We are currently doing measurements to compare the signal to noise ratios (SNR) of the PPG signals that we record with various setups. While doing so, we noticed that although our current software implementation does perform a calibration of the slow 32kHz clock, it does NOT perform a calibration of the fast 32MHz clock. Now we are wondering if this negatively influences the quality of the measured signals, and if so, to what extend it influences the signal quality. Can you please clarify this for us?
The 32MHz clock is responsible for providing LED timing. If you don’t perform a calibration of this clock then it’s possible that, from part to part, you will see slight variations in LED pulse widths. This will result in part to part variation in signal levels in the application. For example, if you are providing 2us LED pulses to two different parts and one part has a 32MHz oscillator running a little fast and the other a little slow, the fast part will have slightly shorter LED pulses and the slow part will have slightly longer LED pulses. The part with slightly longer LED pulses will end up with a higher signal level than the part with the slightly shorter pulses. This won’t affect the SNR performance of either part, but there will be part to part variations in signal level for the same LED timing settings. If this doesn't matter in the application then you don’t have to bother with clock calibration of the 32MHz clock.