1. Experience with the commissioning and operation of the new CERN Digital LLRF family M. E. Angoletta, A. Findlay, M. Jaussi, J. Molendijk, J. Sanchez Quesada, J. Simonin (CERN), C. Schmitzer‎ (MedAustron) LLRF Workshop, Shanghai, 3-6 November 2015

2. Outline 1.DLLRF overview 2.PSB LLRF commissioning & operation 3.PSB Finemet ® tests and planning 4.MedAustron and medical machines 5.Conclusions & outlook Additional slides

3. DLLRF overview  It is the 2 nd generation Digital LLRF family for CERN’s small machines (Meyrin site) →  Developed within the framework of the LHC injector upgrade project (→ symbol), will apply to non LHC injectors, too.  Hardware/firmware mostly described by J. Molendijk @LLRF13 (Lake Tahoe). →  In 2014 it was successfully deployed at CERN’s PSB (4 rings) and on the medical machine MedAustron → next sections. →  Used by RF group for LLRF + diagnostics (future ELENA machine). Adopted by CERN Beam Instrumentation (BI) group for orbit, Schottky & intensity measurement systems. (→ additional slides.) M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 1

4. H /w family: roadmap to other CERN machines WhenWhatGroup 2015 LEIR DLLRF upgradeRF Discussions about possible PS LLRF upgrade.RF 2016 AD orbit systemBI ELENA LLRF + long. diagnostics from magnetic PUs (baseline) RF ELENA orbit systemBI ELENA long. diagnostics from distributed TPU (study)BI LEIR Schottky systemBI 2017 (?)LEIR orbit (if approved)BI >= 2017AD LLRF + long. diagnostics (when manpower allows)RF ????TSR @ Isolde LLRF (it might not take place)RF >= LS2PSB operation with Finemet ® HLRFRF M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 2

5. Main characteristics of current implementation  Direct RF sampling, baseband I,Q processing  Double-tagged clock for system-wide phase synchronisation. Can be distributed over fiber.  Modular:  System clocked proportionally to f REV (tracking) with “gear change” (i.e. change of MDDS harmonic) to cope with large frequency swing. M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 3 NB: this might change as we’re considering going to a fixed clocking scheme (future improvement when manpower allows).  VXS crate, fast intercomms between boards  3 types of daughtercards (functionally: MDDS, DDC, SDDS) in FMC format. Two can be hosted on a FMC-carrier board.  2 FPGAs (Main + FMC) + 1 DSP hosted on FMC carrier board.  Several boards typically interacts to make a LLRF system.

6. Outline 1.DLLRF overview 2.PSB LLRF commissioning & operation 3.PSB Finemet tests and planning 4.MedAustron and medical machines 5.Conclusions & outlook Additional slides

7. PSB overview: the machine  Proton accelerator, LHC injector + experimental areas. 4 rings superimposed. Cycle duration: 1.2 s  Varied RF gymnastics & cavities combinations  Varied intensities (10 9 to 10 13 per ring)  Upgrades after Long Shutdown 2 (LS2): Current PSB cycle parameters. PSB machine schematic view  Injection from Linac4 (T E = 160 MeV, f REV,I = 1 MHz)  Extraction @ 2GeV (f REV,E ~ 1.82 MHz) M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 4

8. PSB overview: beams and DLLRF  One DLLRF system per ring (independent). → Additional slides for LLRF operation during typical PSB cycle  DLLRF sampling period = 10 µs (baseband processing).  “Ring 0” LLRF: additional DLLRF system connected to PSB Ring 4.  Controls Ring 4 beam in PPM with Ring 4 DLLRF.  Used to validate new implementations and for machine studies.  Used for Finemet ® R&D in 2014-2015 (→ see part 3 of this talk). M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 5

9. PSB HLRF ferrite layout and characteristics DLLRF R = 25 m C04 (all rings) C16 (all rings) C02 (R2+R4) C02 (R1+R3) M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 6

10. PSB LLRF picture Ring 0 Ring 4 Ring 3 Ring 2 Ring 1 M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 7

11. PSB LLRF operational layout (one ring) M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 8

12. PSB DLLRF commissioning & operation  PSB restarted after 15 months shutdown on June 2 nd 2014 with DLLRF as THE operational LLRF.  LLRF team worked 7/7 for 3 weeks to commission the system.  Beam captured, accelerated (phase + radial loop) & synchronised on Ring 3 by Thu June 5 th. All rings setup (h =1, basic integration in controls infrastructure) by Sun June 8 th.  More beams and features commissioned during the summer (ex: splitting in August 2014). M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 9

