Design considerations for the RF phase reference distribution system for X-ray FEL and TESLA Krzysztof Czuba *a, Henning C. Weddig #b a Institute of Electronic Systems, Warsaw University of Technology, Warsaw Poland b DESY, Hamburg, Germany ABSTRACT The RF Phase Reference Distribution System (PRDS) must deliver a highly RF phase stable signal to many various RF subsystems of the X-ray Free Electron Laser (XFEL) and in the future the TESLA linear collider. The required phase synchronization corresponding to the short term stability of 1ps must be guaranteed. Taking into consideration large amount of devices to be synchronized, long distances and necessity of delivering different frequencies, the design of PRDS becomes a very difficult and challenging task. This paper describes the main considered issues. Such parameters as distribution frequency, waveguide attenuation, multiplier noise and temperature influence on the system are taken into account. The advantages and disadvantages of coaxial cables and optical fiber as the distribution medium are compared. The feedback system stabilizing long term phase drifts is presented and the structure of PRDS which may fulfill the design requirements is proposed. Keywords: FEL, TESLA, phase noise, frequency stability 1. INTRODUCTION Big European experiments like XFEL and (in future) the TESLA [1] are being developed in DESY [2] in Hamburg, Germany. Both experiments are based on superconducting linear accelerator structures where electron bunches will be accelerated to the extremely high energies 500 GeV is planned for TESLA. The basic element of the accelerating structure is 9 cell superconducting resonator cavity. The operating frequency is 1.3 GHz. Groups of cavities are boosted by 10MW klystrons controlled by the low lever RF system which must assure acceptable amplitude and phase stability so that the electron beam is accelerated properly. Large amount of RF system devices will be located along the accelerating structure which length will reach 33km. Phase stable reference signal must be provided to mentioned devices to make possible the operation of the machine. Taking into account the required stability, the distribution distance and the number of synchronized devices (may reach 1000 in TESLA) one will find the design of PRDS very challenging and difficult task, but not impossible. The paper describes considerations carried out in the early stage of the design. The use of coaxial cable and fiber optic as the distribution media is considered. The most important system parameters like the choice of distribution frequency and phase noise consideration are pointed out. The feedback scheme for stabilizing slow phase drifts is proposed and system layout which may fulfill the requirements described. The proposed system is built and will be tested in the Tesla Test Facility 2 (TTF2). 2. SYSTEM DESIGN REQUIREMENTS The reference signal phase stability requirements derive from the low beam energy spread requirement (s E /E<7x10-4 ) and the timing requirement that determines the arrival of bunches at the interaction point [3]. The interaction position should not exceed one bunch length (1 mm). Both requirements lead to the maximum allowed reference signal phase fluctuation of less than 0.5 o at 1300 MHz corresponding to the timing error equal 1 ps. The long term stability should not exceed 10ps within days. The main system frequency is 1300 MHz (exactly 1299.9996 MHz but for simplicity the numbers are rounded) but there are several devices that require other frequencies and must be synchronized with the master oscillator of the system. Power level required at outputs of PRDS is 0 dbm or +10 dbm so it was assumed that each output will provide +10dBm. * kczuba@elka.pw.edu.pl; phone +48 22 660 7663, # Henning-Christof.Weddig@desy.de; phone +49 40 8998 4910
Next important requirement is the reliability of the system. During 10 years of operation there will be only few days per year allowed for service and repair. This leads to the requirement of less than 0.1% system failures per month which will require accelerator shutdown for immediate repair. More details about main requirements are listed below:? system length - 33 km TESLA, 4 km XFEL, 280 m TTF2;? RF phase stability (short term, minutes) - 0.5 o at 1.3 GHz, corresponds to 1 ps in time domain;? RF phase stability (long term, days) - 5 o at 1.3 GHz (10ps);? RF phase noise - 0.5 o at 1.3 GHz in 1MHz bandwidth;? distributed frequencies - 1 MHz, 9 MHz, 13.5 MHz, 27 MHz, 81 MHz, 108 MHz, 1300 MHz, 2856 MHz? number of outputs - ~ 1000 TESLA? power level - +10dBm each output 3. DESIGN ISSUES The study on the system was performed to find a solution that may fulfill requirements listed above. Most important issues were described in this section:? distribution medium attenuation, thermal phase stability, immunity to EMI;? master oscillator frequency;? frequency multiplier/divider parameters;? frequency to be distributed;? number and location of outputs;? power level of signal at each location;? amplifier parameters - noise and nonlinearities;? output separation;? power supply parameters;? reliability? overall system cost Many of listed issues are related and it is difficult to provide separate discussion on each point and find optimum solution. Therefore basic issues are discussed below having in memory that the description is not complete. Nevertheless it helps to find and understand main problems and can be a very good base for further development. 