Nancy Mazza

Outside Sales – Newark Electronics

Richard Cuplin

Senior Engineer at Casco Products

Russ Schinzing

President at ERC, LLC

I see a lot of SMPS (switch mode power supply) designs in my consulting.  Some are executed well, many however are not.  A poorly designed supply can cause many problems, at forefront of the list is radiated emissions.  I’d like to share the top three things I look for in assessing an SMPS layout:

1. I like to see a continuous ground plane under the circuit with every ground point tied to it right at their respective ground pins.

Why? A continuous ground plane provides a low inductance path for return currents.  Return currents want to return to their source by the lowest inductance path (which is directly under the signal trace) unless they are forced to take a longer path by poor layout practices (no ground plane, cuts in the ground plane, etc).  The higher the inductance of a current loop with high di/dt, the greater EM fields are generated.

2. Physically small switch node.

Why? The switch node is the business end of the switching supply.  It has the fastest changing voltages and currents of the entire circuit.  Making it small by arranging inductor, rectifier, and switch appropriately gives it smaller parasitic inductance, and less radiation efficiency.

3. Low mounted inductance caps placed close to the input.
Why?  You want all of the high di/dt current to be sourced from your input caps, not from the upstream power source.  When current is sourced from the physically close input caps, the high di/dt loop is also small.  Often a series inductor or ferrite bead is placed between the input caps and the upstream power to make sure the high spectral content power is supplied by the input caps.

There are certainly other things you need to pay attention to when designing a switching supply like bootstrap circuits, component sizing, and feedback loop compensation, but following the above guidelines will definitely improve your success rate!

DEB Associates now has a seat of Altium Designer for circuit design and PCB layout.  As a consultant I sometimes need to use a client’s licenses for PCB design.  I have executed a dozen or more commercial designs in KiCAD (an awesome open source tool developed by Jean-Pierre Charras), including a motor controller for automotive, a 10 layer impedance controlled interface board for aerospace, and several others.   I have also had opportunity also execute designs in Mentor Graphics PADS / DxDesigner Toolchain, as well as PCAD, which is the precursor to Altium Designer.  I decided to pick up Altium Designer due to it’s popular use in industry and its integration of all the design functions into a single unified tool.  PADS (layout) and DxDesigner (schematic capture) are two separate programs that interface with each other.  Though powerful, the Mentor tool can be a little clumsy to work with.  KiCAD also consists of a handful of programs to make up its tool chain, but they are well integrated and have a consistent look and feel.  Though I love KiCAD, most of my clients don’t use it and have already invested in a major commercial tool like Altium or PADS.  Another advantage to using Altium is the ability to import designs from other major tool vendors.  Something I can’t do with KiCAD.

Charging my Droid from a 12V supply.

I had some extra board space on another project so I threw down a design for a DC-DC buck USB charger for my Samsung Droid Charge featuring an LTM8025V uModule.

3A USB Charger using the LTM8025

The LT8025V is a self-contained DC-DC converter that needs only a couple of passive external components for such things as setting switching frequency and output voltage.

I was curious how hard it would be to reflow solder the 72 LGA package onto a PCB.  So using our favorite $20 hotplate we solder pasted and reflowed two units at our Grand Rapids area electronics engineering lab.  The main concern in using a part like this for us is about solder bridges and locating the part properly, since we do a lot of hand build prototypes.  Surprisingly the part self-locates very well.  When the board was above the reflow temperature, I could tap (lightly) on the top of the 8025 and actually feel the solder surface tension pulling the part to its pads.  The part moves slightly with the tap of SMD tweezers and then snaps back to its pads.  The first board fired up perfectly.  The second one needed to be heated a second time to make it functional.

Bare PCB showing the 72 LGA footprint on the left side

We’re always interested in trying new parts that may show up in one of our electronics engineering projects for a client in the future.  It’s nice to have fully vetted a solution prior to implementing for a customer.  So based on this Droid charger implementation with the LTM8025V, Linear Technology gets another thumbs up from the DEB Associates Electronics Engineering Lab here on the Lakeshore of Michigan!

“DEB Associates was able to assist us on short notice with our needs with prompt professional service.”

We have a seat of Alibre Design Expert 3D CAD modelling software and can read files made with any major CAD system.  Need help with a your mechanical product design?  Need PE certification?  Contact us.

Below is a picture of the power distribution assembly.  On the left is a Blue Sea 150A Dual Power Bus.  GX11 on the right with a Blue Sea Systems fuse block.  Driver circuitry built on 3″X2″ perf board.  This all gets incorporated in a 19″ 2U rack mount case along with 15 custom circuit boards.

