aLIGO LHO Logbook

H1 CDS

jonathan.hanks@LIGO.ORG - posted 08:31, Tuesday 01 October 2019 - last comment - 09:21, Tuesday 01 October 2019(52231)

Redundant recording of EPICS data, available on h1nds3 (WP 8381)

h1nds3 is acquiring all the epics channels that the regular daq acquires and is recording them as of 1253978688.

As a backup to the regular daqd system h1nds3:8088 can be used to read the epics data stored to frames (VE, FMCS, ...).  This is only the slow data, no fast data is being recorded here.

I will be finishing up some spot checks on the data to make sure they match the production stream.  If there are not problems seen we will leave the machine alone until the end of the vent.

Comments related to this report

jonathan.hanks@LIGO.ORG - 08:50, Tuesday 01 October 2019 (52232)

Link

I picked a channel from the VE overview and streamed the data from h1nds1 and h1nds3.  The data is the same, h1nds3 is able to stream it back a little faster.  This is the difference between directly writing the data on h1nds3 and traversal from h1dc0 to h1nds1 + some processing time involved in that transfer.

See attached screen shot.  The top is the feed from h1nds1, the bottom is from h1nds3.

Images attached to this comment

jonathan.hanks@LIGO.ORG - 09:15, Tuesday 01 October 2019 (52234)

Link

Some timing information for future reference.

The minute trend frames are taking around 4s to write, the raw frames are taking 2s to write.

The raw frames are 20MB. The second trends are 103MB.  I need a longer sample period to find the size for the minute trends (need to get a frame file that is full, not zero filled for the start).

I've attached a screen shot of a strip tool showing some stats on h1nds3.

Images attached to this comment

jonathan.hanks@LIGO.ORG - 09:21, Tuesday 01 October 2019 (52235)

Link

Issues I ran into with setting this up.

The main issue I ran into was with the ini file parser in the standalone_edc.  It did not like trailing spaces, so when it checked to make sure channels were only being allowed at 16Hz and as floats, trailing spaces gave it issues.  The parser should be replaced with the one daqd uses to parse the master file.  Another issue that I ran into in testing is that when DCU ids where not set the same between the ini file and the standalone edc, data wasn't recorded in the daqd.  This should be fixed by pulling the dcuid from the ini file.