On Tuesday June 11th I took squeezing measurements at different nonlinear gains on the homodyne.  The overall message is that the level of squeezing seen on the homodyne is consistent with expectations for the level of nonlinear gain we have, and there isn't a meaningful constraint on the phase noise level from this data.  

Green power measurements:

Before getting started with squeezing measurements I checked green power levels with a good power meter and updated some calibrations.  I measured 23mW coming out of the SHG, and 1.45mW onto SHG power monitor diode, so I updated the responsivity of the SHG power diode from 0.431 to 0.297A/w, and changed the pick off mirror reflectivity to 6.3% from 4.7%, that the the reported power is more acurate.  Before the pickoff for the SHG fiber launch diode (after AOM, Farady, EOM, PBS) I measured 19mW of green, and 18.6mW into the fiber.  I measured 0.83mW on the launch diode itself and updated the reponsivity to 0.21A/W, and the pick off ratio to 4.4%.  In reflection off the unlocked OPO, I measured 6.3mW at the bottom periscope mirror, and 3.4mW at the REFL diode.  I did not update the calibration of the REFL diode since it was already pretty close.  I also spent some time adjusting the alignment and polarization into the pump fiber on ISCT6, I was only able to make a small improvement by walking the alignment.  I decided to check the coupling efficiency into the first fiber by measuring the power (with clean new fiber power meter attachment) at the patch panel, the transmission was 87%, while the transmission of the whole green fiber chain is 19%

Homodyne balancing:

The first time I did the homodye balancing (49786), I did it by adjusting the beamsplitter angle to bring the difference current to zero when both diodes were well aligned.  When I measured the visibility I noticed that the visibility was different in the two arms of the homodyne, which made me suspect that the balancing wasn't great.  This time I started out by trying to make the current in the two paths equal (measuring each while the other was blocked) and found that this did not do a good job of balancing the difference photocurrent at all.  I found that I had to make the DC power measured while eah detector was blocked fairly imbalanced to get a good zero difference current.  This seems like an offset problem with the way I was measuring the difference current, but it isn't because zeroing the difference current results in a flatter shot noise spectrum.  While all of that is a little puzzling, the intensity noise cancelation seems to be good enough.  I ended up with a slightly better shot noise spectrum on June 10th than on the 11th, but the data on the 11th is good enough to get a measurement of the squeezing level. In the configuration used for these measurements, I measured a homodyne visibility of 98% in one arm (the one that results in positive voltages) and 96% in the negative arm. 

Results:

The first attachment shows the squeezing and anti squeezing measurements, in the order they were taken. The shot noise spectrum here was taken at the begining of all the measurements, I retook it at the end of all the measurements and it hadn't changed.  I looked back at the variation in the power level on the homodyne throughout all the measurements, and I don't think it is large enough to cause any real confusion.  (It also wouldn't cause a systematic bias in the results, because I took the nonlinear gains out of order and the power drifted up and down, the changes in shot noise level would be pretty random).  The squeezing and anti squeezing level was estimated by taking the median of each trace, shown with dashed lines.

I used some code from pyDARM and some off the web to try an MCMC fit of this data for total efficiency (eta) and phase noise.  With this homodye visibilty we expect the known losses to come to 10%.  If I allow the total efficiency to go from 0 to 1, the fitting results in unrealistically low losses, near 5%, and phase noise of a few hundred miliradians, which seems to large.  The second attachment shows the fitting results with eta restricted to the range 0 to 0.9, which gives more reasonable results.  The fit does predict consistently lower anti squeezing than we measured, which I don't think is due to fluctuations of the shot noise level.