CERN Accelerating science

LHC-CP sub project on RT controls for the LHC

Meeting no.1 : Friday 09-02-2000 - reported by Thijs Wijnands & Mike Lamont

Present : M. Jonker, Q.King, R. Jones, A. Butterworth, T.Wijnands, P. Ribeiro, JJ Gras, P. Charrue, R. Lauckner, P. Anderssen, M. Lamont.

Minutes :

ML gave a summary on the RT requirements as they were presented to the SLTC on 25/10/2000 (ppt). Slides used in the meeting here (ppt).

• Timescales for correction : can we allow a display that locks for a few seconds ? No.
• hard or soft real time applications ?
• RL : RT Controls should NOT interfere or be used to guarantee the safety of the machine. It should be used to make machine operation as efficient as possible.
• We need a clearer definition of what we mean by real-time. This is never easy, see manifesto/discussion here for example. But it is clear that if we are to do feedback at 10 Hz, or provide fast RT displays, or RT knobs then the control system must be deterministic enough to allow guarenteed acquistion, alogrithm, correction at a maximum rate of 10 Hz
• It was concluded and agree by all present that real-time was necessary and that a general, global solution was to be found. (As opposed to dedicated solutions for each feedback loop etc.)

Concerning ML's slide titled "Implications" :

• JJG : BI will provide measurements of the orbit : 10 Hz for the global orbit, 100 Hz for the BLM's. In real-time to the gateways.
• It is not yet known at what rate tune and chromaticity will be available. In any case, the tune loop should be the responsibility of BI. Although supporting functionality (such as trimming quads) should be part of a global solution.
• QK : Power converters will accept RT trims at 100 Hz.
• AB : RF feedbacks are all local, so no rt connectivity data is expected. However, for radial steering it is required to have knowledge about the global orbit (avoid cross talk)
• PA : Rt communication across the site is required. If this is done by the IT group, ethernet is the technology to use. PC : Is rt communication the responsibility of CO or of IT ? Now there was some discussion about ethernet technology :
• PA : Present day switches are much faster, they do routing as well and never drop a packet.
• PR : Latency is important, if you build a rt system you must have reliable components from A to Z. Hence the reason for investigating ATM technology. Nevertheless, if ethernet has to be used, we can always work out something that will work. < In this context we MUST NOT exclude a separate network for the RT controls domain >. Note that : the TCP/IP stack at the end nodes does not support RT communications.
• PA : it can all be done with ethernet provided we implement it properly. There will not be a separate network for machine controls in any case.

MJ : Made a presentation of which stressed the importance of the logical architecture. We should look at the logical architecture of the system.

Conclusions

• Real-time capabilites will be required to operate the LHC.
• The demands on speed appear to be reasonable, and data acquistion, or measure & correct appear to be necessary at a maximum of 10 Hz. Bandwidth requirements are also moderate. At the maximum rate of 10 Hz, we can compensate for changes of the beam at a closed loop bandwidth of 1 Hz. This is sufficient to control perturbations due to ground motion and persisitent currents (decay and snap back) .Contributions to the time delay are expected from the equipement (BI and PO), the network (distance!) and the centralised processing.
• Both Beam Instrumentation and Power Converter systems are already planning on providing real-time acquistion and correction from the front-ends to at least the gateway level.
• Ethernet by over resourcing appears to be capable of providing the necessary level of performance.
• A real-time environment will be necessary at the high-level.
• An understanding of the global requirements in terms of a logical archietecture is required to allow the implementation of a well-designed, flexible, global, integrated system. To be followed up.