I assembled one of Anchal's revision of Zach's M2ISS transimpedance amplifier for shot noise limited measurements of the transmitted light power. To keep it simple I did not populate the switching logic for switchable transimpedance gain, and also omitted the whitening stage. The 5V bias voltage for the PDs is buffered by individual OP27. Since the photocurrent will be no larger than ~4mA I omitted the BUF634 on the bias side. I did add BUF634 in the transimpedance stage so the circuit would be able to drive 50Ohm loads if needed.

I was using an LT1128 at first, as it should deliver the lowest nominal noise floor for a transimpedance gain of 1kOhm, but I was observing oscillations at ~1.2 MHz which I couldn't suppress without seriously affecting the bandwidth of the detector. Koji pointed out that the LT1128's open loop gain has small phase margin at this frequency, as visible in the data sheet

I replaced the LT1128 with an OPA827 which performs only marginally worse according to Disha's LISO simulations. This remedied the oscillations. I still had to put 22pF in parallel with the 1kOhm transimpedance resistor for loop stability. I boxed the satellite amplifier and mounted two 3mm InGaAs diodes in 4" mounts that I had extruded from Zach's M2ISS design. Using an SR560 I measured the PD noise, which is shown in Attachment 5. I need to check why the west path is worse at low and high frequencies, but the baseline NEP for both is about 6 pW/rtHz, pretty much on point with Disha's LISO simulation.

I changed the optical configuration of the experiment back to sending the undiffracted beams from the AOMs into the cavities. Attachment 6 shows the transfer function from input to the commercial AOM driver to output of the new custom transmission PDs.

I also completely modded the existing ISS box to host two of Anchal's new ISS servo. I finished debugging the circuits and now both are doing exactly as asked - noise measurement pending. Attachment 7 shows the open loop gain of just the servo boards, and Attachment 8 the product of servo TF and plant TF. I haven't yet closed the loop, but it looks like some external attenuation would be enough to make the system stable. I'll try this tomorrow.