- Email: [email protected]
Waveguide monitors—a new type of beam position monitors for the TTF FEL
ARTICLE IN PRESS
Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
Waveguide monitors—a new type of beam position monitors for the TTF FEL U. Hahna, T. Kampsb, R. Lorenzc, W. Rieschd, H.J. Schreiberd,*, F. Tonischd a
DESY Hamburg, 22607 Hamburg, Germany b BESSY Berlin, 12489 Berlin, Germany c Westdeutscher Rundfunk, 50600 Koln, . Germany d DESY Zeuthen, Platanallee 6, 15735 Zeuthen, Germany Received 11 August 2003; received in revised form 15 March 2004; accepted 15 March 2004
Abstract The operation of a Free Electron Laser at the TESLA Test Facility requires the electron trajectory to be aligned with precision of few micrometer in overlap with the photon beam. To achieve this goal precise position monitoring of the electron beam in the narrow undulator gap is mandatory. A beam position monitor based on a new microwave concept was designed, built and tested. It consists of four ridged waveguides which couple by small slots to the magnetic field of the electron beam. The paper reports the concept, the design and tests of prototypes in a test bench in the laboratory and with beam. TTF FEL measurement results are also presented and possible improvements are discussed. r 2004 Elsevier B.V. All rights reserved. PACS: 29.27F Keywords: Beam position measurement; Waveguide monitor
1. Introduction In 1995 an international collaboration proposed the construction of the superconducting accelerator of the TESLA Test Facility (TTF) with an integrated Free Electron Laser (FEL) at the Deutsches Elektronensynchrotron DESY in Hamburg [1]. This facility was set-up, commissioned and brought into operation, for a review see e.g. [2]. First lasing of the FEL was observed in 2000 [3]. *Corresponding author. Tel.: +49-33-762-7298. E-mail address: [email protected] (H.J. Schreiber).
The operation of a Free Electron Laser in the Self Amplified Spontaneous Emission mode (SASE) requires the electron trajectories to be aligned with very high precision in overlap with the photon beam. Simulations showed that the mean variation between the photon path (mainly defined by the undulator axis) and the electron trajectory must be kept below 20 mm in order to reach saturation in the undulator [4]. To ensure this overlap, the undulator modules of the TTF FEL were equipped with beam position monitors (BPM) and steering coils. They have to be integrated into the undulator vacuum chamber inside the small undulator gap. For the beam
0168-9002/$ - see front matter r 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.nima.2004.03.178
ARTICLE IN PRESS U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
BEAM BEAM y S2
S1
S3
S4
241
z
S3
S1
S2
S4
x
x Fig. 1. Schematics of a waveguide BPM with four channels and four coupling slots (profile, top view).
position monitors two schemes with different sophisticated designs were followed. Pick-up monitors were installed into the first two undulator modules, while the third module was equipped with a new type of waveguide beam position monitors. An innovative feature of this monitor is the way the coupling to the beam field is realized, in combination with the special shape of the waveguides. Both types of BPMs were designed for position measurements with a precision of 10–20 mm: For phase 2 of the TESLA Test Facility with a design electron beam energy of 1 GeV; the electron bunches are compressed to approximately 50 mm in order to provide a peak current of 2:5 kA required for FEL [5]. To measure the position of very short bunches accompanied by large peak currents, waveguide monitors are favorable due to their simple coupling to the beam via a small slot. Discharges, which might occur in pick-up monitors between the antenna and the vacuum chamber material are strongly suppressed or even excluded for the waveguide monitor concept. Both complementary monitor schemes were developed within the TESLA collaboration and experiences will drive the future solution. Other monitor types e.g. resonant devices [6] are not suitable due to possible intolerable wakefields. The paper summarizes the development of the waveguide BPMs in its various stages up to their application at TTF FEL phase 1 commissioned in 2000. It includes the mechanical design of the monitor essentially predetermined by the small undulator gap, the cross section of the vacuum chamber and the undulator magnets. Prototypes
were constructed and tested in the laboratory and with beam. Finally, the third TTF FEL undulator chamber was equipped with waveguide BPMs and data were taken. The results obtained in each stage of the monitor development will be discussed, as well as problems which were observed and possible solutions. In the past, several notes on selected topics of the waveguide monitor such as construction and electronics aspects or calibration questions were presented. However, neither of them summarized all aspects of this type of monitor in a compact manner. The paper also updates and supersedes results already published.
2. Design and basic characteristics of the waveguide monitor Basic mechanical design parameters of the waveguide monitor are determined by their position within the small undulator gap of 12 mm; with only horizontal access. The BPMs have to be an integrated part of the vacuum chamber, taking also into account the beam pipe with 9:5 mm inner diameter. The shape of the waveguides is designed in a way which reduces the size and realizes at the same time an effective coupling to the beam field. At the end of each waveguide, a coaxial adapter with a vacuum feedthrough is flange mounted. All waveguide structures were fabricated using Electrical Discharge Machining (EDM). The principle design of this new BPM type is sketched in Fig. 1. A fractional part of the electromagnetic beam field is coupled by four
ARTICLE IN PRESS 242
U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
slots1 into ridged waveguides arranged symmetrically around the beam pipe. Due to restricted vertical space, two waveguide pairs (S1–S3 and S2–S4) were separated by approximately 40 mm in beam direction. Each waveguide is inclined by about 35 with respect to horizontal axis to achieve highest possible vertical separation of the coupling slots. Each coupling slot (Fig. 2) of a BPM can be regarded as a radiating magnetic dipole oriented parallel to the magnetic field lines of the passing beam. The T-ridged waveguide enables a coupling to the magnetic field of the beam. The signal amplitude equals 1% of the induced magnetic field amplitude for a centered beam. An adapter is necessary to match the impedance of the waveguide to that of a standard 50 O coaxial cable. This adaptor is realized in two stages resulting in a lowloss transmission line. In the first part a short rectangular coaxial line (with a length of l=4) is connected to the T-ridged waveguide, followed by a round coupling to match the inner conductor of the vacuum feedthrough. Fig. 3 sketches the three transition steps. An improved version of the monitor (Prototype II) was designed as an integral part of the vacuum chamber.2 It is shown in Fig. 4, with the waveguide profile and position of the slots marked. The electron beam passes the flat chamber through the vacuum flanges at both ends of the prototype. This design provides a better broadband transition between waveguide and vacuum feedthrough due to a l=4-transformer (Fig. 3b) produced in one step together with the waveguide ridge structure. The extended pin of the coaxial vacuum feedthrough plugs into a groove in the transformer matched to fit the pin. This technique provides good electrical contact as well as correct positioning and vacuum tight mounting of all elements [7]. 1
Dimensions of these slots are 3:2 3:2 mm2 ; with a depth of 0:3 mm: 2 For prototype I (see also Fig. 3a), the inner part of the waveguide ridge was an integrated part of the vacuum flange inserted into the rectangular waveguide structure. Due to machining tolerances electrical contact between the inner part of the ridge and the vacuum chamber was not always given. The manufacturing process turned out to be complicated and not well suited for a mass production of monitors.
SLOT
WAVEGUIDE
FEEDTHROUGH TRANSFORMER SIGNAL
MAGNETIC FIELD ELECTRIC FIELD BEAM
1/4 VACUUM CHAMBER
Fig. 2. Sketch of the coupling mechanism of the waveguide monitor.
Prototype II also achieved a less coupling amplitude variation than prototype I. Detailed MAFIA [8] simulations were performed in order to obtain optimized parameters for the coupling slot and transition to the vacuum feedthrough. More details on design parameters and their variation obtained from simulations and analytical estimations can be found in Ref. [9]. The upper working frequency of the waveguide is limited by the beam pipe cut-off at 17:9 GHz: For higher frequencies, fields generated by interaction between the beam and its surrounding material can couple into the waveguide and might lead to interferences with beam-induced signals. The waveguide working frequency was designed to be 12 GHz; mainly to use commercially available signal processing components. This frequency is sufficiently above the waveguide fundamental cutoff frequency at 9 GHz and below 13:2 GHz of its next higher order mode. When the electrons pass the undulator beam pipe wakefields are generated. A thereby introduced additional energy spread in the electron bunch might influence SASE. The impact of the waveguide BPMs has been estimated with the result that the overall beam spread is tolerable for TTF FEL phase I operation [9,10].
3. Measurements in the laboratory Laboratory measurements of new beam diagnostics elements are useful to confirm basic predictions of simulations, to study specific char-
ARTICLE IN PRESS U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
(a) PROTOTYPE I
243
(b) PROTOTYPE II & FINAL DESIGN FEED THROUGH
WAVEGUIDE
WAVEGUIDE FEEDTHROUGH
TO BEAM PIPE
TO BEAM PIPE WAVEGUIDE RIDGE
WAVEGUIDE RIDGE
λ/4 ELECTRIC FIELD
T-RIDGE WAVEGUIDE
SHORT PIECE OF RECTANGULAR COAX WAVEGUIDE
EXTENDED INNER CONDUCTOR OF VACUUM FEEDTHROUGH
Fig. 3. Transition steps between the T-ridged waveguide and feedthrough.
