Frank Ludwig, DESY

Content :

1 Stability requirements for phase and amplitude for the XFEL

2 Next LLRF system for optimized detector operation

3 Limitations from noise and non-linearity

4 Down-converter prototype for CW-modulation scheme

5 LLRF phase noise budget

6 Summary and Outlook

CERN, LLRF05 Workshop 12/10/05

F. Ludwig, M. Hoffmann, G. Möller, S. Simrock / DESY

‚Precision low-noise field detectors‘

T. Filipek, R. Romaniuk / Warsaw University

Frank Ludwig, DESY

Actual LLRF control system using a switched LO-signal :

RFfLOf

time

500us/div

voltage

0.5V/div

time

5us/div

voltage

0.5V/div

voltage

2mV/div

ACC5, ProbeDCW, AN-36

time

100ns/div peak-to-peak comparison

05.0

XFEL2TTF U10U

Amplitude stability : 5105 A

Aδ

Phase stability : 01.0δφ

Stability requirements of the cavity field vector sum :

rmsXFEL V50U (normalized to A=1V)

- 86dB dynamic range of signal-to-noise.

Stability requirements on phase and amplitude for the XFEL

MHzf 10- Bandwidth for transforming the

squared LO-signal :

- Rotation of the LO-signal in four 90o

steps using a squared LO-Signal.

Frank Ludwig, DESY

Where comes the noise from and whatdo we measure after the down-converter ?

Next LLRF system for optimized detector operation

We measure all noise sources of the loop for a finite gain,but not the residual jitter between beam and reference!

How can we improve this ?

Microphonics

. . .

- Conceptional improvements using noise reduction methods, e.g. filtering and averaging.

- Sort the priorities: low noise, low drift, high linearity, absolute accuracy.

- Improve each component.

- Minimize residual jitter by increasing or decreasing the loop gain.

Frank Ludwig, DESY

Jitter transformation :

tf

fT

IF

RF

fst 10

fs160T

MHzf 1Measuring bandwidth :

Precise synchronization system.

Averaging reduces ADC-noise and no aliasing effects.

+

-

Narrowband filtering the IF-signal reduces distortions from mixer non-

linearities. +

No noise from LO-driver.+

Proposed LLRF control system operating with a CW LO-signal :

FPGA

Master-Oscillator

klystron

high-power cavity

down-converter

LO-input

RF-input

ADC-clock

1300-81

LOf

1300

RFf

81=9 x 9IFf

DAC-clock

1300-81

Next LLRF system for optimized detector operation

+ Higher harmonics and disturbancies using bandpass filters can be suppressed.

Frank Ludwig, DESY

Down-converter limitations from noise and non-linearity

Passive Mixer + FET: Active Gilbert-mixers:

+ High linearity

+ Low NF- Large LO drive needed

(additional phase noise)

- High LO/RF crosstalk

+ High conversion gain

+ Low LO drive needed

+ Low LO/RF crosstalk

- Normal NF

- Additional 1/f-noise

Compromise between noise and linearity :

Signal distortions :

- intermodulation effects (IP3)

- higher harmonics (IP2)

IP3

RFP

IFP

dB1,OUTP

Non-lin

earit

y

Noise

floor

Noise

IP2

Spurious FreeDynamic Range (SFDRout)

Dynamic Range (DRout)

Multi-channel detector board :

- Gilbert cell mixer- Linearization during beam pauses- Filtering of distortions

Filtering of distortions :

f

Filter

Signalinformation

Frank Ludwig, DESY

HznV

S AMPU

/7,

HznV

SU

/70

MHzfLO 1300

MHzfRF 13005.8v

HznV

SU

/5.4

,

2,, )( vSSS AMPUUU

Actual down-converter

First mixer stage determines the noise performance.

Actual down-converter performance:

Noise from actual down-converter :

MHz1f ,U2U XFEL

MHz10f ,U6U XFEL

-

-

,U2U XFEL ,U6U XFEL

,U6.0U XFEL

,MHz10f

,MHz1.0f ,MHz0.1f (CW-LO-Signal)

(Switched LO-Signal)

(Cavity filtered)

LLRF05: G.Möller, [43] Multichannel down-converter board for cavity field detection at the TTF.

Frank Ludwig, DESY

14

IF Output

Matching

Circuit

BPF

9MHz

SMD-Filter

Input

Matching

Circuit

Output

Matching

CircuitLNA

Evaluation

Board

AD6645

14 Bit, 80 MSPS,100fs jitter

BPF

BPF

1300 MHzStripline Filter

1291 MHzStripline Filter

Attenuator

LO-input

RF-input

Low-Noise-AmplifierLO Input

Matching

Circuit

Active Mixer

LT5522

1300 MHz

1300-9 MHz

80 MHzOversampling

80 MHz

ADC clock

-5dBm

MHz2f 19dBGain

Down-converter prototype for CW-modulation scheme

Frank Ludwig, DESY

-110

-100

-90

-80

-70

-60

-50

-40

-30

-20

-10

0

-120

-110

-100

-90

-80

-70

-60

-50

-40

-30

-20

-10

0

-1200 2 4 6 8 10 12 14 16 18 20 22 24 28 30 32 34 36 38 40 0 2 4 6 8 10 12 14 16 18 20 22 24 28 30 32 34 36 38 40

dB

FS

dB

FS

Frequency - MHz Frequency - MHz

f = 80MHzf = 9MHz @ -1dBFSs

a

f = 80MHzf = 81MHz @ -1dBFSs

a

0 5 10 15 20 25 30 35 40

0 5 10 15 20 25 30 35 40

Down-converter prototype for CW-modulation scheme

Oversampling : Undersampling :

Oversampling promises betterSNR than undersampling.

