Photodiode Amplifier - BW Problem

Hi!

I'm new in the forum. I was reading several post about Photodiode Amps. I'm working on a project trying to design on of them and I have a BW problem.
I'll appreciate if you can help me or give me some clue.

***
I'm designing a TIA to detect small amplitude sinusoidal signal up to 10MHz, this signal is mounted on a high intensity light source (laser, 3mW/632nm). So basically I have a modulated laser with low modulation index (1/500).

I've simulated and build the circuit below:



The transmittance obtained with TINA-TI (free software) app. was the following:



But when I've measured at the Lab, I found a different behavior. Take a look below:





As you can see, the behavior of the device is different from the simulated one. However if an optical attenuation is implemented, it seems works correctly. This is not due the photodiode, I checked it alone (without amplification) and worked in an acceptable way up to 10 MHz (it is specified for 7ns).

By the way, the set up is the following:


Do you know which could be the problem?
I'm suspecting the opamp goes off ideal condition due the high photocurrent (800uA max, measured with 50ohm Broadband load) but I'm not sure.


Please let me know if you need more detail.

Thanks in advance!
Horacio

ps: I cannot post link yet, so I attached the images. Sorry.

Your feedback circuit doesn't make sense, although I don't think that is your problem.

I would suggest you look at the DC voltage levels before and after the laser is switched on. Look at both inputs and the opamp output. It should help to understand what is going wrong. I will check datasheets later.

Keith

keith1200rs said:

Your feedback circuit doesn't make sense, although I don't think that is your problem.

I would suggest you look at the DC voltage levels before and after the laser is switched on. Look at both inputs and the opamp output. It should help to understand what is going wrong. I will check datasheets later.

Keith

Click to expand...

Thanks a lot for your answer Keith,

The feedback circuit shown is only for simulation purpose. The R2 100Mohm
resistor is not present in the PCB. I added it to the simulation just to
avoid an error from the app. I used a Tee C network in order to achieve a lower
equivalent Capacitor (Ceq=C1.C2/(C1 C2 C3). Actually the Rf is 2.2k ohm.

I will try to check the voltage as you suggested. It's a bit complicated due the device is mounted in the optical system, but I will find the way to accomplish it. Do you suggest something with this? By the way, the signal is not clipping at the output.

Thanks a lot in advance for any other help.
Horacio

I have had a chance to look at the data sheet and there may be other possibilities (I was thinking of a simple overload problem or connection problem, which may still be the case).

The opamp is way faster than I would have thought was necessary for 10MHz bandwidth and you need to be very careful in the layout and construction to ensure stability. Some 100pF capacitors for decoupling would be a good idea - normal 10nF capacitors will not be very effective at 1GHz. A 100MHz opamp would do the job. A FET input isn't required either - you have 1k or 2.2k to provide the bias current for a bipolar opamp.

Look at the output with the fastest oscilloscpe you have to make sure it is not oscillating.

In my simulations of your circuit with 1k it would be very peaky and hence potentially unstable. A transient analysis shows significant ringing.

Keith.

Hi HoracioG,

I didn't check the data sheet in detail - however, I think the opamp chosen is NOT compensated. That means, the voltage gain must not be too small (probably larger than 10). In this context, I doubt if a feedback resistor of 1 k is appropriate.
What about trying a larger one?
LvW

Do you believe that op amp is suitable for driving a 50-ohm load
directly? If you simulated the lumped element load as a lossy
transmission line instead, would you get a more realistic simulation?

Thanks a lot to all for your answers.

keith1200rs said:

I have had a chance to look at the data sheet and there may be other possibilities (I was thinking of a simple overload problem or connection problem, which may still be the case).

Click to expand...

What do you mean with overload or connection problems. Are you refering to PCB issues?

keith1200rs said:

The opamp is way faster than I would have thought was necessary for 10MHz bandwidth and you need to be very careful in the layout and construction to ensure stability. Some 100pF capacitors for decoupling would be a good idea - normal 10nF capacitors will not be very effective at 1GHz. A 100MHz opamp would do the job. A FET input isn't required either - you have 1k or 2.2k to provide the bias current for a bipolar opamp.

Click to expand...

Yes, the opamp is faster for a 10MHz BW, but I'm thinking to make this design work and then use the same with a faster photodiode. In fact I'll be glad it works with Vishay BPW24R (tr=tf=7ns ~50MHz BW).
I choose this opamp because two reasons:

- High gain bandwidth product to recover the small modulating signal (1/500 modulation index) from DC to 10 Mhz (100MHz if possible).
- The FET input because I need a low input bias current and low noise.

Would you please tell me if my criteria is still wrong?

