We performed the characterization of the ALS DARM/CARM by measuring the actuation efficiency and tuning the gain of the LSC LOCK.
The reference was the measurement results of the 14th July, klog30398.
(Fig.1.) TM : the TFs from test of ETMX TM TEST to ALS_DARM was increased 1.9 dB, so we put the gain FM (FM4) to
gain(1.255)gain(0.80352)
(Fig.2.) IM : the TFs from test of ETMX IM TEST to ALS_DARM was increased 2.9 dB, so we put the gain FM (FM3) to
gain(1.685)gain(0.7161)
(Fig.3.) MN : the TFs from test of ETMX MN TEST to ALS_DARM was increased 3.1 dB, so we put the gain FM (FM3) to
gain(1.455)gain(0.6997)
After changing the gain, we tried to lock the ALS_DARM and it seemed stable (depending on the situation of the PLL Y )
We also measured the OLTF for both ALS DARM(Fig.4.) and CARM(Fig.5.)
(Additional comments)
The reference plots we used were measured around 7/14 00:00 UTC
The measured results attached on the original report were measured before we changed the LOCK FBs.
- TM:
- The gain increased by ~1.9dB @ 5Hz
- Removed gain(1.3) and gain(0.677) from the FB "gain" noted here: klog30817
- Added gain(0.80352)
- IM:
- The gain increased by ~2.9dB @ 5Hz
- Removed gain(0.677) from the FB "gain"
- Added gain(0.7161)
- MN:
- The gain increased by ~4.9dB @ 3Hz
- Removed gain(0.677) from the FB "gain"
- Because there was a gain 1.23 when the ref measurement was performed.
- So we added 1.23*0.5689(=-4.9dB) = gain(0.6997)
For further tuning, we measured TF from ISC_INF_OUT to DARM signals when LSC_LOCK guardian was ALS_CARM_LOCKED state.
Blue line in fig1 shows the current heirarchical actuator TF.
Combined transfer function seems not good in the following viewpoints:
1. Some notches appeared around 2-3 Hz and resonant peak cannot be seen in the TF, which makes the gain around 2-3Hz lower.
2. Lower gain at low frequency.
So, we redesigned the filters and made a new herarchical actuators.
Design concept for new filters are as follows:
1. Crossover frequency between TM and IM is 4Hz.
2. Crossover frequency between IM and MN is 1Hz.
Red line in fig1 shows the TF from ISC_INF_OUT to DARM with new herarchical actuators.
In the new TF, resonant peaks can be seen around 2-3 Hz and higher gain at low frequency, so seems much better.
However, phase rotation around 2Hz is larger than before, and ALS_DARM cannot be locked stably with new herarchical actuators.
So, we gave up to update the actuators at this moment.
To use new hierarchical actuaotrs, we need to redesign the DARM filter as well.
I estimated equivalent TFs from LOCK_L_EXC to ALS_DARM by using TFs from TEST_L_EXC to ALS_DARM measured in klog#32464. These TFs from LOCK_L_EXC contains all differences between TM, IM and MN paths, so crossover frequencies of these TFs are same as ones of TM, IM, and MN actuators. (IM and MN are strictly different because DRIVEALIGN is not taken into account in my estimation and the non-diagonal elements are not zero for IM and MN.)
Fig.1 and Fig.2 shows the TFs from LOCK_L_EXC with old filters and with new filters. In the case of old filters, MN (red curve) and IM (blue curve) cross each other around 2.46Hz which is a common resonant peak of IM and MN. In the case of new filters, they seem to cross each other around 2.2Hz (below a common resonant peak) though crossover frequency is designed around 1Hz. If we can see the instability around 2Hz, instability might be able to be mitigated by reducing MN gain around 2.2Hz. e.g. if we add 3Hz pole for MN loop, MN gain around 2.2Hz can be reduced as 2dB though phase margin is also reduced as ~35deg. Reduction as 35deg in phase margin might be acceptable because the phase margin around 1Hz is ~60-70deg.
Memo for same work in future
For considering the design of the DARM loop, we need to know the TF from LOCK_L_IN2 to ALS_DARM. On the other hand, reversing LOCK_L and/or taking DRIVEALIGN into account is complicated, so it would be better to use LOCK_L instead of TEST_L for same work in the future. When there is a change only in overall gain, checking relative gain between latest and reference TFs is enough. But if there are some changes in zero/pole. only checking relatve gain of TF doesn't always make stable loops.
