Stefan, Kiwamu
As a part of the beat noise study, we locked the main infrared laser to the X arm cavity by feeding the reflection signal back to the MC length.
Although we succeeded in locking the IR, the beatnote tonight was not great at all. It was roughly 500 Hz / sqrtHz in a band from 1 to 900 Hz when the infrared was locked. Plus, the mode hopping was so frequent that I was not able to integrate the spectrum lower than 1 Hz.
IR locking to the X arm:
Because the green light tonight was not stable enough for us to do the CARM hand off, we instead started locking the IR to the X arm. This is the first time for us to directly lock the infrared main laser to an arm cavity without an aid of the ALS technique. We started this from estimating the PDH optical gain of REFLAIR_RF9 by simply looking at the free swinging wave form. According to Alxea's math (alog 7054), we got ~ 1.21 counts/Hz for the optical gain. Also the MC2 M3 stage was approximated to be 18.2 / f^2 Hz / counts. Using these two information, we set the LSC gain to be 0.46 as an initial guess. Note that we cranked up the whitening gain of REFLAIR9 to its maximum. The demod phase was also adjusted to maximized the PDH signal in in-phase. We then fed it back to MC2.
A good servo gain was then empirically found to be about -0.2. Even though we didn't have a transmitted DC or reflected DC signal, we could tell if we grabbed a sideband resonance or carrier resonance by looking at the size of the Q-signal. This allowed us to detemie the right control sign.
The locking procedure is as follows:
One major difficulty we had during the IR lock was that a mechanical resonant mode at 41 Hz (bounce? roll? of MC2) rang up so much that we could not keep locking the IR. We often let the IMC alone to allow it to settle down for a while and this helped a lot. Also Stefan could introduce the additive offset path on top of the MC length control. He had to crank up the input gain of both common mode board and IMC board to the maximum of 31 dB. It seems that we could still go higher although we didn't quantitatively evaluate it yet. This additive offset technique should serve as a good mitigation for this 41 Hz mode issue while maintaining a high control bandwidth.
The attached is a screen shot of the IR locking configuration.
Noise of the green beatnote was far from good:
The noise performance of the green beatnote was not great. Something must be wrong. The smallest noise floor I could get was approximately 500 Hz/sqrtHz in a band from 1 Hz to 900 Hz. I tried not include a mode hop during the spectrum integration. Even so, as you can see in the attached spectrum, it was suspiciously flat. I used H1:ALS-C_COMM_PLL_CTRL for measuring the beat noise. The calibration was done by injecting a known RF frequency into the PLL and measuring the ADC counts. It was measured to be 7.916 Hz/counts. Also I added a z40/p1.6 filter to cancel the VCO response.
Additional Note:
PRM is currently in its parking position to avoid a power-recycling fringe at the reflection port.