it to something you might have a better chance of remembering: echo_server.key. We ca use this file to create our own CSR.

openssl req -new -key echo_server.key -out echo_server.csr

[Back to Top]

[C] Special sauce

Create a file in our working directory, echo_config.txt and add the following text. Adjust the DNS/IP entries to match the values you have set for your NAS.

• DNS entries: add as many as you want here — whatever friendly label you want to associate with your NAS.

  • You can use just a simple hostname (i.e., echo) or you can use a ‘more proper’ FQDN (i.e., echo.lan)

  • Note that you can have a bunch of names here as well (although most people, I imagine, will be okay with one or two)

• IP entries: whatever IP addresses are associated with your NAS, add them here.

  • If you want to plan for the future, you can preemptively add IP addresses for planned-upgrades down the road here too! What I mean by this is, for example, my NAS (DS1817+) has a PCIE expansion slot and right now, I have a 1x10Gig NIC in there (192.168.15.1) — down the road, I might add replace this NIC with a 2x10G so it might be prudent for me to also preemptively add the IP for that second NIC (192.168.15.2)

---

 

authorityKeyIdentifier=keyid,issuer
basicConstraints=CA:FALSE
keyUsage = digitalSignature, nonRepudiation, keyEncipherment, dataEncipherment
subjectAltName = @alt_names

[alt_names]
DNS.1 = echo
DNS.2 = echo10
IP.0 = 192.168.10.1
IP.1 = 192.168.10.2
IP.2 = 192.168.10.3
IP.3 = 192.168.10.4
IP.4 = 192.168.15.1
IP.5 = 192.168.15.2


 

To break down the values I used here in my example, my NAS currently has five NICs, spread across two DNS/host entries:

• I have ‘echo’ and ‘echo10’ representing the labels I want to associate with my NAS for accessing via 1G and 10G.

• I added four IPs for the four 1G NICs on my NAS (192.168.10.1, 192.168.10.2, 192.168.10.3, 192.168.10.4)

• I added one IP for the single 10G NIC currently on my NAS (192.168.15.1)

• Since I know I want to one day upgrade by 1x10G NIC for a 2x10G NIC, I preemptively added an additional IP for that (192.168.15.2)

 

This is a V3 certificate extension config file and as it might suggest, this allows us to associate link all of those DNS/IP entries to the certificate we are about to create.

[Back to Top]

[D] Sign the NAS certificate

Now we have everything in place to finally create our signed NAS certificate:

openssl x509 -req -in echo_server.csr -CA ca.pem -CAkey ca-key.pem -CAcreateserial -out echo_signed.pem -days 3650 -sha256 -extfile echo_config.txt -passin file:pass.txt

[Back to Top]

Recap

So at this point, we have a bunch of files in our working folder:

• pass.txt - This file contains the password that we secure our private key with.

• ca-key.pem - This is the private key for the CA

• ca.pem - This is the public facing certificate

• ca.der - This is the public facing certificate, in a different format. For Windows users, this will be the file we install on each machine

• echo_server.key - This is the server.key file downloaded from our NAS when we clicked 'Create certificate signing request (CSR)

• echo_server.csr - This is the CSR we created from the echo_server.key file (instead of using the file in the archive we downloaded)

• echo_config.txt - A custom file we put together to define the different DNS/IP values to be associated with the signed certificate

• echo_signed.pem - This is the signed certificate! This is the echo_server.key file, signed by our CA

[Back to Top]

Step 4 - Add signed certificate to DSM

In DSM, we can now add the signed key:

1. Log into the NAS

2. Open Control Panel

3. Open Security

4. Select Certificate

5. Click Add

View fullsize
View fullsize
View fullsize

Note

I’ve noted sometimes that the add operation can go into a limbo state where it’s ‘working’ — give it a minute and if there’s no progress, reload the portal page and re-add it.

After you’ve added the cert, there’s just one step needed. Under Settings ➞ Advanced, set (or verify) the System default certificate is the certificate you just added.

If you have multiple NAS units that you want to fix this for, repeat Step 3-4 for each one. When you reload the browser, the insecure warning will be gone!

Since you’re not likely going to deal with this for a long time, I’d recommend making some copious notes itemizing what all of the files are and backing them all up somewhere because in 10 years, you’ll come back to this and have no idea what any of it is.

[Back to Top]

TLDR Steps

More-so a note for me when I have to revisit this down the road:

1. Install OpenSSL somewhere, navigate to the installation folder \bin

2. Create CA

   • Create a file, pass.txt with the strong password for private key in it

   • Create the private key: openssl genpkey -algorithm RSA -pass file:pass.txt -out ca-key.pem -pkeyopt rsa_keygen_bits:4096 -aes256

   • Verify the private key: openssl rsa -in ca-key.pem -check -noout -passin file:pass.txt

   • Use the private key to make our public root cert: openssl req -new -x509 -sha256 -days 3650 -passin file:pass.txt -key ca-key.pem -out ca.pem

     • Specify the country

     • Specify the Common Name

     • Everything else can be blank

   • Convert the root cert to the easier DER format: openssl x509 -outform der -in ca.pem -out ca.der

   • Install the ca.der certificate on the machines we will be accessing the NAS portal from

3. Get the CSR package from DSM

   • Log into DSM ➞ Control Panel ➞ Security ➞ Certificates

   • Settings ➞ Advanced ➞ Create certificate signing request (CSR) ➞ download this file (server.zip)

   • From server.zip, extract server.key to our working folder & rename it to something that makes more sense. The other file doesn’t matter

4. Create our own CSR: openssl req -new -key echo_server.key -out echo_server.csr

5. Create EXT file (echo_config.txt): Scroll up to see what goes in it

6. Sign the server key (from DSM) with our CA cert: openssl x509 -req -in echo_server.csr  -CA ca.pem -CAkey ca-key.pem -CAcreateserial -out echo_signed.pem -days 3650 -sha256 -extfile echo_config.txt -passin file:pass.txt

7. Add the certificate to DSM (these steps may change in 10yrs ha, but for now in DSM7 this is what I did)

   • Log into DSM ➞ Control Panel ➞ Security ➞ Certificates

   • Add ➞ Private key = echo_server.key, Certificate = echo_signed.pem

   • Settings ➞ Advanced ➞ set System Default to the certificate we just added

[Back to Top]

---

Product links may be affiliate links: MinMaxGeek may earn a commission on any purchases made via said links without any additional cost to you.

]]>Fixing Synology ‘Untrusted Certificate’TechDPThu, 10 Nov 2022 13:00:00 +0000https://www.minmaxgeek.com/blog/networking-upgrade-2022-46020ac0b21a9b86f0577e29e:60535320d9a09304764a163f:62fabced3163b53a97ddcdd2Earlier in the series, I tackled the problem of “what if the ISP goes down” and in this update, I’m back to tackle the counterpoint: “what if the power goes out”? While we don’t get a lot of extended power outages, I do know that the utility company has some planned outages for infrastructure upgrades in the near future. Since my wife works from home, it’s important that we maintain as much uptime as possible. I have a UPS on the network stack but in this post, I’m going to work on scaling back the power draw to maximize uptime.

---

 

Networking Series

This post is part of a series, check out the other posts!

---

 

Quick Links

• Starting point

• Endurance mode

  • Thinking out loud...

  • Implementation attempt (and failure!)

  • So what gives?

  • New plan

• Looking ahead

---

Starting point

In a previous update, network failover was something I addressed and generally, it works really well - during as blip, the network seamlessly fails over to cellular and comes back to fiber/cable when the connection is restored. We are in reasonable shape to deal with the next telecommunications outage but this begs the question — what if the telecoms are fine but there’s a localized power outage?

During an extended power outage, we could transition to using laptops but would still need to get network access. For convenience, and because the main switches would be powered down, the easiest way to go about this is to drop down to a single access point and all the critical devices can connect wirelessly. One limitation of my current networking stack is just how much of it needs to be powered on to implement this. Here, I’m referring to the number of devices (and their associated wattage) that I need to support with a UPS:

 

 

Not counting any power conversion inefficiencies, my UPS would have to provide somewhere in the neighborhood of 70-91W (using maximum power draw values) just to provide Wi-Fi to the house. According to CyberPower’s runtime calculator, this should give me 1.5-2 hours of runtime. Anecdotally, this is consistent with the actual runtime I was seeing during power outages but more is always better right?

I set a mental goal of at least 4 hours runtime so that Wi-Fi uptime could keep up with the runtime of the laptops on battery power and all of this, without expanding/upgrading my UPS at this time. As such, the only way to eek extra runtime out of the network-stack is to reduce its power draw.

Enter Endurance Mode.

[Back to Top]

Endurance Mode

With Endurance Mode, I wanted to cut out all the devices that wasn’t crucial to providing basic Wi-Fi coverage to the house. The easiest way to swap the entire network stack out for a single travel-router; the travel-router already has a built-in Wi-Fi access point, but depending on where I place the travel router (likely magnetized to the side of the rack), it’s very likely that signal quality throughout the house won’t be great. So I’ll plan for possibly [re-]using an existing access-point to provide house-wide coverage.

 

 

Beyond reducing the power-consumption, there are two soft-requirements for this project:

1. To minimize the amount of additional network gear ($) required to implement this and,

2. Make it so the actual process of cutting over is dead simple: during a power outage, I want to be able to cut over in seconds and I want the process to be brainless

[Back to Top]

Thinking out loud...

When I say keep the process simple, I mean that I don’t want to be fumbling around, unplugging and replugging network cables in the dark — everything should be pre-connected and ready to go so that it’s just a matter of flicking some switches.

At a high level, I wanted to find a way to fully connect both networks and use power-switches to control which set of equipment was live - something like this:

 

 

When I built the Network Rack Mk 2, I specifically went out of my way to put everything on independent power switches to facilitate this: during an extended power outage, I could toggle the power switches to turn off all of the blue devices and turn on the grey devices; after a quick network drop, we would be able to carry on our day with laptops over wireless. The magic sauce is in the dotted-red lines— a way to simultaneously connect two Routers to a single connection port on the Fiber ONT and also to connect my main access point to two devices at once.

[Back to Top]

Implementation attempt (and failure!)

If you search online, you can find a plethora of “ethernet splitter” type devices. They all come with very obvious caveats and disclaimers:

This Ethernet splitter allows two computers to share one Ethernet line ONE AT A TIME, but it DOESN'T SUPPORT Both computers to connect onto the internet Simultaneously. Just ONE output port is working.

Please note-the output port of this RJ45 network splitter does not support two computers connected to the network at the same time. Only one output port is working at a time.

The splitter cable expand one RJ 45 outlet to two standard 8P8C design. Please noted that only one port could be used at one time.

It does not support two devices to access the internet at the same Time. Only one device can be connected to the internet at a time.

So far so good. I specifically only want one network to be powered up at a given time anyways, so this is perfect! Being a sucker for this, I bought two of these too-good-to-be-true dongles: one to split the connections for the routers and one to split the connector to the access point.