3.1 Distribution medium Coaxial cable and optical fiber were considered as the distribution medium. Both have similar phase length versus temperature coeffic ients ~10ppm although the temperature coefficient in fiber is due to the change in refractive index, not the change in physical length as in coaxial cable. The value of coefficients affects the long term stability of PRDS. For example over 15km of coaxial cable the phase of 1.3 GHz signal would drift ~206 o / o C what leads to the temperature stability requirement for the cable equal 0.002 0 C - impossible or at least very difficult to meet. The temperature in the TTF 2 tunnel was measured and we expect 2 3 o C changes per day and up to 10 o C change over long time summer/winter. There are also significant temperature variations along the tunnel, especially near the entrances. One could find temperature compensated optical fiber but using it in so large experiment would lead to unacceptable rise of system cost. Feedback system compensating the phase drifts is required to meet the system requirements. It will be described in section 4. But it must be mentioned here that the system provides phase compensation at the end of the link and since there are temperature variations along the link, the performance may not be met at tap points placed along the distribution line of significant length. Next important parameter of the distribution medium is the attenuation. For typical optical fiber this parameter has a low value <0.3 db/ km which is independent of the RF frequency. The coaxial cable attenuation is high and frequency dependent e.g. 3.6 db/km at 9 MHz, 11.2 db/km at 100 MHz and 49 db/km at RF frequency for the 7/8 Heliax type. Therefore the use of coaxial cable is limited to short distances or use of cascaded sections with amplifier at the input of each section. Important are also physical parameters. The optical fiber cable has the advantage that it is much smaller in diameter and its installation is much easier. The next advantage of optical fiber is excellent EMI immunity but in the accelerator surrounding radiation will be present which may affect fiber parameters. This phenomenon will be tested. Considering the costs, the optical fiber is relatively cheap but the laser transmitter satisfying
system requirements (DFB type with thermo electric cooler and temperature controller) and the phase controlling system are relatively high. Since optical links are suited for point-to-point operation, the PRDS can not be based only on optical fiber because large amount of links would cause unacceptable cost. 3.2 Master oscillator Oven controlled crystal oscillators available currently in the market easily satisfy the stability requirements. It can be found that the best oscillator performance is obtained for frequencies around 10MHz. The parameters are still satisfying up to 100MHz but the performance decreases significantly. Therefore the frequency must be chosen within this range but its influence on the system performance must be considered. The accelerator operating frequency equals exactly 1299.9996 MHz. Divided by 144 (must be an integer number because of worse phase noise parameters of fractional frequency mu ltipliers) it gives the exact value 9.027775 MHz for the best obtainable quartz crystal oscillator. This value makes it difficult to synchronize with common atomic standards frequency like GPS. The SSB noise of the 9MHz is shown in fig. 1. -100-110 phase noise dbc/hz -120-130 -140-150 -160-170 1 10 100 1000 10000 100000 1000000 offset from carrier frequency Hz Fig. 1. SSB phase noise level required for the master oscillator. 3.3 Frequency multipliers and dividers Master oscillator phase noise level will increase of 20logN (if multiplier, in band noise [4, 5]) where N is the multiplication factor. Therefore the lower the frequency the higher noise performance degradation e.g. ~43 db for the 144 times multiplier. The noise presented in fig. 1 degraded about 43 db will still satisfy the design requirements. Conversion between different phase noise units, e.g. SSB level to time jitter, can be found in [7]. Other multiplier parameters like long term stability and power supply amplitude noise to RF phase noise conversion must be taken into account. multipliers are being developed and will be tested soon. Special attention will be paid to phase noise, temperature effects on stability, phase of the output signal related to the input after turning the device on and signal phase drifts between different devices outputs connected to one signal source. Currently available digital frequency dividers offer excellent noise performance with noise floor below -150 dbc/hz. The problem may appear when turning off and on divider connected to frequency higher than frequency. The divider output signal will have constant phase shift related to signal with value determined by the number of periods of the high frequency signal between the turn on time and the zero -crossing of the signal. Procedures eliminating this effect must be worked out if multiple frequency dividers will be used. 3.4 Distributed frequency Considering the problems pointed out in sections 3.1 3.3 one can find the choice of distribution frequency a very difficult task. There were different frequencies specified in section 2 but 1300 MHz is the most important frequency so further, in this paper only this value will be considered to be provided to the destination points. Figure 2 shows three basic distribution schemes suited to distribute different frequencies. If the distributed frequency equals 1300 MHz, fig. 2a, then one of the advantages is use of only one multiplier. But such a high frequency will lead to very high attenuation in coaxial cable of ~49 db/km and necessity of using high power RF amplifiers which may have poor noise performance. Turning to the lowest frequency of M.O equal 9 MHz, fig. 2b, low attenuation in coaxial cable will be met and much smaller RF power will be required to cover distribution distances but separate multipliers with high multiplication factor (N) must be used at each destination point. As it was mentioned above the higher N the worse multiplier parameters and therefore there may be significant signal degradation at each location.
