The LHC LLRF P Baudrenghien CERN BE RF reporting for

The LHC LLRF P. Baudrenghien CERN/BE/RF reporting for the LHC LLRF team Desy, March 29, 2010

Content 1. Presentation of the LHC RF 2. Beam Structure and Transient Beam Loading 3. LLRF Architecture 4. A word on Implementation 5. Results with (first) beams Desy, March 29, 2010 The LHC LLRF 2

1. Presentation of the LHC RF Desy, March 29, 2010 The LHC LLRF 3

Layout • Two independent rings • 8 RF cavities per ring all installed at point 4 • Klystrons and Cavity Controllers in a cavern ~150 m underground • LLRF for Cavity Controllers in two Faraday Cages not accessible during operation • Beam Control equipment in a surface building in SR 4 (Synchronization SPS-LHC transfer, Beam based Loops) Desy, March 29, 2010 The LHC LLRF 4

Cavities • 8 RF cavities per ring at 400. 790 MHz: – Super Conducting Standing Wave Cavities, single-cell, R/Q = 45 ohms, 6 MV/m nominal – Movable Main Coupler (20000 < QL < 180000) • 1 MV /cavity at injection with QL = 20000 • 2 MV/cavity during physics with QL = 60000 – Mechanical Tuner range = 100 k. Hz Desy, March 29, 2010 The LHC LLRF 5

2. Beam structure and transient beam loading Desy, March 29, 2010 The LHC LLRF 6

It takes 12 SPS cycles to fill one LHC ring The SPS receives 2 to 4 batches from the PS, with each batch containing 72 bunches Nominal LHC bunch pattern. Notice the 3 ms long gap for the risetime of the beam dump kicker and the ~1 ms long gaps for the risetime of the LHC injection kicker. Shorter gaps (225 ns long) are needed to accommodate for the risetime of the SPS injection kickers. There is much flexibility with the injector chain and the RF must accommodate with it. Desy, March 29, 2010 The LHC LLRF 7

Transient Beam Loading during filling • Cavity filling time • Q=60000, 0. 5 A DC -> each cavity presents a 2. 7 MW impedance at resonance -> 2. 7 MV beam induced voltage compared to 2 MV RF voltage (per cavity) The resulting phase modulation on the RF voltage is [Boussard] • • • This effect is minimized by using superconducting cavities: Test in SM 18: modulated beam current and measured beam induced Low R/Q and high voltage with QL=60000 (tf=48 ms) During filling, worst case is 50% duty cycle Ib, tgap~50 ms and we get 10 degrees phase modulation for 0. 5 A DC and 16 MV [Boussard] D. Boussard, RF Power This causes an injection phase error if injection phase is Requirements for a High Intensity Proton kept constant, and then emittance blow-up through Collider, PAC, San Francisco, 1991 filamentation Decision: Keep voltage constant over one turn Desy, March 29, 2010 The LHC LLRF 8

Modulation of collisions point • During physics phase modulation on the RF voltage causes displacement of the collision point • With the nominal scheme, the longer gap is the 3 ms long abort gap and it causes an displacement of the collision vertex by 1. 25 mm. Acceptable • So we could let transient beam loading phase-modulate the Cavity voltage during physics and we will modulate the voltage set-point accordingly in the future Desy, March 29, 2010 The LHC LLRF 9

Coupled bunch instabilities • Narrow-band resonant impedance threshold • At 7 Te. V, the impedance must be below 0. 3 MW (1. 0 e. Vs) or below 0. 9 MW (2. 5 e. Vs=nominal) • At fundamental, with QL=60000, the total cavity impedance is 21. 6 MW • Goal: Reduce effective cavity impedance by ~ 100 on resonance Narrow-band impedance threshold Rsh (solid line) and imaginary part of the broad-band impedance threshold Z/n (dotted line) during the acceleration ramp with constant 16 MV for different longitudinal emittances. Reproduced from [Shaposhnikova] E. Shaposhnikova, Longitudinal beam parameters during acceleration in the LHC, LHC project Note 242, Dec 8, 2000 Desy, March 29, 2010 The LHC LLRF 10

