DIY a Portable, Accurate, Low Cost, Open Source Air Particle Counter

The plot shown is for a smoke test inside of our test chamber, where the concentration was allowed to decay naturally over ~2 hours. We see very good correlation against the MetOne HHPC-6 (a $5000 instrument), and also between the 3 MyPart prototypes.

Additional experiments were conducted with calibration particles of known sizes, as well as in outdoor ambient environments. For more in depth information about the tests we conducted, please seehttps://github.com/rutian/MyPart/wiki/Tests.

The overall size of the inner sensing chamber is 18mmx38mmx45mm. These dimensions include an onboard 400mAh battery. The components related to the sensing are completely separated from the outer casing, which allows various form factors to easily be explored, developed, and shared.

The MyPart sensor consumes ~2mA while sleeping, and ~70mA during sampling.

The total cost for the bill of materials is around $75. Granted, this doesn’t take into account the digital fabrication tools required to make the components (3D printer and CNC mill), or the time required to assemble a full device.

The BOM prices for electrical components are listed for quantities of 1 or 2. These prices drop dramatically when purchased in bulk. If you are making a small quantity of devices, many of the ICs (ADC/LDO/OP Amp/Humidity sensor) can be sampled from TI for free.

One major source of cost is the precertified microcontroller module (RFduino). The tradeoff is that for small volumes, the design and hand assembly of the PCB becomes much simpler.

All of the original design files/source code can be found here. We hope that the MyPart will give people the base from which to make and modify their own sensors, to set up sensing in their own communities, and to generate reliable air quality data. There are still many improvements that can be made to this sensor, but we hope that this project will act as a starting point for individuals and communities to actively engage with air quality.

How does optical particle counting work?

A laser and photodiode are arranged orthogonally such that the focal point of the laser is located directly above the photodiode. A small fan draws air through the system and across the photodiode. Particles in the air stream that intersect the path of the laser scatter light onto the photodiode; the resulting voltage signal from the photodiode is amplified by an operational amplifier circuit and sampled by a micro-controller. Peaks in the resulting waveform correspond to particles crossing the photodiode, and can be counted. The amplitude of the peaks can be used to roughly approximate the size of particles (higher peaks correspond to larger particles).

What does each part do?

• Top air channel - Contains the main flow channel, a light trap for the laser light, and the air inlet.
• Bottom air channel - Contains features to hold the fan, and the air outlet
• Analog cap - shields the sensitive analog circuitry from ambient light
• Fan - pulls air through the channel
• Laser - focused light source to illuminate particles in the airstream
• Laser holder - aligns the laser to the photodiode

Design goals

Many of the features on the mechanical flow channel and case are designed to minimize the amount of ambient light leakage and promote 'smooth' airflow through the channel. In addition, the laser should be well aligned with the photodiode, and the overall size was iterated on heavily to maximize the packing density of the components. For more information, please seehttps://github.com/rutian/MyPart/wiki/Design-Rati...

Limitations

Optical scattering: The quantity and direction of light scattered by a particle is dependent on the size, composition, and shape of the particle, as well as where it strikes the laser beam. Because of these factors, accurate sizing of particles tends to be difficult with optical scattering sensors. However, rough size cutoff bins can still be produced by using the amplitude of signal peaks.

Design

The mechanical components were designed in Autodesk Inventor (free for anyone with a .edu email), the electrical schematic/layout was designed in Eagle, and the firmware was written using the Arduino IDE.

Files
The full bill of materials can be found here. The fabrication files, as well as the original design files can be found here.

Mechanical

The mechanical components include ABS sheet stock for milling, self tapping plastic screws, and filament for 3D printing.

Electrical

The more expensive items are the RFduino, laser, fan, and battery. Most of the ICs we used can be sampled from Texas Instruments if you have a .edu email. We sampled our op-amp, GPIO expander, low noise voltage regulator, humidity sensor, and the ADC from TI.

