Last night we had a lot of problems with fast ALS glitches. Today Richard and co went around and inspected the fibers, tightened connections, and reduced strain in some areas. We've had no problems with this, but since its so tough to diagnose, it would be good to develop some code to find this glitching issue and report it on the ALS screen as a kind of Fault condition. The next time that the wind blows some sand on the fiber or a coyote bumps the laser safety tag, the glitching might come back.
In the second-trend plot, you can see that there are a lot of downward dips in the Green Y transmission, but not the X. We think that this is a fast phase shift in the fiber which is recovered by the FSS at the end.
In the DTT time series plot:
lower left: 3 minutes of glitches
upper left: time series AC coupled at 80 Hz
upper/lower right: zoomed in time series
From the zoomed in time series, you can see that the glitch has a time scale of ~2-5 ms.
It would be a very useful thing to make a tool that could run while the arms are locked on green and report on the number of such glitches in each arm. As a second step, being able to find the glitch in the ALS-X/Y_FIBR_ERR_OUT_DQ would be good. If we assume the glitch corresponds to 1/2 of the green linewidth...
By comparison, the free running NPRO noise should have an RMS of a few kHz at this timescale, so it may not be possible to see it with simple bandpass filtering.
Green arm linewidth is 275 Hz.
Attached is a plot that shows that at least when the laser is locked to the arm, we can see these glitches in the PLL signal when the arm is locked in green. This is probably since the arm locking feeds back to the PLL. We just tried tried a couple of times when the glitching was very bad to unlock the arm, and we don't see anything obvious in the PLL signals with the arm unlocked.
For some background/history:
These seem very similar to the glitches which were attributed to the laser safety tag hitting the fiber 36645.
Here is a very similar request from 2015 when we had these problems, perhaps some of this code could be restarted and or added to summary pages. 17576
