I replaced the old ALS QPD output matrices (measured empirically by Elli and Thomas several weeks ago) with new ones based on Cartesian coordinates. I obtained these by ray-tracing on the ALS table. Then I measured the input matrices with a Matlab script (/svn/cdsutils/trunk/ALS).
Now the inputs of the IP_POS and IP_ANG filter modules should be calibrated in meters and radians.
These are the matrices:
INPIT =
-0.001934944012837 0.000295830132597
-0.000025683755141 -0.000141436038465
INYAW =
0.001338191005451 -0.000236296533156
-0.000010597031521 0.000352368394300
OUTPIT =OUTYAW=
1.0e+07 *
-0.746993867828206 2.240981603484618
0.048486940975147 -0.800820822925440
The loops have UGF of about 10 Hz with these gains:
IP_POS_PIT= -8; IP_ANG_PIT=-8; IP_POS_YAW=-8; IP_ANF_YAW=-4
Attached are some spectra measured with the loops either open or closed.
The TMS table relative lateral stability seems to comply with the requirements (=100urad RMS). On the other hand the angular stability seems to be a bit worse than desired (=1urad RMS).
Long term stability (12+ hrs) still has to be evaluated.
There seems to be a factor 2-3 gain peaking at 20Hz. The gain peaking seems to be responsible for most of the remaining rms. There are also lines at 60Hz and just below that are fairly large in the spectrum. Is this real motion or just sensor noise? Also: pdf of 2nd file
I multiplied the input matrices by 1e6 so that we read out um and urad at the input of the IP_POS and IP_ANG filter modules.
On all modules I enabled a filter called "cal" that divides that factor out.