 

 

On paper, this looks like it’s a perfect fit for my needs: during normal operations, the blue devices are powered on and the grey device is off — which meets the requirement of only having one device connected at a time. During endurance operations, the blue devices power down and the grey device powers up, also maintaining the single-device requirement.

This did not work as intended, I couldn’t get either port to work consistently (in the manner I wanted) and in the limited scenarios I could get it to work, the signal was reduced from 1000-Mbit to 100-Mbit. Obviously this is a stupid flop. This entire category of products almost feels maliciously labelled as ethernet splitters.

[Back to Top]

So what gives?

The big clue is the drop to 100-Mbit. When you look at these devices as ethernet mux/demux, it makes a lot more sense. So under the hood, ethernet cables generally have 8-wires running in them. For gigabit (or faster) networks, all 8-wires are utilized however, an older (slower) standard only requires 4-wires. With this bit of extra knowledge we can surmise that under the hood, these devices are splitting a [faster] 8-wire connection into two [slower] 4-wire connections.

So these “ethernet splitter” devices need to be bought in pairs (which is not always evident in product descriptions) and they drop your line speed to 100-Mbit (which is also deceptively ignored in most of the documentation). A better label would be to call it an ethernet multiplexer/demultiplexer (or compressor/decompressor)

The takeaway

Just don’t buy these types of products. They are too good to be true and will never do what you hope they will.

[Back to Top]

New plan

I originally balked at adding a switch to the picture because of the additional power draw but I stumbled upon the D-Link DGS-105 unmanaged switch that uses an astonishing 1.85W (and is wall-mountable which is a bonus); this way I can keep my normal-router and travel-router plugged in at all times, toggling between them using power switches.

One downside of this plan is that I need to have a dedicated access point for endurance mode (added cost). Finding a solution for the access point is a bit more complicated. I have a few options:

• Do nothing, just use the built-in WiFi on the travel router. This costs me nothing (power-wise) and initial tests suggested that it does have an astonishing broadcast strength in ideal conditions, where I plan to mount the unit (magnetized to a sheet of steel), the performance will suffer substantially

• Get another dedicated EAP225 (13W). I know the entire [wireless] network can be driven off a single EAP225 and this would be one of the cheapest (money and power) option

• Move my current EAP225 (13W) to be used in endurance mode and get a new access point to use as my ‘main AP’

This last option now lets me consider getting the first WiFi 6 (or even 6E) support in the house. There are a few candidates:

• EAP690E HD - 51W(!!), WiFi 6E, quad-band, massive throughput, not available yet

• EAP680E HD - power values not known yet (guessing 40W), WiFi 6E, tri band, not available yet

• EAP660 HD / EAP670 - 22W, WiFi 6, dual-band only

• EAP610 / EAP620HD / EAP650 - 15W, WiFi 6, dual-band only

 

---

 

Ultimately, I settled on the EAP690E HD as it was the no-compromise, check-all-the-boxes option. At time of writing, it’s not available yet, but I wanted to get everything ready so that, if the product reviews were positive, I could simply buy it, install it and be done. This is where I ran into the only (amusing) hiccup. The AP is damn huge. I was planning on stuffing it into the drop-ceiling, between joists but at over 11” in diameter, the specific spot I was planning on, didn’t work out.

I opted to move the AP location to a nearby spot and threw together a quick platform to simply place the future AP (for now, since the EAP225 is acting as my ‘main’ AP, I have that wired up there).

For the time being (while I wait for the access point to be released), I’ll need to do one additional step of moving the access point from the main network to the endurance network (and then back, when the power is restored). Once the AP is released, this won’t need to be done.

[Back to Top]

Looking ahead

As a follow up to Project Endurance, I’d like to look at splitting up the UPS coverage for the network to provide dedicated power coverage for certain devices (i.e., each NAS to be caught by a dedicated UPS).

The next network purchase will likely be the TP-Link EAP690E HD (again, assuming no glaring issues with reviews) followed by a router upgrade. The router I was waiting for, the TP-Link ER8411 is due to be released soon but I’m not in a rush to transition since my ISP doesn’t yet offer faster-than-gigabit speeds anyways.

Of course, in a never ending push of technology (and father time), my current main NAS (DS1817+ with 10G) is definitely getting long in the tooth, particularly re: running VMs — I really want to have VMs running off NVME SSDs for the next NAS, so this is something that I’ll need to keep an eye on over the next few years. In the meantime, I think I’d like to get a 2x10G NIC for my NAS so that I can keep the 2.5G (or less) traffic from congesting the 10G flow. While I’m in there, I might upgrade the fans in the NAS to Noctuas not because I’m having issues with the stock fans but to get in front of issues that will inevitably come along.

[Back to Top]

---

 

 

Product links may be affiliate links: MinMaxGeek may earn a commission on any purchases made via said links without any additional cost to you.

]]>Upgrading my Networking Gear (2022.4)TechDPFri, 04 Nov 2022 21:10:00 +0000https://www.minmaxgeek.com/blog/networking-upgrade-2022-36020ac0b21a9b86f0577e29e:60535320d9a09304764a163f:636478e39f327858eba24886When I had left off with the last update, I had a few bits of pre-requisite infrastructure work left to be done most of which can be summarized as “run a bunch of ethernet cables all over the place”. Since then, this work has been completed and all of the major networking pieces are in place (for now). This means I can work on refining the configuration and usability of the network.

---

 

Networking Series

This post is part of a series, check out the other posts!

---

 

Quick Links

• Hardware changes

  • Ethernet runs

  • Tuning the outdoor coverage

  • Indoor APs

• Ad-Blocking

  • DNS Sinkholing

  • Raspberry Pi

  • Getting sidetracked by tragedy

• Looking ahead

---

Hardware changes

Ethernet runs

At the end of the previous update, I was waiting on getting the rest of the house wired up; we had determined that the easiest/cleanest way to run the ethernet was via the outside of the house so I had run ethernet lines from my network hub to a staging area near the side of the house. Pre-terminating the ethernet ends and using couplers gave me two benefits:

1. I could finish getting the connections in the rack complete and get all of the runs in the ceiling tidied up beforehand and,

2. If there was ever a problem with a line running out of the house, I could troubleshoot more easily (rather than having to dig inside my rack) and if needed, I could (in theory), replace individual lines more easily

I made extensive use of Cord Dotz to ensure that all of the runs were as well-labelled as can be — in the event of failure, I didn’t want to have to sift through a sea of identical-cables to determined which was which.

Ultimately, I ran seven lines out of the house:

• Three for cameras

• One for an outdoor access point

• Two going to my wife’s upstairs office

• One going to the living room to provide a hard-line to the living room TV

Pro-tip

When choosing conduit, always size up — trying to go with a smaller conduit to reduce the ‘aesthetic impact’ is totally not worth the exponential increase in effort trying to snake a crap-ton of ethernet cables around bends etc.

 

 

[Back to Top]

Tuning the outdoor coverage

The one last step to do was to tinker with the orientation of the antennae to maximize coverage in the backyard. This was quite tedious actually as there were so many permutations to try out; I settled on trying the antenna at 45° intervals (angle and rotation) and for each configuration, I took a heat-map.

It makes sense that tinkering with the angle and rotation of the antennae would make a difference in signal quality but I was quite surprised to see just how substantial the difference would be, visually. I used WiFiman app to take the heat-maps.

Something to note: while your first instinct might be to maximize the signal throughout the entire space, you may want to reconsider and have the fall-off happen as you start to near the edge of your property (both for security and general WiFi ‘tidiness’ reasons). This is where having a heatmap can really help you visualize where to start cutting off your signal.

[Back to Top]

Indoor APs

For the living room and the upstairs office, I made use of wall access points (specifically the TP-Link EAP-235 Wall) I primarily chose this as a means of ‘getting a bunch of connections to the wall with just a single cable’ and the increased overall wireless coverage was purely a side benefit. Obviously, connecting three downstream devices simultaneously would cut the available bandwidth down to (roughly) 333 Mbit/s (at worst) but in some applications it may not matter.

View fullsize
Compact formfactor View fullsize
Takes a single ethernet through the back View fullsize
Provides WiFi and three downstream connections

For my applications, I was planning on only using one port at a time and the other two would just be a bonus; in the case of the upstairs office, I had a second cable run to provide 2.5G/10G as a dedicated higher speed line.

[Back to Top]

Ad-blocking

DNS sinkholing

Nobody likes ads and while I can understand the need for ads, what I dislike more are ‘bad’ websites - and I don’t mean “we’re going to hack you” (although those aren’t nice either) — I mean websites that are just done poorly. Also: why is it that the “important” websites (i.e., banking, government, taxes etc.) all just coincidentally happen to be built like crap? On the angle of ads, while I don’t mind the occasional ad, my god, some sites are just absolutely littered with them and it always seems to be super obnoxious ads too.

With the rest of the centralized networking in place, it made sense to finally look at centralizing ad- and malware-blocking. I went with a two prong solution of Pi-Hole and a private DNS provider like NextDNS.

Pi-Hole is a system that I manage/control and NextDNS (or similar) is a third party provider: both of these systems work by filtering web requests based on the domain-name: so for example, if you know that ads are being served by ads.somewebsite.com, you can block requests going to that url. The advantage here is that any device that connects through Pi-Hole (or NextDNS) will be ‘protected’ without any additional configuration.

Why use two systems?

The big differentiator here is that the Pi-Hole is really only [easily] accessible from within my network. For devices like my phone which I use in and out of the home, I wanted a solution that would provide me some ad- and malware-blocking out of the home as well.

It’s possible to use Pi-Hole out of the home but (at least for how I have it configured, using via VPN), it’s not a great experience. Next DNS has a free tier that allows for 300K queries per month which is perfect for running two phones.

[Back to Top]

Raspberry Pi

Pi-Hole can be run on a variety of platforms and a popular option is to run it using Docker however I opted to run Pi-Hole as a standalone appliance using a dedicated Raspberry Pi.

A love-hate (almost entirely -hate) relationship with Docker

I really like the idea of Docker: run a bunch of lightweight things that spin up and down easy-peasy. It’s all sunshine and rainbows except either [a] I want to encapsulate things that aren’t services (i.e., give me Photoshop as a Docker container) or [b] the stuff I want to run wants all-the-performance (i.e., video encoding) where the “low-power, low-footprint” ethos doesn’t really mesh well.

I like the idea of a dedicated, single-tasking thing that does one thing (DNS stuff) and does it well. I went with a Raspberry Pi 4 (8G) which I nabbed before prices went too stupid. I used a passive mountable enclosure to let me magnet-mount it to the side of the rack to keep the clutter down.

  View fullsize
Totally passive case View fullsize
Magnet mount  

I didn’t need the top of the line model but with the inflated prices, it was only a few bucks to max out the RAM so it made sense. At some point, I’d like to get a second Raspberry Pi and run a second [dedicated] Pi-Hole instance: this way I can do updates/maintenance on one Raspberry Pi without impacting the network as a whole. Given the chip (and Pi-specific) shortage, it’ll be awhile before I can tackle this.

