SYNCHRONIZATION SYSTEMS FOR ERLS Stefan Simrock, Frank Ludwig, Holger Schlarb DESY Notkestr. 85, 22603 Hamburg News, Germany Corresponding author: Stefan Simrock DESY Notkestr. 85 22603 Hamburg, Germany Phone: +49 40 8998 4556 FAX: +49 40 8998 4303 e mail: stefan.simrock@desy.de ABSTRACT The next generation light sources requires synchronization of (soft and hard) x rays to beamline and end station lasers. The relative timing jitter required will be of the order of the photon pulse length down to a few fs between sources separated by ~100 to 500m. Also the synchronization of electron beam production at the photo injector, the beam acceleration and compression in the linac, the electron arrival time at the insertion devices and some of the beam diagnostics in the linac must be synchronized with a precision of the same order of magnitude. The concept of synchronization is described and limiting factors for synchronization in Energy Recovery Linacs are discussed.
PACS: 07.50.Hp, 29.17.+w, 41.60.Cr, 42.55.Vc, 42.55.Wd, 43.50.+y Keywords: Synchronization, Timing, Recirculating Linacs 1. INTRODUCTION The use of ultrashort light pulses to study coherent interactions in atomic and molecular systems has advanced rapidly in recent years. The development has driven the development of x ray sources based on accelerator technology such as Free Electron Linacs (FELs) and Energy Recovery Linacs (ERLs). The production of short photon pulses implies the generation of short electron bunches with a precisely defined arrival time in the case of pump probe experiments. In order to produce the electron bunch with the required properties a precise synchronization of various timing critical subsystems is required. To clarify the meaning of synchronization serveral definitions found in a thesaurus can be used: 1. Coordinating by causing the same time. 2. An adjustment of that causes something to occur or recur in unison. 3. The relation that exists when things occur at the same time. If there are many systems to be synchronized it becomes important to determine a common reference to which the systems are synchronized. In the case of accelerators an ultrastable rf master oscillator which is common reference to many subsystems should be used to avoid the difficulty in defining the synchronization with respect to a moving target as it may be respresented by the electron beam which is subject to arrival time jitter in various part of the accelerator. 2. REQUIREMENTS FOR SYNCHRONIZATION 2
The requirements for synchronization of subsystems of X FELs are reasonably well understood and technical performance and operational experience have been documented at various facilities [1,2]. They show significant commonality with the synchronization requirements for ERLs and are therefore discussed first. A generic layout of an FEL with the timing critical subsystems is shown in Figure 1. The subsystems to be synchronized are Photocathode laser RF Gun RF acceleration system before the bunch compressor (rf amplitude and phase stability). Harmonic cavity used in bunch compression scheme (rf amplitude and phase stability). Linac rf system (rf amplitude and phase stability). RF master oscillator with several frequency outputs. Frequency distribution system with multiple reference locations and multiple frequencies. Laser and rf references for diagnostics systems. Laser for pump probe experiments. The requirements are derived from the beam parameters which are: Energy stability and energy spread Emittance Bunch length Bunch arrival time The subsystem requirements are specifications for timing stability specified in terms of residual time jitter of the individual subsystems usually with respect to a common master oscillator. In some cases specifications of stability between certain subsystems are needed in addition. One must also distinguish between correlated and uncorrelated errors since low frequency correlated 3
