OPA858: Another Photodiode Amplifier for Short (1-2ns) Laser Pulse

Hello Raymond,

   Thank you for using our reference designs and app notes to design your TIA! Would you also be able to share your input pulse current and rise time. I see the input source parameters, but not too sure which order they are in. Regarding the output impedance matching, you could have the 50 ohm output and increase the gain to double (2k) to accommodate that loss there.

   The problem probably isn't your feedback resistance being too low, but the input capacitance being a bit low for this amplifier. At these low levels, it is hard to get the feedback capacitance at around 0.34pF (like you calculated). The capacitive T network does help with this, that is a good design choice, but there are parasitic capacitance on board that might increase this feedback capacitance a bit as well. This might have to be tuned.

    Also there is another possible problem area as well which is the one inch trace between the photodiode and the input of the amplifier (thank you for including this detail!). Each inch of trace will give an estimate of 1-3pF if there is a ground and power plane present underneath. Are there internal plane cutoff at the input and output of the amplifier? Would you be able to share your layout here; if not, you can send it to over private message on E2E as well.

Thank you,

Sima 

Hi Sima,

Apologies, the parameters for the pulse is 80uA, 200ps rise time, 200ps fall time, 2.4ns on time. The frequency is low so you can think of this as a one shot pulse. In simulation, 80uA works great, but in application, I think the current is a lot higher. I actually don't have an exact current number because my output is saturating right now. I decreased the feedback resistance from 1k to 453ohms in hopes of fixing this problem, and it helps, but the oscillation (scope capture above) gets worse even after tuning the capacitors with the GBWP calculator. I believe you're correct, the added parasitic capacitance was not accounted for during simulation or in the calculator.

Regarding increasing the gain to accommodate for the loss, I think the problem there is the peak output of the OPA858. The OPA858 starts clipping for me around 1.5V peak. From what I can see, the matching divider network (169 and 71.5ohms) is acting as a voltage divider.

That's a good idea. Is there a recommendation on how best to tune to feedback capacitance? Given standard values I have, I can decrease the capacitance, but is overcompensating for the parasitic capacitance better? Or will that introduce more ringing?

I have a ground pour below the top layer. I followed the recommendation and removed all the copper underneath Pins 1 and 2. Power is on layer 3. Layer 4 is mostly ground pours as well.

Thanks again!

Raymond

Hi Raymond,

hmm, doesn't look very well:

raymond_opa858.TSC

Kai

Hi Kai,

Thanks for putting together the TINA simulation! The 690pF tee capacitance is the best I could do in the lab. Based on the GBWP calculator and the 16pF in the tee, that value really should be 721pF to get the 0.34pF that was estimated. I placed that in and it made the oscillation even worse. To put some more damping in, that capacitor had to decrease to under 250pF.

I modified the board to the below configuration and this is the scope capture I get when I power my board.

Raymond

Hi Raymond,

unfortunately, decreasing the cap to down to 100p will -as unwanted side effect- degrade the bandwidth:

Btw, what bandwitdh do you desire?

Kai

Some more simulations:

raymond_opa858_1.TSC

Kai

A feedback capacitance of 400...500f is not all too difficult to achieve. Just put two 1p caps in series and remove the solid ground plane below these caps.

And to decrease the unwanted stray capacitance of 1k feedback resistor put two 500R resistors in series.

By the help of these "tricks" I'm sure you can omit this "nasty" T.

Kai

Hi Kai,

Thanks for running these simulations. My requirement is 1GHz, but I accounted for the bandwidth degradation given the feedback value's network. I'd be okay if I can get >500MHz BW out of this.

Raymond

Hi Kai,

I went ahead and modified my board so I'm not longer using the capacitive-tee network. I have two 1pF capacitors in series along with the 1k resistor. I am not able to remove the solid ground plane below the caps, and I cannot use two 500R resistors as I don't have the pads for them. This is what I see:

I lose out on some BW, but should I increase the feedback capacitance until the oscillation goes away?

Raymond

Hi Raymond,

the OPA858 is decompensated and is really really fast. It does not forgive even the least mistake. To demonstrate how sensitive this OPAmp is to layout mistakes see these two simulations:

All looks fine so far. But let's move the photodiode one inch away from the input of OPA858. One inch -or about 25mm- copper trace not routed over a very close ground plane means an inductance of about 25nH following the 1nH/mm rule of thumb:

So please move the photodiode back to the OPA858 as much as possible

If there's absolutely no way to do so, you can insert a small damping resistance:

Another issue is what you do with the output of OPA858 during your measurements. Can you please post a schematic showing how you connect the scope probe to the output of OPA858?

By the way, I suggest such a ground spring for doing HF measurements:

Kai

Hello Kai,

Yup, I'm beginning to see how unforgiving this part is to design with. I can adjust the feedback resistance and capacitance to a value that has very low ringing, but I the output becomes underdamped. As the board is already printed, I am not able to place a damping resistor where you have shown sadly.

