I measured beat signal, after damping most of the optics, realigning the beams to the cavities, measured the slope of error signals and applied it to the measured detection noise. Acoustics peaks around 200Hz to 1kHz still present.
Fig1: beat measurement, I added shot noise and electronic noise from both cavities to a single trace called detection noise (from measurement).
I turned off the HEPA fans on the table and on the clean bench before measured the beat signal (after I finished, I turned on the fans as usual).
The peak at 58 Hz shows up this time. I think this is the peak from beam line motion of the stacks, see PSL:716. (I think that was before we switched to the softer springs, I'll double check). Note that the air springs were not activated during the measurement, we can try using it and see if there is any improvement.
There is a good improvement on minimizing the acoustic peaks, although still not enough. Also, increasing the modulation depth seems to help with the flat noise part at high frequency, we may really sit on detection noise.