I have looked at all the A2L data that we have since the last time the alignment was significantly changed, which was Monday afternoon after the PSL PZT work (alog 31951). This is the first attached plot.
The first data point is a bit different than the rest, although I'm not totally sure why. Other than that, we're mostly holding our spot positions quite constant. The 3rd-to-last point, taken in the middle of the overnight lock stretch (alog 32004) shows a bit of a spot difference on ETMX, particularly in yaw, but other than that we're pretty solid.
For the next ~week, I'd like operators to run the test mass a2l script (a2l_min_lho.py) about once per day, so that we can track the spot positions a bit. After that, we'll move to our observing run standard of running a2l once a week as part of Tuesday maintenence.
The second attached plot is just the last 2 points from the current lock. First point was taken immediately upon lock, second was take about 30 min into the lock. The maximum spot movement in the figure appears to be about 0.2mm, but I think that is within the error of the A2L measurement. I can't find it right now, but once upon a time I ran A2L 5 or 7 times in a row to see how consistent the answer is, and I think I remember the stdev was about 0.3mm.
The point of the second plot is that at 30W, it doesn't seem to make a big difference if we run a2l immediately or a little later, so we can run it for our once-a-days as soon as we lock, or when we're otherwise out of Observe, and don't have to hold off on going to Observe just for A2L.
In case you don't have it memorized, here's the location of the A2L script:
- cd /opt/rtcds/userapps/release/isc/common/scripts/decoup
- ./a2l_min_LHO.py
A2L: How to know if it's good or bad at the moment.
Here is a dtt template to passively measure a2l quality: /opt/rtcds/userapps/release/isc/common/scripts/decoup/DARM_a2l_passive.xml
It measures the coherence between DARM and ASC drive to all test masses using 404 seconds worth of data.
All references started 25 seconds or so after the last a2l was finished and 9 or 10 seconds before the intent bit was set (GPS 116467290).
"Now" is actually about 15:00 UTC, 7AM PT, and you can see that the coherence at around 20Hz (where the ASC feedback to TM starts to be dominated by the sensing noise) significantly worse, and DARM itself was also worse, so you can say that the a2l was worse AT THIS PARTICULAR POINT IN TIME.
Thing is, this might slowly drift around and go better or worse. You can run this template for many points in time (for example each hour), and if the coherence seems to be consistently worse than right after a2l, you know that we need a2l. (A better approach is to write a script to plot the coherence as a time series, which is a good project for fellows.)
If it is repeatedly observed over multiple lock stretches (without running a2l) that the coherence starts small at the beginning of lock and becomes larger an hour or two into the lock, that's the sign that we need to run a2l an hour or two after the lock.
