1. CLIC Crab Synchronisation
Sam Smith and Amos Dexter
Eucard2 WP12 Annual Meeting
14th March 2017
Hello, so I’m going to be presenting work done on the CLIC crab cavity synchronisation and phase measurement and control. This work has been done by myself, Amos Dexter and others at Lancaster university.

2. Outline CLIC requirement
CLIC Crab cavity synchronization requirements - background
RF Distribution options
Planned solution for CLIC
Microwave Interferometry
Test measurement and control scheme
Signals with phase noise
Estimation of phase measurement precision
Results on phase measurement precision (problems encountered)
Development of RF front end, data processing
Digital sampling and control
Test boards
Calibration and correction
Phase shifters
Requirement and design
Actuation
Performance
Here’s a brief outline of the things I’m going to cover, starting with the requirements for CLIC and the proposed solution, then preliminary results on microwave interferometry from tests at the lab, development of the RF front end and calibration and finally development of high power phase shifters for use in the project.
Achievements

3. Synchronisation Requirement
collision point has small displacement
late electrons
timely positrons
If crab cavities are synchronised then rotation centres synchronised
If crab cavities are not synchronised then they completely miss each other
Cavity to Cavity Phase synchronisation requirement (excluding bunch attraction)
So in order for maximum luminosity to be maintained the crab cavities rotating the bunches have to be synchronised. Here we can see that for CLIC the calculated requirement is 4.4femtoseconds which translates to around 19 milli degrees across a 156us pulse length. This is important is if this is not met, the bunches can arrive at different times and possibly miss each other.
We can see here that even if the electrons arrive late, as long as the crab cavities are synchronised, the bunches will still collide, but the collision point will just be shifted slightly.
Target max. luminosity loss fraction S
f (GHz)
sx (nm)
qc (rads)
frms (deg)
Dt (fs)
Pulse Length (ms)
0.98
12.0 45
0.020
0.0188
4.4
0.156

4. RF Distribution Options
Single klystron with high power RF distribution to the two crab cavities.
Klystron phase jitter travels to both cavities with identical path lengths.
BUT
Requires RF path lengths to be stabilised to 1 micron over 40m.
Option 2:
Klystron for each cavity synchronised using LLRF/optical distribution.
Femto-second level stabilised optical distribution systems have been demonstrated (XFELs).
BUT
Requires klystron output with integrated phase jitter <4.4 fs.
There were two possibilities for meeting this requirement, the first being a single Klystron with high power distribution two both crab cavities which would mean that both cavities have the same phase jitter, but then path length would have to be stabilised to 1um over the 40m waveguide length. The second option was to use two different Klystrons synchronised using optical distribution but this would require klystron outputs with an integrated phase jitter of less than 4.4fs which is very difficult to achieve and so therefore it was decided that option 1 was the best option for CLIC

5. CLIC Synchronisation Proposal
RF path length is continuously measured and adjusted
4kW 30ms pulsed
GHz Klystron repetition 2kHz
Cavity coupler 0dB or -40dB
Cavity coupler 0dB or -40dB
Waveguide path length phase and amplitude measurement and control
Forward power long pulse 1 kW
Forward power main pulse 12 MW
Piezo phase shifter
Mechanical phase shifter
LLRF
-30 dB coupler
-30 dB coupler
LLRF
Magic Tee
Reflected power main pulse ~ 600 W
Reflected power long pulse ~ 500 W
Here we can see the proposed synchronisation method. We have the main Klystron in green at the bottom providing 48MW to the cavities through the magic tee pulsing at 50Hz. We would then have a second low power klystron seen here in blue pulsing at 2kHz This klystron feeds the magic tee from the other port splittint the power between the two lines. It runs at at a frequency of which is outside the cavity bandwidth meaning that the pulse is reflected. These reflected pulses will then measured using directional couplers, mixed down to a sampling frequency and then the phase of the waves will then be compared so that a fast high power piezo phase shifter can be used to adjust one arm to bring the phase difference between the two lines back to zero.
The high power phase shifters required for the project are currenly under development and we’ll come back to them a bit later in the presentation.
Single moded copper plated Invar waveguide losses over 35m ~ 3dB
Phase Shifter
Main beam outward pick up
Main beam outward pick up
From oscillator
48MW 200ns pulsed
GHz Klystron repetition 50Hz
Vector modulation
12 GHz Oscillator
Control

