Final Product

The video above demonstrates the fully functional UV-C Sanitizing Chamber. The system is able to fully sanitize a disposable N95 face mask in just 17 minutes in order to prolong its usable life. In those 17 minutes, it applies a little over 1.0 J/cm^2 of UV-C radiant flux across the front and inside surfaces of the N95 face mask by continuously rotating and tilting the mask up and down in front of a Low pressure Mercury Vapor UV-C Lamp. I estimate the cost to build this sanitizing chamber currently to be around $60 USD.

Functions of the system demonstrated in the video include, Green LED flashing warning and not starting if the start button is pressed and the lid is not closed, pressing the start button with the lid closed starts the sanitizing process and indicated this by turning on the red LED, and stopping the system and turning off the UV-C bulb if the lid is open wile it is sanitizing. NOTE: For the second scene of the video, I am manually defeating the safety limit switch by holding it down while the lid is open, only to demonstrate the UV-C bulb and Mask rotation mechanism in motion.

To-Do List/Future Improvements:

• Create detailed fabrication/assembly documentation
• Clean-up the CAD model, Bill of Materials and Electrical Schematic.
• Create a single low-cost integrated control PCB, that would include:
  • a lower cost microcontroller or perhaps a decade counter to control the systems and timing.
  • a real time clock (RTC) for more accurate timing
  • integrated control button and LEDs (no more daughter PCB)
• Use single external DC power supply to power both the lamp and control PCB and motor.
• Use a low gear ratio N20 gear motor
• Change the position of the safety limit switch to help prevent a user from defeating it by manually holding it down.

The video above shows the partially assembled final version operating. For this test run I have the UV-C bulb power supply turned off, and after pressing the start button I am simulating the lid being closed by holding the safety limit switch with my finger.

The images above show some of the final assembly of the unit, including wire routing. The wire routing for the daughter board and limit switch in the vertical channel was a little tricky to feed through. As the UV-C bulb is hard wired to its power supply I had to cut and solder on 2.5mm barrel connectors so the power supply could be plugged into the unit and to run the power inside the unit through the relay.

I cut out and attached heavy duty aluminum foil to all interior surfaces with double sided VHB tape. These foil reflectors really help spread the unabsorbed UV-C energy all around the chamber. I measured with my UV-C meter the ambient flux within the chamber/behind the mask to be about 1mW/cm^2. I am also directing more energy directly to the mask surface with the elliptical reflector around the bulb. I am not including this ambient energy and the elliptical reflector concentrations in my calculations for how long it will take to sanitize the masks (See part 3), so this extra unaccounted for energy will serve as my factor of safety.

Programming was fairly straight forward, and yes there is likely a better way to go about this as I am not an programming expert, but this code works correctly as intended. The Arduino code I wrote is shown below. I initially tired to use interrupts and interrupt pins for the safety limit switch but I could not get it to work right so I just made a loop that checks the status of the limit switch every millisecond. You will notice the timer1 amount is not exactly 17 minutes, as the trinket has no RTC (real time clock) so I had a do a few trials to calibrate the timing to an actual 17 minutes.

int greenled = 2;
int motorpin = 0;
int uvcrelay = 4;
int button = 1;
int limsw = 3;

long timer1 = 0;

void setup() {
  // initialize the digital pin as an output.
  pinMode(greenled, OUTPUT);
  pinMode(motorpin, OUTPUT);
  pinMode(uvcrelay, OUTPUT);
  pinMode(button, INPUT);
  pinMode(limsw, INPUT);
}

void loop() {      
      if (digitalRead(button) == LOW && digitalRead(limsw) == LOW) {
            digitalWrite(greenled, LOW);
            analogWrite(motorpin, 255);
            delay(20); 
            while (digitalRead(limsw) == LOW && timer1 < 977260) {
              digitalWrite(uvcrelay, HIGH);
              analogWrite(motorpin, 100);
              delay(1); 
              timer1++;
            }
            if (timer1 = 977260){
              analogWrite(motorpin, 0);
              digitalWrite(greenled, HIGH);
              digitalWrite(uvcrelay, LOW);
              timer1 = 0;
            }
            else {
              analogWrite(motorpin, 0);
              digitalWrite(greenled, LOW);
              digitalWrite(uvcrelay, LOW);
              timer1 = 0;
            }
           }
        if (digitalRead(button) == LOW && digitalRead(limsw) == HIGH) {
          digitalWrite(greenled, HIGH);
          delay(50);
          digitalWrite(greenled, LOW);
          delay(50);
          digitalWrite(greenled, HIGH);
          delay(50);
          digitalWrite(greenled, LOW);
      }
      if (digitalRead(limsw) == HIGH && timer1 == 0) {
         digitalWrite(greenled, LOW);
      }
}

