Post on 31-Jan-2016
description
transcript
CLIC meeting, CERN, 20.04.2012
Vertical emittance minimization at the SLS through systematic and
random-walk optimizationCLIC meeting, CERN
20.04.2012
M. Aiba, M. Böge, N. Milas, A. Streun, PSI
CLIC meeting, CERN, 20.04.2012
Introduction
• Vertical emittance minimization is motivated by:– increase of brightness and transverse coherence– operational margin for small gap insertion device
(possibly even smaller undulator gap)– TIARA WP6 SVET (SLS Vertical Emittance Tuning)
• Collaboration: CERN-INFN-PSI+Maxlab• Establish VET means at SLS, for CLIC DR and SuperB
→ Fine corrections of betatron coupling and y, and maintaining small emittance during operation
• Beam size monitor R&D → NM’s talk• Intra Bunch Scattering studies
CLIC meeting, CERN, 20.04.2012
Swiss Light Source
Swiss Light Source – 3rd generation light source– 18(+2) beam lines– 2.4 GeV, 400 mA (top-up)– C~288 m
SLS
CLIC meeting, CERN, 20.04.2012
SLS vertical emittance• What was expected and what is achieved
– Achieved much better than expected, thanks to • 30 more skew quads installed (6 skew quads initially)• Better alignment on girder than expected• Girder realignment in 2011• Elaborated model based corrections• Random-walk optimization
Emittance ratio ≡ v/h
h~ 5 nm
(Insertion devices off)
Application of these methods achieved 0.9 pm!
1.8 pm in March 2011
CLIC meeting, CERN, 20.04.2012
Key component 1
– BBGA initially planned(see Backup slide)
– Switch to BAGA (later slides), more robust approach
• Magnet girder
CLIC meeting, CERN, 20.04.2012
Key component 2
CLIC meeting, CERN, 20.04.2012
Girder realignmentMotivation and approach
• BAGA (Beam Assisted Girder Alignment)– Remotely align girders based on survey data– Confirm the result online with beam and fast orbit feedback running
Girder discontinuity estimation from “corrector pattern”
CLIC meeting, CERN, 20.04.2012
Girder realignment result
BAGA resulted in:
-Gaussian like corrector kick distribution
-About half corrector kick
-About half dispersive skew correction
-Similar non-dispersive skew correction (Qaud rotation dominant?)
Result for Sector 1
Result for all sectors
CLIC meeting, CERN, 20.04.2012
Model based correction 1
CLIC meeting, CERN, 20.04.2012
Model based correction 2
ORM
* Contribution of BPM tilts subtracted
*
CLIC meeting, CERN, 20.04.2012
Random walk optimization
• Limitations in model based corrections…– Beam measurement errors– Model deficiencies
• Apply multi-variable optimization– Random-walk would be the best algorithm
• Model independent correction• The curse of dimensionality is avoided (#Knobs=12/24/36)• The optimum solution is within easy “walking-distance” after systematic
correction• Minimal effort to implement• Potential of online optimization, i.e. keeping small emittance during the
operation• NB: the optimization needs a target function, which is the measured vertical
beam size in our case
RAndom Coupling Correction (RACC) ~ Luck (in Japanese pronunciation…)
CLIC meeting, CERN, 20.04.2012
How it worksTypical successful step(Figure from first test)
Configure the parameters of RWO
Measure target function, f0
Generate random corrections (RC)
Add RC to the present correction variables
Measure target function, f
Update f0=f Remove RC
f<f0 f≥f0
Reconfigure RWO if necessary
Flowchart
CLIC meeting, CERN, 20.04.2012
MD on 6th Dec. 2011
Corrected vertical dispersion
• First dedicated MD after BAGA– v~1.3 mm rms with model based correction!– v~1.2 pm at the end of model based correction– v~0.9±0.4 pm with RWO in addition!
(Only ND skew quads optimized)– Better coupling correction with RWO was confirmed with ORM before and
after
1000 2000 3000 4000-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
I (
A)
Time (sec)
ND skew Q currents during RWO (plotted 4 out of 24)
Beam size measurement during MD
CLIC meeting, CERN, 20.04.2012
Model deficiencies?• What limits the model based correction?
– Step size of RWO was determined from the continuation of model based correction after it converged → Fluctuation of correction due to measurement error
KL~10-5 m-1 rms → Step size chosen in a range of KL=0.5~2×10-5 m-1 – However, RWO arrived at KL=7×10-5 m-1 rms – BPM and corrector tilts may be the sources
Simulation showed residual tilts of a few mrad already explain this
0 1 2 3 40
5
10
15
Re
sid
ual s
kew
qu
ad. K
L, r
ms
(10
-5 m
-1)
Residual corrector tilt, rms (mrad)
0 1 2 3 40
5
10
15
Res
idua
l ske
w q
uad.
KL,
rm
s (1
0-5 m
-1)
Residual BPM tilt, rms (mrad)
CLIC meeting, CERN, 20.04.2012
Summary
• Ultra low vertical emittance of 0.9 pm is achieved at the SLS !– BAGA + Model based corrections + RWO
• RWO– Successfully demonstrated, a good performance
booster– Potential applications:
• Beam lifetime of the SLS, SwissFEL undulator BBA etc.• Online optimization
Maintain small emittance during operation in CLIC DR and SuperB
CLIC meeting, CERN, 20.04.2012
Backup slides
CLIC meeting, CERN, 20.04.2012
Beam Based Girder Alignment
CLIC meeting, CERN, 20.04.2012
BAGA seen by the HLS in Sector 2