[Back to Top]

Getting sidetracked by tragedy

Once I got my Pi-Hole in a good steady state, I made the move to run it from a SD card to a USB key and then finally a SSD. I’m not a Linux expert by any means so one thing I did learn to do was to make sure I had multiple backup copies of everything.

When setting up initially, I went for the full-fat GUI experience — the only thing this Pi was going to do was run Pi-Hole so there was plenty of horsepower left to run VNC and a GUI. Leveraging the easy-mode “SD Card Copier” makes things so much less annoying than trying to brute-force copy SD-cards. Fully expecting SD cards, USB keys and even SSDs to fail, I made multiple copies of everything as cold-spares

Nevertheless, something did go wrong. One time, when I was powering down the Pi, I tried powering it back up and was in Kernel Panic, unable to mount partition mode. I was not in the mood to troubleshoot this (esp. as a relative noob), so I went with the nuclear option: just reimage the SSD back to a backup, I lose a bit of configuration but I’m back up and running shortly. This was a very long night and even though I ultimately got back up and running, I’m not entirely sure why it broke or how it is working again now with effectively no change..

 

---

 

From previous testing with the equipment involved, I had determined the Raspberry Pi is unable to boot using the [SATA - USB B] and I quickly determined that this time, the SD Card Copier was unable to deal with [SATA - USB A]. After a handful of hours, I eventually got everything back and working again (albeit with a month old configuration). I had been meaning to get backups and synchronization setup but just didn’t get around to it. This was the push I needed 😉

To do this, I created a shared folder on the NAS as well as a new set of credentials with access to this share. Then, on the Raspberry Pi, create a convenient script (I put mine as /usr/local/bin/pi-backup) and I added the following:

 

#!/bin/sh

sudo umount ~/some_folder_name || /bin/true

sudo mount -t cifs -o username=NAS_user,password=NAS_password,uid=1000 //NAS_IP/NAS_shared_folder_name ~/some_folder_name

cd ~/some_folder_name/Pi1

pihole -a -t

This should mostly be straightforward, the only two things of note are:

1. The || /bin/true is appended to ignore any errors unmounting the NAS drive (i.e., if the drive wasn't mounted in the first place)
2. I added a subfolder Pi1 so that each individual Raspberry Pi could have its own distinct configuration backup

I modified the file for execute privileges (sudo chmod +x /usr/local/bin/pi-backup) and now I can quickly create a backup of the Pi-Hole configuration with just a simple call. Easily adjustable so that each [future] Raspberry Pi has a copy of this to be able to back it up. This leverages the built-in Teleport command to perform the configuration backup.

[Back to Top]

Looking ahead

I’d like to go down the rabbit hole of Raspberry Pi a bit more; at the very least, I’d like to run another unit as a failover — this way I can do maintenance/updates on one (or suffer some kind of failure) without being a pain on the network as a whole. I’d also like to get a couple more units to play around with Home Assistant and as a VPN box. With this many units, I’d be running out of space to magnet-mount them on the side of the rack so I would probably look at something like a POE rack-mount.

Over the upcoming year-end break I’d like to spend some time to tackle (or at least start looking at) two things:

1. Replacing the fans on my NAS units with some Noctua fans; nothing currently wrong with the fans on there but they are more than a few years old now and I’d like to get ahead of any potential problems while I have a break

2. Following the weirdness (for lack of a better word) with the Raspberry Pi and also as a follow up to my Data Loss, I’d like to spend some time to refine my data backup strategy to make it more robust.

All of this is me trying to avoid the big problem on the horizon: that I’ve effectively ran out of rackspace. At some point (next year perhaps?), I’ll consider going for a four-post rack and “finally” be done. Ish.

I have an idea for an upcoming [networking] mini-project thought, ‘Project Endurance’. That’ll be the next update!

[Back to Top]

---

 

 

Product links may be affiliate links: MinMaxGeek may earn a commission on any purchases made via said links without any additional cost to you.

]]>Upgrading my Networking Gear (2022.3)TechDPFri, 16 Sep 2022 20:55:00 +0000https://www.minmaxgeek.com/blog/smart-watch6020ac0b21a9b86f0577e29e:60535320d9a09304764a163f:63247a0b5e6e2e303b255b42My wife bought me a Galaxy Watch S3 Classic a ages years ago when it first came out and after recently getting the new phones setup, I was revisiting the thought of getting a new watch. Going against all the advice, recommendations and reviews out there, I ended up going with the OnePlus Watch. I’m only a few days into the new watch but I acclimatize quickly — and the reviews are right: the feature set is a bit bare so there isn’t a whole lot to have to get used to.

---

Quick Links

• Selecting an upgrade

  • Samsung’s offerings

  • Re-evaluating my needs

  • What about the OnePlus watch?

• So how does the OnePlus watch fare?

• Side quest to fix Chat/Hangouts notifications

• The final verdict

---

Selecting an upgrade

Samsung's offerings

After enjoying the Gear S3 watch, I bought my wife the successor, the Galaxy Watch. I initially bought her the larger 46mm variant (prioritizing battery life) but that ended up being a bit too bulky so I returned that for the smaller 42mm variant. Of course, while setting it up, I took the time to play with it as well and it was noticeably more responsive. The 46mm and 42mm variants only really differed in their physical size (and thus the battery capacity) and were otherwise identical in performance

 

	A	B	C	D
1		Gear S3	Galaxy Watch (46mm)	Galaxy Watch (42mm)
2	Year Released	2016	2018	2018
3	Weight	59g	63g	49g
4	Screen size	1.3"	1.3"	1.2"
5	Screen resolution	360x360	360x360	360x360
6	Always On Display	yes	yes	yes
7	Special Input	Rotating Bezel	Rotating Bezel	Rotating Bezel
8	Water/Dust Resistance	IP68	IP68	IP68
9	Protection	Gorilla Glass SR+	Gorilla Glass DX+	Gorilla Glass DX+
10	OS	Tizen 5.5	Tizen 5.5	Tizen 5.5
11	Chipset	Exynos 7270	Exynos 9110	Exynos 9110
12	CPU	Dual-Core 1GHz	Dual-Core 1.15GHz	Dual-Core 1.15GHz
13	RAM	0.75GB	1.5GB	1.5GB
14	Wi-Fi Support	Wi-Fi 4 (N)	Wi-Fi 4 (N)	Wi-Fi 4 (N)
15	Bluetooth Support	4.2	4.2	4.2
16	Battery Size	380 mAh	472 mAh	270 mAh
17	Charging	Wireless	Wireless	Wireless

 

Although the newer watch was definitely snappier, at the time (2018), there wasn’t a compelling reason to upgrade for the sake of upgrading since my watch was still plenty fine. When we compare to Samsung’s modern offerings we can start to see some substantial improvements:

 

	A	B	C	D	E
1		Gear S3	Galaxy Watch (46mm)	Galaxy Watch 4 Classic	Galaxy Watch 5 Pro
2	Year Released	2016	2018	2021	2022
3	Weight	59g	63g	52g	47g
4	Screen size	1.3"	1.3"	1.4"	1.4"
5	Screen resolution	360x360	360x360	450x450	450x450
6	Always on display	yes	yes	yes	yes
7	Special Input	Rotating Bezel	Rotating Bezel	None	None
8	Water/Dust Resistance	IP68	IP68	IP68	IP68
9	Protection	Gorilla Glass SR+	Gorilla Glass DX+	Gorilla Glass DX	Sapphire
10	OS	Tizen 5.5	Tizen 5.5	Android Wear	Android Wear
11	Chipset	Exynos 7270	Exynos 9110	Exynos W920	Exynos W920
12	CPU	Dual-Core 1GHz	Dual-Core 1.15GHz	Dual-Core 1.18Ghz	Dual-Core 1.18Ghz
13	RAM	0.75GB	1.5GB	1.5GB	1.5GB
14	Wi-Fi Support	Wi-Fi 4 (N)	Wi-Fi 4 (N)	Wi-Fi 4 (N) Dual-Band	Wi-Fi 4 (N) Dual-Band
15	Bluetooth Support	4.2	4.2	5.0	5.2
16	Battery Size	380 mAh	472 mAh	361 mAh	590mAh
17	Charging	Wireless	Wireless	Wireless	Wireless Fast (10W)

 

Looking at the modern watches, we can finally start to see a meaningful jump with the latest Galaxy Watch 5 Pro — which, of course, is the most expensive — $500 at time of writing!! There are a few things of note:

• Chargers

  It’s not immediately apparent just looking at a spreadsheet but Samsung changed the style of charger in 2019 so all my existing chargers would not work with any of the current generation watches.

  This is annoying because Nicole’s watch would still be using the older charger so we would have to have multiple sets of chargers lying around (at least until I buy her a new watch). The second annoyance is that I like to keep multiple chargers in different places so it would be an added expense to deploy those chargers for the new watch.

• Goodbye rotating bezel

  One of the best features of the Samsung watches was the brilliant use of the rotating bezel as an input mechanism - you had a solid, tactile thing that you could interact with to (literally) scroll through the watch interface with. I’m sure that the modern watches work just perfectly fine just using gestures and taps.

• Android Wear

  No more Tizen OS! I’m not sure how I feel about this. In theory, Google Wear is a great thing as it makes apps and support so much less tedious. That being said, I never really had any problems with Tizen and our use case is pretty simple.

[Back to Top]

Re-evaluating my needs

In light of the substantial asking price and the hassle of having to figure something out for charging the watch, I needed to figure out what I actually from my watch. It ends up being stupid simple:

1. Show me notifications from apps of my choosing

2. Provide a mechanism for “find my phone”

3. Be water resistant

4. Give me good battery life

You will note that this list is devoid of the biggest talking point of the entire smartwatch and wearable space: health and fitness. Almost every smartwatch will offer some kind of basic fitness crap: heart rate monitoring, steps etc. so even if I cared about this stuff, the ‘basics’ should be present on almost every option. For the most part, I’m treating the health and fitness functionality as whimsical party-tricks.

The big takeaway here is that, since health and fitness barely registers as a requirement, this opens up the doors to a lot more options (read: cheaper options).

[Back to Top]

What about the OnePlus watch?

Coming off the heels of getting the phones setup, the first entry on my radar was the OnePlus watch. From the way the watch is presented, it’s clear they focused on style over substance but if the watch manages to meet my needs and happens to be stylish? That would be a win all around.

 

	A	B	C	D
1		Gear S3	Galaxy Watch 5 Pro	OnePlus Watch
2	Year Released	2016	2022	2021
3	Weight	59g	47g	45g
4	Screen size	1.3"	1.4"	1.4"
5	Screen resolution	360x360	450x450	454x454
6	Always on display	yes	yes	yes
7	Special Input	Rotating Bezel	None	None
8	Water/Dust Resistance	IP68	IP68	IP68
9	Protection	Gorilla Glass SR+	Sapphire	None
10	OS	Tizen 5.5	Android Wear	Custom RTOS
11	Chipset	Exynos 7270	Exynos W920	STM32
12	CPU	Dual-Core 1GHz	Dual-Core 1.18Ghz	Not-specified
13	RAM	0.75GB	1.5GB	1GB
14	Wi-Fi Support	Wi-Fi 4 (N)	Wi-Fi 4 (N) Dual-Band	None
15	Bluetooth Support	4.2	5.2	5.0
16	Battery Size	380 mAh	590mAh	402mAh
17	Charging	Wireless	Wireless Fast (10W)	Wireless Fast

 

We don’t have a lot of visibility into the CPU or the operating system so it’s difficult to do an on-paper comparison (all three phones are using different operating systems too). Going by online reviews, consensus seemed to be that the OnePlus watch was very snappy, charged quickly and had incredible battery life. For downsides, the watch is disliked for being too basic and not offering very much by way of apps, watch-faces, customization etc.