the images above show the final CAD model of the unit. Over all dimensions are 198mm [7.8″] wide, 275mm [10.8″] deep and 287mm [11.3″] high. The walls of the chamber are made of a single piece of 0.5mm PS sheet that I had leftover from another project. The sheet is held in place by being fit in the slots in the base (light Teal) and top frame (Dark Teal). All internal surfaces will be lined with aluminum foil to help reflect the light all around. The design uses only a few screws with the rest of the parts nesting or being press fit into each other. You can see the user control panel with its green & red LEDs and black button, at the top of the unit near the opening lid (light red). The height and diameter of the chamber were sized to accommodate the four N95 masks that I modeled, while still getting the masks as close to the UV-C bulb as possible. It should also be possible to fit many other disposable N95 masks types in here as well.

The above image shows the cross section view of the unit. You can see how the UV-C lamp and the vertical channel (light yellow) are captured between the top frame and base. The vertical channel has a hollow opening inside of it to run the wires for the daughter control board and the safety limit switch down to the main control breadboard (green) in the base. The light purple part is the big internal tooth gear that is pressed onto the U-shaped rotating part shaft (light teal, center). I added flats to any shafts that have gears pressed on them so there is no chance of anything slipping. The gray TT motor in the bottom right has a small pinion gear on its top shaft which mates to in the internal gear teeth. Again all joints and rotating parts use small 1060 bearings.

In the above image I made a number of the parts transparent for clarity. You can see the elliptical reflector shape around the UV-C lamp which helps to try and reflect some of the extra light on the bulb backside around directly to the mask. On the bottom back side of the unit, you can see the two DC barrel jack plugs and the trinket usb micro connector port. As the unit will be fairly light, the large Lid has hard stops (as shown) to prevent shifting the center of gravity and tipping the unit over, while still providing access to the mask.

The above video shows the next version of the mask rotation and tilting mechanism fully operational. I moved the linkages to the outside of the U-shaped rotation mechanism. The U-shaped rotation part is connected to a big internal tooth gear inside the base (basically the diameter of the base). the TT motor drives this big gear with smallest pinion gear possible so to show the entire mechanism down further. In the video I am running the motor at 3.3V, but I can slow it down a bit by running at 2V and it still operates.

With each full rotation the single tooth gear fixed to the base moves its mating fully toothed gear by 3 teeth, which through the reciprocating mechanism, tilts mask tilts up or down about ~10-30 degrees depending on the mask position. The fully toothed gear has 20 positions, but as we are moving an odd amount of teeth each time, we skip around in tilt angle positions, but we do reach all of the angle positions, thought not in order, which probably is good to jump around.

To hold the masks in place to the tilt mechanism, I did not want to have to physically attach to the masks themselves as that would block some of the filter area, even if only a little bit. So instead we attach a small amount of the straps to the tilt mechanism. The above image shows two part magnetic clip that clamps the mask straps in place. Both parts are toothed so the straps can be gripped, and each half has embedded magnets with enough force to hold everything in place but easy enough to pull apart. You just have to take care to make sure the straps are at least semi-taught so the mask correctly moves with the change in tilt angle.

All along the outside edge of the black base, between the large round teal part, there is actually a small gap so a flat plastic sheet material (~0.5mm) can be inserted to form the walls of the chamber. This will take much less time to produce and be will lower cost compared with printing all of the chamber walls. I also can sandwich in aluminum foil so the UV-C rays can reflect all over the chamber too. Image above shows me quickly mocking the up to see and i looks like it will work out quite well.

For the next iteration of the design, I will need to add where the aquarium UV-C bulb will mount. This will be just outside the the turn radius of the mask rotation mechanism so we get it as close to the mask surface as possible. Since the bulb is basically just a quartz tube, it will just sit in a circular pocket in the top of the base. But I will also make the shape of the base around the bulb, and thus the chamber plastic sheet walls/aluminum reflector, to form an elliptical reflector so I can try and focus more of the bulb energy directly to the mask. See the diagram above for the difference between ray paths for elliptical and parabolic reflectors.