0
2
4
6
8
10cm
SLOT
WAVEGUIDE PROFILE 0
2
4
6
8
10cm
Fig. 4. Photo of Prototype II. Top: total view; bottom: view onto sidefront with two waveguide ports and tapped wholes for mounting to the vacuum feedthroughs.
acteristics under idealized conditions and to develop calibration algorithms. A laboratory set-up was built, where the electromagnetic field propagating with an electron
beam has been emulated by a coaxial system of a stretched copper–beryllium wire (250 mm diameter) inside the monitor. An impedance transformer was inserted between the signal generator
ARTICLE IN PRESS 244
U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
and the BPM prototype to minimize reflections of the electromagnetic wave. At the end of the setup the electromagnetic power had to be damped by an absorber to prevent reflections. The BPM itself was moved with respect to the wire by two stepping motors, with a position repeatability of better than 1 mm: Detailed studies were performed to provide good electrical contact, correct impedance matching, best absorption and minimal signal losses [9]. Signals excited in the waveguides were filtered at 12 GHz and amplified. A LABVIEW application running on a PC was used to control the power meter read out as well as movements of both stepping motors. The schematic layout of the electrical facet is shown in Fig. 5. Narrowband measurements were carried out to check the linearity of the monitor and to measure the transfer characteristics of the four channels. A signal generator delivered a 12 GHz cw-signal, at a power level of 14 dBm: Cross scans around the monitor center were performed on both the x- and y-axis, with a typical range of 1–2 mm in steps of X10 mm: The normalized signal of an ideal BPM channel can be calculated using the image charge model [11]. This model describes the locally induced wall current density in a conducting vacuum chamber surface for a pencil beam. This wall current density is equivalent to the local magnetic field, to which the BPM couples with the slots. The magnetic field induced by the electrons can be computed using Biot–Savart’s law and results for a point-like
charge to b2 r2 ; b2 þ r2 2br cosðfi yÞ
Si ¼
for slot i at radius b and angle fi and a beam at position ðr; y). The assumptions made in the model are to a good approximation valid in our case since 1 the slots can be considered as point-like (only 10 of the beam pipe circumference) and beam movements only occurred near the center of the monitor. For large off-sets (several mm) one has to integrate over the complete slot size. Due to individual transfer characteristics of each channel the measured voltage of a channel depends on the response function Si modified with the scalar channel gain gi : Vi ¼ Agi Si : Here, A is a factor describing the amplitude of an input signal. The gain factors are introduced to adjust imbalances among individual transfer characteristics of the channels i ¼ 1–4. Gain parameters were estimated by measuring induced voltages Vi;j ði ¼ 1; y; 4Þ for j ¼ 1; y; n wire positions in the BPM aperture. A fit to the image charge model Mi;j provides the coupling factor for each channel Mi;j ¼ gi
b2 r2j b2 þ r2j 2brj cosðfi yj Þ
i ¼ 1; y; 4 and a ¼ ðg1 ; g2 ; g3 ; g4 Þ;
where a is the array of fitting parameters to be determined.
WIRE
RF IN
RF CABLE
PC WITH GPIB INTERFACE AND LABVIEW
FILTER & AMPLIFIER STEP MOTOR CONTROL
Mði; j; aÞ;
j ¼ 1; y; m
RF OUT
SIGNALGENERATOR
ð1Þ
RELAY
POWERMETER
CONTROL
READ OUT CONTROL & READ OUT
Fig. 5. Schematics of the electrical facet of the laboratory test bench.
ARTICLE IN PRESS U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
The waveguide signals were filtered, amplified and measured. In order to exclude effects from imbalances among amplifiers, cables and filters, detected voltages were divided by their individual amplification factors. At each position, four voltages were recorded such that for sufficient position measurements the number of data points exceeded significantly the number of free parameters. The gain results normalized to g1 ¼ 1 are listed in Table 1. Their variation of few times 102 is to a great extend ascribed to manufacturing tolerances. Simulation studies provided demands for mechanical tolerances [9]. For example, to keep the root–mean–square spread of the coupling amplitudes below a level of 2%, the slot width and the height have to be manufactured within 750 mm and the depth of the slot must be within 7100 mm: Since it is difficult to achieve such tolerances, calibration for each waveguide monitor is needed. The image charge model was also used to estimate the angle of each coupling slot by using the gain factors obtained in the previous step. Results of this fit are in fair agreement with the Table 1 Gain values normalized to g1 ¼ 1 and coupling slot angles as derived from a fit to the image charge model Channel
Gain
Slot angle (deg)
1 2 3 4
1:000070:0018 0:905470:0017 1:066870:0017 1:009470:0019
35:9470:13 33:4670:09 35:9270:14 34:2070:13
design value of 35:51 (with respect to x-axis), as shown in Table 1. Slight systematic differences can be accounted for by a small tilt between the BPM reference frame and the test bench system. We noticed that the gain parameters of Table 1 are unchanged within uncertainties when the fitted fi values instead of the nominal slot angles are used in the fitting procedure. The sensitivity describes the change of an induced signal amplitude for a given beam offset. Induced voltages of channels 1–4, normalized with corresponding amplifications and damping factors, were used to construct virtual signal functions L ¼ S2 þ S3 and R ¼ S1 þ S4 for the horizontal (U ¼ S1 þ S2 and D ¼ S3 þ S4 for the vertical) direction, see also Fig. 1. The sensitivity Sx (Sy ) for the horizontal (vertical) direction is then defined as the slope between set position in x (or y) and the logarithmic ratio of L and R (or U and D) L U xSx ¼ 20 log and y Sy ¼ 20 log : ð2Þ R D Values for log-ratios as function of the wire position are plotted in Fig. 6. A linear performance for jxj; jyjp500 mm around the BPM center can be observed. The resulting sensitivities are listed in Table 2, with results from MAFIA SParameter computations and from the image charge model. Good agreement between test bench measurements and expected values is noted. In addition, broadband measurements up to 16 GHz were performed to check matching stages,
PROTOTYPE I
PROTOTYPE II 3
1 0 -1
vertical horizontal
-2 -3 -600 -400 -200 0 200 400 POSITION [micron]
600
LOG AMPLITUDE RATIO [dB]
LOG AMPLITUDE RATIO [dB]
3 2
245
2 1 0 vertical -1
horizontal
-2 -3 -600 -400 -200 0 200 400 POSITION [micron]
Fig. 6. Measured sensitivities for both prototypes (test bench).
600
ARTICLE IN PRESS 246
U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
to observe cut-off frequencies and to look for trapped resonances in a waveguide. The transfer characteristics for each BPM channel were studied by measuring the transmission parameter S21 using a network analyzer. The transmission was rather small until the waveguide cut-off frequency of about 9 GHz was reached. Above this frequency high-pass characteristic was observed for all channels with acceptable differences caused by imperfect mechanical manufacturing.
4. Prototype measurements with beam Measurements under real beam conditions were considered as an experimental proof of the operation principle of the waveguide monitor. In addition, basic signal processing electronics using commercially available components were implemented and tested successfully. First beam test data were obtained with Prototype I at the CLIC Test Facility at CERN [12]. Results on strength measurements of the four coupling amplitudes and on sensitivities in x and y directions were consistent with test bench data and MAFIA simulations [9]. Prototype II was installed in the S-Band Test Facility at DESY (SBTF) [13]. Typical beam parameters relevant for single bunch measurements are summarized in Table 3. Table 2 Sensitivities as derived from Prototype II measurements on the wire test bench, MAFIA simulations and the image charge model
Sx (dB/mm) Sy (dB/mm)
Measurements
Simulation
Model
5:7970:04 4:1070:04
5.78 3.88
5.78 4.10
Table 3 S-Band Test Facility beam parameters during measurements Energy Repetition rate Bunch charge Bunch length Bunch size
E fr q sz sx;y
46:7 MeV 1 Hz 0:2 nC 4–5 mm B1 mm
The electron beam with desired time structure and energy was generated and accelerated by three sections. Prototype II has been installed in a drift region between two quadrupole triplets where the beam envelope should be equal for horizontal and vertical planes. A two-dimensional stepping motor frame with the BPM integrated was computer controlled to move the monitor in well-defined steps perpendicular to the beam. After centering the beam with quadrupole scans, first measurements were performed by displaying beam induced signals on a digital sampling scope. For the scope trigger, a beam related signal was taken from a stripline BPM downstream. With this set-up, 12 GHz components of beam related induced signals were clearly visible. In a next step, downconverted monitor signals for all four channels were recorded. The 12 GHz LO signal needed for down-conversion was derived using the 4th harmonic of the accelerating frequency (2:99 GHz) as induced in the stripline BPM. This signal had a constant amplitude and was phase locked to the beam. After filtering, the remaining IF signal was amplified. Signals were recorded for several waveguide monitor positions transverse to the beam direction.3 After accounting for different amplitude factors of the BPM channels—attenuation in nearly all components as well as electronics gain—the signal function Sx as defined in Eq. (2) was derived from a linear fit to the data. In Fig. 7, Sx is plotted versus transverse beam position for a rough scan (step width Dx ¼ 500 mm) and a fine scan (Dx ¼ 50 mm). The results can be reasonably interpreted using the image charge model (continuous curves). The horizontal sensitivity derived from the fine scan results to Sx ¼ ð5:2770:14Þ dB=mm which is somewhat lower than the wire test bench value of ð5:7970:04Þ dB=mm: This can be explained by the fact that the beam size is larger than the linear region of the BPM. We have also compared the beam position in x; determined by using an algorithm developed for beam-based calibration [14], with the set beam positions by the stepping 3 Steering along the vertical axis was not possible due to a system failure of one stepping monitor.
