Holger Schlarb, DESY, 16 th DESY-MAC 10.11.2010 FLASH LLRF & Synchronization System + Roadmap On...

transcript

Holger Schlarb, DESY, 16th DESY-MAC 10.11.2010

LLRF & Synchronization System+ Roadmap

On behalf of the LLRF and

Laser based Synchronization team

Presented by H. Schlarb

Holger Schlarb, MSK, DESY, Advanced Techniques in LLRF control for XFEL, Cracow 18.04.2011 2

Outline

• LLRF upgrades and developments at FLASH- Decision to keep MO frequency at 9MHz

=> Software & Hardware will be identical to XFEL

=> Implementation of uTCA based system at FLASH 2011/2012

=> XFEL software development has basically started!

• Highlights from FLASH operation

• Overview on synchronization + some recent results

• Roadmap (Tuesday …)Roadmap (Tuesday …)

Overview on LLRF systems at FLASH

Gun ACC1 3rd ACC2 ACC3 ACC4 ACC7

A A A A

BC2BC3

Station until 2009 After upgrade 09/10 Shutdown 11/12

RF gun Simcon3.1 SimconDSP uTCA

ACC1 SimconDSP (250kHz)

SimconDSP (250kHz)

ACC39 - SimconDSP(54MHz)

ACC23 DSP(250kHz)

SimconDSP (250kHz)

ACC45 DSP/ATCA(250kHz/54MHz)

SimconDSP (250kHz)

uTCA(semi-dist.)

ACC67 DSP(250kHz)

SimconDSP (250kHz)

ACC4 ACC4

LLRF LLRF LLRF LLRF LLRF

Upgrade of LLRF system

Upgrade of all RF stations using SimconDSP controller- Gun/ACC23/ACC45/ACC67- IF=250kHz, IQ-sampling scheme- Sampling rate 81MHz (use averaging)

RF control for 3.9GHz - Probe, forward and reflected signals- New RF down converter & LO generation with

IF=54MHz, non IQ-sampling, LO = 3954MHz- Sampling rate 81MHz

10 Channel 14bit ADCs

81 MHz clock rate

8 DAC, 14 bits

2 Gigalinks

FPGA: XILINX Virtex II pro

Upgrade of all RF stations using SimconDSP controller RF control for 3.9GHz New cabling in injector racks

-MO1 & MO2 cabling completed-New rack & cabling for RF gun-New rack & cabling for ACC1/ACC39-Enclosed racks for better temperature stability-Parallel cabling for development system-Careful noise investigation and power level

adjustment of LO and RF signals

ACC1 ACC39 RF gun

MO ACC39 DWC

Upgrade of all RF stations using SimconDSP controller RF control for 3.9GHz New cabling in injector racks Upgrade & unified FPGA controller firmware

- Multiple feed forward table (main/beam loading/correction)- Multiple setpoint table (main/beam based correction)- Model based Multiple In Multiple Out (MIMO) controller- Charge correction & intra-train beam based feedback- Exception & Error handling, limiters- Error and status displays

Feed forward table architecture

Scheme of LLRF RF controller

Upgrade of all RF stations using SimconDSP controller RF control for 3.9GHz New cabling in injector racks Upgrade & unified controller firmware Unified and new control software

- New C++ architecture for front end server- LLRF library based on SysML approach- Unified naming convention- Automatic firmware downloads- Finite State Machine for automation- High level software: diagnostics, calibration…- DAQ integration- Model based learning feed forward (LFF)- Loop phase/gain correction- Piezo control for cavity detuning comp.- …

LFF off

LFF on

Phase [deg]Voltage [MV]

1st bunch

Upgrade of all RF stations using SimconDSP controller RF control for 3.9GHz New cabling in injector racks Upgrade & unified controller firmware Unified and new control software Beam signals integrated

- Charge signals- Bunch Arrival time- Pyro signal- Real time FB with matrix- Limiter on Ampl/Phase corr.- IIRF filters- Rep. rate adaption- Charge scaling of beam load

compensation table

Control software ~ 90 % completed

Most of features are commissioned

Upgrade of all RF stations using SimconDSP controller RF control for 3.9GHz New cabling in injector racks Upgrade & unified controller firmware Unified and new control software Beam signals integrated Piezo drivers

