Using a photodiode in near darkness

[Artlav]

Hello.

How can get maximum sensitivity out of a photodiode?

So far the best result i got is by using a current-to-voltage amplifier:

But with R at 1 MOhm it's still not good enough, and any more causes the thing to start oscillating out of control.

Would adding another amplifier after that improve the situation, or would it just amplify some noise? If yes, what kind of amplifier should it be?

Is there a way or point to remove whatever noise or oscillation this circuit gives with R>1MOhm, and keep adding resistors?

Or, how to do it some other (proper?) way?

 

[dlgoff]

According to the AD820 Data Sheet, for a single supply, the voltage can be +3 V to +36 V. You could try a higher supply voltage than the 5 V you're showing. i.e. if you're looking for a larger output voltage.

 

[yungman]

Just by the fact that the Rf=1M tells me that your photo diode give a lot of current...in close to 1uA range. This is no where even push any boundary unless you are looking for 100MHz + bandwidth.

You are right into transimpedance amp that we had two post talk extensively about. Tell us what is the condition, frequency response requirement etc.

For oscillation, first you have to tell us what is the capacitance of the photo diode. It is all about adjusting the close loop poles and zeros to tame the amp.

Post the spec and we'll talk more. I have a suspicion that you have parasitic problem that has nothing to do with the circuit in the schematic. eg, if the lead from the detector to the amp is very long, you'll have a serious problem.

More info.

What happen, all of a sudden transimpedance amp is in style?!

 

[Artlav]

I should mention that i had no prior experience with op-amps, and all i do now is based on a few days worth of reading the theory and experimenting.

You could try a higher supply voltage than the 5 V you're showing. i.e. if you're looking for a larger output voltage.

As far as i understand, supply voltage defines the range of the output, not amplification?
So, if I'm trying to tell millivolts apart, what good would it be?

Just by the fact that the Rf=1M tells me that your photo diode give a lot of current...in close to 1uA range.

Is it normal to have MOhm range resistance in this kind of application?
I.e. is there a limit on how low a current can be amplified that way?

Tell us what is the condition, frequency response requirement etc.

Light level measurement, 10-100 samples per second ideal.
With the provided circuit i get output indistinguishable from zero most of the time.

For oscillation, first you have to tell us what is the capacitance of the photo diode. It is all about adjusting the close loop poles and zeros to tame the amp.

Not specified, nothing to measure with.

Post the spec and we'll talk more. I have a suspicion that you have parasitic problem that has nothing to do with the circuit in the schematic. eg, if the lead from the detector to the amp is very long, you'll have a serious problem.

The oscillations are white noise or close it it.
In case of 2M and 0.1uf it looks similar to 1M and no capacitor.
With 1M and 0.1uf there is no detectable noise.

Lead length is supposed to be half a meter, but the problem appears with the diode sitting right next to the amp.

 

[skeptic2]

Here is an application bulletin from Burr Brown, a manufacturer of very high quality op amps. You might try some of the ideas in it.

http://www.ti.com/lit/an/sboa061/sboa061.pdf

 

[yungman]

I should mention that i had no prior experience with op-amps, and all i do now is based on a few days worth of reading the theory and experimenting.

As far as i understand, supply voltage defines the range of the output, not amplification?
So, if I'm trying to tell millivolts apart, what good would it be?

Is it normal to have MOhm range resistance in this kind of application?
I.e. is there a limit on how low a current can be amplified that way?
I worked mostly with value much bigger than that!
Light level measurement, 10-100 samples per second ideal.
With the provided circuit i get output indistinguishable from zero most of the time.

Not specified, nothing to measure with.
The capacitance of the diode should be given in the data sheet.

The oscillations are white noise or close it it.
In case of 2M and 0.1uf it looks similar to 1M and no capacitor.
With 1M and 0.1uf there is no detectable noise.

Lead length is supposed to be half a meter, but the problem appears with the diode sitting right next to the amp.

How do you come up with the 1M as feedback resistor? What is the range of current you are trying to detect?

You put a 0.1uF cap across the 1M resistor, your frequency response is 1.59Hz! That is too slow. Transimpedance amp is more than just design basic op-amp circuit. Read the article Skeptic2 posted.

If your lead is 1/2m, you might have to put a 50Ω resistor in series with the photo diode close to the input of the op-amp to give some isolation.

 

[Artlav]

Here is an application bulletin from Burr Brown, a manufacturer of very high quality op amps. You might try some of the ideas in it.

Thanks, that's certainly of use.

The capacitance of the diode should be given in the data sheet.

Indeed it is in full one, and i was checking abbreviated one.
Diode Capacitance:
(VR = 0 V, f = 1 MHz, E = 0) 70 pF
(VR = 3 V, f = 1 MHz, E = 0) 40 pF

Also, I've drawn the diode wrong way around in the schematic above.

How do you come up with the 1M as feedback resistor?

V=IR, and 1M is the largest resistor I've seen on sale.
So i guesstimeted it to be close to the limit or something.

Now, a Giga-Ohm resistor?
Not sure where to buy, not straightforward to make, nothing to measure with directly.
10M ones seem to be available, so about 100M is the highest i know i can go at the moment.

What is the range of current you are trying to detect?

Apparently, below 5nA.
Assuming 5nA to be the upper level (at 5V), 20pA would be the lower sensitivity level.

You put a 0.1uF cap across the 1M resistor, your frequency response is 1.59Hz! That is too slow. Transimpedance amp is more than just design basic op-amp circuit.

Sounds about right.
If i understand correctly that it's 1/(2*pi*R*C), then for 100Hz i'll need C about 1.5nF?

I don't have anything from this range at hand today, so can't check.
The noise is what troubles me - with 2M it exist with any C i tried (100pF, 47nF, 100nF) and is white noise, while with 1M it only appear at 100pF and less and looks like mains hum with something about 4Hz added over.