The crazy amount of battery life on the watch changes the way I care about the charging situation — I don’t necessarily want (or care) to have watch-chargers in various places anymore — I can just keep one charger at my desk and charge it once a week for 15-minutes and be good to go. This has the added benefit of decluttering the limited space I have on my night stand. It all comes down to whether or not the watch, with all it’s missing functionality, can manage to meet by basic needs or not.

[Back to Top]

So how does the OnePlus watch fare?

I bit the bullet and ordered the OnePlus watch and I picked up a matching watch band to go with it. From an aesthetic perspective, I love it: they nailed the shape, curves, angles and the choice and finish of materials is spot on. I quite like the use of plastic for the underside as well: I personally like the feel and I imagine it’ll be more comfortable to wear on super warm/cold days. They definitely hit a home run for the way it looks and how it feels.

My comments are with respect to version 11_B.70.

• Watch Faces

  This is the first area where you start to see that the prime directive of the watch was to prioritize battery life and responsiveness, everything else be damned.

  The selection is abysmal. Full stop. There are 54 available watch faces to choose from and perplexingly, before you can set your watch to use a given face, you must preload that watchface onto the watch - of which you have 14 available slots. From an OS design perspective, I get why they did this but it’s just a little jarring when coming from a full-experience smart watch.

  My main complaint though is that there are not a lot of simple and clean [digital] options. I did manage to find three really clean analog options for which I am very grateful, but if you’re looking for a digital watch face, I only found ~1.5 options.

• Always On Display

  From what I’ve read, this wasn’t available at launch and was provided via a software update and people were a bit miffed about that. Thankfully I had the update available so I had access to this. Enabling the AOD though leads to a few perplexing things:

  Firstly there exactly four (4) always-on watch faces: two analog, two digital. Of the four, only one of the analog faces is clean and while the two digital faces are mostly clean, they have odd background shapes on them (one has a single skewed triangle, the other has a boatload of circles). I understand the need for hardcoded AOD watchfaces - it’s easier to manage resources when the faces are predefined but why can’t we get one uncluttered digital display as well?

  This brings me to a related problem: with the AOD displaying a watch, when I flick my wrist to wake the watch up, it then wakes to the watch-face I pre-loaded and then selected. It would be nice to just keep the same watch face for both (even if has to be the predefined one). All of this is very odd from a design perspective.

  Note: you can only set the AOD watchface using the watch directly which is odd.

• Interacting with the display

  With an emphasis on battery life, the watch has a maximum delay of 8 seconds before the watch turns the display off (or switches to the AOD).

  This is more of an oddity with the “why even bother having a main watchface”. After all, both the main and AOD watch face show the same things (actually my AOD displays the date, something the main watch face doesn’t show me).

• The Update Process

  Out of the box I had an update available which was nice, but the update process was peculiar. Firstly, the update process needed me to have 40% battery life remaining before it would allow me to start. For a watch that has a marketing-advertised number of two-weeks that seems excessive. Even with active-GPS, the battery life is advertised as 25 hours.

  The update process is pretty standard: download the update on your phone, transmit the update to the watch and then transmit a go signal to the watch to tell it to do it’s thing. The 40% battery requirement prevents you from even transferring the update to the watch (even if the watch is plugged in). The transmit process took a lot longer than I expected (feels like 15 minutes to transfer) and the phone app tells you not to use the watch or phone during the update.

  That warning isn’t messing around - I accidentally replied to a message on my phone and even though the watch seemed to finish the update, I had to redo everything (including the transmit) because reasons.

  All in all, a very peculiar update process.

• Swiping to 'go back'

  This is a minor one but on the OnePlus phones, to go back, you swipe from the right of the screen towards the left. On the OnePlus watch, it’s the other way — you swipe from the left of the watch face towards the right.

  Would it have hurt to have them do the same thing (or even allow swiping from either side to trigger a ‘go back’)

• Watch 'applet list'

  When you open up the watch applet list to see what you can do with your watch, the list of applets is static and predefined (which makes sense if you understand that they did this for RTOS battery and performance reasons)

  It’s just a bit of a pain for me since I have to scroll through a bunch of fitness crap I don’t care about to get to the app I want; this is also one of those times when I really miss the rotating bezel dial on my old watch. Thankfully, you can very easily bind the alternate [physical] button to any one of those apps.

  Pro-tip: when you’re setting up your watch, set the alternate button to launch the Settings app!

Most of these are relatively minor quirks of the watch and I can live with all of them if it delivers on the primary reason I got a watch in the first place — to pass along notifications.

Broadly speaking, notifications worked like a charm:

• GMail notifications

• Phone and text notifications

• Skype and Teams chats

• Camera alerts

One notable missing entry is Google Chats (formerly Hangouts) notifications - I struggled to get them to get them to the watch. Very intermittently, I would get notifications to the watch (maybe one notification for 50-100 messages). Poking a bit further, the problem isn’t the watch or it’s software — any notification that made it to the phone was faithfully passed along to the watch but Chat/Hangouts notifications just weren’t getting to the phone.

[Back to Top]

Side quest to fix Chat/Hangouts notifications

The problem ultimately is Google’s Chat/Hangouts trying to be clever. The underlying issue has some merit: if you are using a computer to chat and happen to have your phone on you, should you get notifications on your phone? Google seems to think ‘no’ and I can certainly see the motivation to prevent duplicate notifications. Where this falls apart is that it relies on ‘something’ figuring out which system you are ‘actively using’ at any given moment and how fast the cutover should be.

Getting the notifications to work for me involved a few things to check for:

• Prerequisites:

  • Make sure the app on the phone has notifications enabled and set the way you want

  • Make sure OnePlus Health app is set to synchronize that application’s notification to your phone

  • If needed (likely is),

    • lock the OnePlus health app so it doesn’t get closed to aggressively

    • disable power optimization for the OnePlus health app

• Google Chat (on Desktop). I happen to use the standalone app:

  • Check the Settings Gear window and uncheck ‘Allow chat notifications’

  • If you’re missing notifications for just one person (or perhaps a select few people), in the chat conversation view, click on their name at the top (it’s a dropdown) and make sure it the setting says ‘turn off notifications’ (meaning notifications are on) and not ‘turn on notifications’ (meaning the notifications are off)

• Google Chat (app): under settings, make ‘Mobile notifications’ is turned on and double check that your entire phone isn’t running in ‘Do not disturb’ mode (at least while you troubleshoot)

Once you get notifications working reliably, you can revisit whether you specifically need to have an application locked to stay running all the time or not.

But I want notifications on all devices (including PC)? Let me manage ‘duplicates’ on a per-device basis

Unfortunately it doesn’t look like you can do this. Google’s assumption is that you don’t have a bunch of computers in front of you all connected to chat and that you may want to reply back from any given one.

Now that I have [chat] notifications on the watch, I find the notifications to be bland — they are functional but that’s it. Their update indicated that ‘Enabled notification app icons for the most frequently used apps’ which is neat but it seems like Chat is not one of them, so my notification icons are generic. A minor gripe for now.

[Back to Top]

The final verdict

After tinkering with Chats notifications to get them to work, I can say that the primary requirement of the watch has been met: it gets me notifications reliably of the things I care about. As for some of the other [limited] things you can do with the watch:

• The alarm app, while basic, is perfectly fine

• I love the Do-not-disturb / bedtime mode of the watch (being a totally separate setting from any Do-not-disturb setting you might have on your phone)

• I don’t take it seriously, but for fun, sometimes I check out the sleep-tracking

• The remote camera trigger (for your phone’s camera) is neat albeit gimmicky

• The find-my-phone functionality works as advertised

• The built-music app can be run in two different modes: you can either use it as a standalone music player that plays music on your watch or you can use it as a remote control for media playing on your phone. This is actually really awesome if you happen to be listening to music on your phone but have your phone on your desk or something. My headphones have a button that lets me pause/resume but on the watch, I can also change songs and adjust the volume. This works for video as well (i.e., YouTube) which is awesome. I have no interest in using the watch as a local standalone music player.

While I originally wanted to have a few chargers (say, one for a travel bag), I think with the ridiculous battery life of this watch and fast charging, this bypasses any charge-anxiety concerns.

Don’t get me wrong, the software experience of this watch is definitely much rougher than it ought to be in 2022 and the software feels unpolished but for my primary objective of being a cost effective notification relay, this watch checks all the boxes I need it to.

[Back to Top]

---

Product links may be affiliate links: MinMaxGeek may earn a commission on any purchases made via said links without any additional cost to you.

]]>A smartwatch as a notification relayTechDPSun, 11 Sep 2022 17:25:00 +0000https://www.minmaxgeek.com/blog/new-phones6020ac0b21a9b86f0577e29e:60535320d9a09304764a163f:631ce2e0c60ee464ddb06ec1To make tech support and maintenance easier on me, both my wife and I use the same phone (happens to be the OnePlus 7T which we nabbed a few years ago for a great price). Generally speaking, we have no complaints with it although now in 2022, it’s just starting to get a bit long in the tooth. OnePlus published what I thought was a fairly reasonable maintenance/update policy back in 2021 where their flagship-tier phones would receive support for 3-major android releases and security updates for 4-years.

The phone launched in 2019 with Android 10 and as such, I would expect that support would go to Android 12 and/or 2023. At the time of writing, Android 12 is in beta-testing and we’re rapidly approaching the 2023 cutoff for security updates. I had a plan to upgrade the phone(s) in 2023 sometimes (depending on what models were on sale etc.). My wife’s luck had different plans though and she had the misfortune of cracking her phone screen with the most insignificant of drops.

Or maybe it’s an excuse to buy new phones. I’ll never say.

---

Quick Links

• 'Recovering' the old phone

• The contenders

• Enter: the refurb market…

• So how was the refurbished experience?

• No plan survives contact with the enemy...

• OnePlus 9 Pro (T-Mobile) setup, start to finish

  • [01] Prerequisites

  • [02] Converting to EU and unlocking the bootloader

  • [03] Updating the phone via OTA

  • [04] Preparing for Android 12

  • [05] Major update: moving to Android 12

  • [06] Doing updates on our terms

• Now to fix all the dumb...

• Not enough excitement

• Looking Back

---

'Recovering' the old phone

By the time I got to the phone, the AMOLED had leaked everywhere and almost the entire phone was blacked out. While we could walk away from phone with relatively little data loss (at most, 2-days worth of content not backed-up), how hard could it possibly to access/use a phone almost totally blind? Right?

