Status of FONT5 upstream
feedback
Glenn Christian
Robert Apsimon, Philip Burrows, Doug Bett, Neven
Blaskovic, Michael Davis, Young-Im Kim, Colin Perry
John Adams Institute, Oxford University
Javier Resta Lopez
IFIC,Valencia
26 June 2012
ATF2 Project Meeting
FONT5 upstream feedback system @ ATF2
•Two phase FB (position and angle) system to stabilise beam to the 1 micron level at entrance
to FF
•Bunch-by-bunch system (measure first bunch, correct subsequent bunches in train)
•3 stripline BPMs (on movers), 2 stripline kickers
P1
P2
P3
To dump
K1
QD10X
QF11X
K2
FB board
DAQ
QD12X
QF13X
QD14X
QF15X
FONT5 upstream feedback system @ ATF2
•Two phase FB (position and angle) system to stabilise beam to the 1 micron level at entrance
to FF
•Bunch-by-bunch system (measure first bunch, correct subsequent bunches in train)
•3 stripline BPMs (on movers), 2 stripline kickers
P1
P2
P3
To dump
K1
QD10X
QF11X
K2
QD12X
QF13X
QD14X
QF15X
FB board
Witness BPMs
DAQ
FONT5 Hardware
Analogue Front-end
BPM processor
FPGA-based digital
processor
Kicker drive amplifier
Strip-line kicker
Strip-line BPM with
mover system
FONT5 Hardware
Analogue Front-end
FPGA-based digital
processor
BPM processor
System Resolution (BPM processor)
<1 m
System Latency
<150 ns
Amplifier/ Kicker Bandwidth
~20 MHz
Dynamic Range of feedback system
+/- ~100
m (>46
dB)
Dynamic range of the BPM system
+/- ~500
m (>60
dB)
System parameters
Strip-line BPM with
mover system
Kicker drive amplifier
Strip-line kicker
16 April 2010
Feedback Performance (2) – Jitter
Reduction @ P2 (16 April 2010)
Measured bunch-to-bunch
correlations:
Bunch 1 – Bunch 2 : 98 %
Bunch 2 – Bunch 3 : 89 %
Bunch 1 – Bunch 3 : 85 %
Bunch 2 result implies resolution
of ~ 300 nm!
16 April 2010
Feedback Performance (2) – Jitter
Reduction @ P2 (16 April 2010)
Measured bunch-to-bunch
correlations:
Bunch 1 – Bunch 2 : 98 %
Bunch 2 – Bunch 3 : 89 %
Bunch 1 – Bunch 3 : 85 %
Bunch 2 result implies resolution
of ~ 300 nm!
16 April 2010
Feedback Performance (2) – Jitter
Reduction @ P2 (16 April 2010)
16 April 2010
Feedback Performance (2) – Jitter
Reduction @ P2 (16 April 2010)
Measured bunch-to-bunch correlations:
Bunch 1 – Bunch 2 : 98 %
Bunch 2 – Bunch 3 : 89 %
(Bunch 1 – Bunch 3 : 85 %)

'


2
cov
n
,
n

1
)
2
n
2
n
2
n

1
Bunch 2 result implies resolution of ~
300 nm!
16 April 2010
Feedback Performance (3) – Jitter Reduction @
P3 (16 April 2010)
Measured bunch-to-bunch
correlations:
Bunch1- Bunch2 = 84%
Bunch2 - Bunch3 = 87%
(Bunch1- Bunch3 = 94%)
Summary of FONT data-taking visits 2012
•
March 2012 – 1 week [Bett, Davis, Blaskovic]
–
–
•
April 2012 – 1 week [Bett, Davis, Blaskovic]
–
–
•
2 shifts
First tests of new FB firmware (with online phase compensation) – see later!
May 2012 – 2 weeks [Davis, Blaskovic]
–
–
•
1 week, 2 shifts.
Further investigations of phase jitter effects.