Tools

• 3D printer (we used the Printrbot Simple Metal)
• CNC mill (we used the Othermill)
• Laser cutter
• End mills (1/8th inch and 1/16th inch. check BOM for details)
• SMD soldering supplies: iron/solder/flux/tweezers/etc..
• 5 minute epoxy (optional: syringe for precisely dispensing epoxy)
• Oscilloscope or logic analyzer
• Adjustable power supply

All of the fabrication files can be found here. This includes the post-processed gcode for the Othermill, the STLs for 3D printing, and gerbers for a barebones PCB. A list of potential design improvements can be found here.

We got our circuits using the barebones quick turn service from Bay Area Circuits. However, a solder mask is highly recommended for easier soldering. Though it is possible, I would not recommend milling the board in house due to the high number of vias that may need to be manually through hole plated. A matte black soldermask would be ideal to absorb the most light, since the bottom of the PCB acts as the top of the air flow channel. For our sensor, we laser cut the PCB outline from black printer paper and used that as the 'soldermask' for the PCB.

In addition, another smaller PCB must be made to mate with the programming cable connector. We milled these boards in house.

The CNC parts include the top and bottom air flow channels, laser holder, and an eccentric cam lock to help with gluing the laser. The top air channel is milled from the 3/8in ABS, and all other parts are milled from the 1/4in ABS. The post-processed gcode for the Othermill can be found here. For postprocessing to other CNC machines, or tweaking with the toolpaths, the CAM files are in thisdirectory.

This includes a ‘light shield’ for the analog portion of the circuit, the main body of the case, the lid, and a small button. For desktop 3D printers, we recommend printing in PLA for minimum warping/splitting.

• For the light shield, we recommend printing at 100% fill using black PLA for maximum ambient light rejection.
• There are currently four case designs: carabiner, wire loop, side mount, and bottom mount.
• Only the side and bottom mount cases have a slot for the FPC cable that can be used to control the sensor from a computer.
• Additional cases can be created by modifying the MyPart_V0_Cases.ipt file.
• The small button can be printed from a white filament, so that the indicator light underneath it can show through the case. Alternatively, a more opaque color can be used, and a small hole drilled into the button to let the light through.

The prints may need some cleanup depending on the printer you are using; more details will be provided in the assembly instructions. We tested our device using black PLA for all of the case components. Other colors for the case and lid can be used, though they might need a coating of mirror spray/spray paint on the inside to be more opaque. Other colors have yet to be tested in our outdoor evaluations.

In this step, we will use the PCB to align and glue the laser assembly. This step should be easier if the board has yet to be soldered, though we did it after soldering. At the end of this step, the laser will be glued onto the laser holder with epoxy, and will need 24 hours to fully set.

1. Chamfer the holes that the self tapping screws will go into. (2 holes on the laser, center hole on the cam lock). This helps the screw find the hole, and also prevents the self tapping screw from raising a small bump on the surface near the hole. This can be done by hand with a countersink, or a D bit.
2. Push dowel pins into the laser holder
3. Place black construction paper on the PCB, and push the laser holder (with dowel pins) into the PCB. screw in 2 6mm self tapping screws.
4. Take off heat shrink from back of laser (this will help us clear the corner of the air channel more easily
5. Unscrew the front barrel of the laser to be roughly 3.5 mm distance between the front barrel and the back.
6. Push the front face of the laser to mate with the laser holder. Assemble the cam lock onto the PCB using another 6mm self tapping screw. Use the cam lock to secure the laser onto the holder. The lock is slightly elliptical such that when it rotates, it will push the laser against the holder.

Now that the laser is in place, we need to adjust its focus such that the focal point is directly above the photodiode. This can be done by screwing/unscrewing the front barrel of the laser. Bringing the two parts of the laser body closer to each other will move the focal distance farther away, and moving them farther apart will move the focal distance closer. (The thread pitch was measured to be roughly .35mm per turn.)

