Since the boys found that the ETM dewhitening transient was kicking the IFO out of lock we
decided to investigate.
First, we wrote a script to diagnose and then tune the DC gain of the dewhitening filters'
digital compensation filter (a.k.a. FM9 or SimDW). It is in the scripts/SUS/ directory
and is called dwgaintuner. It puts in an offset on each coil's DAC channel and
then reads back the Vmon on the coil driver with the DWF on and off. It reports the ratio of these
voltages which you can then type into the FM9/SimDW filter's gain field. We learned that the
difference between the analog DWF path and the bypass path was ~3% (which is consistent with
what you expect from the use of 1% resistors). We need to repeat this for all of the rest of
the suspended optics except for MC1 and MC3.
This Vmon method is better than what's used at the sites so we will export this new technology.
The attached plot shows some switching transients with only the local damping on:
BLUE: Output of filter bank during an FM9 turn off. This is the transient which goes to the DAC.
The transients are mostly of the same magnitude as this.
RED: This is the input of the filter module during another such transient.
GREEN: Tried another switch; this time I filtered the time series in DTT by typing the SimDW into
the Triggered Time Series filter field. This should be simulating what comes out of the
output of the DW board - to convert to volts multiply by 15/32768.
PURPLE: Same kind of filtering as the GREEN, but with also a double 30 Hz highpass to remove the low
frequency damping control signals. You can see that the total transient is only ~5 counts
or ~1 mV at the coil driver output. This is comparable to the relative offset in the bypass
and filter paths.
|