Page tree

Present: Bolko Beutner Maria-Elena Castro Carballo Lars Froehlich Raimund Kammering Unknown User (yuhui) @Torsten Limberg Sascha Meykopff @Sergey Tomin Josef Wilgen

FEL Optimization

We list potential tuning knobs for optimizers and gather ideas on automatic FEL tuning.

Tuning knobs/potential optimization loops:

  • Adaptive orbit feedback
  • Intra-undulator orbit (air coils)
  • Phase shifters
    • Should they be optimized one-by-one or in sets, via scan or via extremum-finding? The most time-efficient method should be preferred.
    • As far as we know, Matthias is preparing a tool for fast, continuous phase shifter scans.
    • There have been reports that hysteresis is an issue. Yuhui is going to gather some information.
  • Undulator taper
    • Could be optimized automatically, but has not been tried yet.
    • We should verify if the taper settings recommended by the LCLS tool agree with reality.
  • RF chirp & curvature, gun phase
    • Can be tuned via optimizer, but requires manual selection of bounds and actuators so far.
    • Automatic optimization based on BCM/XGM correlations is possible.
  • Injector trim coil
    • Is a strong tuning knob, believed to have an impact on the beam tilt (z-x or z-y correlation).
  • Trajectory in pre-BC2 sections
    • The orbit in the low-to-medium energy part of the machine can have significant impact on the FEL process.
    • Insufficient understanding for automatic tuning.
  • Laser heater (intensity, position)
  • Optics (quadrupoles upstream of undulators)

Required diagnostics & tools:

  • Gain curve measurement
    • A fast measurement of the gain curve could help us better understand where and what to tune. We have no clear picture how far Lyuba's efforts have come.
  • Online spectrometer
    • An online spectrometer would also help with tuning. The quality of the spectrum might be a direct input to optimization algorithms together with the intensity.
  • K-monochromator
    • K values should be measured and equalized once a lasing trajectory has been found. Other labs have reported that a fast scan with a photodiode works well – why do we use slow image processing?

Next steps:

  1. GUIs with launch parameters: We'll try to make existing optimizer settings more easily accessible. For example, the adaptive orbit feedback for SASE2 could be started directly by a button on a DOOCS panel. Another button could start the optimizer with a preset for optimizing the SASE1 quads.
  2. Autonomous scripts: Well-working setups can then be automated further so that they can be started and stopped without additional user interaction.
  3. Server integration: Well-working scripts could be replaced by servers or started by a master-optimization server in a fashion similar to Florian's PSYCO.

Action items