HOM Studies at the FLASH(TTF2) Linac

Nathan Eddy, Ron Rechenmacher, Luciano Piccoli, Marc Ross

FNAL

Josef Frisch, Stephen MolloySLAC

Nicoleta Baboi, Olaf HenslerDESY

FLASH Facility (formerly TTF2)

• 1.3 GHz superconducting linac– 5 current accelerating modules, with a further two planned for

installation.– Typical energy of 400 – 750 MeV.

• Bunch compressors create a ~10 fs spike in the charge profile.– This generates intense VUV light when passed through the

undulator section (SASE).• Used for ILC and XFEL studies, as well as VUV-FEL

generation for users.

dumpbunch compressors

collimator section

bypass line5 accelerating modules

gunundulators

FEL beam

Higher Order Modes in Cavities

• In addition to the fundamental accelerating mode, cavities can support a spectrum of higher order modes.

• Traditionally they are seen as “bad”.– Beam breakup (BBU), HOM heating, …

• Here we investigate their usefulness,– Beam diagnostics– Cavity alignment– Cavity diagnostics

TESLA Cavities

• Nine cell superconducting cavities.• 1.3 GHz standing wave used for acceleration.• Gradient of up to 25 MV/m.

– Addition of piezo-tuners and improvement of manufacturing technique intended to increase this to ~35 MV/m.

• HOM couplers with a tunable notch filter to reject fundamental.– One upstream and one downstream, separated by 115degrees

azimuthally.– Couple electrically and magnetically to the cavity fields.

Response of HOM modes to beam

1.6 1.7 1.8 1.9 2 2.1 2.2 2.3 2.4 2.5

x 109

107

108

109

Raw Spectrum

TE111dipole band

TM011Monopole band

TM110dipole band

Sample HOM Spectrum

• Beam Position Monitoring– Dipole mode amplitude is a function of the bunch charge and

transverse offset.– Exist in two polarisations corresponding to two transverse

orthogonal directions.• Not necessarily coincident with horizontal and vertical directions due

to perturbations from cavity imperfections and the couplers.• Problem – polarisations not necessarily degenerate in frequency.

– Frequency splitting <1 MHz (of same size as the resonance width).

• Beam Phase Monitoring– Power leakage of the 1.3 GHz accelerating mode through the

HOM coupler is approximately the same amplitude as the HOM signals.

• i.e. Accelerating RF and beam induced monopole modes exist on same cables.

– Compare phase of 1.3 GHz and a HOM monopole mode.

HOMs as a Beam Diagnostic

Narrow-band Measurements

• ~1.7 GHz tone added for calibration purposes.

• Cal tone, LO, and digitiser clock all locked to accelerator reference.

• Dipole modes exist in two polarisations corresponding to orthogonal transverse directions.

• The polarisations may be degenerate in frequency, or may be split by the perturbing affect of the couplers, cavity imperfections, etc.

• May be difficult to determine their frequencies.

Method

• Steer beam using two correctors upstream of the accelerating module.– Try to choose a large range of values in (x,x’)

and (y,y’) phase space.

• Record the response of the mixed-down dipole mode at each steerer setting.

BPMsaccelerating module

1 8

HOM electronicssteering magnets

electron bunch

Singular Value Decomposition (SVD) to Find Modes

• Collect HOM data for series of machine pulses with varying beam orbits

• Use SVD to find an orthonormal basis set.– Select 6 largest amplitude modes– Calculate mode amplitudes

• Linear regression to find matricies to correlate beam orbit (X,X’,Y,Y’), and mode amplitudes

• Use SVD modes and amplitudes to measure position on subsequent pulses

Singular Value Decomposition

jjj

j

jjnjn

j

VV

VV

S

S

UU

UU

X

1

11111

1

111

0

0

• SVD decomposes a matrix, X, into the product of three matrices, U, S, and V.– U and V are unitary.– S is diagonal.

• It finds the “normal eigenvectors” of the dataset.– i.e. “modes” whose amplitude changes independently of each other.– These may be linear combinations of the expected modes.

• Use a large number of pulses for each cavity.– Make sure the beam was moved a significant amount in x, x’, y, and y’.

• Does not need a priori knowledge of resonance frequency, Q, etc.– Similar to a Model Independent Analysis.

Predict position at one cavity from positions at adjacent cavities

X resolution ~ 6.1µm Y resolution ~ 3.3 µm

Cavity Alignment ACC5

• X: 240 micron misalignment, 9 micron reproducibility• Y: 200 micron misalignment, 5 micron reproducibility

Multi-Bunch Processing

Multi-Bunch Processing

Multi-Bunch Initial Results

HOM as BPM in DOOCs

VMEHOM

Front-End

DOOCsMatlab Vectors

Calibration Constants

DataBase

DisplayX, X’Y, Y’

HOM BPM Details

knk

n

jnj

n

kjk

j

AA

AA

XX

XX

VV

VV

1

111

1

111

1

111

n

n

n

n

knk

n

k

k

yy

yy

xx

xx

AA

AA

MM

MM

1

1

1

1

1

111

1,441

1,111

11111

Mode VectorsRaw Data

Amplitudes

Calibration Matrix

4D Position

k ~ 6, j ~ 100 to 4k

DESY System

• Need to read out raw data for mod*cav*coupler channels at 4k to 10k data points per for multibunch then perform dot products to determine mode amplitudes

• This requires a lot of I/O in the front-end (slow) and then a bunch multiply accumulates which must be done sequentially on the front-end processor

