We measured the transfer functions to the cavity length from M0 POS, L1 POS and L2 POS, when the cavity was locked and only M0 was damped.
At the same time we also measured the transfer functions from the same actuation points to the OPLEV signals.
Two main goals of this were:
1. To see if L2 stage (penultimate mass) drive was working fine. There has been speculations but no definitive answer.
2. To provide a set of measured data for SUS so hierarchical control effort could be accelerated.
Anyway, if you're only interested in the plots see attached. Frequency points are kind of sparse and not even (the former is constrained by time, the latter is by the fact that I'm throwing away low coherence data).
Plots as well as data files etc. are all under /ligo/home/controls/keita.kawabe/OAT_2012/ETM_M0_L1_and_L2_POS_to_L3
Everything was checked into svn: /ligo/svncommon/SusSVN/sus/trunk/QUAD/H2/ETMY/Common/Data/2012-08-27_H2SUSETMY_M0_L1_L2_POS_to_L3
[Update 13:30-ish 28/Aug/2012]
The plots are now normalized by the L2L element of the drivealign matrix, as that was 1 for M0 and L2 (as it should be) but 10 for L1 for whatever reason.
Two things that are obvious from the plots:
(Updated 13:30-ish Pacific, 28/Aug/2012) 1. L2 drive is working. It is about a factor of 120 or so weaker than L1, and L1 is about a factor of 6 weaker than M0 (see page 1).
1. L2 drive is working. It is about a factor of 12 or so weaker than L1, and L1 is about a factor of 60 weaker than M0 (see page 1).
2. Cavity length to angle coupling could be problematic at resonances (see page 4). At DC for M0, it seems to be 0.1rad/m in a ball park, and and even if we feed back 1um RMS this is 0.1urad RMS, which sounds OK.
One thing that is not obvious from the plot:
For L2 drive, I had to use a ridiculously large excitation (+-120000 counts, half about a quarter of the range of 18bit DAC considering the output matrix of 0.25) with ridiculously long integration time (e.g. 160 seconds) to get a good coherence for f>1Hz. The background noise is too large.
This practically means that, as others pointed out, L2 is going to be railing if the ALS signal is fed back to L2 with a UGF of 1 Hz.
0.1Hz might be possible, but 1Hz, not likely.
Other things:
When the measurement was done, L2 stage driver FM2/3/5/6/7 were on while FM1 was off.
EUL2OSEM output matrix elements for M0 (for two lower face coils F2 and F3) were (0.5, 0.5).
EUL2OSEM output matrix elements for L1 and L2 (for all four coils) were both 0.25*(1, 1, 1, 1).
Update Aug/31/2012
In the above entry,
"When the measurement was done, L2 stage driver FM2/3/5/6/7 were on while FM1 was off."
this was incorrect but I cannot edit it any more, it seems. It should read
" L2 stage driver FM2/3/6/7/8 were on while FM1 was off."
J. Kissel, B. Shapiro I attached plots comparing Keita's transfer functions to what I expect from the model. Executive summary: 5 of the 9 transfer functions measured match my model exquisitely -- All L to L TFs, and the TOP to TST, and PUM to TST L to P TFs. Of the remaining TFs: I don't expect the model to predict the L to Y coupling well at all, but I'm still baffled as to why the UIM to TST L to P transfer function doesn't match up. Comments / questions / concerns welcome. I really haven't yet been able to get a warm and fuzzy feeling about a lot of this data. So, take it with a grain of salt. You'll notice that among the series of plots is the predicted maximum range for each stage. Please don't read too much into these numbers, I haven't yet verified them against Norna's numbers (see T1100595), taking into to account the differences between her numbers and mine (mostly the maximum range of the coil driver, updated to use the real, recently measured, transconductance of the coil drivers times the 10 [V] DAC range.) BUT I know that frequency response is accurate, because it uses the latest and greatest measured responses. Notes / Details: - There are fudge factors that I don't yet understand. They're explicitly called out in the legend, but they're summarized here: %L P Y driveAlignGain = [-1 -1 -1;... % M0 -5 -5 -5;... % L1 1 1 1]; % L2 meas(iStage,iDOF).tf = meas(iStage,iDOF).tf / driveAlignGain(iStage,iDOF); As Keita mentions, I expect the L1/UIM fudge factor to be 10 not 5, from the driveAlign gain. I'm NOT really that surprised that we got the sign wrong on M0 and L1, but I don't know yet where it lies. - The modeled M0-only damping loops are not *exactly* representative of what Matt tuned a month or 3 ago, but they should be close enough. I expect the overall gain to be different, and I expect the low frequency bump filters to be different, but otherwise they should match pretty well.