13. PSB DLLRF commissioning & operation– cont’d  Lack of manpower: integration in controls infrastructure lag behind actual implementation in low-level part. “the beam never lies”  Sometimes malfunctioning after system firmware update (“the beam never lies”). If problem cannot be solved quickly, we revert to previous version (< 30 min operation).  Operators can setup the LLRF & optimise beams: overall very satisfied.  System flexibility triggered new functional requirements (ex: C16 blowup) that will be deployed soon. M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 10  DLLRF reliability:  DLLRF rings 1-4 operate successfully 24/7 since June 2014 (excluded Christmas stop, 15 Dec 2014 – 2 Feb 2015).  July 2015: problem on ring 4 DLLRF (giga-bit links h/w fault on a board). Degradation over time, difficult to point clearly to LLRF as culprit. 2 days downtime for DLLRF ring 4, could be replaced by DLLRF Ring 0 → no downtime to machine due to LLRF.  Short-term malfunctioning due to loose/broken cables (bad quality in stores).

14. Performance with respect to previous LLRF Fully PPM, all settings can be archived & copied between rings/users. Shot to shot reproducibility greatly improved Drifts in parameters requiring adjustment all but eliminated (was daily intervention with previous LLRF). Synchro at extraction much more reproducible and smooth. A huge number of signals are available on CERN standard signal observation system (virtual scopes and samplers). Beam blowup with C16 more flexible & powerful (→ additional slides), less voltage needed. A big investment but it has paid off. Training:  LLRF stand-by service people still to be fully trained on new system  Operators: basic training done, enough for simple setting up, signal observation and debugging. M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 11

15. Future work on the PSB operational system  Integration with the new interlock system  Rings synchronisation at injection M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 12 Short-term (2016) Medium term (depends on manpower)  New requirements to be developed & deployed after the Xmas technical stop.  New interface with the Btrain (white rabbit-based) to be deployed and tested in 2016.  Preparation for Linac4 injection:  Improved FPGA-DSP data transfer using DMA.  Clocking scheme implemented with fixed sampling clock Long long term: 3rd generation LLRF. Ex: FPGA with embedded processor (Xilinx ZYNQ SoC FPGA family to replace Main FPGA + DSP).

16. Beam results: synchronisation @extraction  Bunch to bucket. transfer to PS machine  Several algorithms available & switchable in PPM fashion.  Typically: frequency steering to beating frequency (100 Hz) then synchro phase loop closed.  Excellent synchro quality. M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 13 Beam extracted Beam-to-extraction reference phase, after reaching the beating frequency. Green trace: new algorithm White trace: previous algorithm Synchro phase loop closed Waterfall plot of the last 35 ms before extraction (new algorithm implementation). The synchro phase loop closure causes no disturbance to the beam.

17. Beam examples: hollow bunches (studies) Hollow bunch in the PSB towards the end of acceleration. Picture & MDs from A. Oeftiger & S. Hancock  Hollow (flat-topped) bunches generated in the PSB to gain space charge margin by reducing the bunch peak current.  Important for the PS (i.e. next machine) which has transverse space charge limitations at injection for the beams foreseen for HL-LHC.  Generated via beam phase loop modulation to shake particles out of centre. M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 14  C16 blowup to smooth-up distribution. http://tomograp.web.cern.ch

18. Outline 1.DLLRF overview 2.PSB LLRF commissioning & operation 3.PSB Finemet ® tests and planning 4.MedAustron and medical machines 5.Conclusions & outlook Additional slides

19. Finemet ® studies in PSB  Reliability run  Operation as h=1, h=2, h=1+2. Beam blowup as C16 (h=10).  Simulations to study longitudinal emittance and future beams. Extensive MD plan carried out in 2014-15  Aim: validate replacement of ferrite-based HLRF with Finemet ® -based.  10 Finemet ® cells (7 kV) installed in PSB Ring 4 → J. Molendijk’s talk “Digital Receiver & Modulator Architecture for Multi-harmonic RF Finemet Operation”, this workshop.  Simulations showed Finemet ® validity also for higher-intensity beams.  International review (September 2015). Outcome: after LS2 Finemet ® will replace ALL PSB ferrite-based systems (C02/C04/ C16). M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 15 NB: no way back (ferrite cavities will be removed)

