Fast intra-train beam-based feedback/feed-forward systems Feedback On Nanosecond Timescales (FONT) Philip Burrows Douglas Bett, Neven Blaskovic Kraljevic*, Ryan Bodenstein, Talitha Bromwich, Glenn Christian**, Colin Perry, Rebecca Ramjiawan John Adams Institute, Oxford University Now at: *CERN; **DLS in collaboration with: KEK, KNU, LAL
Outline • Introduction: linear electron-positron colliders • Fast intra-train collision feedback • KEK Accelerator Test Facility: ILC collision FB prototype Coupled-loop FB system ATF IP feedback system • CLIC drive-beam phase feed-forward prototype 2
International Linear Collider (ILC) c. 250 Ge. V / beam 31 km 3
CLIC layout (3 Te. V) 1. 5 Te. V / beam 4
High luminosity is critical Event rate R = = luminosity x cross-section L σ σ is ‘probability of something being produced’: fixed by nature to maximise R need to maximise L For a collider, L = N 1 N 2 frep / ( 4π σx σy ) N 1 particles/bunch of type 1 collide with N 2 /bunch of type 2 frep is the bunch collision frequency σx (σy) is the bunch size in x (y) at the interaction point 5
Beam parameters ILC (500 Ge. V) Electrons/bunch Bunches/train 2 10**10 1312 Train repetition rate 5 Hz Bunch separation 554 ns Train length 730 us Horizontal IP beam size 474 nm Vertical IP beam size 6 nm Longitudinal IP beam size 300 um Luminosity 1. 8 6 10**34
Beam parameters ILC (500 Ge. V) CLIC (3 Te. V) Electrons/bunch Bunches/train 2 1312 0. 37 10**10 312 Train repetition rate 5 50 Hz Bunch separation 554 0. 5 ns Train length 730 0. 156 us Horizontal IP beam size 474 40 nm Vertical IP beam size 6 1 nm Longitudinal IP beam size 300 44 um Luminosity 1. 8 5. 9 10**34 7
Beam parameters ILC (500 Ge. V) CLIC (3 Te. V) Electrons/bunch Bunches/train 2 1312 0. 37 10**10 312 Train repetition rate 5 50 Hz Bunch separation 554 0. 5 ns Train length 730 0. 156 us Horizontal IP beam size 474 40 nm Vertical IP beam size 6 1 nm Longitudinal IP beam size 300 44 um Luminosity 1. 8 5. 9 10**34 8
Like firing bullets to hit in middle … 9
Except that … 10
Beam trajectory Ideal case: In reality, magnet and accelerating cavity offsets: Can cause: emittance growth, orbit perturbations beam instabilities 11
Beam delivery system design (ILC) 12
ILC interaction region Door Cavern wall Si. D 13
Collider Final Focus: aligned 14
Collider Final Focus: misaligned 15
Luminosity loss vs. beam separation Gaussian beam 16
Luminosity loss vs. beam separation Including beam-beam EM interaction 17
Interaction point feedback system e+ e. Crossing angle: 14 mrad (ILC) 20 mrad (CLIC) 18
Interaction point feedback system Ground vibrations cause beams to miss one another e+ e- 19
Interaction point feedback system Ground vibrations cause beams to miss one another e+ e. Outgoing beams deflected 20
Interaction point feedback system Ground vibrations cause beams to miss one another e+ e. Outgoing beams deflected Measure beam position 21
Interaction point feedback system Ground vibrations cause beams to miss one another Feedback circuit e+ e. Outgoing beams deflected Measure beam position 22
Interaction point feedback system Kick beam back into collision Ground vibrations cause beams to miss one another Feedback circuit e+ e. Outgoing beams deflected Measure beam position 23
System parameters ILC CLIC BPM resolution ~ 1 um Dynamic range +- 300 nm +- 10 nm Bandwidth > 10 MHz as large as possible Latency < 300 ns as fast as possible 24
FB latency beam Kicker BPM Amplifier Signal processor FB board 25
IP Feedback latency 1. Beam flight time IP BPM 2. Signal processing, FB calculation 3. Cable delays 4. Amplifier risetime 5. Kicker fill time 6. Beam flight time kicker IP 5 6 4 3 1 26 2
