Light to Frequency converter using LM331

[ChAndrea]

Hi all,

I designed a discrete light to frequency converter using a LM331 in a "precision voltage to frequency converter" configuration.

I attached the schematics. This circuit works well when I use a controlled voltage (and a resistor) or a current source as the input, but when I put the photodiode the frequency output slowly drifts over time in response to a constant light (3 leds inside a shielded box with the photodiode).

The drift is sometimes positive and sometimes negative, with a really slow dynamic. The problem is I could not find any deterministic explanation to this phenomena (maybe a temperature drift, I read is around 1-2% of the signal every 5C°). The drift is usually higher if the current flowing is higher, around 10-50 nA per 10secs with a 5uA current, slowly stabilizing after a light step.

The photodiode is ideally unbiased, but during operation each time a "reset" occurs a low reverse bias is applied because of the finite op amp GBW. This could affect accuracy but cause a drift (with such a slow dynamic)?

I checked the bode plot of op amp integrator using a classical photodiode model, with a shunt resistor and capacitor. Everything should work fine.

The photodiodes are PIN (I tried several of them). To measure the frequency I'm using a Tektrotronix 2022c oscilloscope, but I obtain the same results using a microcontrollor as a frequency meter.

I'm working with high resolution (10Khz per 1uA circa) so this could be a normal behaviour which I could not see using a transimpedance amp.

I'd like to know if you ever encountered something similiar, and if I'm missing some pheomena.

Thank you for your help and time.

Andrea.

 

[FvM]

Did you try to apply a larger reverse bias to the photodiode by raising the voltage at the n.i. OP input?

I think, a transient up-swing of the inverting input and respective reverse photodiode bias won't be a problem, but succeeding positive bias in the transient response will eat-up part of the input current.

 

[ChAndrea]

Actually I put a voltage offset compensation network on the n.i. op amp input (+-500uV). I'll try with a larger bias.
Anyway I checked the inverting input voltage several times, and I did not see any transient positive bias (the transient reverse bias is maximum 5mV then drops to zero).
Just to be sure I'll try it, thanks.

 

[ChAndrea]

Well, today I put a reverse bias across the photodiode. Unfortunately the drift was still there. Sometimes positive sometimes negative. I am a bit lost. Anyway thank for suggestion.

 

[FvM]

How do you know that the light intensity is actually constant?

 

[ChAndrea]

I can't; in fact I start to believe is just normal. This is a simple application and i cant find any weirdness. Anyway, I checked the inverting pin voltage and I saw a small forward bias during the transient, so adding a small reverse bias was a good suggestion. I'll leave the circuit in this configuration. Thank you.

 

[schmitt trigger]

I'm not saying that the circuit is wrong, but I haven't seen a photodiode/LM331/transimpedance amp configuration like you have.

Most app notes indicate that, to operate in photocurrent mode, to have a constant negative bias. This bias is around -3 to -5volt.

 

[ChAndrea]

Well, I started with a TIA coupled with a microcontroller ADC.
However I needed higher resolution and a wider dynamic range at low cost, while speed was not important. Furthermore I was interested in signal integration (charge), but over long period of time (no charge sensitive amplifier).
I had some LM331 in the lab, and following https://www.ti.com/lit/an/snoa594b/snoa594b.pdf I designed this. This is not probably the newest and best solution, but interfacing the output to a micro I could just count the total number of pulses, obtaining the information.
There are some integrated light to frequency converter, but I'd like to change the sensor freely.
Using a better op amp (with an higher GBW) could improve performance in response to the reset circuit.
Anyway some time ago I even found the same circuit (with just few mV of bias to stay "reverse biased" during reset, photovoltaic) used to detect some fluorescence or similar. Right now I can't find it.

 

[FvM]

Basically, avoiding current-to-voltage conversion promises a higher dynamic. In so far the circuit topology looks reasonable.

As already stated, I don't see what causes the drift. As long as there's no transient overload (e.g. unintentional forward biasing of photo diode and other junctions) and the charge balancing principle is fulfilled, even limited GBW won't cause drift.

 

[KlausST]

Hi,

I just made an LTspice simulation on a simple I-f converter circuit.

circuit:

Q1, V2, R1, V3 is just to simulate a constant current here. You don´t need it. Replace it with your photodiode.

Your photodiode discharges C1 down to a dedicated threshold (adjusted with R2, R3, R4), then the comparator output turns to HIGH.
With R5 and D1 (low leakage) the capacitor C1 is charged to a dedicated level (adjusted with R2, R3, R4) then the comparator output turns to LOW.

The comparator is for a constant time HIGH, but the LOW time depends on photodiode current.
With well calculated device values there will be an almost constant I to f ratio.

Here the result of the simulation:


Hope this helps

Klaus