NARIT visit April 2018. Questions and agenda - NewtonFund/NewTCS GitHub Wiki

Error handling and reporting

Some system somewhere (not the ACS) must have a state model that decides things like, "I cannot start the exposure until the focus gets to the correct position, but I can be moving the science fold pick-off mirror at the same time as the focus."

ACS detects reports state transitions. It could provide more information on the cause. e.g. AG timeout. Would this be of use or is this already handled by the TCS?

Publish/subscribe paradigm. I have no objection to pub/sub. My question is how to handle delivery failures i.e. subscriber not listening so ACS/TCS state models become out of sync.

AG guide packets are always interpreted as Az,Alt errors, never rotation, flexure or optic tip-tilt errors?

Provide NARIT use cases that we can collaboratively encode into this command set there in the meeting rather than just using our ListOfUseCases (UseCaseImplementations).

Clarify each side of the ICD's understanding and response to commands arriving out of sequence. (NodeID vs MsgSequence)

Do axis encoders need calibrating every night? I.e., the POINT CALIB command.

Look at TrackInit command

What is Az bearing?

Answers

What is demand cadence for controllers? Do not really know yet. It may be more complicated but for now we are agreed that the baseline is just three axis positions at of order 100Hz.

Scope of this ACS is TNT only? NARIT say yes. TNT only

NARIT says do not use tip-tilt M2 during the night. M2 is actively driven, but only to keep the star in the same place to compensate for flexure etc. ACS does not have any M2 control.

Need "track here" command? NARIT says no. They do not want "TRACK HERE" commend

NARIT say time stamps on all states (encoder values, weather values etc) are set in the command centre being the time the command centre received the data from the underlying source. Not a time stamp generated in the underlying source itself.

Expected timescale to use the second Nasmyth? NARIT says they want this ACS to handle both Nasmyth from the start. Second Nasmyth should be installed over the coming year.

Single autoguider or multiple autoguiders? NARIT say there is no AG for first year but but they want the ability for a single AG to exist. As soon as they get the second Nasmyth then they will install the AG as well.

NARIT will not use Windows event logger in the new software stack

Demands will be three axis positions "now" being sent at of order 100Hz. Baseline spec is 400Hz, but we do not know that exact number

We have two design paradigms on the table. One is a command 'set' and the other is to wrap all requests into just SET/GET requests. Discuss the pros and cons of each and agree on one. ACS command set has been re-cast in terms of Pakawat's SET/GET/SUBSCRIBE

Is the 'TCS' machine guaranteed to be a single physical hardware including both TCS and ACS functions? NARIT say Yes, all on one physical machine.

Is comms to/from the ACS always from the same place? I.e., are all comms to the ACS via some kind of command broker / command centre or can there be messages from distributed systems? Autoguiders? GUI? Axis Controllers? NARIT say the only think that ACS will ever talk to directly is the 'Command Centre'.

How fast does the UI want to poll the ACS? What is UI screen update rate? Are there ACS efficiency issues making polling or publish-subscribe preferable? Agreed baseline plan for axis demands subscribe at 400Hz and data feeds to the UI typically subscribe at 1Hz.

What is the cable wrap range for TNT? : [There is no unwrap command. There is TrackDuration in TrackInit, but do we need to be able to force an unwrap or force a maximum wrap in either direction? "Unwrap" is a TCS function, not ACS. Force an unwrap through the ACS by sending TrackInit with a long Duration, but note that does not necessarily correctly retain all offsets that have been applied. We calculate that given a 540deg range on the cable wrap, it is almost (not quite) impossible to ever need to force a wrap.] Data from the EOS delivery acceptance tests: Az ±220deg (440deg range). -224deg < Rot < 256deg (480deg range).

AG guide packets are always interpreted as Az,Alt errors, never rotation, flexure or optic tip-tilt errors. Agreed yes this is right.

Do axis encoders need calibrating every night? I.e., the POINT CALIB command. No. TNT does not use a nightly startup pointing test.

What is Az bearing? Definitely a roller bearing. Not 100% certain and will check for us, but pretty confident it is a simply, grease packed, caged roller bearing. Inserting grease at the nipple is the only routine maintenance. They generally always park in a fixed place and have never worried about denting the bearing surface there.

As well as agenda, who will attend? I wrote something about this a while ago, but cannot remember where. I am not sure if we are going to be able to get everyone together the entire time, but we do at least need access to input from individuals who represent i] software (Pakawat?), ii] controllers (TGI?), iii] night operations (a telescope operator?).