Reports 1-1 of 1 Clear search Modify search
DetChar (General)
hirotaka.yuzurihara - 16:50 Thursday 20 June 2024 (30013) Print this report
Usage manual was updated for the latest Omicron (v4.1) 

[Ishikawa(Aogaku), Oshino, Yuzurihara]

We tried using the latest Omicron (v4.1) installed by the IGWN conda and found that it's important to select the output format `hdf5` (not `root` format) to obtain almost the same triggers as the Omicron (v3.2). See the attached plots.
I updated the usage manual for Omicron (v4.1) to include an example and caution about the output format.

Detailed memo (for future workers)

  • For the previous Omicron (v3.2), we often used the output format of XML. But, the output format of XML was deprecated in the latest Omicron (v4.1). Omicron (v4.1) has two options for output format: root and hdf5. Initially we used `root` format and found that the number of triggers is significantly different between the `root` format in Omicron (v4.1) and `xml` format in Omicron (v3.2). This is because the clustering option will not be used in `root` format. (see the Omicron document) Due to this reason, I recommend using the `hdf5` output format instead of `root` in Omicron (v4.1).

  • We confirmed that the numbers of triggers are almost consistent between the `hdf5` format in Omicron (v4.1) and `hdf5` format in Omicron (v3.2). Actually, there is a slight difference. I guess the clustering method or something internal was updated over the version-up.

  • Currently, the Omicron with the old version (<<< v4.1) is running for the summary page. At some point in the future, we need to consider replacing the Omicron with a newer (or the latest) one.

Images attached to this report
Comments to this report:
shoichi.oshino - 9:00 Friday 21 June 2024 (30033) Print this report
We also compared the difference between 'root' and 'hdf5' result of v4.1 omicron.

Without the clustering option, the number of triggers of 'root' and 'hdf5' are the same.
However, for the values of snr, phase, and amplitude, the 14~15 decimal places differ. This is outside the precision range of the double type, so it is not considered to be a problem.
Then enable the clustering option and set the CLUSTERDT value to 0.1 and the 'hdf5' file will be written with the clustered results.
In this case, time, frequency, tstart, tend, snr, q, amplitude, and phase are the values of the trigger with the highest SNR in the cluster, fstart is the value of the lowest fstart in the cluster, and fend is the value of the highest fend in the cluster.
takahiro.yamamoto - 9:44 Friday 21 June 2024 (30035) Print this report

We confirmed that the numbers of triggers are almost consistent between the `hdf5` format in Omicron (v4.1) and `hdf5` format in Omicron (v3.2). Actually, there is a slight difference. I guess the clustering method or something internal was updated over the version-up.
Would you check the difference of triggerd glitches without clustering in new and old versions? Difference in triggered glitches may affect a significance of the coincidence between two channels. This means that veto-segment by hveto may depends on the omicron version. If difference comes from the bug fix in old version or the improvement (on ROC etc.) of omicron, it's no problem. But if not so, it's not so preferred situation.

hirotaka.yuzurihara - 17:44 Friday 21 June 2024 (30048) Print this report

I counted the number of triggers during the 60s chunks by hand. The trigger numbers by Omicron of both versions are always different for each chunk.

Details

  • I generated triggers by using the Omicron of the new and old versions without the clustering option. I randomly selected the data (Gps: 1368964818~1368965818).
  • From the log message, I counted the number of triggers during the 60s chunks. I attached the result (Fig).
  • The total number of triggers is 75970 for the new version and 75672 for the old version. The trigger numbers by Omicron of both versions are always different for each chunk.
  • I'm not sure what caused the differences. I will ask the developer about the update points from the old version to the new version and which version we should use.
Images attached to this comment
hirotaka.yuzurihara - 15:53 Thursday 27 June 2024 (30101) Print this report

[Ishikawa, Yuzu]

Last week, I asked the developer(Florent Robinet) about the issue of the different results with different versions.

  • The past klog post includes the the typo (sorry). Actually, I used the Omicron (v3.1) instead of v3.2. (Fig)
  • The release note of Omicron is available at https://git.ligo.org/virgo/virgoapp/Omicron/-/releases . The developer recommended that we use the latest version (v4.1) because the bug was reported in v3.1 and was fixed in v3.2.
  • Although the developer mentioned that the number of triggers should be consistent with the hdf5 format of v4.1 and the hdf5 format of v3.2 (the version after fixing the bug), I could not reproduce the consistent results.
Images attached to this comment
Search Help
×

Warning

×