I've looked over MT-050 article from the thread below, and the diode's capacitance seems to be the C1, R1 is the lead resistance, and C2 and R2 are out feedback/gain ones?
But I'm too slow at the moment to get just what does this give me, and how does the stark change in the noise kind come in.

Read the article Skeptic2 posted.

That is certainly of use.
Also, this thread that you seem to have mentioned, is an interesting read.
https://www.physicsforums.com/showthread.php?t=577647

If your lead is 1/2m, you might have to put a 50Ω resistor in series with the photo diode close to the input of the op-amp to give some isolation.

Where does the number come from?

 

[yungman]

Thanks, that's certainly of use.

Indeed it is in full one, and i was checking abbreviated one.
Diode Capacitance:
(VR = 0 V, f = 1 MHz, E = 0) 70 pF
(VR = 3 V, f = 1 MHz, E = 0) 40 pF
That's not too bad. Your circuit bias a 0V, so 70pF it is.
Also, I've drawn the diode wrong way around in the schematic above.

V=IR, and 1M is the largest resistor I've seen on sale.
So i guesstimeted it to be close to the limit or something.
http://search.digikey.com/scripts/DkSearch/dksus.dll
Now, a Giga-Ohm resistor?
Not sure where to buy, not straightforward to make, nothing to measure with directly.
10M ones seem to be available, so about 100M is the highest i know i can go at the moment.

Apparently, below 5nA.
Assuming 5nA to be the upper level (at 5V), 20pA would be the lower sensitivity level.
For 5nA to full scale of 5V, you need R=V/I=1GΩ
Sounds about right.
If i understand correctly that it's 1/(2*pi*R*C), then for 100Hz i'll need C about 1.5nF?
f=\frac 1 {2\pi R C}
For 1G and 100Hz, C=\frac 1 {2\pi R f}=1.6pF
Again, look in Digikey. They ship fast.

I don't have anything from this range at hand today, so can't check.
The noise is what troubles me - with 2M it exist with any C i tried (100pF, 47nF, 100nF) and is white noise, while with 1M it only appear at 100pF and less and looks like mains hum with something about 4Hz added over.

I've looked over MT-050 article from the thread below, and the diode's capacitance seems to be the C1, R1 is the lead resistance, and C2 and R2 are out feedback/gain ones?
But I'm too slow at the moment to get just what does this give me, and how does the stark change in the noise kind come in.

That is certainly of use.
Also, this thread that you seem to have mentioned, is an interesting read.
https://www.physicsforums.com/showthread.php?t=577647

Where does the number come from?

Yes, you need to learn more as you have to taylor the circuit. Use the last op-amp suggest in that post.

 

[Artlav]

Good news, I've figured out where the white noise came from - for the 2+ MΩ tests i put several 1MΩ resistors in series, but the capacitor spanned only the first one!

With that fixed, i got it up to 10MΩ, and things still look good.
Looks like finding these mythical 1GΩ resistors could actually be worth the trouble. :)

Thanks for all the info, that'll take some time to digest.

 

[yungman]

Good news, I've figured out where the white noise came from - for the 2+ MΩ tests i put several 1MΩ resistors in series, but the capacitor spanned only the first one!

With that fixed, i got it up to 10MΩ, and things still look good.
Looks like finding these mythical 1GΩ resistors could actually be worth the trouble. :)

Thanks for all the info, that'll take some time to digest.

Look up Caddock resistor. 1G is nothing! Make sure to get a resistor with longer body because the surface leakage might present problem if the body is too short. If you don't want to deal with 1G, do the 100M and follow with the next stage with gain of 10. Read the post you put in the last one, read the link in detail. You need exactly the gain as that one. All the discussion implies to you. You are lucky that you don't need high speed like the other post, for you, it would be much easier. The other poster is running into wall at the moment because his detector is higher capacitance, and he need much higher speed.

What! Transimpedance amp and diode detector are the latest fashion? All these post about this topic!

 

[Artlav]

Noise is a problem.
On a breadboard with 50 MΩ (5x10M) and 10pF there is no detectable noise.

But for real, with almost a meter long leads there is noise.
Adding 100Ω in series near the amp didn't help.

In the dark the noise looks like this (FFT at the output, 0-100Hz):

Does this look familiar?
Any ideas how to fix it?

Also, I've never worked with resistance this high before - anything to be aware of? Like leakage across a moisture film along the board or something?

 

[skeptic2]

With leads a meter long, are you using shielded cable? In the Burr Brown Bulletin, the oval loop tied to ground represents shielded cable.

 

[yungman]

Noise is a problem.
On a breadboard with 50 MΩ (5x10M) and 10pF there is no detectable noise.

But for real, with almost a meter long leads there is noise.
Adding 100Ω in series near the amp didn't help.

In the dark the noise looks like this (FFT at the output, 0-100Hz):

Does this look familiar?
Any ideas how to fix it?

Also, I've never worked with resistance this high before - anything to be aware of? Like leakage across a moisture film along the board or something?

You are picking up noise on the long input wires. Look into what Skeptic2 suggested.

As for resistance, don't be afraid, it's just a resistor. I gave you the hint that you need long body resistor. When I did the transimpedance amp with high value resistor, I used DIP package and I bent the -ve input up and solder the resistor on the pin up in the air. You need very good cleaning spray, compressed air to blow away the solvent right after cleaning. 1G is not that high, you do this, you'll likely be fine.

If you measure down to pA, watch out for vibration. I worked on Mass Spectrometers and the vibration from the turbo pumps introduce vibration noise. We end up using rigid coax for input and mount the amp on the frame to lower the vibration to the amp.

 

[Artlav]

You are picking up noise on the long input wires. Look into what Skeptic2 suggested.

The most annoying part is a ~2-4Hz regular-looking oscillation, with up to +-0.5V of amplitude at output. That one happens on a breadboard as well, with no extra wires.
It's not in the light or surroundings - happens from many lightsources and in many places, don't happen in none.
The "no detectable noise" part was about full darkness and near the upper limit.

