1. Experience with X band klystrons and solid state modulators at CERN
Gerry McMonagle
& Xbox team
International Workshop on Breakdown
Science and High Gradient Technology
(HG2015)

2. At the CLIC workshop in 2007 CERN decided that to complement the testing being done at KEK and SLAC a high power X band power source is necessary for testing structures on site at 12GHz
Market survey showed that there was no source available from industry
Initial contact with SLAC klystron department for a collaboration to develop a 12GHz klystron based on the existing 11.4 GHz tube was positive
Wheels in motion for a collaboration between CERN and SLAC with financial contributions from CERN (1 year administrative delay).
PSI and TRIESTE joined collaboration.
5 klystrons ordered (1 CERN and 2 each PSI and Trieste)

3. Installation and Requirements
Klystron
50 MW
1500 ns
50 Hz
test slot
200 MW, ns or
100 MW, ns
phase
modulation
hybrid
Modulator
pulse compression
CTF2
Gallery
Specification for modulator based on XL5 parameters form SLAC
In kind contribution for modulator from France (CEA Saclay)
Only limitation in call for tender was size and volume, not the technology
Scandinova won the bid with solid state technology

4. XL5 X Band Klystron spec Parameters Specifications units RF Frequency
GHz
Peak RF power 50 MW
RF gain dB
Efficiency %
RF pulse length
1.5 µs
Pulse repetition rate
50 (100) Hz
Duty cycle
Klystron voltage kV
Klystron current A
Inverse klystron voltage
100 max
Heater voltage
20 max V
Heater current
30 max
Heater start up surge current
?max
Preheating period 60
min

5. Modulator delivered to CERN (only tested in dummy load waiting for klystron)

6. XL5 performance at SLAC (Feb 2010)
Voltage, Current and rf waveforms at 57MW 1.25us
Tube was damaged in manipulation after tests.
2nd tube delivered to CERN late 2010

7. Klystron installation in modulator with help from our SLAC collaborators
Many teething problems with modulator calibration, interlocking etc. But we got first pulse on klystron before end of the year, stopped to do some modifications to modulator

8. Klystron testing Vacuum leak developed in klystron
Repaired in situe at CERN (several months)
Several issues with believing current and voltage calibration of the modulator
But tube was pulsing !!!
If we believe the figures from the voltage and current waveforms the perveance on the tube is lower than what was seen at SLAC
There was a problem with the tube but diagnostics were difficult as we could not believe our measurements.
Modulator was not tuned for klystron perveance, lots of overshoot and not a flat pulse.
Did not help that the current measurement transformers were on the ”low” side of the secondary pulse transformer windings enhancing the overshoot at the beginning of the pulse due to capacitive coupling.
Eventually did a mechanical squeeze of gun to correct perveance

9. While repairing klystron some additional measurement systems were implemented in the modulator to help with our calibration doubts
Started pulsing klystron at short pulse width on 30th August 2011
All seems fine running at around 370kV 50Hz
By 2nd September we got to near nominal pulse width at 50 Hz and at approximately 370kV

10. Some comments on the operation of Xbox1 modulator
Nice pulse to pulse stability
Pulse shape better once klystron problems were resolved but not yet optimised as we never got to nominal voltage due to testing schedule.
GUI
Not very user friendly
Hexidecimal inputs and readings for interlock levels etc.
No realtime and date stamp for interlock events that occurred.
No accumulative counter for high voltage hours
Only counts from when modulator is turned on and reset if powered down
Initially no ΔT body temperature measurement, resolved by Scandinova

11. Xbox2 modulator and Klystron
CERN decided to buy commercially produced XL5 klystron from industry (CPI)
New modulator from Scandinova (after call for tender)
Current measurement transformer now on high side of secondary windings 
Still unfriendly GUI 
Again late delivery of the tube meant that factory acceptance test was done with resistive load and also initial site acceptance tests.
First klystron tests were performed using SLAC XL5 (limited peak voltage and pulse width)

12. 50 Hz rep rate klystron operates in diode ok
1150 V IGBT ,
50 Hz rep rate klystron operates in diode ok
Pulse width reference 0.5µS
Vkly 348kV
Pulse width reference 0.6µS
Vkly 348kV
Pulse width reference 0.8µS
Vkly 343kV
Still some modulator issues to be resolved.
To limit overshoot IGBTS were delay triggered for this mode of operation.
Charging voltage had to be increased to compensate.
Confident that klystron can run at this level until new CPI one arrives for crab cavity tests.

