Using a photodiode in near darkness

[Artlav]

No, your original posted circuit had the photodiode in the correct orientation:

Ok, i seem to be have been confused by the datasheet, where the pins are named A and C, which are actually pins A and C, not Anode and Cathode...

So it works?

Very much so.
Thank you for helping.

Hey if the resistor work for you and you have no problem buying it for the product, go for it. Remember the parts availability is very important.

It's a hobby project, so parts availability is a matter of annoyance.

 

[yungman]

I cannot answer that as I don't have the data sheet. I assume you connect according to the drawing! I doubt you go this far if you connect it up side down!

I am impressed you are doing it as a hobby. As much as I have the passion in electronics, I don't think I'll do something like this out in the blue now...Yes, I did a lot when I was young!

 

[Redbelly98]

So it works?

Very much so.
Thank you for helping.

It's a hobby project, so parts availability is a matter of annoyance.

Glad it's working out

I was going to suggest -- if your signal is too low, another way to increase it is to use a photodiode with a larger area, which would collect proportionately more of the ambient light. The active area of the photodiode is often given in the spec sheet.

But if you got it working, no worries.

 

[yungman]

Bigger diode is a double edge sword, the capacitance goes up and noise gain goes up and a bigger feedback cap is need to control the noise, this will slow the amp down. Actually 10GΩ is not that bad. I have seen 100G used. You just need to have MOS input op-amp. About speed, he is using a less than good resistor that has parasitic capacitance. There are resistors from Caddock available that don't have capacitance like this.

Actually if design allows, I prefer using two PD circuit in parallel ie two smaller PD opamp circuit and sum together. When summing, you get double the PD signal, but the noise being uncorrelated, only increase by 1.41 times. This obviously won't work for system that the light is fed by fiber-optics.

 

[Artlav]

Well, the problems never end. :(

The new confusing one is what can be described as sticky zero.
Before there was always some amount of ambient light that i kind of subtracted, and everything worked beautifully.
Then, I've been improving light insulation of the box, and there came a point where the values read are essentially 0 for prolonged period of time.

The problem is, that 0 sticks.
Once the value reaches zero (which is below about 2pA for PD or 5mV of output for ADC) it gets stuck there.

Giving it some light does not change the value. Giving it a bit more light tears it loose, and the sensitivity is back - i can keep varying the light around there and all is properly registered.
I can get it down to almost 0 and back, and all is read right.
But once it touches zero again, it stays there until light goes up enough.

It's not an ADC issue, I've also measured the output voltage directly.
Go below about 18-12mV, and it snaps to 4.3mV and stays there, unless enough light is added to get it into about 40-50mV range.

What the could this be?

 

[Artlav]

Actually, one thing i can think of is hitting the rail, which as far as i read, could cause it to latch up for a moment, and that is what happens.

So, should i add a negative side supply, or somehow prevent the current from reaching the dead zone?
If i got this right, providing -5V would allow it to go to and over 0, thus eliminating the sticking?
Any alternatives?

 

[yungman]

latch up only happen if you drive the input to -0.7V or so. Even saturation of the output transistor only take a few uS to come out of it.

You might have an offset problem! Think about if you have a +ve offset, the output want to go negative. Of cause it won't as you don't have a -ve supply. If you have a -ve supply, you might see the output actually sit at negative a few mV. the reason you need to give a little light to "unstick" it because you need the light to overcome the few mV offset, not exactly latching at the negative rail. Putting a -ve supply alone is not going to help this problem.

You do need a -5V supply. One way to deal with it is put a 5K potentiometer with one side of the pot resistor grounded and the other side to a 500K resistor to -5V. You have to use another of the high value resistor and solder on side to pin 2 of the opamp, the other side to the wiper of the potentiometer. This will give you about +50mV offset. Adjust the pot in total darkness until you see the LSB of the ADC toggle. This show you null the offset to zero. You should not see the sticking anymore.



Try it a report back.

 

[Artlav]

You might have an offset problem! Think about if you have a +ve offset, the output want to go negative. Of cause it won't as you don't have a -ve supply. If you have a -ve supply, you might see the output actually sit at negative a few mV. the reason you need to give a little light to "unstick" it because you need the light to overcome the few mV offset, not exactly latching at the negative rail. Putting a -ve supply alone is not going to help this problem.

Not really clear.
There is a +4.3mV offset i can see, and unstick point is about ten times that.
If it's just an issue of it trying to go below zero, then why does it work properly until it touches zero?

i.e. if i give the light in a curve of 10-0-10, then it would go from 10 to 0 and stay at 0 forever (unless more light added), but if i give it 10-1-10, then it would go down to 1 and then back to 10 all right.

This show you null the offset to zero. You should not see the sticking anymore.

Having trouble getting it.
The point is to shift the - of the op-amp down by enough to zero the offset, so that there would be no +4.3mV when the current is 0?
How is that related to sticking?

The offset itself does not sound like an issue - all i should do is subtract it when i get the output, no?
Loosing 1-2% of the range is not that much of a problem.

And if i'll be adding more big resistors, why not just add it to the other side to give a little current to keep it always above the same 20mV, without the need for a negative voltage?

 

[Artlav]

Haven't tried anything so far - i can only get unusual parts (and charge pump chip for negative voltage is not something i'd call usual) on weekends without much trouble.

