OPA818: Ripples and lower bandwidth than designed/simulated for TIA configuration

Hello Dharani,

  Thank you for following our guides and using our calculators for your design!

  Really good question, that is very strange that the bandwidth changes drastically with supply voltage, I would actually expect it to have no change or minimal degradation in the opposite direction (with lower voltage range compared to higher). 

  Would it be possible to share your schematic especially including the diode configuration + its reverse bias voltage value? And, a snippet of the layout of the device and surrounding components to check the grounding. 

   Also, I know isolation resistor is recommended in the datasheet, but also try replacing it with a solder short or 0Ohm instead. 

Thank you,

Sima 

Hi Sima,

I appreciate your response, I will try to remove the isolation resistor and see if that solves the issue. I'm using a reverse bias voltage of around 63V and my schematic looks like this :

The layout is:

This layout has a voltage divider providing 0v to input+, but I have also directly grounded the input+ via to the ground plane next to it to show the first waveform I attached in the question. The ground clearance for the feedback traces and everywhere is 0.5mm. I have the GND vias placed for heat dissipation and so that unnecessary ground loops aren't formed . 

Look forward to your opinion and advice on this!

Kind Regards,

Dharani

Hi Dharani, You listed the bias on the APD but not what you think its capacitance is under that bias. Can't go foward without that. 

Hi Michael, thanks for the response. The capacitance is 2.2pF for that bias

Ok, running that through a butterworth design solutions suggests a 0.15pF feedback C and 66MHz resulting BW. 

Those tee networks are tricky, typ surface mount R's have 0.18pF parasitic, why don't you removed those T network Cs and check if the parastic Cf on the resistor gives you what you want. 

Thanks Michael,  is this snapshot from a different sheet than the TIA calculator that I used? Also, out of curiosity, what are the kind of issues seen with tee networks usually?

I was wondering if a T network can actually cause the sort of weird overshoots/undershoots which seem to vary with supply voltage and input+ supply that I am seeing. Thanks again!

I had developed all those tools originally for myself, this is the active one I still maintain and have open most of the time

Thanks Michael, I will eliminate the T-network and see if that helps out. I will keep you posted on this thread!

Hi Michael and Sima,

Thanks for your inputs, I did try both the solutions:

1) Removing the isolation resistor (10R): No change in the output, the ripples are still present and the bandwidth is much lower than expected (input+ node of the OPA818 is connected to GND):

2) Removing the feedback network: Seems to cause the OPAMP to oscillate and causes major overshoot/undershoot (input+ node of the OPA818 is connected to GND):

Any further comments/inputs/suggestions will be appreciated!

Hi, this issue is still unresolved, would really appreciate any input/advice!

Hello Dharani,

  I apologize for the long delay here. Thank you for trying out the suggestions. One more suggestion, would you happen to have around 150-250 fF capacitor to try out as a singular feedback capacitance instead of the capacitive network since removing it led to oscillation? 

  Thinking about this more, an explanation I could come up with is that a combination of lower supplies + keeping the same output, might be pushing the amplifier a bit more putting it into a slew boost mode which would decrease rise-time and fall-time and then cause those overshoot/undershoot you are seeing. I will check this with our designer and get back to you by tomorrow.

Thank you for your patience! 

Sima 

Hi Sima,

Thank you for your response! Sure, I will give that a go. Also, I wanted to understand more about what sort of effect the T-network has and why it is not recommended. How do you recommend connecting the OPAMP supply? I have placed two de-coupling capacitors of 10nF close to the chip, do you suggest anything else to reduce unwanted effects on the circuit performance?   

Look forward to hearing back from the designer and hope to get more information!

Kind Regards,

Dharani

Hello Dharani,

  Thank you for giving it a try! I have also suggested capacitive t-networks in the past, like the datasheet suggests, to realize very small capacitive values; however, sometimes this does bring a new set of issues due to the additional parasitic in the feedback or introduction of resonance. To avoid that, it is better to use a single capacitor if it is above 100fF. 

  I asked my team today about this issue, and these are some of the suggestions:

1. What you exactly brought up about decoupling capacitors. In the schematic it only shows one decoupling capacitor for each supply. When you say you have two de-coupling capacitors, does this mean one for each supply or two for each supply? If it is the first case, would you be able to keep the 10nF for each supply, and then add one or two larger capacitors, at 0.1uF - 0.22uF and 1uF - 2.2uF to each supply. And a differential capacitor between the supplies. The larger capacitors does not have to be directly close to the chip as the 10nF. Something similar to what is done in the OPA818EVM: 
   
                                                                                                                                                
   
   Or, if possible to implement a pi-filter at the supplies (Low Pass PI Filter Link). I have seen in previous designs where if the power supply is a bit noisy, it could cause ripple effects in the frequency response of the amplifier. Maybe the higher the supply voltages, the issue becomes more apparent which could be the cause of the bandwidth reduction. 
   