2 shifts per week
Troubleshooting problems with new firmware & first look at IP cavity signals
June 2012 – 2 weeks [Kim, Bett, Blaskovic, Perry (wk2), Christian (wk2)]
–
–
2 shifts per week
First week mainly monitoring IP cavity BPM signals, second week tests of new IPFB kicker
installed at IP.
•
More details can now be found on ATF Twiki/elog system.
•
Attention slowly shifting from upstream dual-phase FB system to IP feedback
–
See talk on IP feedback (P. Burrows) Thursday 14:40 JST!
Summary of FONT data-taking visits 2012
•
March 2012 – 1 week [Bett, Davis, Blaskovic]
–
–
•
April 2012 – 1 week [Bett, Davis, Blaskovic]
–
–
•
2 shifts
First tests of new FB firmware (with online phase compensation) – see later!
May 2012 – 2 weeks [Davis, Blaskovic]
–
–
•
1 week, 2 shifts.
Further investigations of phase jitter effects.
2 shifts per week
Troubleshooting problems with new firmware & first look at IP cavity signals
June 2012 – 2 weeks [Kim, Bett, Blaskovic, Perry (wk2), Christian (wk2)]
–
–
2 shifts per week
First week mainly monitoring IP cavity BPM signals, second week tests of new IPFB kicker
installed at IP.
•
More details can now be found on ATF Twiki/elog system.
•
Attention slowly shifting from upstream dual-phase FB system to IP feedback
–
See talk on IP feedback (P. Burrows) Thursday 14:40 JST!
Summary of FONT data-taking visits 2012
•
March 2012 – 1 week [Bett, Davis, Blaskovic]
–
–
•
April 2012 – 1 week [Bett, Davis, Blaskovic]
–
–
•
2 shifts
First tests of new FB firmware (with online phase compensation) – see later!
May 2012 – 2 weeks [Davis, Blaskovic]
–
–
•
1 week, 2 shifts.
Further investigations of phase jitter effects.
2 shifts per week
Troubleshooting problems with new firmware & first look at IP cavity signals
June 2012 – 2 weeks [Kim, Bett, Blaskovic, Perry (wk2), Christian (wk2)]
–
–
2 shifts per week
First week mainly monitoring IP cavity BPM signals, second week tests of new IPFB kicker
installed at IP.
•
More details can now be found on ATF Twiki/elog system.
•
Attention slowly shifting from upstream dual-phase FB system to IP feedback
–
See talk on IP feedback (P. Burrows) Thursday 14:40 JST!
Upstream FB status
• Since ~2010, aim with upstream system has been to demonstrate
similar level of performance in K2 loop as seen with K1, and observe
correction downstream.
– Programme of modification to processors in 2011 (Better matching of
cable lengths into hybrid, suppression of pickup on ADC sampling
clocks, BP-filtering FPGA clock)
• Focused on studying performance limitations of the system:
processor noise, systematic effects of LO phase jitter wrt bunch, and
understanding discrepancy between observed FB performance and
expected limit:
– Observed correction level at FB BPMs < 0.4 microns
– Measured (apparent) resolution of BPMs: 1 – 3 microns!!
• Previously knew about differing sensitivity of FONT BPMs to phase
jitter between bunch and LO (especially at P1), and strongly
suspected this was limiting contribution to apparent resolution, but
only earlier this year realised strong correlation between measured
position jitter and LO phase jitter
– Driven programme for early this year – trying to compensate for LO
phase jitter in the feedback loop.
Dec 2011
December 2011 FB results
Feedback BPMS
Dec 2011
December 2011 FB results
Witness BPMS (1)
Dec 2011
December 2011 FB results
Witness BPMS (2)
Dec 2010
Feedback examples (14 Dec 2011)
Second Bunch
Run6_141211
Upstream FB status
• Since ~2010, aim with upstream system has been to demonstrate
similar level of performance in K2 loop as seen with K1, and observe
correction downstream.
