nlc - the next linear collider project control and feedback for rf linacs marc ross rf control and...

NLC - The Next Linear Collider Project

Control and Feedback for RF Linacs

Marc Ross

RF Control and MonitoringFeedback

Like most modern ‘plants’ there are loops within loops and feedforward

Special issues: precision ‘handling’ of microwave and ~high bandwidth

Author NameDate

Slide #2

Joint Accelerator School - 2002 Marc Ross – SLACNovember 14, 2002

CLIC J/NLC TESLA

Typical controls and feedback loops

• Accelerating vector – phase and amplitude– Low Level (long distance) Distribution– Source– High Power distribution– Structure beam loading & thermal… Feedback– Environment

• Transverse– Position– Emittance

• Longitudinal– Energy– Energy spread & z

…and protection systems for high power linacs

Author NameDate

Slide #3


CLIC J/NLC TESLA

Phase and amplitude

tolerances – NLC example

Config # 1 E1 (GeV)

2 E2 (GeV)

3

7 18 30 5 335 -30

8 20 30 7 320 -30

9 22 30 9 300 -30

A parameter that characterizes the strength of the wakefield relative to the focusing is the BNS energy spread needed for autophasing:

Autophasing is the condition where the chromatic growth of a beam performing a coherent betatron oscillation exactly cancels the wakefield growth and thus the beam oscillates as a rigid body.

BNS phase offsets imply tight phase stability tolerances (+ extra gradient)

Author NameDate

Slide #4


CLIC J/NLC TESLA

Phase and amplitude tolerances - NLC example (2)

Parameter Accuracy Stability& Resolution

Units

Energy profile

0.5 0.1 % voltage

Energy gain knowledge

5 0.1 % voltage

Phase readback

1 0.1 degree

Phase stability

N/A 0.1 degree

X-band:

11.424 GHz

= 26.3 mm

/360 = 73 um

/360*0.66 = 50 um

(0.66 for plastic cable signal speed)

Author NameDate

Slide #5


CLIC J/NLC TESLA

RF stabilization speeds

• Two kinds of linacs:– Pulse width is long compared to the transit times ‘within the

pulse’ feedback is necessary• Superconducting

– Stabilize microphonics

• Warm proton linacs– Pulse width is short compared to transit times ‘within the pulse’

feedback is not possible• Warm electron linacs• Interpulse feedback is required

– Stabilize thermal effects

• Beam loading

Long Pulse RF Control (Proton linacs and cold electron linacs)

S. Simrock - DESY

Author NameDate

Slide #7


CLIC J/NLC TESLA

Linac RF block diagram

3 basic loops:

1. Long baseline distribution

2. High power amplifier (Klystron)

3. Beam – based

1

1

2

3

SL

AC

RF

Dis

trib

uti

on s

chem

atic

- 1985

PAD = Phase and Amplitude Detector

Author NameDate

Slide #9


CLIC J/NLC TESLA

• RF phase stability at the 1 degree X-band (0.2 picosecond) level over the ~30 kilometer length of the machine.

• The RF timing requirement corresponds to a L/L stability of <2.510-9, which would be impractical without feedback.

• ( the timing distribution system will use the same hardware as the RF distribution system.)

• It is assumed that RF phase measurements relative to the electron beam will be used to obtain long term stability.

•The RF distribution system needs to maintain the RF phase to within 20 degrees X-band (<510-8) for long periods of time when the beam is not running.

RF long baseline distribution system specifications – NLC

‘Common mode’ effects are worst

NLC RF Distribution System

NL

C R

F D

istr

ibu

tion

tes

t - 2001

Author NameDate

Slide #13


CLIC J/NLC TESLA

High Power Phase and Amplitude Detection and Control

• High power couplers (40-70 dB)

• Cables

• Diode/Mixer detectors

• Attenuators

• Phase shifter

• Phase measurement

• Control system architecture

Hig

h P

ower

X-b

and

W

aveg

uid

e co

up

ler

– 60

dB

Hig

h P

ower

S-

ban

d B

eth

e h

ole

Wav

egu

ide

cou

ple

r

RF input

Cutaway detector diode – showing failed connection

RF Amplitude: Diode Detectors

‘Video’ out

Diode junction

Simple illustration of RF detector diode operation

Output matching network

Author NameDate

Slide #16


CLIC J/NLC TESLA

-10 -5 0 5 10 15100

101

102

103 10270b

dBm

mV

-10 -5 0 5 10 1510

0

101

102

103 10270a

dBm

mV

Matched Detector Diodes from Agilent

Power in

Vol

ts o

ut

Showing deviation from square law at moderate power – Low signal output is proportional to the square of the incoming RF AC voltage (V out P)

