Electron Beam Diagnostics at the Radiation
Source ELBE
P. Evtushenko, U. Lehnert, P. Michel, C. Schneider, R. Schurig,
J. Teichert
Radiation Source ELBE,
Research Center Rossendorf, Postfach 510119, 01314 Dresden, Germany
Abstract. In the research center Rossendorf, the radiation source ELBE, based on a super
conducting LINAC, is under construction. In the year 2001 the first accelerating module was
commissioned. The electron beam parameters like emittance, bunch length, energy spread were
measured. Here we present results of the measurements as well as the methods used to make the
measurements. In the ELBE injector, where electron beam energy is 250 keV, the emittance was
measured with the aid of a multislit device. Emittance of the accelerated beam was measured by
means of quadrupole scan method and is 8 mmxmrad at 77 pC bunch charge. Electron bunch
length was measured using the coherent transition radiation technique. At the maximum design
bunch charge of 77 pC the RMS bunch length was measured to be 2 ps. A set of online
diagnostic systems is also under development. One these include a system of stripline beam
position monitors is also described here. A BPM resolution of about 10 um was achieved using
logarithmic amplifier as the core element of the BPM electronics. A system of beam loss
monitors based on the RF Heliax cable working as an ionization chamber is intended to be
another online diagnostic system.
TRANSVERSALE EMITTANCE MEASUREMENTS
Introduction
The ELBE accelerator is a conventional design with an injector section followed by
an accelerator section. The injector itself consists of the thermionic gun and a beam
line section wherein two bunchers, focusing and steering elements as well as
diagnostics are installed. The bunched beam is injected into the first accelerating
module with an energy of 250 keV. After the accelerating section the beam is
available with energies up to 20 MeV; in a later stage of improvement up to 40 MeV
are planned. For some of the applications at ELBE, the emittance for a specific current
or bunch charge is an important input. Two commonly used approaches are applied at
ELBE to determine the emittance. The sampling of partitions of the phase space
named multislit or pepper pot method. The determination of the beam matrix through
the measurement of the beam diameter at different values of a focusing magnetic lens
namely a quadrupole or solenoid scan method. In the injector section both methods
CP648, Beam Instrumentation Workshop 2002: Tenth Workshop, edited by G. A. Smith and T. Russo
© 2002 American Institute of Physics 0-7354-0103-9/02/$19.00
010
313
mulitslit / pepper pot and solenoid scan are used while for the accelerated beam the
quadrupole scan method is used to determine the emittance.
The Multislit Method
In the injector section of ELBE different masks can be put into the beam to register
slices of the beam on a luminescent screen. In the past we have used slit masks with
slit dimensions of 100 jim and slit distances of 1 mm and 3 mm, while now we us
pepper pot masks to measure vertical and horizontal emittance in one step. The image
on the luminescent screen is recorded by a vidicon camera, which delivers the image
to a frame grabber card in a PC. From the projection of the beam through the slits onto
the screen the space location and the divergence of the phase space samples can be
measured. These parameters are extracted by a multi Gauss fit from a profile slice of
the image data. From the fit parameters a RMS emittance is calculated.
Quadrupole Scan Method
A focusing magnetic lens changes the beam diameter at a fixed distance. If the
beam diameter on the screen is recorded for different magnetic strengths of the lens
the elements of the beam matrix in front of the lens can be extracted. From the matrix
elements the transverse emittance can be calculated. In the injector we use solenoids
while after the accelerator cavities quadrupoles are used as focusing elements. In the
injector the multislit and the solenoid scan method agree to within a 20% percent
level, which gives a hint to the precision of the emittance determination.
Results
On Fig. 1 all the relevant emittance measurements are summed as a function of the
bunch charge. The squares show older measurements performed with the multislit
method in the injector at a micro pulse rate of 13 MHz. They agree well with model
calculations of our thermonic gun [1] (solid curve) especially in the range of higher
bunch charges. For applications in the radiation physics an attempt was made to
reduce the emittance by cutting off the outer part of the beam through apertures, (see
the triangles). With a transmission of 10% the emittance can be improved significantly
but the micro pulse rate has to be raised to reach the claimed mean current. Repeating
the multislit measurements (circles) at the injector in a next stage of extension it was
found that the emittance has improved. From an image of the cathode it could be seen
that the emission characteristic has changed, now the main part of the electrons come
from a smaller region of the cathode surface. These data compared with model
calculations performed with a smaller cathode radius (2.5 mm) agree well with this
assumption (dashed curve). The measurements after the accelerating section with the
quadrupole scan method (stars) show that within the errors of both methods there is no
visible increase of the emittance due to the accelerating cavities.