– Programme of modification to processors in 2011 (Better matching of
cable lengths into hybrid, suppression of pickup on ADC sampling
clocks, BP-filtering FPGA clock)
• Focused on studying performance limitations of the system:
processor noise, systematic effects of LO phase jitter wrt bunch, and
understanding discrepancy between observed FB performance and
expected limit:
– Observed correction level at FB BPMs < 0.4 microns
– Measured (apparent) resolution of BPMs: 1 – 3 microns!!
• Previously knew about differing sensitivity of FONT BPMs to phase
jitter between bunch and LO (especially at P1), and strongly
suspected this was limiting contribution to apparent resolution, but
only earlier this year realised strong correlation between measured
position jitter and LO phase jitter
– Driven programme for early this year – trying to compensate for LO
phase jitter in the feedback loop.
Oct-Dec 2011
BPM resolution tests (parasitic)
October 2011 – 3 processors on P2 (Charge ~1000-1500 cnts,
Jitter 3-4 microns)
Proc1
Proc2
Proc3
0.55
0.56
0.60
0.56
0.54
0.51
0.53
0.40
0.35
0.50
0.35
0.33
0.45
0.44
0.35
0.50
0.43
0.36
Minimum resolution based
on noise alone
Board #
Method
P1 soln
P2 soln
P3 soln
BOTH
1 3-BPM fit
2 3-BPM fit
2-on-1 pairwise
3.01
1.49
0.39
0.61
0.79
0.67
0.61
0.80
0.40
BOTH
1 3-BPM fit
2 3-BPM fit
2-on-1 pairwise
3.38
2.25
0.39
0.70
0.77
0.53
0.70
0.78
0.36
 /m
December 2011 – 2 processors on P1,P2,P3

Upstream FB status
• Since ~2010, aim with upstream system has been to demonstrate
similar level of performance in K2 loop as seen with K1, and observe
correction downstream.
– Programme of modification to processors in 2011 (Better matching of
cable lengths into hybrid, suppression of pickup on ADC sampling
clocks, BP-filtering FPGA clock)
• Focused on studying performance limitations of the system:
processor noise, systematic effects of LO phase jitter wrt bunch, and
understanding discrepancy between observed FB performance and
expected limit:
– Observed correction level at FB BPMs < 0.4 microns
– Measured (apparent) resolution of BPMs: 1 – 3 microns!!
• Previously knew about differing sensitivity of FONT BPMs to phase
jitter between bunch and LO (especially at P1), and strongly
suspected this was limiting contribution to apparent resolution, but
only earlier this year realised strong correlation between measured
position jitter and LO phase jitter
– Driven programme for early this year – trying to compensate for LO
phase jitter in the feedback loop.
Dec 2011
LO phase scans example (02/12/11)
Dec 2011
LO phase scans (Nov-Dec summary)
Dec 2011
BPM/LO correlations
(1000 pulse parasitic dataset 13/12/11)
P1: ρ=0.86
P3: ρ=0.31
P2: ρ=0.01
FONTP1
FONTP2
20
FONTP3
10
15
P1 POS
P1 LO
P2 POS
P2 LO
P3 POS
P3 LO
5
15
10
0
10
-5
5
5
-10
0
0
-5
-15
0
100
200
300
400
500
600
700
800
900
-20
1000
0
100
200
300
FONTP1
500
600
700
800
900
-5
1000
5.5
0
6
5
5.5
-1.5
-2
-2.5
4.5
4
3.5
-3
10
15
20
2
-20
400
500
600
700
800
900
1000
4
3.5
3
2
2.5
P1 POS / m
300
2.5
3
5
200
4.5
5
P3 LO phase /deg
