Reports until 14:05, Monday 24 March 2025
H1 SQZ
sheila.dwyer@LIGO.ORG - posted 14:05, Monday 24 March 2025 (83457)
nlg sweep with IFO: 67% total efficiency for squeezing

I'm starting to look at the nice data set Camilla took with different non linear gains in the interferometer,  83370.  A few weeks ago we took a similar dataset on the homodyne, see 83040.

OPO threshold and NLG measurements by both methods now make sense:

Based on the important realization in 83032 that we previously had pump depletion while we were measuring NLG by injecting a seed beam, we reduced the seed power for these two more recent datasets.  This means that we can rely on the NLG measurements to fit the OPO threshold, and not have the OPO threshold as a free parameter in this dataset.  (In the homodyne dataset, I had the OPO threshold as a free parameter, and it fit the NLG data well.  With the IFO data, we want to fit more parameters so it's nice to be able to fit the threshold independently.). The first attachment shows a plot of nonlinear gain measured two ways, from Camilla's table in 83370.  The first method of NLG calculation is the one we normally use at LHO, where we measure the amplified seed while scanning the seed PZT, and then block the green and scan the OPO to measure the unamplified seed level (blue dots in 1st attachment).  The second method is to measure the amplified and deamplified seed while the seed PZT is scanning, (max and min), and nlg = {[1+sqrt(amplified./deamplified)]/2}^2 (orange pluses in attached plot).  The fit amplified/ unamplified method gives a threshold of 158.1uW OPO transmitted power in this case, while the amplified/ deamplified method gives 157/5uW. 

Mean squeezing lump and estimate of eta:

The second attachment here shows all the spectra that Camilla saved in 83370.  As she mentioned there, there is something strange happening in the mean squeezing spectra from 300-400 Hz, which is probably due to the ADF at 322Hz was on while the LO loop was unlocked.  In the future it would be nice to turn off the ADF when we do mean squeezing so that we don't see this. This could also be adding noise at low frequencies, making the mean squeezing measurement confusing.

Mean squeezing is injecting squeezing with the LO loop unlocked, which means that it averages over the squeezing angles, and if we know the nonlinear gain (the generated squeezing) the mean squeezing level is determined only by the total efficiency.  With x = sqrt(P/P_thresh)

Rp = 1 + 4*x*eta / ((1-x)**2)
Rm = 1 - 4*x*eta / ((1+x)**2)
R_mean = (Rp + Rm)/2
x_factor = 2*x*(1/(1-x)**2-1/(1+x)**2)
eta = (R_mean-1)/x_factor
 
Despite the confusing effects that might be from the ADF below 400 Hz, these estimates of total efficiency are all in good agreement with each other at high frequency, and they do suggest a slight frequency dependence to the total efficiency.  Above 1kHz these suggest a total effiicency from 64-67%. 
 
Fitting squeezing and anti-squeezing.
I chose two high frequency bands that seem to be free of technical noise, and estimated the squeezing, anti-squeezing and mean squeezing levels for each, and used them separately to fit for OPO threshold, total efficiency, and phase noise, see plot. This is without doing any subtraction of non quantum noise. The two frequency bands agree well with each other, with a threshold of 158uW, a total efficiency of 67%, and basically no phase noise.
 
Loss Budgeting:
Tallying up the losses from the google sheet, we have 83% total efficiency expected, if we also include the 6% of unexpected losses seen on the homodyne (which is probably due to crystal losses), this leaves us with an additional 14% unexplained losses related to the squeezing injection into the IFO. 
  • opo_esc= 0.985
  • SFIs = 0.99**3
  • FC_WFS = 0.99
  • ZM456 = 0.99
  • OFI = 0.99*0.995
  • SRC_loss = 0.99
  • OM1 = 0.9993
  • OM3 = 0.985
  • OMC_QPD = 0.9904
  • OMC = 0.956
  • QE = 0.98
This is a larger loss estimate than, in 82097, where the estimated total efficency was 72%, where we had to infer the nonlinear gain.  With this total efficiency of 67%, if we were to reduce the losses in HAM7 by 6% by moving the crytsal we would see an improvement at high frequency from 4.6 dB to 5.1 dB, if we keep the OPO trans at 80uW. 
 
We can also use this data set to look at how much technical noise is limiting our squeezing.  However, we can already see that since the mean squeezing and the anti-squeezing agree very well with this level of loss, which explains our squeezing very well at high frequencies, it seems that phase noise and non-quantum noise like noise from the CLF is not an important limit to our high frequency squeezing.  Another thing to note about this data set is how obvious the SRCL detuning is at the high NLG squeezing measurement.   The code used is attached, and is available in this git repo along with the dtt file.

 

 

Images attached to this report
Non-image files attached to this report
ifo_nlg_sweep.py
ifo_nlg_sweep.py