Hysteresis Compensation

Dedicated MD 2024-09-04

3 min read

Logbook entries

https://be-op-logbook.web.cern.ch/elogbook-server/GET/showEventInLogbook/4137207 https://be-op-logbook.web.cern.ch/elogbook-server/GET/showEventInLogbook/4137284 https://be-op-logbook.web.cern.ch/elogbook-server/GET/showEventInLogbook/4137293 https://be-op-logbook.web.cern.ch/elogbook-server/GET/showEventInLogbook/4137303 https://be-op-logbook.web.cern.ch/elogbook-server/GET/showEventInLogbook/4137345 https://be-op-logbook.web.cern.ch/elogbook-server/GET/showEventInLogbook/4137384 https://be-op-logbook.web.cern.ch/elogbook-server/GET/showEventInLogbook/4137393 https://be-op-logbook.web.cern.ch/elogbook-server/GET/showEventInLogbook/4137399 https://be-op-logbook.web.cern.ch/elogbook-server/GET/showEventInLogbook/4137428 https://be-op-logbook.web.cern.ch/elogbook-server/GET/showEventInLogbook/4137431

Trims are saved to ~/cernbox/hysteresis/dipole/notebooks/MD20240904/trims_cycle for SFTPRO1 and SFTPRO3.

Previous MD used MD20240508/tft_v1.ckpt

Dedicated hysteresis compensation MD has been shifted from 2024-08-01 to 2024-09-04.

  • Follow up with Foteini for the MD dates [priority:: medium] [due:: 2024-07-05] [completion:: 2024-07-29]
  • Move data buffering from local sps-app-hysteresis to UCAP [priority:: medium] [due:: 2024-08-23] [completion:: 2024-09-06]
  • Adapt sps-app-hysteresis to new PETE architecture and implement RDP in the application. [completion:: 2024-09-06]
  • Implement autoregressive predictions in sps-app-hysteresis [completion:: 2024-09-06]
  • Query Francesco 2024-08-28 max intensity extracted to the TED [priority:: high] [due:: 2024-08-28] [completion:: 2024-09-06]

MD goals

Proof of concept of hysteresis compensation on all cycles.

  • Eddy current correction on all cycles ( T accuracy)
  • Remove MD1 in supercycle (or with dipoles on idle current)
  • Hysteresis compensation on all operational cycle types
    • 2x SFTPRO1 in same supercycle with higher energy cycle
  • Autoregressive prediction + compensation
  • Evaluate effects of pre-cycle
  • Potentially: quadrupole predictions with hall sensor data (but requires revamp of prediction infrastructure)

Supercycle requests

  • Ask Francesco for TED max intensity [completion:: 2024-10-03]

Cycle requests

  • MD2: MD1, but with MBIs at idle current in LSA - we need this for the logging and tracking. This cannot be replaced with ZEROs because we need the higher order magnetic cycles intact.

  • SFT_PRO_MTE_L4780_2024_V1_Clone: 5e12 intensity, whatever standard beam, extract to TED

  • MD_26_L7200_Q26_North_Extraction_2024_V1: standard beam

  • AWAKE_Q20_2024_V1

  • LHC_PILOT_Q20_2024_V1: no beam, magnetic cycle only. standalone is OK.

Supercycles

PRECYCLE:

  • LHC_PILOT_Q20_2024_V1
  • MD2
  • LHC_PILOT_Q20_2024_V1
  • MD2
  • LHC_PILOT_Q20_2024_V1
  • MD2

NA_PHYS_LHC:

  • SFT_PRO_MTE_L4780_2024_V1_Clone
  • MD2
  • SFT_PRO_MTE_L4780_2024_V1_Clone
  • MD2
  • SFT_PRO_MTE_L4780_2024_V1_Clone
  • SPS_TIMING_2021_V1
  • LHC_PILOT_Q20_2024_V1, no succeeding ZEROs for padding
  • MD3

NA_PHYS:

  • SFT_PRO_MTE_L4780_2024_V1_Clone
  • MD2
  • SFT_PRO_MTE_L4780_2024_V1_Clone
  • MD2
  • SFT_PRO_MTE_L4780_2024_V1_Clone
  • MD3

HIRADMT_PHYS

  • SFT_PRO_MTE_L4780_2024_V1_Clone
  • HIRADMT_PILOT_L8400_Q20_2024_V1
  • MD2
  • SFT_PRO_MTE_L4780_2024_V1_Clone
  • HIRADMT_PILOT_L8400_Q20_2024_V1
  • MD2

AWAKE_PHYS

  • SFT_PRO_MTE_L4780_2024_V1_Clone
  • AWAKE_Q20_2024_V1
  • MD2
  • SFT_PRO_MTE_L4780_2024_V1_Clone
  • AWAKE_Q20_2024_V1
  • MD2

The MD will only trim on SPSBEAM/BHYS.

Schedule

Focus SFTPRO Focus MD3 flat bottom If MD3 works, test HIRADMT & AWAKE operational cycles Switch between 3 supercycles, with stable spill. Try autoregressive and non-autoregressive

Action

  • Parallel MD on Friday before to set up applications [completion:: 2024-08-28] No MD BOOKES
  • Ask Pablo to set up spill for SFTPRO2 with MTE core only [priority:: medium] [due:: 2024-08-23] [completion:: 2024-08-23] Pablo will make an appearance and try to optimize the spill quality during our MD. We will query Francesco the max intensity extracted to the TED.

Game plan 26 Aug - 30 Aug

Monday UCAP event building initial design Hall sensor placements and initial measurements

Tuesday UCAP event building polish and finish Dipole measurements Fit calibration function from measured data (v6)

Wednesday Adapt sps-app-hysteresis to new transformertf flow and PhyLSTM and TFT inference. Drift correction on dipole measurements

Thursday Implement autoregressive predictions in sps-app-hysteresis Model training / fine-tuning

Friday Bugfix sps-app-hysteresis and make sure it runs Evaluate and the trained models

Graph View

Backlinks