Have i missed something again?

 

[yungman]

This is unique in your situation. You have to chase down the 2 to 4 Hz noise.

Can you explain:

It's not in the light or surroundings - happens from many lightsources and in many places, don't happen in none.
The "no detectable noise" part was about full darkness and near the upper limit.

Is it a regular periodic frequency or random of about 2 to 4 Hz? If it is happening on the breadboard, then put Faraday cage around as see whether it is EM or not. Also tape the top of the photo diode so no light can go through and see whether that change the amplitude.

 

[Artlav]

Regular and periodic, but not real - going from 100 samples per second to 200 turned it into a much higher frequency one, also periodic. Something is interfering or plain broken in the serial link part.

Time-out, i shouldn't be asking questions until i rule out all the obvious possibilities.

 

[yungman]

Regular and periodic, but not real - going from 100 samples per second to 200 turned it into a much higher frequency one, also periodic. Something is interfering or plain broken in the serial link part.

Time-out, i shouldn't be asking questions until i rule out all the obvious possibilities.

Now you are talking. Good thing you can produce the problem on the breadboard, try different shielding method including blocking light, copper tapes and report back. Use a scope to look at the signal, don't rely on the digitizing result, it might fool you.

One thing of interest, replace the photo diode with a 70pF cap and see whether you pick up the same noise.

 

[Artlav]

50Hz, mains hum.
At 100 samples/sec it was synced up almost perfectly, giving ~2Hz from drift, but at higher and non-divisible read-out frequencies it's plain and obvious.

It's RF - any light source indoors - hum.
Outdoors - no hum, no detectable noise at any light level with a flashlight.

How do you filter out mains hum...?
Any better ideas than a Faraday cage?

 

[yungman]

50Hz, mains hum.
At 100 samples/sec it was synced up almost perfectly, giving ~2Hz from drift, but at higher and non-divisible read-out frequencies it's plain and obvious.

It's RF - any light source indoors - hum.
Outdoors - no hum, no detectable noise at any light level with a flashlight.

How do you filter out mains hum...?
Any better ideas than a Faraday cage?

Any way to take pictures of the real circuit and the breadboard? One look might worth more than a thousand words. 50Hz, are you in Europe? I can't even get a sandwich out of you even if I can help...you too far away!

 

[yungman]

Hold your horses! Did you do what I asked to tape the window of the detector diode? If you did not, the 50Hz is likely from the light! Forget the picture, answer this first.

If it is from the light source to the detector, you need to talk to the one designing the optics. If that cannot be avoid, you might have to do cancellation. I did it on one of the transimpedance amp that has switching supply noise. We needed a Faraday Cup amp floating on 200V. I decided to use a small switching supply build inside the FC amp box!. To cancel the noise, take the ground noise and sum with the output of the FC amp and cancel or reduced the switching noise, it worked.

If so, and there is no way out as your signal has the 50Hz, then you might have to generate the 50Hz of opposite phase and sum to cancel the noise.

 

[Artlav]

Did you do what I asked to tape the window of the detector diode? If you did not, the 50Hz is likely from the light!

Both, actually.
With any mains lights (bulbs and lumenescent ones) i always get noise to one degree or another.
In total darkness there is no noise unless i get right next to a fridge or something.

With a flashlight i get various amount of noise depending on where i stand - away from any wires there is no noise.
It's also light level dependent - if there is no light the noise is tiny, but if i shine a flashlight while in the same spot, it can be +-0.5V at the output.

Weird, i thought incandescent light bulbs didn't pulsate.
Not sure yet if it's a problem or a result.

Everything is battery powered, so as far as i understand there is no reference to cancel it out on the analog side.

50Hz, are you in Europe?

More or less - I'm in the western part of Russia.

 

[Antiphon]

First things first- your diode in the schematic looks reversed. If you are using a one-sided supply this will give erratic results.

 

[dlgoff]

First things first- your diode in the schematic looks reversed. If you are using a one-sided supply this will give erratic results.

The diode is in the correct orientation. This circuit diagram came from the Application Bulletin - DESIGNING PHOTODIODE AMPLIFIER CIRCUITS WITH OPA128 that skeptic2 provided in post #5.

Note also that it's a "one-sided" supply.

 

[yungman]

Both, actually.
With any mains lights (bulbs and lumenescent ones) i always get noise to one degree or another.
In total darkness there is no noise unless i get right next to a fridge or something.

With a flashlight i get various amount of noise depending on where i stand - away from any wires there is no noise.
It's also light level dependent - if there is no light the noise is tiny, but if i shine a flashlight while in the same spot, it can be +-0.5V at the output.

Weird, i thought incandescent light bulbs didn't pulsate.
Not sure yet if it's a problem or a result.

Everything is battery powered, so as far as i understand there is no reference to cancel it out on the analog side.

More or less - I'm in the western part of Russia.

1)Well so much for the idea of power supply. So the unit is totally isolated from everything...it's a handheld? But how do you measure the 50Hz? You must have to hook up the scope ground! Tell me more about your device if you need help as I can't trouble shoot with the info given. How do you digitize and get the result? Is all the digitizing circuit in the same unit and all power by battery?
2) Want to verify that if you tape the diode, you don't get any noise unless you are close to the fridge.
3) If you are in the room that you have lights on, you get 50Hz, but if you leave the unit in the exact place and turn off all lights, the noise goes away?
4)Then with the unit at the same place as 3) with no lights on, but when you shine a flash light, the noise come back even the room light is off and is total darkness other than the flash light?

1, 2 and 3 points to the 50Hz is from the light source. BUT 4 don't make sense at all. Assuming the flash light is battery powered, you should not have 50Hz.

Take a picture and post it anyway.