Analogue or digital? Time structure of bunch train: ILC (500 Ge. V): O (1300) bunches w. c. 500 ns separation CLIC (3 Te. V): O (300) bunches w. c. 0. 5 ns separation Feedback latency: ILC: can have bunch-by-bunch FB with O(100 ns) latency allows digital approach CLIC: can have intra-train FB with O(10 ns) latency requires analogue approach 27
ATF 2/KEK: prototype final focus Goals: 1) 37 nm beam spot (41 nm achieved) 2) Beam spot stabilisation at nanometre level 28
FONT 5 ‘intra-train’ feedbacks 29
ILC IP FB prototype Kicker BPM 1 e- Drive amplifier Analogue BPM processor Digital feedback 30 BPM 2 BPM 3
ILC IP FB prototype Kicker BPM 1 e- Drive amplifier Analogue BPM processor Digital feedback 31 BPM 2 BPM 3
Stripline BPMs 154 ns Excellent temporal resolution 32
Stripline BPM readout 33
BPM analogue signal processor 34
BPM signal processor latency Bench latency meas: 10 ns or 15 ns with amplifier stage 35
FONT 5 digital FB board Xilinx Virtex 5 FPGA 9 ADC input channels (TI ADS 5474) 4 DAC output channels (AD 9744) Clocked at up to 550 MHz (phase-locked to beam)
BPM system resolution Resolution = 291 +- 10 nm (Q ~ 1 n. C) 37
38
BPM spatial resolution: update Dec 2016 Two technical improvements to BPM signal processor: • 6 d. B attenuator before sum mixer used for high-charge operation • No-PLL firmware used to remove FONT 5 A board sample timing jitter relative to the beam Resolution = 157 +- 8 nm (Q ~ 1. 3 n. C)
FONT 4 drive amplifier • • • FONT 4 amplifier, outline design done in JAI/Oxford Production design + fabrication by TMD Technologies Specifications: +- 15 A (kicker terminated with 50 Ohm) +- 30 A (kicker shorted at far end) 35 ns risetime (to 90%) pulse length 10 us repetition rate 10 Hz 40
ILC IP FB prototype Kicker BPM 1 e- Drive amplifier BPM 2 BPM 3 Analogue BPM processor BPM resolution ~ 0. 15 um Latency ~ 130 ns Drive power > 300 nm @ ILC Digital feedback 41
Feedback correction of jitter Bunch 1 Incoming bunch jitter ~ 2 um Bunch 2 stabilised to 600 nm Bunch 3 stabilised to 450 nm 6
Coupled-loop feedback P 2 & P 3 used to drive K 1 & K 2 Beam position and angle stabilisation
Time sequence Bunch 2: corrected FB off FB on 44
Bunch 1 – bunch 2 correlation FB off FB on 45
Cavity BPM system near IP 46
IP cavity BPM system 47
Low-Q cavity BPMs Design parameters 48
Cavity BPM signal processing I I’ Q Q’ bunch charge 49 Honda
Cavity BPM outputs (2 -bunch train) Single-shot measurement I 204 ns Q 50
Interaction Point FONT System Analogue Front-end FPGA-based digital processor BPM processor Latency ~ 160 ns Kicker drive amplifier Beam Cavity BPM Strip-line kicker
IP feedback results Stabilising beam to ~70 nm
Recent updates • Major effort on cavity BPM resolution + multi-sample averaging in FB firmware better signal/noise better performance • REAL-TIME BPM resolution ~ 20 nm • 1 -BPM beam FB ~ 50 nm • 2 -BPM beam FB ~ 40 nm 53
CLIC layout (3 Te. V) 1. 5 Te. V / beam 54
CTF 3/CERN Beam control: feed-forward systems Precision cavity + stripline BPMs Beam size diagnostics Beam tuning techniques 55
CTF 3 phase FF prototype 56
CTF 3 phase FF prototype 1 mrad kick 8 degrees at 12 GHz 0. 2 degree resolution 57
Closed-loop FF tests (Dec 2016) Correction of phase along pulse: Performance vs. frequency: Reduction in mean phase jitter:
Closed-loop FF tests (Dec 2016) Reduction in mean phase jitter: Correction of phase along pulse: CLIC goals met! Performance vs. frequency:
Closed-loop FF tests (Dec 2016) Correction of induced phase oscillations: 1. 7 degrees 0. 3 degrees
