CERN Accelerating science

Real Time RMS prototype

[for discussion!]

As has been established, there will be a number of real-time processes in action, including real-time feed-forward of corrections based on measurements coming from the reference magnets. Inter-system dependencies, settings management, common requirements for access to various parameters strongly suggest a common approach. Indeed central control for a given class of correctors to take care of: summing trims from different inputs, field to current calculation and hysteresis loop crossing would appear to mandatory. 

The logical architecture shown below is not aimed the feedback loops themselves, but at providing common facilities for changing beam and/or magnet parameters.

• Assumptions
• Possible Nominal Sequence
• Suggestion for Prototype 
• Logical Architecture Data Flow
• Trim Controller Data Flow

Assumptions & Questions

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 is therefore done centrally.
• Power converters can execute pre-loaded functions. Could drive correctors during injection? Will drive them during the ramp.

Magnets

• 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).
• Open question: Do real-time measurements from reference magnets come in as absolute values? What is input? How?
• A "current forecast" model: this would look at the multipole history of the reference magnets, and the powering history (in terms of I and T) of the main dipole chains, and predict the multipole behaviour of the main magnets. This "current forecast" could be used: 
  1. at pre-injection to anticipate the magnitude of the persistent currents at the start of the injection plateau and subsequent decay on the injection plateau.
  2. just before the start of ramp to anticipate the depth of the snapback, given the adjustments made on the injection plateau, the measured chromaticity, and length of time on plateau.

Beam Measurements

• Chromaticity feedback not available routinely.

Possible nominal sequence for handling b3

1. Pre-injection: The off-line MF model produces prediction of persistent current and their decay. Predicted b3 behaviour written to database. If different enough from last fill, prediction converted into time dependant corrector functions and downloaded to power converters. Possibility to incorporate empirical trims to predicted corrections established from experience.
2. Ramp to injection level
3. Timing event to power converters which start driving correctors with pre-loaded functions.
4. Enable reference magnets, which feed measurements to MF-RT feed- forward process.  The latter compares with respect to reference and produces correction if necessary.
5. Inject pilot
6. Measure chromaticity, correct with main sextupoles.
7. Before ramp, assuming random length at injection: call MF model to make prediction about snapback depth/b3 behaviour (and other multipoles), produce corrector functions. Stop power converters executing injection plateau functions. Load power converters with ramp functions, start ramp. MF-RT feed-forward clearly has to be kept carefully informed about reference values during this phase (if it needs to be on).

Prototype

Within agreed framework:

• First model to produce predicted persistent current and decay. Need to decide what parameters will be exchanged.
• Convert to correction functions.
• Download to power converter.
• Dummy up reference magnet to provide RT multipole measurement.
• Compare with reference. Compute correction. Compute current change. Feed to power converter.
• Try running the two together!
• Offline model to produce predicted snapback behaviour. Start ramp. Test RT-MF b3 to current plus hysteresis loop crossing.

Data Flow

Trim Controller (e.g. b3 and feedback)