I don't think the noise is electrical, but humor me, use some kitchen aluminum foil and cover the circuit and make sure it is grounded. See whether there is any improvement. At least whether you get rid of the noise when you are close to the fridge with the diode taped.

What is your application, do you have to work in the lights? If all else fail and the 50Hz is from the light source that you have to work with, then:

1) can you lower the bandwidth to below 25Hz? If so, at least you can use steep low pass or notch filter the get rid of the 50Hz.
2) If you need to cover to over 50Hz, active cancellation is the way to go. Set up another amp with diode but point the diode at place that you are not measuring but still can detect light and sum.

Tell more info.

 

[Artlav]

First things first- your diode in the schematic looks reversed. If you are using a one-sided supply this will give erratic results.

The diode is in the correct orientation.

I've mentioned it somewhere in the thread- it should have been drawn other way around.
I'm using it in photovoltaic mode, not backwards.

But how do you measure the 50Hz?

In the output.
Most of it seems to come from the lights.
None if the PD is covered, unless really close to the wires or motors of some sort.
None at all in the sun outside, little to none in sunlight inside, as above dependant on the presence of wiring.
So far that's fine for me.

Thanks for help so far, i'll post in details if i run into something unsolvable.

 

[Artlav]

All right, here is the next problem.

I got the 2GΩ resistors, and suddenly all the flickers at the edge of detection became a flood of noise.

The mains hum was omnipresent, as RF.
That was solved by wrapping the box into tin foil, grounded at the common ground of the batteries.

The only noise left now is from the stepper motors.
Details.

The thing is a 60x40x30cm box, with a coordinate table out of two steppers inside. The photodiode is mounted on the caret, the electronics are all around on the walls. The power is from two sets of batteries with common ground. Everything is ruled by an Arduino, with a USB cable going to a laptop.

Inside: http://orbides.1gb.ru/img/innards.jpg
Overview: http://orbides.1gb.ru/img/overview.jpg
The foil around the box is attached to the common ground.
Without it mains hum dominate the signal top to bottom, with it it's not here at all.

The amplifier board is on the side, far from everything else.
http://orbides.1gb.ru/img/amp.jpg

The problem:
With the motors at idle or not powered, there is no noise in the photodiode reading.
Once the motor starts running, there is a flood of white noise (Only one motor runs at a time).
At 50MΩ that noise was undetectable, at 2GΩ it dominates.

What i tried:
-Wrapping the wire to the PD in tin foil and grounding it to the box - no effect.
-Choke on the same wire - no effect.
-100nF capacitors between the motor wires - no effect.

What i can't think of how to do or can't be done:
-Decouple the ground of the motor and the sensor - no conceivable way to sync up the motion and readings.
-Reduce the PD bandwidth to leave the noise out - by now the requirement is >=700Hz, and the noise is all over it.

Any help would be appreciated, if that's not too far off-topic by now.

 

[skeptic2]

Could your amp be picking up noise from the motor through the power supply?

 

[yungman]

I just have time to take a quick look only. This really bring the meaning of one picture speak a thousand words. The first and fatal problem is you did not use a board with ground plane. You need to have a shielded cable from the PD all the way to the ground plane on the board and ground the shield on the amp board. BUT don't ground or even touch the shield on the ground on PD side. I can't see you stepper motor control boards, but don't look like you are using ground plane either. This will also cause noise to emitted out. Always use ground planed boards. When you do pcb, make sure you use ground plane. For simple amp circuit like this, I can easily do it on a two sided pcb so it will be cheaper.

If you have a torroid, put at the power feed to the board and one on the output wires. This will get rid of some noise if it comes in from the output line or the power line.

This is the general bandaid first. I am not surprised you pick up so much noise looking at the picture.

We literally have a metal box for the amp alone. Stepper motor is very noisy!

To be honest, if you have a cable to the PD, and your circuit is so small, I would really consider putting the amp board right on the place with the PD and just have the power feed and the output line running out.

Now you getting to problem that there usually no one magic fix. It is more step by step working on the ground and shielding. Looks like you have a lot of space to mount the board right on where the PD is. You can make the board very small, the size mainly governed by the size of the 2G resistor. Also, I don't see any capacitor on from the power to ground of the op-amp.

 

[Artlav]

Could your amp be picking up noise from the motor through the power supply?

Possibly, but not all of it.
I tried running the stepper from a separate power source, ground and all, and the noise on PD is still here and looks pretty much the same.

Always use ground planed boards. When you do pcb, make sure you use ground plane.

Perfboards is all i have, no custom PCBs.

You need to have a shielded cable from the PD all the way to the ground plane on the board and ground the shield on the amp board. BUT don't ground or even touch the shield on the ground on PD side

A little slower please.
Ground plane is also the the ground to everything else?
The point the shielding is connected to it should be on the amp side, not PD side?
How can it be approximated? A piece of metal under the board?

I'm afraid i don't understand the concept yet.

If you have a torroid, put at the power feed to the board and one on the output wires. This will get rid of some noise if it comes in from the output line or the power line.

So far i gather that most of the noise is RF - running the stepper from a separate power supply does not reduce it noticeably.
I've put the choke on them - three wraps around a 2cm ferrite torroid - no noticeable effect.

I would really consider putting the amp board right on the place with the PD and just have the power feed and the output line running out.

I can put it on the black horizontal caret, but putting it right on the vertical one where PD is would be tricky.
First case would only reduce the cable length by the "shielded" part.

Also, I don't see any capacitor on from the power to ground of the op-amp.

What kind of capacitor should there be?
Tried 1000uF anti-hum one, no changes.

 

[yungman]

Possibly, but not all of it.
I tried running the stepper from a separate power source, ground and all, and the noise on PD is still here and looks pretty much the same.

