Yesterday Kentaro and I took squeezing measurements in the homodyne at various CLF levels with the ISS's running. The homodyne should still have the high ~.98 visibility established Sunday by Nutsinee and Kentaro. This was taken with NLG=5.91 (OPO Trans monitor 1.8x), corresponding to 11.74db of generated antisqueezing.
shows the various squeezing spectra and references up to 4kHz. You can see immediately that the reference isn't flat. The compensation filters aren't shaped quite right. The noise becomes worse at high frequencies due to rolloff of the whitening and relative increase in bit-noise (see D1700403). We could potentially increase the gain of those boards (or downstream?) to help with that, although improving the homodyne isn't too essential.
At around 2kHz you can see we can see ~4-5 db squeezing. There is some additional noise at low frequencies which is dependent on the CLF power level. This is alarming if it is also affecting the interferometer squeezing. At the higher CLF powers, it eliminates the gains from squeezing. Interestingly, it does not degrade the antisqueezing spectrum, illuminating some mechanisms. Since it only affects squeezing, it could potentially be CLF-level dependent phase-noise, but the SQZ spectrum has shape and phase noise will (almost) always create a broadband degradation to squeezing. For this reason we suspected noise sidebands on seeding, which would indeed have both a frequency and CLF-level dependence.
From here we switched operating modes. After the ALS fiber pickoff point and AOM frequency adjustment to stable 80MHz, the SQZ laser can be stably offset in frequency to the PSL. This allows us to shift the LO frequency w.r.t the squeezer carrier and seeding, to look for beatnotes. The LO loop was unlocked, and the VCO was shifted by 200Hz, creating a 400Hz offset of seed light and LO. Any seeding should create a 400Hz noise peak on the homodyne. The total integrated noise power excess over shot-noise indicates the total power in the seeding.
HD_SEEDING.png shows the studies here.
The main curves to look at are to compare the w/ SEED and w/CLF. The w/ SEED curves show the homodyne with the CLF fully blocked, and the SEED semi-blocked (unblocked flipper, but leaking through the fiber-switch). This can purely generate the 400Hz peak and serves as a good reference, especially given the very strange 800Hz harmonic peak caused by the CLF light.
The CLF seeding curves are generated with the CLF having 40uW on OPO-M2, held to that level by the CLF-ISS from a maximum of 70uW given the picomotor settings.
The further studies show that the CLF 800Hz peak depends on the nonlinear gain. The OPO power was decreased to .42x on transmission. This reduced the seed and CLF levels. The SEED level dimished because the seeding light should be scaled approximately by the NLG. The CLF noise level scaled by more than the true seeding. Furthermore, in the bottom plots, the OPO pump was zerod, and the OPO was manually moved on resonance (it is extremely stable, and holds for a while). The resonace was checked by maximizing the LO-CLF beatnote level while the CLF was on. During this mode, the w/SEED still shows a similar level of true seeding, but the w/CLF shows no 800Hz peak, and only a small excess of true seeding at 400Hz, so indeed the CLF has very little (few attowatts) of actually seed light in this configuration. The missing 800Hz during 0NLG shows the dependence on parametric gain of the CLF noise mechanism.
It is very difficult for the OPO to generate 800Hz in any optical manner, since the pump/parametric gain frequency is matched to the seed/CLF, nothing in the OPO "has knowledge" of the 400Hz offset, only the LO light carries the 400Hz offset. To generate 800Hz, there must be some nonlinearity in the homodyne. My theory is that the RF level of the CLF should be oscillating at the 800Hz. The RF level depends on the squeezing angle due to the unbalanced CLF sidebands. The 400Hz LO-offset then has the squeezing ellipse rotating at 400Hz, and there is a similar ellipse determining the CLF3 sideband level and phasing (it is offset from the sqz ellipse ~20deg due to the 3.125MHz and OPO bandwidth). With the ellipses spinning, the RF level then gets doubled from the mirror symmetry of the ellipse. I think the noise peak is from finite power supply rejection and the oscillating RF power.
This means that the homodyne electronics could probably be improved. It also means that the degradation should likely be coherent with the RF3-Q signal. We will look today once the squeezer is back (some high-voltage for PZTs is down at the moment).
For the interferometer this may or may not have implications. The RF gain of the OMCPDs is smaller and the electronics diffferent, so difficult to predict. The OMC lowers the CLF power to 1%, but the carrier is 20x the homodyne. In the past, we have characterized the frequency-doubled peak as OPO conversion of optical backscatter (unique to IFO compared to homodyne, which has much lower backscatter). If this mechanism is present on OMCPDs, then we may have to revise backscatter measurements. Once we have IFO time, we can enable/disable the CLF and perform the same test as here with the IFO to check.