20. Future PSB HLRF layout  Finemet will be in sections now allocated to C02 and C04.  Multi-harmonic operation until ~ h = 18  Total voltage available: 24 kV Finemet ® M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 16

21. DLLRF - Finemet ® operation in 2014-15  Feedback loops (I/Q coordinates) in LLRF to compensate beam loading.  Currently operational harmonics h=1,2,3,5. Four additional harmonics under commissioning.  h=1 & h=2 controlled with voltage programs. Phase also controlled by phase function and rotations.  All other harmonics servoed to 0 V.  Operation of Finemet ® as C16 (beam blowup) validated as well.  Protection against HLRF overdrive.  DLLRF controls gap relay in PPM. Gap open Gap closed Intensity [E10] B field [G]  Ring 0 DLLRF controls Finemet HLRF in PPM with normal operation. → additional slides for Ring0 DLLRF layout. M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 17

22. Operation experience  DLLRF ring 0 operated ISOLDE beam reliability run (July 2015 onwards), LHC-type beams & several MD cycles.  Flexible and powerful system, integrated in controls infrastructure.  LEIR (since 2005) and MedAustron experience with Finemet ® control helped the development. Also, useful previous tests (2012) with LEIR-style hardware to control Finemet ® in the PSB.  Operators can switch between Ring 4 DLLRF (ferrite HLRF only) and Ring 0 DLLRF (ferrite + Finemet ® ) autonomously. They can also switch ON/OFF Finemet HLRF.  Beam performances (beam intensity & characteristics) similar to what is obtained with ferrite cavities. → Additional slides for beam and servoloop results. M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 18

23. Results: beam loading suppression Servoloops OFF. 920E10 protons extracted.Servoloops ON. 960E10 protons extracted.  Finemet ® as h=1 with/without servoloops @harmonics 1,2,3,5 for ISOLDE.  h=1 servoed to 7.2 kV; other harmonics servoed to 0.  > 900 E10 protons accelerated in both cases → strong beam loading. Vertical scale for h=2,3,5: 10 V/div Vertical scale for h=2,3,5: 200 V/div B field M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 19

24. Servoloop future improvements  Several improvements already planned  Servoloops implemented in FPGA to obtain lower group delay + larger BW (target BW = 80 kHz).  Additional harmonics. The final solution will use multi-harmonic LO sources scheme → J. Molendijk’s talk “Digital Receiver and Modulator Architecture for Multi-harmonic RF Finemet Operation”, this workshop  Feedforward compensation table of Finemet ® transfer function to increase phase margin (hence available gain).  Now we’re running with only four, non-optimised cavity servoloops (8 kHz single-sided BW, ~36 dB loop gain) …and still performance with beam is satisfactory ! 960E10 protons accelerated with Finemet as h=1 Intensity [E10] Bfield [G] M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 20

25. Outline 1.DLLRF overview 2.PSB LLRF commissioning & operation 3.PSB Finemet tests and planning 4.MedAustron and medical machines 5.Conclusions & outlook Additional slides

26. MedAustron (MA): the synchrotron  Ion-therapy & research centre in Wiener-Neustadt (Austria)  Proton & Carbon ion therapy, clinical + non-clinical research  Treatment of first patient expected in 2016. → Additional slides for currently achieved milestones.  Protons: ≤ 1 · 10 10 / pulse  Carbon ions : ≤ 4 · 10 8 / pulse  Rep. rate: 0.5 s)  Synchrotron circumference = 77.64 m MedAustron synchrotron parameters. M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 21

27. M. E. Angoletta “Digital LLRF: achievements & LS1 plans” LIU 2013 event 22  Collaboration with CERN on both HLRF (Finemet ® ) and LLRF.  MA LLRF:  CERN experts have contributed to the definition of the LLRF functional specs and controls integration.  Common development: MedAustron staff contributed to the system development and hardware tests.  2008 – 2013: (some) MA staff located @CERN (offices + labs).  Same h/w and similar firmware + s/w as for PSB DLLRF.  → See spare slides for MA LLRF layout.  H/w modules designed specifically by CERN specialists for MA LLRF: Btrain + Tune Kicker interfacing.  Collaboration now completed. CERN - MedAustron collaboration