- new driver for ACC1 / ACC7- DAQ server for detuning measurements- Several piezo studies performed- Active compensation of ringing- DC voltage added for static detuning

Control software ~ 80 % completedBut many features not fully commissioned

LLRF Control Tables – software philosophy -

Controller- + Rot

FF_CORR

FF_BLC

FF-total

≤≤

Operator & FSM & LLRF expert

Setpoints: A, & Parameters: timing, …

Learning

Feed forward≤ Bunch

Pattern

SP_USER

table≤SP_CORR

table≤

Model based

FF & SP tables

Rota bc d

Field detection

Beam signals -

SP_BBF

Beam based

SP correction

- As larger impact on

cavity/coupler

the more restriction

on table/table generation

- Separation of physics

cause of effect

- Easier exception handling

Static detuning

correction

Dyn. detuning

correction

SP filling correction

Resonance filling

Slow BBF correction

VM offset corr.

DCW calibration

BLC adaptation

Process control via FSM

Slow FB loops for parameter

optimization

On-crest acceleration phase Definition and adjustment (min. energy spread/ max acceleration) Drifts of down converter + cables (~1-2deg) -> hardware changes 2011 Difference between Setpoint and Vectorsum -> software done To some extend drifts from laser arrival time + conv. -> hardware changes 2011 On-crest phases a dominated by operation set-point (support panel)

Remark:

1% ACC1 gradient change

3.5 phase difference for all

downstream modules

FLASH results: Learning Feed Forward

4 ps 0.5 ps

FLASH results: Learning Feed Forward

FLASH results: with MIMO (ACC39)

Limited by ADC bit noise

Value Repetitive error (aver. 100 macro-pulses)

Abs. Rms PKPKPulse to pulse (each time stamp)

LFF off

Arrival time measurements

Typically values

60-100fs rms from injector

60-80fs rms behind BC2

50-60fs rms exit LINAC

Pulse to pulse

about factor of 2 better

than last year

Across bunch train

dA/A~7e-4

(LFF was off

250kHz@0.5nC)

RMS timing jitter

FLASH results: Performance LLRF

shutdown shutdown

FLASH results: Gradient stability ACC1/ACC39

Holger Schlarb, MSK, DESY, Advanced Techniques in LLRF control for XFEL, Cracow 18.04.2011

FLASH results: Performance at 4.5mA operation

Flat gradient solution achieved– 4.5 mA beam

• Characterisation of solution by scanning beam current– model benchmarking

Beam Current (mA)1 2 3 4 5

Gradient Tilts vs Beam Current (ACC7)

Intended working

15 consecutive studies shifts (120hrs), and with no downtime

Time to restore 400us bunch-trains after beam-off studies: ~10mins

Energy stability with beam loading over periods of hours: ~0.02%

Individual cavity “tilts” equally stable

Energy stability over 3hrs with 4.5mA

~0.02% pk-pk

9 Feb 2011

Concept with toroid based BLC scaling worked excellent (at least up to 4.5mA)

FLASH results: Performance at 4.5mA operation

FLASH results: Beam base FB (with MIMO & LFF)

Exit of linac & out-of-loop

• Both intra-train FB on• MIMO controller• Repetitive pkpk deviation < 100fs

< 22 fs

~ ~Laser

BC2 BC3

Latency of system

Open software developments:

Next steps (till end of FLASH run)• Firmware & Software for RF gun (consistent to SRF)• Automation and permanent usage of DC/AC piezo operation • MIMO controller including BBF• Setpoint correction during filling / resonance filling using phase slopes• Forward peak power reduction• Improved error handling• Rapid VS calibration• Improved DAQ implementation and long term statistics • Error budget management & optimization of loops for LLRF parameters• LO table correction

1. Project phase (19“ modules, uTCA without backplane)

Complicated cable management - LLRF RTM backplane concept

Layout of uTCA LLRF system

TT Interne DiskussionFrank Ludwig, Tomasz Jezynski, DESY

Higher risk of signal degradation … (for main linac eventually only)

2. Project phase (19“ modules, uTCA with backplane)

Layout of uTCA LLRF system

Activities and milestones scheduled for Jan. 11 -> Jun. 11

Main components of uTCA based LLRF system:Typically 2-3 revision required for each, 3-6 month per revision

