Advanced Topics of Cavity BPMSimulation and Analysis
17.01.2012 A. Lyapin et al, BPM workshop
1
A. Lyapin, S. Boogert, G. Boorman, F. Cullinan, N. Joshi (JAI/RHUL, UK)A. Morgan, G. Rehm (Diamond Light Source, UK)
M. Ross (Fermilab, USA)A. Aryshev, Y. Honda, T. Tauchi, N. Terunuma, J. Urakawa (KEK, Japan)
A.-Y. Heo, E.-S. Kim, H.-S. Kim, Y. I. Kim (KNU, Korea)J. Frisch, D. McCormick, J. Nelson, T. Smith, G. White (SLAC, USA)
Nquestions > Nanswers
ΔNquestions ~ eNanswers
What this talk is about
17.01.2012 A. Lyapin et al, BPM workshop
2
• Some problems we encountered while working at the ATF/ATF2, none of which are of a deciding importance, but taken together define the performance of a CBPM system Tuning, timing and thermal variations Calibration and beam jitter Timing and gain variations and long-term stability Non-linearities and electronics noise Closely spaced bunches
• Will try demonstrating the importance of system simulations, including the whole chain Beam motion Cavity Electronics Digitisers Digital processing
Accelerator test facility
17.01.2012 A. Lyapin et al, BPM workshop
3
• Low-emittance facility, test system for 35 nm beam size next LC beam delivery system• Very dense with instrumentation: wire scanners, OTRs, laserwires, laser interference BSM• Relies mainly on cavity BPMs, currently ~ 40 in total
Cavity beam position monitor system
17.01.2012 A. Lyapin et al, BPM workshop
4
BPM test areaStrip line/Cavity BPMs(mounted rigidly)
C-band BPMs(mounted on movers)
S-band BPMs(movers)
IP region(4 BPMs)
System resolution
SFs, Large BBA offset
200 nm
IP1
Lines indicatecut, at which BPM is labelled
bad
No attenuators in this region
40 nm
IP1
yx
17.01.2012 5A. Lyapin et al, BPM workshop
• SVD using a few BPMs surrounding the one of interest and calculate the residual• Usually a high residual signals for a re-calibration• In some cases it indicates more fundamental problems
Large offsets (between the BPM and quad) and consequent saturation• This display is now an online tool for operators
Cavities+Electronics
A. Lyapin et al, BPM workshop
6
C-band S-band
• Single stage image reject mixer, converting down to 20-30 MHz• Front-end LNA in C-band, all but 3 attenuated• Digitise at ~100 MHz
Parameter C-band S-band
Frequency, MHz 6422 2888
QL ~6000 ~1800
x-y isolation, dB
45 30 (prev. 16)
• C and S-band cylindrical cavities with 4 symmetric couplers• Slot-coupled structure for monopole mode rejection, based on cavities
previously used in NanoBPM experiment• Tuners for adjusting x-y coupling
17.01.2012
Digital processing
• Digitised signal is processed– Digital IQ mixer– Digital filtering (Gaussian filter)– LO frequency tuned to IF frequency for each channel– Same processing for position and reference
• Amplitude and phase are sampled at one point• Position phasor normalised by the reference to remove the
charge and length dependency, and reference the phase to the beam arrival
• The real and the imaginary parts of the resulting phasor are referred to as I’s and Q’s (in phase and in quadrature phase with the reference)
• I and Q carry information on position, angle and tilt (separated using calibration)
17.01.2012 A. Lyapin et al, BPM workshop
7
Amplitude
Phase
Tuning and timing
• The frequency of the LO signal used in digital demodulation needs to be tuned precisely to the frequency of the cavity
• Set a relatively large offset to make S/N high
• Look at the phase of the demodulated signal trying to flatten it adjusting the LO frequency
• If the signal is saturated, the sampling point slides to the right, the amplitude must be extrapolated, but the phase apparently stays the same
17.01.2012 A. Lyapin et al, BPM workshop
8
Trigger jitter/drift
17.01.2012 A. Lyapin et al, BPM workshop
9
• Due to small differences between the position and reference cavities, changes of the trigger timing cause changes of the phase, even when the phase is flattened along the waveform
• Measuring the beam arrival time for each beam pass and referring the sampling point to the arrival time, it’s possible to compensate for this effect
€
VpVr
=ApAre−ΔΓ(ts −t0 )e jΔω (ts −t0 )
t0
ts
RF/trigger adjustments
• Problem when using DR RF ramp for dispersion measurements
• Both the accelerator RF-derived LO (locked to 714 MHz) and the beam trigger change when activating the ramp
• Problem exaggerated by a large frequency difference between the position and reference cavities in the S-band system (~50 MHz)
• Less pronounced, but still visible in C-band system (Δf~1-2 MHz)
17.01.2012 A. Lyapin et al, BPM workshop
10
S. Boogert, RHUL
RF/trigger adjustments
• Investigated several filters and chose the one providing a few points on the front of the pulse for the t0 reconstruction while not smoothing it (10 MHz)
• Correction works a lot better but much worse than wished for• Thinking of replacing the reference!