Summary • • Fast intra-train beam feedback + feed-forward systems Built + tested prototypes at KEK/ATF 2 and CERN/CTF 3 Low latency: High-power: High bandwidth: High resolution: • ILC collision FB: • ATF coupled-loop: • ATF IP feedback: • 130 - 350 ns 10 - 50 k. W 20 - 50 MHz 20 – 150 nm single-shot dynamic range +- 300 nm / 250 Ge. V beam stabilised y, y’ 450 nm beam stabilised 40 nm CLIC drive-beam phase FF: beam arrival stabilised 50 fs 62
FONT Group alumni • • • • • Gavin Nesom: Simon Jolly: Steve Molloy: Christine Clarke: Christina Swinson: Glenn Christian: Glen White: Tony Hartin: Ben Constance: Robert Apsimon: Javier Resta Lopez: Alex Gerbershagen: Michael Davis: Doug Bett: Young-Im Kim: Neven Blaskovic: Davide Gamba: Jack Roberts: Riverbed Technology, California UCL faculty ESS staff SLAC staff BNL staff JAI staff Diamond LS SLAC staff DESY staff CERN Fellow start-up Cambridge CERN Fellow Lancaster Marie Curie Fellow, Cockcroft PSI postdoc CERN staff UBS, London CERN Fellow JAI, Oxford IBS Daejon CERN Fellow staff JAI
Backup material 64
High-power + B/W, low-latency amps NLCTA ATF 2 CTF 3 ATF 65
Beam-beam deflection vs. offset Vertical deflection angle (urad) 50 nm 400 urad Example for CLIC 380 Ge. V – qualitatively same for ILC, CLIC at all energies discussed 66 Vertical beam-beam offset Δy (um)
Zoom in Vertical deflection angle (urad) 1 nm 40 urad Example for CLIC 380 Ge. V – qualitatively same for ILC, CLIC at all energies discussed 67 Vertical beam-beam offset Δy (um)
ATF 2 design parameters 68
Latency issues • BPM – kicker separation: beam flight time at speed c cable delays at speed f * c, f = 0. 1 0. 9 • BPM signal processing time resolution vs. speed of response • Feedback calculation digitise information? very fast processors (FPGA) run up to ~ GHz • Amplifier risetime speed vs. power and bandwidth 69
Latency 70
FB system latency Kick to bunch 2 Latency: < 154 ns
FB system dynamic range Kick to bunch 2 Max. kick = 75 um (1. 3 Ge. V) = 400 nm (ILC 250 Ge. V beam) = 66 σy (ILC 250) meets ILC requirement of 50 σy
Dynamic range of feedback Incoming beam position scan = +-50 σy for ILC (design value) 6
Stabilising beam near IP 1. Upstream FB: monitor beam at IP 2. Feed-forward: from upstream BPMs IP kicker 3. Local IP FB: using IPBPM signal and IP kicker
FONT 5 system performance Bunch 1: input to FB FB off FB on 75
FONT 5 system performance Bunch 1: input to FB FB off FB on Bunch 2: corrected FB off FB on 76
Jitter reduction Factor ~ 3. 5 improvement 77
Feedback loop witness 78
Feedback loop 79 predict
Witness BPM: measure predict 80
Predicted jitter reduction at IP y FB off FB on y’ 81
Predicted jitter reduction at IP Predict position stabilised at few nanometre level… How to measure it? ! 82
Measured beam-size reduction at IP 83
Sample Integration and Resolution • Best resolution results were collected May 2017, at a charge of ~0. 5× 1010. • The geometric and fitted resolution can be improved through integration. For these results the integration was performed over 10 samples. • Geometric and fitted resolution show good agreement with each other. Geometric resolution – single sample (nm) 47 BPM Fitting resolution – single sample (nm) Geometric resolution – integrated (nm) 20 Fitting resolution – integrated (nm) IPA 47 20 IPB 47 20 IPC 62 21 Data with thanks to T. Bromwich Tuesday 23 rd January 2018 Rebecca Ramjiawan 84
IP Feedback Results – 1 -BPM Mode • Tuesday 23 rd January 2018 Rebecca Ramjiawan 85
1 -BPM Feedback Results Tuesday 23 rd January 2018 Rebecca Ramjiawan 86
IP Feedback Results – 2 -BPM Mode • Tuesday 23 rd January 2018 Rebecca Ramjiawan 87
2 -BPM Feedback Results Tuesday 23 rd January 2018 Rebecca Ramjiawan 88
CLIC drive-beam phase feed-forward 89
Скачать презентацию Fast intra-train beam-based feedback feed-forward systems Feedback On Nanosecond
538f15b6a1e1bb5fee66138270012071.ppt