ARTICLE IN PRESS U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252 FINE SCAN
ROUGH SCAN -0.10
0.6 DATA IMAGE CHARGE MODEL
0.2 0 -0.2
-0.15 NORMALIZED SIGNALFUNCTION
NORMALIZED SIGNALFUNCTION
0.4
DATA LINEAR MODEL
-0.20 -0.25 -0.30 -0.35
-0.4
-0.40
-0.6
-200
-2.0 -1.5 -1.0 -0.5 0 0.5 1.0 1.5 2.0 2.5
(a)
247
POSITION x [mm]
(b)
0 200 400 600 POSITION x [micron]
800
Fig. 7. Relative signal function Sx for the horizontal plane: (left) for a broad scan width of Dx ¼ 500 mm; (right) for a narrow scan with Dx ¼ 50 mm: Three values around 400 mm were excluded from the fit because of strong beam position jitter.
motor. Linearity for off-sets up to about 72 mm has been observed with high precision [9]. The measurements at the S-band Test Facility were limited by the granularity of the internal ADC of the oscilloscope. For operation at the TTF FEL an ADC was developed for optimum pulse response with significant better resolution (see Section 5.1).
FEL operation was identical to the Prototype II design. Special mounting, alignment and support systems for the flat and long vacuum chamber allowed for a proper insertion of the chamber in the undulator gap and its positioning within a tenth of a millimeter in vertical and horizontal direction. 5.1. Signal processing electronics
5. TTF FEL measurements For phase 1 of the TTF FEL the undulator [15] was segmented into three sections. The vacuum chamber of the third section was equipped with ten microwave monitors as an integrated part of the chamber. The vacuum chamber is a flat 4:5 m long structure with the cross-section of 11:5 128 mm2 of extruded aluminium, with a central aperture of 9:5 mm to guide the electron beam. In order to guarantee a tough control of the beam orbit inside the undulator one BPM and one correction coil were required per FODO quadrupole of the undulator magnet structure. The correction coils were located in between the BPMs and allowed for horizontal and vertical steering of the beam. The waveguide holes with their relative complex contours and the small coupling slots at the beam aperture were fabricated by EDM. At the end of each waveguide a coaxial adapter with a special UHV compatible RF feedthrough was flangemounted. The design of the monitors for TTF
The signal processing electronics have been divided into a 12-to-1 GHz down-converter followed by a 1 GHz amplitude demodulator and an analog-to-digital converter (ADC). In the following we describe the electronics in general, for more details we refer to Ref. [16]. The four signals of a monitor were grouped together into one down-converter circuit box, positioned close to the vacuum chamber to minimize signal attenuation. This down-converter was designed for a total gain of about 5 dB with a Noise figure of 20 dB: After passing semi-rigid cables of 1 to 2 m length, signals from each BPM were processed within the down-converter in a parallel way by four identical hardware chains. This is shown in Fig. 8 for one channel in a simplified manner. The broadband monitor signal was filtered at 12 GHz by a 5-pole interdigital bandpass with a bandwidth of 140 MHz: For a bunch charge of 1 nC the signal amplitude after the bandpass was estimated to be 35 mV at 50 O
ARTICLE IN PRESS U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
248
DOWN-CONVERTER (12 GHz /1 GHz)
12GHz Bandpass
Limiter ACLM 4571F
Attenuator MC 2710
Amplifier NE 32484
Mixer 1GHz MC 2710 Bandpass f0 = 1 GHz B = 64 MHz -10 dB -4.5 dBm
ATT: DAT=150
+11 dB NF= 0.6 dB
NF=11.1 dB
-9.5 dBm
-8 dBm
-5.5 dBm
-5 dBm
REF: +8 dBm, 12.3 GHz
-6 dB
-1.5 dB
-15.5 dBm
-2.5 dB
-4.5 dB
LO : +8 dBm, 11.29 GHz
+19 dB NF = 3.8 dB
0 dBm
f0 = 12.29 GHz B=140 MHz
-0.5 dB
Linedriver INA-10386
-19 dBm
20dBDir.Coupler
G = 5 dB NF= 18.6 dB
Fig. 8. Block scheme for the 12-to-1 GHz down-converter electronics.
impedance. A 20 dB directional coupler before this filter allowed to introduce an 8 dBm 12:3 GHz reference signal for calibration purposes. The generator stabilized by a dielectric resonator was integrated in the down-converter box, too. After the bandpass filter a limiter protected the subsequent components from signal levels above a certain threshold, which might be caused by e.g. large mis-steering of the beam. The limiter is followed by an attenuator and an amplifier to avoid saturation in the down-converter circuit as well as to ensure an acceptable signal-to-noise ratio. 12-to-1 GHz down-conversion was realized within a combination of a mixer and a 1 GHz bandpass filter. The 8 dBm local oscillator signal at 11:3 GHz needed for the frequency downconversion was generated by a circuit identical to the generator mentioned above. After output signal amplification the signal was passed to the demodulator in the racks outside the tunnel via a 25 m long 50 O coaxial cable. Down-converted 1 GHz signals were processed by amplitude demodulation. The simplest asynchronous (diode) demodulation was realized, with the disadvantage of a relatively small linear range of the demodulator transfer function.4 A block diagram of the demodulator for one BPM channel is shown in Fig. 9. The first two circuits represent a
variable amplifier with a range of 1379 dB; so that an optimal signal level for the ADC input range can be adapted. The signal was then detected and filtered by a 25 MHz bandpass. At the end, the demodulated pulse was amplified again and passed to the ADC. The four channels of a single BPM were grouped together and installed in a VME–RF cassette. The final signal with a width of about 25 ns has to be as jitter-free as possible and bunch-synchronized transferred to the ADC. With a noise rms of about 6 mV; the beam displacement resolution caused by the signal processing electronics only was estimated to be about 12 mm: The noise from the demodulator alone (no input signal) was measured to about 300 mV rms and, hence, negligible compared to the overall noise level. Three signal attenuation possibilities for dynamic range optimization in the electronics were realized to adjust the small dynamic range of the demodulator: (i) at the entrance of the 12-to1 GHz down-converter (fixed, any value), (ii) before the 12 GHz amplifier (remote via I 2 C; max. attenuation 20 dB) and (iii) at the entrance of the demodulator (remote via I 2 C; max. attenuation 20 dB). Careful signal attenuation control is therefore required during data taking in order to obtain undisturbed pulses.5 The reference signal
4 In a first attempt we applied an I=Q demodulation using a commercial circuit. However, the obtained signals were very noisy resulting to unacceptable position resolutions of about 100 mm or more [17].
5 In principle, an extension of the dynamic range of the electronics is possible by means of software correction procedures which have to rely on the transfer characteristics of each BPM channel.
ARTICLE IN PRESS U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
249
DEMODULATOR Driver MAX 4213
BP -8.2 dB
+4.6 dBm (PEP)
-7.4 dB +12 dBm (PEP)
ATT: DAT = 0 /255
-12 dBm (PEP)
71 m Vpp / 50 Ohm
SNR :70 dB
G = +24 dB, NF = 3.8 dB -2 / -20 dB
Bandpass 25 MHz
+5.6 dB
G = +12 dB
+2 dBm (PEP) SNR : 40 dB
Detector BAT14/03W
0.28 Vpp / 50 Ohm
Amplifier INA-10386
-3.6 dBm (PEP)
Attenuator HMC121G8
Fig. 9. Block scheme of the demodulator electronics.
for calibration purposes was used to remove the electronics offset in advance of data taking. Signals from the analog electronics were fed into an 8-channel high speed ADC with a resolution of 14 bit: It converted the signals at a rate of 1 MSample=s and stored the data in an on-board memory (depth 128 kSample=channel; [18]). The ADC was triggered and clocked by the TTF timing system. In order to sample the signal at the peak the clock was adjusted using a digital delay integrated on the ADC board, with 5 ns step size. The input range of the ADC was set to 71 V with an input impedance of 50 O: The ADC values were read out by the standard TTF data acquisition system.