17.01.2012 A. Lyapin et al, BPM workshop
11
S. Boogert, RHUL
Temperature changes
• CBPMs are mounted on the quads in ATF2, so where the current is high, temperature variations can be considerable
• Gradients of tens of kHz/0C are expected, in this example 48 kHz for a 2.9 GHz BPM
• Eventually, can use cavity’s phase gradient for the temperature measurement, but need to have an idea of the scale of the variations
17.01.2012 A. Lyapin et al, BPM workshop
12
N. Joshi, RHUL
Calibration
• Cavity BPMs need to be calibrated in order to determine: position scale IQ rotation of the position signal
• suppress angle/tilt• Can calibrate by either:
moving the beam• may introduce angle
moving the BPM• more precise• need precision movers
• Calibration: position changed in steps I and Q averaged over several
beam passes fit Q vs I to get the rotation fit rotated I (I’-position) to get the
scale
17.01.2012 A. Lyapin et al, BPM workshop
13
Calibration and long-term stability
17.01.2012 A. Lyapin et al, BPM workshop
14
2010:
Calibration and jitter subtraction
17.01.2012 A. Lyapin et al, BPM workshop
15
• Easiest approach – collect multiple pulses, average for each step
• Works well, until you face large excursions• Can collect some pulses before doing the
calibration to establish correlations with the BPMs upstream, and then subtract the jitter
• Even better results when running the correlations on the calibration data including the mover/bump position in the matrix
F. Cullinan, RHUL
Calibration stability - Scale
17.01.2012 A. Lyapin et al, BPM workshop
16
F. Cullinan, RHUL
• Histograms for repeated calibrations processed using all three approaches• Clearly, using the SVD directly on the calibration gives the best results• Going down from 10-15% in X and 5% in Y to <1%
Calibration stability – IQ rotation
17.01.2012 A. Lyapin et al, BPM workshop
17
F. Cullinan, RHUL
• Similar results for the IQ rotation• The improvement is not as impressive as for jitter, if mainly positional, does not affect the
rotation as much, and the large variations due to timing had been fixed
Calibration stability - Summary
17.01.2012 A. Lyapin et al, BPM workshop
18
F. Cullinan, RHUL• Scales vary by less than 1% - is that enough?• Measuring 100 μm x 0.01 = 1 μm precision. Limited by movers?• Probably need to calibrate each BPM several times• Use the active gain/phase monitoring system at that level
Non-linearities and electronics noise
17.01.2012 A. Lyapin et al, BPM workshop
19
• Electronics noise well understood• Need to keep in mind the digitisers are noisy too• Non-linearities are less well explored, even the cavity is not perfectly linear when it comes
to measuring nm• A measurement of resolution as f(x,y) across the dynamic range would be interesting,
and can be done at ATF, so we should probably include it in our program
Closely spaced bunches
17.01.2012 A. Lyapin et al, BPM workshop
20
• 2 bunches separated by 182 ns
• BPM moved in 100 um steps
• Contribution from the 1st bunch is propagated to the sampling point of the 2nd and phasors are subtracted
N.Joshi, RHUL
System simulations
17.01.2012 A. Lyapin et al, BPM workshop
21
Beamdynamics
BPMresponse
Analogprocessing
Digitisation
Digitalprocessing
Conversion toposition
position,size, length
modes,sensitivities
gain, NF, non-linearity
non-linearity,sensitivity
granularity,noise
calibration errors
• Hopefully, it is clear by now that in systems aiming at high resolution, precision and accuracy all the elements of the processing chain need to be analysed at least at a basic level
• Not always there is an obvious “bottleneck” issue limiting the resolution, but a number of factors contributing
• Propagating the signal through the whole system looks like the right idea• Depending on the aim of a simulation, the models included may be less sophisticated
System simulations: Example 1
17.01.2012 A. Lyapin et al, BPM workshop
22
• Beam is tracked using a lightweight tracker code (Serpentine)• Beam jitter is calculated at each CBPM location and its effect on the calibration is
estimated
F. Cullinan, RHUL
System simulations: Example 2
17.01.2012 A. Lyapin et al, BPM workshop
23
Y.I.Kim, KNU
• Start with the beam optics and get bunch position at each BPM
• Propagate the errors and noise (without doing the full signal simulation)
• Do resolution calculation as for the beam data• Describes the system’s behavior pretty well…
Without electronics noise
With electronics noiseBeam measurement
Diagnostic simulation package?
17.01.2012 A. Lyapin et al, BPM workshop
24
• Probably all beam diagnostics work in a similar way: certain properties of the beam are transformed into electrical signals
• There are certain sensitivity, noise and background signals• The output is digitised at some point• Additional digital processing and calibration may be applied to provide the actual
measurement• Maybe we should think of a package simulating diagnostic tools?• A lightweight version of it could be useful for beam dynamics simulations• Anyway, just a thought…