CERN Accelerating science

LHC real-time

See also Jens's page on SPS prototyping 

We will have large multipole swings during the injection plateau and snapback...tight physics constraints & the low tolerance to beam loss.  Beam stability will depend totally on feed-forward from the reference magnets, feedback on the key beam parameters tune, orbit, energy, chromaticity...

• Reference magnets: feed-forward of corrections to machine etc.
• Feedback-loops: Global Orbit, Local Orbit - e.g. orbit stability at collimators, Tune Chromaticity
• Real-time knobs: typically tune, orbit bumps extension of feedback, operator close loop, need...
• Display: Orbit Beam loss (not for abort, but display & possible feedback) Luminosity, Beam sizes, Lifetimes�
• Timing like actions (possibly) e.g. post-mortem trigger
• Distribution of real-time data - possible use of the LHC real time data by multiple users.

Tuesday 16th August 2004

Use Case: Orbit feedback during nominal cycle [write-up c/o Steve Page]

Tuesday 9th August 2004

Jens's presentation showing outline of implemented system and performance.

Ralph's presentation showing: the motivation for; the challenges of; details of LHC orbit feedback; and test bed results.
 

• Various applications of RT feedback, feed forward and display enumerated - agreed that common components be used where possible.
• Jens's present implementation takes as input setup info (input UDP , filter, matrix, output UDP config.) plus initial parameters (filter coefficients, matrix). The later are adjustable via CMW set to the server. Modular implementation within a server which runs as one process.
• Interesting performance analysis shown by Jens with results from 1 M cycles. Generally good results with occasional dropouts of 3-4 consecutive cycle every 20 minutes.
• Ralph explained (among other things) the proposed breakdown of functionality into controller and service unit.

Lengthy discussion...

Issues:

• Simultaneous use of acquisition channels - in particular 100 k turn orbit acquisition, the need to slice etc.
• Simultaneous use of acquisition channels - OP acquire
• Change of reference
• QOS - dedicated channel for RT - majority in favour.
• What data passes between slow controls and RT service unit/controller - see list below
• Limits on applied corrections, power converter limits etc.
• Control of the controller
• Reference management, catalogs, reference versus time (in a fill), rollback.
• Fill-to-fill feed forward of RT corrections, format, parameters etc.
• Hysteresis loop crossing in MCS correctors - around 1 Hz - complicated, do it at the high level
• RT feed to display from Service Unit. Not faster then 10 hZ , print, screen grab, scales, speed, ilog-jview (lionel).
• Jdataviewer...   have to sort this out. OK we don't use the Dataviewer.
• BLM at 1 Hz.  Concentrator required. Different types of monitors - turn by turn.  Spec required. Versus threshold, energy etc.
• Select set of elements show time evolution over ring.
• BCT. Bunch by bunch.
• Power converter - summary. State, reference, ... alarms. Everything gateway block all properties for publishing.
• Tune
• Luminosity
• Real-time knob - e.g orbit bump, disable appropriate region - watch correction rate. KISS. Change reference in real-time... One reference - Service Unit takes care of multiple inputs.
• Data distribution: DIP - CMW should be able handle UDP push to multiple clients.
• Standard HW for the controller - FPGA for the future... interface between SU and controller fixed.

Future

• Analysis of controller and Service Unit
• preliminary controller - long term tests.
• someone in controls to collaborate... prototype end 2005.

Failure scenarios (incomplete list)

1. BPMs go to the bad
2. Controller doesn't respond, Controller crashes
3. FE crashes
4. PC gateway failure
5. PC off

Agreed that we should Use Case use of RT in various operational scenarios to establish interplay between slow controls and RT channel.

Data exchange from business layer to RT:

• Optics change during squeeze
• Twiss/matrices
• Transfer functions
• Energy
• Pre-programmed I(t) in corrector circuits
• Circuit configuration
• Knob definitions
• Correction limits
• Reference orbits
• References in general.
• Control of the control
• Control loop parameters

Data exchange from RT to business layer:

• Trim history
• RT feed of corrections to high level

2 machines for SU (RT OS) and controller

Data management - big issue for Service Unit.

• 5th June 2003 Meeting to discuss RT in the context of the Multipole Factory: Notes plus outline 

Real-time meetings             

• 18th March 2003: Planned MD - first look at MF Notes
• Feb 2003: Publication of MD note (AB Note 2003-019)
• 26th November 2002: MD results Notes      Mariana's network test results from SM18 (ppt)
• 8th October 2002: MD Planning Notes
• 19th April 2002: Another meeting to discuss progress of SPS prototype project Notes
• 1st March 2002: Meeting to discuss progress of SPS prototype project Notes
• 5th September 2001: Presentation and acceptance of SPS prototype project to/by SLTC
• 6th July 2001: Meeting to discuss SPS prototyping with dedicated pickups Note
• 8th June 2001:  Meeting to discuss SPS prototyping with dedicated pickups Notes
• 27th April 2001: 7th Brainstorming meeting Minutes
• 30th March 2001:   6th Brainstorming meeting Minutes
• 23rd March 2001:   Meeting with MTA to discuss multipoles factory Minutes
• 16th March 2001:   4th Brainstorming meeting Minutes
• 9th March 2001:   3rd Brainstorming meeting Minutes
• 16th February 2001:  2nd Brainstorming meeting Minutes
• 9th February 2001:   Brainstorming meeting to follow up on reactions to LHC-CP proposal. Minutes.
• 8th November 2000: Need for real-time established and accepted by SL-TC. LHC-CP given the go ahead to start working group to refine requirements and investigate possible architecture. Slides from the meeting: html powerpoint
• April 2000 Summary of discussion with P.Ribeiro/T.Wijnands SL-CO/FE courtesy Michel Jonker for the LHC-CP workshop. here.