In recent times IMC (and PMC) is loosing the lock systematically when MC is moving in order to put RFC in resonance. Actually this action determines a variation of laser frequency, performed by the 'thermal correction' (in the data, BsX_ML_TH_CORR). This is an auxiliary actuator, which replaces the main one (ML_PZT), when the request of frequency variation is too large. The 'reallocation' of the actuation is performed by the board which take care also of the Beam Pointing Control (known by DAQ as BsX). The loop is simply an integrator whose input is BxS_ML_PZT_CORR and the output is BsX_ML_TH_CORR, sent to a DAC connected to ML_TH actuator. The loop can be very slow if only the natural drift of the frequency has to be compensated, but for the ramp needed in order to lock RFC it cannot be too slow. Currently the UGF could be somewhere between 0.1 and 1 Hz, but a characterization has never been done. With the current setting, the effect of the ramp on IMC length is that the thermal correction follows fast enough, but not so fast to keep PZT correction close to zero: usually it reaches about 0.5 V. At the beginning (5 years ago or so) the gain of the loop was lower, and the maximum value of PZT_CORR during the ramp was 1 V or a bit more. After some event of lock loss during the ramp (3 years ago or so), we evaluated that 1 V was to close to the tolerable limit of PZT_CORR, so the gain of thermal loop was simply doubled.
Today we investigate about the possible relationship between the lock losses during the MC ramp and some malfunctioning/large values/other stuff/ of TH/PZT loops.
The typical pattern of IMC lock loss during RFC locking has been replicated (fig 1): 50 um of MC displacement in 30 s, producing Delta TH_CORR = 40 mV and a maximal excursion of PZT out of zero < 1 V. The lock loss typically does not occur when PZT_CORR is larger.
A larger excursion of PZT_CORR (> 1V) has been produced in two ways, with TH loop disabled (fig 2): at first, a small offset (10 mV) has been applied to TH_CORR; then, a small displacement of IMC length (10 um) has been performed. No lock loss. PZT_CORR should not be the responsible of the problem.
Could be TH_CORR the responsible? A fast step (in 1 sample), has been applied (fig 3). 15 mV and no more, because the compensating action of PZT is more than 1 V. No lock loss.
We tried with the same 50 um IMC displacement, performed slowly (fig 4). At first, 60 s, then 120 s. Correction were doing their work quite smoothly. In both the cases, IMC unlocked with the same pattern.
One possible conclusion is that who annoys IMC (PMC) lock is the variation of frequency itself, and not the way that variation is performed.