I've tried to apply RF PCB design concept and follow TI datasheet recomendations.

keith1200rs said:

Look at the output with the fastest oscilloscpe you have to make sure it is not oscillating.
In my simulations of your circuit with 1k it would be very peaky and hence potentially unstable. A transient analysis shows significant ringing.

Keith.

Click to expand...

I only have a 60 MHz BW oscilloscope, I'll check if can find a better one.
Could you please tell me which program did you use for simulation? Did you used the last revision of the Opamp model?
Thanks a lot for your help.

LvW said:

Hi HoracioG,
I didn't check the data sheet in detail - however, I think the opamp chosen is NOT compensated. That means, the voltage gain must not be too small (probably larger than 10). In this context, I doubt if a feedback resistor of 1 k is appropriate.
What about trying a larger one?
LvW

Click to expand...

Thanks for your recomendation. Actually I'm working with a 2.2kohm resistor.
I guess I cannot use a bigger resistor due the max value of the photocurrent. 800uA. It will saturate the opamp DC gain.

dick_freebird said:

Do you believe that op amp is suitable for driving a 50-ohm load
directly? If you simulated the lumped element load as a lossy
transmission line instead, would you get a more realistic simulation?

Click to expand...

The datasheet recomends a shunt resistor around 60 ohms for a capacitive load of 20pF, corresponding to the oscilloscope input stage in my case.

I think the simulation is more realistic but with some limitations of course.


I will measure the DC voltages and let you know about that.
Thanks a lot again.
Horacio

I mention overload problems because your performance is not consistent with amplitude. It could be just a slew rate limitation but you really ought to check the DC levels to make sure the circuit is operating correctly - it seems a logical first step in checking a circuit.

Connection problems? Yes, PCB issues. The circuit doesn't work as expected to a connection error or bad connection is a possibility.

I don't see why you need 1.6GHz amp to recover a "small" 10MHz signal. A 1GHz amplifier won't make the 10MHz signal any larger than a 100MHz amp.

Why do you need low input bias current? You have 1k or 2.2k feedback so you could tolerate 500nA or 1uA of bias current.

Low noise? FET opamps don't actually have very low noise voltage. Where they excel is low current noise. However, with 1k transimpedance that is not likely to be a huge issue. The OPA657 has 4.8nV/rt(Hz) but there are plenty of opamps down to 1nV/rt(Hz). You do need to watch out for the current noise contribution and input bias current though.

I use SIMetrix for simulation and yes, I used the latest model from TI - revision E, which they cleverly put in a zip file called revision C.

You say you have applied RF PCB design concepts but only include one decoupling capacitor. I use between 2 and 4 depending on the circuit. You can buy "microwave" 100nF capacitors which work well up to GHz but they are not cheap, nor easy to get hold of so I still use multiple values starting at 47pF or 100pF. I have ample measured evidence to show that they ARE necessary.

I have had my fair share of oscillating amplifiers and without test equipment to see it, it can be difficult to diagnose. A couple of clues are: strange and unexplained DC operating points (they are not really DC - you just cannot see the oscillations) and performance changes with amplitude. Changes in performance when you probe the circuit can also be a clue, although that can often happen with a circuit that works properly.

I have had spurious oscillations up to several GHz. That shouldn't be the case with the OPA657, but it could still be >100MHz. If you have access to a spectrum analyser, that may help.

Keith.

HoracioG said:

Hi!

I'm new in the forum. I was reading several post about Photodiode Amps. I'm working on a project trying to design on of them and I have a BW problem.
I'll appreciate if you can help me or give me some clue.

***
I'm designing a TIA to detect small amplitude sinusoidal signal up to 10MHz, this signal is mounted on a high intensity light source (laser, 3mW/632nm). So basically I have a modulated laser with low modulation index (1/500).

I've simulated and build the circuit below:

View attachment 49779

The transmittance obtained with TINA-TI (free software) app. was the following:

View attachment 49783

But when I've measured at the Lab, I found a different behavior. Take a look below:

View attachment 49780
View attachment 49782


As you can see, the behavior of the device is different from the simulated one. However if an optical attenuation is implemented, it seems works correctly. This is not due the photodiode, I checked it alone (without amplification) and worked in an acceptable way up to 10 MHz (it is specified for 7ns).

By the way, the set up is the following:
View attachment 49781

Do you know which could be the problem?
I'm suspecting the opamp goes off ideal condition due the high photocurrent (800uA max, measured with 50ohm Broadband load) but I'm not sure.


Please let me know if you need more detail.

Thanks in advance!
Horacio

ps: I cannot post link yet, so I attached the images. Sorry.

View attachment 49779View attachment 49780View attachment 49781View attachment 49782View attachment 49783

Click to expand...

hi
put ur design and i'll solve ur poblem

commeng said:

hi
put ur design and i'll solve ur poblem

Click to expand...