Author NameDate

Slide #17


CLIC J/NLC TESLA

Non-’Square Law’ detector:

Thermionic Diode

Used in original SLAC phasing system to extend dynamic range Closer to linear

Very radiation hard

Str

uct

ure

Ph

asin

g sy

stem

- S

LA

C (

1965

)

High Power RF attenuator

RF Phase shifter

In

Out

Capacitance change in ‘varactor’ diode moves effective reflection point

Conductance change in ‘PIN’ diode changes reflection coefficient

Author NameDate

Slide #20


CLIC J/NLC TESLA

Phase measurement

• Mixer output – if both RF / LO ports are basically the same frequency: ARF*ALO* cos()

– neglect usual mixer issues (intermod, compression)

– worry about others – offsets/diode matching

• Phase ambiguity and offsets:1. Nulling + dither to measure sign of derivative

• Wobbler (+/- 180)

• active synchronized wobbling to monitor offset

2. I/Q

• calibrated ‘double channel’

• 5 parameters – two gains/offsets and 1 angle

Author NameDate

Slide #21


CLIC J/NLC TESLA

‘PAD’ phase detector (and shifter) circuit

Nulling + Wobbler

Author NameDate

Slide #22


CLIC J/NLC TESLA

Controls Architecture

Phase and Amplitude Detector

SLAC - 1983

Sampled RF waveforms

One point digitized/pulse (120 Hz) with 30 MHz bandwidth

Cal. RF amplitude-

Fitted for energy gain estimate - lattice feedforward

RF amplitude- vs klystron drive atten. Klystron saturation

Beam Volts Modulator timing

SLAC - 1985

Author NameDate

Slide #24


CLIC J/NLC TESLA

motorized phase controller – 1 klystron

temp

temp

manual drive line length

What the long term feedback is doing…

Common mode error – either injection or distribution system

Single klystron – environmental (e.g. leakage through insulation)

Author NameDate

Slide #25


CLIC J/NLC TESLA

“Precision” microwave

• High power RF controls and monitoring + beam position monitors + beam phase monitors– SC Cavity tuning at TTF; lorentz force compensation + coupling

control

– Bunch length and Beam ‘tilt’ monitors

• programmed phase control– NLC ‘Delay Line Distribution System’

– Beam loading feedforward for short pulse linacs

2002… Modern RF controls

Author NameDate

Slide #26


CLIC J/NLC TESLA

Digital Low Level RF

• Precision I/Q determination– Phase/amplitude calculations with very low (bit) noise

– Use of complex math linearizes v/v amplitude and phase

1. Home-made – outgrowth of DSP based multi-bunch storage ring feedback systems

– TTF & SNS (DSP/FPGA based)

2. Commercial– Echotek ‘Digital Down-Conversion’ (Digital receiver)

• (Within the last 10 years) Biggest challenge integration with the control system &

diagnostics

Cold Linac LLRF – TESLA / TTF Simrock, DESY

System Block Diagrams

Analog (CEBAF ~1994)

Digital (TTF ~ 1998)

Author NameDate

Slide #28


CLIC J/NLC TESLA

Synchronous Digital Sampling – Direct down conversion

sampling clock effectively LO- importance of sampling clock stability

Digital RF How it really works

Author NameDate

Slide #29


CLIC J/NLC TESLA

TTF LLRF Drive Controls D

AC

DA

CReIm

Cavity 32

......8x

Cavity 25

klystronvector

modulator masteroscillator

1.3 GHz Cavity 8

......8x

Cavity 1

cryomodule 4

...cryomodule 1

. . . .

LO 1.3 GHz+ 250 kHz

250 kHz

AD

C

f = 1 MHzs

. . . . ...

vector-sum

( )aba -b

1 8( )ab

a -b

25( )ab

a -b

32( )ab

a -b

setpointtable

gaintable

feed

tableforward

+ digitallow pass

filterImReImRe ImRe

clock

LO

AD

C

LO

AD

C

LO

AD

C

S. SimrockDESY

Also have tuners, coupling etc.

