Reports 1-1 of 1 Clear search Modify search
CAL (General)
takahiro.yamamoto - 7:23 Wednesday 11 December 2019 (12063) Print this report
Computational error check
[Darkhan, Yokozawa, Yamamoto]

We checked the current computational error on K1:CAL-CS_PROC_DARM_DISPLACEMENT_DQ (DISPLACEMENT) and K1:CAL-CS_PROC_C00_STRAIN_DBL_DQ (STRAIN).

[Related with Sensitivity monitor]
At first we tried to increase whitening filter stage for DISPLACEMENT.
DISPLACEMENT is saved as a single precision floating point number with 4-stage z1;p10 whitening filter.
This channel is used for monitoring on DTT sensitivity monitor in the control room.

After increasing the filter as 5-stage z1;p10, but DISPLACEMENT but there is no changes on the sensitivity monitor (Attachment 1).
This means that the 4-stage whitening filter is enough for current sensivity.

[Related with Inspiral Range]
Next we tried to apply the HPF for STRAIN signal.
STRAIN is saved as a double precision floating point number without any supplemental filters such as whitening filters.
This channel is used for computing Inspiral Range and provided online h(t) mainly for the data analysis activities.

We applied high pass filters (cutoff=7Hz, n=2 and 4) for STRAIN channel.
Sensitivity contamination is drastically improved with a n=2 filter and there is no improvement by changing from n=2 filter to n=4 filter (Attachment 2).
So n=2 filter is enough for current sensitivity.

[Updated Inspiral Ranges]
After filter applying, Inspral range reach ~20kpc.
Unfortunately we cannot apply the 2nd whitening filter stage for AS17 PD now.
I calculated by using the past data around 6am on Dec. 10th (in this time, the 2nd whitening filter stage was turned ON), inspired range reached around 40kpc.

Images attached to this report
Comments to this report:
tomotada.akutsu - 8:17 Wednesday 11 December 2019 (12064) Print this report


By the way, does the same matter happen when we do other FFT-like calculations such as coherence with the DARM signal or similar wide-range singlas? How about bruco, for example??

takahiro.yamamoto - 10:14 Wednesday 11 December 2019 (12066) Print this report
Bruco and some similar tools were affected this issue if K1:CAL-CS_PROC_C00_STRAIN_DBL_DQ was used.
If K1:CAL-CS_PROC_DARM_DISPLACEMENT_DQ or K1:LSC-DARM_IN1_DQ were used, they were not affected.
(Basically, these three channels has the same information only for the coherence analysis.)

I asked Yokozawa-san that which channel he used, and his answer is DARM_DISPLACEMENT and diaggui cannot treat C00_STRAIN (double precision).
So fortunately many measurements were not affected by this issue in this time.

But we must take care about this issue. We have a 10^{3}~10^{4} times improvement of sensitivity until achieving design sensitivity.
So dynamic range will also increase 10^{3}~10^{4} times.

darkhan.tuyenbayev - 12:33 Thursday 12 December 2019 (12090) Print this report

TYamamoto, Sawada, Darkhan,

The CALCS model was updated (Dec 10-11 night): a filter block has been added in front of the STRAIN output.

This filter block has been populated with a high-pass (7 Hz) 2nd order butterworth filter to reduce the digital dinamic range of the output. The detailes of the modifications are shown on the attached screenshots.

The CAL_LIB library and the updated MEDM screen were committed to SVN.

Images attached to this comment
Search Help