314
14-,
• 13 MHz multi slit (old data)
A 260 MHz (3mm aperture with 10% transmission)
——Simulation of the electron gun
- - - - Simulation with rcath= 2.5 mm
• 6.5 MHz quadrupol scan
• 13 MHz multi slit
1
0,1
10
100
bunch charge / pC
FIGURE
FIGURE1.1. Emittance
Emittanceas
asaafunction
function of
of bunch
bunch charge
charge for
for different
different settings
settings of
of ELBE.
ELBE.
BUNCH
BUNCH LENGTH
LENGTH MEASUREMENTS
MEASUREMENTS
Introduction
Introduction
TheELBE
ELBE electron
electron gun
gun is
is aa thermionic
thermionic triode,
triode, which produces electron pulses with
The
charge of
of up
up to
to 100
100 pC
pC and
and an
an electron
electron energy
energy of 250
250 keV. Electron pulse length is
aacharge
about of
of 450
450 ps
ps RMS
RMS at
at the
the gun
gun output.
output. The electron bunch is compressed in the
about
injector down
down to
to 10
10 ps
ps with
with the
the help
help of
of two
two buncher
buncher cavities.
cavities. The first sub-harmonic
injector
buncheroperates
operates atat 260
260 MHz,
MHz, and
and the
the second
second one
one works
works at the fundamental frequency
buncher
of the
the LINAC
LINAC which
which isis 1,3
1,3 GHz.
GHz. The
The ELBE
ELBE accelerating
accelerating module
module consists of two
of
TESLA type
type nine-cell
nine-cell cavities.
cavities. Since
Since the
the electrons
electrons become relativistic in the first
TESLA
first
cavity, the
the phase
phase of
of the
the RF
RF field
field there
there is
is the
the key
key parameter
parameter influencing
influencing the bunch
cavity,
length atat the
the output
output of
of the
the accelerating
accelerating module.
module. Bunch
Bunch length
length behavior was studied at
length
different
bunch
charges
as
a
function
of
RF
field
phase
in the
the first
first cavity.
cavity. The bunch
different bunch charges as a function of RF field phase in
length
was
measured
during
commissioning
of
the
first
acceleration
module using
length was measured during commissioning of the first
well-known
coherent
transition
radiation
(CTR)
techniques.
The
method
uses a
well-known coherent transition radiation (CTR) techniques.
Martin-Puplett
interferometer
to
measure
an
autocorrelation
function
of
the
CTR
Martin-Puplett interferometer to measure an autocorrelation
pulse.
The
measurements
were
done
at
electron
beam
energy
of
about
12
MeV.
Here
pulse. The measurements were done at electron beam energy of about 12 MeV. Here
we present
present results
results of
of the
the measurements,
measurements, description
description of the experimental setup, and the
we
dataevaluation
evaluationprocedure.
procedure.
data
315
The Coherent Transition Radiation Technique
The coherent transition radiation technique is already well described in a number of
publications [2] -[4]. Here we just want to remind one of the key issues of this method.
As soon as a charged particle crosses a boundary of two media with different
dielectrics or magnetic constants, transition radiation (TR) is produced [5], If an
electron bunch consisting of N electrons crosses such a boundary, each electron of the
bunch radiates the TR. For a wavelength shorter than the bunch length, the radiation
power is proportional to N, since for every electron there is an electron radiating in
opposite phase and the coherent term is equal to zero. Part of this radiation lies in the
optical range and is used nowadays very widely for beam profile measurements. For a
wavelength longer than the bunch length, all electrons radiate almost in one phase and
since the phase difference is constant, the radiation is coherent and therefore the power
of the radiation is proportional to N2. Of course, there is a transition region when the
spectral power density goes from N to N2. Obviously, the position of this transition
depends on the bunch length; hence measurements of the transition radiation spectrum
can give information about the bunch length. We want to point out here also that for
the ELBE bunch charge 77 pC N is about 5.5xl08, which leads first to a huge
difference of N and N2 and, moreover, almost all power of the transition radiation is in
the coherent part. PARMELA simulations of the electron beam transport predict the
bunch length to be in the picosecond range.
