jeffrey.kissel@LIGO.ORG - posted 16:17, Friday 23 August 2019 - last comment - 16:49, Friday 23 August 2019(51472)
Re-occuring ALS EX / EY SDF DIFF Caused by Floating Point Precision Comparisons. Fixed By Rounding Values to 1e-6 Precision
J. Kissel There have been many times since we re-adjusted the initial alignment settings (last done Thursday Aug 15 2019; see LHO aLOG 51307), that the ALS EX and ALS EY front-end models show SDF differences at the 1e-17 - 1e-18 level. This is a result of the python scripts we use to set the initial alignment references -- in particular the transmon (TMS) green QPD offsets -- /opt/rtcds/userapps/trunk/als/h1/scripts/ setEndGreenQPDOffsets_XARM.py setEndGreenQPDOffsets_YARM.py not rounding before installing to any reasonable precision before writing the numbers to EPICs. As a quick fix, I manually rounded and entered the numbers in to the respective EPICs channels. Attached are screenshots of the SDF diffs after rounding, mostly to prove to you and me that I didn't make a typo in doing so (which would botch the initial/global alignment references -- a think I *definitely* don't want to do at 4p PDT on a Friday). The long term fix will be to edit the above scripts, but since we don't plan on updating the initial alignment references again any time soon, the quick fix should suffice until we get to editing the script.
Images attached to this report
Comments related to this report
J. Kissel, S. Dwyer The future is now! With the watchful eye of Sheila, I edited the above mentioned X and Y arm codes such that once the transmitted (green) Arm Power is above 0.9 [normalized counts], and the I phase outputs of the (green) WFS A and B are less than 10 [DAC counts], then (a) read the current offsets, (b) round the QPD offsets to a precision of 1e-6 (c) write those rounded offsets, and (d) break out of the while loop (a), (b), and (c) are all done in one line via the magic of ezca python bindings, and (d) is new -- the script used to just ask the user to CTRL+C out of the script. This code change is completely untested, and again we likely won't use it for another 3-6 months, but hopefully I've left enough breakcrumbs in - this aLOG, - the instructions that reference these scripts, LHO aLOG 50968, - the comments of the code, and - in the log of the svn commit that people will come find me when it doesn't work. Apologies in advance!