P2 LO phase /deg
-1
0
100
FONTP3
6.5
-3.5
-5
0
FONTP2
0.5
-0.5
P1 LO phase /deg
400
-15
-10
-5
P2 POS / m
0
5
10
1.5
-5
0
5
P3 POS / m
10
15
Dec 2011
Original residuals – drift subtracted
(1000 pulse parasitic dataset 13/12/11)
FONTP1 Bunch1 Resolution Residuals from Least Squares Fitting
FONTP2 Bunch1 Resolution Residuals from Least Squares Fitting
250
FONTP3 Bunch1 Resolution Residuals from Least Squares Fitting
300
250
: -0.04
 : 2.71
: -0.03
 : 1.19
: 0.01
 : 1.26
250
200
200
200
150
150
150
100
100
100
50
0
-10
50
50
-8
-6
-4
-2
0
2
Size of Residuals / m
4
6
8
0
-5
FONTP1 Bunch1 Resolution Residuals from Transfer Matrices
-4
-3
-2
-1
0
1
Size of Residuals / m
2
3
4
5
0
-4
FONTP2 Bunch1 Resolution Residuals from Transfer Matrices
300
200
200
200
150
150
150
100
100
100
50
50
50
4
6
8
0
-8
3
4
: 0.01
 : 2.17
250
-2
0
2
Size of Residuals / m
2
: -0.01
 : 2.17
250
-4
-1
0
1
Size of Residuals / m
300
: 0.01
 : 2.17
-6
-2
FONTP3 Bunch1 Resolution Residuals from Transfer Matrices
300
250
0
-8
-3
-6
-4
-2
0
2
Size of Residuals / m
4
6
8
0
-8
-6
-4
-2
0
2
Size of Residuals / m
4
6
8
Dec 2011
Resolution residuals – LO phase
jitter subtracted + drift removal
FONTP1 Bunch1 Resolution Residuals from Least Squares Fitting
FONTP2 Bunch1 Resolution Residuals from Least Squares Fitting
300
FONTP3 Bunch1 Resolution Residuals from Least Squares Fitting
300
250
: 0.00
 : 0.44
: 0.00
 : 0.43
250
250
200
200
150
150
100
100
50
50
: 0.00
 : 0.42
200
150
100
0
-1.5
-1
-0.5
0
0.5
Size of Residuals / m
1
0
-1.5
1.5
50
-1
-0.5
0
0.5
Size of Residuals / m
1
0
-1.5
1.5
-1
-0.5
0
0.5
Size of Residuals / m
1
1.5
FONTP3 Bunch1 Resolution Residuals from Transfer Matrices
FONTP1 Bunch1 Resolution Residuals from Transfer Matrices
FONTP2 Bunch1 Resolution Residuals from Transfer Matrices
250
250
250
: -0.01
 : 0.43
: -0.01
 : 0.43
: 0.01
 : 0.43
200
200
150
150
100
100
50
50
200
150
100
50
0
-1.5
-1
-0.5
0
0.5
Size of Residuals / m
1
1.5
0
-1.5
0
-1.5
-1
-0.5
0
0.5
Size of Residuals / m
1
1.5
-1
-0.5
0
0.5
Size of Residuals / m
1
1.5
Dec 2011
Resolution residuals – LO phase
jitter subtracted + drift removal
FONTP1 Bunch1 Resolution Residuals from Least Squares Fitting
FONTP2 Bunch1 Resolution Residuals from Least Squares Fitting
300
FONTP3 Bunch1 Resolution Residuals from Least Squares Fitting
300
250
: 0.00
 : 0.44
: 0.00
 : 0.43
250
250
200
200
Original - Fitting
150
: 0.00
 : 0.42
FONTP1
200
FONTP2
150
3.26
150
2.11
Original – Nominal matrices
100
100
2.56
50
1.23
50
Original (drift sub) – Nominal matrices
0
-1.5
-1
-0.5
0
0.5
Size of Residuals / m
1
0
-1.5
1.5
-1
LO subtracted - Fitting
-0.5
0
0.5
Size of Residuals / m
0.47
LO subtracted – Actual matrices
150
150
LO + drift sub - Fitting
0.52
250
-0.5
0
0.5
Size of Residuals / m
1
1.5
0.46
FONTP3 Bunch1 Resolution Residuals from Transfer Matrices
: -0.01
 : 0.43
200
200
0.48
150
0.42
0.43
LO + drift sub – Actual
100 matrices
100
-1
0.49
: 0.01
 : 0.43
200
0
-1.5
1.5
FONTP2 Bunch1 Resolution Residuals from Transfer Matrices
LO subtracted – Nominal