The above image is the initial prototype layout of the electronics for the sanitizing chamber. There are probably different and better ways to implement this, but since I am not an electronics expert and for this first version this is plenty sufficient to prove out the concept. I am using all components that I had on hand already for ease of prototyping (except of course the Aquarium UV-C Bulb).

I am using a 3.3v Adafruit Trinket as the micro controller and using all 5 of its pins. A 3.3v wall power supply will drive the microcontroller, motor, relay and leds, etc. This is actually a little lower than what is recommended (4.5v) for the 3.3v trinket version but seems to work fine. The low pressure mercury vapor UV-C Aquarium bulb has its own little power supply rated at 24-36V, which I am assuming a range is listed because it’ll draw differently at startup than when already ignited. Ideally of course one would certainly only want a single power supply for a product, but for ease of integration and quick implementation for now, this the bulb power will be a separate subsystem turned on/off by the relay. So the bottom of the cleaning chamber will have 3 plug ports: 3.3v, Aquarium bulb power supply and the trinket usb micro so I can tweak the code.

For safety, I do not want a user to be able to see the UV-C light while it is running if they were to open the lid, so I am adding a little limit switch so the system knows that the lid is closed (laid out here on the green breadboard). If the lid is opened the system will immediately stop everything and turn the UV-C bulb off via the relay.

The motor to drive the mask rotation system is a common hobby “TT Motor”. I wanted a gear motor to have enough torque and also to make all the mask rotation be as slow as possible. I would prefer to use something like an N20 gear motor as they come in a range of gear ratios but they are actually kind of expensive in small quantities. So, This 1:120 “TT” gear motor is readily available and can run at 3V. I am using a transistor with it so I can pwm’d down as slow as possible while still being able to drive everything.

The above image shows the little daughter board for the main user control interface which will be located near the top of the machine, separate from the main board. The plan is when the top door is closed (checked via limit switch) and the user presses the button, the red light is on during the sanitizing cycle. One the the cycle is complete, the green light stays on until the door is opened. If the door is open and the user presses the start button, the green light will flash (signifying an error). Since there are only 5 pins on a trinket, I have run out, so the Red LED is actually inline with the motor pin, which works because the motor is only turning when it is sanitizing. The solder joints/traces on the back side are pretty crude, but they get the job done.

The above image shows the much cleaned up and refined electronics. I was able to squeeze all of the wires and components on to the single white breadboard. This breadboard will be actually be somewhere in the bottom section of the cleaning chamber (not in the actual UV-C illuminated part). The green and red breadboards are just place holders to represent that there will be a cable harness connecting to the daughter control board and a cable harness connecting to the safety limit switch for the lid.

I am using a 3.3v Adafruit Trinket for the microcontroller as I had it on hand already. I really like the Trinket platform as they are just so very small, easy to use, Arduino compatible and I have used them on a ton on other product prototyping projects before with great success/integration. I actually really wanted to try out the newer and more powerful M0 Trinket version but unfortunately they have been sold out for quite some time. I also leaned of the Seeeduino XIAO recently, which looks very interesting too, as it is very small as well, but offers even more I/O pins and a low cost.

Compared with a static cleaning chamber where both the mask and lamp are fixed, I want the mask to rotate in my cleaning chamber as this will ensure there are no shadows and UV-C light can get into every nook and cranny on the masks surface. As an added measure I would also want the mask to change its angle up/down each time it makes a full revolution. This way we can ensure full coverage, reaching every exterior and interior part of the mask surface.

One of my favorite places to look when I need kinematic mechanism inspiration is Nguyen Duc Thang’s youtube channel; thang010146. His channel has been around a long time and is constantly updated with all kinds of kinematic mechanisms that he himself designs and documents. It is really easy to get sucked into watching one video after another in his channel! I really appreciate his work. Looking around, one mechanism really stood out to me that uses a single motor to both rotate and angle an object up and down; check out his “Drawing a spherical helix” video above.

I would need to modify this mechanism because if left as-is, the mask would always follow the same exact helical path (orange line in the video). What I want to do is have a single motor rotate the mask, and then also once per revolution, increment the mask angle down or up in a reciprocating range. That way every up or down angle increment of the mask gets one full rotation in the chamber before changing. This way we are always exposing every little nook and cranny on the mask. So from the video the idea would be to get rid of the big gray gear, and then make the fixed purple gear only have a few teeth so that the yellow gear only turns a little bit with each full rotation of the system. This yellow gear in turn would need to change the angle of the mask.

