Details
-
Type:
Story
-
Status: Invalid
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: ts_main_telescope
-
Story Points:0
-
Epic Link:
-
Sprint:TSSW Sprint - Apr 26 - May 10
-
Team:Telescope and Site
-
Urgent?:No
Description
It would be helpful to have a timestamp field added to the telemetry struct output by the low-level MT Rotator controller. My preferred format is that used by SAL: TAI unix seconds as double precision (float isn't accurate enough).
This would provide a reference point for reported velocity.
It would also allow computing a measured rotator velocity using (pos1 - pos0) /( t1 - t0) – which seems likely to avoid the oddities we see in the two actual velocities that are presently being reported.
I assume the low-level Rotator controller has easy access to TAI time; it certainly should, as it is necessary in order to properly handle tracking commands.
Attachments
Issue Links
Activity
If you want to use "(pos1 - pos0) /( t1 - t0)" to calculate the velocity, you may need to be very careful because this formula will be affected by the noise pretty much.
Thank you. I had forgotten about the header. Yes, that will do, though I would much rather have TAI time than unix (UTC-ish) time. TAI is our standard time system. See also DM-26451 where it is much more important.
To get TAI you just have to call to clock_gettime(CLOCK_TAI, &unixTAI).
Do you think (pos1 - pos0) /( t1 - t0) will be worse than the two velocities you are presently providing? It seems to avoid the weird behavior around 0 velocity in your graphs, and those two velocities are also noisy.
If you are willing to report TAI instead of standard unix time then can we use this ticket for that purpose? Feel free to edit the title and description. Or we can close this and open a separate ticket if you prefer.
As long as the time in the telemetry header is the correct time at which the position and velocity were measured, I agree it is much better to use the time in the header than to add a new time field to the telemetry.
The following is the timestamp maintained in the rotator Simulink model:
https://github.com/lsst-ts/ts_rotator_controller/blob/master/src/smlnkModel/rotator_smlnk_Controller.c#L778
Not sure this is what we want or not. May need to check with Simulink model to see the details.
Based on the code, this rotatorTelemetry_signal.Time_UnixSec is the same value of ddsTlm.header.mjd. Maybe this is the value you are looking for:
https://github.com/lsst-ts/ts_rotator_controller/blob/master/src/sys.c#L380