Goal: Keep RF voltage constant over one turn = full compensation of transient beam loading RF feedback Operational 1 -T feedback Installed but not operational With an RF feedback the minimal effective impedance Rmin and closed-loop bandwith Dw scale with loop delay T Desy, March 29, 2010 The LHC LLRF 1 -T feedfwd Postponed 11

Tuning Cavity linked to klystron via a circulator • Incident wave current (transformed at the gap): Equivalent circuit. Forward and reverse waves transformed at the cavity gap • Klystron power (superconducting case: Qe=QL) with Desy, March 29, 2010 The LHC LLRF 12 Vector diagram with zero stable phase

Half detuning • We keep RF voltage constant during beam-on and gap. • To keep power constant over one turn we must detune the cavity for the half-beam-intensity • We implemented an algorithm that automatically converges to identical klystron power during beam-on and gap segment [Baudrenghien] Klystron current with optimal (half) detuning, during beam-on segment (I 1) and during beam gaps (I 2) [Baudrenghien] P. Baudrenghien, The Tuning Algorithm of the LHC 400 MHz Superconducting Cavities, Cern-AB-2007 -011, Feb. 2007 Desy, March 29, 2010 The LHC LLRF 13

Movable Coupler • At injection: • • During physics: • 8 MV to match the 0. 8 e. Vs bunch from the SPS Low QL favorable for fast damping of momentum/phase errors For 0. 5 A DC • Optimal QL ~ 30000 • Detuning -4. 5 k. Hz • Klystron power 133 k. W Desy, March 29, 2010 • • The LHC LLRF Lifetime limited by intra-beam scattering. Emittance must be blown up to 2. 5 e. Vs 16 MV needed for the 2. 5 e. Vs emittance For 0. 5 A DC • optimal QL ~ 60000 • Detuning -2. 25 k. Hz • Klystron power 264 k. W 14

3. LLRF Architecture Desy, March 29, 2010 The LHC LLRF 15

Four sub-systems • Beam Control: Slow loops (clocked at Frev) using beam-based measurements. Controls the average energy of the beam via the RF frequency, and the phase of the average voltage (vector sum of 8 cavities). Phase loop, radial loop and synchro loop • Cavity controller: Potentially fast loops (clocked at 40 MHz bunch frequency) using cavity or waveguide measurements. Individual control of the field in each cavity, its tune and the klystron gain/phase shift • Longitudinal damper: During the sequence of injections, damps the phase and energy error (dipole oscillation) and the bucket mismatch (quadrupole) by modulating the field in the cavities. Used only at injection • RF Synchronization: Pilots the bunch into bucket transfer from SPS to LHC Desy, March 29, 2010 The LHC LLRF 16

Layout Desy, March 29, 2010 The LHC LLRF 17

Cavity Controller Loops Fast loops (clocked at 40 MHz bunch frequency) using cavity or waveguide measurements • RF Feedback Loop: Reduces the cavity impedance at the fundamental (by 20 linear for Q=20000, by 180 at Q = 180000). Precision of RF voltage, transient beam loading and longitudinal stability • Klystron Polar Loop: Compensates for the klystron gain/phase changes. (HT drifts and ripples). • 1 -T Feedback: Adds factor 10 reduction on the revolution frequency side-bands. (Transient beam loading + longitudinal stability) • Tuner Loop: Minimizes klystron current. (Half detuning) • Set Point: Customizes the voltage for each bunch. (Each bunch slightly displaced with respect to a constant spacing) Desy, March 29, 2010 The LHC LLRF 18