matrices
250
: -0.01
 : 0.43
1.20
2.14
1
FONTP1 Bunch1 Resolution Residuals from Transfer Matrices
250
1.99
100
2.44
Original (drift sub) - Fitting
50
FONTP3
0.41
100
0.41
50
50
0
-1.5
50
-1
-0.5
0
0.5
Size of Residuals / m
1
1.5
0
-1.5
0
-1.5
-1
-0.5
0
0.5
Size of Residuals / m
1
1.5
-1
-0.5
0
0.5
Size of Residuals / m
1
1.5
Mitigating against bunch phase
jitter wrt LO
•
Understand why see good correction at FB control
BPMS but not witness BPMs
–
•
FB system will couple relative phase jitter back into the beam
(e.g. synchrotron motion turns into vertical beam jitter)
Mitigation options
1.
Remove effects (eg synchrotron motion) in DR
–
2.
Immunise against effects in DR
–
3.
Feed-back/forward on beam in DR Hard
Feed-forward on the LO to track the bunch phase Easier
Subtract the phase jitter from position data OFFLINE, and
correct the feedback signal ONLINE
–
–
Easiest, OFFLINE already done, ONLINE requires firmware mods
Proposed solution in first instance
   k Q 

y  G
 I

Phase jitter compensation firmware
• Problem with new firmware discovered in the K1 FB loop
(K2 appears to be fine!) – compounded with several
other issues: amplifier failure, cable failure, DAQ
problems …
• Glitches appear to be firmware related (standard ‘nonphase compensating’ FW works fine)
• Problem needs careful debugging in lab with controlled
inputs to the board.
Summary
•
•
Feedback performance determined by three quantities: bunch-to-bunch
correlation (beam), resolution (processor), and gain (system)
Over the past year or so spent a lot of time and effort in understanding and
mitigating effects limiting resolution
– Minimising processor sensitivity to LO phase jitter – optimising the path lengths
to hybrid
– Reducing ADC noise pickup – timing jitter on ADC clocks
– Removing BPM sensitivity to phase jitter
• Now see very good resolution ~400 nm, in all BPMs, and perfect agreement between
machine model and fitting beam trajectory
•
Feedback – goal has been to reproduce excellent correction previously
seen in P2 at P3 also, and maintain this correct downstream
– Very good results obtained for P2,P3 (down to ~500-600 nm) correction factor 35 , but in general not preserved at witness BPMs
– Should be able to see better downstream corrections from the removing the
phase sensitivity of the BPMs in the feedback correction.
– Firmware with phase jitter compensation in the FB been tested, but showed
glitching in one FB loop – needs thorough testing in lab.
Spares
Oct 2010
ATF Damping Ring Multi-bunch Diagnostics
Modified feedback hardware for multi-bunch turn-byturn DAQ from ATF damping ring
• Up to 3 bunches,3 channels, from up to 2 BPMs
• Records 131,071 samples per pulse (up to 15% of
damping period for single bunch, single channel)
• Can record n-turns-in-m to vary time window and
resolution
Nov 2011
Bunch phase oscillations at extraction wrt LO
SumQ/SumI – ‘raw’ data
3 distinct frequencies:
0.03
10.8 kHz – synchrotron
0.02
0.01
434 Hz slow oscillation (unknown)?
0
735 kHz fast oscillation – aliased?
