Eurizon Task 4.1 general update and Topic 3: pyAT errors

Europe/Paris
zoom (remote)

zoom

remote

Eurizon Monthly meeting

Done:
M4.1.1 (M30 July 2022, DESY) Identify common beam physics interests, define necessary software developments
M4.1.2 (M30 July 2022, DESY) Definition of the work organisation for beam dynamics studies
M4.1.3 (M32 September 2022, ESRF) Define work organisation and contributors to the tasks identified in M4.1.1
M4.1.4 (M36 Jan 2023, DESY) Beam Diagnostics: Definition of the work organization for beam diagnostics studies


To do:

D4.1.1 (M38 March 2023, ESRF) Preliminary technical report with proof of principle validation on simplified test cases

M4.1.5 (M42 Jul 2023, ESRF) Beam Diagnostics: Selection of the shaker device

D4.1.2 (M48, January 2024, DESY) Final technical report with detailed beam dynamics studies and documentation for EBS and/or PETRA IV

D4.1.3 (M48, Januray 2024 DESY) Technical report on beam diagnostics studies with detailed documentation.

 

 

Eurizon Task 4.1


participants:

SL Simone Liuzzo  YES
LH Lina Hoummi YES
SW Simon White YES
LC Lee Carver YES
NC Nicola Carmignani YES
IA Ilya Agapov NO
JK Joachim Keil YES
TH Thorsten Hellert YES
BV Bianca Veglia YES
LM Lukas Malina YES
EM Elaf Musa NO
MB Micahel BOESE NO


For diag meetings:
HS Holger Schlarb  NO
GK Gero Kube   NO
SJ Szymon Jablonski  NO
SP Sven Pfeiffer   NO
SM Sajjad Mirza  NO
KS Kees Scheidt NO
BR Benoit Roche  NO
EB Elena Buratin NO
FE Friederike Ewald NO


discussion:

 

Thorsten presentation on SC

SC for ALS, used for petraIV
Try to translate SC to python for PETRAIV future users

outline of capabilities and design of software
may be need a dedicated meeting just on technical details

collective effects in AT/pyAT, why converting to elegant?
SW: pyAT has collective effect. also matlab AT has. needs only impedence model. advantge of python is better user interface. all pass methods are there.

SL May SC replace MML?
TH: testing SC direct interaction with control. kind of works. Will work on this for next contract at ALS-U. toolkit designed as a simulation software, not for easy communication to the machine. there are fundamental differences with MML.


SC has a manual, difficult to translate it as well?
TH: no much effort to produce a new manual.


SL atpass also for trajectory?
LF: ringpass and linepass both call atpass. actually very good to keep.
LM: may be more functions used: 8 methods from AT all together!
atpass
tgetfieldvalues
findorbit
fidsopos
atlinopt
atmatch chrom delta
twiss line
atloco

 

 

SL Frequency error (RFCavityPass)?
TH: RFCavityPass already there

SL input from magnetic data for multipole errors?
TH: shows example, see pictures.

SL very large orbit errors (mm at start)?
TH: no option for this. Could be an interesting addition. Adding custom start for orbit. May be not too hard to implement.

SL aperture expected in the lattice or from SC?
TH: can do both, but in the lattice in the end.

SL which version of AT is christoph staier using?
TH: 2.0! (last release is 2.4) SC is using latest version. master from git.

SC checked for new releses of AT?
TH: no check, trust. few functions.
may need to add unit tests for pySC developments

SL cluster interfaced for DA/TLT, etc...?
TH: not natively. but script exists. already slurm.

SL which corrections can be done within SC?
TH: many corrections, all, up to ID tuning.
DA, MA, IE,


Lukas Malina, presentation on conversion

troubles with implicit code, quite some solution to be found manually

python pointer like approach is to be considered.

full package is transalted. 1/3 not really working. some parts already producing results

L.C. do you think you saved time with this method?
L.M. about a factor 3 in efficiency estimated. may be realistically a factor 2.
T.H. Also jus copying by hand time is spared.
L.M. 5000 lines matlab code. 1month alone, 2 weeks with ChatGPT.
L.C. is  the python code nice/elegant after conversion? is it pythonic? or is the code running but could look better.
L.M. surprisingly nice. parsing function by function. this is a limitation. stitching by hand.
for looping not removed by chatGPT. not doing vectorization. sometimes, it forgets (in the long functions) and just copy matlab bits. Also hidden pitfalls. matlab-numpy corresponding methods give different results (SVD for example).  

L.M. easier to read the translated python code than the matlab one.

N.C.: will the comments go back?
L.M.: at some point.


if you are polite with chatGDP you get better translation?
NO

how can we contribute from ESRF side to the conversion?

documenting the code

could plan a meeting for technical meeting dedicatd : 9th March

L Farvacque big difference in python and matlab. in python pyAT sphynx, you have to write doc with RESTRUCTURED text.
L.M. like sphynx since also pyAT uses it.


MDTs are confirmed at PETRA III 22nd march and 26th April 2023

N.C. will run some simulations of TbT to see precision of method on simulation. will organize a separated meeting.

discussion on deliverable status

presentation by L.Farvacque on Observables and Variables in pyAT.

meeting closed

There are minutes attached to this event. Show them.
    • 1
      Eurizon Task 4.1 status
      Speaker: Simone Liuzzo (ESRF)
    • 2
      Simulated Commissioning Introduction
      Speaker: Dr Thorsten Hellert (DESY)
    • 3
      SC conversion to python
      Speaker: Lukas Malina (DESY)
    • 4
      pyAT recent error developments
      Speakers: Dr Laurent Farvacque (ESRF), Simon White (ESRF)
    • 5
      Eurizon Task 4.1 Deliverable D4.1.1 status

      must be completed between 1st March and 31March. We assume 15th March to leave some (little) time to the other management levels to review and do what has to be done for submission

      Speaker: Simone Liuzzo (ESRF)
    • 6
      Status of future MDTs

      planning of MDTs in next EBS / PETRAIII runs

      22nd March Badge @ PETRA III
      26th April TbT @ PETRAII

    • 7
      Any other bussiness

      status of Topic 1
      next meeting