Experimental Setup
An aluminum foil as thin as 10 jim, stretched to a frame, was used to produce the
CTR. The view screen is oriented 45° to the beam direction. Thus the backward CTR
part is propagating perpendicular to the electron beam. We use a crystal-quartz
window for the output of the CTR from the beam line. A Martin-Puplett
interferometer is used to measure the autocorrelation function of the CTR pulses. A
parabolic aluminum mirror with focal distance of 200 mm is used to transform the
divergent transition radiation into a quasi-plane wave, which then goes to the
interferometer. Wire grids are used as a polarizer, analyzer and also as a beam splitter
in the interferometer. The grids are made of gold covered tungsten wires, with
diameter of 20 jim. The grid period is 100 jim. Another parabolic mirror at the output
of the interferometer focuses the radiation on the input windows of the detectors. We
have used two Golay cell detectors for the measurements with the interferometer. The
theory of the Martin-Puplett interferometer is well developed and shows that an
interferogram is the autocorrelation function:
(t-T))- dt
(1)
of the incoming into interferometer radiation pulse f(t) [6]. The Wiener-Kintchine
theorem proves that the Fourier transform of the autocorrelation function is the power
spectrum. Hence we measure the power spectrum of the CTR pulse. The power
spectrum defines uniquely the amplitude of the components of the frequency domain
representation of the pulse. But information about the relative phases of the different
316
components is lost in the interferometric measurements. This way a direct pulse
components isfrom
lost the
in power
the interferometric
measurements.
This way a direct pulse
reconstruction
spectrum is not
possible.
reconstruction from the power spectrum is not possible.
Bunch Length Estimation From The Interferogram
Bunch Length Estimation From The Interferogram
The following procedure is used to evaluate the measurement data. First of all an
The following procedure is used to evaluate the measurement data. First of all an
assumption about a longitudinal charge distribution is made. Then the power spectrum
assumption about a longitudinal charge distribution is made. Then the power spectrum
of the distribution is calculated. A filter function of the interferometer is applied to the
of the distribution is calculated. A filter function of the interferometer is applied to the
power spectrum to get the modified power spectrum. This modified power spectrum is
power spectrum to get the modified power spectrum. This modified power spectrum is
compared with the measured power spectrum, which in turn is the Fourier transform
compared with the measured power spectrum, which in turn is the Fourier transform
of the interferogram. Finally we have to change parameters of the assumed charge
of the interferogram. Finally we have to change parameters of the assumed charge
distribution so long, until the calculated power spectrum fits well to the measured one.
distribution so long, until the calculated power spectrum fits well to the measured one.
In practice we assume a Gaussian distribution of the charge, which makes it possible
In practice we assume a Gaussian distribution of the charge, which makes it possible
toto do
the calculations analytically. The filter function we use corresponds to a low
do the calculations analytically. The filter function we use corresponds to a low
frequency
and the
the wire
wire grid
grid spectral
spectral response
response as
as
frequency cut
cut off
off which
which comes
comes from
from the
the detector
detector and
well
as
from
the
crystal-quartz
window
transmission.
The
function
was
chosen
well as from the crystal-quartz window transmission. The function was chosen
empirically
empirically to
to be:
be:
ν 
− 
ν0 
4
(2)
F filtrer (ν ) = 1 − e
(2)
here
such aa filter
filter function,
function, all
all experimental
experimental
here the
the Vo=0.1
ν0=0.1 THz.
THz. The
The criterion
criterion is
is that
that with
with such
data
have
a
good
fit.
Under
these
assumptions
we
have
the
fitfunction:
data have a good fit. Under these assumptions we have the fitfunction:
2
ν 
 ν 

−  
−

Q


ν
F (ν , Q, σ ) = 1 − e  0   ⋅
⋅ e σ⋅ 2 
 2π



(3)
(3)
which
measured power
power spectrum.
spectrum.
which was
was used
used for
for the
the bunch
bunch length
length extraction
extraction from
from the
the measured
Result
length. Some
Some examples
examples of
of
Result of
of aa fit
fit isis the
the aσ parameter,
parameter, which
which is
is the
the RMS
RMS bunch
bunch length.
the
are shown
shown on
on Fig.