-0.01
-0.02
-0.03
-0.04
-0.05
Last 200 turns before extraction
0
2000
4000
6000
8000
10000
12000
All pulses : Bunch 3
5-15 kHz BPF
0.03
0.01
0.025
0.02
0.005
0.015
0.01
0
0.005
0
-0.005
-0.005
-0.01
-0.01
-0.015
-0.02
1.178 1.18 1.182 1.184 1.186 1.188 1.19 1.192 1.194 1.196 1.198
0
2000
4000
6000
8000
10000
12000
4
x 10
BPM processor resolution and FB
performance limitations
• Standard 3-BPM resolution method gives 'average' resolutions of 1
– 2 micron across 3-BPM system, however FB system performance
in P2-K1 loop show ~300 nm.
–
–
–
Believe we were lucky with processor at P2, and that all
processors have different resolutions due to different sensitivity
to LO jitter
Largest effect due to path length imbalance to hybrid (unique for
each processor) – larger residual from subtraction, more
susceptible to LO jitter
All processors optimised, to be tested in Autumn
– Even if resolution 'perfect', system performance still determined by
beam jitter conditions
–
–
Measured bunch-to-bunch correlations of >94% needed to make
useful correction on ~3 micron beam jitter (50 % needed to
break even)
Bunch 3 assumed to be on edge of ~310 ns EXT kicker pulse
Processor Improvements (2011)
• Hypothesis that discrepancy between FB results and resolution due
to sensitivity of measured position to LO phase jitter
– All processors/BPMs exhibit different sensitivity to LO jitter wrt beam.
(P2 just happens to be least sensitive.)
– Effects cancel for measurements using just one BPM, for example FB,
whereas measurements involving correlating positions across several
BPM, appear to have poor resolution.
– Largest effect due to path length imbalance to hybrid (unique for each
processor) – larger residual from subtraction, more susceptible to LO
jitter
• All processors optimised (summer 2011)
– Input cables optimised for matched path length at hybrid
– Sum loopback cables re-made to phase sum and difference channels
• Also, discovered and fixed problem with sampling jitter caused by
noise pickup on ADC clocks from FPGA (affected correlated
measurements across more than one BPM, hence contributed to
effective resolution)
Dec 2011
Feedback examples (14 Dec 2011)
Run5_141211
Dec 2011
Latency (Dec 2011) – not-optimised
Latency P3-K1:
~154 ns
Dec 2011
Kicker K1 gain scan (14/12/11)
Dec 2011
Kicker K2 gain scan (14/12/11)
Dec 2011
Original residuals
(1000 pulse parasitic dataset 13/12/11)
FONTP1 Bunch1 Resolution Residuals from Least Squares Fitting
FONTP2 Bunch1 Resolution Residuals from Least Squares Fitting
250
FONTP3 Bunch1 Resolution Residuals from Least Squares Fitting
250
: -0.05
 : 3.29
250
: 0.21
 : 2.30
200
: -0.12
 : 2.09
200
200
150
150
150
100
100
100
50
50
50
0
-15
-10
-5
0
5
Size of Residuals / m
10
15
0
-8
FONTP1 Bunch1 Resolution Residuals from Transfer Matrices
-6
-4
-2
0
Size of Residuals / m
2
4
6
0
-8
FONTP2 Bunch1 Resolution Residuals from Transfer Matrices
300
200
200
200
150
150
150
100
100
100
50
50
50
6
8
10
12
Size of Residuals / m
14
16
18
20
0
-20
6
8
: 10.12
 : 2.45
250
4
4
: -10.12
 : 2.45
250
2
-2
0
2
Size of Residuals / m
300
: 10.12
 : 2.45
0
-4
FONTP3 Bunch1 Resolution Residuals from Transfer Matrices
300
250
0
-6
-18
-16
-14
-12
-10
-8
Size of Residuals / m
-6
-4
-2
0
0
0
2
4
6
8
10
12
Size of Residuals / m
14
16
18
20