FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH 1. FLASH / FEL BeamDynamics Seminar

Transcription

1 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH 1 FLASH / FEL BeamDynamics Seminar Beam Based Alignment in the FLASH Undulators Using the LCLS Method (a Preliminary Status Report) Mathias Vogt (DESY MFL)... with a lot of help and inspiration from Jin Hyunchang & Duan Gu! The (modified) LCLS Method SASEBBAGUIMain The Measurements in FLASH A Long FODO Channel w/ Optimized Phase Advance FLASH Free Electron Laser in Hamburg

2 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH 2 The LCLS Method (H.Loos/SLAC) here: for one Phase Plane, say (x, p x ) At P different Energies {E k } k=1,p Given : M BPMs with offsets R M for each k : Actual orbit X k, measured Orbit Y k with random errors ξ k Y k = X k + + ξ k Given : N perturbations(= misaligned quads) and/or correctors(= movers) : misalignments d R N independent of energy!!! For each k : initial cond.(= launch) z (x 0, x 0) T k X k = L k z k + O k d LaunchResponseMatrix (LRM) L k OrbitResponseMatrix (ORM) O k

3 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH 3 The LCLS Method (H.Loos/SLAC) (2) Now join over all P energies : X := ( X T 1,, X T P ) T, Y :=..., z :=..., ξ :=... L := diag(l 1,, L P ) R P M P 2 O := (O T 1,..., OT P )T R P M N U := (1 M M 1,, 1 M M ) T R P M M P Y = L z + O d + U + ξ or: Y = A v + ξ with A := (L, O, U) and v := ( z T, d T, T ) T Possibly add more objectives (soft constraints) : either 0 = P M i=1 i & 0 = P M i=1 s i i i = (0) + (1) s i or 0 = P N i=1 d i & 0 = P N i=1 s id i d i = d (0) + d (1) s i or {0 = i ref i or {0 = d i d ref i } i=1,...,m } i=1,...,n P M+ x constraints for 2P + N + M variables... each constraint can be given it s own weight!

4 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH 4 SASEBBAGUIMain Interlude: LIVE DEMO

5 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH 5 jplot/plotdialog(s)

6 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH 6 Measurements night n n (first 3h) e log : Achievements e log : Achievements Took one set (3 data sets) of undulator saved (2x) 3 undulator BBA data sets BBA measurements in the files prepared Monday/Late (600,720,900 MeV) AFTER CORRECTION (Jin Hyunchang) (100,1000 shots) for (600,720,900) MeV / with quad movers as is. for all three energies one establish a Preliminary evaluation of the data suggests a correction which we applied le time!!!!! launch with decent losses in relatively litt- manually. This strongly suggests that the correction computed by Jin does really straighten Took another set of Undu BBA measurements w/ quad mover corrections the beam!!!!! applied. Offline analysis will follow.

7 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH n / 1-st set / all 3 orbits (Energies)

8 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH n / 1-st set / Plain Singular Values

9 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH n / 1-st set / SingVal s after restricting the Movers Not all quads have movers! First quad misalignments are equivalent(degenerate) to initial condition Last quads: no influence on SASE and too few BPMs downstream

10 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH n / 1-st set / SingVal s after removing BPM2UND3 ( sucks!) In principle: all BPM2UNDx & BPM4UNDx could be given lower weights also BPM9DUMP beam sweeper!

11 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH n / 1-st set / Computed Quad Alignment e log Hor.: quad index (as in list) ; Vert.: misal. (in m) too high scale??

12 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH n / 2-nd set / After applying corrections (scaled!) Y measured = A v machine + ξ error

13 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH n / 2-nd set / Comparison: Measured (L) vs. Predicted (R) Orbit v calculated = e A τ, w,... Ymeasured Ypredicted = A v calculated

14 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH n / 2-nd set / Calculated Quad Misalignments

15 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH n / 2-nd set / A hint of Horizontal Convergence? 0.1 "y13m01d09n_ms2_600_720_900_01.resx" u 1:3 "y13m01d09n_ms2_600_720_900_04.resx" u 1:

16 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH n / 2-nd set / A hint of Vertical Convergence? 0.1 "y13m01d09n_ms2_600_720_900_01.resy" u 1:3 "y13m01d09n_ms2_600_720_900_04.resy" u 1:

17 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH 17 Example: FAKED Quad Misal. s Calc. X Orbit Reproduced Quad-Misal. s? 600 MeV 1/4 β tron oscillation,, 720 MeV 1/6 900 MeV 1/ "FAKE_MS2_e3_002.orbx" every 1:3::0::0 u 2:3 "FAKE_MS2_e3_002.orbx" every 1:3::1::1 u 2:3 "FAKE_MS2_e3_002.orbx" every 1:3::2::2 u 2:

18 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH 18 Example: FAKED Quad Misal. s Calc. Orbit Reproduced Quad-Misal. s? input, reproduced with τ = small artefacts, reproduced with τ = % "FAKE_MS2_e3_001.resx" u 1:3 "FAKE_MS2_e3_003.resx" u 1:3 "FAKE_MS2_e3_004.resx" u 1: e e

19 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH 19 MODEL: Long (20xPer) FODO Section w/ non degenerate Phase Advance mu_fodo = k*(1/12+1/41) 0.5 k=1 (039deg) k=2 (078deg) k=3 (116deg) 0.4 [mu] < 1/2? [mu] :1-[mu] period

20 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH 20 FODO: Faked Orbit (w/o errors!) from Assumed Misalignments N.B.: there s no noise in this simulation.

21 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH 21 FODO: SingVal s after Movers have been Adjusted τ (SVD cut off) should not be above

22 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH 22 FODO: Reconstructed Quad Misalignments Input reproduced on the level of 10 8 ; orbit residue (least square s fit) Remember: there s no noise in this simulation.

23 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH 23 FODO: Reconstructed BPM Offsets (should be 0.D+0) into the noise

24 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH 24 FODO: Horizontal Input vs. Reconstructed Misalignments/Offsets/Launches 1 "FAKE_FoPe20Bpm4_E3_10.resx" u 1:3 "FAKE_FoPe20Bpm4_E3_12.resx" u 1:

25 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH 25 FODO: Vertical Input vs. Reconstructed Misalignments/Offsets/Launches 1 "FAKE_FoPe20Bpm4_E3_10.resy" u 1:3 "FAKE_FoPe20Bpm4_E3_12.resy" u 1:

26 FLASH- / FEL BD- Sem : / M.Vogt DESY MFL : LCLS style undulator BBA at FLASH 26 Summary First indications are that the modified LCLS approach can be applied (in some way) to FLASH But so far no fully convincing prove for a real improvement was found in FLASH: n : algorithm seemed to converge, but no improved transmission (quick-n-dirty!) n : slightly wrong optics model used, still better transmission! Hmmm... This is work in progress ; so much in progress that I have to apologize for the ugly slides! Maybe also try sflash & FLASH2?!?! Long structures are beneficial EU-XFEL