Fig. 2.
2.
themeasured
measured data
data with
with their
their fit
fit functions
functions are
4
0,2
0.3
frequency, THz
100
105
110
Cavity 1 phase, degree
FIGURE
FIGURE2.2. Measured
Measured power
power spectrums
spectrums (points)
and
and their
their fit
fit functions
functions (curves).
317
Bunch length
length vs.
vs. cavity
cavity one
one phase.
phase,
FIGURE 3. Bunch
Results Of The Measurements
As was mentioned above, during commissioning of the first accelerating module, it
was important to study methodically the work of the accelerator. During this phase we
have measured the RMS bunch length as a function of the RF field phase in the first
cavity at different bunch charges. The results of the measurements are shown on Fig.
3. The bunch length was measured to be a minimum about 2 ps RMS at 77pC bunch
charge. The results are in good agreement with the PARMELA simulations [7].
BEAM POSITION MONITORS
Introduction
ELBE will be used for experiments in radiation physics, nuclear physics and
neutron physics. It will also be the driver for the infrared free electron laser (PEL).
Obviously, an accelerator needs a system for the beam position measurements. Also
the position of the electron beam has to be controlled at the target in any experiment
and inside the undulator of the PEL. In the case of the ELBE accelerator, the required
resolution of the beam position measurements is about 100 um. We decided to use
stripline BPM, since it is well known that with the BPM one can easily achieve the
resolution. The BPM system has to work in all possible modes of the accelerator,
which are:
• The PEL mode: repetition rate of 13 MHz, CW or macropulsed beam with
macropulse length from 100 jis up to 37 ms. The maximum bunch charge in
this mode is 77 pC, which corresponds to an average current of 1 mA.
• The mode for the radiation physics experiments: 260 MHz repetition rate, the
bunch charge of about 0.4 pC or average current of 100 jiA.
• The diagnostic mode: the repetition rate from 13/128 MHz up to 13 MHz, CW
or pulsed beam, maximum bunch charge of 77 pC.
Stripline BPM Design Experience
As reported before [8,9] a system of stripline beam position monitors (BPM) is
under construction for the ELBE accelerator. The BPM from the JLab PEL machine
was a prototype for the first version of the ELBE BPM. The construction was only
changed to be appropriate for the ELBE beam line diameter and for the accelerator
fundamental frequency of 1.3 GHz. Two BPMs were manufactured and successfully
tested at ELBE. The BPM was used during the injector characterization. A resolution
of about 10 jim at 1 mA beam current was achieved which is about ten times better
than required.
However, we have faced some problems during the BPMs manufacturing. First of
all, the SMA feedthroughs, which have to be electron beam welded to the BPM, are
very sensitive to mechanical stress. About half the feedthroughs were broken in the
first welding attempt. Another problem is the length of the electrodes. To form the
318
electrodes, a pipe with a diameter exactly as the beam line has to be made with four
cuts. Because of mechanical stress in the beam pipe, the electrodes can change their
position with respect to the pipe center. This leads to an impedance of the transmission
line slightly different from the designed 50 Q. The electrode displacement also
imposes some danger for the feedthrough. Besides all of this we could not put the
BPM at some desirable places because of its length and the allowable space. All these
reasons required some changes in the BPM design. The length of the BPM electrode
in general is chosen such that the BPM has maximum sensitivity to the fundamental
frequency of an accelerator. To satisfy this condition the length should be
L=(2n+l)A/4, where n is integer number and A is the wavelength of an accelerator
fundamental frequency (1.3 GHz in our case). The first ELBE BPM has the electrode
length of about 3A/4. The main idea of the redesign was to make the strip length A/4.
At first, an electrical model of the A/4 BPM was built and tested on a wire test bench.
It was demonstrated that the resolution of the model is not worse than the 3A/4 BPM
resolution. After this successful test we have constructed a BPM with the electrode
length of about A/4. In the construction we used another type of the feedthrough,
which is not welded to the BPM but sealed to it with a CF flange. Brazing is used in
the construction instead of expensive electron beam welding. The feedthrough can be
replaced any time, thus the BPM can be repaired without removing it from the beam
line, which can save a lot of time. Total length of the BPM is 85 mm from flange to
flange. The A/4 BPM is in factor two cheaper than the first version. For beam tests two
A/4 BPMs were manufactured without any technical and technological problems. The
new BPM was tested at ELBE and has demonstrated resolution and dynamic range not
worse than the first version of the BPM with an electrode length of 3/4A. Finally, it
was decided to use the A/4 BPM at ELBE.