He has already posted his design

So what is the solution?

Keith.

yes i know but i want him to post the modelling coz i want to run it firts, coz really i am so busy to moddel it again

commeng said:

yes i know but i want him to post the modelling coz i want to run it firts, coz really i am so busy to moddel it again

Click to expand...

Hi Commeng,

Thanks a lot for your help. Please see the attached circuit to this reply. You can download the TINA-TI free version here: TINA-TI Free
Attachment: View attachment PDAMP_V2.5_20-Sep-2010_TI.rar

Keith,

Thanks a lot again for your answer. Please see my comments below.

keith1200rs said:

I mention overload problems because your performance is not consistent with amplitude. It could be just a slew rate limitation but you really ought to check the DC levels to make sure the circuit is operating correctly - it seems a logical first step in checking a circuit.

Click to expand...

I think you're right. I have measured the following voltages:
Without Laser (Turned off)

Vout: -2.75mV
NC(Pin1): - (Disconnected from GND)
V-(pin2): 0 V
V+(pin3): 0 V
-Vs(pin4): -4.97 V
NC(Pin5): - (Disconnected from GND)
+Vs(pin7): 4.97 V
NC(Pin8): - (Disconnected from GND)

With Laser (Turned on)

Vout: 1.39V
NC(Pin1): - (Disconnected from GND)
V-(pin2): 0 V
V+(pin3): 0 V
-Vs(pin4): -4.67 V
NC(Pin5): - (Disconnected from GND)
Out(Pin6): 2.78 V
+Vs(pin7): 4.96 V
NC(Pin8): - (Disconnected from GND)

keith1200rs said:

Connection problems? Yes, PCB issues. The circuit doesn't work as expected to a connection error or bad connection is a possibility.
I don't see why you need 1.6GHz amp to recover a "small" 10MHz signal. A 1GHz amplifier won't make the 10MHz signal any larger than a 100MHz amp.
Why do you need low input bias current? You have 1k or 2.2k feedback so you could tolerate 500nA or 1uA of bias current.
Low noise? FET opamps don't actually have very low noise voltage. Where they excel is low current noise. However, with 1k transimpedance that is not likely to be a huge issue. The OPA657 has 4.8nV/rt(Hz) but there are plenty of opamps down to 1nV/rt(Hz). You do need to watch out for the current noise contribution and input bias current though.

Click to expand...

Please correct me if I'm wrong:
1.- 1GHz amplifier should have more Gain-Bandwidth Product than a 100 MHz one.
2.- I need low input bias current because the index modulation is very low (1/500 - 1/1000) so the modulated current could be around 1uA with a mean intensity value of 1mA. So I need a lower input current than the modulated current.
Could you please recommend me another opamp you consider better for this application?

keith1200rs said:

I use SIMetrix for simulation and yes, I used the latest model from TI - revision E, which they cleverly put in a zip file called revision C.

Click to expand...

You're right. .
I have downloaded the free SIMetrix and simulated the circuit but couldn't find what you said. Could you please help me to obtain the same result or maybe send me the file to test by myself?

keith1200rs said:

You say you have applied RF PCB design concepts but only include one decoupling capacitor. I use between 2 and 4 depending on the circuit. You can buy "microwave" 100nF capacitors which work well up to GHz but they are not cheap, nor easy to get hold of so I still use multiple values starting at 47pF or 100pF. I have ample measured evidence to show that they ARE necessary.

Click to expand...

Well, actually I said "I tried to apply RF PCB design concepts". I sorry but I don't have too much expertise. I'm trying to learn from people like you. I really appreciate your help.
You're right with the capacitors. I just only followed the recomendations from the datasheet that says to use a 0.1uF capacitors nearest the Power Supply pins.
Anyway I don't have too much space on the PCB and I thought that these would be enough.
Regarding the "microwave" capacitors, I think they're a good option, but as I'm in Argentina and is not so easy to find them independently of the price.

keith1200rs said:

I have had my fair share of oscillating amplifiers and without test equipment to see it, it can be difficult to diagnose. A couple of clues are: strange and unexplained DC operating points (they are not really DC - you just cannot see the oscillations) and performance changes with amplitude. Changes in performance when you probe the circuit can also be a clue, although that can often happen with a circuit that works properly.

I have had spurious oscillations up to several GHz. That shouldn't be the case with the OPA657, but it could still be >100MHz. If you have access to a spectrum analyser, that may help.

Keith.

Click to expand...

I think you're right. I think there is a signal around 150MHz present at the output. I couldn't check it correctly due the BW of the Oscilloscope. I will try to check with a Spectrum Analyzer.
Later I made another device, just with another photodiode (with similar Cd) and I saw clearly a high frequency signal at the output without input signal. I think you're right when you said it's oscillating.
In that case, what should I do? Should I test with a trimmer capacitor at the feedback loop in order to adjust the compensation?

