This is an update on the quantum noise compairson to models started in 80747. Apoloiges for the large numer of plots. The main message is that we have a decent model of the quantum noise from October (see this and this plot ) can get a fairly independent constraint on homodyne angle and SRC detunings from these squeezer data sets without the filter cavity, and some loose estimates for arm power based on squeezer losses. There is a discrepancy between the model and the data in the traces taken near anti-squeezing without the filter cavity that is interesting, and I haven't tried to model mode mismatches yet.
Outline of how this model is made:
- start with an assumption about arm power (365kW). Later we can check how this model changes with arm power changed.
- use noise budget estimate of technical noise at 1-2 kHz, measured ASD and assumption about arm power to fit IFO readout efficenecy
- (82% here, compare to known readout losses from google sheet ((1-0.005)*(1-0.00072)*(1-0.015)*(1-0.0096)*(1-0.036)*(1-0.02)= 91.6% expected from OFI 1 pass, OM1, OM3, OMC QPD pick off, OMC throughput, DCPD QE). This implies 10.4% unknown readout losses, including IFO to OMC mode mismatch, but this depends on the arm power we start with.
- use the best sqz and anti-sqz (either with or without filter cavity), subtracting noise budget estimate of technical noise, at kHz to estimate total squeezer efficiency (72%) and nonlinear gain (14.8). Set the squeezer injection losses so that the total squeezer efficiency is injection losses * IFO readout losses.
- This gives us 12.1% injection losses, compared to 8.2% known injection losses from the google sheet (1-0.015)*(1-0.01)**3*(1-0.01)*(1-0.01)*(1-0.01)*(1-0.01) for OPO, 3 SFI passes, FC WFS pick off, ZMs, OFI sqz injection, and SRC losses), this gives us 4.25% unknown injection losses.
- A mode mismatch could cause a different readout loss for the IFO beam than for the squeezer, so this assingment of losses could be incorrect if there is a large mode mismatch between the IFO and the OMC.
- The code fits the squeezing angle for anti-squeezing and squeezing traces to either maximize or minimize quantum noise.
- For the other squeezing angles, (mid squeezing, +/-10 from sqz or a sqz), it uses a fit. This fit starts with an estimate of the non-quantum noise made by subtracting the measure DARM asd without squeezing from the quantum noise model without squeezing, from 2055 Hz to 2300Hz, and adds that to a quantum noise model to fit the squeezing angle. The procedure up to this point should force all the models to match the measurements at 2kHz, but not at mid or low frequencies.
The first two attachments show data from mid October (80664), separated into two different plots (one with mid squeezing, one with +/-10 degrees from sqz and a sqz) to make it easier to read. We can look at the mid and low frequencies to help us constrain the model for homodyne angle, SRC detuning, and maybe mode matching and FC parameters. With the filter cavity, the traces that are +/- 10 degrees from squeezing and anti-squeezing don't provide much information, we can probably skip these in future data sets, but the ones +/-10 degrees from anti-squeezing without the filter cavity are interesting.
Arm power: We can get some constraints on arm power by assuming that all of our unknown loss is injection loss, or all readout loss, this gives us a range of 327kW-381kW with a homodyne angle of 10.8 degrees and no SRC detuning. Attached are plots made for the high and lower power models, by clicking back and forth you can see that the lowest powers seem to disagree with the low frequency data, implying that some of our unknown losses truly are readout losses, but this isn't a very precise way to estimate arm power.
SRC detuning: Without the filter cavity, the mid squeezing traces are most clearly sensitive to SRC detuning, shown in this attachment where the SRC detuning of 0.057 degrees is clearly too large. You can flip through these 5 SRC plots which suggests that our SRC detuning at the time of these measurements was between 0 and -0.057 degrees (I'll leave it at -0.029 degrees, or -1mrad in the model) .
Homodyne angle: In 80747 we noticed that anti-squeezing without the filter cavity is very sensitive to the homodyne angle, and that we've been using the wrong sign for our modeling in O4. Here are plots that reaffirm that the negative homodyne angle is wrong, with mid squeezing (as we did in 80747) and with squeezing angle +/-10degrees from squeezing and anti-squeezing, which also clearly shows the problem for those traces near anti-squeezing. It's also nice to notice that the mid sqz traces aren't especially sensitive to homodyne angle in the mid frequency range taht we were using to set the SRC detuning, so this data set without the filter cavity can give us fairly independent constraints on SRC detuning and homodyne angle. Here is a plot of the homodyne angle of 10.7 degrees (based on 71913), without the squeezing traces so that it's easier to see what's happening. There is a large discrepancy between the model and FDAS -10 degrees, from 30-55Hz, and there is also a smaller discrepancy for FDAS+10 degrees in this frequency range. This discrepancy can be redcued a bit by making the homodyne angle larger, but that doesn't seem to be a good model of the data, so this discrepancy is probably due to something we aren't modeling correctly here.
Filter cavity detuning: We have used -33Hz as the filter cavity detuning, but data from mid squeezing with the filter cavity suggests that the detuning in this data set closer to 27 Hz. Here is a series of plots to show how mid sqz is sensitive to FC detuning.
The last two plots are with the final model paramters, it is a reasonably good description of the data, with the main discrepancy remaining in the anti-squeezing without the filter cavity. The scripts, parameters and data used to make these plots is in quantumnoisebudgeting commit 92f7e965. The quantum noise paramters used are here.