errors between subsystem are not critical. For the rf systems amplitude and phase stability must be specified due to conversion of beam energy changes in arrival time changes as a result of longitudinal dispersion in the beam lines. In the example of the FEL in Figure 1, typical requirements for the electron bunch length at the undulator are 100 fs, a arrival time jitter of the order of the bunch length (also 100 fs), and an intra bunch and bunch to bunch energy spread of the order of 10 4. As one can imagine, the timing requiremements for most other subsystems are also of the order of 100 fs (corresponding to a rf phase stability 0.05 deg. (@ 1.3 GHz). The requirements for the photocathode laser are somewhat relaxed since timing jitter will be compressed in the bunch compressor since the energy chirp is introduced by the accelerator section. The requirements for the rf acceleration system before the bunch compressor are quite stringent due to the large R56 in the bunch compressor which converts momentum errors induced by rf amplitude and phase errors into bunch length and arrival time errors. 3. CHALLENGES FOR TIMING AND RF CONTROL IN ERLs A typical layout of an energy recovery linac is shown in Figure 2. The injector is similar to that of a linac based FEL but the average beam current and duty cycle are much higher. The concept of energy recovery requires that the beam accelerated in the main linac returns to the linac with a phase shift of 180 degrees [3]. With close to 100% beam transmission the rf power required for the linac can be quite small. Only power reserve for the control of microphonics, residual beam losses, photon production, power losses due to wakefields and pathlength errors for the returning beam has to be provided. Additional challenges in the ERL based scheme are: Possibility of beam break up with thresholds in the ma range. 4
Timing jitter at the undulator from magnet power supply ripple in the first linac arc due to bunch compression in the arc. Beam disruption in the insertion devices and subsequent beam scraping in the second arc. The resulting beam current fluctuation result in heavy beam loading fluctuation in the linac cavities and may require significant rf power for control. Time jitter of returning beam resulting in phase errors and subsequent beam loading variations. This also will lead to beam loading fluctuations which must be controlled by the low level rf system. The threshold for beam breakup can be increased with bunch to bunch beam feedback or by proper settings of the beam optics. Amplitude and phase fluctuations of the recovered beam must be held small to ensure that sufficient rf power and feedback gain in the low level rf system are available to control the cavity fields. 4. SOURCES FOR TIMING JITTER The typical sources for timing error, bunch length variations and energy spread are: Laser timing jitter (reduced in bunch compressor). RF stability (rf gun, harmonic cavity, injector rf, linac rf). Stability of magnets (bunch compression, phase for energy recovery). The main sources of timing jitter are the phase noise of the master oscillator and the frequency distribution system, phase drifts (usually of thermal nature) between outputs of the frequency distribution system at various locations in the accelerator, noise induced by electromagnetic emissions of high power equipment in sensitive electronics (EMI) and the noise of low level electronics in the associated subsystems. The requirements for the phase noise of the master 5
oscillators are usually specified in the frequency range from 1 Hz to several MHz. Phase noise close to the carrier increase rapidly due to the 1/f characteristics. Typical single sideband (SSB) phase noise of a low noise oscillator at 1.3 GHz is of the order of 130 dbc at 10 khz from carrier improving to 150 dbc for frequencies > 100 khz from carrier. This results in an integrated timing jitter of less than 10 fs for a frequencies offset > 10 khz. For frequencies closer to the carrier (1 Hz 10 khz) the timing jitter becomes larger and can reach several 100 fs. Fortunately this jitter will be the same for all subsystems which must be phase locked to the master oscillator such that the relative jitter between subsystem remain to be very small. Differences in loop bandwidth of the phase locked loop may however result in timing jitter between systems. Nowadays crystal oscillator, dielectric resonators and optical fiber laser oscillators with very low phase noise are available to support timing stability of accelerator subsystem to better than 100 fs and even approaching 10 fs. Also frequency distribution systems which in the past have been realized with thermally compensated and thermally stabilized coaxial cables and are supplemented or replaced by fiber optic distribution systems which can achieve a similar or better long term stability of the order of 100 fs over distances of several kilometers. 5. DETECTION OF TIMING JITTER Timing jitter of the various subsystems in the accelerator is usually measured with respect to the master oscillator but can be measured also between subsystems. The types of subsystems are Optical systems such as lasers for photocathode rf guns, diagnostics, seed laser and pump probe lasers, as well as fiber optics laser for frequency reference. RF systems such as master oscillators, frequency reference outputs of frequency distribution systems, rf gun, harmonic cavity and superconducting cavities in injector or linac. 6