TTF (DESY) DSP based I/Q controller

(Simrock)

loadableKCM

loadable

KCM

loadable

KCM +

+

Input latch

Iset

Qset

Mux,sign

Input trace

buffer512x16

++

+ +

+

restart 4-cycle accumulate

+

+

xor

+

+

feedforwardbuffer512x8

OutputDAC

saturate

reg

integrate enable

xor signfeedback enable

-

clear

regerr1 err2

samp

setp

regcum

regavg

regerr3

int1

reg

10

11

10

Host

Host

Host

Host

to 10 bitssaturate

16

+

+

regint2

loadable

KCM

loadable

KCM

loadable

KCM +

+

Input latch

Iset

Qset

Mux,sign

Input trace

buffer

512x16

++

+ +

+


+

+

xor

+

+


OutputDAC

saturate

reg

integrate enable

xor sign

feedback enable-

clear

regerr1 err2

samp

setp

regcum

reg

avg

regerr3

int1

reg

10

11

10

Host

Host

Host

Host

to 10 bits

saturate

16

+

+

reg

int2

Larry Doolittle – LBNLSNS Low Level RF Digital Feedback FPGA

(LINAC 2002 proceedings)

SNS Low level ‘within the pulse’ feedback Gate Array program schematic

Larry Doolittle – LBNLSNS Low Level RF Digital Feedback -1

loadable

KCM

loadable

KCM

loadable

KCM +

+

Input latch

Iset

Qset

Mux,sign

Input trace

buffer

512x16

++

+ +

+


+

+

xor

+

+


Output

DAC

saturate

reg

integrate enable

xor sign

feedback enable-

clear

regerr1 err2

samp

setp

regcum

reg

avg

regerr3

int1

reg

10

11

10

Host

Host

Host

Host

to 10 bits

saturate

16

+

+

reg

int2

• Input Sampler

• Diagnostic buffer

• Averaging

• Set point subtraction

loadable

KCM

loadable

KCM

loadable

KCM +

+

Input latch

Iset

Qset

Mux,sign

Input trace

buffer

512x16

++

+ +

+


+

+

xor

+

+


Output

DAC

saturate

reg

integrate enable

xor sign

feedback enable-

clear

regerr1 err2

samp

setp

regcum

reg

avg

regerr3

int1

reg

10

11

10

Host

Host

Host

Host

to 10 bits

saturate

16

+

+

reg

int2

Larry Doolittle – LBNLSNS Low Level RF Digital Feedback -2

• I / Q gain scaling and recombination ‘KCM’

• System calibration input

• DAC driver

• Integrator loop for fine error zeroing and feedforward input

Uses ~ 20% of the $20 FPGA

FPGA

DAC

ADC

SNS Digital LLRF prototype circuit - LBNL

Small,

Simple hardware,

~ simple software (EPICs)

Easily tested

Commercial Digital I/Q receiver

Integrated by Echotek

S. Smith, SLAC

Programmed phases/amplitudes used to switch outputs and compensate for beam loading

Delay Line Distribution System

Linac LLRF Drive

TW

T

Kly

stro

n

TW

T

Kly

stro

n

TW

T

Kly

stro

n

TW

T

Kly

stro

n

TW

T

Kly

stro

n

TW

T

Kly

stro

n

TW

T

Kly

stro

n

Modulated11.424 GHz

TimingSystem

High SpeedDDS

DDSClock

300 MHz

DDS UpdateMemory

TWT

Kly

stro

n

MIXER

IF189.25MHz

Lowpass120 MHz

TransitionTime Amplitude Frequency Phase df/dtState

LO1625 MHz

MIXER

IF2714MHz

Bandpass714 MHz

LO210.7 GHz

Bandpass11.4 GHz

xxx xxx xxx xxx xxx1



: : : : ::

DLDS

DDSStates

100 MHz

NLC RF compression system control using Direct Digital Synthesis – waveform memory

NLC Linac LLRF

Measurement Requirements

Parameter Value Details

Bandwidth > 100 MHz at -3 dB

Rise time < 5ns 10% to 90%

Phase resolution 1 degree At 11.424 GHz

Dynamic Range > 20 dB

Amplitude Resolution 10-3 of full scale

Beam phase wrt RF 1 degree At 11.424 GHz

Beam signal / RF -40 dB (!)

Reflected power detector max input < 100 mW Peak

Reflected power detector rise time < 10 ns

DLDS Waveforms with Beam Loading Compensation

0 500 1000 1500 2000 2500 3000

Am

plitu

de a

t eac

h D

LD

S O

utpu

t

Time (ns)

1

2

3

4

5

6

7

8

Each of 8 klystrons is programmed and combined to give independent outputs for each of 8 structure groups

Klystron 3.2 us

Structure 400 ns

Author NameDate

Slide #40


CLIC J/NLC TESLA

Bunch length• Streak cameras

– resolution limited to ~ 1mm– space charge, calibration

• Coherent radiation– stronger signal with shorter beams– asymmetry difficult (use power spectrum – phase info lost)

