OPA855: Wideband composite photodiode TIA (4KOhm, 500MHz) oscillates

morning Sven, 

While conceptually reasonable your circuit is likely not going to make it at these speeds (one way to see that is to add package and layout delay into that 2nd stage in simulation, if you can, use a transmission line, if not use a all pass filter). Just a guess but the package is maybe 0.3nsec and layout another 0.2nsec. 

I tried a test with your 50pF and the OPA855, can't get much more than 100ohm feedback and hold >800Mhz. If you have not back biased your diode to reduce it's capacitance, that would be a very useful next step. 

And then once you get a reasonable gain out of your first stage by reducing the detector C, add gain with outside the loop postamplifiers. 

Dear Michael,

Thank you for your quick reply!
Is there that much delay in layout & package? Critical traces are only few mms? I would be interested to know how this actually works.

Is it possible to realize a stable loop with only OPA855 up to >500MHz? Or do I run into the same issue as right now, including layout effects. If it would work, I could indeed cascade voltage amps if I can get high enough gain such that voltage noise of the cascade amp does not dominate.

Do you think there is anything to win with building composite loop amps using SIGe tech, or does the package/layout bring a fundamental limit in composite structures?

The detectors are already completely depleted so 50pF is the lowest we can get. They are rather large in size. We will be testing on smaller devices in the future, but with these results I'm not sure whether we can get faster amplifiers that actually work on a PCB...

Cheers,
Sven

Probably need to consider a bootstrap then, 

There have been numerous discussion of that here, here is one, 

https://e2e.ti.com/support/amplifiers-group/amplifiers/f/amplifiers-forum/980967/opa656/3624536?tisearch=e2e-sitesearch&keymatch=bootstrap%252525252525252520transimpedance#3624536

Michael,

I did consider bootstrapping, but we have 4 detectors that share cathode. I want to read them simultaneously, so bootstrapping does not seem to be possible. 

Hi Sven,

your circuit is instable, unfortunately.

But you could give this scheme a try:

sven_opa855_2.TSC

To increase the bandwidth you can slightly decrease C1 and/or C3.

Keep in mind though, that it is rather difficult to exactly hit C1=200f. A bit too high and the bandwidth goes down, a bit too low and the circuit becomes instable. So it might be easier to adjust C3 to get proper results, and not C1.

With the given component values the phase stability analysis shows sufficient phase margins:

sven_opa855.TSC

sven_opa855_1.TSC

I haven't understood why bootstrapping doesn't seem to work. Can you show how the four photodetectors are wired?

Kai

Helly Kai,

Thanks for your response, I will have a closer look in the lokal Cf. 

We have a 10mm diameter detector with 4 pixels. They share cathode connection. You would need to bootstrap the cathode for every pixel, but only have one connection.

Sven