Today we started to work on the lock of CARM (at very low frequency) in order to keep the RFC in resonance, by using the RFC error signal at 8 MHz (as the 6 MHz one gets suppressed a lot during lock acquisition).

As preliminary actions, the BS alignment loop was restored in full bandwidth already in LOW_NOISE_1, after a check and a small tuning of the demodulation phase of the SDB2_B5_QD2_56MHz sensor; then we started to work on the foreseen topic:

• first thing, while still being in LOW_NOISE_3_SQZ, we tried to change the modulation frequency at the level of the LNFS in order to check if this was doable without making (too big) glitches and/or unlocking; this is actually the same as the initial FmodErr tuning done by INJ_MAIN, but during a full lock; in any case this was successfull;
• after a very preliminary test we collected a (probably unrelated) unlock, then we proceeded with the ITF only in LOW_NOISE_1, as this is the latest stage we need to already control CARM (before we try to lock the OMCs):
  • we tuned the demodulation phase of the SIB2_RFC_PD1_8MHz_I signal (which is already propagated as LSC_RFC_8MHz to the CARM algorithm); this was found to be -2.795, and we saved it in the CARM_DARM_LOCK.ini file, section [CARM_SLOW]; mind that as of now nothing reads or writes it, it is just stored there;
  • using a draft filter we closed the loop, at first by helping it out to reach the resonance by moving the IPs in order to get the correct sign and tentative gain, and then we tuned the gain;
  • the filter is unconditionally stable for very low gains so we could engage it with an arbitrarily low gain, then we iteratively increased the gain to study its behaviour (up to -12); the loop behaved properly but, not being optimized yet, it is quite noisy on the correction despite still having quite a low bandwidth; moreover, for this reason we could not engage it directly with the high gain as it would saturate the MAR corrections and unlock, and we found -0.6 to be a safe value;
  • given the noisy correction that was impressed on DARM (Figure 1), during these and later steps we disengaged the dampers on the Y and TZ degrees of freedom of the Fabry-Perot mirrors, as they use DARM_ERR and could become unstable in case of noisy corrections;
  • this, however, was still too high in order to proceed with lock acquisition towards DC Readout and LOW_NOISE_2/3;
  • as a last test, we locked the ITF completely with the CARM loop open as usual, then we engaged it in LOW_NOISE_3_SQZ but with very low gain (-0.01), after disengaging the FmodErr slow loop: in this case it worked, despite being very slow with such a gain;
  • at this point we tried to tune FmodErr by hand, again by acting directly on the LNFS with small steps (of 10 mHz) and by looking at the usual Sc_BS_fModErr signal as a probe (Figure 2); we could manage to zero the (out of loop) error signal with no issues or glitches, and also keeping a good sensitivity; unfortunately, we couldn't make it to start and increase the gain, as the ITF unlocked and went to an instability period, thus ending the shift.
• in conclusion, this was a quite good starting point and the next steps will be to design a better suited filter and to find a compromise between getting better accuracy and not spoiling the correction levels.