6. Feasibility Measurement Scheme
+30 dBm
-10 dBm
Attenuator
Diode Switch
short
RF1 = GHz RF2 = GHz
short
Piezo activated phase shifter
phase shifter
Isolator
DC pulse
IF (46 MHz)
Magic Tee RF
control voltage for pulse LO
Directional coupler
Directional coupler
actuator control
National Instruments PXI Crate
16 bit ADC at 120MS/s on 4 channels
channel 1 = IF signal from Magic Tee Delta Port
channel 2 = DC signal DBM (range ~ 20o )
channel 3 = IF signal from right waveguide arm
channel 4 = IF signal from left waveguide arm
DBM ws ws
LPF
In order to check the feasibility of the measurement scheme a small scale test using around 4m of waveguide was devised to check whether it was possible. The RF is generated by a commercial PLL and goes through a diode switch which can be turned on and off in order to run in pulsed mode. This then passes through a 3W amplifier and into the magic tee where it is splits and reflects of the two shorts at the ends which act as the cavities in the system. Two directional couplers then tap a portion of the signal which goes into the phase measurement board for mixing.
There are three methods for measuring the phase difference between the two lines, the first is the DC offset given by the double balance mixer seen in green, as this is amplified it only has a range of around 20degrees. We then have the two signals from the directional couplers which are mixed down using the local oscillator and then asynchronously sampled, the phase difference between them is then calculated. Finally, there is the output from the 4th port of the magic tee which is mixed down using the LO to 46MHz and sampled using the PXI crate.
We then have a piezo phase shifter on one of the arms which adjust the length of the waveguide with high precision.
Amplified immediately to reduce noise – LPF part of amplifier ws
Circuit Board ws
ws = Wilkinson splitter
LO = GHz
Diode Switch
Attenuator
+20 dBm
-10 dBm

7. RF feed via amplifier and isolator
Electronics
RF feed via amplifier and isolator
Mixing PCB
From Magic Tee
NI PXI Spec.:
16bit ADC, 120MS/s, 4-channel digitizer
2.2GHz Celeron module with FlexRIO FPGA module
Pin diode switch to pulse RF
Optimise power levels 13dbm – attenuate for best performance – filters on inputs 100mhz for high harmonics
Replace 50mhz to 40 LO RF
Locked DI Instruments Oscillators ($700)
National Instruments PXI crate

8. Waveguide runs around edge of RF lab – phase shifters and attenuators on each arm

9. Digital Sampling Hardware
LabVIEW used for acquisition
Front panel allows for control of calibration, pulsing timings, piezo actuator and ADC clocking frequency.
GUI on host computer allows for real time viewing of signal spectrums
NI PXI Spec.:
16bit ADC, 120MS/s, 4-channel digitizer
2.2GHz Celeron module with FlexRIO FPGA module
Here is a small section from the LabVIEW code developed for the test. Through the user interface the calibration, pulsing parameters, data acquisition and piezo actuator can all be controlled. The GUI on the host computer provides real-time viewing of the signals and of the signal spectrums

10. Test Boards
Test board 1
Multiple test boards have been developed to achieve optimum performances at 10-12GHz.
Test board 1 examined:
Choice of optimum transmission line (Microstrip, CPW, GCWP, Stripline, …)
Impedance matching and transmission responses
SMA connector type, bends, etc.
And measured wavelength
Test boards 2 & 3 have:
A mixer
2 TL lines differing in length by λ/4
A Wilkinson splitter
A ring resonator
Test boards 2 & 3 examined:
The ‘real’ PCB ɛr
The effect of soldermask
Improved pads, via locations, copper-to-edge, etc.
Test board 2
A number of test boards have been developed for the project so that the optimum phase measurement board could be made. The first had a number of different lines including CPW, microstrip and stripline. From this it became apparent that GCPW was the best choice for the board. Test board 2 and 3 had some of the components used on the board such as a splitter and mixer and a resonator so that the permittivity of the board could be compared to simulation.
Test board 3

11. Connector Simulations
Excessive reflection from test tracks
CST simulations model 3D configuration of connectors , tracks and via positions.
Track width and taper optimised to match connector and launch pad minimising reflection at operating frequency
Track width taper at launch pad
So we were getting much more reflections than expected were seen on the straight lines on the test boards when they were measured. A number of COST simulations were done to try and model the interaction between the connectors and the launch pads. We found that the connectors were not matched to the tracks at the launch pad, the track width was then optimised using a taper in order to provide a better match, this taper was then used in the final phase measurement board.