If we were to make another batch of sensors, we would probably design a jig to help with the alignment. For now, follow these steps to manually adjust the laser focus:

1. Power the laser with a power supply. Lower the voltage of the power supply until the laser light is very weak, and the point at which it is focused can be easily gauged by eye. For us, this voltage was 1.8 volts.

2. Put a piece of white printer paper perpendicular to the laser beam. The spot size changes as the piece of paper is moved closer or farther away from the laser.

3. Observe the spot size on the paper when it is positioned at the same distance before and after the hole for the photodiode. We want the spot size to be symmetric at these two positions. This will mean that the focus point is centered on the hole for the photodiode.

4. Undo the Cam lock, and screw/unscrew the front lens of the laser to adjust the focal length. Lock the laser back in place again.

5. Repeat steps 3-4 until the spot size is visually symmetric before and after the photodiode hole.

6. At this point, the laser is ready to be glued. We used 5 minute epoxy to glue the laser to the ABS laser holder.

7. A syringe is prefered for precisely dispense the epoxy. With the plunger removed, the two part epoxy can be dispensed into the syringe from the back and mixed using a thin rod. Alternatively, a toothpick (or similar) can be used to brush the epoxy onto the joint between the ABS and laser.

8. Wait 24 hours for epoxy to set.

After the epoxy has set, remove the laser assembly from the PCB.

Following the Eagle layout, solder all the components onto the board.

Some things to note:

• A solder stencil is recommended, but we were able to solder each board by hand in 2-3 hours.

• The photodiode should be soldered as quickly as possible since it is very heat sensitive.

• The temperature/humidity sensor has no leads, and is difficult to solder. It was a little easier for us since there was no soldermask, and we could solder the pad of the device by heating up the trace leading up to it. Of the 3 sensors we made, all of them have fully functional temp/humidity sensors.

1. Chamfer the holes for the self tapping screws.
2. Put the fan in; it should drop in easily with very little resistance.
3. Apply some double sided sticky tape to the outside pocket of the fan channel. This will help us with our wire routing.
4. Tape the fan wires onto the bottom channel as shown. The yellow wire (a tachometer for the fan) was not used, and was cut off.
5. This configuration provides strain relief for the wires, keeps them away from the antennae, and helps with our wire routing.
6. While re-attaching the laser assembly, feed the laser’s wires through the bottom left hole of the PCB.
7. Stack the top channel, bottom channel, and PCB.
8. Put one 12mm screw through the hole next to the RFduino. We used a small plastic washer cut in half so that the screw would not directly push down on the RFduino.
9. Screw in the other 12mm screw in the analog region of the board.
10. Screw in the two 6mm screws. A washer is needed for the one of these screws to prevent shorting out the traces around the screw hole. If you have a soldermask, this will not be necessary.
11. Attach the 400mAh battery to the back of the top channel with double sided tape.
12. Place the wires from the fan and battery into the slot on the PCB.

Solder the leads for the fan and laser. For the battery, solder temporary leads from the board so that power can be clipped on and off for testing. In the current PCB, the leads for the laser and fan are a little hard to get to, and may need some fiddling with tweezers.

Using the Arduino software, follow these instructions for installing the RFduino libraries.

Download the libraries for the I2C port expander, and the temperature/humidity sensor from the Github repo. A standard FTDI cable can be used to program the RFduino. However, a .1uF capacitor needs to be added inline to the DTR line. (reply #8) The official programmer from RFduino can also be used.

Connect the cables from the FTDI to the small FPC adapter to the FPC connector on the MyPart. The firmware can now be loaded through the Arduino IDE.

We now need to find the noise floor for the analog signal. This can be found by turning on the analog components and the laser, keeping the fan off, and taping the inlet and outlet of the sensor so there is no airflow. When the fan is on, any peaks above the noise floor will be considered a particle, and binned based on the height of the peak.

Solder temporary test leads to the test pins for analog ground, output of the transimpedance stage, and the final amplified output.