I was hoping that it was just a visual problem — that I was just working blind. My plan was to fire up my identical phone and walk through the process with the two phones side by side and eyeballing where I had to touch on her phone. No dice, almost the entirety of the touch screen was non-responsive as well: there was only about a postage stamp worth of responsive touch screen in the bottom right corner. We got really lucky that the selfie-camera was not impacted by the broken screen I was able to use Face Unlock to get into the phone. By touching the screen every so often, I could keep the lock screen from triggering.

I was able to remotely install apps to the phone; specifically, I went with droidVNC-NG. I was hoping to use voice activation to launch the application (since I would have no way of launching it via the screen), but due to the ‘odd’ name of the app, we couldn’t figure out a way to get OK Google to voice-launch the app. Luckily, I happen to create a folder on the desktop in the bottom right and my launcher allows me to ‘Add Apps’ to this folder. The second stroke of luck was that the ‘Add Apps’ interface is a single-column list of app the apps. Of course, finding the app in this list was still a challenge but as a third stroke of luck, the app actually has a longer name: droidVNC-NG Admin Panel — and I was just barely able to make out the text in a small patch of visible screen. By connecting a keyboard to the OTG port, I was able to add the app to the folder.

I had to briefly switch back to a mouse in order to launch the app because the app icon was placed ever so slightly outside of the still-functioning touch range. Getting the app launched only a small step into the journey — an app like this requires a whole bunch of permissions! From here on, I switched back to the keyboard and bounced between the two phones side by side, counting how many tabs it took to select focus on the elements I wanted to select before pressing [enter] to click it:

1. Click the START button

2. An accessibility popup would show —> figure out how to click YES

3. This would then take me to a list of installed apps —> find droidVNC-NG from this list and click on it

4. This would take me to the [accessibility] details page of the app —> find a way to toggle the accessibility switch to on

5. This would trigger a confirmation popup —> find a way to click ALLOW

6. Clicking ALLOW just applies the setting and dismisses the popup — I’m still stuck in the app-details page and have to find a way back to the droidVNC-NG app

7. Getting back into the droidVNC-NG app, there would be a second popup waiting for me (file permissions) —> find a way to click YES

8. This would trigger an Android permissions confirmation —> find a way to click ALLOW

9. This would trigger an Android screen casting confirmation —> find a way to click START NOW

View fullsize
Ooof View fullsize
Finally able to launch the droidVNC-NG View fullsize
Somehow clicking the Start button View fullsize
If you manage to click Yes View fullsize
If you manage to click on your app View fullsize
If you manage to toggle the switch View fullsize
Getting back into the app View fullsize
After confirming that dialog View fullsize
Last dialog! View fullsize
Finally, no more nuisance dialogs View fullsize
Erureka

At this point, I’m able to VNC into the phone from a desktop or, comically, even another phone. I already had the phones hooked up to be able to access the NAS so I was able to quickly dump files to the network and also manually fired the backup/sync tools. Given the current state of this phone, I took this time to also:

• Triple verify that VNC is set to auto-start on boot up: I don’t want to have to do this again if the phone randomly decides it wants to reboot for whatever reason

• Disable all alarm clocks

• Disable the lock-screen password. Note: you can still authenticate the lock screen over VNC, but given the new nature of this phone, I wanted to streamline the process

• Turned off Bluetooth (to break any connections to accessories that wouldn’t really be used with this phone anymore)

• Extend the display sleep time to 30 minutes (not that it really matters)

• I enabled USB debugging on the phone and authorized my PC

[Back to Top]

The contenders

Before the OnePlus 7T, we previously used the OnePlus 3T and OnePlus 5 so we had an accumulation a few of the specialty chargers and cables over the years so, unless there was a monumental reason otherwise, I wanted to limit the search to OnePlus phones (or alternatively, OPPO or Realme phones if I couldn’t find a suitable option).

What’s so special about these phones?

All three brands use a proprietary charging system called ‘VOOC’ (sometimes branded as Warp-, Dash- or Dart-charge). You may have heard of USB-PD (power-delivery) or the popular Qualcomm Quick Charge; VOOC provides the same rapid-charge functionality as those other systems but instead of ramping up voltage during the charging process, it ramps up current.

All of these systems provide a means of rapidly-charging batteries but, charging via VOOC generally keeps the heat in the power brick rather than the phone (and power bricks generally can handle and dissipate heat a bit better than compact batteries in tight enclosures… The second notable advantage of VOOC charging is that the charging performance isn’t nearly as affected by what you’re doing on your phone while charging: you could charge while gaming or watching videos and the charge time would mostly be the same.

What I wanted

• Must: not be stupid-priced

• Must have: A phone that uses Dash/Warp/VOOC charging

• Must have: A performance uplift compared to our current phone (Snapdragon 855+ with 8GB RAM)

• Ideally have: some form of official dust/water resistance

• Ideally have: WiFi 6 support (none of the phones I looked at yet support WiFi 6e)

• Ideally have: better battery capacity than my current phone (3800mAh)

• Ideally have: for cameras, not be a major step back in the main, wide and macro camera

• Would be nice but not a deal breaker: 256GB storage. Apps, media etc. have really bloated over the years

• Would be nice but not a deal breaker: wireless charging (I probably won’t be able to use it anyways since I use magnetic mounts)

• Would be nice but not a deal breaker: dual-SIM

What was available from OnePlus at the time:

• OnePlus 10 Pro - $1030 CAD

• OnePlus 10T - $850 CAD (promotional pricing)

• OnePlus 9 Pro - $1170 CAD

• OnePlus 9 - $840 CAD

• OnePlus 8 Pro - $950 CAD

• Nord N20 - $380 CAD

For context, we bought the OnePlus 7T in 2020 for an absolute steal price of $350 CAD. Phone prices have definitely jumped a lot in the interim! With the exception of the Nord N20, each successive phone generally is more performant than the previous (and certainly more performant than our current phones). I threw together a quick sheet to group all of the specs that matter to me, highlighting substantial standouts:

	A	B	C	D	E	F	G	H
1		OnePlus 7T	Nord N20	OnePlus 8 Pro	OnePlus 9	OnePlus 9 Pro	One Plus 10T	OnePlus 10 Pro
2	Price available to me	Paid 350CAD	380CAD	950CAD	840CAD	1170CAD	850CAD	1030CAD
3	Year Released	2019	2022	2020	2021	2021	2022	2022
4	Last Year of Security Updates	2023	2025	2024	2025	2025	2026	2026
5	OS Support	Android 10-12	Android 11-12	Android 10-12	Android 11-13	Android 11-13	Android 12-14	Android 12-14
6	Max Cellular	4G	5G	5G	5G	5G	5G	5G
7	CPU	Snapdragon 855+	Snapdragon 695	Snapdragon 865	Snapdragon 888	Snapdragon 888	Snapdragon 8+ Gen 1	Snapdragon 8+ Gen 1
8	RAM	8GB	6GB	12GB	8GB	12GB	16GB	12G
9	Storage	128GB	128GB	256GB	128GB	256GB	256GB	256GB
10	Storage Type	UFS 3.0	UFS 2.2	UFS 3.0	UFS 3.1	UFS 3.1	UFS 3.1	UFS 3.1
11	Latest Wi-Fi	Wi-Fi 5 (AC)	Wi-Fi 5 (AC)	Wi-Fi 6 (AX)	Wi-Fi 6 (AX)	Wi-Fi 6 (AX)	Wi-Fi 6 (AX)	Wi-Fi 6 (AX)
12	Latest Bluetooth	5	5.1	5.1	5.2	5.2	5.2	5.2
13	Battery	3800mAh	4500mAh	4510mAh	4500mAh	4500mAh	4800mAh	5000mAh
14	Wired Charging	30W	33W	30W	65W	65W	65W	65W
15	Wireless Charging	No	No	30W	15W	50W	No	50W
16	Reverse Wireless Charging	No	No	3W	No	Yes, speed unknown	No	Yes, speed unknown
17	USB Power Delivery	No	No	No	Yes	Yes	No	Yes
18	Camera - Main	48MP f/1.6, 26mm, OIS	64MB, f/1.8, EIS	48MP f/1.6, 26mm, OIS	48MP, f/1.8, 23mm, EIS	48MP, f/1.8, 23mm, OIS	50MP, f/1.8, 24mm OIS	48MP, f/1.8, 23mm, OIS
19	Camera - Wide	16MP f/2.2, 17mm	None	48MP f/2.2, 14mm	50MP, f/2.2, 14mm	50MP, f/2.2, 14mm	8MP, f/2.2	50MP, f/2.2, 14mm
20	Camera - Tele	12MP f/2.2, 51mm (2X)	None	8MP, f/2.4, 75mm (3x), OIS	None	8MP, f/2.4, 77mm (3X), OIS	None	8MP, f/2.4, 77mm (3X), OIS
21	Camera - Front	16MP f/2.0, 25mm	16MP, f/2.4, 25mm	16MP, f/2.5, 25mm	16MP, f/2.4, 25mm	16MP, f/2.4, 25mm		32MP, f/2.2
22	Camera - Other		2MP, f/2.4 macro	5MP, f/2.4 color filtering camera	2MP, f/2.4 black & white	2MP, f/2.4 black & white	2MP, f/2.4 macro
23	Max Video	4K60, 1080@240, 720@960	1080@30, 720@120	4K60, 1080@240, 720@480	8K30, 4K60, 1080@240, 720@480	8K30, 4K120, 1080@240, 720@480	4K60, 1080@240, 720@480	8K24, 4K120, 1080@240, 720@480
24	Screen - Resolution	1080x2400	1080x2400	1440x3168	1080x2400	1440x3216	1080x2412	1440x3216
25	Screen - Refresh	90	60	120	120	120	120	120
26	Screen - Protection	Gorilla Glass 5	Yes, unknown, assuming v5	Gorilla Glass 5	Gorilla Glass 5	Gorilla Glass 5	Gorilla Glass 5	Gorilla Glass Victus
27	Screen - Always On Display	No	Yes	No	Yes	Yes	No	Yes
28	SIM Slots	2		2	2	2	2	2
29	Dust Resistance	Nothing official	Nothing official	IP68	Nothing official (on the model available)	IP68	Nothing official	IP68
30	USB Type-C	3.1	2.0	3.1	3.1	3.1	2.0	3.1
31	USB OTG	Yes	Yes	Yes	Yes	Yes	No	Yes

 

---

 

To begin the elimination process:

• Without even looking at prices, the OnePlus 8 Pro (released 2020 with Android 10) was out because it doesn’t really give us any more breathing room on software-support — OS support would only extend to Android 12 (no improvement) and security updates would only go another year.

• I eliminated both the OnePlus 10 Pro and OnePlus 9 Pro purely on the grounds of just being too damn expensive.

• The Nord N20 is a really interesting option — I would only be buying one and this would act as a stop-gap phone for Nicole until we both upgrade in 2023. At that time, I would end up buying two identical phones during the upgrade cycle because I don’t really want to maintain two separate hardware platforms. I ended up not going this route because it’s simply too much money for such a drop in everything across the board especially knowing that I would be spending big-bucks in less than year’s time.