28. MA LLRF validation & commissioning M. E. Angoletta “Digital LLRF: achievements & LS1 plans” LIU 2013 event 23  MA control system targeted at operation, not commissioning: some parameters were not meant to be changed.  Btrain system did not work, hence pre-calculated B-train played with the LLRF embedded function generator was used.  Rapid turnover of MA staff (especially controls integration specialists) slowed down system integration. Knowledge transferred is then lost!  Not yet finalised full system documentation from CERN (manpower limitation).  Other technical “teething troubles” …  Initial & partial system validation done on the MedAustron Finemet ® installation @CERN.  LLRF team travelled to Austria several times for in-situ intervention.  Remote sessions carried out from CERN on MA LLRF system, with the collaboration and agreement of MA staff.  Initial commissioning @MA slow because:

29. MA LLRF validation & commissioning – cont’d M. E. Angoletta “Digital LLRF: achievements & LS1 plans” LIU 2013 event 24  Processing share between DSP and FPGA very useful, especially for future system upgrade (if required)  Many features developed for the PSB or for lab development very useful for MA, too. For instance:  Btrain generation from FPGA algorithm OR from embedded function.  Fixed frequency operation before moving to standard Btrain-driven frequency program.  FPGA: comms, function generation, housekeeping. It does not change once finalised and debugged.  DSP (C language): beam & cavity loops. Can be customised by MA staff.  For some beam results → additional slides

30. Medical machines & CERN Knowledge Transfer (KT)  CERN Knowledge Transfer: CERN’s technical expertise & innovative technologies are available for scientific/commercial purposes through various technology transfer opportunities → http://knowledgetransfer.web.cern.ch/http://knowledgetransfer.web.cern.ch/  2004-2010: collaboration with CNAO (Centro Nazionale for Adronterapia, Pavia, Italy)  2010-2015: collaboration with MA (design/development/commissioning of MA LLRF) M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 25  CERN can provide the existing LLRF technology that was developed in common with MA, independently of MedAustron.  Future medical machines very much interested in Finemet HLRF.  LLRF needed, too!  Experience in LLRF for medical accelerators @CERN RF group :  Finemet ® -based HLRF companion (but could work with other HLRF, too).  In 2016 CERN RF group will apply for a KT project funding to develop a simplified & optimised DLLRF for medical accelerators.

31. Conclusions and outlook  A new DLLRF family for small machines developed @CERN.  Very flexible and powerful, will be used also for diagnostics by RF+BI.  DLLRF successfully commissioned in 4 rings of PSB. M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 26  DLLRF essential to operate PSB Finemet ® HLRF. Successful MDs have lead to decision of replacing all ferrite cavities in the PSB with Finemet ® ones.  DLLRF successfully commissioned in medical synchrotron MedAustron.  Experience with medical machines within CERN RF group.  CERN will make available to medical machines existing DLLRF and possibly (depending on KT funding) a new simplified&optimised version. Very aggressive planning for the next years + limited manpowerBUT great enthusiasm, interest for new developments & team spirit will grant successful deployment(s).

32. Additional slides M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15

33. PSB-style DLLRF: not only LLRF applications!  Evolution of LEIR DLLRF (1 st generation), deployed in 2005.  H/w family aimed at LLRF applications for Meyrin synchrotrons  Project started in 2009.  Includes LHC experience.  PSB-style DLLRF is 2 nd generation DLLRF system.  Modular & flexible system: initially aimed at LLRF only, it has been adopted by BI for diagnostics applications. Ex: AD orbit, to be deployed in early 2016. M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15

34. ELENA: the (future) decelerator ELENA ring parametersELENA location within the AD ring AD ring ELENA ring ELENA cycle – qualitative view  Further decelerates AD pbars.  To be constructed in AD hall.  Ring commissioning: 2016.  Connection to AD + beam physics: 2017. M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15