Coil Economizers

Prior to selecting the GX11, I was looking at another product that uses a technology called “coil economizer”.  It works by PWMing the input voltage after pull-in to reduce power consumption.  I’m sure this is a fine product in the right application, however, since my application involves data acquisition, I wanted a dead quiet (electromagnetically speaking) contactor solution.  The coil economizer’s switching electronics would have injected way too much electrical noise into the box.

Regulating Voltage to the GX11

Since coil current is proportional to input voltage in any contactor, the coil driver circuit must regulate it within a fairly narrow band – too low and the contacts don’t pull in, too high and the coil overheats.  My application must work over a fairly wide input voltage range of 9-32V so regulation was a must.  I chose to base my driver on a Linear Technology’s LT3080 linear regulator set at 12V output.  Since it can handle around 1.5A for short periods, it supports the pull-in/inrush requirement of the GX11.  It was also the lowest (1.3V) dropout voltage unit I could find which also supports my voltage and current requirements.  Most others in that range are upwards of 2V.  I’ve found that Linear Technology is usually not the cheapest, but almost always has the best performing  parts available.  I wanted to be able to actuate the GX11 all the way up to 32V, but found the internal current limiting of the LT3080 limited inrush to <0.5A – too low to pull in the contacts.  On the other end, the GX11 guarantees pull-in at >9V.  This limits low voltage actuation to >10.3V (9V+1.3Vdo).  Once actuated, holding the contacts in is no problem over the entire 9-32V range.

Beware Flyback Voltage

For this driver circuit a fly back diode across the coil is required to keep the voltage across the LT3080 within it’s absolute maximum Vin – Vout rating which is 40V.  The problem comes during switch off.  The GX11′s coil flyback voltage drives the LT3080 output negative by 20V.

32V – (-20V) = 52V > 40V = goodbye LT3080

The GX11 does have internal flyback suppression, but not enough for this application.  A diode in my driver circuit clamps the flyback at about -1V.

Finishing Touch

Lastly, a small transistor circuit drives the indicator LED on the main power switch.  I could have driven it from the coil voltage, but I wanted to give positive assurance of contact pull-in.  The solution was to connect the base of a small NPN transistor to the GX11 output, which doesn’t have any voltage on it until the contacts pull in.  The NPN then drives the power switch LED from the regulated supply.

Need custom test equipment?  Contact us.

Tortoise SVN

I’ve been wanting to get Subversion (open source version control) running on my development machine to make it easier to make code changes and be able to revert to an earlier version if there’s a problem.  I found a Windows version of Subversion called TortoiseSVN that integrates right into Windows explorer.  Here’s how to get it up and managing your code in five minutes:

1. install TortoiseSVN, blah, blah, restart.
2. make a folder called “Repository”,
3. right click on it and select:TortoiseSVN:Create repository here…
4. right click on the folder with your current code in it and select:import
5. in the dialog box that appears, select the folder called Repository, click OK
6. your code is now stored in the Subversion repository “Repository”
7. create a folder called “Working Software”
8. right click on it and select:checkout
9. you can now work on the code in the “Working Software” folder
10. when you’re done right click on “Working Software” folder and select:commit
11. You’ve just created your first revision!

Drain Source Voltage

Drain to source voltage spikes in hard-switched power converter designs can cause anything from immediate failure on the power MOSFET or (more insidiously) a delayed failure due to repetitive micro-damage.

What Causes D-S Spikes?

Any time a MOSFET is hard-switched (switching time measured in 10s-100s of nSec) current is changed abruptly.  If you are simulating your design, just add a couple of nH in series with the drain of the control side (top) FET and measure drain to source voltage.  What was a clean square wave is now a ringing mess that probably exceeds the Vds capability of the part.  So where is this imaginary inductor in your design?  It’s in the PCB layout!  Just trace the return path from the top drain through the drain cap, and back to the lower FET source.  How long is it?  If you have any significant trace length in the ground return of the top FET (think half-bridge / synchronous buck) you’ve added parasitic trace inductance.

Better Layout = Less Ringing / Spiking

There are two things I make sure of when I layout a half-bridge: 1) the top FET drain and the bottom FET source need to be adjacent to each other and 2) Put low ESL capacitance across those two points.  The layout criteria (#1) minimizes the return loop and thus the parasitic inductance.  The capacitance isolates the top FET drain from parasitic inductance in the input circuit.  These two points are not hard to incorporate and make a huge difference in the Vds safety margin.