The OnePlus 10T has some controversy with it:

• The elimination of the much beloved alert slider (not a big deal for me — on all my phones, I set the alert slider once when I get the phone and never touch it again

• For 2022 as a flagship-tier phone, the camera situation is … peculiar. I find, for the types of things I need to take pictures of, I primarily use the wide-angle camera which takes a 50% resolution hit compared to my current phone. I know ‘megapixels aren’t everything’ but if I’m going to pay that much money for a phone, I’m not really keep for that kind of a compromise

But comparing the OnePlus 10T against the OnePlus 9 is an interesting dilemma; for what is essentially the same price, the OnePlus 9 is probably the better phone but the OnePlus 10T gives me a longer runway before I have to revisit getting a new phone. At the time, the pricing of the 10T was with a limited time launch-promo (get the higher-spec variant for the lower-spec-price). There was no way I was entertaining this when the promotion ended.

Overall, I just didn’t like any of these options.

[Back to Top]

Enter: the refurb market...

Not liking any of the options I found: they were way too pricey and/or had way too many compromises, I started looking at the refurbished market (I wasn’t about to touch the used-market with a 10ft pole!) This ended up being a great idea as it gave me some options:

• OnePlus 8 @ $380 CAD

• OnePlus 8T+ @ $450 CAD

• OnePlus 9 @ $510 CAD

• OnePlus 9 Pro @ $640 CAD

Again, throwing these into a spreadsheet:

	A	B	C	D	E	F
1		OnePlus 7T	OnePlus 8	OnePlus 8T+	OnePlus 9	OnePlus 9 Pro
2	Price available to me	Paid 350CAD	380CAD	450CAD	510CAD	640CAD
3	Year Released	2019	2020	2020	2021	2021
4	Last Year of Security Updates	2023	2024	2024	2025	2025
5	OS Support	Android 10-12	Android 10-12	Android 11-13	Android 11-13	Android 11-13
6	Max Cellular	4G	5G	5G	5G	5G
7	CPU	Snapdragon 855+	Snapdragon 865	Snapdragon 865	Snapdragon 888	Snapdragon 888
8	RAM	8GB	8GB	12GB	8GB	12GB
9	Storage	128GB	128GB	256GB	128GB	256GB
10	Storage Type	UFS 3.0	UFS 3.0	UFS 3.1	UFS 3.1	UFS 3.1
11	Latest Wi-Fi	Wi-Fi 5 (AC)	Wi-Fi 6 (AX)	Wi-Fi 6 (AX)	Wi-Fi 6 (AX)	Wi-Fi 6 (AX)
12	Latest Bluetooth	5	5.1	5.1	5.2	5.2
13	Battery	3800mAh	4300mAh	4500mAh	4500mAh	4500mAh
14	Wired Charging	30W	30W	65W	65W	65W
15	Wireless Charging	No	No	No	15W	50W
16	Reverse Wireless Charging	No	No	No	No	Yes, speed unknown
17	USB Power Delivery	No	No	No	Yes	Yes
18	Camera - Main	48MP f/1.6, 26mm, OIS	48MP f/1.6, 26mm, OIS	48MP f/1.6, 26mm, OIS	48MP, f/1.8, 23mm, EIS	48MP, f/1.8, 23mm, OIS
19	Camera - Wide	16MP f/2.2, 17mm	16MP f/2.2, 14mm	16MP f/2.2, 14mm	50MP, f/2.2, 14mm	50MP, f/2.2, 14mm
20	Camera - Tele	12MP f/2.2, 51mm (2X)	None	None	None	8MP, f/2.4, 77mm (3X), OIS
21	Camera - Front	16MP f/2.0, 25mm	16MP, f/2.4, 25mm	16MP, f/2.4, 25mm	16MP, f/2.4, 25mm	16MP, f/2.4, 25mm
22	Camera - Other		2MP, f/2.4 macro	5MP, f/2.4 macro	2MP, f/2.4 black & white	2MP, f/2.4 black & white
23	Max Video	4K60, 1080@240, 720@960	4K60, 1080@240, 720@480	4K60, 1080@240, 720@480	8K30, 4K60, 1080@240, 720@480	8K30, 4K120, 1080@240, 720@480
24	Screen - Resolution	1080x2400	1080x2400	1080x2400	1080x2400	1440x3216
25	Screen - Refresh	90	90	120	120	120
26	Screen - Protection	Gorilla Glass 5	Gorilla Glass 5	Gorilla Glass 5	Gorilla Glass 5	Gorilla Glass 5
27	Screen - Always On Display	No	No	Yes	Yes	Yes
28	SIM Slots	2	2	1	2	2
29	Dust Resistance	Nothing official	Nothing official	IP68	Nothing official (on the model available)	IP68
30	USB Type-C	3.1	3.1	3.1	3.1	3.1
31	USB OTG	Yes	Yes	Yes	Yes	Yes

With these price points, the options become exciting (and viable) again!

• The OnePlus 8 for the absolute lowest-cost

• The OnePlus 8T becomes the budget option that sneaks in [official] dust- and water-resistance

• The OnePlus 9 is still a bit of an odd child that arguably is a bit worse than the 8T but gives you a +1 year on the support runway — with a small premium of $60 CAD, that was fair

• The OnePlus 9 Pro being the least compromises option

Unlike the online/new/retail pricing, the refurbished pricing separates the products with meaningful price jumps/drops. Deciding between the OnePlus 8T and the 9 Pro was a bit of a challenge: I ended up going with the One Plus 9 Pro. I figured the price premium for it is justified by the extra support runaway and having better specs all around.

[Back to Top]

So how was the refurbished experience?

That’s the one uncertainty of going with refurbished products. In theory, it should be fine, but there’s always a gotcha! I specifically with Amazon Renewed as a bit of a safety net (on top of the ease of returning things as a Prime member) for easy of returns if anything was even a bit out of place. The presentation of the Renewed program is pretty enticing:

The phones arrived the next day in generic boxes without any visible damage of any kind - they even came with a OnePlus (30W) charger and cable! I was expecting a generic phone charger so this was a great surprise. Both phones powered up fine and, as advertised, were both carrier-unlocked (to be fair, I was only able to test connectivity with the Telus network, but the phones did indicate they were network-unlocked). I was expecting this to be a walk in the park - I’d spend a couple hours tinkering and Nicole would have her phone up and running in no time.

Am I foreshadowing too much?

[Back to Top]

No plan survives contact with the enemy...

The first clue that something might be amiss was seeing the specific model was the LE2127 — this is the T-Mobile variant although initially I wasn’t phased by this. After all, our current phones (HD1907) were also T-Mobile variants and I didn’t have any major issues with it: the phone shipped with the normal Oxygen OS operating system and I was able to take updates just fine. At this point, the only thing that I knew was that the T-Mobile variant was a single-SIM model only; this was only a small disappointment rather than anything earth shattering.

When I sat down to start running the phone through the gamut of setup tasks, I rebooted it and finally saw what I had missed on the first boot-up. I would have been totally okay with having a jarringly bright-pink splash screen on boot-up, if the OS that booted up was Oxygen OS (the OS that comes with OnePlus phones). Nope. It was a T-Mobile bastardized version of the OS that would only accept updates provided by T-Mobile. This had to go.

This is very much ‘standard phone setup’ for me so at this point, I’m irked but not concerned:

1. Install drivers and prerequisites

2. Enable Developer Options and turn on USB debugging and OEM unlocking

3. Reboot to the bootloader and unlock it. This wipes the phone

4. After the phone comes back, redo step 2 (because the phone was wiped)

5. Reboot to the bootloader and flash a temporary root environment

6. Side-load Magisk and root the device permanently

7. Side-load Oxygen Updater and do a local update to the latest OS

8. Once you’re done all of this, now you can log into the phone and setup/restore your apps etc.

When I jumped into the bootloader, I saw a message I haven’t seen in forever:

 

> fastboot oem unlock

FAILED: (remote: 'Please flash unlock token first.')
fastboot: error: Command failed

Thinking I had somehow forgotten to enable OEM unlocking and annoyed, I rebooted into the OS and saw that I hadn't forgot to do this at all: OEM unlocking was enabled. Despite the ‘OEM unlocking’ setting being set to ON, this was not going to be an easy-peasy setup. Turns out there’s a huge dance you’re “supposed” to do.

 

1. First, get the unlock code
   > fastboot oem unlock

   (bootloader) Serial Number:
(bootloader) ====================================
(bootloader) [8-character-string]
(bootloader) ====================================
(bootloader) Unlock code:
(bootloader) ====================================
(bootloader) [32-character-string]
(bootloader) [32-character-string]
(bootloader) ====================================


2. Now go to the unlock website

3. Fill in your IMEI, serial number (- thats the 8-character string) and the unlock code. It's not eimmediately obvious but you have to combine the two 32-character-strings into a single 64-character code
4. Wait 7-14 days and they will/may get back to you with an unlock code of some sort, I'm not entirely sure.
5. You're supposed to save the unlock code as a file (i.e., unlock_token.bin)
6. Now you can unlock the right-to unlock your bootloader:
   > fastboot flash cust_unlock unlock_token.bin

   target reported max download size of [SOMENUMBER] bytes
sending 'cust-unlock' (0 KB)...
OKAY [ 0.005s]
writing 'cust-unlock'...
(bootloader) Device is unlocked. OKAY [ 0.005s]
finished. total time: 0.010s

7. And now you can unlock your bootloader:
   > fastboot oem unlock

   ...
OKAY [ 0.025s]
finished. total time: 0.026s


That’s right. You are expected to wait 7-14 days to do something that should take milliseconds to do. And this is one of the easier options! Some carriers don’t even provide this and some other brands (i.e., Samsung, LG) may not even allow unlocking of the bootloader period.

[Back to Top]

OnePlus 9 Pro (T-Mobile) setup, start to finish

On the off chance you stumbled on this page trying to set up your OnePlus 9 Pro, the following is a fairly complete documentation of everything I needed to do to get the phone into an acceptable state. There wasn’t a chance in hell I was going to wait for T-Mobile to get back to me with the unlock code — we’re going bypass all of this. If you’re following along, you’ll need two things:

1. A Windows PC and

2. To accept that your phone will get wiped and it’s possible for things to go really wrong!

 

---

 

At a high level, I need to do a few things:

1. Replace the T-Mobile OS with Oxygen OS by any means necessary (reference XDA thread)

2. Get the phone rooted (reference XDA thread)

3. Get call recording setup (reference GitHub repo)

4. Finish setting up the phone

5. Fix any stupid-behaviour or quirks as I encounter them

[Back to Top]

[01] Prerequisites

Basic Connectivity

You’ll want to download a few things before we get started:

1. Download the latest Android SDK Platform tools. Extract them somewhere; for this example, we’ll use C:\android-platform-tools

2. Download and install the OnePlus drivers. If Windows update doesn’t sort you out, you can get some drivers from here

3. Download and extract the Google USB driver

4. Download the Qualcomm USB drivers. I got mine from here. Extract this somewhere

Before we continue on, we need to prepare a few things:

• Plug your phone in now, select File Transfer / Android Auto as the USB connection mode

• If there is a missing driver in Device Manager, point the drivers to the OnePlus drivers you installed above. You should be able to open My Computer and navigate into the phone’s internal storage without issue

• Enable Developer Options. On normal phones, you’d tap on Build Number a bunch of times, but for the T-Mobile, you’ll need to tap on Software Version

• In Developer Options, enable the following settings:

  • OEM Unlocking

  • Advanced Reboot

  • USB Debugging — if your phone is plugged in at this point, you will get a confirmation if you want to trust the RSA thumbprint (of the computer); go ahead and check the box and trust it

Verify ADB Works

• Open command prompt and navigate to the platform tools you extracted earlier (C:\android-platform-tools). Run adb devices and you should see a single device listed. Keep this command prompt window open until we are all done this guide.