So with this idea, the yellow gear would need to be connected to a reciprocating kinematic linkage. To help play around with different kinematic linkages and sizing I came across a really simple to use program called GIM. GIM allows you to quickly layout and animate different linkage shapes and journal bearing points and then map out the movement of the entire mechanism and target end effector linkage. (Yes this can be done with Soildworks, but this is a faster and simple implementation). The animated gif above shows a side view of the simple linkage I came up with to have small a continuous rotation at the bearing on the bottom cause reciprocating motion of around +50 and -50 degrees which the mask would be attached (large T-shaped bar at the top). That small bearing at the bottom would be the incremental continuous rotation of basically the yellow gear in the youtube video mechanism.

The above image shows the first prototype mechanism. The large circular part at the bottom is fixed and if you look at the center it has a right angle gear with only 1 tooth. So as the big U-shaped part rotates around the fixed circular base, that other fully toothed right angle gear is incremented a little bit. That fully toothed right angle gear is that connected by a small shaft to two linkages (one of each end of the U-shaped part) that I developed in the GIM simulation above. So the angled end parts of that mechanism will connect to the masks somehow (I need to continue to work on this). I tired to put the linkages on the inside of the U-shaped part to try and minimize the width of the rotating section so we can be slightly closer the UV-C source, but I don’t think this will work and it will interfere with however we hold the mask, so I will move it to the outside. Also the little single toothed right angled gears in the bottom right of the image were for getting the fit just right so the linkage mechanism would not rotate on its own when not contacting the single tooth gear. For all the linkages and rotating sections, I used these very compact little 1060 bearings that I have had lying around for a long time.

Now that we have 2D contour plots of the energy ranges emitted by the bulbs at different distances/locations, we need to turn those plots into 3D models. We also need to 3D model the actual masks that will be sanitized so we can intersect the UV-C bulb energy model with the mask model to determine how much mask surface area each energy range covers at a given position (yes!). From various projects over the years, I ended up with an assortment of a couple of leftover N95 and N100 disposable masks. I 3D modeled the 4 mask types above using Solidworks.

Next after cleaning up a bit first in Illustrator, I imported the SVG vector lines from the 2D contour plots into Solidworks and extruded them. I made sure to curve them from the center to the ends of the bulb in order to account for the power drop off I also measured. I then colored each contour section to match the 2D plot colors to easy lookup as shown in the images above. With my 4 different size/shape mask models I was able to determine the closest distance possible to the bulb while being able to rotate the mask and keep from hitting the bulb, which turned out to be about 85mm (more on that in a subsequent post, but I intend to have the mask rotating and moving up/down in the chamber for maximum UV-C coverage).

Now that we have the distance between the mask and the bulb, we need to determine how much energy reaches the mask surface when the mask is at different angles relative to the bulb (simulating movement in the chamber). I measured the energy received by each of the 4 different masks in 3 configurations: Bulb perpendicular to the front of the mask, ~70 degrees to the side, and 45 degrees to the side while also being 30 degrees raised. As you can see in the above images, I am showing the intersections between the 3D mask models and the bulb output range 3D model. I then used the measuring tool to measure each of the cut surfaces in the bulb model to determine how much area in that particular energy band range gets to the surface.

I then take all that surface area and energy data, for all the different bulb position conditions and mask types, and determine a nice weighted average energy output in the average weighted coverage area. The above image is an excerpt of these calculations. The Toothbrush bulb outputs an average of 0.85 mW/cm^2 in an average mask coverage area of 43.3 cm^2. The Aquarium bulb outputs 2.50 mW/cm^2 in an average mask coverage area of 211.4 cm^2.

Now since I modeled all the masks, I know the total area of the front and inside surfaces that need to be sanitized. With the initial requirement of 1 J/cm^2 to sanitize a mask, and with all this data I can calculate the time it will take to get 1 J/cm^2 to all surfaces of the mask. For the small toothbrush bulb this comes to 2 hours and since there are two sides (due to the rotation I plan on having), we double this number for a total of 4 hours required to clean an N95 mask with this bulb. For the large Aquarium bulb it comes to 8.5 minutes per side, so doubling that is only 17 minutes to sanitize an N95 mask! Given that the aquarium bulb is much more powerful and covers much more area, I will proceed with using this for my automated UV-C mask cleaning chamber.

