In preparation for acceptance of H1 SUS TMSY, I've been asked to model its performance. I accepted given that we've got a state-space model and parameter set that's been confirmed by measurement and it's straight-forward to adopt the loop performance figures of merit from the software suite developed for the QUAD and BSFM Level 2 designs. Here I merely show results that confirm the model's accuracy using the current filters before I move forward with the Level 2 filter design.
The new software suite for the TMTS (which will be useful for the OMCS when it comes online soon as well), can be found here:
${SusSVN}/sus/trunk/TMTS/Common/FilterDesign/
design_damping_TMTS_20130313.m (script where overall design is done)
- plottmtsdampingcontroldesign.m (function that takes in newly designed filters and computes the performance figures of merit)
- plottmtsactuatornoise.m (sub-function that computes the actuator noise component of the model)
In addition, I've added the TMTS requirements (based on Fig 18 of E1100537) to the function
${SusSVN}/sus/trunk/Common/MatlabTools/SUS_reqs.m
I won't bore you with all of the plots that this suite produces since this has only been run on the Level 1 design of the filters, but I will post results of this morning's transfer functions, and a tease of why the TMTS filters need some tweaking. I'll focus on Pitch, since that (with Yaw) are the DOFs with the most stringent requirements for the TMS, and since L and P were the only DOFs I had time to measure.
Three plots:
(1) dampingfilters_TMTS_20130313_loopmeas_P.pdf A comparison between the modeled and measured open loop gain transfer function. It confirms the stability, and we now see the model is only off by the usual known quantities:
- ~50% under in magnitude (mostly likely due to a systematic scale factor error in the model calibration -- no biggie),
- a tiny fraction in resonant frequencies (due to modeling the "ideal" suspension properties as opposed to the "real" properties)
- some unmodeled time delay-like feature in the phase.
(2) dampingfilters_TMTS_20130313_loopdesign_P.pdf A loop design plot showing all of the useful figures of merit, the open loop gain transfer function (G), the suppresion (1 / (1+G)), and the closed loop gain transfer function (G / (1+G)). Note, I use the standard convention for loop gain with the explicit -1, such that the stability criteria is that lower unity gain crossings must stay away from +180, and upper unity gain crossings must stay away from -180, with no phase wraps in between the first LUGF and the last UUGF.
(3) dampingfilters_TMTS_20130313_totalbudget_P.pdf A noise budget based on the loop design.
------
For the record, the filters currently in place (and ON) are (using [zeros:poles:gain] notation, and "pair" to represent a complex pair of zeros or poles at frequency of the first argument, and phase of the second. Both arguments are rounded to the nearest integer for brevity)
L FM1 "damp30" [0:30,100:1], FM5 "from_um" [::43.478], FM10 "ELF10" [pair(16,86),pair(36,56):pair(6,51),pair(10,79):1], G = -20.0
T FM1 "damp30" [0:30,100:1], FM5 "from_um" [::43.478], FM10 "ELF10" [pair(16,86),pair(36,56):pair(6,51),pair(10,79):1], G = -10.0
V FM1 "damp30" [0,1:0.3,30,100:3.51521], FM5 "from_um" [::43.478], FM10 "ELF10" [pair(16,86),pair(36,56):pair(6,51),pair(10,79):1], G = -20.0
R FM1 "damp30" [0,1:0.3,30,100:3.51521], FM5 "from_um" [::43.478], FM10 "ELF10" [pair(16,86),pair(36,56):pair(6,51),pair(10,79):1], G = -0.05
P FM1 "damp30" [0,1:0.3,30,100:3.51521], FM5 "from_um" [::43.478], FM10 "ELF10" [pair(16,86),pair(36,56):pair(6,51),pair(10,79):1], G = -0.5
Y FM1 "damp30" [0,1:0.3,30,100:3.51521], FM5 "from_um" [::43.478], FM10 "ELF10" [pair(16,86),pair(36,56):pair(6,51),pair(10,79):1], G = -0.1
which is notably different that when I'd spruced up the TMTS, SUS style back in March (see LHO aLOG 5754, and LHO aLOG 5781), when the different Yaw filter in FM2, and any boosts that were found in FM2 were ON. I didn't see any aLOG of this configuration change, so my guess is this was an old configuration that stuck after a reboot (which is surprising because I specifically captured a new safe...)
I've copied the foton file representing this configuration over to the userapps repo, and committed it to rev 4641
208-69-129-56:filterfiles kissel$ svn info H1SUSTMSY.txt
Path: H1SUSTMSY.txt
Name: H1SUSTMSY.txt
URL: https://redoubt.ligo-wa.caltech.edu/svn/cds_user_apps/trunk/sus/h1/filterfiles/H1SUSTMSY.txt
Repository Root: https://redoubt.ligo-wa.caltech.edu/svn/cds_user_apps
Repository UUID: e3bfa956-ff0e-4416-9af7-00b7a258cde5
Revision: 4650
Node Kind: file
Schedule: normal
Last Changed Author: jeffrey.kissel@LIGO.ORG
Last Changed Rev: 4641
Last Changed Date: 2013-06-06 14:35:30 -0400 (Thu, 06 Jun 2013)
Text Last Updated: 2013-06-06 14:45:31 -0400 (Thu, 06 Jun 2013)
Checksum: e5032b936a0f52fd58cfd2ec15dd771b
------
Also, the data (for L and P only thus far) was taken with the following templates:
2013-06-06_1511_H1SUSTMSY_M1_L_WhiteNoise_OLGTF.xml
2013-06-06_1511_H1SUSTMSY_M1_P_WhiteNoise_OLGTF.xml
which notably improve on the standard SUS white noise DTT excitation: in order to get more coherence at low frequency, I added a zpk([0.5;0.5],[0.05;0.05],100,"n") to the Filter field in the Excitation Tab of the measurement. This adds a factor of 100 gain at low frequency (filtered out by 1 [Hz]) where the coherence is the worst, and we're not plagued with DAC saturations because on the anti-dewhitening filters in the COILOUTF banks (which start adding gain above 1 [Hz]). This way I can jack up the overall gain of the measurement at low frequency where we need it, without saturating the DAC nearly as quickly.
