Greg Mendell, Patrick Brockill, Aaron Viets I restarted the primary and redundant GDS pipelines at GPS time 1164999500. The latency was ~10 seconds and the CPU usage is ~80%, both in the usual range. This restart picked up gstlal-calibration-1.1.0, which includes a bug fix that should solve the problem that Greg discovered last week with non-identical h(t) produced by primary vs redundant pipelines. This required two changes: 1) Removal of start-time dependence from the demodulation routine used in computing the time-dependent corrections. 2) Removal of the Gstreamer stock element audiocheblimit, which I replaced with lal_firbank. The audiocheblimit element appears to have a bug which makes its output irreproducible when run on the very same data in some instances. With this fix, h(t) is identical (as far as I can tell--to 16 digits) between runs of different start times after sufficient filter settling time. The settling time depends on the state of the interferometer at the time of start (whether it's locked and the coherence of the calibration lines). Also incorporated into this version is additional gating in the kappa_tst calculation. See this aLOG for more information: https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=31911 The first two plots show the h(t) spectrum compared between GDS and CALCS. The third plot is a time series of the kappas. The filters have not been changed. See this aLOG for information on the filters: https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=31712