12. Added to improve isolation on DC mixer
New Board 2016
The phase board must split signal as with minimal reflection, track lengths are careful set so that reflections cancel. The board must be compact and dimensionally stable.
Down conversion mixers
DC to 1 GHz low noise Amplifier
IF output 1
RF input1
DC out
LO input
Mix to DC
IF output 2
RF input 2
Here we have the latest phase measurement board that does the downmixing and provides the DC measurement. The two RF signals from the left hand side are split, half go into another mixer which provides a DC voltage which depends on the phase difference between the lines, this is then amplified using a low noise amplifier and recorded by the PXI crate.
The other halves are mixed down to 46MHz using the local oscillator input here and then sampled by the PXI crate. Low pass filters were added to the board to provide better isolation between the DC and If mixers.
IF filters
Added to improve isolation on DC mixer
Green Solder Resist removed from above CPW
Tappers introduced to PCB plan for improved connector matching

13. Notes on Synchronisation
DBM = double balanced mixer
DBM1
DBM2
DBM3
RF LH
RF RH LO
For synchronisation DBM1 controlled to zero.
Measurement independent of oscillator phase noise.
Corrections on phase measurements require knowledge of amplitudes.
Very small ‘d.c.’ voltage pulses lasting the length of the RF pulse must be measured.
Offsets can be determined between pulses and then removed.
High amplification on DBM1 means that 360o cannot be measured (PXI input limitation).
Direct sampling of DBM2 and DBM3 allows:-
360o to be determined
course phase variation on each arm to be monitored
phase differences to be brought to range of DBM1
calibration of DBM1 output – tells us how close to zero we are
monitoring separate arms of interferometer needs RF and LO to be locked
DBM2 and DBM3 are used to calibrate the output from the mixer at DBM1.

14. Phase Determined from IF
Choice of 16 bit 120 MS/s ADC forced use of asynchronous sampling
Deduce phase using
where
Simulated ADC errors (assumes noiseless input to the ADC)
ADC Aperture jitter = 80 fs (rms) and noise = random +/-15
We were forced to use asynchronous sampling as the frequency had to be outside of the cavity bandwidth, IQ sampling would have put the IF frequency inside the cavity BW. We use this equation here to calculate the phase difference between the two signals.
At the bottom we can see the simulated phase error for the ADC aperture jitter and random noise, you can see that it’s around 100 millidegrees.

15. Expected verses Actual Performance for IF
The results on this slide show the simulated phase error for asynchronous and IQ sampling. Some of the first results for the measured IF phase are shown at the bottom. Originally this was found to be around 400 millidegrees, nearly 4 times

16. Amplitude Determined from IF
Amplitude determined from adjacent sampled points yo and y1 on IF waveform using
Originally notch filters are applied to the raw data to remove the IF frequency of 46 MHz and other spurious frequencies
The amplitudes of the intermediate frequency can also be calculated from the asynchronous sampling formulas, here we can see the measured amplitudes for the two arms. One arm had slightly more noise than the other due to the different feeding ports of the mixers.
Expected spread on measured values ~ 35, actual ~ 60, (left slight more noisy than right)

17. Magic Tee - Amplitude Dependence of Phase
The interferometer launches on port 1 has a return signals on ports 2 and 3 with slightly different amplitudes and phase.
We require phase difference q3 - q2
For the perfect Magic Tee we have
Left =3
Right =2
H =1
Measuring amplitude V4 from port 4 we have
As well as the DC and IF measurements we also have the measurement from the magic tee. The phase difference between ports 2 and 3 which is the phase difference between the two arms of the interferometer. We can see that the accuracy of the magic tee measurements depends on how well we can measure the amplitude of each arm.
The phase difference between ports 2 and 3 depends on input amplitudes to the ports as well as output on port 4.
The accuracy of determining the phase difference between returning signals on the left and right arms of the interferometer depend on accuracy of our measurement of amplitude

18. Magic Tee and IF Measurements Compared
Vertical range is 1 degree for both graphs
Here we have some preliminary results for the magic tee compared with the IF. The magic tee measurement was found to be more noisy than the IF amplitudes, probably due to a different mixer being used to downconvert the signal.
The Magic Tee measurements did not use the phase board and SIM24 MH+ mixer, this might explain the slightly high level of noise.