35. ELENA: the (future) decelerator ELENA layout  RF workpackage includes:  LLRF: new DLLRF family. Frequency program, beam phase/radial loops. Injection & extraction synchronisation loops. Cavity servoloop. Multi-harmonic system.  HLRF: Finemet based, max Vp = 500 V. Operational: 25 mV to 100 V.  Longitudinal diagnostics (ultra-low-noise beam transformer + processing) LLRF Finemet cavity Ring beam transformer (HF + LF) Extraction line beam transformers (LF) M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15

36. ELENA: LLRF + long. diag. system (baseline) M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15

37. Longitudinal bunched processing  Intensity from Fourier analysis of RF harmonics (AD-style method). Bunch-shape dependence. New bunched-beam digital signal processing.  New method will integrate over bunch shape & subtract the baseline. RF: ELENA longitudinal diagnostics processing Bunched beam Bunched beam: intensity Debunched beam Debunched beam: intensity, dp/p and from Schottky signals analysis Debunched beam processing  Windowing & FFT averaging. M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15

38. BI: ELENA orbit + long. diag. system (study) Beam orbit and distributed electrostatic PU system for longitudinal parameters measurement.  Innovative idea: add Σ signal from orbit transverse PUs to obtain a wide-BW distributed electrostatic PU.  Digital acquisition + processing with LLRF h/w.  Risk of head ampli saturation + very innovative processing: not baseline system for ELENA (… yet, but it could become such in the future). M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15

39. PSB: typical LLRF cycle operation Actionctime Beam capture (Bdot ≠0)275 Phase loop ON275.5 Radial loop ON276 Bunch splitting(if required)765 Extr. synchro loop ON780 C04 phase servoingall cycles C16 long. blowup (ffwd) ~200 ms window (1) (1) (2)(2) (3)(3) (4) (4) (5) (5) M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 2

40. PSB: C16 beam blowup M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15

41. PSB Ring 0 LLRF layout (2014-2016) Under commissioning Servoloops at h =1,2,3,5 currently operational

42. M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 Results: cavity servoloop step response Red trace: voltage reference for h=1 Yellow trace: detected voltage @h=1  Step response @fixed f REV (0.5 MHz, 1 MHz, 1.5 MHz) for various step amplitudes and PID gains.  Measured loop BW ~ 8 kHz for gain used in operation.  More servoloop results → additional slides Response to a 1 kV step in voltage reference for h=1 and fixed f REV = 1 MHz. Green trace: Finemet ® driving signal from DLLRF 2 sampling periods delay

43. M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 Results: cavity servoloop normalised response  Fixed f REV : 0.5 MHz, 1 MHz, 1.5 MHz. Servoloop harmonics: 1,2,3,5.  Network analyser (NA) excites the HLRF by sweeping the band around each considered harmonic.  Gap return input to NA and transfer function measured.  Servoloop PID settings are those used operationally. ~ -36 dB reduction in cavity impedance (system gain) 3 dB BW: ~ 16 kHz (double sided) Non-optimum phase margin

44. M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15 PSB Finemet results: Finemet as h=1+2  ISOLDE beams operationally have the maximum intensity available, hence generate the maximum beam loading in the cavities.  Voltages: 3.5 kV on both harmonics of the Finemet®, 4.5 kV on both C02 & C04 and 0.6kV on C16.  We ramp Finemet® H=2 down to zero on the extraction FT but keep 2.5kV on C04 to stop it from dropping out.  It was possible to get 950-980E10 accelerated and to get similar longitudinal parameters @extraction to the operational beam.

45. PSB Finemet results: beam loading suppression < 60 V detected with servoloops ON @ h=1,3,5. Error on Finemet ® shaping voltage < 80 VFinemet driving signal Finemet ® detected voltage  Finemet ® as h=2 with servoloops ON as used operationally for ISOLDE reliability run. Typical beam intensities ≥ 800 E10 protons. M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15

46. MedAustron main milestones (so far)  26 May ‘14: first capture  20 Oct ‘14: first extraction to IR3 @62.5 MeV  6 Nov ‘14: 10E8 protons extracted to IR3 @250 MeV  4 August 2015: phase jump extraction scheme successfully validated MedAustron synchrotron Irradiation Room 3 (IR3) MedAustron complex

47. MA LLRF system layout M. E. Angoletta “Commissioning & operation of the new CERN DLLRF family” LLRF15

48. MDDS harmonic change Zoom on phase jump before slow extraction MedAustron beam results