A way to measure the UV-C energy was needed calibrated to the correct wavelengths. I have used Ophir Starlite digital power meters in industry in the past with Lasers and LEDs for wavelengths in and around this range, but they are prohibitively expensive. Luckily I came across the BLAK-RAY J-225 Short Wave UV meter which specifically meant for measuring UV-C bulb power output (pic above). Its measurement sensitivity range is from 220nm-290nm with peak sensitivity at 254nm, which is just what I need. You can find lots of these units used on eBay. I was able to find one in great shape with a relatively recent calibration for only $40!

It is a neat little analog device, and seems to be self-powered. The measurement probe is on top of the unit and has a rectangular area that collects the UV-C light. This probe is actually removable and can be connected with simple banana jack cables for remote use in smaller areas. Output is measured in micro Watts per square centimeter. There is a switch on top of the unit to switch between the A (top) and B (bottom) measurement ranges on the analog meter.

To make a contour plot of the energy emitted from the bulbs, I made a semi-circle pattern and printed it out with 22.5 degree radials with 10mm spacing along each up to 100mm. I placed this pattern on my little flat granite block measuring surface plate and put the bulb at the center as shown in the images above. I mounted the UV-C probe onto a small linear translation slide so I could adjust the Z height if needed. I then moved the probe to the 90 points and manually recorded them in excel.

The 90 data points were laid out in a table at their respective X and Y coordinates. Unfortunately when making contour plots in Excel, it cannot automatically fill-in/interpolate data between X/Y points, so the plot does not really work. So I turned to the really awesome python program, Plotly. With a lot of massaging of the data first in Excel and Notepad++, I was able to modify one of their online contour plot demos with my data to produce the really nice plots below. Also Plotly lets you save plots as vector SVG file format, which helps a lot to make 3D CAD model of the out energy; more on that further down.

Above is the contour plot of the Toothbrush bulb with reflector. The X and Y axes are position distance in mm, intensity scale on right side is μW/cm^2. I cleaned up the reflector path since light will be blocked directly to the right and left of the bulb. There is approximately a 12.5% drop off in power between the mid plane Z-height of the bulb and the ends.

Above is the contour plot of the Aquarium Lamp. This bulb has significantly higher output than the toothbrush bulb. The X and Y axes are position distance in mm, intensity scale on right side is μW/cm^2. The Lamp is a cylinder so the output is even radially (I did not clean this up here in the 2D plot, but I did in the 3D CAD). There is approximately a 33.3% drop off in power between the mid plane Z-height of the bulb and the ends.

Given the current shortage, extending the usable life of existing disposable N95 respirators is one of the very highest priorities. One of the most effective and rapidly deployable methods is to expose the masks to a measured amount of UV-C radiation.

A number of papers were published after the first SARS outbreak many years ago on how to decontaminate N95 masks in order to prolong their usable life during a future pandemic outbreak and supply shortage. Below is a listing of relevant papers on this decontamination concept as well as a summary New York Times article. Somewhat ironically, a number of these older papers cite that we need to take steps to increase our emergency stockpile of N95 masks for a future outbreak.

• [2011] A pandemic influenza preparedness study: Use of energetic methods to decontaminate filtering facepiece respirators contaminated with H1N1 aerosols and droplets
• [2011] Effectiveness of Three Decontamination Treatments against Influenza Virus Applied to Filtering Facepiece Respirators
• [2015] Effects of Ultraviolet Germicidal Irradiation (UVGI) on N95 Respirator Filtration Performance and Structural Integrity
• [2018] Ultraviolet germicidal irradiation of influenza-contaminated N95 filtering facepiece respirators
• [2020] N95 Filtering Facepiece Respirator Ultraviolet Germicidal Irradiation (UVGI) Process for Decontamination and Reuse
• [2020] Technical Report for UV-C-Based N95 Reuse Risk Management

You can download all of these papers here: UV-C Papers

A summary New York Times article can be found here

A number of the papers detail experimentally measured required UV-C wavelength energy and soak times to fully decontaminate the masks. Typically the consensus it that masks need to be exposed to at least 1.0 J/cm^2 centered around 254nm wavelength in order to be fully decontaminated. Another paper discusses the structural integrity of N95 respirators after UV-C radiation and finds that they are still effective at filtering particles, though depending on manufacturer, may have a limit to how many times they can be cleaned before the materials begin to breakdown. The last two papers are were written recently as a response to the pandemic to summarize and steps on decontamination methods.