    E2E Thread: https://e2e.ti.com/support/amplifiers-group/amplifiers/f/amplifiers-forum/1014831/opa855dsgevm-opa855-as-high-speed-tia-oscillating  
    E2E Thread: https://e2e.ti.com/support/amplifiers-group/amplifiers/f/amplifiers-forum/1203828/opa657-opa657-as-tia-output-noise-osillates
   
   
2. Would you be able to match the output Vpp at +/-5V. It looks like It is smaller value than measured at +/-3V. 

Thank you,
Sima

Hi Sima,

Thank you very much for the detailed response! Before I send you the waveforms for 2. as requested, just wanted to respond to the first point regarding the decoupling caps:

I have one for each supply (10nF) placed very close to the chip. I will add the extra capacitors as mentioned, thanks for the suggestion. Should the capacitor between the two supplies (differential cap) be in the bottom layer? (mine is a 2 layer design) and also should I add the ferrite beads as well?

Hello Dharani,

  I would include the ferrite beads if it is possible. This will create a LC filter which will help filter out the high-frequency noise. For the differential capacitor, if you are able to add it to your circuit, this can be on either layer. We mostly focus/recommend keeping the power supply filtering as close as possible to the amplifier. Here are more resources on these concepts:

1. E2E Thread on power supply filtering
2. PI filter/LC Filter
3. PI filter

  Also, as a note, we usually do recommend a four layer design, but it looks like your PCB doesn't have too many components; therefore, you could have enough PCB area for power supply planes without many disconnections, and a dedicated ground plane. 

  Let us know how it works out with the decoupling changes on your current PCB! 

Thank you!
Sima

Hi Sima,

Thank you for your response and tips! I really appreciate it. Just to follow-up with your earlier request from the designers , here are the two waveforms with +/-3V supply and +/-5V supply to the OPA818 with almost equal Vpp (it is hard to regulate the laser better than this). This is the OPA818 with the non-inverting input set to 0V. I will check with the new PCB providing more decoupling capacitors and see if it helps. I am still concerned about why when the non-inverting node of the OPA818 is grounded, we see completely different results.

Another question: What type of probe and oscilloscope impedance would you recommend to see the output with minimal probe and loading effect? The output is DC coupled and should be capable of driving a 500ohm load and buffered by a 33ohm resistor. 

Thanks a lot and look forward to your advice/input.

Hello Dharani,

  Thank you for sharing these results! I wasn't in the office today, I will be for the rest of the week. I will confirm with the team, but it does look like increasing supply voltages causes higher overshoot/undershoot which suggests instability or power reflections. The grounding could suggest that since there isn't a dedicated ground plane, it could cause increased ringing. 

  That is a good question, since that is one source commonly linked to instability. When probing the output, we recommend using a probe that is very low in capacitance, and probing after a small isolation resistor which you have done with the 33 ohm resistor. Also, the ground clip on probe should be short. 

  Here is another resource on probing and decoupling + layout: https://e2e.ti.com/support/amplifiers-group/amplifiers/f/amplifiers-forum/1121367/ths3092-harmonics-everywhere-in-the-circuit

Thank you,
Sima 

Thanks Sima, the bottom layer of the PCB is almost completely grounded to act like a dedicated ground plane (apart from one or two traces away from the OPA818):

Thanks for the tips on the probing and decoupling practices, will keep you posted if the decoupling capacitors help! Is it recommended to probe the output through an SMA cable into a 50ohm oscilloscope? 

Hello Dharani,

  Sorry, I forgot to ping this thread earlier. That is correct, it is recommended to measure high speed applications through SMA cable with traces on board matched to 50Ohm to a 50Ohm oscilloscope. This is because it avoids power reflections, and instability caused by a probe due to the capacitance. 

 You are right, those two traces shouldn't affect it, that is strange this issue is occurring on a dedicated ground plane. I hope the decoupling capacitors helped with your issue!

Thank you,
Sima 

Hi Sima,

Thanks for your response. Adding the decoupling capacitors and removing the feedback capacitance did help the design quite a bit, and I was able to get a good bandwidth of around 60MHz with the parasitic capacitance and a gain of 22k. However, the OPAMP seems to have heavy overshoot when I increase the input optical power(to check how to OPAMP saturates):

This is how it looks otherwise (at lower input optical powers):

Is the parasitic capacitance too low and maybe pushing the GBP of the OPAMP or do you think its any other reason? Would really appreciate any advice or input you may have!

Kind Regards,

Dharani

Hello Dharani,

 Thank you for trying out the suggestion, it does look much better!

 Your assumption is correct, increasing the input optical power looks like it decreases your fall/rise time compared to your second scope shot 

  Your design is right at the edge of 60MHz with the input capacitance (including amplifier's internal cap) and feedback resistance. The first scope shot is showing 3-4ns of rise/fall time which would even suggest it is producing higher speeds than a closed-loop bandwidth of 60MHz. You would need a higher GBP amplifier such as our OPA85x family. I would recommend trying out our OPA858 in your next design. Or you can play around with your gain by lowering it, then add a second stage for additional gain if noise is not too critical in your design. If you need a differential output, you could add a FDA (Fully differential amplifier) which you can also incorporate additional gain.

Thank you,
Sima 

Thanks Sima, I've started looking into the OPA858 but just wanted to know how I can measure OPA818 or OPA858's noise performance over temperature? The chip has to show good performance up to 100C, with the APD we are using being cooled by a TEC, so its just the TIA noise that is being evaluated. Have these measurements been done by TI and any idea how I can test it?

Any advice would be appreciated!

Thanks Sima, this is very helpful. Will have a look and let you know if I have any more questions!