19. DC Measurement <20 milli-degrees within a single pulse achieved
Inter-pulse drift ~ 20 milli-degrees – phase shifter can be used to remove this
DC and MT measurements are calibrated using the down converted phase.
DC has less noise but its usefulness depends on amplitude correction
Finally we have the DC measurement which was giving us the best result from the three. You can see that it was significantly better than the IF measurement, and was found to be able to give the 20 millidegrees required for synchronization.
But the question was, why were the other measurements so bad in comparison?

20. IF problems - Spectrums
Unwanted frequencies at 18,28,56MHz
We started to investigate the spectrums for the signals and found a number of extra frequencY components in the signals that we had not expected, here we can see that the IF and MT had extra frequency components at 18,28 and 56 MHz but that these frequencies were not present in the DC.

21. IF problems - solutions
Unwanted frequencies generated by mixing between the harmonics of the IF frequency and the 120MHz clock. Other harmonics not present on DC as 50MHz filter is better than others.
IF frequency in DC removed with notch filter.
120 - (2x46) = 28MHz
120 - (3x46) = -18MHz
240 – (4x46) = 56MHz
100MHz filters on IF
50MHz filter on DC
Although we had analogue filters in place on the PXI crate these frequencies were the product of higher order harmonics of the intermediate frequency of 46MHz and the sampling clock at 120MHz. 28 generated by *IF, 18 generated by 120-3*IF and finally 56 generated by 240-4*IF. The extra frequency components were not seen on the DC measurement as there was a better 50MHz filter on the line. A bandpass filter around 46MHz was applied to the IF and MT measurements to remove the extra frequencies and a notch filter was applied to the DC to remove the IF leaking into the channel. A mixer from the test boards

22. IF problems + solution Noise spread ~ 40
Measured phase error ~ 140 milli-degrees
Measured phase error ~ 100 milli-degrees

23. Phase Shifter Requirement and Design
Phase shifter must:
Work in high power conditions ~20 MW
Give at least 4 degrees of fine tuning and half a wavelength of coarse tuning
Have fast response times – 2 degrees of phase shift in 20ms and 0.1 degrees in 4ms (time between pulses is 20ms )
Suitable for automation and integration into a control loop
SHIM
Highest piston position
Some extra work that has been going on is the development of high power phase shifters for use in the project. Relatively few examples of waveguide phase shifters that were capable of operating in the MW range were found so new designs had to be developed. A prototype phase shifter based off a 3dB coupler design by Alexej Grudiev. It utlisies a piezo actuator to convert the voltage applied into a precise phase shift. This is the phase shifter currently being used in the tests shown here.
Lowest piston position
Prototype high power phase shifter built at Lancaster university, being used for current testing
3dB Hybrid design – adapted from Alexej Grudiev
CLIC – Note – 1067

24. Prototype Phase Shifter Performance

25. Prototype Phase Shifter Performance
PI controller implemented to control phase shifter – Localised heating applied to waveguide at the same temperature – different gains for controller tested
Controller shown to be able to bring phase back to phase set point using phase shifter
More work to be done to optimise controller
Seeing random phase spikes – source needs to be found and eliminated

26. Lasted Phase Shifter Design
Converts waveguide TE10 mode to two polarisations of the TE11 circular waveguide mode.
Design – Alexej Grudiev
CLIC – Note – 1067
New high power phase shifter developed at CERN for CLIC. Design allows for integration of a stepper motor and piezo actuator giving solutions for the fast and slow phase shifters required.
Provides 20 degrees per mm, giving a piezo range of 6 degrees, enough for expected thermal expansion.
Drawings are finished and manufacturing will begin soon.
Flange allows for 2 attachments – motor for slow movements and piezo actuator for fast response

27. Achievements
A PCB for mixing RF signals to d.c. and simultaneously mixing to an IF frequency has been developed. A key feature of the board is the management of path lengths to cancel reflections.
An X-band waveguide interferometer has been set up with Piezo-activated phase shifters to control arm lengths.
A National Instruments PXI data acquisition and control system has been set up to measure the phase difference and amplitude of signals returning on the interferometer arms.
Measurement of phase differences at the resolution of 10 milli-degrees for 30 micro second pulses X-band has been demonstrated.
Drawings for high power phase shifter completed – will be manufactured soon
Controller implemented to keep phase difference constant

28. Thanks for listening! Future work Optimize controller
Put full waveguide length into environmental chamber to monitor temperature drifts
Investigate vibration effects on and near waveguide
Possible high power tests at CERN of new phase shifter in waveguide when completed
Thanks for listening!