My goal is to create a rapidly depolyable, low-cost, and 3D printable UV-C N95 Mask decontamination chamber with a measured output to conform to the 1.0 J/cm^2 UV-C requirement.

UV Wavelengths

Germicidal Wavelengths fall generally in the 200nm-300nm range. The vast majority of Germicidal Sanitizing lamps emit peak energy around the center of that range at 254nm. Note: Your typical UV curing oven for photocurable resin 3D Printers (SLA, etc), and nail salon curing lamps, emit only within the UV-A and UV-B ranges and thus they will not disinfect anything.

UV-C light Safety Warning – UV-C light is very harmful to your vision, even for short periods of time. Always wear safety glasses that filter out these wavelengths. I have a pair of UVEX-S1933X which are inexpensive and readily available and have measured data on the specific wavelengths they filter out, (enlarge the above picture). Also it can be harmful to your skin like other UV wavelengths, so it is a good idea to wear long sleeves, and gloves when handling active bulbs for long periods of time.

Ozone Safety Warning – Ozone is very bad to breathe and will harm your lungs. It has a very distinct smell. Ozone is produced around 185nm wavelength. Cheap Low-Pressure Mercury Vapor UV-C Lamps emit a range of wavelengths centered around 254nm including wavelengths around 185nm. Higher quality mercury vapor lamps have AR coatings to filter out these ozone producing wavelengths. Here is a great summary Article from the EPA on the dangers of breathing in Ozone. If you are using one of these cheap bulbs that produces Ozone, and you will definitely be able to smell its unique smell, be sure to test it outside or with a externally venting fume extractor.

There are generally two common bulb types that can emit light in the UV-C range: Low pressure Mercury-Vapor Lamps and LED. Mercury Vapor lamps are the most common and have been around a long time. They are available in a wide range of shapes and sizes and typically use higher voltages to excite the gas in the quartz tubes (sort of like florescent bulbs). These lamps emit a wide range of wavelengths but the peak energy emitted is centered around 254nm. LEDs that emit in this wavelength are a relatively newer technology and are thus much much more expensive. These LEDs produce a very very narrow wavelength range and are “binned” (sorted) for their peak wavelength output. They are also generally have much lower power than the gas lamps, so many will be needed to achieve the same effect. UV-C range LEDs increase in ubiquity, volume and thus price every year. So eventually they will likely overtake the mercury vapor lamps.

Low-Cost and Readily available Bulbs

I will focus on low pressure mercury vapor lamps thier cost/ubiquity and also lamps that come with their own power supply (for ease of integration).

The above Toothbrush Sanitizer is available across the internet in various versions for around $20. I chose it for the bulb size ~100mm long. It also came with a 9v power supply and I could possibly reuse the internal PCB and modify its decade counter timer. The bulb is integrated into a nice reflector for focusing the energy. One drawback is that this cheap bulb emits lots of Ozone.

I later found that there are much larger Aquarium Sanitizer UV-C lamps available in a range of sizes. I bought the highest power version I could find on eBay pictured above (13W) for around $15. It comes with its own ~36v DC power supply. I do not smell any obvious ozone, so I think it may have an AR coating.

Update 04/01/2020: I have added a top shield piece, see details at the end of the post.

Minimum Viable Product Face Shield v1.0

Minimum Viable Product Face Shield: The goal is to fabricate and get as many Face Shields to hospitals as possible in the least amount of time while still achieving the intended function (Help block bulk aerosol particles from reaching the face). This is achieved with two-fold; the face shield design was paired down and optimized as much as physically possible in order to minimize print time on FDM/PJP/FFF type printers, and the part was printed “stacked” to maximize print density in a 24 hour period. Readily available transparency/report cover film with a standard 3-hole punch pattern and rubber bands comprise the other components of the mask. We are able to output 120 functional face shields, across 2 printers, in a little under 24 hours.

Top Down View of Frame features

Design: The print is optimized for minimizing print time, which generally means less print material is used also. The entire part is two path widths wide, around 1.20mm, except for the 3 pegs and the hook features for the rubber band which are solid. It is 15 layers, at 0.33mm layer height, approximately 5mm high. The pegs are radially spaced at the standard US 4.25 inch 3-hole punch distance. The pegs on the end are spaced slightly further out so as to keep the transparency/report cover face shield taught. There are two built-in spring features at the side pegs, to also help with tension but to take up any extra slack from the ~6-8mm diameter hole range on different US 3-hole punches. Hook features at the back side hold the standard 7″ x 1/8″ rubber bands. You will possibly need to tweak the 1.20mm width for your particular printer to ensure that the two paths that form the structure of the part do touch fully for maximum strength.