Perfboards is all i have, no custom PCBs.
Buy copper plated FR4 boards on ebay and build on top of it. Something like this:http://www.ebay.com/itm/5pcs-FR4-Copper-Clad-Circuit-Board-Single-PCB-70-x-90-x-1-5mm-/251002427171?pt=BI_Electrical_Equipment_Tools&hash=item3a70e91723

A little slower please.
Ground plane is also the the ground to everything else?
No, just for the amplifier, it is very small.
The point the shielding is connected to it should be on the amp side, not PD side?
Yes, prefer the ground shield don't even touch the mounting of the PD.
How can it be approximated? A piece of metal under the board?

I'm afraid i don't understand the concept yet.

So far i gather that most of the noise is RF - running the stepper from a separate power supply does not reduce it noticeably.
I've put the choke on them - three wraps around a 2cm ferrite torroid - no noticeable effect.

I can put it on the black horizontal caret, but putting it right on the vertical one where PD is would be tricky.
First case would only reduce the cable length by the "shielded" part.

What kind of capacitor should there be?
Tried 1000uF anti-hum one, no changes.

You should have 0.1uF parallel with a 10uF cap across the power pins of the op-amp. The connection from the PD to the amp board should be completely shielded. With the ground plane board and shielded cable input, try using aluminum foil to cover the circuit. You should see improvement.

 

[Artlav]

No, just for the amplifier, it is very small.

Slower...
Ground as in electrical ground or separate thing to ground to?
I.e. it's a piece of metal that is only connected to the cable shielding, or a piece of metal connected to the ground wire?

Because ground for the amp = ground for the batteries = ground for everything.

 

[sophiecentaur]

Good news, I've figured out where the white noise came from - for the 2+ MΩ tests i put several 1MΩ resistors in series, but the capacitor spanned only the first one!

With that fixed, i got it up to 10MΩ, and things still look good.
Looks like finding these mythical 1GΩ resistors could actually be worth the trouble. :)

Thanks for all the info, that'll take some time to digest.

Have you considered that the internal resistances of the OP amp will be far lower than your suggested 1GΩ resistor? These resistance paths will provide a limit to the gain (being 'in parallel' with any external resistor you add). I should stick to resistors that are available and look at the application notes to see what values are used there. You are unlikely to do better. You won't do better than the open loop gain.
Also, increasing supply volts can worsen the noise performance and hence, the stability.

 

[Artlav]

Have you considered that the internal resistances of the OP amp will be far lower than your suggested 1GΩ resistor? These resistance paths will provide a limit to the gain (being 'in parallel' with any external resistor you add). I should stick to resistors that are available and look at the application notes to see what values are used there. You are unlikely to do better. You won't do better than the open loop gain.

Not sure what internal resistance there is (nothing named so in the spec), but i got a 2.2GΩ resistor across it now, and the sensitivity is between 10 and 50 times higher than with the 50MΩ one, with no extra noise on it's own.
So, it seem to work.

 

[yungman]

Slower...
Ground as in electrical ground or separate thing to ground to?
I.e. it's a piece of metal that is only connected to the cable shielding, or a piece of metal connected to the ground wire?

Because ground for the amp = ground for the batteries = ground for everything.

Ground is electrical ground, the ground you ground the +ve input of the op-amp, or the ground on the cathode of the PD. You need a plane to provide better noise shielding.

 

[Artlav]

Ground is electrical ground, the ground you ground the +ve input of the op-amp, or the ground on the cathode of the PD. You need a plane to provide better noise shielding.

Ok, would mounting a metal plate below the board do it, or there is something more delicate at work?
I can just cover the bottom of the board with tin foil, and route all the ground things to it, is that a proper solution?

In other news - I've grounded the stepper's casing across a 0.1uF capacitor, and the noise roughly halved.
It should also be noted that the noise is less with the caret away from the motor, and more up close.

 

[sophiecentaur]

@artlav
There has to be a resistive path through any device. It is made of real material and definitely not an insulator so pure logic tells you that its resistance is bound to be less than the maximum obtainable in a purpose built 1GΩ resistor. At the very most, it would be limited by the resistivity of the package of the IC but, more likely, by the resistivity of undoped silicon (?).

 

[yungman]

Ok, would mounting a metal plate below the board do it, or there is something more delicate at work?
I can just cover the bottom of the board with tin foil, and route all the ground things to it, is that a proper solution?

In other news - I've grounded the stepper's casing across a 0.1uF capacitor, and the noise roughly halved.
It should also be noted that the noise is less with the caret away from the motor, and more up close.

First of all Sophiecentaur Asked a valid question about the op-amp, way at the beginning, I asked you to look at the other post and use the op-amp the other poster chose. I assumed you followed that. What is the op-amp you are using? You need to get low bias current of 2pA or less or else you are not going to get accurate result. Right now, you are drowned by the noise and you're not going to see the problem, when you fix the noise problem, this is going to show up.Regarding to ground plane, you connect to the signal ground. Sooner or later, you are going to have to have one with ground plane. This is very standard practice. Designing the circuit is the easy part, grounding is where the meet is, this is the hard part of engineering.

If you really don't want to redo the board( which is easy), get a small metal box and put the whole board in. Connect the board ground to the box. When you have the shield cable from the PD going into the box, make sure you ground the shield to the box. Make sure you isolate the other side of the shield(at the PD end) NOT touching any metal conductors. You don't want to connect grounds from other part of the assembly as you might create ground loop. At the output of the box, make the battery ground and the output signal ground the same. Put a toroid on this output wires. With this, you only have one ground connection...from the battery and output side.

Grounding is very important, designing circuits are the easy part. Grounding is where you earn your keep! If you know how to do grounding, circuits from cook books and textbooks usually work! People complained about following the circuits from the books and it did not work...mostly are how they build the circuits...grounding! Op-amp circuits in books and application notes work...If you can keep it from oscillating and keep the noise out. We talk about all sort of designs here, but when the rubber hits the road, it's the grounding!

