One problem with my current code for very low frequencies is that the frequency resolution, or the width of each FFT bin in the frequency domain, is pretty poor. With my current code, we're collecting 64 samples at a sample rate of around 40 kHz. 64 samples in the time domain yields 64 frequency bins after an FFT. With a sample rate of 40 kHz, you can resolve frequencies 20 kHz and below (per the Nyquist sampling theorem). 20,000 Hertz / 64 bins = 312 Hz/bin. This means that all frequencies between 0 and 312 Hz are grouped into one number, all frequencies between 313 and 625 Hz are grouped into one number, etc. You can increase the number of samples, but it will then take longer to collect those samples. Once you find a good tradeoff between sampling time and frequency resolution, you should just be able to look at only the frequency bins below 50 Hz and use the Arduino Map function to control the RGB intensity values like I do in the example code. Let me know if you have any more questions!

Glad to hear everything went well and it was annoying enough to have someone call maintenance to fix it! Let me know if you decide to do anything else with the Annoy-O-Bug

Glad to hear you're helping this circuit live up to its annoying potential ;) I didn't think that magnets could affect the piezo speaker; my understanding was that they don't use magnets and instead have a piezoelectric crystal as the actuator. Perhaps you're covering the hole of the buzzer with the magnets? My recommendation would be to hot glue it by where the positive battery terminal is soldered. Anyways, let me know what your co-workers think!