Aritomi, Eleonora, and Yuhang
Last Friday, we had an issue with homodyne alignment. We found the frequency of peaks in shot noise is exactly the frequency of bench EW direction motion. At some point, we were thinking this is because of a bad way of aligning homodyne, but this seems not to be the case in the end. We figured out it should come from leaked LO backscattering.
The leaked LO mainly comes from the AR coating of cubic BS, then it goes to a sensor card in the squeezing path and the sensor card causes backscattering. About the leaked LO from homodyne, we measured power as following
from cube BS |
2.5uW |
from flipping mirror |
0.85uW |
from lens/PD |
almost no |
total |
3.35uW |
And the sensor card is shown in the attached figure 1, which has a plastic shining surface. It will reflect a little bit scatter light which contaminates the shot noise spectrum. Besides, this sensor card is put in the position which makes it easy to sense the motion of bench in EW direction. So we got the spectrum as in the attached figure2. (In this case, we are using our homodyne as a very sensitive sensor of reflective surface motion)
We tried to characterize the effect by putting it farther and farther, but the contamination doesn't change. Until we put it after a Faraday isolator, which means we have an FI to isolate this backscattered light. We got a reduction of peaks by 15dB. See attached figure 3. But the backscattering noise didn't disappear.
But we need to notice that if the reflected surface doesn't move, we will not have this backscattering noise. For example, when we don't put anything in-between OPO and homodyne, although there should be light back-reflected from OPO to homodyne, we don't have backscattering noise.
Then we tried to send this leaked LO to filter cavity, we got the homodyne shot noise spectrum as shown in the attached figure 4. This measurement suggests that our low-frequency bump in FDS is from this backscattering noise. Also, the phase noise of squeezing will not introduce the increase of floor by almost 30dB because the anti-squeezing is only ~15dB.
Additional: we measured the backscattering noise from the sensor card when turbopump is on and off. We could see the difference from attached figure 5, which means it really introduces more bench motion.