Beam pick up from button, strip line or cavity position or current monitors providing timing signals from the beam. X ray photon beam produced by insertion devices. Various methods exist to measure timing jitter between rf signals, optical signals, and between optical and rf signals. 5.1 RF BASED TIMING JITTER MEASUREMENTS Timing jitter measurements between rf systems are usually performed as phase measurements using double balanced mixers or active multiplier circuits. Different frequencies can be compared if they are converted to the same frequency first. If they are harmonically related this can be achieved by use of frequency dividers or multipliers. If they are not harmonically related on must synthesize appropriate reference frequencies for up or down conversion. In general it is favorable to perform the phase measurement at high frequencies since this will improve the signal to noise ratio of the measurement. With typical mixer sensitivities of 10 mv/deg. of rf phase one can achieve a sensitivity of around 3 uv/ fs at 1 GHz. With a thermal noise floor of 174 dbm/hz and assuming a measurement bandwidth of 1 MHz this has be compared to a noise level of 114 dbm which is equivalent to 0.5 uvrms which would allow a resolution of better than 1 fs. Usually the phase noise of the reference will be dominant (typ. 140 dbc at a frequency offset > 10 khz for a reference at 1 GHz) and increase the noise floor to around 20 uv allowing for a resolution of 10 fs at a signal to noise ratio of 1. The measurement of long term drifts is usually difficult due to temperature dependent offset drifts in the double balanced mixer or active multiplier circuits. Also cables and amplifier are susceptible to temperature drifts which can reach the order of 1 ps / deg. 7
C. Careful selection of components, temperature stabilization of the circuits and temperature compensation are necessary to achieve a long time measurement stability of the order of 10 100 fs. For the measurement of the timing jitter of optical signal with respect to a rf reference the optical signal is converted to an rf signal. In the case of short laser pulses (femto to picosecond duration) with a repetition rate of a subharmonic of the rf reference, this can be accomplished by with a photodiode. A beam splitter send a part of the optical signal to a photodiode which creates short electrical pulses (of the order of a few hundred ps for a diode bandwidth of several GHz). This pulses contain all harmonics of the laser repetition frequency and a bandpass filter is used to generate a ringing at the desired harmonic. The bandwidth of the filter should be selected according to the desired measurement bandwidth. Typical signal levels are of the order of a few mv at the desired frequency which are amplified to desired rf level of the order 10dBm to 0 dbm. The time jitter measurement is then based on a phase measurement of the rf signal. The accuracy of this method is limited by amplitude to phase conversion in the photodiode and requires highly amplitude stable optical signals of the order of 10 4 for a timing jitter resolution of 10 fs. The long term stability is also limited by the temperature dependency of the photodiode. 5.2 OPTICAL TIME JITTER MEASUREMENTS Different optical systems, for example, the optical clock, the photo injector laser, lasers for beam diagnostics or pump probe lasers must be synchronized to each other. Several solutions for an optical to optical synchronization have been investigated, for example using the balanced cross correlator technique via sum frequency generation [4]. Another approach is performed by using a Mach Zehnder interferometer scheme consisting of phase modulators, balanced diode detectors and a VCO in a phase lock loop configuration to provide an optical to rf conversion [5]. The 8
