Summary
- DIFFp line amplitude need to be reduced by at least a factor 1.5 to avoid glitches due to saturation
- Would be good to check if saturation flags are able to veto these glitches
- A sin^2 transition instead of linear could reduce the amplitude of glitches, maybe by a large factor.
- Updating the blending filters of the B1 photodiode could reduce the amplitude of the glitches by a factor few.
There have been periods of loud glitches that were reducing the range over the weekend. These were reported by the operator.
Figure 1. Many of them are due to B1 photodiode audio channel saturation. The B1 fluctuations are dominated by the DIFFp lines (4.8Hz and 5.2Hz) and by the SDB1/OMC TX line (4.3Hz)
Figure 2. A week ago the B1 peak to peak fluctuations were about a factor 2 smaller, but now the fluctuations are all the time close to saturation.
Figure 3. There are glitches also without saturation, so there might be an additional problem than just B1 photodiode saturation.
Figure 4 Compares the Blended PD signal and the DC channel spectrum. When the audio channel saturates the DARM (B1) signal uses the DC channel only, which is much more noisy, hence a glitch is unavoidable. Hower the spectrum of the glitches looks very different.
Figure 5 shows a zoom on transition when the switch from blended to DC channel happens. It needs to be fast, but also has sharp kinks. Using a sin^2 shaped relay instead of linear could reduce the glitch amplitude due to the transient.
FIgure 6 the blending filter is not very good, it deviates by more than 10% in the 2-3Hz region. This could makes the transition between DC and blend channels larger, even though it is actually very small (0.1% as shown on figure 5).