• The current system is unable to report a position for every pulse at 5Hz for single bunch even with only a few cavities per module enabled

Custom FPGA Based Board• Extreme flexibility inherent in

FPGA– Algorithms and functionality can

be changed and updated as needed

– Code base which can be used for multiple projects

– Intellectual Property (IP) cores provide off the shelf solutions for many interfaces and DSP applications

• The speed of parallel processing– Can perform up to 512

multiplies using dedicated blocks

• The Pipeline nature of FPGA logic is able to satisfy rigorous and well defined timing requirements

Dot Product FPGA Implementation

• Store mode vectors in FPGA RAM• Perform dot product (multiply accumulate) in FPGA for

digitized data as it arrives from ADC• Simply read out mode amplitudes which are available as

soon as data has arrived• Can perform calculation on all channels in parallel• Also able to store raw data in internal RAM

ADC

FPGA

xn*vn,jCoupler Data

Dedicated HOM BPM Digitizer

• Dedicated HOM Digitizer– Provide amplitudes in real time– Reduce front-end processor I/O and load by orders 3-

4 orders of magnitude• Maximum rate still limited by front-end I/O

– Provide bunch by bunch data for every pulse

• Design dedicated 8 channel digitizer– Modify existing design ~ 6 months– Commissioning time (have prototype already)– Conservative estimate of $200 per channel

Broadband System

• Broadband (scope-based) system– Monitor HOM modes up to 2.5GHz – Several simultaneous channels (4 or 8)– Limited dynamic range (8 bit scope)– Use for Phase measurement

Monopole Spectrum• Data taken with fast scope. Both couplers

for 1 cavity shown

Note that different lines have different couplings to the 2 couplers: More on this later

Monopole lines due to beam, and phase is related to beam time of arrival

Fundamental 1.3GHz line also couples out – provides RF phase

Analysis of Monopole Data

• Lines are singlets – frequencies are easy to find• Find real and imaginary amplitudes of the

waveform at the line frequency• Find phase angle for each HOM mode• Convert phases to times• Weight the times by the average power in each

line• Correct the scope trigger time using this

weighted average of the times• Calculate the phase of the 1.3GHz fundimental

relative to this new time

Beam Phase vs. RF Measurement During 5 Degree Phase Shift

Measure 5 degree phase shift commanded by control system

See about 0.1 degrees of rms

Summary• HOMs are useful for diagnostic purposes.

– Beamline hardware already exists.– Large proportion of linac occupied with structures.

• Beam diagnostics.– Accelerating RF and beam induced monopole HOM

exist on same cable.• No effect from thermal expansion of cables.• Can find beam phase with respect to machine RF.

– Dipole modes respond strongly to beam position.• Can use these to measure transverse beam position.

• Cavity/Structure diagnostics.– Alignment of cavities within supercooled structure.– Possibility of exploring inner cavity geometry by

examining HOM output and comparing to simulation.

Backup Slides

Intuitive modes?• This calibration matrix, M, shows how much of each SVD mode

contributes to the modes corresponding to x, y, (x’, y’).• Therefore, can sum the SVD modes to find the intuitive modes.

– Lack of calibration tone in the reconstructed modes, as expected.– Beating indicates presence of two frequencies, i.e. actual cavity modes

are rotated with respect to x and y.– Could rotate these modes to find orientation of polarisation vectors in the

cavity…

• Using SVD on the (n x j) cavity output matrix, X, produces three matrices.– U (n x j), S (j x j, diagonal), and V (j x j)

• V contains j modes.– These are the orthonormal eigenvectors.– “Intuitive” modes will be linear combinations of these.

• The diagonal elements of S are the eigenvalues of the eigenvectors.– i.e. the amount with which the associated eigenvector contributes to the

average coupler output.– It can be shown that the largest k eigenvalues found by SVD are the largest

possible eigenvalues.

• U gives the amplitude of each eigenvector for each beam pulse.

Using SVD

jjj

j

jjnjn

j

VV

VV

S

S

UU

UU

X

1

11111

1

111

0

0

Theoretical Resolution

• Corresponds to a limit of ~65 nm– Included 10 dB cable losses, 6.5 dB noise figure, and 10 dB

attenuator in electronics.• Need good charge measurement to perform

normalisation.– 0.1% stability of toroids, to achieve 1 um at 1 mm offset.– Not the case with the FLASH toroids.

• LO has a measured phase noise of ~1 degree RMS.– This will mix angle and position, and will degrade resolution.– LO and calibration tone have a similar circuit, and cal. tone has

much better phase noise.• Therefore, should be simple to improve.

2

2q

Q

RU

TkU bth 2

1Energy in mode – Thermal noise –

HOM Calibration Overview

DOOCs

VMEHOM

Front-EndMatlabControlCode

MatlabAnalysis

Code

TTF2CorrectorsBPMs, Etc

Display

Display

Matlab Data

Structure

Matlab Vectors

Steering Plots

Apply calibration to a different dataset

Practical system

• Can use 2 LO frequencies to mix both the 1.3GHz, and the 2.4GHz Homs to a convienent IF (~25MHz).

• Digitize with same system used for dipole HOM measurements.

• Filters will greatly improve signal to noise– Dual bandpass for 1.3GHz, and 2.4GHz– Risk introducing phase shifts from filters

• System low cost – couplers already exist, electronics is inexpensive.

HOM Downmix Board

Input and sample out

Bandpass filters

Pre-amplifier

Mixer

IF outputamplifier