The board I'm testing with is fairly similar to the OPA858DSGEVM. I have a SMA connector on the output that I'm feeding directly into my oscilloscope. That was actually another question I had. I know the EVM and the datasheet both reference a matching resistor network on the output of the opamp, but that's attenuating my peak output voltage. Is it okay if I omit the 169 and 71.5ohm resistors?

Thanks for the advice on the ground spring!

Raymond

Hi Raymond,

if your scope has 50R input impedance, then you should remove the 71R5 resistor and replace the 169R resistor with a 49.9R resistor. Also, use a good 50R cable. I prefer semi-rigid cables. They are the best. Yes, I know they are expensive

Kai

Hello Kai,

Thanks for clarifying that. This is a silly question, but I assume it's expected that my output voltage read by the scope will drop by half once I replace the 71R5 resistor with a 49.9R resistor? Is that just a trade off I have to live with?

Raymond

Hi Raymond,

I'm confused now. Why would you want to replace the 71R5 resistor with a 49.9R resistor?

Please remove the 71R5 resistor and replace the 169R resistor with a 49.9R resistor.

Let's first make the OPamp work properly. In a later step we can discuss the unwanted dampening :-)

Kai

Hello Kai,

Sorry for the confusion, that was a typo. I meant to say the 169R resistor was replaced with a 49.9R resistor. The 71R5 resistor is unpopulated (captured in TINA sim screenshot below). If I was reading a peak output before of 1.5V, now I'm reading a peak output of 0.75V.

I modeled the above sim using the layout below which is what my current layout is based from. I managed to get the values above to run properly in simulation and the lab, but the system appears to be underdamped. Not sure if that's a consequence of using a lower feedback resistance? I also noticed if I even tweak the capacitance a few fF in either direction, the oscillation returns.

...

Raymond

1447.raymond_opa858.TSC

Hello Raymond,

  From your first reply, it seems like replacing the feedback resistance from 1kOhms to 500Ohms seemed to help a bit. This suggest there is above 300pF capacitance at the feedback of the amplifier. Can you try taking off the feedback capacitance entirely? If this doesn't help, try also taking off the Riso at the input of the amplifier as well. 

   Thank you for the all of the simulations Kai, they are very helpful. As Kai mentioned, there will be inductance between the photodiode and the input of the amplifier. Another solution might be to place a small 0.3-0.5pF capacitance at the input of the amplifier (close to the amplifier). This can help solve the problem since the photodiode capacitance might become isolated from the amplifier at high frequencies due to the inductance. 

  Having a low feedback resistor does make the part more prone to oscillation, we can move to this solution next if the above does not work. It is best for now to match input/output impedances for high frequency parts. Sadly, high frequency and decompensated parts are difficult to stabilize. Once we can stabilize it, we can tune the feedback resistor to avoid saturation. 

Thank you,

Sima

Hi Sima,

Oh I never tried using a 500R feedback resistance. I found that even if I could reduce the oscillation with 500 ohms, the output signal once excited has a significant amount of ringing. It's extremely underdamped, and changing the feedback capacitance slightly in the wrong direction would cause it to oscillate again.

I can try putting a capacitor at the input of the opamp, but I'd have to run a wire to a ground pad. Would that cause any issues?

Thanks!

Raymond

Hello Kai,

  As an update due to some of this conversation was taken offline. After further inspection of the layout, it looks like adding an input capacitance won't help in this case due to the inductance occurring in the feedback loop. As pointed out in the OPA858 datasheet, it is best for longer distances for the feedback loop to be closed near the APD rather than near the input of the amplifier. This will avoid the issue of isolating the input capacitance introduced by the APD from the amplifier at high frequencies due to the parasitic inductances. However, it looks like at distances maybe half an inch to an inch of trace length, this method loses its benefits and leaves too high of parasitic inductance in the feedback of the amplifier. Thank you Raymond for following the datasheet closely; unfortunately in this case it did not help with oscillations. Suggested increasing damping resistor as you showed in your simulations. Hopefully they will get rid of most of the oscillations, but will probably be some ringing. 

Thank you,
Sima

Thank you for the feedback, Sima

Raymond, can you please post a photo of your setup showing the APD and OPA858?

Kai

Hello Kai,

Hopefully this helps.

Raymond

Well, Raymond, that's way too far away from the OPA858 for my taste...

This might properly work with a 2MHz OPAmp but not with a 5.5GHz rocket.

Can you please post a full schematic, showing the Vbias generation as well?

And please another photo from the whole circuit arround the OPA858

Kai

Keep in mind, that with improper Vbias decoupling the photodiode is twice the distance away from the OPA858, effectively spoken...

So we have to improve the situation here, urgently!

Kai

Hello Kai,

Can I inbox you that information?

Thanks,

Raymond

I have sent you a private message

Kai

Hi Raymond,

well, the major issue here is that the photodiode is distanced too far away from the OPA858. Move the photodiode as much to the OPA858 as you can

Also, I would ground the metal housing of photodiode.

Kai