Don't think "ground", think of this is the return path of the signal and you respect the ground just like you respect the signal. Yes, I am making a big sting about this because I spend my career pretty much specialized in grounding and noise design. All the electronics knowledge are just matter of fact, you just handle it. The grounding is where the arts of electronics. You are dealing with low frequency stuff, as frequency goes up, circuit layout ( grounding), pcb design( grounding) become 40% of the design. If I have to convey the first and most important thing from my 30 years of design experience.....RESPECT THE GROUND!

Remember I asked for the picture, The moment you talk noise, I want to see your ground!

 

[Artlav]

There has to be a resistive path through any device. It is made of real material and definitely not an insulator so pure logic tells you that its resistance is bound to be less than the maximum obtainable in a purpose built 1GΩ resistor. At the very most, it would be limited by the resistivity of the package of the IC but, more likely, by the resistivity of undoped silicon (?).

Makes sense, but I'm not sure what exactly to look for. Does the bias current define internal resistance in some way?

First of all Sophiecentaur Asked a valid question about the op-amp, way at the beginning, I asked you to look at the other post and use the op-amp the other poster chose. I assumed you followed that. What is the op-amp you are using? You need to get low bias current of 2pA or less or else you are not going to get accurate result.

I use AD820, spec says input bias current is 2pA typical.
I haven't noticed you suggesting using the same one he used.
If it's lt1012, then the spec says it have ten times more bias current.

Right now, you are drowned by the noise and you're not going to see the problem, when you fix the noise problem, this is going to show up.

What kind of noise would that be? Without the steppers i don't get any signs of noise.

If you really don't want to redo the board( which is easy)

Not really an issue of not wanting to redo it, but an issue of not wasting the effort for nothing - i want to know what causes what and how it should look like, then redo it properly and permanently.

And i kind of like the idea of making the board as small as possible, putting it into a grounded box with the PD sticking out of that, and mount the whole thing onto the caret without any extra wires. But for that i need to know that i won't need to fiddle with it later.

get a small metal box and put the whole board in. Connect the board ground to the box. When you have the shield cable from the PD going into the box, make sure you ground the shield to the box. Make sure you isolate the other side of the shield(at the PD end) NOT touching any metal conductors. You don't want to connect grounds from other part of the assembly as you might create ground loop. At the output of the box, make the battery ground and the output signal ground the same. Put a toroid on this output wires. With this, you only have one ground connection...from the battery and output side.

Ok, if i understood this right.
The toroid (choke?) separates the ground "domains", so that shielding on the outside is not interfering with the shielding on the amp? Don't really understand how this is working - what is noise?
We want to avoid different parts of the "ground" side of the circuit from having different potentials?

On the attempt:
I've wrapped the amp into duct tape, then into foil, grounded the whole thing to the amp's ground and added a choke at the wires going from it to the battery.
I've also wrapped most of the wire into the foil, not letting it touch any metal parts.

That's what it looks like now:
http://orbides.1gb.ru/orbf/inside_120228.jpg

Apparently, i didn't understood it well, since now i get nonsense instead of noise.
Best described as a threshold - there is less noise, but there is either nothing or a swift climb into something if light is added. Before the "shielding" it was way more gradual.
I.e. it now responds to a very, very narrow range of currents from PD with the same output it had from a much wider range before.

What have i missed?

 

[yungman]

Makes sense, but I'm not sure what exactly to look for. Does the bias current define internal resistance in some way?

I use AD820, spec says input bias current is 2pA typical.
That's good enough.
I haven't noticed you suggesting using the same one he used.
If it's lt1012, then the spec says it have ten times more bias current.

What kind of noise would that be? Without the steppers i don't get any signs of noise.
Anything that is not supposed to be the desired signal is noise. Stepper motor controller, RF transmitter, florescent lights...anything.
Not really an issue of not wanting to redo it, but an issue of not wasting the effort for nothing - i want to know what causes what and how it should look like, then redo it properly and permanently.
It is my advice as a starting point. You want to start with a solid platform. I am sure you can make some ground connection and filtering and make it look better. But if you don't do the ground correctly, any accidental change in other area unrelated to the amp might cause the noise to come back. When I do this kind of design, I put in all the precaution at the beginning, I never have to spend time in fixing noise issue.
And i kind of like the idea of making the board as small as possible, putting it into a grounded box with the PD sticking out of that, and mount the whole thing onto the caret without any extra wires. But for that i need to know that i won't need to fiddle with it later.
There is never a guaranty on that, that's grounding for you. Read the long post I posted before, this is the hardest part. If I have my hands on the system, it would be 10 times easier. Just by looking at the picture is very hard.
Ok, if i understood this right.
The toroid (choke?) separates the ground "domains", so that shielding on the outside is not interfering with the shielding on the amp? Don't really understand how this is working - what is noise?
I am trying to isolate the amp and the diode. The only ground connection is from the battery and the output line. It is important that the PD and the shield of the input cable do not tough any of the metal in the system, that if you disconnect the connector of the battery input and the output line, you measure infinite resistance from the box to the rest of the system.
We want to avoid different parts of the "ground" side of the circuit from having different potentials?
Avoiding ground loop, I don't want any return current from other electronics go through the shield and the ground of the amp. I am trying to design the ground so there will be no other return path from other electronics. That's what cause the noise.

On the attempt:
I've wrapped the amp into duct tape, then into foil, grounded the whole thing to the amp's ground and added a choke at the wires going from it to the battery.
I've also wrapped most of the wire into the foil, not letting it touch any metal parts.

That's what it looks like now:
http://orbides.1gb.ru/orbf/inside_120228.jpg
If you make sure there is no contact to ground, it will do for now.
Apparently, i didn't understood it well, since now i get nonsense instead of noise.
Best described as a threshold - there is less noise, but there is either nothing or a swift climb into something if light is added. Before the "shielding" it was way more gradual.
I.e. it now responds to a very, very narrow range of currents from PD with the same output it had from a much wider range before.

