purple_air

Code for displaying Air Quality from a Purple Air device or Web endpoint.

Read and display Air Quality from a local Purple Air device--or from a Purple Air device that is on the web--on a M5Stack M5StickC "finger" computer.

The local version Uses the local JSON endpoint's built-in 120 second average, which is updated every 10 seconds.

The web version uses the default values from api.purpleair.com, which does not indicate what this represents (instantaneous? some average?). This value is read every 50 seconds.

See my blog post for more information on the Purple Air unit, my design decisions, and my unique writing style. :)

Setup

Set up the M5StickC for UIFLow using the M5Burner utility. Once this is done, the software needs to be loaded onto the M5StickC device. There's a few different ways to do this. Alternatives include rshell, or you can setup Visual Studio Code and install the m5stack extension.

If you have a 1.5.x or greater version of UiFlow, I've set up a relatively simple set of steps to use the UiFlow GUI to make the copy.

No matter how you get the code onto the M5StickC, you will need to provide either the IP address of your local unit or the "sensor index" of the station that you want to monitor and an API key. To get the web "sensor index", click on the station you want to track, and the URL will change to something like this:

https://www.purpleair.com/map?opt=1/mAQI/a10/cC0&select=38889#15/37.25548/-121.8908

This shows you the sensor index of 38889 (and the station's Latitude and Longitude).

While you only need an IP address to read the data locally, you need the sensor index and an API Key to read the data from the web. You will need to email [email protected] to get yours. They will send you back two keys: A "Read Key" and a "Write Key". As we will be reading data, that's the one you want.

If you don't use the copy_for_web.py or copy_for_local.py scripts, you will need to modify aqi.json for local and aqi_web.json for the web: They are pretty self-explanatory: Fill in sensor_location with your IP address for local, or your sensor index and read_api_key for web.

Run

Restart the M5StickC by holding down the button on the left (the power button) while holding down the front button (the "A" button): It will show "Program".

Use the button on the right (the "B" button) to navigate to APPS.

Press the A button to enter APPS. Use the B button to scroll down to either LocalAQI or WebAQI depending on which one you have chosen.

Press A to start the app. Now set the M5StickC on its side--doesn't matter which one--and after a bit it should show WiFi connected,

Then start displaying the air quality.

.

Brightness and Correction Settings

• Brightness can be adjusted, and is remembered if the unit is restarted: Hit the B button, use the B button to adjust, and then A to save and return.
• There are several different correction factors: Hit the A button, use the B button again to cycle through the corrections, then use A to return.The numbers and colors change as you cycle through so that you can see what they would be.

Hardware Abstractions

• All of the hardware-specific code is abstracted to m5stick.py, so it is possible to port to another platform. (Well, at least in theory: The code is designed around the M5StickC's capabilities).
• The sensor interface has been abstracted so that different devices, or even different ways to access the same device, can be easily implemented.

Other Features

• Uses accelerometer to display text correctly with device on either side.
  • Note that this pauses during sub-modes (brightness & corrections).
• Uses the WiFi credentials that were setup in M5Burner.
• Heartbeat shows that the unit is working, even if the AQI isn't changing.
• There is also a "marching ants" chaser to show that it is working. See below.
• If one is in a sub-mode (brightness & corrections) and forgets to exit, it will exit back to the main menu eventually.

"Marching ants"

I'm currently on the fence about the "marching ants" animation. This animation is a hollow rectangle that runs around the edge of the display clockwise. Since I couldn't decide if I liked it or not, I put a constant in the aqi.py file. If CHASER is True, then the animation will run; otherwise it won't. It is True in this iteration.

Correction factors

The technology used by the Purple Air monitors can read somewhat high, particulately :) for fire smoke. There have been several studies that suggest correction factors.

None

There are 2 versions of "none", although one of them only works with the local web interface.

The local web interface provides AQI and color values directly. This is represented by "R" on the display.

AQI and color can be calculated from PM2.5, and both the local interface and the internet provide that value. This is represented by "N" on the display.

EPA

The EPA just recently updated their correction factor to use [separate equations for low and high PM2.5]( (https://youtu.be/G7CNziDkUok?t=1779), and this is now used. This is represented by an "E" on the display. (The original study on the accuracy of Purple Air sensors).

LRAPA

The Lane Regional Air Protection Agency in Springfield, Oregon has created a correction factor for wood smoke. This is represented by "L" on the display.

AQ&U

The University of Utah did a study. This adjustment is probably good for large urban areas like Salt Lake City. This is represented by "A" on the display.

Lawrence Berkeley Lab & UC Berkeley

The results from this study are "it depends on the event", but their current correction factor is pretty close to EPA and LRPA, so I didn't include it.

PM2.5

Sometimes you just want the raw number, maybe to see if things are working. The raw PM2.5 value is represented by "P"; the color is just gray.

Development

If you want to hack on this, read on.

urequest, I request, punting redirects

The Purple Air site appears to sometimes throw redirects, which the M5stack urequest module punts on. I have modified it to support redirection, based on code from a urequest that supports redirection.

Simulating the hardware

It is painful to develop on the M5StickC, so I created a hardware simulation using pygame. Pygame is easy to use, and it has primatives that are very similar to the micropython lcd module, making the simulation fairly straight-forward.

This allows you to put in print statements and see them, as well as seeing any logic errors without having to copy to the device every time.

In the pygame window, 'a' and 'b' stand in for the hardware buttons; 'q' quits the simulation.

The programs automatically detect if they are on the M5StickC and use the simulation if they are not.

Simulator requirements

1. python 3.
2. pip install pygame to get pygame for python.
3. DejaVu Sans Mono font.
4. You will need to modify PYTHONPATH to run the simulation (see below).
5. It's probably easiest to run out of the flash directory:

   purple_air\flash>python3 apps/LocalAQI.py
   

   or

   purple_air\flash>python3 apps/WebAQI.py
   

PYTHONPATH

In a command window:

echo %PYTHONPATH%.

If this gives you a path, use:

Set PYTHONPATH=C:\Path\to\purple_air\flash;%PYTHONPATH%

If it doesn't, do not add ;%PYTHONPATH%: windows will keep it as %PYTHONPATH% and everything will break!

I am indebted to this stackoverflow post, which I don't have enough "credit" to upvote.

Linux and Mac are more reasonable here: You always do the same thing:

export PYTHONPATH="/path/to/purple_air/flash;$PYTHONPATH"

Here is a handy way to see if the path looks right:

python -c "import sys; print('\n'.join(sys.path))"

Unit Tests

I wrote a few unit tests. To run these, you will need mock and parameterized:

pip install mock
pip install parameterized

Then, just:

python -m unittest