of the horizontal beam displacement achieved by different correction coil currents Ix ; for a bunch charge of about 1 nC: For waveguide channel notations we refer to Fig. 1. Voltages S1 and S4 decrease with increasing Ix (negative off-sets in x), whereas S2 and S3 grow, as expected. Fig. 10b shows the beam intensity defined as the sum of all four voltages against corrector current. The sudden change of the intensity near Ix ¼ 0 is directly reflected to the voltage behavior in Fig. 10a since the signals are directly proportional to the bunch charge. Quantities proportional to the beam displacement in horizontal and vertical directions are obtained by combining left/right and up/down signals
5.2. Measurement results
L ¼ S2 þ S3 ; D ¼ S3 þ S4 ;
Measurements at the TTF FEL facility were performed to study the response of all monitors and their electronics as a whole system. Systematic studies were done for bunch charges of 0.5–5 nC at fixed beam position and for different beam displacements at a given beam intensity. Usually, the results presented were averaged over a number of individual bunches (typically 100) in order to avoid charge and position fluctuations during data taking. Data with extreme fluctuations were rejected from the analysis. Tuning of the electronics by proper choices of attenuation respectively gain values for each BPM channel was needed to match the signals to the ADC input range. The four signal voltages Si ði ¼ 1; y; 4Þ of a particular BPM are plotted in Fig. 10a as functions
R ¼ S1 þ S4 and U ¼ S1 þ S2 ;
from which norm differences in x- and y-directions are derived Fx ¼ ðR LÞ=ðR þ LÞ;
Fy ¼ ðU DÞ=ðU þ DÞ:
Another possibility is to combine the diagonal signals like P ¼ ðS4 S2 Þ=ðS2 þ S4 Þ; S ¼ ðS1 S3 Þ=ðS1 þ S3 Þ; to provide the skew differences Dx ¼ S P;
Dy ¼ S þ P:
Both, norm and skew variables have the advantage of being independent of the bunch charge, so that beam intensity fluctuations cancel
ARTICLE IN PRESS 250
U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252 1350 S2
500 VOLTAGE [mV]
SUMVOLTAGES [mVpp]
600 S1
400 S3
300 200 S4
100 -30
-20
-10
(a)
0
10
20
1200 1150 1100
-30
30
-10
1.50
0.60
1.00
SKEW
0.20 0.00
0
10
20
30
10
20
30
Ix [A]
0.80
0.50 0.00 -0.50
-0.20 -0.40 -30
-20
(b)
Ix [A]
0.40 NORM
1250
1050
0
(c)
1300
-20
-10
0
10
20
Ix [A]
-1.00 -30
30
(d)
-20
-10
0 Ix [A]
Fig. 10. Measured signals of a BPM against correction coil current. (a) Signal voltages; (b) Beam intensity; (c) Norm difference Fx and (d) Skew difference Dx :
out. Figs. 10c and d show Fx and Dx ; respectively, as functions of the steering current within 730 A: This corresponds to beam displacements in horizontal direction up to about 71 mm: We observe an approximately linear behavior of both quantities within this range. It can also be noted that the slope of the skew difference Dx is larger than that of the norm difference. This implies a larger change of the induced signal for a given beam displacement (sensitivity) of Dx : The same statement holds for the skew difference Dy : To determine position resolutions for all BPMs, the response of each monitor system was at first studied for e.g. 3:5 nC bunch charge with correction coil settings of the so called FEL mode. These settings were achieved to observe self-amplified spontaneous emission in a free-electron laser [3]. For this mode, the beam is believed to be positioned close to the center of the monitors and these positions define the point x ¼ y ¼ 0 in the following. The raw signals for the ‘center’ position were modified by the constraint of being equal and normalized to the response value of port 1. In this way, differences among the transfer characteristics were adjusted. Then, the beam was displaced in x and y in a controlled manner within
each BPM6. Fig. 11 shows, as an example, the skew differences Dx and Dy for xðyÞ within 71 ð70:5Þ mm for the second monitor in the vacuum chamber. In good approximation, a linear performance for both quantities is observed. In addition, the monitor also reveals a large robustness of Dx when y is modified and vice versa, Dy vs. x; as expected when Dx (Dy ) is measured for beam steering in the y ðxÞ direction. Also, the sensitivity of the signal function in the horizontal direction is greater than in the vertical direction, as observed for both prototypes. This result is expected from the slot position differences within the BPM (see Fig. 1). Correlations of the skew differences might be deduced from Fig. 11 (right), where Dx and Dy are plotted as a function of the x–y displacement plane. Typical 1-standard deviation errors for the skew differences Dx and Dy were found about 0.01. This together with the slope of the skew versus displacement variation (calibration constant) results to the following beam position resolution of 6
Changing the current in preselected correction coils, predictions of relative beam positions within the BPMs are possible.
ARTICLE IN PRESS U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
251
Fig. 11. Skew differences Dx and Dy against relative beam displacements x and y (left) and the skew differences in the x–y plane (right).
the waveguide monitors as installed into TTF FEL: DxC25 mm;
DyC30 mm:
These numbers involve fluctuations of the monitor readings due to noise and non-linearities of the electronics and residual fluctuations of the bunch charges and positions, after averaging over many bunches.
6. Improvements for further measurements Measurements discussed in the previous section clearly underlined the importance of selecting proper calibration values during data taking. If e.g. calibration data for 3:5 nC bunch charge were included in Dx or Dy measurements as functions of the beam intensity between 0.5 and 3:0 nC; variations of Dx and Dy were observed while charge independence would be expected. Also, measurements in a multi-bunch regime (with bunch spacing of 1 ms; 10 bunches within a train and a 1 Hz train repetition) revealed a signal voltage variation with increasing bunch number. Hence, improved electronics with significantly larger dynamic range of the transfer characteristics are highly desirable. In particular, demodulator module improvements are needed by e.g. applying a quadrature demodulation instead of the asynchronous demodulation.
In order to achieve a few micrometer beam position resolution, this new demodulator has to have a noise level about 20 times lower than the I=Q demodulator of our first electronics (see Section 5.1).
7. Summary and conclusions The operation principle and the design of a new waveguide BPM has been described. The waveguide concept relies on four waveguides with small slots arranged around the beam path. In order to fit in the narrow undulator vacuum chamber special ridged waveguides were designed. For each BPM, waveguides and slots were split into two symmetric pairs separated by 40 mm along the beam direction. Each pair was positioned at 735 with respect to the horizontal plane. The coupling factor of a single slot was designed to be about 1% at 12 GHz: At the end of the waveguide, the electromagnetic waves are transferred to a coaxial ultra-high vacuum feedthrough. Two BPM prototypes were manufactured and exhaustively tested on a laboratory test bench and with beam. It was found that the RF-characteristics were in agreement with expectations from theory. Results on transfer characteristics, gain factors and sensitivities obtained from laboratory and beam measurements were in good agreement
ARTICLE IN PRESS 252
U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
with MAFIA simulations and the image charge model. They encouraged us to use Prototype II design for the ten waveguide monitors installed in the third undulator module at the TTF FEL. The results obtained are summarized as follows: *
*
*
*
measurements performed after careful signal adjustments revealed good linearity of quantities proportional to the beam displacement, within a range of about 71 mm, beam position resolutions were determined to DxC25 mm and DyC30 mm; the 12-to-1 GHz down-converter met the requirements of the Noise Figure and sufficiently large dynamic range, a relatively small linear range of the demodulator transfer characteristics was observed.
Possible improvements of the waveguide BPM system are suggested. They rely on application of an I=Q demodulator which together with sufficiently low noise would allow for few micrometer beam position resolution, also for multi-bunch operation, and would resolve the signal variation for bunches in a bunch train. Based on MAFIA simulations and analytical calculations, stringent mechanical tolerances for the waveguide monitors were derived, see Section 3 for some examples. It turned out to be difficult to achieve these requirements in practice. In particular, the repeatability of the design parameters within the specifications required significant effort. In summary, a high-precision beam position monitor based on a new microwave concept has proven its functionality. By improving the signal processing electronics as proposed the new BPM system would be an usable device for micrometer beam position measurements. This monitor type has a large potential for beam position monitoring in cases of extremely short electron bunches (t50 mm) at restricted space conditions in an accelerator.
Acknowledgements The authors would like to express their gratitude to P. Castro, H. Henke, M. Sachwitz, G. Schmidt, R. Steinbrecher, H. Thom, K. Trutzschler . and M.
Wendt for important contributions as well as valuable and encouraging discussions during different stages of the project.
References [1] D.A. Edwards (Ed.), TESLA TEST FACILITY—Design Report, DESY Hamburg, TESLA-Note 95-01, March 1995; A VUV Free Electron Laser at the TESLA Test Facility Linac-Conceptual Design Report, DESY Hamburg, TESLA-FEL 95-03, June 1995. [2] TESLA Technical Design Report, DESY 2001-011, Part II, p.48 ff. [3] J. Andruszkow, et al., Phys. Rev. Lett. 85 (2000) 3825. [4] B. Faatz, J. Pfluger, . Y.M. Nikitina, Nucl. Instr. and Meth. A 393 (1997) 380. [5] The TESLA Test Facility Team, SASE FEL at the TESLA Test Facility, Phase 2, TESLA-FEL 2002-01, June 2002. [6] e.g. R. Lorenz, Cavity Beam Position Monitors, Beam Instrumentation Workshop, SLAC, Stanford, USA, 1998. [7] U. Hahn, P.K. den Hartog, J. Pluger, . M. Ruter, . G. Schmidt, E.M. Trakhtenberg, Nucl. Instr. and Meth. A 445 (2000) 442. [8] R. Klatt et al., MAFIA—a three-dimensional electromagnetic CAD system for magnets, RF structures and transient wake-fields calculations, IEEE Proceedings of the LINAC86 Conference, Geneva, Switzerland, 1986. [9] T. Kamps, Ph.D. Thesis, Humboldt University, Berlin, 2000. [10] M. Dohlus, et al., TESLA-FEL-98-02, 1998. [11] J.H. Cuperus, Nucl. Instr. and Meth. 145 (1977) 219. [12] J.P. Dalahaye, The CLIC study of a multi-TeV eþ e linear collider, Proceedings of the PAC99, New York, USA, 1999; CLIC Team, CTF2 Design Report, CERN/PS-96-29, 1996. [13] R. Brinkmann, et al., Conceptual design of a 500 GeV eþ e linear collider with integrated X-ray laser facility, Vol. II, DESY 1997-048, ECFA 1997-182; M. Schmitz, Status of the S-band test facility, Proceedings of the LC97, Zvenigorod, Russia, 1997. [14] T. Kamps, Proceedings of the FEL99 Conference, Hamburg, Germany, 1999. [15] Y.M. Nikitina, J. Pfluger, . Nucl. Instr. and Meth. A 375 (1996) 3825. [16] U. Hahn, et al., TESLA Report, TESLA-FEL 2003-04, 2003. [17] T. Kamps, W. Riesch, F. Tonisch, Proceedings of the EPAC2000, Vienna, Austria, 2000. [18] J. Andruszkow, P. Jurkiewicz, F. Tonisch, 8-channel fastADC with 14 bit resolution, http://tesla.desy.de/ doocs/hardware/8ch 10MHz ADC manuel.pdf
Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
Waveguide monitors—a new type of beam position monitors for the TTF FEL U. Hahna, T. Kampsb, R. Lorenzc, W. Rieschd, H.J. Schreiberd,*, F. Tonischd a
DESY Hamburg, 22607 Hamburg, Germany b BESSY Berlin, 12489 Berlin, Germany c Westdeutscher Rundfunk, 50600 Koln, . Germany d DESY Zeuthen, Platanallee 6, 15735 Zeuthen, Germany Received 11 August 2003; received in revised form 15 March 2004; accepted 15 March 2004
Abstract The operation of a Free Electron Laser at the TESLA Test Facility requires the electron trajectory to be aligned with precision of few micrometer in overlap with the photon beam. To achieve this goal precise position monitoring of the electron beam in the narrow undulator gap is mandatory. A beam position monitor based on a new microwave concept was designed, built and tested. It consists of four ridged waveguides which couple by small slots to the magnetic field of the electron beam. The paper reports the concept, the design and tests of prototypes in a test bench in the laboratory and with beam. TTF FEL measurement results are also presented and possible improvements are discussed. r 2004 Elsevier B.V. All rights reserved. PACS: 29.27F Keywords: Beam position measurement; Waveguide monitor
1. Introduction In 1995 an international collaboration proposed the construction of the superconducting accelerator of the TESLA Test Facility (TTF) with an integrated Free Electron Laser (FEL) at the Deutsches Elektronensynchrotron DESY in Hamburg [1]. This facility was set-up, commissioned and brought into operation, for a review see e.g. [2]. First lasing of the FEL was observed in 2000 [3]. *Corresponding author. Tel.: +49-33-762-7298. E-mail address: [email protected] (H.J. Schreiber).
