photodiode dark current at zero bias

I'm trying to detect very low levels of light using a PN photodiode. The
photodiode is connected to input of a CMOS op-amp and the bias voltage
is less than +-0.3mV. When I remove all light by placing the detector
in a sealed box, there still remains about 200pA of output current.

Disconnecting the photodiode showed the circuit stray leakage current to
be less than 5pA, so that cannot be blamed.

My question is does dark current supposed to exist at zero bias ? If not
then is the current caused by IR radiation from small temperature
differences between the photodiode junction and surrounding surfaces ?

Adam

 

Might be DC offset of the opamp driving a little leakage in the pd.
Add an offset trimpot maybe?

You might have thermocouple effects, driven by opamp self-heating, but
I'm guessing that's unlikely. A common PN photodiode will be blind to
longwave (thermal) IR.

John

 

I believe the sensitivity of a silicon junction has a peak in the
(high) IR region...

 

and most have to be used with a blue filter which reduces transmission ...

have you also checked to see whether the opamp itself is oscillating ? that
is, this in itself can cause offset problems -- the photodiode's junction
capacitance is related to the surface area so the opamps get a little
troublesome to compensate. check out Burr Brown (TI's) website for ideas on
compensating photodiode amplifiers.

 

Silicon photodiodes are not sensitive above about 1100 nm. The peak
(in amps/watt) is typically around 950 nm, which could be shifted if
filters are used but it can't go above 1100 nm (photon energy becomes
less than band gap energy).

This is all way below room-temperature thermal IR.

Mark

 

I've also had problems connecting a photodiode directly to an op-amp.
Mind you, these were just cheap 741's. But I did find that putting a
small resistor between the photodiode and opamp helped (I think I used
100 ohm, might have been 10 ohm). I assumed the problem was with the
input offset voltage, but didn't analyze it in detail once I got the
thing to work.

Try a resistor in series with the photodiode.

Mark

 

Hmm. No, that doesn't seem right at first blush. How about the following?

1. (Tire-kicking)
I assume you're using the usual op amp TIA, with the + input and the PD
anode grounded, the - input connected to the PD cathode, and a big
feedback resistor connected between the - input and output of the op
amp. How big is the feedback resistor, and how are you measuring the
output voltage? Scope? DVM? ADC card and Labview? (involuntary
shudder) Are you using split supplies, or an op amp with common-mode
range including the negative rail? How sure are you that the input
offset voltage is < 300 uV at your actual bias point? (You probably
didn't learn that by hanging a DVM off the summing junction.)

2. (PD details)
If you're trying to detect very low levels, you're probably using a huge
PD, right? Diodes at zero bias have a large but finite resistance.
300 uV/200 pA = 1.5 Mohm, which is certainly on the low side even for a
big PD. What exactly are you using for a PD, and does it have a spec
for zero-bias resistance? (A solar-cell type diode might very well have
a zero-bias resistance of this order.) A definitive test for PD
leakage would be to build a voltage divider, using a 10k resistor to a
variable *negative* supply voltage, and a 10-ohm resistor to ground;
connect the PD to the tap and see if you can make the output move around
by changing the PD bias by a few millivolts.

Does warming the PD very slightly make it noticeably worse? Besides the
zero-bias resistance, the thermocouple potential between silicon and any
metal is gigantic--400 uV/K or thereabouts, which under some
circumstances might dwarf the 300 uV input offset voltage.

3. (Light leaks)
Or you might be using a plastic cover that looks black but lets 800-1100
nm light pass through--the window filter of an IR remote control is like
that, and it's not unknown in other plastics--some very early plastic IC
packages had that problem, in fact. If your box is plastic, try
wrapping it in aluminum foil. In fact, try wrapping it in aluminum foil
anyway, to make sure it isn't light leaks that you're seeing.

4. (Instability)
Make sure you have a capacitor in parallel with your feedback resistor,
to reduce the noise BW and eliminate the instability that the other guys
are warning about--though I'd expect an oscillation to result in at
least hundreds of millivolts of dc shift, even if you're measuring the
output with a DVM.

Cheers,

Phil Hobbs

 

And if you could put a pv device inside a box, and have it generate
voltage from the IR emitted thermally by the walls of said box, it
would violate conservation of energy and a law or two of
thermodynamics.

John

 

Thanks Phil, for your 4 points. I just logged on to reply with the
problem I found causing the small DC shift of ~20mV. Instability.

After attaching a CRO to the output of the op-amp I noticed an amazingly
high frequency 20mVpk-pk signal, almost in the VHF range. Who would
think a slow GBW 1.5MHz LMC6482 could cause such a thing ?
Actually I think the oscillation was due to the 2nd op-amp in that
package which was driving a light feedback stabilized LED circuit.
Adding some bypass capacitors fixed the problem.

The output voltage was measured with DVM, the op-amp feedback was 20Mohm
in parallel with 47nF.

 

Nope!

 

Or do you refer to Johnson noise? Or have you invented perpetual
motion?

John

 

Actually, perpetual motion was invented the day that politicians were
invented...

 

Politicians weren't invented. They are the spawn of satan.

 

Well, it's obviously Doomsday. Here, in one day, I've found myself:
(A) laughing out loud at a joke that made Jim Thompson Sno-o-o-ort,
and
(B) Agreeing with Michael A. Terrell about politicians.

Will someone please send me some better drugs?

Thanks,
Rich

 

So who was the mom?

Cheers,

Phil Hobbs

 

Rosemary.

 

The Mother of Everything, of course:
http://www.godchannel.com/quest.html