One may find a compromise between attenuation in cable and multiplier parameters by choosing intermediate distribution frequency like 81 MHz. The multipliers with N=16 may have better parameters than with N=144 and the attenuation may be below 10 db/km. x144 a b x144 x144 x144 x9 c Fig. 2. Basic system layout for depending on distribution frequency: a) 1300 MHz, b) 9 MHz, c) 81 MHz. Test version of the coax cable distribution system is built in TTF2 and each of proposed schemes will be evaluated experimentally. With the use of optical fiber as the distribution medium described problems with attenuation are eliminated and each RF frequency can be transmitted. But the system cost may rise to unacceptable levels because of very large amount of optical links. 3.5 Number of outputs, power level at each output, amplifier parameters and output separation All the parameters listed in the section title are related and must be considered simultaneously. The output power of the amplifier at the input of distribution line must ensure required power level at each destination point. One must consider that the signal power will decrease along the cable and there may be necessity to attenuate or amplify the signal locally. Dividing the system into cascaded sections with power amplifier at the input will have influence on the system phase noise [8] which must be studied. It is also valid for optical fiber links because the signal from the optical receiver will have to be amplified. Amplifier nonlinearities should be examined to find amplitude to phase noise conversion range. Output separation must be assured in order to prevent of phase jumps in the system after disconnecting one of the target devices. It can easily be obtained by the use of directional couplers but it causes significant output power decrease of e.g. 20 db. Amplifier noise and amplitude to phase nois e conversion will be studied and proper parameters for the amplifiers and power supplies will be specified. 4. PHASE STABILISATION FEEDBACK SYSTEM For both coaxial cable and optical fiber phase stabilization feedback system can be built based on the detection of the phase difference between signal transmitted into the line and signal reflected from the end of the line. The signal travels back and forth through the temperature sensitive medium. Reflection can be easy obtained by opening the end of coaxial cable or putting a mirror at the end of optical fiber. The signal from the phase detector can be used to adjust electronically controlled phase shifter inserted into the transmission line. Interesting scheme of the feedback system for phase stabilization in long optical fiber links is shown in fig. 3. The idea was developed for SLAC accelerator [9]. The measured output phase change at the end of 15 km fiber link as a response for 5 o C temperature step was below 2 o X- band (11.4 GHz). The phase step without the feedback would have been 12500 degrees! Considering the 1.3 GHz frequency the phase step would be about eight times smaller 1560 degrees. Therefore the feedback system would probably ensure stabilization on the level below one degree. But even two degrees satisfy required short term phase drift requirement of 5 degrees. The performance will be verified very soon in DESY because such system is being built and. Interesting component of the system is the optical phase shifter made of optical fiber spool in a temperature controlled oven. It seems to be the best solution for required range of phase shift. The 1560 degrees RF signal phase shift corresponds roughly to over 6 x10 5 degrees of light phase shift in optical fiber. Such phase changes can not be obtained with the use of commercial optical phase shifters offering maximum delays corresponding to 20 times 2p (7200 degrees). The oven solution seems to be very slow but it has satisfying parameters since the temperature in the accelerator tunnel will drift rather slowly.
RF Signal in DFB Laser Tx Directional Coupler 5 km single mode fiber in temperature controlled oven Circulator Long Link Mirror Directional Phase Shifter Coupler RF Phase Detector Fig. 3. Feedback system for phase stabilization in long optical fiber link Controller Phase stable RF Signal 5. CONCLUSIONS Most important issues of the RF phase reference distribution system were discussed in the early stage of the design. Such parameters like master oscillator frequency, distribution frequency, noise performance and long term stability were discussed. Coaxial cable and optical fiber were considered as the distribution media. Basing on performed considerations distribution system layout is proposed (fig. 4) which may fulfill the design requirements. The system will consist of N sections bas ed on coaxial cable supplied by long, phase stabilized optical fiber links using the feedback scheme described in section 4. Each cable section will also include feedback system locking the very end signal phase to the input signal phase. Test system is prepared for the 300 m long TESLA Test Facility 2 in order to study issues described in this paper and, if needed, change the system concept. N-1 phase stabilized fiber optic lines xn 1st section 2nd section N th section Master Oscillator Fig. 4. Proposed RF phase reference distribution system for XFEL and TESLA REFERENCES 1. http://tesla.desy.de. 2. http://www.desy.de. 3. A. Gamp, M. Liepe, T. Plawski, K. Rehlich, S. N. Simrock, Design of the RF phase reference system and timing control for the TESLA linear collider, XIX International Linear Accelerator Conference, Chicago, 98. 4. D. Wolaver, Phase-locked loop circuit design, Prentice Hall, New Jersey 1991. 5. M. Curtin, P. O Brien, Phase-locked loops for high frequency receivers and transmitters part 2, Analog Dialogue 33-5, Analog Devices 1999. 6. T. H. Lee, A. Hajimiri, Oscillator phase noise: A tutorial, IEEE Journal of Solid-state Circuits, vol. 35, no. 3, March 2000. 7. Phase noise theory and measurement, Application note1, Aeroflex, New York. 8. K. V. Puglia, Phase noise analysis of component cascades, IEEE Microwave Magazine, December 2002. 9. J. Frisch, D. G. Brown, E. L. Cisneros, The RF phase distribution and timing system fort the NLC, XX International Linac Conference, Monterey, California.