Desy, March 29, 2010 The LHC LLRF 19

RF feedback closed loop • Closed-loop BW ~ 700 k. Hz (2 -sided), independent of QL, fixed by loop delay (650 ns) • Impedance reduced to 45 k. W per cavity, independent on QL • Feedback gain must scale as sqrt[QL] RF feedback alone reduces the total impedance to 8 x 45 k. W = 270 k. W. Comparing to the Narrowband impedance instability threshold, we are marginally stable for 1 e. Vs stable with margin of 3 for 2. 5 e. Vs 1 -T feedback will add further impedance reduction by 10 linear Desy, March 29, 2010 The LHC LLRF 20

Step response Vcav Q Vcav I Ig 400 k. V in 10 ms, 2 ms per div 100 k. V in 3 ms, 2 ms per div • • 1 MV in I, step in Q Q=60000, 1 MV asks for ~ 40 k. W Observe Vcav I and Q plus klystron drive @ 400 MHz Linear regime: 70 k. V in 1 ms Phase modulation of the voltage will be used for longitudinal damping (dipole mode) Desy, March 29, 2010 The LHC LLRF 21

Klystron Polar Loop Compensation for HV ripples (phase) Phase Noise Phase compensation Loop open. Phase noise Ig-Ref: Mainly 100 Hz and 600 Hz due to HV ripples. Calib 10 m. V/dg @ 400 MHz. ~3. 5 dg pkpk (10 m. V/div, 5 ms /div) Loop closed. Red trace = phase noise Ig-Ref. Calib 10 m. V/dg @ 400 MHz. ~0. 2 dg pkpk (2 m. V/div, 5 ms /div). Blue trace = phase compensation. • Measured with HP 8508 Vector voltmeter • Calibration: 10 m. V/dg @ 400 MHz Desy, March 29, 2010 The LHC LLRF 22

Klystron phase ripples PSD in d. BV 2/Hz, 10 d. B/div, DC to 1 k. Hz. Phase noise Ig-Ref: Bright trace = klystron loop On, background trace = loop off. Measured reduction 30 d. B @ 600 Hz • • • PSD in d. BV 2/Hz, 10 d. B/div, DC to 200 Hz. Phase noise Ig-Ref: Bright trace= klystron loop On, background trace = loop off. Measured reduction 50 d. B @ 100 Hz Measured with HP 8508 Vector voltmeter plus HP 3562 A spectrum analyzer Calibration: 10 m. V/dg @ 400 MHz Observe 30 d. B reduction @ 600 Hz and 50 d. B reduction @ 100 Hz Desy, March 29, 2010 The LHC LLRF 23

Cavity Field phase noise • • • Coast at 7 Te. V/c with 16 MV, 2. 5 e. Vs (fs 0=23 Hz). – 1 ps rms white noise just compensates synchrotron radiation damping [Tuckmantel] Crossing the 50 Hz fs during ramp Cut-off of – during ~ 1 min, 50 Hz falls inside fs band. Dangerous RF fdbk Coast at 450 Ge. V/c with 8 MV, 0. 7 e. Vs (fs 0=63 Hz). closed loop BW – 1 ps rms white noise now gives 0. 1 % loss after 1 hour. Notch at the synchrotron freq due to phase loop 1. 18 ps rms 10 Hz-10 MHz 0. 818 ps rms 10 Hz-10 MHz Power Spectral Density (PSD) of the Cavity Sum phase noise measured with circulating single bunch pilot at injection 450 Ge. V (left) and at 3. 5 Te. V. Notice the action of the phase loop at the synchrotron frequency , ~ 65 Hz (left) with 8 MV and 28 Hz (rigth) with 12 MV. The 350 k. Hz single-sided BW of the RF feedback is clearly visible on the spectra. Ring 2. March 23, 2010 [Tuckmantel] J. Tuckmantel, Simulation of LHCbunches under influence of 50 Hz multiple lines on the cavity field, LHC project Note 404, June 2007 Desy, March 29, 2010 The LHC LLRF 24