What have i missed?

I really can't answer this as I don't know your circuit. Grounding really don't change your circuit, it only give you noise if you don't do it right.

Could it be your gain is too high so either it is 0 or rail? You get rid of the noise, make sure you have all the connections, make sure the +ve input of the opamp is grounded, no open circuit. Also are you sure you want to have single supply with Vee of the opamp grounded only? I always work with dual supply.

Grounding is very involved, you can write a book just on this. You can read books in Design with EMC. That's the problem with school, all books just put ground as a triangular symbol as if you put the signal there, it will magically go away.

If you build the stepper controller with non ground plane boards, you should redo it with ground plane as bad grounding will cause more emission. Put if power bypass cap like you said one cap on those circuit improve it, do it.

 

[Artlav]

When I do this kind of design, I put in all the precaution at the beginning, I never have to spend time in fixing noise issue.

Easy to do when you know what is at work...

I am trying to isolate the amp and the diode. The only ground connection is from the battery and the output line.

Devil in the details?
I have a hub where all the ground wires come together, and one wire go out to the battery. Is that normal, or should the ground wire from the amp go on a more direct route?

Avoiding ground loop, I don't want any return current from other electronics go through the shield and the ground of the amp. I am trying to design the ground so there will be no other return path from other electronics. That's what cause the noise.

One thing i want to clarify - i tried earlier to run the stepper from a separate power supply, and the noise remained mostly unchanged.
Is that a different kind of beast, or the same currents-going-around kind of thing?

Also are you sure you want to have single supply with Vee of the opamp grounded only? I always work with dual supply.

Can't answer that.
What would having dual supply give me?
Would i have to take care to avoid negative voltage going into the ADC?
If that is related, the spec on the amp claims "True Single Supply Operation".

Grounding is very involved, you can write a book just on this.

As a matter of fact, can you recommend any good books on the subject?

I really can't answer this as I don't know your circuit. Grounding really don't change your circuit, it only give you noise if you don't do it right.

Well, i traced it back to the "shielded" cable, apparently wrapping tin foil around it does something other than just shielding.

Now I'm trying to get the things done in a cleaner fashion.

Can you define exactly what a shielded cable for this kind of connection is?
Am i to look for a two-wire cable with a mesh around them?
Would there be any undesirably large capacitance from the shielded cable?

Then, i redid the board - put the pieces closer together to remove the extra wires, and put all the ground exits onto a piece of aluminium duct tape glued to the bottom.
Below: http://orbides.1gb.ru/orbf/bottom_120229.jpg
Above: http://orbides.1gb.ru/orbf/top_120229.jpg
Does that make sense, or should i settle for a simple metal box?

 

[yungman]

I'll try to do a simple system ground drawing later on.

First I want you to try one experiment, before doing it, I want you to double check all the connection to make sure you have good solid connections and the circuit is working.

Then with stepper motor off, double check there is no noise at the output. Use a stick to touch and move the wrapped cable from the PD to amp. If there is a lot of noise just by moving the cable. You have a secondary problem. I want you to use a stick to make sure your fingers don't create a circuit path. This kind of problem is common as the shield vibrate and induce enough noise to the amp to disturb the reading.

The usual cable we use is RG59 coax cable and just use MHV connector to hook up to the amp. So the cathode just connect to the shield and the anode connect to the center connector.

Using a coax will cut a lot of the problem of noise due to the vibration of the shield. But there still can be a problem. If RG59 is not even good enough, I would strongly suggest you to consider putting the whole amp with the PD. Or else, rigid coax with solid tube as shield would be the next, or the coax with double shield. you don't want neither one.

Try that first. If it is a problem, this is a totally different issue. At the mean time, I'll try to draw the diagram.

I saw the new picture, the ground foil looks good. On the top view, I would suggest you to bend the pin 2 of the opamp up and solder the big resistor and the 100 ohm directly onto the pin. You put the connection onto the board, you can have leakage current that ruin your reading. You might not see it now, but when you clear up all the grounding and shielding problem, it might show it's ugly head. For 1G resistor, it is not go just solder onto the board. You see articles of ground guards that you are not doing. Bent pin 2 up, and solder the two resistor onto pin 2 up in the air.

One more question, Is the PD moving by the stepper motor during operation?

 

[Artlav]

Ok, i'll try it all over tomorrow (GMT+4 here).

This kind of problem is common as the shield vibrate and induce enough noise to the amp to disturb the reading.

Vibration disturbance as in physical vibration transferred to the amp, or vibration of the shield itself making electrical noise?

If the amp is sensitive to vibration then there would be other problems to consider, including for a case of putting it to the shaky place where the PD is.

One more question, Is the PD moving by the stepper motor during operation?

It does. All the motors do is moving the PD around.
Incidentally, there is more noise when the PD is closer to the motor.

 

[yungman]

Ok, i'll try it all over tomorrow (GMT+4 here).

Vibration disturbance as in physical vibration transferred to the amp, or vibration of the shield itself making electrical noise?
I believe it's the vibration of the shield itself making electrical noise.
If the amp is sensitive to vibration then there would be other problems to consider, including for a case of putting it to the shaky place where the PD is.
I don't think the amp care about the vibration, it's the shied.
It does. All the motors do is moving the PD around.
Incidentally, there is more noise when the PD is closer to the motor.

You really have to consider putting the opamp with the PD. Vibration is a big problem with these kind of amp that I never brought up so far. We had to have rigid coax to run from detector to the amplifier module because we had turbo pumps running and they vibrate less than stepper motor. This is going to be a big problem.

Before I was only suggesting about putting the amp with PD for lower induced noise, but you might have vibration problem.

Did you do the experiment with the stick? I bet you'll see noise if you touch it. It should be easy to replace the wires and foil with a coax to test if it is acceptable when motor is running. Cross your finger that it is ok. Make sure the shield and the PD is isolated from all the metal parts.