Since the lid of the MyPart can’t be put on while test wires are attached, put the sensor under a cardboard or opaque plastic box to block ambient light. If you can, add a lip to the bottom of the box for the wires to pass under while still preventing light entrance.

Attach the probes of a logic analyzer/oscilliscope to the test wires, and record the output.

The output of the first stage should be around 0.2 volts. If the output is substantially higher, light may be leaking in. The output of the last amplification stage should be near ground as well, since the signal has passed through a DC block. The noise floors for two of our three devices was around 0.2V. For the third device, the noise floor was around 0.4V. One very important potential improvement is upgrading the analog front end for the sensor so it is more consistent and more noise immune. This seems like it would be a very good starting point.

This stage one of the largest pain points. Some sort of a wire bundle or test probe jig would make this a great deal simpler. Ideally, the circuitboard will be inserted into the test jig, which will make all the test connections via pins and provide a light-isolated container. In this way you will not need to assemble the entire sensor before testing, nor solder on specific test leads, and the conditions will be controlled and repeatable.

When everything seems to work, the temporary leads for power can be removed. Clip off the JST connector from the battery, and solder the battery directly to the board.

The battery can be charged through the micro USB port.

1. Using a sharp XActo knife, remove any 3D printing artifacts from the case, analog cap, button, and lid.
2. Mount the analog cap to the PCB by removing the 12mm screw on the USB port side of the board, aligning the analog cap with the board, and re-fastening the 12mm screw.
3. Check the fit of the lid with the PCB and analog cap around the entire perimeter of the lid/PCB. The outer rib of the lid should sit flat along the PCB.
4. Because some of the features are at the limit of sizes that can be accurately resolved with a desktop 3D printer, some additional parts may need to be cut off with an XActo knife for step 3.
5. Slide the air channel/PCB assembly into the case. This sensor should be able to slide in and out the case with very little resistance. Double check that the wires are not snagged on the side of the case, and that all extra bumps/strings have been removed from the case.
6. Push a small length of 1.75mm filament through the smaller hole on the button. I was able to get the filament in relatively easily by cutting the tip of the filament at an angle to help guide it through the hole. Pliers may be required.
7. Trim off all excess on the side with the button that protrudes through the case
8. Trim off excess on the other side, leave roughly 1 mm. Since the 3D printed button is directly above the LED, this protrusion off to the side allows the 3D printed button to press the actual button on the PCB.
9. [optional] Use the shims to secure the sensor to the case. Place 3 shims in the locations indicated.
10. Place the button onto the lid, and place the lid onto the case. The lid should snap nicely into the case.

At this point, the sensor should be able to give relative particle counts.

The code can be modified to map the values to the RGB LED, or broadcast over Bluetooth to a mobile phone.

To test and calibrate your sensor(s) against a reference instrument, we recommend building a test chamber to collect lots of data across a broad range of particle concentrations. The instructions for building a simple chamber from a ‘weather-tight’ storage bin can be found here.

Using this chamber, we can test with smoke, as well as calibration particles of a known size. For a more in depth description of the tests we conducted using this chamber, please see: https://github.com/rutian/MyPart/wiki/Tests.

Many potential improvements can be made to the sensor. Some highlights are below, and the full list can be found athttps://github.com/rutian/MyPart/wiki/Potential-Im...

Improvements

• More sophisticated/less noisy analog circuitry. Maybe something along the lines of: http://www.ti.com/tool/TIDA-00378#1
• More sophisticated particle counting scheme. The same link above has information about this as well.
• Reroute placement of leads for the laser and fan. They are currently in places that are difficult to solder.

• Microcontroller IC on board rather than pre-certified module.

Other thoughts

• Adaptive sampling? (sample more when particle concentration is low to get a better reading), sample less when it is extra high) - optimizing for accuracy/power consumption.
• Fully laser cut assembly for the air flow channel? Since most features are straight extrusions.