BPM Electronics
Stripline BPMs are similar to directional couplers. Energy from the fields generated
by the electron beam is coupled out. The signal power on each of the four ports is
proportional to the beam current and the beam position with respect to each "coupler"
in a x/y-coordinate system. The stripline BPM electronics has to convert the power
signal from the BPM into a voltage, which is then fed into an ADC system. The BPM
output power ranges from -80 dBm (lowest detectable beam current with no beam
displacement) up to -25 dBm (maximum beam current, 10 mm off axis). The signal
flow through the different conversion stages is shown on Fig. 4 (one channel). The 1.3
GHz fundamental signal is selected by an external 5-pol inter-digital filter (ID1300)
having a 3 dB bandwidth of 8 MHz. Signal attenuation of the coax cables from the
accelerator cave to the diagnostics room plus filter attenuation is about of 5 dB, but
exactly measured for each channel and considered in the software for calibration. The
first stage inside the electronics housing is an integrated MMIC amplifier with 20 dB
gain and a high IP3 level to avoid spurious signal generation. The central part is the
AD8313 logarithmic detector made by Analog Devices Inc., converting the RF signal
into a DC voltage. The dynamic input range of this detector is -65 dBm to -10 dBm
for less than 1 dB conversion error. The two following rail-to-rail operational
amplifiers LT1630 from Linear Technology shift and amplify the detector voltage to -
319
5 V to +5 V which meets the input range of the ADC. Low offset voltage and
temperature drift are essential here. In order to minimize crosstalk problems, each
PCB-board contains the electronics for two channels housed in milled boxes, ensuring
proper shielding against stray signals from the nearby klystrons, which may badly
affect the measurement at low input levels (note, that the maximum klystron output is
10 kW, or +60 dBm and the minimum detectable electronic input is -85 dBm, so any
leakage path must be prevented). Two such boxes and a power supply make the
system for one BPM. Three units have been built and tested during commissioning of
the first part of the ELBE accelerator. Sufficient sensitivity along with ruggedness
against environmental effects, and long-term stability were proven. Crosstalk between
channels was found to be lower than expected. Therefore, the next revision will house
the electronics for four channels in one box and 10 such boxes in one 19" rack. It is
necessary to include a sample and hold (S&H) stage to make the system work at
repetition rates below 13 MHz, which was not intended at the beginning of the
development. A version with S&H has been built but not tested yet.
ID1300
VNA25
AD8313
LT1630
LT1630
AD783
ADC
—o
gam
S&H
CPCI510
12 bit
FIGURE 4. The BPM electronics scheme.
ACKNOWLEDGMENTS
We would like to thank Kevin Jordan from JLab for all useful discussions of the
electron beam diagnostic at ELBE. Bernt Wustmann did the biggest part of the BPM
technical design. We are very grateful to Frank Gabriel for his help with the BPM
electronics design. Roland Jainsch and Dieter Proehle helped us a lot with the BPM
data acquisition system and the BPM software design.
REFERENCES
1.
2.
3.
4.
5.
6.
J. Teichert, et al., FZ Rossendorf, Wiss.-Tech. Ber. 1999-2000, 7-8
Happek U., et al., Phys. Rev. Lett. 66, 1967 (1991)
Lihn H. C., et al., Phys. Rev. E53, 6413 (1996)
Geitz M., et al., Proceedings of the 1999 Particle Accelerator Conference, New York, 2172-2174
Jackson J. D., Classical Electrodynamics, Second Edition, New York, 1975
Lesurf J., Millimeter-Wave Optic, Devices & Systems, New York, Adam Hilger, Bistrol and New
York, 1990
7. A. Buechner, private communication
8. Evtushenko P. et al., FZ Rossendorf, Wiss.-Tech. Ber. FZR-319, 2001, 13.
9. Evtushenko P. et al., Proceedings of the 5-th European Workshop on Diagnostic and Beam
Instrumentation, Granoble, 2001.
320