The operation of a Free Electron Laser in the Self Amplified Spontaneous Emission mode (SASE) requires the electron trajectories to be aligned with very high precision in overlap with the photon beam. Simulations showed that the mean variation between the photon path (mainly defined by the undulator axis) and the electron trajectory must be kept below 20 mm in order to reach saturation in the undulator [4]. To ensure this overlap, the undulator modules of the TTF FEL were equipped with beam position monitors (BPM) and steering coils. They have to be integrated into the undulator vacuum chamber inside the small undulator gap. For the beam
0168-9002/$ - see front matter r 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.nima.2004.03.178
ARTICLE IN PRESS U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
BEAM BEAM y S2
S1
S3
S4
241
z
S3
S1
S2
S4
x
x Fig. 1. Schematics of a waveguide BPM with four channels and four coupling slots (profile, top view).
position monitors two schemes with different sophisticated designs were followed. Pick-up monitors were installed into the first two undulator modules, while the third module was equipped with a new type of waveguide beam position monitors. An innovative feature of this monitor is the way the coupling to the beam field is realized, in combination with the special shape of the waveguides. Both types of BPMs were designed for position measurements with a precision of 10–20 mm: For phase 2 of the TESLA Test Facility with a design electron beam energy of 1 GeV; the electron bunches are compressed to approximately 50 mm in order to provide a peak current of 2:5 kA required for FEL [5]. To measure the position of very short bunches accompanied by large peak currents, waveguide monitors are favorable due to their simple coupling to the beam via a small slot. Discharges, which might occur in pick-up monitors between the antenna and the vacuum chamber material are strongly suppressed or even excluded for the waveguide monitor concept. Both complementary monitor schemes were developed within the TESLA collaboration and experiences will drive the future solution. Other monitor types e.g. resonant devices [6] are not suitable due to possible intolerable wakefields. The paper summarizes the development of the waveguide BPMs in its various stages up to their application at TTF FEL phase 1 commissioned in 2000. It includes the mechanical design of the monitor essentially predetermined by the small undulator gap, the cross section of the vacuum chamber and the undulator magnets. Prototypes
were constructed and tested in the laboratory and with beam. Finally, the third TTF FEL undulator chamber was equipped with waveguide BPMs and data were taken. The results obtained in each stage of the monitor development will be discussed, as well as problems which were observed and possible solutions. In the past, several notes on selected topics of the waveguide monitor such as construction and electronics aspects or calibration questions were presented. However, neither of them summarized all aspects of this type of monitor in a compact manner. The paper also updates and supersedes results already published.
2. Design and basic characteristics of the waveguide monitor Basic mechanical design parameters of the waveguide monitor are determined by their position within the small undulator gap of 12 mm; with only horizontal access. The BPMs have to be an integrated part of the vacuum chamber, taking also into account the beam pipe with 9:5 mm inner diameter. The shape of the waveguides is designed in a way which reduces the size and realizes at the same time an effective coupling to the beam field. At the end of each waveguide, a coaxial adapter with a vacuum feedthrough is flange mounted. All waveguide structures were fabricated using Electrical Discharge Machining (EDM). The principle design of this new BPM type is sketched in Fig. 1. A fractional part of the electromagnetic beam field is coupled by four
ARTICLE IN PRESS 242
U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
slots1 into ridged waveguides arranged symmetrically around the beam pipe. Due to restricted vertical space, two waveguide pairs (S1–S3 and S2–S4) were separated by approximately 40 mm in beam direction. Each waveguide is inclined by about 35 with respect to horizontal axis to achieve highest possible vertical separation of the coupling slots. Each coupling slot (Fig. 2) of a BPM can be regarded as a radiating magnetic dipole oriented parallel to the magnetic field lines of the passing beam. The T-ridged waveguide enables a coupling to the magnetic field of the beam. The signal amplitude equals 1% of the induced magnetic field amplitude for a centered beam. An adapter is necessary to match the impedance of the waveguide to that of a standard 50 O coaxial cable. This adaptor is realized in two stages resulting in a lowloss transmission line. In the first part a short rectangular coaxial line (with a length of l=4) is connected to the T-ridged waveguide, followed by a round coupling to match the inner conductor of the vacuum feedthrough. Fig. 3 sketches the three transition steps. An improved version of the monitor (Prototype II) was designed as an integral part of the vacuum chamber.2 It is shown in Fig. 4, with the waveguide profile and position of the slots marked. The electron beam passes the flat chamber through the vacuum flanges at both ends of the prototype. This design provides a better broadband transition between waveguide and vacuum feedthrough due to a l=4-transformer (Fig. 3b) produced in one step together with the waveguide ridge structure. The extended pin of the coaxial vacuum feedthrough plugs into a groove in the transformer matched to fit the pin. This technique provides good electrical contact as well as correct positioning and vacuum tight mounting of all elements [7]. 1
Dimensions of these slots are 3:2 3:2 mm2 ; with a depth of 0:3 mm: 2 For prototype I (see also Fig. 3a), the inner part of the waveguide ridge was an integrated part of the vacuum flange inserted into the rectangular waveguide structure. Due to machining tolerances electrical contact between the inner part of the ridge and the vacuum chamber was not always given. The manufacturing process turned out to be complicated and not well suited for a mass production of monitors.
SLOT
WAVEGUIDE
FEEDTHROUGH TRANSFORMER SIGNAL
MAGNETIC FIELD ELECTRIC FIELD BEAM
1/4 VACUUM CHAMBER
Fig. 2. Sketch of the coupling mechanism of the waveguide monitor.
Prototype II also achieved a less coupling amplitude variation than prototype I. Detailed MAFIA [8] simulations were performed in order to obtain optimized parameters for the coupling slot and transition to the vacuum feedthrough. More details on design parameters and their variation obtained from simulations and analytical estimations can be found in Ref. [9]. The upper working frequency of the waveguide is limited by the beam pipe cut-off at 17:9 GHz: For higher frequencies, fields generated by interaction between the beam and its surrounding material can couple into the waveguide and might lead to interferences with beam-induced signals. The waveguide working frequency was designed to be 12 GHz; mainly to use commercially available signal processing components. This frequency is sufficiently above the waveguide fundamental cutoff frequency at 9 GHz and below 13:2 GHz of its next higher order mode. When the electrons pass the undulator beam pipe wakefields are generated. A thereby introduced additional energy spread in the electron bunch might influence SASE. The impact of the waveguide BPMs has been estimated with the result that the overall beam spread is tolerable for TTF FEL phase I operation [9,10].
3. Measurements in the laboratory Laboratory measurements of new beam diagnostics elements are useful to confirm basic predictions of simulations, to study specific char-
ARTICLE IN PRESS U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
(a) PROTOTYPE I
243
(b) PROTOTYPE II & FINAL DESIGN FEED THROUGH
WAVEGUIDE
WAVEGUIDE FEEDTHROUGH
TO BEAM PIPE
TO BEAM PIPE WAVEGUIDE RIDGE
WAVEGUIDE RIDGE
λ/4 ELECTRIC FIELD
T-RIDGE WAVEGUIDE
SHORT PIECE OF RECTANGULAR COAX WAVEGUIDE
EXTENDED INNER CONDUCTOR OF VACUUM FEEDTHROUGH
Fig. 3. Transition steps between the T-ridged waveguide and feedthrough.