 

> adb devices

[8-character-string] device

Verify Fastboot Works

• Go ahead and run the command adb reboot bootloader. This should reboot the phone to a scary looking screen. Check Windows Device Manager to verify that there aren’t any devices that are missing drivers. If you need drivers, use the Google USB driver you grabbed earlier (or you can also get this from Windows Update —> Optional Updates —> Driver Updates)

 

 

• Run the command fastboot devices and this should also return a single device listed

 

> fastboot devices

[8-character-string] fastboot

You can either press the power button or run fastboot reboot to reboot the phone back to the normal desktop

Verify EDL Works

Getting your phone into EDL mode can be a bit tricky and since the screen is off, there isn’t an easy confirmation that you did it right. Before we get started, you’ll want to have Windows Device Manager open and have easy access to the Qualcomm Drivers you extracted earlier.

1. Power the phone off

2. Leave the USB cable plugged into the computer and unplug it from the phone

3. Hold down the volume-up and volume-down buttons simultaneously. I hold mine down for a 3-count before proceeding. Do not let go of these buttons yet

4. While still holding the volume-up and volume-down buttons down, plug the USB cable into your phone

5. Continue holding the volume-up and volume-down buttons for a 2-count and then release them

If you did everything right, the screen will not turn on. Quickly go to Device Manager, and check for devices that might be missing drivers — should be named QHSUSB_BULK or similar. Point it at the Qualcomm drivers you extracted earlier. Once the drivers install, the device should now be recognized as Qualcomm HS-USB QDLoader 9008.

To get out of this special mode, I held the Volume Up and Power button for 15-seconds. Once you’ve verified that EDL works, we can actually get started!

[Back to Top]

[02] Converting to EU and unlocking the bootloader

1. Disconnect your phone from the computer and power off your phone

2. As per the post, download the two files we will need

   • Conversion package (MD5: 8CC3E5E81D15E28BB10DC5F1E7BD4133). Extract this to a folder somewhere

   • T-Mobile modems (MD5: 1F23347CA9E2924EC4E9D67982857732). Extract the file to our platform tools folder, C:\android-platform-tools

3. From the conversion package, run MsmDownloadTool V4.0.exe. It takes a moment to launch so be patient!

4. Once it launches, you will be presented with a login dialog, select Other and click Next to continue

5. Once you have logged in, the main application will launch. Click Start to set the application in listening mode. By putting the application in listening mode, once the phone connects, the application will start to do it’s thing without any further interaction from the user

6. Confirm with yourself that you want to continue

7. Do the funky volume-up and volume-down button-press thing to get the phone into EDL mode and connected to the computer. If you’re successful, the MSM application will detect it and do it’s thing all on its own. If you miss the timing and the phone just boots up normally, go ahead and power down and try again

8. Walk away for 10 minutes and let the MSM application do it’s thing. The phone will reboot automatically when it’s done: you’ll come back to a “Welcome to Android” wizard

9. Once you’ve successfully got to the “Welcome to Android'“ wizard, go ahead and click Stop in the MSM app and then you can close the app

  View fullsize
MSM App - Login View fullsize
MSM App - Startup View fullsize
MSM App - Listening  

At this point your phone is has been converted to an EU edition of the phone (which is a good thing). You no longer have a locked bootloader and you can now get all of the updates that have been available for a long time (and not ‘approved’ or made available by T-Mobile).

Go ahead and finish the Welcome to Android wizard — just do the bare minimum to get to the desktop. Don’t get too excited and go and log into the phone and install apps or anything — we’re going to be wiping the phone at least once more so you’ll have to do this all over again.

1. Connect to Wi-Fi or cell

2. Disconnect and reconnect your phone to the computer and select File Transfer / Android Auto as the connection mode

3. Enable Developer Mode and turn on:

   • OEM Unlocking

   • Advanced Reboot

   • USB debugging (and re-authorize your computer)

4. In the command prompt dialog, reboot to bootloader using adb reboot bootloader

 **\> fastboot oem unlock** >... OKAY [ 0.025s] finished. total time: 0.026s

You will be prompted on your phone on whether you want to proceed or not - unlocking will wipe all data on the phone. Go ahead and confirm to continue. The phone will self-wipe and reboot on its own.

After the phone finishes and reboots, complete the ‘Welcome to Android’ wizard again. You know the drill by now:

1. Connect to Wi-Fi or cell

2. Disconnect and reconnect your phone to the computer and select File Transfer / Android Auto as the connection mode

3. Enable Developer Mode and turn on:

   • OEM Unlocking - this time confirm that you can’t toggle this because the bootloader is already unlocked!

   • Advanced Reboot

   • USB debugging (and re-authorize your computer)

 

 

The last thing we need to do now is to restore the T-Mobile modem. In the command prompt dialog, reboot to bootloader using adb reboot bootloader and run the following commands:

 

> fastboot flash modem_a NON-HLOS.bin

Sending 'modem_a' (214892 KB) OKAY [4.992s]
Writing 'modem_a' OKAY [0.614s]
finished. total time: 5.628s

> fastboot flash modem_b NON-HLOS.bin

Sending 'modem_b' (214892 KB) OKAY [4.992s]
Writing 'modem_b' OKAY [0.693s]
finished. total time: 5.709s

Now you can reboot your phone by pressing the power button or running fastboot reboot.

[Back to Top]

[03] Updating the phone via OTA

Now that you have a real OS that can get the widely available updates, we’re going to want to go ahead and get the phone updated to latest:

1. Connect to Wi-Fi

2. Use the built-in updater (Settings —> About Device) to go ahead download and install the updates. After each update finishes, you’ll need to restart

3. After each update, reboot into the bootloader via adb reboot bootloader

4. Flash the modems:

   • fastboot flash modem_a NON-HLOS.bin

   • fastboot flash modem_b NON-HLOS.bin

5. Reboot back to the desktop using the power button or fastboot reboot

6. Once you’re back to the desktop, check the current build version — we’ll want to stop when we get to 11.2.10.10 — this is the last Android 11 update.

Stop before updating to Android 12

If you’re not paying attention to the exact version numbers, at some point, during one of the updates, you should get a caution prompt ‘the update you’re about to apply is a major update’. That update is the update to Android 12.

Don’t apply the update yet, go do the next step first,

[Back to Top]

[04] Preparing for Android 12

Prerequisites

Once you’re on version 11.2.10.10, the next update will push us to Android 12 and there are a few things we’ll want to do in preparation. On your computer

• Download the last Magisk installer from here

• Download a root checking application from here

• Download the latest Oxygen Updater from here

• This Untouched/Magisked (BA) package for version 11.2.10.10 (MD5: 936F2D6C5ABB19908DD20AEFA5AFF052). From this archive, extract the magisk_boot.img file to the platform tools folder, C:\android-platform-tools. If you find yourself on a slightly different version than 11.2.10.10, have a look at this XDA thread, you may be able to find a matching package — just remember to look for the ‘BA’ variant since this is now an EU phone.

Side load the applications:

1. Connect your phone to the computer for file-transfer

2. Copy the APKs to your phone (the \Downloads folder is fine)

3. Using a file manager like Google Files, navigate to that folder and run those APKs to install them (you’ll be prompted to grant rights to access that folder as well as perform the installation)

Booting to Temporary Root

1. [Optional] Using the root checker app, verify that we don’t have root

2. Using the command prompt dialog, reboot to the bootloader: adb reboot bootloader

3. Once you get to the bootloader, we can go ahead and reboot into a temporary rooted-environment

 

> fastboot boot magisk_boot.img

Sending 'boot.img' (98304 KB) OKAY [2.300s]
Booting OKAY [0.280s]
finished. total time: 2.708s

Once you’re back in the desktop, you can use the root-checking application to confirm that you have root — you’ll get a prompt asking you for root permissions (you’ll need to grant it for toot to be detected)

  View fullsize
No root! View fullsize
Root!  

Getting Permanent Root

The phone only has root until it reboots — and we have a lot more rebooting to do! To get permanent root we need to do the following:

1. Launch the Magisk app

2. In the first box, ‘Magisk’, click Install

3. You will get some options, choose Direct Install (Recommended)

4. Once it finishes, go ahead and reboot

5. Once you’re back, run the root checking application to confirm that you still have too

Fixing the Modems

1. On this XDA thread, download the Data_Retain.zip attachment. Copy this to your phone

2. Launch the Magisk app

3. At the bottom, select the Modules tab

4. Click Install from storage and install the Data_Retain.zip attachment

5. After the package is flashed, go ahead and reboot

[Back to Top]

[05] Major update: moving to Android 12

Using the system update (Settings —> About Device), go ahead and apply the major update (at the time of writing, the version number of the update happened to be 11.C.47_1470_202203102117).

DO NOT REBOOT AFTER THE UPDATE.

After the update completes, we need to apply root to the new OS:

1. Launch the Magisk app

2. In the first box, ‘Magisk’, click Install

3. Select Direct Install (Recommended) —> DON’T REBOOT

4. Go back and select Install to Inactive Slot (After OTA). It will prompt you that you’ll need to reboot for it to take apply. Continue with the installation

5. Now you can reboot

After you’re rebooted, go ahead do some quick checks

• Verify the OS is Android 12

• You may need to re-enable developer mode and/or re-enable the USB Debugging setting

• Verify that you still have root

Once that’s done, download the local update application from here (MD5: ADF932B2FD4C2A2B379C9427197B6B6A) and copy it to your phone and install it. This is needed because OnePlus decided to remove the ability to do local updates with their Android 12 OS. This APK restores that functionality.

[Back to Top]

[06] Doing updates on our terms

These steps are the process I use for long term maintenance of the phone. At a high level: grab latest package from Oxygen Updater —> Update (don’t reboot) —> Preserve root —> Reboot.

1. Launch the Oxygen Updater application. If it’s your first time running the application, walk through the wizard; make sure that your device is set to OnePlus 9 Pro (EU) and that you’ve selected the Stable (full) updates. Once finished, go ahead and download the latest update. At the time of writing this was 11.C.62_1620202206251349.

