Sheila, Alexa, Elli, Rana, Evan
Summary
Building on this afternoon's work of transitioning control of the X arm from ALS COMM to sqrt(TRX) [LHO#16250], we have been able to recover the handoff of CARM from ALS COMM to sqrt(TRX+TRY).
Next steps are to try to reduce the CARM offset further, transition to RF DARM, transition to the in-vacuum transmon QPDs for CARM, transition to digital REFL9I for CARM, and then reduce the CARM offset all the way to 0 pm.
Details
The transition sequence is largely the same as for the X arm case. I repeat the sequence here, with modifications as appropriate:
- Adjust the ALS COMM VCO so that the TRX and TRY signals are around 1 ct in magnitude.
-
Engage the sqrt(TRX) path with a 1 Hz pole and a 40 Hz zero, an offset of −1.0 ct in TR_CARM, and an overall TR_CARM/REFLBIAS/REFLDCBIAS gain of 50 ct/ct. The common-mode board has an input gain of 0 dB (ALS COMM has an input gain of −4 dB). Addtionally, we have the usual filters engaged in TR_CARM/REFLBIAS/REFLDCBIAS:
- a 1 kHz pole and a pair of complex zeros at 35 Hz (with a Q of 1) which are meant to make sqrt(TRX+TRY) fall off as 1/f out to high frequency
- a 1.6 Hz pole and a 40 Hz zero which imitates the response of the ALS COMM VCO.
- Turn off the COMM VCO frequency servo. Tune the VCO voltage so that the sqrt(TRX+TRY) output (as monitored via REFLDCBIAS_OUT) is close to zero.
- Engage a 0 Hz pole and 1 Hz zero in the sqrt(TRX+TRY) path, so that sqrt(TRX+TRY) has a true integrator.
- Turn off the ALS COMM PLL boost (pole at 1.6 Hz, zero at 40 Hz).
- Ramp down the analog gain of ALS_COMM on the summing node board to −32 dB. Attempting to switch off the input will blow the lock, so we ramp down the analog gain on the common-mode board by about 20 dB. But both inputs are left on.
- Remove the 1.6 Hz pole and 40 Hz zero engaged earlier, by engaging FM4 in REFLBIAS (it is an anti-boost which undoes the effect of FM3). It is an unnecessary boost which prematurely cuts off the lower edge of the phase bubble of the TR-controlled CARM loop.
- Turn the gain of TR_CARM/REFLBIAS/REFLDCBIAS to 35 ct/ct, so as to place the UGF in the best portion of the phase bubble.
Sheila has put these steps into the Guardian, but they haven't been extensively tested yet.
The attached plots and data give the OLTF of CARM controlled by sqrt(TRX+TRY).
Some lock loss times when we were on TR CARM.
7:26:10 Jan 25 UTC
8:11:20
8:13
5:54:35
A crucial point is that we were never able to actually shut off the ALS. Although we reduced the ALS gain by 30-40 dB after bringing on the Thorlabs/Transmon signal, actually switching off the ALS signal using the IN2 on/off switch on the CM board broke the lock each time. Also turning the gain down further using the gain slider on the summing board (before the CM board) also broke the lock.
At the time of the lock losses, the CARM UGF was ~100-150 Hz, so it seems like the ALS would have no effect on the loop. Since we're able to hear the gain slider steps on the CARM audio (because of our great new speakers), we suspect that the ~mV level offsets in the gain slider and the on/off switches are the problem.
Talking with Daniel today, we think one strategy is to increase the level of the TR_CARM signal (either digitallly or analogly) before it gets to the CM board so that mV offsets don't matter as much. We can compensate this gain increase by reducing the Additive Offset gain slider in the IMC board from its current value of +16 dB, to something more like -20 dB. We know from log #16251 that the required AO range is going to be less than 100 Hz.