0
2
4
6
8
10cm
SLOT
WAVEGUIDE PROFILE 0
2
4
6
8
10cm
Fig. 4. Photo of Prototype II. Top: total view; bottom: view onto sidefront with two waveguide ports and tapped wholes for mounting to the vacuum feedthroughs.
acteristics under idealized conditions and to develop calibration algorithms. A laboratory set-up was built, where the electromagnetic field propagating with an electron
beam has been emulated by a coaxial system of a stretched copper–beryllium wire (250 mm diameter) inside the monitor. An impedance transformer was inserted between the signal generator
ARTICLE IN PRESS 244
U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
and the BPM prototype to minimize reflections of the electromagnetic wave. At the end of the setup the electromagnetic power had to be damped by an absorber to prevent reflections. The BPM itself was moved with respect to the wire by two stepping motors, with a position repeatability of better than 1 mm: Detailed studies were performed to provide good electrical contact, correct impedance matching, best absorption and minimal signal losses [9]. Signals excited in the waveguides were filtered at 12 GHz and amplified. A LABVIEW application running on a PC was used to control the power meter read out as well as movements of both stepping motors. The schematic layout of the electrical facet is shown in Fig. 5. Narrowband measurements were carried out to check the linearity of the monitor and to measure the transfer characteristics of the four channels. A signal generator delivered a 12 GHz cw-signal, at a power level of 14 dBm: Cross scans around the monitor center were performed on both the x- and y-axis, with a typical range of 1–2 mm in steps of X10 mm: The normalized signal of an ideal BPM channel can be calculated using the image charge model [11]. This model describes the locally induced wall current density in a conducting vacuum chamber surface for a pencil beam. This wall current density is equivalent to the local magnetic field, to which the BPM couples with the slots. The magnetic field induced by the electrons can be computed using Biot–Savart’s law and results for a point-like
charge to b2 r2 ; b2 þ r2 2br cosðfi yÞ
Si ¼
for slot i at radius b and angle fi and a beam at position ðr; y). The assumptions made in the model are to a good approximation valid in our case since 1 the slots can be considered as point-like (only 10 of the beam pipe circumference) and beam movements only occurred near the center of the monitor. For large off-sets (several mm) one has to integrate over the complete slot size. Due to individual transfer characteristics of each channel the measured voltage of a channel depends on the response function Si modified with the scalar channel gain gi : Vi ¼ Agi Si : Here, A is a factor describing the amplitude of an input signal. The gain factors are introduced to adjust imbalances among individual transfer characteristics of the channels i ¼ 1–4. Gain parameters were estimated by measuring induced voltages Vi;j ði ¼ 1; y; 4Þ for j ¼ 1; y; n wire positions in the BPM aperture. A fit to the image charge model Mi;j provides the coupling factor for each channel Mi;j ¼ gi
b2 r2j b2 þ r2j 2brj cosðfi yj Þ
i ¼ 1; y; 4 and a ¼ ðg1 ; g2 ; g3 ; g4 Þ;
where a is the array of fitting parameters to be determined.
WIRE
RF IN
RF CABLE
PC WITH GPIB INTERFACE AND LABVIEW
FILTER & AMPLIFIER STEP MOTOR CONTROL
Mði; j; aÞ;
j ¼ 1; y; m
RF OUT
SIGNALGENERATOR
ð1Þ
RELAY
POWERMETER
CONTROL
READ OUT CONTROL & READ OUT
Fig. 5. Schematics of the electrical facet of the laboratory test bench.
ARTICLE IN PRESS U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
The waveguide signals were filtered, amplified and measured. In order to exclude effects from imbalances among amplifiers, cables and filters, detected voltages were divided by their individual amplification factors. At each position, four voltages were recorded such that for sufficient position measurements the number of data points exceeded significantly the number of free parameters. The gain results normalized to g1 ¼ 1 are listed in Table 1. Their variation of few times 102 is to a great extend ascribed to manufacturing tolerances. Simulation studies provided demands for mechanical tolerances [9]. For example, to keep the root–mean–square spread of the coupling amplitudes below a level of 2%, the slot width and the height have to be manufactured within 750 mm and the depth of the slot must be within 7100 mm: Since it is difficult to achieve such tolerances, calibration for each waveguide monitor is needed. The image charge model was also used to estimate the angle of each coupling slot by using the gain factors obtained in the previous step. Results of this fit are in fair agreement with the Table 1 Gain values normalized to g1 ¼ 1 and coupling slot angles as derived from a fit to the image charge model Channel
Gain
Slot angle (deg)
1 2 3 4
1:000070:0018 0:905470:0017 1:066870:0017 1:009470:0019
35:9470:13 33:4670:09 35:9270:14 34:2070:13
design value of 35:51 (with respect to x-axis), as shown in Table 1. Slight systematic differences can be accounted for by a small tilt between the BPM reference frame and the test bench system. We noticed that the gain parameters of Table 1 are unchanged within uncertainties when the fitted fi values instead of the nominal slot angles are used in the fitting procedure. The sensitivity describes the change of an induced signal amplitude for a given beam offset. Induced voltages of channels 1–4, normalized with corresponding amplifications and damping factors, were used to construct virtual signal functions L ¼ S2 þ S3 and R ¼ S1 þ S4 for the horizontal (U ¼ S1 þ S2 and D ¼ S3 þ S4 for the vertical) direction, see also Fig. 1. The sensitivity Sx (Sy ) for the horizontal (vertical) direction is then defined as the slope between set position in x (or y) and the logarithmic ratio of L and R (or U and D) L U xSx ¼ 20 log and y Sy ¼ 20 log : ð2Þ R D Values for log-ratios as function of the wire position are plotted in Fig. 6. A linear performance for jxj; jyjp500 mm around the BPM center can be observed. The resulting sensitivities are listed in Table 2, with results from MAFIA SParameter computations and from the image charge model. Good agreement between test bench measurements and expected values is noted. In addition, broadband measurements up to 16 GHz were performed to check matching stages,
PROTOTYPE I
PROTOTYPE II 3
1 0 -1
vertical horizontal
-2 -3 -600 -400 -200 0 200 400 POSITION [micron]
600
LOG AMPLITUDE RATIO [dB]
LOG AMPLITUDE RATIO [dB]
3 2
245
2 1 0 vertical -1
horizontal
-2 -3 -600 -400 -200 0 200 400 POSITION [micron]
Fig. 6. Measured sensitivities for both prototypes (test bench).
600
ARTICLE IN PRESS 246
U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
to observe cut-off frequencies and to look for trapped resonances in a waveguide. The transfer characteristics for each BPM channel were studied by measuring the transmission parameter S21 using a network analyzer. The transmission was rather small until the waveguide cut-off frequency of about 9 GHz was reached. Above this frequency high-pass characteristic was observed for all channels with acceptable differences caused by imperfect mechanical manufacturing.
4. Prototype measurements with beam Measurements under real beam conditions were considered as an experimental proof of the operation principle of the waveguide monitor. In addition, basic signal processing electronics using commercially available components were implemented and tested successfully. First beam test data were obtained with Prototype I at the CLIC Test Facility at CERN [12]. Results on strength measurements of the four coupling amplitudes and on sensitivities in x and y directions were consistent with test bench data and MAFIA simulations [9]. Prototype II was installed in the S-Band Test Facility at DESY (SBTF) [13]. Typical beam parameters relevant for single bunch measurements are summarized in Table 3. Table 2 Sensitivities as derived from Prototype II measurements on the wire test bench, MAFIA simulations and the image charge model
Sx (dB/mm) Sy (dB/mm)
Measurements
Simulation
Model
5:7970:04 4:1070:04
5.78 3.88
5.78 4.10
Table 3 S-Band Test Facility beam parameters during measurements Energy Repetition rate Bunch charge Bunch length Bunch size
E fr q sz sx;y
46:7 MeV 1 Hz 0:2 nC 4–5 mm B1 mm
The electron beam with desired time structure and energy was generated and accelerated by three sections. Prototype II has been installed in a drift region between two quadrupole triplets where the beam envelope should be equal for horizontal and vertical planes. A two-dimensional stepping motor frame with the BPM integrated was computer controlled to move the monitor in well-defined steps perpendicular to the beam. After centering the beam with quadrupole scans, first measurements were performed by displaying beam induced signals on a digital sampling scope. For the scope trigger, a beam related signal was taken from a stripline BPM downstream. With this set-up, 12 GHz components of beam related induced signals were clearly visible. In a next step, downconverted monitor signals for all four channels were recorded. The 12 GHz LO signal needed for down-conversion was derived using the 4th harmonic of the accelerating frequency (2:99 GHz) as induced in the stripline BPM. This signal had a constant amplitude and was phase locked to the beam. After filtering, the remaining IF signal was amplified. Signals were recorded for several waveguide monitor positions transverse to the beam direction.3 After accounting for different amplitude factors of the BPM channels—attenuation in nearly all components as well as electronics gain—the signal function Sx as defined in Eq. (2) was derived from a linear fit to the data. In Fig. 7, Sx is plotted versus transverse beam position for a rough scan (step width Dx ¼ 500 mm) and a fine scan (Dx ¼ 50 mm). The results can be reasonably interpreted using the image charge model (continuous curves). The horizontal sensitivity derived from the fine scan results to Sx ¼ ð5:2770:14Þ dB=mm which is somewhat lower than the wire test bench value of ð5:7970:04Þ dB=mm: This can be explained by the fact that the beam size is larger than the linear region of the BPM. We have also compared the beam position in x; determined by using an algorithm developed for beam-based calibration [14], with the set beam positions by the stepping 3 Steering along the vertical axis was not possible due to a system failure of one stepping monitor.