locked VCO output signal drives the phase modulator of a second interferometer connected to the laser to be synchronized. Furthermore direct phase noise measurements of optical signals using the interferometer can be performed by commercial rf phase noise instruments. 5.3 RF FIELD STABILITY MEASUREMENTS The cavity field detection can be accomplished with traditional amplitude and phase detectors or with IQ detectors which operated directly at the rf operation frequency or at an intermediate IF frequency which contains the amplitude and phase information. Another possibility is a scheme employing digital IQ detection where the IF (or the RF signal) is sampled directly by an ADC which usually samples alternating the real and imaginary components of the cavity. This of course requires correct timing of the data acquisition. With the rapid development of the telecommunication market industry a variety of single chip solutions for amplitude detection, phase detection, and IQ detection based on analog multipliers have been developed. Examples are: AD8343, RF2411, LT5522, LT5526 analog multiplier AD8361 linear video detector (temperature stabilized) AD8302 logarithmic video detector and phase detector HMC 439 digital phase detector AD8347, LT5516 IQ detector HMJ7 1 high level FET mixer The same circuits are also used to detect the incident wave and reflected wave vectors usually described as forward and reflected power. 6. SYNCHRONIZATION OF ACCELERATOR SUBSYSTEMS 9
The basic building block for a synchronization system is a phase locked loop (PLL) as shown in Figure 3. It consists of a phase detector to detect the timing error between the two systems to be synchronized, a controller filter which amplifies and filters the error signal, and an actuator for control. This can be a voltage controlled frequency source (rf or optical) in the case where a frequency source is phase locked to a reference, or a phase shifter or vector modulator in the case of rf field control. In the latter case also the amplitude of the field must be controlled. PLLs can be simulated using [6]. Concerning phase noise, respectively timing jitter, Figure 4 shows the difference, so called residual phase noise spectrum between a reference signal and the phase locked signal. Phase noise can be reduced within the loop bandwidth when using a feedback. The residual uncontrolled noise depends critically on the loop bandwidth and the noise generated in the detector for the phase measurement. Typically the master oscillator should be a source with the lowest phase noise available. Very important is the phase noise close to the carrier because all other subsystems will be locked to this reference. The lock bandwidth is typically chosen to be of the order of a few khz since the phase noise for frequencies >10kHz from the carries is negligible i.e. < 10 fs. Broadband phase locked loops would add unnecessary noise into the synchronized system. The typical frequency roll off of a phase locked loop is 20 db/decade since the voltage controlled oscillator (VCO) provides the integrator function in the loop. An additional integrator is required if temperature dependent offsets in the stages following the phase detector must be suppressed. In the case of phase control in the rf cavities one must distinguish between pulsed or cw operation of the cavities. While usually an integrator is applied in cw operation one avoids the integrator in pulsed operation to guarantee a constant error during the short pulses. 7. ARCHITECURE OF SYNCHRONIZATION OF ACCELERATOR SYSTEMS 10
The basic components for accelerator synchronization systems are: RF oscillator (crystal, surface acoustic wave, dielectric resonator) Mode locked laser oscillator (fiber, Ti Sa etc.) Optical to rf and rf to optical converts Coaxial frequency distribution Fiber optic distribution (optical pulses or rf modulated optical cw laser) Beam pick ups (beam phase, beam position) and beam diagnostics (bunch length etc.) Low level rf control systems The concept of a precision synchronization system for large scale accelerators will employ most of the above elements in a combination which will give the utmost stability of the subsystems which have to be synchronized. A general scheme of this concept is shown in Figure 5. The master oscillator makes use of a fiber optic laser which provides low phase noise and its optical distribution system which is stabilized by interferometric methods. Particularly for the master laser oscillator, when using passively mode locked Er/Yb glass lasers with sub 20 fs timing jitter have been demonstrated recently [7]. The long term stability of the fiber laser is guaranteed by a crystal oscillator and atomic reference which both provide the long term stability. The optical references are distributed to strategic locations where they are converted to rf signals to serve as reference for the low level rf control systems. Optical systems such as the lasers for the photocathode, diagnostics or pump probe experiment should be synchronized directly with optical methods. Finally the beam information will be used to guarantee the long term stability. 8. STATE OF THE ART PERFORMANCE 11