In the mean time, there is another problem to clarify.
I've been trying another photodiode, and this one have 1200pF of capacitance.
When the light goes off the scale on the upper side, it stays at maximum brightness for some time before returning.
i.e. bright light on, reads full, light off, stays full for a tenth of a second, then gets back to correct reading.

Am i right to attribute this effect to the increased sensor capacitance?
It's not there on the other one with low capacitance, but just about everything is different between the two.

 

[jrive]

To comment on a previous post, there is nothing wrong with th way the diode is drawn in the original schematic. In that case, you´re using the diode in photovoltaic mode, with zero bias across it (because of the virtual ground provided by the op-amp). This connection mode is the quitest method, but slows down the response when compared to the reversed bias configuration. The greater the reverse bias across the photodiode, the faster the response, but also the noisier the circuit becomes.

Current flows "into" the diode, so the although the transimpedance amp looks like an invertere, since current is flowing into the photodiode, the output will be positive going as light strikes the photodiode.

Stabilizing it should be no problem by ensuring the pole created by the feedback cap cancels out the zero created by the junction capacitance of the photodiode. There's a ton of literature on how to size this cap based on this and the desired bandwidth.


Another option for you, if you need higher gain (without requiring the huge feedback resistor whichi adds noise) and faster response is not necessary is to use a phototransistor instead of the photodiode. Not sure what your application is, but if you're dealing with low light, and speed is not an issue, they are better suited. Another option yet is to use a T-network feedback scheme which will enable you to increase your gain significantly without having to resort to huge valued resistors...

 

[Artlav]

Another option for you, if you need higher gain (without requiring the huge feedback resistor whichi adds noise) and faster response is not necessary is to use a phototransistor instead of the photodiode. Not sure what your application is, but if you're dealing with low light, and speed is not an issue, they are better suited.

Phototransistors are kind of rare for non-visible and ir bands, and the idea of my project was to use various photodiodes to see in various bands, from ~5um to ultraviolet.

Basic description of the project is here:
http://orbides.1gb.ru/photobot.php?lng=eng

The bandwidth needed is only about 1KHz - the motors don't go faster.

The sensitivity at 2.2GOhms is already somewhat too much, actually - it gets off the scale in broad sunlight. At the same time, indoors it's almost too faint to see, so i tried some higher values with little success due to resistor's structure.

Another option yet is to use a T-network feedback scheme which will enable you to increase your gain significantly without having to resort to huge valued resistors...

I can't google up much on "T-network feedback", can you provide any details?

 

[yungman]

I have been busy as my wife just had hip replacement on Monday. Really don't have time to think about this for now.

Regarding on the sticky at 0, I have no idea as the circuit recovery time is so much shorter than what you are describing.

Regarding to too much gain in bright light and not enough in dark, use two different feedback resistors and use relay to kick it in an out. Say you use two feedback resistor, one is 500M, the other is 5G. One end of both resistor connect directly to pin 2 of the opamp. The other end connect to the relay. The relay is to switch the output of the opamp to either the 500M or 5G. You then switch the relay depend on the brightness of the surrounding. Easy. That's what we do all the time, you don't get enough dynamic range from one resistor.

 

[Artlav]

I have been busy as my wife just had hip replacement on Monday. Really don't have time to think about this for now.

No problems, I'm in no hurry. Good health to her.

When you get some time, or someone else look at this, here is he situation so far.
I've been trying to get negative supply for the amp. The best idea i had is MAX660 charge pump chip, turning +5V into -5V. 10uF cap for the pump, 100uF one to filter the output voltage, but I'm still getting a load of noise just from the thing being here. Also tried at different frequencies and with different caps - the noise does not go away.

So, is there either a way to set up the negative voltage properly, or a way to remove sticking without using that?

 

[Artlav]

The solution have found the problem.
Not surprisingly, it was all in the manual.

AD820 have two NULL pins for offset voltage compensation, by linking these to +5V with 4.7K and 10K resistors, i got the output 0 to be in the 9 to 30 range (1=5mV), depending on the PD used.
And presto, nothing sticks to anything, without any troublesome negative voltages.

 

[yungman]

Glad you find the problem. Are you going to put in the relay to extend the range for different situation? I think it should work as you can have one setting for indoor or night, and the other setting for bright outdoor.

The T-Network is a common scheme that use a voltage divider with low value resistor and the big resistor connects to the divider. eg. If you need a 1G resistor. You can use a divider using 9K and 1K to divide the output by 10, then you only need to use a 100M resistor to connect to the divider to pin 2 and get the same gain. The big down side is the noise is going to increase by 10 times and the offset will increase by 10 times. The good point is you get to use a smaller feedback resistor and therefore increase speed.

I did not suggest the T-network because you are dealing with noise already at the time. It is a standard practice to avoid using T-network unless you really run into speed problem. This is no different than using a 100M resistor and then put a X10 gain stage following it. All the problems are amplified by 10 times using this.

 

[Artlav]

Glad you find the problem. Are you going to put in the relay to extend the range for different situation? I think it should work as you can have one setting for indoor or night, and the other setting for bright outdoor.

Something like this, quite likely. A simple manual switch with three positions should suffice - sunlight, overcast, indoors. 500M, 2.2G, ~10G respectively. It's not moving around fast enough to warrant the complexity of a relay.