Thanks a lot for your kind help!
Best regards
Horacio

Here are the simulation results I get with 1k. Be careful with the Pspice model from TI - they have the pins in a non-standard order.

With 2.78V on the output, assuming 2.2k feedback you have nearly 1.3mA of DC photocurrent. That shouldn't be enough to cause a problem but the loaded output swing of the opamp is only +/-3.3V typical, so you are close. It might be worthwhile working with a lower light level or lower transimpedance first in case that is the problem.

I think you are misunderstanding the way a transimpedance amplifier works and the important characteristics. A low bias current is important because transimpedance amplifiers frequently have very high transimpedances and with a high bias current they would supply saturate trying to supply the opamp bias current. Other than that, there is no benefit. A lower bias current does not amplify small signals any better.

TI applications notes sboa060 and sboa055a might be worth a read.

A 1GHz opamp will not make a 10MHz signal any large than a 100MHz one (dependingon the transimpedance). As an example, here is a 28MHz opamp with 2.2k feedback and 9MHz bandwidth. I am not saying it is the best one to use - just illustrating that 1GHz isn't necessary. Ideally you would want your signal to not be at the 3dB point.

For capacitors, just stick to normal ones but you do need to use something small such as 100pF when working with 1GHz opamps, even if the data sheet shows simply 100nF. I don't know why they do that - a typical 100nF capacitor will resonate at a few MHz.

If the circuit is oscillating then it could be tricky to track down. I think I would look for an opamp with maybe 100MHz bandwidth - it will be a lot easier to work with.

Keith.

Just a short comment on the original circuit. I fear, it's a stupid idea to use a decompensated OP with a capacitive voltage divider instead of universal compensated type with a single integration capacitor. I'm rather sure, that it doesn't bring any benefit, but a lot of parasitic circuit effects. As the datasheet suggests, the OPA657 can be useful for a highspeed TIA, if the source (photodiode) capacitance already allows a gain > 10.

One other comment - a TIA essentially runs with the opamp in unit gain mode. This assumes everything else is perfect (no photodiode capacitance, no input capacitance, no leakage resistance) but it means that a 10MHz opamp in a perfect system would have a 10MHz bandwidth in a TIA. I have attached an example.

Also, here is the SIMetrix file with the OPA657. I have imported the model and corrected the pin order. You can see the model if you hit F11 when looking at the schematic in SIMetrix.

Keith.

hi
i installed the software and i told u i have no time to rebuild the circuit so plz if u r free build it and attach it. tnx and sorry

commeng said:

hi
i installed the software and i told u i have no time to rebuild the circuit so plz if u r free build it and attach it. tnx and sorry

Click to expand...

Hi commeng,

I have attached the file on a preview reply. I attach it here again. Please see below.
Attachment:

FvM said:

Just a short comment on the original circuit. I fear, it's a stupid idea to use a decompensated OP with a capacitive voltage divider instead of universal compensated type with a single integration capacitor. I'm rather sure, that it doesn't bring any benefit, but a lot of parasitic circuit effects. As the datasheet suggests, the OPA657 can be useful for a highspeed TIA, if the source (photodiode) capacitance already allows a gain > 10.

Click to expand...

Hi FvM,

Thanks for your comments.
Let me tell you I have build another device with universal capacitor (Rf=2k2, Cf=1pF) and the same result was seen. I agree with you about the Tee Network, but I'm trying to found a solution to this behavior.
What do you mean with " if the source (photodiode) capacitance already allows a gain > 10." Do you mean low capacitive photodiodes?

Regards,
Horacio

---------- Post added at 04:32 ---------- Previous post was at 04:30 ----------

keith1200rs said:

One other comment - a TIA essentially runs with the opamp in unit gain mode. This assumes everything else is perfect (no photodiode capacitance, no input capacitance, no leakage resistance) but it means that a 10MHz opamp in a perfect system would have a 10MHz bandwidth in a TIA. I have attached an example.

Also, here is the SIMetrix file with the OPA657. I have imported the model and corrected the pin order. You can see the model if you hit F11 when looking at the schematic in SIMetrix.

Keith.

Click to expand...

Hi again Keith,

Many thanks for all the information you sent. I will check/read everything, do the measures with the spectrum analyzer and then post a reply again.
I hope have clearest ideas then.
Best regards and thanks again!
Horacio

"Optical Sensing Techniques and Signal Processing" by Tudor E. Jenkins, ISBN 0-13-638107-3 is a useful read, although it is out of print I think. You may be able to get a copy from your library depending on where you are.

Keith.