4 X 9 Print Stacking Method

Print Stacking Method: Only 4 frames can be “traditionally” laid out on the print platform by nesting them in X & Y. In order to produce the most amount of parts in the least amount of time, it is not ideal to have to sit by the printer, remove the 4 parts and start another build, as all that time really adds up and is prone to additional errors/issues. The most ideal scenario would be to have full build take a little under 24hrs, and make as many frames as possible in that build, so their is only 1 build per day, significantly reducing build turnaround time and labor.

To achieve this we must stack the frames vertically one on top of another in the Z direction. I modified the physical model to have a single layer of support material (HIPS) in between each (ABS) frame so they separate cleanly. While stacking the parts without any support material in between may be possible, you are much more likely to have these thing parts break when trying to peel them apart (more on that in the next section). You can see the white HIPS support material in between the black ABS frames. With this method, we can print as high as our Z-axis. I am printing 15 frames high, so 4 stacks of 15 yield 60 frames per printer in a little under a 24hr period.

Close-up of Print Stacking

Easy Peeling off Single Support Layers

Dual or Single Extruder Print Stacking Method: While the particular printer I am using (dimension 1200es) has a dual extruder, it is still possible to print stacked with a single extruder with PLA. Rather than the single support material layer being a different material, that layer could be selected as a being compromised as a perforated layer. This layer would be similar of the print process at the top of automatic single material supports with the “support material” option selected in the desktop printer slicer. Since PLA has minimal warp, this perforated layer should allow the stack to print correctly, yet still be separated when the build is complete.

100 Face Shield Frames in ~19 hours

Production Timing: The printer I am using is not particularly super fast, and it takes roughly about ~22 minutes to print one frame. It is very likely that a desktop printer could be tweaked to print much faster than this. I am trying to time it for one ~23hr build per day to minimize turn over time. I can achieve this by printing 15 frames high, so my 4 stacks of 15 yield 60 frames per printer. I am running two of the same printer, so I can output 120 complete Face Shields per day with this frame design and print stacking method.

Readily available and Low cost materials

Materials: The goal was minimize any custom parts and use as much readily available parts and materials as possible. Make sure you have the clear face shield aligned straight when you attach it to the frame.

• Clear Face Shield Material:
  • Best Option (Clarity and Price), Presentation Covers. [note that it is a little oversized than 8.5″x11″ so you may want to tweak the stop position on you 3-hole punch so it is more even: https://www.amazon.com/gp/product/B0015ZXIL2
  • Marginally Acceptable Transparency Film (maybe only as a last resort), this brand only, as it is clear, but there is a lensed pattern in the film which makes it harder to see through straight: https://www.amazon.com/gp/product/B001GXD2A0
  • Do not buy this, cheap Transparency Film, has a similar pattern to the other, but is really bad you cant read text like a few inches from your face: https://www.amazon.com/gp/product/B003V1BPXG
• 7″ x 1/8″ Rubber bands:
  • https://www.amazon.com/gp/product/B001GKO2L6/
• 3-Hole Punch:
  • https://www.amazon.com/gp/product/B01GIJLSGG/

Total cost of materials for each complete assembled Face Shield is around $0.70.

Update 04/01/2020: Based on feedback I have added a top shield piece to protect from falling particles. See image below. Still keep in mind that Face Shields do not take the place of Goggles or safety glasses, they just help block bulk particles. Rather than add a lot of extra print time and make the design a lot less conformable by adding to the frame to cover the top, I opted to add a second simple piece of cut and bent film. This way one can still produce 60+ completely assembled face shields with one printer per day.

Minimum Viable Product Face Shield v1.1

I have added to the download zip folder a printable 8.5″x11″ PDF template to Cut & Bend the top shield pieces. 3 top covers can be made from a single report cover film sheet as laid out in the PDF. Make sure you print with the “actual size” option selected. These are sized for the report covers (which are slightly over sized), so you will notice the ends are cut off, but you can still easily trace all needed dimensions. Cut on solid lines and bend on dashed lines.

Download All Source Files here: MVP Face Shield v1.1. Contains the Frame and Stacked Assemblies in Source Solidworks 2015, Parasolid, STEP and STL formats.