   • If this is your first time, make rough note of the file size and the filename of the update. In my case, the filename was b906713f52384ca3b1f64eaa36c6db0c.zip and was 4.16GB

2. Once the download is complete, you can close the app and System Update application

3. Click the gear in the top right corner to perform the update —> DON’T REBOOT

4. As before, launch the Magisk application

   • Select Direct Install (Recommended) —> DON’T REBOOT

   • Go back and select Install to Inactive Slot (After OTA). It will prompt you that you’ll need to reboot for it to take apply. Continue with the installation

   • Now you can reboot

5. After the reboot, you can confirm that your phone has the expected version

6. Now you can delete the downloaded update using any means you want (either from a file manager on the phone, or you can connect to a computer and delete it that way). The update will always be in the root path of the phone and will always be a .zip file that matches what you noted in Step 2.

[Back to Top]

Now to fix all the dumb...

Now, I’m ready to log into the various services (Google account etc.) and get the phones setup. In the brief time I spent setting up the phones, I’ve found a few really dumb things with Android 12; most of these stupid changes are pushed down by Google, but some of it from the OnePlus implementation of Android 12 (or possibly made worse).

• No local updates

  We already fixed this above but for reasons unknown, OnePlus removed the ability to perform local OS upgrades.

  Fix: download this (MD5: ADF932B2FD4C2A2B379C9427197B6B6A) and install it. This will create a shortcut in your launcher called ‘System Updater’.

• Swapping out the Dialer

  OnePlus is now using a strictly-Google phone stack (phone, contacts, messaging apps). I don’t mind the messaging app but the phone/contacts app is horrendous - says me, someone who averages one call a year - maybe. As an additional nuisance, the Google phone app mangles support for call recording as well. So it’s ugly and nonfunctional.

  I really tried to find a way to not use third-party phone/dialer apps, but after spending a couple hours trying to brute force a pre-Android 12 dialer app into Android 12, I just gave up and settled for something that would be simple and clean.

  I settled on Simple Dialer and it lives up to its namesake.

• Restoring call recording functionality

  The pre-Android 12 OnePlus dialer supported call recording and while it was a bit kludgy to get setup initially, once set up (requiring a Magisk module to activate), sonce setup it just worked.

  The Google dialer supports call recording — but only for certain locales as determined by Google. And the recording has all kinds of issues depending on whether or not the screen has gone to sleep (i.e., because you raised the handset to your ear) before the call has connected or not.

  I ended up going with a dialer-agnostic recorder, ‘Basic call Recorder’. Installation is super straightforward: using Magisk, flash the zip package. OnePlus devices need one extra step: extract the APK embedded in the zip and install that as well.

  Note: if you’re going down this route, it’s probably a good idea to double check the laws in your area regulating this.

• Restoring the Power Menu

  Android 11 introduced a really cool thing to the long-press Power Menu — the ability to add smart-home shortcuts there! Think of having a button right there for interacting with any of your Google Home devices! (link). Since this was such a great feature, the natural thing to do was to remove it in Android 12.

  I feel like this problem may be magnified by the OnePlus OS skin: I can see the old power menu controls but when I go to try and add controls, it just refuses to let me.

  Fix: Install Classic Power Menu which replaces the OS power menu where I can customize some of the reboot options available. The headline feature however, is that it brings back the smart-home controls to this menu. I disabled the entry for Google Pay/Wallet since that’s not supported by my bank in Canada anyways.

  This app uses the Accessibility service to hijack the power-button long-press event to display its own custom view and because of this reliance on the Accessibility service, there is a delay — you see the stock power menu for 100ms or so and then the override kicks in. The author provides a way to avoid this delay using an XPOSED module, but I couldn’t get it to work at all when the module was active. This feels like an OxygenOS thing so maybe I’ll revisit this in a few months. For now, I can live with the slight delay.

• No more support for LT2P

  Android 12 dropped support for anything but IKEv2 VPNs which, really, isn’t a bad thing per se, just an inconvenient thing - it’s super awesome to just have an L2TP VPN baked into the pulldown menu that I can very quickly access.

  On my networking side, it looks like support for this is coming/already-here; it just happens that my specific equipment isn’t supported yet. Something to look forward to I suppose.

  In the meanwhile, the workaround is to reconfigure my network to use OpenVPN. On the phones, I can create two easy shortcuts “VPN ON” (which auto connects the specific OpenVPN profile) and “VPN Off”. There’s only one caveat (so far) of not being able to access my custom private-DNS provider whilst on VPN, but [a] that’s a fairly minor inconvenience and [b] technically can be circumvented by toggling the private DNS provider.

[Back to Top]

Not enough excitement

Because all of this complexity and hoop jumping wasn’t enough fun, I managed to [accidentally] self-sabotage myself. The best part is that this happened after I had got the phone updated to the latest Android 12 - I was ready to finally log into my Google account and get apps installed etc. I was going to call it a night (it was 4AM) before tackling the apps in the morning so I did one final reboot just sanity check that I was in a good state.

And the phone refused to boot: it was stuck on the very first power-on screen (before the OS starts to boot) with a Qualcomm crash-dump. It struggled to even load the Android Recovery environment - the only environment that it would load with any kind of consistency was the bootloader.

I’m not sure what exactly happened but this is what I think happened:

• After updating to Android 12 (and doing multiple reboots in the process), I did one last reboot. This was while the phone was plugged into the computer

• While it rebooted, I unplugged the phone to bring it closer to me

• Somehow during that reboot, the phone kicked into EDL mode

• I still had the MSM app running in the background in listening mode. After all, to my understanding, the app would only do it’s thing if you went through the annoying process to get the phone into the special EDL mode

• Unbeknownst to me, the phone was in EDL mode and the MSM app started to brute force rewrite the phone

• And with perfect timing, I disconnected the phone from the computer

Thankfully, I was able to get the phone back into EDL mode and brute force the entire process again (starting essentially from scratch again).

This is why in my steps above, I explicitly state that after the MSM app has finished, you should (at the least), stop it. Better yet, exit the application to completely remove the possibility of this happening.

[Back to Top]

Looking Back

Was it worth it?

Out of the box, the hardware was fine - great even, considering I got OnePlus chargers when I was expecting generic no-name chargers. The phones appear to be in excellent condition and there was only one oddity I noticed: that the phones (while in T-Mobile mode) refused to take updates (that were pushed by T-Mobile). I recall trying it twice, on each phone: downloading the nearly 4GB update, installing, rebooting and then being presented with the same update again as if it didn’t take.

To answer the question, given that I broke the phones into submission, absolutely it was worth it. The underlying hardware is plenty good and I get the phones to do what I want and how I want it and at a fairly ‘great’ price. There some hairy moments where I wasn’t sure if I was going to be dead in the water with a locked bootloader or if the phone was going to even boot (see: excitement); it was hella frustrating to have to redo everything more than few times as I stumbled along but I eventually got the phones broken down into submission.

If I had to use the phones without breaking them into submission (even assuming that I was able to get the officially supported OS updates working)? Probably not for two reasons

1. T-Mobile is particularly lacklustre on getting updates out in a timely fashion (this isn’t just a T-Mobile things, all carriers are abysmal at it)

2. It doesn’t matter what kind of a deal I got with the phones - if it doesn’t do the things I want, then it’s no good to me at all

But this is me. Everyone is different - perhaps it’s not as big a deal if you’re just using a phone for Instagram-only.

If I had bought the phone directly from OnePlus, then all of this hassle wouldn’t have really mattered — I’d be able to [easily] unlock the bootloader and be on my way. The downside of course is paying nearly twice the price.

Rage against the industry

While this particular phone happens to exhibit problems associated with T-Mobile, they are by no means alone in this. All carriers (in North America it seems) are really, really bad about it. I think toxic might be a better word to describe it. A phone is just a thing that happens to operate a network that provides connectivity. Imagine if your washer-dryer could only receive power from the outlet if you bought the washer-dryer from an appliance store associated with power company A. Or if you were building something but your none of your screws would since into wood because your screwdriver wasn’t the same brand.

I’m not sure about other parts of the world, but in Canada, I can’t even [easily] walk into a store and buy a phone plain and simple. No account, no plan, just pay for hardware and walk out. “Sorry you need to have a plan”. The sheer hoops that people have to jump through to change carrier is nightmarish. Locked SIMs, locked bootloaders, all of it is so anti-consumer.

Root/Jailbreak is ‘dangerous’

This is an amusing one and while it’s not completely false - yes, if you root/jailbreak your phone and then do certain things, you can do bad things. I get that, there’s some truth to that. But how many times have people run into a problem with a computer — an objectively ‘higher tier’ device and the ‘solution’ is “Oh! you need to run as administrator” or “you need to sudo command”. Nobody blinks an eye when you have to sudo/Administrator something on a computer but on a phone? It’s doom and gloom.

And the audacity to try and limit things because you’re rooted/jailbroken? Try logging into a bank website on a computer browser with admin privileges. Huh - you can do it. No fuss. Try connecting a bank app on a rooted device and you get warnings of doom and gloom. This is from ‘reputable’ companies like banks which have such a focus on security that your password has to be a whopping 8-characters long and if you’re lucky, they’ll even make you use upper and lower case characters. /s

There ought to be hoops to jump through to root/jailbreak your device: unwary users can cause problems with their devices and these users tend to be the vocal ones but at the end of the day, phones are just hardware. No, I don’t have a solution for this that doesn’t end in ‘ban stupid users’. I’m glad OnePlus/T-Mobile does come with a mechanism to unlock the bootloaders but there’s no functional reason in the world that shouldn’t be an automated system with a 12-hour turnaround.

If I had bought the phone from OnePlus directly, this wouldn’t have been a problem and I could have just click-clack-presto, bootloader unlocked. OnePlus doesn’t care if I unlock the bootloader and they are the ones making the phone. Why the hell does a carrier, which is nothing more than a glorified sales/middleman care so much? I am grateful that, at least, this is a possibility even if it is a huge PITA. There are some brands (Samsung, LG: I’m looking especially at you) that lock it down at the manufacturer level.

Phones software is fairly indestructible

With the ‘excitement’ of nuking my phone when I was practically at the finish line, I’ve realized that the term ‘bricking’ is way, way, waaaaayyy overused and in fact, phones are much more resilient than people think. Sure, you might need specialized software, skills and/or experience to recover from that scenario but counterpoint: suppose something falls on your windshield and it’s a spidery cracked mess — you wouldn’t say “this car is now a write-off because I can’t personally fix it”.

I’ve noticed that there is a trend to say ‘hard bricked’ and ‘soft bricked’ to try and differentiate between actually totally making a device permanently ruined versus “I screwed something up”.

Why didn’t I go with the screen repair route?

The cost of just the hardware (screen, AMOLED, digitizer) would depend on the just what components needed to be replaced - I knew it was more than just the glass ($30) as the AMOLED had leaked everywhere and the touch functionality was lost on 98% of the screen, pushing the minimum parts-cost to $100 and likely closer to $300. Given that we paid 350 CAD for ours new (and could get a refurbished OnePlus 7T for 330 CAD at the time, it just didn’t make money sense to go down that road.

[Back to Top]

---

Product links may be affiliate links: MinMaxGeek may earn a commission on any purchases made via said links without any additional cost to you.

]]>A New Phone Adventure