• Deflecting RF structures– promising

• Broadband microwave emission– cheap, relative – a given

• accurate monitor critical for short wave FEL

Microwave based beam diagnostics

Author NameDate

Slide #41


CLIC J/NLC TESLA

Transverse deflection

Brute forceCalibratedExpensiveExcellent resolution

SLAC LCLS – Krejcik/Emma (EPAC 02)SLAC/DESY TTF2

Old idea – 1965 ‘LOLA IV’Testing in linac sector 29

Krejcik / Emma EPAC 2002

Author NameDate

Slide #46


CLIC J/NLC TESLA

Beyond Bunch length Correlations

• E – z• y – z• x – y

• Proposed use of simple microwave single cell cavities to estimate correlations

• Most phase space distortions start with a linear correlation– a monitor simple, cheap and accurate compared to a

profile monitor can be more widely distributed and used to pinpoint errors

Microwave based beam diagnostics

Author NameDate

Slide #47


CLIC J/NLC TESLA

Response of Cavity BPMto Point Charge

Q

)sin()( taqtV

S. Smith – SLAC, Snowmass 2001

Response of BPM to Tilted BunchCentered in Cavity

q

2sincos

2)

2(sin

22)

2(sin

22)( ttt t

qat

qat

qatV

Treat as pair of macroparticles:

tq/2

q/2

Tilted bunch

• Point charge offset by

• Centered, extended bunch tilted at slope t

• Tilt signal is in quadrature to displacement

• The amplitude due to a tilt of is down by a factor of:with respect to that of a displacement of (~bunch length / Cavity Period )

2sincos

2)( t

t tqa

tV

)sin()( taqtVy

TVV tt

y

t

24

Example• Bunch length t = 200 m/c = 0.67 ps

• Tilt tolerance d = 200 nm

• Cavity Frequency F = 11.424 GHz

• Ratio of tilt to position sensitivity ½ft = 0.012

• A bunch tilt of 200 nm / 200 m (1 mrad) yields as much signal as a beam offset of 0.012 * 200 nm = 2.4nm

• Need BPM resolution of ~ 2 nm to measure this tilt

• Challenging!– Getting resolution

– Separating tilt from position

• Use higher cavity frequency?

Need 1 mrad tilt sensitivity for linac tuning

Cavity BPMFFTB (Shintake) ATF ext line (Vogel) X-band (Naito)

f 5.712 6.426 11.424 (GHz)position resolution 20 200 200 (nm)Vt/Vy (200um sig_z) 0.6% 0.7% 1.2% (.5 pi sig_t f)achieved 'projected dipole resolution' (200um sig_z) 3.3 29.7 16.7 umachieved 'tilt' angle resolution 17 149 84 mradachieved 'trajectory angle resolution' 3 26 30 uradcavity 'length' 15 15 8 mm

Angled trajectories• A trajectory that is not parallel to the cavity axis also

introduces a quadrature signal (in phase with ‘tilt’ signal)

• Projected ‘dipole’ sensitivity is increased by z/cavity length

– ~ 50

ATF z ~ 8mm gives expected tilt resolution ~ 0.1mrad

y res/y ~ 5%y’ res/y’ ~ 10x

Relative normalized precisionBeam position/beam traj angle

Very good resolution possible – 25 nm achieved in FFTB few nm possible by limiting spatial dynamic range

Wave cavity BPM X-band

12 mm bore

Naito/Li

ATF extraction lineC-band cavityL = 12mm, Radius = 26mm, f = 6426MHz, =46.6mm Movers – x, y, pitch (y-z)

ATF Cavity BPM – V. Vogel / H. Hayano

Author NameDate

Slide #54


CLIC J/NLC TESLA

ReferenceCavity

BPMcavity

C-bandAmplifiers15dB gain

X

X

X

Splitter

6410MHzsource

StriplineBPM

4:1 combinerX4

Scope, 250Ms/s4 channel, external trigger500 samples/ch

20MHz BW limit

Ch1

Ch2

Ch3 (Y)

Ch4 (X)

Tilt monitor electronicsJ. Frisch

Raw ‘mixed – down’ scope data from cavity BPM

Phase and amplitude wrt ref are extracted

(I and Q)

I Q response as the cavity is moved vertically using mover

The angle is arbitrary (phase offset between ref and BPM cavity)

A ‘monopole’ beam with an axial trajectory should give a (0,0) response at some point

Use the cavity ‘tilter’ to observe response to tilted trajectories

(Beam ‘tilter’ was not ready during this test – May 2002)

Compare 35 urad with 26 in table estimate

Author NameDate

Slide #56


CLIC J/NLC TESLA

Concluding remarks

• Digital LLRF field attractive and exciting– Wide variety of application from controls to instrumentation

• Next few years will see the application and success of critical new techniques, opening up new paths to higher brightness, low emittance transport, more stability…

nlc - the next linear collider project control and feedback for rf linacs marc ross rf control and...

Documents