Remember about bending pin 2 up from the board. This is important as it will give you error sooner or later.

Yes, I delt with all these in my days working with transimpedance amp.

 

[yungman]

This is a very simple grounding scheme, basically I want to make sure there is no ground current from other circuits going through the coax shield of the amp module. It is important the shield and the PD body is isolated from any metal of the carrier or else you form a grounding path.

150297[/ATTACH]"]

This is the most preferred way to power up the amp, through the ADC control board. if you cannot do that, a power cable with ferite toroid is absolutely necessary to break the RF ground loop.

The fact you said noise is still there even if the stepper motor is from a separate supply don't necessary mean a lot as, if you have a common ground and you are not careful on the return path, you can still have problem.

 

[Artlav]

Well, the cable idea have ultimately failed.
Haven't got around to test the touch-it-with-a-stick issue.
Any coaxial cable i can get is too inflexible for the use, and wrap-tin-foil-around solution is not worth a permanent presence.
And even in a static fixture with a coax cable i got quite some noise with motor working.

Eventually i decided to call it, and mounted the amp right where the PD is.
That have effectively eliminated 99% of detectable noise, leaving only some flicker just below the lower edge of the range, and a strip of slightly noticeable one when the caret is right on top of the motor.
Good enough for me.

The idea with soldering the big resistor and the op-amp pin above the board wasn't useless as well - i get somewhat more consistent reading now, and the sensitivity got perceptibly higher.

All in all, thank you for helping.


Some more questions:
I suppose that increasing sensitivity isn't going to go well above 2.2GΩ?
What is likely the limiting factor?

-The resistors are wire wrapped into a coil, which got some capacitance. At the moment the resistor i got seems to have enough capacitance to avoid the need of a separate capacitor in parallel.
The bigger ones are of the same build, so I'm likely going to run into unresolvable excess capacitance (=insufficient bandwidth) pretty soon.

-Other issue is thermal noise, which if i got the math right is at 10^-14A level with 2.2GΩ, and wouldn't catch me for some time yet.

-Any other considerations down there?

The fact you said noise is still there even if the stepper motor is from a separate supply don't necessary mean a lot as, if you have a common ground and you are not careful on the return path, you can still have problem.

Separate as in separate batteries, separate controller, separate laptop. No metallic link what-so-ever.

 

[yungman]

That's very good news. We did everything to avoid vibration all the time, it just skip my mind to ask you at the beginning! Should have thought about it sooner to ask whether you are running the motor while taking data.

About resistor, we use Caddock:

http://caddock.com/Online_catalog/high_resistance/high_resistance.html

2.2G is nothing, we use 10G! Caddock is not wire wound and it does not have capacitance like you described.
In production form, you can use a stand-off place very close to pin 2 and solder things on it. Make sure you clean the flux good with de-flux multiple times. Flux residue is the killer because it form surface resistance and likely to be even lower resistance than your resistor.

Even in your setup, you should clean the area around pin 2 with de-fluxer a few times. Even the dirt between pin 2 to pin 1 and pin 2 is enough to ruin the day for you. It is very important now that you got rid of the original problem. You battle in not over yet. Now you are at the point of getting back to the circuit. Believe me, things will make sense.

But one thing, when you using high value resistor like this, settling time from vibration might still be a problem. I'm afraid that you might still see some error not in form or noise, but settle to the ultimate value.

Can you do an experiment after you get the right resistor? I want you to read the data continuously and then stop the motor. Then read the data after the motor stop and compare to the last read value when the motor is running and see whether there's a big difference. This might not be an accurate test as the value you read the moment before the motor stop might take a little bit of time to settle also, but it should be close. I guess you can make it more consistent by making the light for the detector constant throughout the experiment so the output should be static when the motor is running. If you compare to reading after the motor stop, then you can see the difference in reading between the motor running and motor stop.

This might not help you as there is not much more you can do unless you damp the vibration further. You just need to understand what you are getting into and what to expect.

 

[yungman]

I forgot, if you go higher resistance, you might need a new amp. The bias current is 2pA, it will produce 4.4mV offset. The problem is when the temperature goes up, the leakage goes up a little also.

 

[Artlav]

I'm pretty certain there is no settling issues with the 2.2G - the edges are sharp and where they should be.
Might have some delay or offset, but if there is, it is constant.

However, with either 4.7G as one or 7G as two resistors there is already noticeable blurring as if it's not fast enough.
So, for the moment i'll stick with 2.2G - it's quite sensitive enough for outdoor scenes, and there is no source of low-capacitance gigaohm resistors here that won't need transcontinental shipping.

 

[Redbelly98]

First things first- your diode in the schematic looks reversed. If you are using a one-sided supply this will give erratic results.

The diode is in the correct orientation. This circuit diagram came from the Application Bulletin - DESIGNING PHOTODIODE AMPLIFIER CIRCUITS WITH OPA128 that skeptic2 provided in post #5.

Note also that it's a "one-sided" supply.

I've mentioned it somewhere in the thread- it should have been drawn other way around.
I'm using it in photovoltaic mode, not backwards.

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

Since the photo-generated current in a photodiode is in the opposite direction from what it is for a normal forward-biased diode, the current will flow from the opamp output towards the feedback resistor and then towards the photodiode: i.e. the current direction is downward at the photodiode. This is consistent with the opamp output being at a positive potential relative to the -input (current flows to the left through the feedback resistor), as it must be since you are using only a positive voltage supply to the opamp.

If the photodiode is reversed then it will try to drive the opamp output negative, which is not allowed given the power supply configuration you are using.

 

[yungman]

So it works? remember look for a better amp for 4.7G unless you can put up with the offset error I mentioned.

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.

BTW, the direction of the PD in your drawing is the correct way, or else I would have said so right at the beginning and you won't go this far!