Spectrum RGB

Essential REST API service to convert Spectrum to RGB & Colour Temperature to RGB

• Python REST API using FastAPI
• Docker image via github actions (arm and x86)
• Tailoured for frequences from AdaFruti AS7341
• Generate RGB values for specific colour temperature

Theory

Firstly, colo[u]r(1) theory is a large and complex topic. Add in how your brain perceives colours. You have been warned!

Overacall context for this was to take the the sprectrum values from the Adafruit AS7341 and convert that to a RGB value that can be sent to a tri-colour LED or strip. There is a tri-colour LED strip in the office that I wanted to control.

Idea being that the colour of the LED would reflect the outside colour; by adding in the ability to convert colour temperature to RGB is a bonus.

Specifically the I'm aiming for 'emperically perceptually equivalent'. (i.e. it looked ok-ish). This will only ever be an aproximation; but this is not intended for serious scientific use.

AS7341

This outputs counts for a number of frequencies: 415nm, 445nm, 480nm, 515nm, 555nm, 590nm, 630nm, 680nm In addition there near-ir counts, but I've stuck with the visible counts.

As an example here's the distribution of these frequencies currently getting.

The sensor gets these values to a MQTT topic, so the Node-Red flow is processing these already; but it need to get the RGB value to send to the LEDs.

Spectrum to RGB Conversion

A superb post on the SciPython site provided the theory and also some python code that was very close to what was needed. This specifically calculates the RGB for given colour temperature; colour temperature was converted to a spectrum of wave lengths, and then to RGB.

The code here is to a very large extent taken from the code on that page; with two changes:

• The sprectrum it uses is frequencies from 380nm increase in 5nm increments. With only 8 entires in the AS7341 spectrum this need some changes. The cie-cmf.txt file was reduced to just the frequencies I had (the abrdiged-cie-cmf.txt file). Suspect that this is a source of loss of accuracy. How wide the frequency response of the sensor is I don't know.

• Some refectoring to make it easier to handle

REST API

Using the FastAPI it was straightforward to create a simple Python REST api that would accept a simple JSON structure with the frequency counts. Process this via the algorithm above, and return RGB values (both in hex and int styles) in a JSON structure.

Similarly the API can is also converting a colour temperature to a spectrum for conversion to RGB. This is using the full range as per the original code.

That's it really, built into a docker image via github actions - ready to be deployed into my Portainer configuration.

Deployment

Briefly, the system is configured as follows:

• Arduino captures sensor values, and sends via 433Mhz
• RaspberryPi I've called the 'EdgeController' receives the 433Mhz signal, and forwards on via MQTT
• Node-RED flow triggered by MQTT. This takes the spectrum count array, passes to this SpectrumRGB service, and then publishes on MQTT
• PicoW is subscribed to MQTT and does the LED strip control

Development Notes

• Using Poetry for handling the depdencies of the python project; currently seems to be leading tool and handles most situations well.
• As above, this is using FastAPI. Previously used uvicorn, but recently I've had issues with that and HTTP/2, so using hypercord. Purely a pragmatic decision.

Notes