Spitfire compass swing to calibrate the onboard compass from The Royal Air Force in Pictures (1941).

The sensor plotter shows some of the characteristics of the sensors. A few are highlighted here.


The temperature on the CLUE board appears to over-read. This is likely to be from the electronics on the board heating the sensor. The graph below shows a value around 30.8C for a room at 23C.

Plotting temperature showing drop of 0.25C when sensor is not read for a short period while the progam jump scrolls the plot.

The graph clearly shows the temperature readings are not constant and have a distinctive pattern not representative of (random) noise. This screenshot was taken with the program in scrolling mode where it scrolls in jumps of one division. That scroll takes a few hundred milliseconds. It's possible that when the sensor is not being read it cools down a tiny amount but more research is required here.


The pressure is surprisingly accurate in matching local meteorological observations (adjusted for height difference). This may vary per device. The relative pressure change for changes in altitude are capable of measuring changes of a few feet. There is some noise on the sensor visible in the video below. If the vertical movements are known to be slow then filtering could be used to improve the accuracy.


Unlike a traditional wet-bulb thermometer the humidity sensor responds very rapidly to changes in humidity. It responds sub-second and can easily detect human breath nearby. In some circumstances this might be a disadvantage - careful sampling and filtering might be required to ensure a value is achieved that truly represents the local atmosphere.


The volume is calculated from samples taken from a microphone on the board. This means it will pick up physical shocks to the board like button presses. These spikes can cause the auto-ranging to expand too much.

Colour Sensitivity

The colour sensitivity is a bit surprising. The NeoPixel on a Feather M4 Express was used to test the red, green and blue plotting. The blue is clearly the highest of the three despite the green appearing to be the brightest to the human eye. Some ambient light was present but this plus the cross-sensitivity can be seen on the plot - these are not factors here.

Plotting light from NeoPixel illumination from Feather M4 plus a little ambient light.

The APDS-9960 datasheet has a graph showing the sensitivity of the different colour sensors but this seems to run counter to what is observed.

Spectral response of the photodiodes used for colour detection and proximity/gesture sensing from the Avago APDS-9960 datasheet.


The proximity sensor is based on a simple value derived from measuring the infrared light reflected from the object illuminated by its 950nm IR LED. This will vary depending on the infrared reflectance of the object so cannot be considered to be an absolute measure of inverse distance. For example, a matte black, plastic USB connector only registers a maximum proximity of 40. The absolute maximum is 255.


The magnetometer plots on the two CLUE boards used during development of the sensor plotter showed very different values. In the absence of any strong magnetic fields the sensor picks up the Earth's magentic field, therefore the boards must be in the same position and orientation for a valid comparison. A pair of representative samples were:

  • CLUE 1: (-11.1371, -7.04472, 40.2952),
  • CLUE 2: (-46.5653, 20.8272, 21.2803).

Magnetometer sensors are well-known for their need to be calibrated. This non-trivial process is described in Adafruit SensorLab - Magnetometer Calibration.

A set of sample readings for two (uncalibrated) CLUE boards are shown below, comparing the values from the Earth's field and that from a small neodymium magnet placed 8cm away.

This guide was first published on Apr 01, 2020. It was last updated on Apr 01, 2020.
This page (Sensors) was last updated on Jul 03, 2020.