Present state of the art in performance of synchronization of optical system has surpassed the 1 fs level in a laboratory environment [4,8]. However in a noisy accelerator environment this performance needs to be demonstrated. For rf measurements the best results that have been obtained are of the order of 10 fs at a frequency of 1.3 GHz. For rf control the best results are of the order of 10 4 for amplitude and 0.01 deg. for the phase of superconducting cavities operating at 1.5 GHz [8]. 9. DISCUSSION AND CONCLUSIONS In summary, a scalable timing distribution and synchronization scheme for future accelerators, free electron laser facilities and ERLs is introduced. The overall residual beam jitter between high energy photon beam produced by an electron bunch and e.g. pump probe laser depends on various uncorrelated jitter contributions. Especially from the stability of the master clock, timing distribution system, optical converters, stability of laser systems itself, rf systems and the locking performance of each subsystems to each other. A crucial role concerning long term phase stability and jitter contribution plays the performance of phase detectors and feedback systems for noise reduction. State of the art, commercial available low noise phase and amplitude detectors contribute on the fs scale to the jitter. A careful selection of components, low noise electronic design, temperature stabilizations and self calibrating methods are necessary to achieve the long term stability. A variety of detectors specialized for low noise, high linearity and long term stable operation are necessary for feedback systems. Each subsystem has to be noise optimized. The synchronization of optical systems and the signal distribution on the fs scale has to be approved in a noisy accelerator environment. Concerning the residual jitter 12
between the beam and the pump probe laser an overall noise budget for each accelerator including all subsystems is required. The overall jitter is mainly determined by the differences of loop bandwidths from the subsystems and their uncorrelated noise contributions. This requires either a matching of loop bandwidths with moderate phase contribution from master clock or a negligible phase noise contribution from master clock while having a much better variation of loop bandwidths. To reduce the overall jitter minimization of the number of distributed frequencies within the accelerator and its subsystems is desired. However, with timing stabilized fiber links, ultra low jitter mode locked laser, low noise oscillators and detectors a large scale timing distribution and synchronization technique is available in the near future for the next generation of FELs and ERLs. 13
FIGURE CAPTIONS Figure 1. Schematic of FEL with subsystems. Figure 2. Schematic of ERL with timing relevant subsystems. Figure 3. Basic configuration of a phase locked loop. Figure 4. Phase noise budget for synchronized systems. Figure 5. Architecture for synchronization of accelerator subsystems. Figure 6. Example for beam energy and timing stability for the VUV FEL.. 14
REFERENCES [1] L. Merminga and J. J. Bisognano, Energy Stability in a High Average Power FEL, Proceedings of PAC 95 [2] L. Merminga, P. Alexeev, S. Benson, A. Bolshakov, L. Doolittle and G. Neil, Analysis of the FEL RF Interaction in Recirculating, Energy Recovering Linacs with an FEL, NIM A 429 (1999) 58 64 [3] G.R. Neil et. al., Sustained Kilowatt Lasing in a Free Electron Laser with Same Cell Energy Recovery, Phys. Rev. Lett. Vol. 84, No 4, pp. 662 665 (2000) [4] L. Ma, R. K. Shelton, H. C. Kapteyn, M. M. Murnane, J. Ye, Sub 10 femtosecond active synchronization of two passively mode locked Ti:sapphire oscillators, Phys. Rev. A 64, 021802 (R) (2001) [5] J.Kim, F. O. Ilday, F.X. Kärnter, O. D. Mücke, M. H. Perrott, W. S. Graves, D. E. Moncton, T. Zwart, Large Scale Timing Distribution and RF Synchronization for FEL Facilities, Proceedings of the 2004 FEL Conference, 339 342 (2004) [6] M. H. Perrott, PLL Design using the PLL Design Assistant Programm, MIT High Speed Circuits and Systems Group, Cambridge MA, 2002 15
[7] J. B. Schlager, B. E. Callicoatt, R. P. Mirin, N. A. Sanford, D. J. Jones, J. Ye, Passively Mode Locked Glass Waveguide Laser with 14 fs Timing Jitter, Opt. Lett. 28, 2411 (2003). [8] R. K. Shelton, S. M. Foreman, L. Ma, J. L. Hall, H. C. Kapteyn, M. M. Murnane, M. Notcutt, J. Ye, Subfemtosecond timing jitter between two independent, actively synchronized, mode locked lasers, Opt. Lett. 27, 312 (2002) [9] M.Liepe, Overview of LLRF Systems, Proceedings of PAC 05 16
Figure 1 17
Figure 2 18
Figure 3 19
Figure 4 20
Figure 5 21
Figure 6 22