Notes from 27 July 2006 SWMD Telephone Meeting

Notes from 4-July-2006 meeting

---

1. July ESA run
   
   • Processing status
     Still problems with aborts after 2820 events in root2mat converter, email discussions with Jonny/Mark S/Yuri
   • Comparison of slot 3
     See below.
   • Mechanical measurements of collimators? Mover reliability?
     Wait until we meet with Steve M. in few weeks to discuss this.

2. T-480 Data analysis
   
   • Luis plots
     Study of "resolution" and "misalignment" for BPMs, definitions of these given in code. Very large resolutions/misalignments seen for slot 3 sandwich 1 in the second downstream triplet only. Studied effect of not using BPM9, this did reduce resolution (rms of measurements) and misalignment, but made little change in the reconstructed kick.

     Disussion about whether rms is most appropriate way of estimating resolution, Andre will look into impact of using gaussian fit to peak as alternative to reduce sensitivity to single pathalogical data points. (Luis saw such points when lowering charge cuts to >100, which admitted a single additional point.) Resolution should be either given by BPM experts or taken from reference data for preference.

     Argued that this is because the reconstructed kick is a real effect due to scattering from the trailing edge of the long collimator, and the fact that it is only seen at one extreme in collimator y is because the collimator pitch is not perfectly aligned to the incoming beam, therefore at one extreme the beam will be close to (impinge on) the upstream collimator edge, while at the other the beam will scrape the downstream edge and hence lead to more background in downstream BPMs.
     Discussion to be continued. Need to look at the July run data to test whether the same effects are also seen there.
     To help understand this, need to be sure we can all see the same effects!

   • Discussion of specific details! How to converge on results by mid-Sept. (latest!)
     Long discussions during the day, concentrating on a few specific runs that Luis had been looking at in particular for slot 3 sandwich 1 of the April run: 595 (ref 591), 619 (ref 618), 638 (ref 637), 657 (ref 659).
     Adriana has been hindered significantly by PC being damaged by mains voltage phase problem in her CI office, leaving her without Matlab access.
     Concentrated on making sure we had 100% agreement on the results of basic elements such as quality cuts, collimator y, fitting, so that we could then have meaningful comparisons of the data themselves.
     Specific checks of basics to ensure everyone agreeds on quality cuts: bunch charge (>200 units), camac_chan8 (and, for some early data, camac_chan0), and bpm_y. Where a pulse fails e.g. a charge cut on any BPM tested, reject entire pulse rather than reduce number of BPMs used, to simplify analysis and ensure (more or less) stable angular resolution. One implementation of these cuts is GetGoodPulseIndex.m (others have different ways of doing the same thing, this is just as example).
     For conversion of camac data to collimator y, exchanged vectors of y positions in microns for runs 595/599, and exact agreement found using either "binning" or 2nd order polynomial parametrisation of camac_chan0 vs. y (see e.g. CamacToCollimatorY.m.
     For fitting, Luis had used Steve's noplotpolyfit, Jonny had used this as well as lscov. Need to check results again in detail for a single run to ensure we have like-with-like agreement.
     Apparently the BPM resolutions input to the fits have very little effect on the reconstructed kicks - surprising - can someone check fit is behaving correctly?

3. Wakefield Simulations
   • New simulation of collimator 8 - Mikko Karkkainen
     I simulated collimator number 8 using rms bunch length of 300 micrometers. I truncated the structure about 20 mm after the collimator and used indirect integration to calculate the wake potential assuming infinitely long outgoing pipe (comparing with the bunch length, I assume that 1.5 meters is "almost infinite"). My present implementation is with C, so it is much faster than the old matlab-code. See the attached first results with very coarse meshes. I will send you more results with finer meshes if these results make sense. Please comment how they compare with other results (measurements or simulations) that you have.
     Difficult to compare detailed results with Mafia simulations (for which 300 micron bunch length is difficult because of memory limitations), and so far for GdfidL (cpu time/cluster limits). Jonny will reply in detail to Mikko.
   • GdifdL simulation status/plans - Jonny
     Ongoing problems with number of nodes available on Birmingham escience cluster, current jobs needs 20 cpus at a time so delays are significant. Collimator 6 is currently in the queue... From discussion with Warner Bruns, Jonny will try to make greater use of symmetry/boundary conditions to reduce resource requirements, and will also try out a GdfidL development version with a moving (but not conformal) mesh. Also, will investigate reducing computing requirements by limiting transverse dimensions of the problem to a region of the minimum aperture +/- 10 sigma.
     Question about whether long range wakes are a concern, possibly, but difficult to see how this could be calculated explicitly, suggestion to use an eigenvalue solver with Q, R/Q for the structure.

4. Merlin plans
   Implementation of wakefields up to m=4 modes for circular jaws OK, tests for rectangular jaws ongoing, progress slower while recruiting replacement for GK (Adam M. has also left). The function must depend on half gap, taper angle, length of collimator and beam pipe radius (although not all of these are independent variables). Adriana would make a first version of this available to Frank in the next week.

5. Plans
   Assume we request time for next ESA run Jan/Feb 2007. Previous experience shows we should allow 3 months from final "physicist" plan of collimators to having hardware delivered to SLAC, allowing for workshop scheduling, material procurement, iteration on design during drawing.

   Therefore, final "physics" design needed by beginning of October.

   This needs the existing data to be fully analysed, compared with all simulations and calculations available to decide what the purpose of the next tests should be. It also needs any improvements in the reliability of wakefield modelling to be implemented into tracking codes such as Merlin and for the predicted impact such wakefields would have on emmittance to be studied.

   Proposed timescales for us are: first release of wakefield module in Merlin: end of July;

   • Analysis/simulation of T-480
     • Completion of ESA data analysis - output: kick factors per collimator - first week Sept.?
     • Completion of wakefield simulations for ESA collimators, in form of kick factors per collimator - first week Sept.?
     • Comparison of simulations/data from 2006 ESA run - output: aspects of the design that need more data to improve models and refinements to existing calculations 3rd week Sept.?
   • Tracking simulations for ILC collimations
     • Implementing wakefield calculations into Merlin (or Placet) - end of July?
     • Quantify effect on ILC performance?
     • Study effect of more realistic (data-validated) calculations on ILC performance?
     • Iterating on design (more e.m. simulation/calculation - testing effect of revised designs in Merlin)?
   • Manufacturing new insertions
     • Proposal to SLAC for new collimator designs (1 week)
     • Fabrication/shipping to SLAC (3 months)
     • Beam test SLAC, Jan/Feb 2007

6. Damage tests - not discussed, to be picked up again!
   Justin reported that George had initiated discussions with Culham on graphite-metal brazing, with one technique they suggested being "liquid phase bonding" (glue?!), following up how to fabricate a first graphite-titanium collimator, as studied for the damage investigations.

7. Wire tests - not discussed.

8. Next meeting - suggested in week starting Aug 7 as Steve Molloy will be in UK at that time, date t.b.c.