CERN Accelerating science

Real-time control in the context of the Multipole factory

Meeting held on Friday 1-03-2003 -  reported by Mike Lamont

Present : Thijs Wijnands, Mike Lamont, Luca Bottura, Laurent Deniau, Quentin King, Ralph    , Jens Andersen, 

Luca presented the status of possible on-line measurements... 

1. Rotating coils. At present can provide measurements at rate of one per 20 second. The use of slip rings which allow the coils to carry on spinning in the same direction should allow 1 Hz (as in HERA). 

2. Fast measurements from Hall plates  give b3 and b5 at around 10 Hz with less precision than the rotating coils. Should be capable of +/- 0.1 units of b3 after calibration.

3. NMR probe giving b1 at some what less than 1 Hz. New German NMRs are capable of 10^-5 accuracy after calibration. Expected to be available after 1 to 2 years.

Luca brief mentioned the b3, b5 measurements at the Tevatron, and associated improvements of control algorithms controlling sextupoles during snapback.

Following discussion partially based on collation of points that had been raised before. Main points below.

All RT processes need to use the same stuff to:

• Access equipment
•  Make trims
• Settings management
• B to I conversion, algorithms etc.
• Access configuration information
• Access dynamic information

Power converters

• Single point of RT entry into power converters
• System will allow for RT input for a given corrector from more than one point of control, possibly but not necessarily, simultaneously.
• Summing of RT inputs must therefore be done centrally.
• Power converters can execute pre-loaded functions. Could drive correctors during injection? Will drive them during the ramp.

i.e. control of corrector settings & translation between high level parameters and current for a given class of correctors needs to be performed at a single place. That is: strength to gradient conversion, the transfer function look up, taking care of hysteresis loop crossing etc. Changes in the correctors can be tracked and appropriate adjustments applied as required. This process will require access to the transfer functions and beam parameters (such as energy).

Discussion of the potential use of preloaded functions on flat bottom.

QK explained that the low level RST control would smother the granularity of real-time corrections (??).

Somewhat worrying discussion about the effects of a front-end controller crashing (e.g from a SEU) in the process of executing a pre-loaded function (PLF). Correctors in the tunnel will crash. They have the ability to recover in 2 seconds. The PLF is stored in RAM, problems  if this is corrupted. Recovery faster if not using a PLF, which would eliminate the reload time. Quentin suggest using real-time channel instead of PLFs. Mainly of concern for the orbit corrector (tunnel). Sextupoles and decapoles expected to receive something like 200 time less radiation.

Multipole factory

• Energy function drives construction of current functions
• Single point of entry into real-time correction channel, with possible reservation and permit flag (again). Single point of input from MF to control system.
• Hysteresis Loop Crossing long discussion

Problem sketched by Luca. Sign change in the direction of the current  forces a loop crossing (jump in B/I versus I), magnitude dependent effect which will require sophisticated on-line modeling.

Could attempt to pre-stress correctors in attempt to minimise effect.  Push some sets up, some down perhaps.  Not viable for b3 correctors. Increase frequency lots if little trips round the loop. Lots of volts at snapback to cover dynamic effects.

• Some question about size of effect  (100 mT)?
• Transfer function/Hysteresis Loop Crossing model necessarily a component of the control system. Agree we don't go to the Multipole Factory for B to I
• Corrector  Hysteresis Loop Crossing should be more-or-less independent of powering history and shouldn't be dynamic.

Discussion about pursuing some prototyping some stuff to exercise some of the above concepts. Ideally via feedforward from magnetic measurements - ambitious at this stage but possibilities might exist in block 4/SM18. In the first instance:

• Generate predicted b3 function, feed through conversion to current via transfer function and hysteresis loop crossing model. Offline this would give a first test of principle. MTA responsible for the model. See how this deals with snapback.
• Dummy up b3 input from MF and attempt RT like calculation of corrections.
• (Alarms, interlocks, debugging..

Everyone pretty much very busy until Q3 2003 and then some. Agreed to try and get moving after the summer recess.

Other points

Who is responsible for defining the architecture and implementation of the crates/back ends/rt control environment/interfaces for the test reference magnets and eventually the real thing? MTA need some support here.