Components With whom Status ExpecteduTCA crate (EMI/PS noise/…)* indu./indu. rev1/prod. Jun/Jan11

ADC board (16bit/10Ch)* with industry revision 1 Mar. 2011

Low noise DCW* in-house/indu. start rev 1. May 2011

Vector modulator* collaboration routing Mar. 2011

AMC controller collaboration production Feb. 2011

LO generation (19’’)* collabr./in-house design May 2011

LO distribution (19’’) collabr./in-house design Mar. 2011

LO generation (RTM)* industry contacted Aug. 2011

Calibration unit (19’’)* collabr./in-house design Apr. 2011

uTCA back plane* collabr./indu. production May 2011

Piezo driver board collabr. design June 2011

* Indicates ultra high performance (<150dBc/Hz , clk ~200fs , <10fs stab., -80dB cx-talk) usually not available in industry (close collaboration / exchange important / few companies)

Software developments uTCA

Most important is the transition from SimconDSP -> uTCA!!! • Firmware & Communication protocols & Front-end server

• Middle layer server can be move to front-end CPU

• Implementation of new Timing/Clock

New software developments feasible• Down-converter calibrations (directly into performance)

• Make us of Pfor/Pref within controller (e.g. real time quench det./model)

• 30Hz operation with 9MHz tables & 96 channels (DSP???)

• AMTF software developments for routine measurements

• Upgrade of beam base feedbacks …

2013/2014• Control software for 25 RF stations XFEL

• Energy management …

uTCA based LLRF Systems & schedule2011•REGAE (May-July) 1 Crate (8)•FLASH ACC1/ACC39 (June-Sep) 1 Crate (24/12) Shutdown 12.09.•PITZ TDS (Aug-Sep) 1 Crate (8)•CMTB (Sep) 1 Crate (24)•FLASH ACC23/ACC45 (Sep-Dec) 2 Crates (48/48) DWC/ADC?

2012•Freeze final LLRF design•FLASH ACC67 1 Crate (48)•FLASH ACC45 1 Crate (48) Semi-distr.•AMTF (March) 3 Crates (24/24/24)•Final revision (Oct.)•Start mass production

2013•XFEL L0 2 Crates (36/36)

2014•XFEL L1-L3 50 Crates (96)

1) RF distribution

~f ~ 100MHz …GHz

3) Carrier is optically + detection

~f ~ 200 THz

2) Carrier is optically

f ~ GHz

4) Pulsed sourcef ~ 5 THz

Mode locked

t ft f=

Synchronization system approaches

Hybrid system for FEL facilities

Performance

Reliability

RF System

Pulsed systemCW optical

Layout of XFEL Synchronization System

Optical synchronization system

NarrowBand.

Direct/Interferometer

EOMs/Seeding

Two color bal. Opt. cross-corr.

End-station

LaserMLO

Laser pulse Arrival beam/laser DWC/Kly

A & cavity

Desired point-to-point stability ~ 10 fs

Direct

Distribution

Optical link Optical link

<5fs <5fs

Optical link

EDFL, soliton, t~200fs, f=216MHzSESAM, P > 100mW, phase noise < 5fs (1kHz)

Free space distribution+ EDFA

Dispersion comp.,Polarization contr.,Collinear bal. opt. cross-corr.

Other lasers

Main issue: robustness, stability and maintainability Prototype at FLASH09.12.2010, Daresbury, “Rule of lasers in particle beam research” Holger Schlarb, MSK, DESY

Optical Synchronization SystemInstallation at FLASH

• Master Laser Oscillator (RF locked to MO)• Free space distribution system to 16 ports• Optical Links: 6 stabilized using OXC & 1 passive • Front-ends

• 4 Bunch arrival time monitors (BAM) • OXC for INJ / TiSA lasers• RF locked for TiSA (HHG) (not yet completed)

Bunch Arrival MonitorDetector

Bunch Arrival MonitorsFront-end Electronics

Top view

Bottom view

to LLRF

Fiber Link Stabilization (RF based)

Time domain Frequency domain

Scheme RF link

Every odd harmonicdestructively interfere

Amplitude detectionwith mixer (sign) ofhigh harmonics (45th)allow to measure link delay variations

Fiber link stabilization (RF based)