Beam Control Loops Slow loops (clocked at Frev) using beam-based measurements • • • A Voltage Controlled Crystal Oscillator (VCXO) generates the reference RF (one per beam) The Frequency Program generates an RF that is set to the injection frequency on the injection plateau and then follows a function through the acceleration ramp (plus a real-time correction coming from the global orbit feedback. Not operational yet) The synchronization loop locks the VCXO output onto the Frequency Program RF A strong phase loop compares the beam phase (Wide-Band PU) with the Cavity Sum (RF Sum of the Antenna signals of the 8 cavities). It acts back on the input of the VCXO We do not use a radial loop during normal operation but it may be used for commissioning the start of the ramp Desy, March 29, 2010 The LHC LLRF 25

LHC LLRF Beam Control A function sets the RF frequency on the injection plateau and through the ramp Synchro Loop locks VCXO onto Freq Prgm Phase Loop locks Vt onto PU The VCXO generates the RF sent to the Cavity Controllers 26

4. A word on implementation Desy, March 29, 2010 The LHC LLRF 27

The cavity controller: 1 rack, 2 VME crates, 15 VME modules per cavity • VME crates • Custom designed VME cards with RF front-end and digital part (FPGA) • Use of Cadence for schematics and layout • Standardized Design Flow for the FPGA: Visual Elite, Synplify, Cadence [Molendijk] • Built in Observation and Post. Mortem memories for diagnostic • Built in excitation buffers for remote measurement of transfer functions (or step responses) [Molendijk] J. Molendijk, Digital Design of the LHC Low Level RF: The Tuning System for the Superconducting Cavities, LHC Project Report 908, EPAC 06, June 2006 Desy, March 29, 2010 The LHC LLRF 28

Example: The Tuner module RF front-end: 400 MHz signals are mixed with 380 MHz LO. 20 MHz IF is then sampled at 80 MHz FPGA clocked at beam-synchronous 80 MHz: Digital I/Q demodulator, Decimating CIC filters, cross-product to measure cavity phase shift, Min/Max over one turn for half-detuning Desy, March 29, 2010 The LHC LLRF DSP for slow processing (9. 766 k. Hz rate) 29

Volume • 20 racks in the UX 45 cavern plus 15 racks in SR 4 (surface building) • ~ 50 special LLRF VME crates plus 5 standard VME crates • ~ 500 NIM/VME cards of 36 different makes • Series of ~900 modules Desy, March 29, 2010 The LHC LLRF 30

Tools for Setting-up the Cavity Controllers • Automated setting-up of the Cavity Controller Loops using Matlab tools developed at SLAC (LARP collaboration) Klystron Power Sweep Use of the excitation buffers and observation memories built in the VME cards Desy, March 29, 2010 Closed loop setting-up Alignment of Digital and Analog feedback The LHC LLRF 31

5. Results with (first) beams Desy, March 29, 2010 The LHC LLRF 32

RF Monitoring application for the CCR 16 Klystrons One Power Converter feeds 4 klystrons Cavity Controller LLRF Monitoring the cogging…actually wrong at the time… Beam Control Loops Desy, March 29, 2010 The LHC LLRF 33

Capture and Phase Loop transient (2009) Bunch avg phase at inj with phase loop on MR with phase loop OFF and intentional phase error. 4. 8 MV. ~48 Hz synchrotron frequency [Bohl] T. Bohl, LHC commissioning 2009, Note-2009 -39 Desy, March 29, 2010 The LHC LLRF Color coded MR with phase loop ON and intentional phase error. Voltage 34 mismatched.