13. CPI Klystron Installed

14. 233kV on klystron Instability starts
Klystron voltage approximately 220 kV , dark blue trace and current
Pulse width 1.5 μs
233kV on klystron
Instability starts

15. V klystron 236 kV
V klystron 242 kV
V klystron 250 kV
V klystron 262 kV

16. 1000 V reference
Klystron Current
Klystron Voltage
Tank Probe
1 MHz bandwidth
zoom
700 MHz gun oscillation

17. Example of the pulse with ~ 5 MW RF peak power expected:
CPI2-2 Klystron in Xbox2
First diode tests (no RF) of CPI#2 tube showed that 0.7 GHz gun oscillations starts at about 240 kV and even generate RF power from the input cavity:
12 GHz pulse from the input cavity
Fortunately, going higher in voltage, the instability zone moves towards the rise/fall time, so the plat top can be used now.
Example of the pulse with ~ 5 MW RF peak power expected:

18. Operation of XBOX2 modulator at full power
Decided that even with oscillations tube can be run to finish crab cavity tests
Modulator issues
IGBT switch and charging supply interlocks and breakages at nominal klystron voltage
initially caused by delay triggering but also a design modification was needed on the capacitor charging power supplies.
Now operating correctly
Problem with ΔT measurement on body. Very sensitive with instantaneous changes to the input temperature from cooling station. Input became a lot greater or lower than output measurement. Resolved by using temperature controlled chiller on body

19. Gun Oscillation problem with klystron not modulator
No oscillations in XBOX1 with first CPI tube
Only difference is the geometry of high voltage tank housing the pulse transformer and klystron
Proposition for Gun Oscillation Suppression in CERN XBOX-2 Oil Tank from CPI
Method 1: replace existing parts on base of gun with steel parts to provide magnetic loss.
Method 2: in combination with method 1, add ferrite tiles to walls and bottom of oil tank.
Method 3: in combination with method 1, add plates to mimic the internal shape of the XBOX-1 oil tank. Plates are made of steel to additionally provide magnetic loss.
Pieces arrived at CERN, modification and tests to start 29th June

20. XBOX 3 Klystron specification 6 MW , 6μS, 400 Hz ( 4 required)
Decided to do call for tender on a turnkey solution with both klystron and modulator under the responsibility of klystron manufacturer.
Toshiba won contract, with the modulator subcontracted to Scandinova.
1st prototype tube successfully tested at Toshiba and turnkey solution delivered to CERN with the modulator already tuned for pulse shape.
Excellent flat top and pulse to pulse stability

21. XBOX3a pulse shape
Klystron current 92 A
Klystron voltage 152 kV

22. Phase measurement 0.19 degrees fluctuation -
Preliminary RF phase measurements last Thursday
Klystron
LLRF phase
HPRF phase
PXI measurement + -
Difference measured to eliminate TWT phase fluctuations
RMS measurement and mean value over 2 minutes
0.19 degrees fluctuation

23. XBOX 3 New modulator control system (SCANDICAT)
Great improvement on older models 
Nice GUI
User friendly
Time and date stamps of all events memorized
Fixed hour counter for all high voltage time
Interlocks levels can easily be changed via GUI
Commissioning ongoing, may have some issues with klystron or modulator performance, not conclusive yet to report here

24. Some comments on Solid State modulators
Unfortunately the capacitor charging power supplies and switch unit design has evolved and there is no backward compatibility
i.e. extra spares needed for each type of modulator
Flat top on voltage is far superior to line type modulators
Instantaneous adjustment of voltage pulse width is a big advantage
Line type modulators give the possibility of adjusting PFN to compensate for overshoot at lower voltages, solid state does not
But due to the very good flat top and pulse to pulse stability working in the non saturated regime of the klystron is possible

25. Other news
A new solid state klystron driver prototype has just been delvered to CERN
330 W , 6 μS, 40 Hz
Performance will be evaluated in the next few weeks.
CPI klystron number 3 under test.
Once delivered XL5 klystron will be returned to SLAC for repair.
XBOX1 modulator has had two faulty solenoid supplies in the last week (over temperature) still to be evaluated
The supplies are out of their 5 year manufacturer warranty.

26. Conclusions
Progress in high power solid state modulators is still evolving.
Scandinova and hopefully other manufacturers, will listen to our positive and negative feedback and implement the changes required by our community in order to improve the already encouraging performance and reliability of the modulators.

27. Thank you for your attention