J. Kissel

Though we're confident that the only way to improve the X-arm Test Mass angular performance is to improve the BSC-ISI performance between 0.3 and 0.7 [Hz], I tried tweaking the Level 2.1 Damping Loop design to see if I could do any better, by increasing the Q of the boost filters on the L and P degrees of freedom. Further, I moved the frequency of the P boost down from 0.56 [Hz] to 0.51 [Hz] to account for the difference between the modelled and measured frequency of this mode. Regrettably, I could not improve the strength (at the resonances) of the boosts much more than 5 to 10 [dB] without destroying the the stability of the L & P loops -- I'd already pushed the phase and gain margins of the *lower* unity gain frequency pretty far. As such, the improvement was only in the sharp frequency regions over which I'd focused the boost, but little-to-no change in the RMS. 

We'll continue to work on improving the performance of the ISIs. 

For posterity, I post all of the design information and performance comparisons. The good news is, that my MIMO model of the QUADs can now accurately predict the optic angular motion -- especially Pitch -- at all frequencies where not limited by its sensor noise. Check out the pg 2 of the third attachment for proof!

Details:
---------
allquads_2014-01-31_AllGoodFibers_P.pdf 
     -- Shows a collection of transfer functions of all monolithic suspensions. It shows that compared against the modeled first, top mass, P2P mode at 0.56 [Hz], all the measurements show that this mode is at 0.51 pm 0.01 [Hz]. This is what motivated me to focus the sharp P boost at a lower frequency.
2014-01-31_H1SUSETMX_M0_DAMP_Filters.pdf
     -- Bode plots comparing the 2013-06-14, Level 2.1 Filters against the new 2014-01-31 Sharp Boost filters. 
dampingfilters_QUAD_2013-06-14_Level2p1_2014-01-30_H1ISIETMX_Seismic_SelectPlots.pdf
     -- Using current seismic data input, this is a few select plots from the loop design script for the 2013-06-14 Level 2.1 filters. As mentioned above, it predicts the measured P motion extremely well below ~1.5 [Hz]. Interestingly, the Y prediction is not perfect, but I suspect that below 0.5 [Hz] the signal is dominated by L and P motion of test mass, confused as Y. Especially because the two extra modes that appear at ... you guessed it 0.44 and 0.56 [Hz].
dampingfilters_QUAD_2014-01-30_Level2p1_RealSeismic_SelectPlots.pdf
     -- Selected performance plots with the new sharper boost. Bare in mind that I've used the *model* for the predicted transfer functions, which don't get the first P mode at the right frequency. That's why things appear like they will be unstable, and all sort of gain peaky near the lower unity gain frequency. 
dampingfilters_comparison_2013-06-14vs2014-01-31.pdf
     -- A comparison between the two designs. 
2014-01-31_H1SUSETMX_Level2p1vsSharpBoost_Performance_ASDs.pdf
     -- A comparison of the optical lever performance between the two configurations. Of course, the spectra were taken at different times of day, so above ~0.6 [Hz] the input motion is a little different, but below 0.6 [Hz] the motion is the *same*,  and one can see the expected change in shape of the hump, but no change in RMS. Oh well.


Design Scripts:
SusSVN/sus/trunk/QUAD/Common/FilterDesign/Scripts/
compare_quad_dampfilter_design_20140131_NormalvsSharp.m
design_damping_QUAD_20130614.m
design_damping_QUAD_20140123.m
plotquaddampingcontroldesign.m

Filters and Model saved to:
SusSVN/sus/trunk/QUAD/Common/FilterDesign/MatFiles/
dampingfilters_QUAD_2013-06-14.mat
dampingfilters_QUAD_2014-01-30.mat
dampingfilters_QUAD_2013-06-14_Level2p1_2014-01-30_H1ISIETMX_Seismic_model.mat
dampingfilters_QUAD_2013-06-14_Level2p1_RealSeismic_model.mat
dampingfilters_QUAD_2014-01-30_Level2p1_RealSeismic_model.mat