Naoki, Vicky, Dhruva
With the ability to increase power into the OPO launch fiber, we decided to take NLG sweeps on the HD to recalibrate our OPO Trans vs NLG measurement and get an estimate for loss, phase noise and mystery noise on the HD. We managed to increase the NLG to being very close to threshold, where pump depletion started biasing our estimates of the NLG by a lot. So, in order to estimate loss and noise from SQZ levels, we used the OPO trans to estimate NLG. This involves making a linear fit between sqrt(OPO_TRANS) and the inferred single pass squeeze factor from the measured NLG (dropping the final few points that see pump depletion). The estimated NLG from trans is quite sensitive to this linear fit parameter at high NLGs, so there is some uncertainty around the last few points.
We seem to measure up to 5.5dB (with NLG 40) on the HD, our visibilty is around 98% and we think we should be measuring more given the amount of loss we can account for. We do see a phase noise turn-around at very high NLGs (~100) which probably constrains our phase noise to be very low. With the uncertainty in NLG measurement, our loss can be contstrained to 22-25% based on mean SQZ measurements.
Attaching opo power and clf power trends from Dhruva and Naoki sweeping NLG on HD. These observed SQZ levels used low AOM1 RF drive (+11dBm) with all CLF power transmitted through the PBS -- the opposite of our CLF configuration from earlier 67038. It does seem we can squeeze with low AOM1 RF levels, with almost 1.5x higher CLF power at the RF6 PD (~18uW vs prev ~10uW).
In the grander scheme of things, we are trying to understand what about higher CLF levels degrades observed squeezing. I think today's work suggests it is not just the RF/AOM generation of the CLF that reduces broadband observed squeezing. We have yet to pin down the issue with high CLF levels. Some brainstorming might be maybe the OPO doesn't like high CLF (we do observe ~1% pump depletion for >50uW transmitted CLF), or maybe high CLF creates noise through the RF6 PD or following electronics (could check optically attenuating the CLF power at the RF6 pd?). From Dhruva and Naoki's measurements today, we can also begin to constrain the classical technical noise contribution in the homodyne. Sqzing can only reduce the shot noise contribution to HD_DIFF_DC_OUT, but wondering if we bottom out the measured sqz level b/c we are now hitting the noise floor from a non-shot-noise contribution to HD_DIFF_DC_OUT, similar to how dark noise contributes to this in 67186.
I've added a flat techincal noise estimate to the fits. Right now, it looks like it's around 10dB below shot noise