LO generation

Optics section

In-loop Detector branch

Out-of-loop Detector branch

Digital FB loop

Fiber link stabilization (RF based)Error between in-loop and out of loop ~ 0.8fs rms, 4.8 fs pkpk, (30 m long fiber in laboratory, not stabilized, only monitored)

In-loop / 2

Out-of-loop

difference

overcomes AM-PM conversion in photo-detectors several advantages compared to OXC link (low opt. power, monitoring possible, simple disp. comp.) low cost version link with still high performance

38 hours

RF generation from optical pulses

Direct conversion with photo detector (PD)– Low phase noise (to be proven at end-station)– Temperature drifts (0.4ps/C°)– AM to PM conversion (0.5-4ps/W)– Potential for improvement (corporation with PSI)

PD BPFlaser pulses

frepf = n*frep

f = n*frep

Time domain Frequency domain

T = 5ns = 1/frep

Photo DetectorBandwidth PD

100fs Phase noise

Sagnac loop interferometer– balanced optical mixer to lock RF osc.– insensitive against laser fluctuation – Very low temperature drifts Results: f=1.3GHz jitter & drift < 10 fs rms limited by detectionRemark: much easier at hire frequencies …

MZI based balanced RF lock– new scheme, under investigation

Results double balanced MZI-L2RF

But 0.8 ps/Ktemperature dependence

2fs pkpk

Sensitive to environment

But insensitive to laser power

Thanks for your attention

Beam Based FeedbackInstallation

~ ~BAM BAM BCM BCM BAMBAM

Toroid Toroid Toroid

LLRF LLRF

A A A A

Beam Based Feedbacks:• BAM before BC2 corrects phase in RF-Gun• BAM and BCM after BC2 simultaneously correct amplitude and phase in

ACC1 and 3rd harmonic• BAM and BCM after BC3 correct amplitude and phase in ACC23

Results from BBF running at BC2

BC2 BC3

Master Laser Oscillator (MLO)Pulse generation and distribution

•Promising: OneFive ORIGAMI-15

• Repetition rate: 216,66MHz

• Average power: > 100mW

• Pulse duration: p < 150 fs

• Integrated timing jitter < 5 fsin the interval [1 kHz; 10MHz]

• Mechanically robust, easy to maintain

Fiber Link Stabilization (optically)

216 MHzEr-dopedfiber laser

J. Kim et al., Opt. Lett. 32, 1044-1046 (2007) Det 1

Det 2 -Balanced optical cross-correlator

Fiber Link Stabilization (optically) 3 generation of opto-mechanics typical in loop jitter ~ 1-2 fs rms (also smaller)

Experience:

Operate reliablyAmpl. FB to be addSmaller open questions

Dispersion management need to be improvedDelay stage too short for long links and large temp. changes

Courtesy: M. Bock

LLRF Control Tables

ADC10 PeakDetection

Pyro SPTable

-ChargeCorrection

TransferMatrixQnom

Qtoroid

tsample

Control System

ADC9 Charge Measurement

Gating

OpticalLink

LLRF SPTable

SP SignalModulation

Low Level RF Control SystemsIntra-train BBF Implementation

Toroid

BBF CalibrationTransfer Matrix Determination

Gun ACC39ACC1

BAM Pyro

A/ A/ t C/zMonitor systemActuators

ACC1 ACC39scanning

extract

measure

Open issues: observation QL changes

ACC4 ACC5

2.6 -> 3.0 (16%)

Voltage change ACC45 from 67MV to 255MV (~4MV/m -> ~15MV/m)

No change at all

Decay over time

10 min 10 min

17MV/m

Cause need to be investigated (likely main coupler antenna position change due to thermal expansion)

Open issues: Cavity pre-detuning using DC piezo voltage

Is accompanied with Ql changes => detuning via Lorenz force detuning

For some cavities orbit changes are observed reduces SASE, but simple corrector sufficient

global orbit FB essential for reproducibility of machine operation

Several studies on detuning compensation during macro-pulse Since 4 weeks PZT for ACC67 in operation (DC/AC)

Successfully used to change pre-detuning Successfully used to change pre-detuning

Holger Schlarb, DESY, 16 th DESY-MAC 10.11.2010 FLASH LLRF & Synchronization System + Roadmap On...

Documents