CERN Accelerating science

Prototyping in the SPS using pickups equipped with LHC electronics

Meeting held on Friday 26-11-2002 - reported by Mike Lamont

Present : Lars Jensen, Joerg Wenninger, Mike Lamont, Rhodri Jones, Jens Andersson, Kris Kostro, Jean-Jacques Gras, Michel Jonker, Quentin King, Thijs Wijnands, Marianna Zuin.

Agenda : MD results, closing the loop.

MD results

MD note to be produced with following sections:

• Introduction and motivation
• Description of system architecture
• MD setup: beams, conditions
• Network transmission results
• BPM behaviour and calibration
• Behaviour of closed orbit
• Conclusion & future plans

To be shown:

• Extrapolation from MOPOS measurements taken at the same time on the flat bottom to be used in an attempt to cross-calibrate new pickups (offsets/scaling factors).
• Orbit stability through the cycle
• Orbit reproducibility
• Noise Comparison of 1 turn v 255 turn average

Jorg has data (also available on bursa). He will analyse data and knock out some plots to illustrate the above.

The log files are in /nfs/bursa/opdata/bpmdata/datalogs/ with obvious naming and format (time [ms], 2 (# of pickups), H1, H2, V1, V2, H1-sigma, H2-sigma, V1-sigma, V2-sigma). 10 cycles (~200 seconds) per file.

Network

Jens presented some results of investigations he is performing to look at the situation involving many machines. Plans to have access to 20/30 machines including some on the new PS switched network. Triggers provided by GPS and the PS timing system.

* Current test setup, using GPS cards for trigger (LEMO) and timestamps:

http://proj-lhcfeedback.web.cern.ch/proj-lhcfeedback/images/test_lynxos.gif

* Results for 30 packets each from 2 machines (typically <25 ms, except when
disturbed by NFS traffic, but note the slow and shared network, 10Base-T):

http://proj-lhcfeedback.web.cern.ch/proj-lhcfeedback/images/recv_network.gif

* Practical problems with PS machines, such as shutdown and access to
triggering & timing, but working on that.

He aims also to enable multicast and test sending corrections (to crates that have joined the multicast). Utility of multicast debated at length. Will try it but "probably easiest to forget"... (Update Dec. 2nd: Received kernel fix from CES, multicast now works as it should on LynxOS.)

Also mentioned clustering of RT data to improved transmission and processing time.

Update of SPS network not planned.

Marianna, and thus IT, is aware of what's going on. Evidenced by her following a long discussion on the internals of switches, routers, protocols etc. between other members of the group.

Closing the loop

MuGeF - RT channels: Available, but might have to be activated (facility provided by John Brazier a wee while ago). Quentin suggests starting with a stripped down version of ROCS to access MuGeFs in question. Jens to follow up on this. Lars has examples of call structure from Q-loop prototype, to be forwarded.

Input data: Energy, calibration factors to be hacked in for the moment. The "generic architecture" can wait a while

Controller: PID thought to be sufficient. Jorg has first version coded based on the stuff Thijs did. Will use this in first instance.

Tying it all together: Jens fingered.