ARTICLE IN PRESS U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252 FINE SCAN
ROUGH SCAN -0.10
0.6 DATA IMAGE CHARGE MODEL
0.2 0 -0.2
-0.15 NORMALIZED SIGNALFUNCTION
NORMALIZED SIGNALFUNCTION
0.4
DATA LINEAR MODEL
-0.20 -0.25 -0.30 -0.35
-0.4
-0.40
-0.6
-200
-2.0 -1.5 -1.0 -0.5 0 0.5 1.0 1.5 2.0 2.5
(a)
247
POSITION x [mm]
(b)
0 200 400 600 POSITION x [micron]
800
Fig. 7. Relative signal function Sx for the horizontal plane: (left) for a broad scan width of Dx ¼ 500 mm; (right) for a narrow scan with Dx ¼ 50 mm: Three values around 400 mm were excluded from the fit because of strong beam position jitter.
motor. Linearity for off-sets up to about 72 mm has been observed with high precision [9]. The measurements at the S-band Test Facility were limited by the granularity of the internal ADC of the oscilloscope. For operation at the TTF FEL an ADC was developed for optimum pulse response with significant better resolution (see Section 5.1).
FEL operation was identical to the Prototype II design. Special mounting, alignment and support systems for the flat and long vacuum chamber allowed for a proper insertion of the chamber in the undulator gap and its positioning within a tenth of a millimeter in vertical and horizontal direction. 5.1. Signal processing electronics
5. TTF FEL measurements For phase 1 of the TTF FEL the undulator [15] was segmented into three sections. The vacuum chamber of the third section was equipped with ten microwave monitors as an integrated part of the chamber. The vacuum chamber is a flat 4:5 m long structure with the cross-section of 11:5 128 mm2 of extruded aluminium, with a central aperture of 9:5 mm to guide the electron beam. In order to guarantee a tough control of the beam orbit inside the undulator one BPM and one correction coil were required per FODO quadrupole of the undulator magnet structure. The correction coils were located in between the BPMs and allowed for horizontal and vertical steering of the beam. The waveguide holes with their relative complex contours and the small coupling slots at the beam aperture were fabricated by EDM. At the end of each waveguide a coaxial adapter with a special UHV compatible RF feedthrough was flangemounted. The design of the monitors for TTF
The signal processing electronics have been divided into a 12-to-1 GHz down-converter followed by a 1 GHz amplitude demodulator and an analog-to-digital converter (ADC). In the following we describe the electronics in general, for more details we refer to Ref. [16]. The four signals of a monitor were grouped together into one down-converter circuit box, positioned close to the vacuum chamber to minimize signal attenuation. This down-converter was designed for a total gain of about 5 dB with a Noise figure of 20 dB: After passing semi-rigid cables of 1 to 2 m length, signals from each BPM were processed within the down-converter in a parallel way by four identical hardware chains. This is shown in Fig. 8 for one channel in a simplified manner. The broadband monitor signal was filtered at 12 GHz by a 5-pole interdigital bandpass with a bandwidth of 140 MHz: For a bunch charge of 1 nC the signal amplitude after the bandpass was estimated to be 35 mV at 50 O
ARTICLE IN PRESS U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
248
DOWN-CONVERTER (12 GHz /1 GHz)
12GHz Bandpass
Limiter ACLM 4571F
Attenuator MC 2710
Amplifier NE 32484
Mixer 1GHz MC 2710 Bandpass f0 = 1 GHz B = 64 MHz -10 dB -4.5 dBm
ATT: DAT=150
+11 dB NF= 0.6 dB
NF=11.1 dB
-9.5 dBm
-8 dBm
-5.5 dBm
-5 dBm
REF: +8 dBm, 12.3 GHz
-6 dB
-1.5 dB
-15.5 dBm
-2.5 dB
-4.5 dB
LO : +8 dBm, 11.29 GHz
+19 dB NF = 3.8 dB
0 dBm
f0 = 12.29 GHz B=140 MHz
-0.5 dB
Linedriver INA-10386
-19 dBm
20dBDir.Coupler
G = 5 dB NF= 18.6 dB
Fig. 8. Block scheme for the 12-to-1 GHz down-converter electronics.
impedance. A 20 dB directional coupler before this filter allowed to introduce an 8 dBm 12:3 GHz reference signal for calibration purposes. The generator stabilized by a dielectric resonator was integrated in the down-converter box, too. After the bandpass filter a limiter protected the subsequent components from signal levels above a certain threshold, which might be caused by e.g. large mis-steering of the beam. The limiter is followed by an attenuator and an amplifier to avoid saturation in the down-converter circuit as well as to ensure an acceptable signal-to-noise ratio. 12-to-1 GHz down-conversion was realized within a combination of a mixer and a 1 GHz bandpass filter. The 8 dBm local oscillator signal at 11:3 GHz needed for the frequency downconversion was generated by a circuit identical to the generator mentioned above. After output signal amplification the signal was passed to the demodulator in the racks outside the tunnel via a 25 m long 50 O coaxial cable. Down-converted 1 GHz signals were processed by amplitude demodulation. The simplest asynchronous (diode) demodulation was realized, with the disadvantage of a relatively small linear range of the demodulator transfer function.4 A block diagram of the demodulator for one BPM channel is shown in Fig. 9. The first two circuits represent a
variable amplifier with a range of 1379 dB; so that an optimal signal level for the ADC input range can be adapted. The signal was then detected and filtered by a 25 MHz bandpass. At the end, the demodulated pulse was amplified again and passed to the ADC. The four channels of a single BPM were grouped together and installed in a VME–RF cassette. The final signal with a width of about 25 ns has to be as jitter-free as possible and bunch-synchronized transferred to the ADC. With a noise rms of about 6 mV; the beam displacement resolution caused by the signal processing electronics only was estimated to be about 12 mm: The noise from the demodulator alone (no input signal) was measured to about 300 mV rms and, hence, negligible compared to the overall noise level. Three signal attenuation possibilities for dynamic range optimization in the electronics were realized to adjust the small dynamic range of the demodulator: (i) at the entrance of the 12-to1 GHz down-converter (fixed, any value), (ii) before the 12 GHz amplifier (remote via I 2 C; max. attenuation 20 dB) and (iii) at the entrance of the demodulator (remote via I 2 C; max. attenuation 20 dB). Careful signal attenuation control is therefore required during data taking in order to obtain undisturbed pulses.5 The reference signal
4 In a first attempt we applied an I=Q demodulation using a commercial circuit. However, the obtained signals were very noisy resulting to unacceptable position resolutions of about 100 mm or more [17].
5 In principle, an extension of the dynamic range of the electronics is possible by means of software correction procedures which have to rely on the transfer characteristics of each BPM channel.
ARTICLE IN PRESS U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
249
DEMODULATOR Driver MAX 4213
BP -8.2 dB
+4.6 dBm (PEP)
-7.4 dB +12 dBm (PEP)
ATT: DAT = 0 /255
-12 dBm (PEP)
71 m Vpp / 50 Ohm
SNR :70 dB
G = +24 dB, NF = 3.8 dB -2 / -20 dB
Bandpass 25 MHz
+5.6 dB
G = +12 dB
+2 dBm (PEP) SNR : 40 dB
Detector BAT14/03W
0.28 Vpp / 50 Ohm
Amplifier INA-10386
-3.6 dBm (PEP)
Attenuator HMC121G8
Fig. 9. Block scheme of the demodulator electronics.
for calibration purposes was used to remove the electronics offset in advance of data taking. Signals from the analog electronics were fed into an 8-channel high speed ADC with a resolution of 14 bit: It converted the signals at a rate of 1 MSample=s and stored the data in an on-board memory (depth 128 kSample=channel; [18]). The ADC was triggered and clocked by the TTF timing system. In order to sample the signal at the peak the clock was adjusted using a digital delay integrated on the ADC board, with 5 ns step size. The input range of the ADC was set to 71 V with an input impedance of 50 O: The ADC values were read out by the standard TTF data acquisition system.
