CERN Accelerating science

LHC-CP sub project on RT controls for the LHC

Meeting no.5: Friday 23-03-2001 - reported by Thijs Wijnands & Mike Lamont

Present: M. Jonker, A. Butterworth, M. Lamont, O. Berrig, O.Bruening, J. Wenninger, T. Wijnands L. Walckiers, L. Bottura, L. Deniau.

Agenda: LHC-MTA mulitpole factory

This was mainly an information meeting with LHC-MTA to discuss possible implementation of the so-called multipoles factory.

Minutes:

Overview of LHC_CP real time control requirements (M. Lamont)

ML gave a summary of the LHC-CP real-time discussions as covered in the previous real-time meetings, slides here (ppt). Issues like the foreseen requirements, correction rates, latencies, sampling speed and networking triggered some discussion:

LB remarked that there will be another 100 ms of latency because of the L/R time of the magnet cables for the main magnets. However, it was noted that trimming of the main magnet chains is not foreseen.

Then there was again the discussion on the logical architecture. So far, 2 different representations of the logical architectures have been presented. MJ would like to compare and discuss the architectures as they have been presented. There was some more discussion following the slides on reference magnets and MIMO feedback loops. An updated data flow diagram is available here (pdf).

Overview of LHC-MTA multipole factory (L. Walckiers)

LW presented the operational aspects of the multipoles factory as envisaged today (slides here, chamonix write-up here).

There was some discussion on the hysteresis : in particular the problems facing in crossing the hysteresis curve of the corrector magnets. A model will be needed to account for the effects of hysteresis loop crossing.

• The effect is large and a jump of DQ' ~ 10 must be anticipated when the MCS hysteresis curve is crossed during the snapback.
• MJ replied that is easy to account for the dead zone effect with the design of one's controller.
• JW noticed that is not as easy as it might look like. Some correctors will ramp up, some will ramp down. Can we use some of the correctors only for ramp up and the remainder only for ramp down?

The implementation of the multipole factory includes :

1. Cold measurements database
2. History recorder
3. Reference magnets

These components would be used to provide:

• A "current forecast" model: this would look at the multipole history of the reference magnets and main dipoles chains and the powering history (in terms of I and T) of the main dipole chains to predict the multipole behaviour of the main magnets and the required corrections.
• This "current forecast could be used:

1. off-line at pre-injection to anticipate the magnitude of the persistent current effect
2. off-line just before the start of ramp to anticipate the depth of the snapback
3. on-line together with the reference magnets to provide real-time correction
4. on-line to make the connection between beam parameter change and current change in a magnet or magnets.

Then there was a lively discussion on the "current forecasting", much of which focused on getting clear what the proposed functionality was.

• Operationally there is a clear reluctance to invoked anything too complicated (reliability, speed, debugging) in the real-time channel. For example, the conversion of parameter changes to current should be as simple as possible.
• Since the power converters only have a single trim dac the summing of real time corrections have to be done some where outside of the multipole schema.
• LB made some remarks : We cannot guarantee RT measurements - we might need a real time interface which is using non real time data. We should also prepared for fault tolerance (e.g blocking coils). Then the issue of latency came up again. The multipole factory should provide real time data and not be compromised if it needs to access a database.
• OBr asked about the Quality Assurance plan of this installation. How can we be sure that this will work ? Do we have confidence in this ?

ML proposed to have at least another meeting on this subject in the following weeks.

Next Week (23rd March) .... Next RT Meeting 10:00 Pavilion, on the agenda Thijs & Olav.