Analyzing Recordings of a Mobile Phone Lying Still
What is the background “noise” in the sensors of a mobile phone? In the fourMs Lab, we have a tradition of testing the noise levels of various devices. Over the last few years, we have been using mobile phones in multiple experiments, including the MusicLab app that has been used in public research concerts, such as MusicLab Copenhagen.
I have yet to conduct a systematic study of many mobile phones lying still, but today I tried recording my phone—a Samsung Galaxy Ultra S21—lying still on the table for ten minutes. I did this using the Physics Toolbox Sensor Suite by Vieyra Software. There was no particular reason for choosing this software; it just showed up with high ratings in the app store.
The exported CSV files had some formatting issues, but once I solved those, I could do some plotting. The time plot of the accelerometer values shows a consistent noise level:
Plotting the accelerometer data axes against each other shows that the noise is even also in the spatial domain:
The same can be said for the gyroscope data, although there are some spikes there:
The spikes can be more clearly seen in the planar plots:
There is nothing revolutionary to report here. The plots show that the noise levels of the built-in sensors are low, which was what I wanted to confirm.
Update (2022-12-30): I have uploaded the Jupyter notebook with the analysis on GitHub now.