I did some test with the InGaAs photodiode and 940nm LED.

Due to large area (and capacitance) of InGaAs diode, I had some instabilities in the oscillator. I corrected part of the problem adding capacitors to voltage sources and voltage reference. However, sometimes few seconds are necessary before the circuit starts to oscillate right. The next PCB version will correct this, I think... by adding more capacitance, individual voltage regulators, etc.

Here I show some measurements with ambient light only (The flourencent lamp have a very very low 800-1700nm emission) varying light position, facing or not facing to the lamp.

When I turn on the 940nm LED, with 50mA (~50mW) the ADC output goes to ~123000. With earlobe between LED and photodiode the output goes to ~16000. So, a earlobe absorbs 92% of light with 940nm.

The conclusions:

- The ambient/background plus dark current is very low in comparison to the work conditions.

- The circuit is very sensitive and could be better using different settings (smaller 3.3-vref, bigger integration time).

- The ADC max value is enourmous (max output = 4 bi) and, with integration time of 1 sec, I used only a small part (150k) of its maximum, so, I can use integration time of days!! I think it will be enough to get the necessary precision.

- The earlobe absorbs 92% of 940 nm light.