Capture transients Phase loop is fast: “jumps” the field on the beam at injection Synchro loop is slow. No reaction in first 100 turns. Slope gives frequency (energy) error at injection -30 deg in 60 turns -> -15 Hz @ 400 MHz Inj Phase Error 35 deg/45 deg Cavity field “jumps” on the beam in ~ 10 turns Synchro loop brings RF (and beam) back to Freq Prgm reference Phase Loop Error: Beam PU-Cav Sum Synchro Loop Error: VCXO-Freq Prgm -15 deg in 80 turns > -6 Hz @ 400 MHz Desy, March 29, 2010 The LHC LLRF Very slow (seconds) time constant. Boosts DC gain to minimize thermal drifts 35

Check collimator hierarchy by moving beam (500 Hz@400 MHz = 5 mm) before start ramp-> losses Inject Beam 1 Inject Beam 2 Desy, March 29, 2010 Ramping No visible capture loss at start ramp Bunch intensity (Fast BCT) and main bending current during a ramp. 1 pilot per beam The LHC LLRF 36

Bunch length (4 sigma) At 450 Ge. V bunch length increases by ~ 30 ps/h At 3. 5 Te. V bunch length increases by ~ 6 ps/h During ramp bunch length should be multiplied by 0. 6. Observed factor only 0. 75 TO BE STUDIED Single bunch pilot in both rings, ~ 0. 4 -0. 5 e. Vs. Constant 12 MV during acceleration ramp. Desy, March 29, 2010 The LHC LLRF 37

Bunch profile Lifetime ~ 1 ns (4 sigma) at 450 Ge. V Single bunch pilot circulating at 450 Ge. V. Wideband PU (3 GHz). Ring 2. Top: bunch profile 1 ns/div. Bottom: 10 ms/div. Bunch in bucket Ten pilots circulating at 450 Ge. V in each ring. Top two traces: 1 should display in the centre. We beam current. Bottom trace: Intensity lifetime. After injection were 1 bucket off…. transients, lifetime is in the 100 hours range Desy, March 29, 2010 The LHC LLRF 38

Coupler gymnastics observed in Cav 1 B 1 current Klystron Pwr 90 k. W Injection 1 MV/cavity Q=30 k Coupler 30 k->60 k See next slide Klystron Pwr 95 k. W 1. 5 MV/cavity Q=60 k Capture with 1 MV/cav and Q=30 k. Before ramping: First move Q to 60 k, then raise voltage to 1. 5 MV/cav. 1 pilot Desy, March 29, 2010 The LHC LLRF 39

Coupler gymnastics - enlargement Red=coupler pos. Beige=klystron power Green=cavity field Q vs Pos depends on circulator S 22. Being studied When coupler moves, parameters for the tuner loop (phase offset) and for the RF fdbk (phase and gain) are continuously adjusted following individual calibration curves The couplers are moved with circulating beam, RF power ONnand all LLRF loops On (Tuner, Klystron loop, RF fdbk) Desy, March 29, 2010 Tuner phase offset calib RF fdbk Gain and Phase calibration The LHC LLRF 40

Thank you for your attention Acknowledgments: The original designers: D. Boussard, T. Linnecar The LHC RF Project Leader: E. Ciapala Longitudinal dynamics: E. Shaposhnikova, J. Tuckmantel The LLRF and RF Control team: M. E. Angoletta, L. Arnaudon, P. Baudrenghien, T. Bohl, A. Butterworth, F. Dubouchet, J. Ferreira-Bento, D. Glenat, G. Hagmann, W. Hofle, S. Kouzue, P. Maesen, J. Molendijk, J. Noirjean, A. Pashnin, V. Rossi, J. Sanchez-Quesada, M. Shokker, D. Stellfeld, D. Valuch, U. Wehrle, F. Weierud

RF Feedback Open Loop • 20 d. B gain increase at lower frequencies • Optimal Open-Loop gain = 20 (Q=20000), 120 linear (Q=180000) • Notches at +-4. 3 MHz offset. Compensates for the klystron bunching cavity resonance. • 20 d. B (10 linear) gain increase in 4 k. Hz band around centre frequency for precision Full Fdbk OL response 10 MHz band Dig Fdbk OL response 10 k. Hz band Desy, March 29, 2010 The LHC LLRF 42

Desy, March 29, 2010 The LHC LLRF 43