Skip to content

Add PSD to Initial coinc.xml

Instead of uploading the psd.xml.gz file in a separate GraceDB upload, this change now appends the PSD as a REAL8FrequencySeries LIGOLW array in the coinc.xml file - one for each IFO.

The PSD calculation itself has been left unchanged.

Some additional code not used (i.e. commented or under an if False block) has also been trimmed.

This merge request implements #23 (closed).

Details

Examples of the final coinc.xml are provided below: H1L1V1_1187008602_3_369.xml H1L1V1_1187008603_3_606.xml

There are two minor points we may wish to address in our internal review:

  1. Changes in SPIIR's alert latency (i.e. GraceDB upload time).
  2. PSD Epoch Time and Calculation.

1. Latency

The construction of the PSD LIGOLW arrays have about a ~0.025s latency (estimated with a simple datetime.now() call before and after the code block at line 834 to 856).

At time of writing we have not tested the latency of GraceDB upload (i.e. one upload with a larger coinc.xml in this branch, compared to a coinc.xml and a psd.xml.gz uploaded separately) - we have only tested that the output schema is correct on OzStar.

We can test upload latency with a run on CIT if internal review requests it before merging into the current dev branch.

2. PSD Epoch Time

Currently we get the PSD for each IFO from the GStreamer pipeline at the time we construct the XML document and attach an epoch time to this PSD (see Line #845) - this is exactly as it was done in O3.

dfddc4b9

It may be the case that the PSD associated with the upload should be as it was calculated back at the time of the trigger, rather than at upload time (therefore we would need to store the previously calculated PSD and its corresponding epoch time as is done for the SNR series upload).

However, this introduces a lot of complexities in regard to state and memory management, and we should typically expect the PSD to the similar enough as we do not expect the PSD to change over small timescales of less than a minute.

Regardless, we figured it would be worth flagging, even if no changes are required.

Edited by Daniel Tang

Merge request reports