1. LLRF 15 Daresbury Andrew Moss ASTeC, STFC Daresbury Laboratory

2.  Description of why and how we started  Early test results and methods  Specific LLRF installations on ALICE and VELA  EMMA commercial LLRF  CLARA LLRF  Recent developments Contents

3.  The original ALICE energy recovery linac used FZR Rossendorf analog LLRF cards based on Op Amps and mixers  Very fast operation and still in use on the ALICE booster  Limited by being inside the accelerator hall with no possibility to adjust parameters quickly- changes to capacitor and resistor values to tune loop response  During 2009 the group took part on an USPAS course on LLRF and experienced Digital LLRF working as part of the hands on work  We recognised the need for the cavity control to be able to adapt quickly, to use feedback and feed-forward to compensate for beam loading and to be able to remove/measure environmental effects  We already had a number of digital boards including the LLRF4 and so work concentrated on this on our return  All of our work so far has used connectorised components – off board mixers  We very quickly found that temperature changes were to be avoided at all costs to get a stable system Our beginning

4. Initial tests Slow phase loop to lock FPGA to master oscillator 1.3GHz During 2011 tests made using developed software and LLRF4 board on ALICE Buncher NC 1.3GHz cavity Analog card response to beam in buncher cavity – notice phase error is not eliminated Digital system response to beam in Buncher cavity with feedback and with feedback and feed forward 84uS beam at 40 pC Cavity probe response in comparison with MO using Hittite phase detector Hittite phase detector HMC439QS16G

5. Daresbury LLRF Conversion to/from IF of 50MHz via connectorised components Temperature regulation to improve stability EPICS interface to control LLRF4 card Open loop Calibration mode Closed loop Closed loop with ramped fill P & I constants, amplitude and phase Software programs Self excited loop Feedback/feedforward Ramped filling mode to reduce high peak power Control over overshoot and filling time for SCRF cavities

6. ALICE ERL with FEL Installation of digital LLRF on Buncher cavity in 2012 – 1.3GHz 2kW solid state amplifier Phase rms error : 0.024 degrees; amplitude rms error: 0.05%. Measured by a Hittite phase detector and Boonton power meter Test during 2014 on 1.3GHz SCRF booster showing ramped fill mode to reduce waveguide arc trips Stability was comparable with the figures above for operation with the Buncher RF system is at 1.3GHz and is pulsed at 2.5mSec 10Hz Peak power during beam loading Booster and Linac are Tesla type SCRF structures FEL

7.  During January 2015 digital LLRF was installed on both Linac cavities  LLRF systems share a common reference and clock system  1.3GHz master oscillator used by LLRF, preforms DDC to get reference phase, frequency control of VCO clock and LO to lock FPGA to MO ALICE LINAC Linac 1 Linac 2 MO

8.  Machine began a 3 month user run in Feb 2015 with digital LLRF installed  Physicist’s find the systems more stable and using ramped filling has removed waveguide arc trips  Energy recovery display (reduction in forward power during the beam on pulse)is harder to see however and a more user friendly LLRF screen is needed  Overall the performance of the machine is better and more reliable with increased FEL performance ALICE results

9. LLRF 4 MO RF systemLaser system FEL 1.3GHz 81MHz 1.3GHz CLK Photodiode ADC channels FEL stability has been an issue on ALICE, both fast changes and slow drifts LLRF 4 card was used to diagnose problem LLRF clocked at 1.3GHz and measure 81MHz I&Q of laser drive signal and then laser photodiode FEL output level is directly linked to laser phase changes Solution was to install a phase shifter in front of the laser and drive it from the LLRF4 to correct for phase changes FEL stability improved from 5-10% to 1-2% with long term stability much improved Laser has been fixed for the next run but monitoring will continue ALICE FEL stability 81MHz laser signal

10.  S – band 2.5 cell standing wave normal conducting gun  Photo injector laser at 81MHz  10MW modulator  3uS RF pulse at 10Hz  Temperature stabilised to 0.1 degrees at 35C Versatile Electron Linear Accelerator VELA

11.  VELA master oscillator is 2998.5MHz with laser output at 81.04Mhz  LLRF4 board used with IF at 50MHz  Local oscillator at 2.948MHz controlled by LLRF4 in FM mode  Short pulse open loop operation  Calibration system for measurement of cable drifts has been built but not implemented so far VELA LLRF RF down/up conversion chassis on heated plate Testing ideas and performance limitations for future 3GHz machines Second identical LLRF system to drive 6MW Time Deflecting Cavity

12. VELA phase drift The VELA machine has always suffered from a phase drift. The LLRF is locked to the MO using the DDC process so that everything is locked together Both LLRF systems indicated that the phase reference was stable over many months Experiments were performed to locate this problem AD9914 DDS in programmable modulus mode AD8302 phase detectors Using the phase detectors we could monitor the system and found during a 24 hour period that the 81MHz driving the laser was slipping at 5 degrees per hour away from the 3GHz So the LLRF was correct that the 3GHz was stable, but everything driven by 81MHz from the MO was moving Short term solution is that the laser is now driven directly from the AD9914 DDS which is stable and jitter is no worse on the electron beam VELA MO

13. EMMA 19 NC cavities 19 forward and reflected power signals One 100kW IOT amplifier Waveguide power splitting to each cavity Waveguide phase shifter for each cavity Ability to operate off frequency -4MHz to +1MHz around 1.3 GHz synchronised to ALICE 1.3Ghz

14. EMMA Non Scaling FFAG Liberia LLRF system with master/slave units to get all RF channels monitored Global vector sum control of amplitude and phase Beam loading could be seeing in LLRF which helped us to set phase of each cavity 30 minutes required to change the frequency of all cavities using the Liberia Acceleration was achieved in the machine

15.  S band linear accelerator FEL at up to 250MeV  Aiming for 10fS stability for experiments  LLRF will be commercial product because Daresbury LLRF team is currently rebuilding and timescales mean this is the only option  Calibration systems will be needed for long term stability  Beam diagnostics will be used to stabilise the whole facility  CLARA is a test accelerator towards the next UK FEL project CLARA

16.  New LLRF person has started last month  So far he has written his own code to produce a VNA application via firmware and LLRF 4 and 4.6 boards (Spartan 6 version )  Also have IOxOS Spartan 6 boards as used at Swissfel, work will begin soon on these  Using a combination of commercial LLRF and our in house developments we intend to progress towards high quality systems for CLARA  We require a phase noise analyser and would welcome any advice on the available systems Recent developments

17.  We have show the ability to develop our own LLRF systems with a lot of help ! Thanks to Larry Doolittle and Dmitry Teytelman  Most of the what I have presented is the work of LiLi Ma and Peter Corlett over the last 5 years  We have used commercial LLRF and understand the depth of knowledge that is included to make these products and the value for money this offers  In the future we will use both solutions depending on timescales and availability of resources because LLRF takes a huge effort and we are a very small team  All of our accelerators are primarily test facilities and developing future technology is our task, our facilities are available to external companies and laboratories Conclusion