Photodiode+sunlight+lock-in amplifier, something weird happens

[tinska.h]

Hello All,

I have a set-up which consists a photodiode(non biased set-up) -> transimpedance amplifier -> sallen-key band-pass filter -> anti-aliasing filter -> ADC -> digital lock-in amplifier. In analog amplifier section, I'm referencing all ground connections(except op-amp's VSS pin) to a virtual ground. Digital lock-in amplifier locks in to a ~980Hz square signal, sampling rate is 10 000.1Hz. Behavior is very good in normal room condition with curtains closed. Device is working as expected.

However when sunlight hits the photodiode (non directly) I get approx. 500µV shift to the results (depending on sunlight intensity).
I have made sure that:
- Photodiode is not saturating, also since photodiode is operated in non-biased mode, light response should be linear as possible
- Virtual ground impedance is not a problem, since I'm using a rail splitter which has approx <1 Ohm sink/source impedance. Also I have tried to use the real ground, but problem still exists.
- With DSP I have calculated average, min and max values from signal batch which DOES NOT change significantly (only +/- one LSB) when sunlight hits the photodiode.

I've been investigating this whole day and I can't think of any more ideas. So fellow forumers, all expertise, ideas and suggestions are welcome!

Thank you in advance!

 

[tinska.h]

Could capacitors in high pass section encounter any loading effects when constant DC offset is applied?

 

[keith1200rs]

You need to show your circuit.

Keith

 

[tinska.h]

Here is a description of op amp configuration. ADC is AD7980 (16bit) referenced to ground and clean 5V VDD rail. Photodiode outputs a square wave current signal which is converted to voltage, passed through to band pass.

When sunlight hits the photodiode I get increase in results. Maximum pk-pk of amplitude does not change and average does not change when this happens. I guess signal gets somehow distorted which affects my lock-in results? Could filter become loaded when a DC-offset is applied to it?

 

[keith1200rs]

Is your 2.5V shifting?

Keith

 

[tinska.h]

Hi Keith,

I tried to connect anode pin to ground plane, however the result shift amplitude stayed the same. Although when I did connect anode pin to ground, I left positive input at 2.5V this means that photodiode becomes reverse biased which affects linearity and capacitance of photodiode. But then I gain I should seen difference in pk-pk maximum and avg values.

I suspect that either photodiode characteristics change or filter phase response change when sunlight hits it. But what could cause that?

 

[keith1200rs]

The opamp bandwidth could change depending on the DC input voltage. I don't know if that would be sufficient to affect your results though as the speed should be dominated by your external components.

It is possible that the photodiode speed changes with ambient light current. More charge could increase the effective capacitance.

Keith.

 

[keith1200rs]

Just to add, while it isn't the same situation, I had an unexpected effect with APDs where the phase shift varies with light level. While this was partially expected due to small voltage drops due to protection resistors the effect was far greater than the theory would suggest. It was observed with different manufacturers devices but a satisfactory explanation was never received from any of them. However, the effect varied considerably from one manufacturer to another.

Trying a different photodiode may be worthwhile.

Keith.

 

[FvM]

1. I'm missing a clear description. You say "Photodiode is not saturating". That's in fact obvious, because in case, the OP will saturate, not the diode. So what's the DC voltage shift with sunlight applied.
2. I don't yet exactly understand which quantity changes by 500 µV. You say "average, min and max values" don't change. Which value does? In any case, the error should be related to signal amplitudes, that aren't said.

Presumed, all amplifiers are in linear range, there may be nonlinear diode effects. The outer diode voltage is kept constant by the OP, but the inner junction voltage will slightly change according to the bulk resistance.

 

[tinska.h]

Keith, I guess I have to try some other manufacturers and models. I guess I'd have to try some low junction capacitance models.

FvM
1. thats correct, I meant actually the both. DC shift measured after transimpedance amplifier is approx. 200-500mV, depending on how the sunlight hits the photodiode. I have made sure that my op is not saturating. NOR my photodiode is in a state where it can encounter non-linearities.

2. Yes, I'm using a digital lock in amplifier(quadrature) which basically multiplies sin and cos reference signal to the actual signal. Then averaging result with a low pass and making Sqrt(X^2+Y^2) to neglect any need to adjust the phase. What I described is that when I do not apply lock-in technique, signal's minimum or maximum amplitude does not change, and also average does not change. However when I apply the lock-in I get shift in lock-in's result. I think this suggests that the phase response of analogous circuitry changes when sunlight is applied to it.

 

[FvM]

Thanks for the clarification. It would be interesting to calculate the actual phase shift from the quadrature measurement.

I previously considered non-linear diode effects. Reviewing your circuit, I guess it may be also a problem of virtual ground crosstalk. The "rail splitter" impedance can be expected to vary with DC bias. I would try separate blocking of diode and filter stage virtual ground.