of the horizontal beam displacement achieved by different correction coil currents Ix ; for a bunch charge of about 1 nC: For waveguide channel notations we refer to Fig. 1. Voltages S1 and S4 decrease with increasing Ix (negative off-sets in x), whereas S2 and S3 grow, as expected. Fig. 10b shows the beam intensity defined as the sum of all four voltages against corrector current. The sudden change of the intensity near Ix ¼ 0 is directly reflected to the voltage behavior in Fig. 10a since the signals are directly proportional to the bunch charge. Quantities proportional to the beam displacement in horizontal and vertical directions are obtained by combining left/right and up/down signals
5.2. Measurement results
L ¼ S2 þ S3 ; D ¼ S3 þ S4 ;
Measurements at the TTF FEL facility were performed to study the response of all monitors and their electronics as a whole system. Systematic studies were done for bunch charges of 0.5–5 nC at fixed beam position and for different beam displacements at a given beam intensity. Usually, the results presented were averaged over a number of individual bunches (typically 100) in order to avoid charge and position fluctuations during data taking. Data with extreme fluctuations were rejected from the analysis. Tuning of the electronics by proper choices of attenuation respectively gain values for each BPM channel was needed to match the signals to the ADC input range. The four signal voltages Si ði ¼ 1; y; 4Þ of a particular BPM are plotted in Fig. 10a as functions
R ¼ S1 þ S4 and U ¼ S1 þ S2 ;
from which norm differences in x- and y-directions are derived Fx ¼ ðR LÞ=ðR þ LÞ;
Fy ¼ ðU DÞ=ðU þ DÞ:
Another possibility is to combine the diagonal signals like P ¼ ðS4 S2 Þ=ðS2 þ S4 Þ; S ¼ ðS1 S3 Þ=ðS1 þ S3 Þ; to provide the skew differences Dx ¼ S P;
Dy ¼ S þ P:
Both, norm and skew variables have the advantage of being independent of the bunch charge, so that beam intensity fluctuations cancel
ARTICLE IN PRESS 250
U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252 1350 S2
500 VOLTAGE [mV]
SUMVOLTAGES [mVpp]
600 S1
400 S3
300 200 S4
100 -30
-20
-10
(a)
0
10
20
1200 1150 1100
-30
30
-10
1.50
0.60
1.00
SKEW
0.20 0.00
0
10
20
30
10
20
30
Ix [A]
0.80
0.50 0.00 -0.50
-0.20 -0.40 -30
-20
(b)
Ix [A]
0.40 NORM
1250
1050
0
(c)
1300
-20
-10
0
10
20
Ix [A]
-1.00 -30
30
(d)
-20
-10
0 Ix [A]
Fig. 10. Measured signals of a BPM against correction coil current. (a) Signal voltages; (b) Beam intensity; (c) Norm difference Fx and (d) Skew difference Dx :
out. Figs. 10c and d show Fx and Dx ; respectively, as functions of the steering current within 730 A: This corresponds to beam displacements in horizontal direction up to about 71 mm: We observe an approximately linear behavior of both quantities within this range. It can also be noted that the slope of the skew difference Dx is larger than that of the norm difference. This implies a larger change of the induced signal for a given beam displacement (sensitivity) of Dx : The same statement holds for the skew difference Dy : To determine position resolutions for all BPMs, the response of each monitor system was at first studied for e.g. 3:5 nC bunch charge with correction coil settings of the so called FEL mode. These settings were achieved to observe self-amplified spontaneous emission in a free-electron laser [3]. For this mode, the beam is believed to be positioned close to the center of the monitors and these positions define the point x ¼ y ¼ 0 in the following. The raw signals for the ‘center’ position were modified by the constraint of being equal and normalized to the response value of port 1. In this way, differences among the transfer characteristics were adjusted. Then, the beam was displaced in x and y in a controlled manner within
each BPM6. Fig. 11 shows, as an example, the skew differences Dx and Dy for xðyÞ within 71 ð70:5Þ mm for the second monitor in the vacuum chamber. In good approximation, a linear performance for both quantities is observed. In addition, the monitor also reveals a large robustness of Dx when y is modified and vice versa, Dy vs. x; as expected when Dx (Dy ) is measured for beam steering in the y ðxÞ direction. Also, the sensitivity of the signal function in the horizontal direction is greater than in the vertical direction, as observed for both prototypes. This result is expected from the slot position differences within the BPM (see Fig. 1). Correlations of the skew differences might be deduced from Fig. 11 (right), where Dx and Dy are plotted as a function of the x–y displacement plane. Typical 1-standard deviation errors for the skew differences Dx and Dy were found about 0.01. This together with the slope of the skew versus displacement variation (calibration constant) results to the following beam position resolution of 6
Changing the current in preselected correction coils, predictions of relative beam positions within the BPMs are possible.
ARTICLE IN PRESS U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
251
Fig. 11. Skew differences Dx and Dy against relative beam displacements x and y (left) and the skew differences in the x–y plane (right).
the waveguide monitors as installed into TTF FEL: DxC25 mm;
DyC30 mm:
These numbers involve fluctuations of the monitor readings due to noise and non-linearities of the electronics and residual fluctuations of the bunch charges and positions, after averaging over many bunches.
6. Improvements for further measurements Measurements discussed in the previous section clearly underlined the importance of selecting proper calibration values during data taking. If e.g. calibration data for 3:5 nC bunch charge were included in Dx or Dy measurements as functions of the beam intensity between 0.5 and 3:0 nC; variations of Dx and Dy were observed while charge independence would be expected. Also, measurements in a multi-bunch regime (with bunch spacing of 1 ms; 10 bunches within a train and a 1 Hz train repetition) revealed a signal voltage variation with increasing bunch number. Hence, improved electronics with significantly larger dynamic range of the transfer characteristics are highly desirable. In particular, demodulator module improvements are needed by e.g. applying a quadrature demodulation instead of the asynchronous demodulation.
In order to achieve a few micrometer beam position resolution, this new demodulator has to have a noise level about 20 times lower than the I=Q demodulator of our first electronics (see Section 5.1).
7. Summary and conclusions The operation principle and the design of a new waveguide BPM has been described. The waveguide concept relies on four waveguides with small slots arranged around the beam path. In order to fit in the narrow undulator vacuum chamber special ridged waveguides were designed. For each BPM, waveguides and slots were split into two symmetric pairs separated by 40 mm along the beam direction. Each pair was positioned at 735 with respect to the horizontal plane. The coupling factor of a single slot was designed to be about 1% at 12 GHz: At the end of the waveguide, the electromagnetic waves are transferred to a coaxial ultra-high vacuum feedthrough. Two BPM prototypes were manufactured and exhaustively tested on a laboratory test bench and with beam. It was found that the RF-characteristics were in agreement with expectations from theory. Results on transfer characteristics, gain factors and sensitivities obtained from laboratory and beam measurements were in good agreement
ARTICLE IN PRESS 252
U. Hahn et al. / Nuclear Instruments and Methods in Physics Research A 527 (2004) 240–252
with MAFIA simulations and the image charge model. They encouraged us to use Prototype II design for the ten waveguide monitors installed in the third undulator module at the TTF FEL. The results obtained are summarized as follows: *
*
*
*
measurements performed after careful signal adjustments revealed good linearity of quantities proportional to the beam displacement, within a range of about 71 mm, beam position resolutions were determined to DxC25 mm and DyC30 mm; the 12-to-1 GHz down-converter met the requirements of the Noise Figure and sufficiently large dynamic range, a relatively small linear range of the demodulator transfer characteristics was observed.
Possible improvements of the waveguide BPM system are suggested. They rely on application of an I=Q demodulator which together with sufficiently low noise would allow for few micrometer beam position resolution, also for multi-bunch operation, and would resolve the signal variation for bunches in a bunch train. Based on MAFIA simulations and analytical calculations, stringent mechanical tolerances for the waveguide monitors were derived, see Section 3 for some examples. It turned out to be difficult to achieve these requirements in practice. In particular, the repeatability of the design parameters within the specifications required significant effort. In summary, a high-precision beam position monitor based on a new microwave concept has proven its functionality. By improving the signal processing electronics as proposed the new BPM system would be an usable device for micrometer beam position measurements. This monitor type has a large potential for beam position monitoring in cases of extremely short electron bunches (t50 mm) at restricted space conditions in an accelerator.
Acknowledgements The authors would like to express their gratitude to P. Castro, H. Henke, M. Sachwitz, G. Schmidt, R. Steinbrecher, H. Thom, K. Trutzschler . and M.
Wendt for important contributions as well as valuable and encouraging discussions during different stages of the project.
References [1] D.A. Edwards (Ed.), TESLA TEST FACILITY—Design Report, DESY Hamburg, TESLA-Note 95-01, March 1995; A VUV Free Electron Laser at the TESLA Test Facility Linac-Conceptual Design Report, DESY Hamburg, TESLA-FEL 95-03, June 1995. [2] TESLA Technical Design Report, DESY 2001-011, Part II, p.48 ff. [3] J. Andruszkow, et al., Phys. Rev. Lett. 85 (2000) 3825. [4] B. Faatz, J. Pfluger, . Y.M. Nikitina, Nucl. Instr. and Meth. A 393 (1997) 380. [5] The TESLA Test Facility Team, SASE FEL at the TESLA Test Facility, Phase 2, TESLA-FEL 2002-01, June 2002. [6] e.g. R. Lorenz, Cavity Beam Position Monitors, Beam Instrumentation Workshop, SLAC, Stanford, USA, 1998. [7] U. Hahn, P.K. den Hartog, J. Pluger, . M. Ruter, . G. Schmidt, E.M. Trakhtenberg, Nucl. Instr. and Meth. A 445 (2000) 442. [8] R. Klatt et al., MAFIA—a three-dimensional electromagnetic CAD system for magnets, RF structures and transient wake-fields calculations, IEEE Proceedings of the LINAC86 Conference, Geneva, Switzerland, 1986. [9] T. Kamps, Ph.D. Thesis, Humboldt University, Berlin, 2000. [10] M. Dohlus, et al., TESLA-FEL-98-02, 1998. [11] J.H. Cuperus, Nucl. Instr. and Meth. 145 (1977) 219. [12] J.P. Dalahaye, The CLIC study of a multi-TeV eþ e linear collider, Proceedings of the PAC99, New York, USA, 1999; CLIC Team, CTF2 Design Report, CERN/PS-96-29, 1996. [13] R. Brinkmann, et al., Conceptual design of a 500 GeV eþ e linear collider with integrated X-ray laser facility, Vol. II, DESY 1997-048, ECFA 1997-182; M. Schmitz, Status of the S-band test facility, Proceedings of the LC97, Zvenigorod, Russia, 1997. [14] T. Kamps, Proceedings of the FEL99 Conference, Hamburg, Germany, 1999. [15] Y.M. Nikitina, J. Pfluger, . Nucl. Instr. and Meth. A 375 (1996) 3825. [16] U. Hahn, et al., TESLA Report, TESLA-FEL 2003-04, 2003. [17] T. Kamps, W. Riesch, F. Tonisch, Proceedings of the EPAC2000, Vienna, Austria, 2000. [18] J. Andruszkow, P. Jurkiewicz, F. Tonisch, 8-channel fastADC with 14 bit resolution, http://tesla.desy.de/ doocs/hardware/8ch 10MHz ADC manuel.pdf