SFDR = 86dBSNR = 72dB

SFDR = 62dBSNR = 60.7dB

Signal bandwidth:

Bandlimited noise from Mixer

Noise from ADCf

ADC-noise, oversampling, clock phase noise requirements

LLRF05: T.Filipek, Frequency Conversion in Field Stabilization System for Application in SC cavity of linear accelerator.

Frank Ludwig, DESY

)f(Ssg

1 )f(S

s

s )f(S

s )f(S mKLY,

2

12

12mMO,

2

12m,

2

12

12mR,

)f(S mKLY,

)(, mVMφ fS)(, mRu fS)(, mφ fS )(, mMOφ fS

)(, mRφ fS )(, mMICφ fS

0f

12ω

pωg ,ω

The effective noise bandwidth for the down-converter is given by : 1212 ωgω

1st order HP1st order LP 1st order LP

- MO and klystron contributions decreases with gain. - Down-converter contributions increases with gain.

LLRF phase noise budget – Residual jitter

(simplified)

Beam jitter

Frank Ludwig, DESY

Phase noise budget (Switched LO, single cavity)

Phase noise spectra : Contributions to residual jitter :

Measured down-converter noise is largerthan residual noise and beam jitter.

MHz1300f,kHz500f 0p

Hz/dBc 110)f(S mAMP,

Hz200f ,100g 12

Hz/nV70)f(S m,u

)(, mMOφ fS TTF2 (new supply)

MHz10f

MHzf 10

- Noise is filtered by the cavity.

- The down-converter is not a good indicator for the residual jitter!

Frank Ludwig, DESY

Phase noise budget (Switched LO, single cavity)

Phase noise spectra : Contributions to residual jitter :

MHz1300f,kHz500f 0p

Hz/dBc 110)f(S mAMP,

Hz200f ,100g 12

Hz/nV70)f(S m,u

)(, mMOφ fS TTF2 (new supply)

MHz1f

MHz1f

Measured down-converter noise is largerthan residual noise and beam jitter.

- Noise is filtered by the cavity.

- The down-converter is not a good indicator for the residual jitter!

Frank Ludwig, DESY

Summary and Outlook

Outlook : - Design a multi-channel board and test within accelerator environment.

- Beam jitter caused by LLRF should be measured with fs-resolution.

Decrease mixers noise :

- Passive front end structures.

- Parallel structures of detectors (VLSI prefered).

- pHEMT Gilbert mixers (promise higher gain and lower noise).

- Additional „Zero-Phase“ detectors.

Summary : - The CW-modulation scheme combines many advantages, for example :

- No aliasing effects and ADC-noise reduction.

- Filtering of distortions, which allows a linearization with improved SNR.

- For multi-channel systems Gilbert mixers are recommended.

- Oversampling promises better SNR than undersampling.

- The down-converters noise contribution to the beam jitter

is reduced by the cavity transfer function.

- Linearize the down-converters

characteristic within the beam pause.

Increase mixer output :

Frank Ludwig, DESY

Summary and Outlook

Thanks for your attention!

Frank Ludwig, DESY

Summary and Outlook

Backup Slides

Frank Ludwig, DESY

HznVSU /70]10,40[ dBmdBmPRF

+ High LO/RF isolation

- Mixing into baseband

causes additional noise

dBmPLO 58-channels from cavity probe : 8-channels to ADC-Board : LO-Input :

(Designed by G.Möller/DESY/MHF-p)

Actual down-converter

LLRF05: G.Möller, [43] Multichannel down-converter board for cavity field detection at the TTF.

Frank Ludwig, DESY

Choice of LLRF system for optimized detector operation

Actual LLRF control system using a switched LO-signal :

MHzf 10

Bandwidth for transforming250kHz squared pulses :

but required regulationbandwidth is only :

MHzf 1

Phase and amplitude detection of the cavity field vector :

Rotation of the LO-signal in four 90o steps,using a 250kHz squared LO-Signal.

Re

Im(+I,+Q)

(+I,-Q)(-I,-Q)

(-I,+Q)RFf

LOf

Down-converter output IF-signal :

time

500us/div

voltage

0.5V/div

time

5us/div

voltage

0.5V/div

Frank Ludwig, DESY

SNR gain from ADC oversampling :

Down-converter prototype for CW-modulation scheme

- The signal within the bandbass filter, respectively noise from mixer stage will not be averaged.

Oversampling

Undersampling

f/1 sf/1

Measuring bandwidth

Optimal IF frequency, clock phase noise requirements

LNA ADC

LO-input

ADC clock

x N

LLRF05: T.Filipek, Frequency Conversion in Field Stabilization System for Application in SC cavity of linear accelerator.

SNR gain from averaging within the measuring time :

Quant. noise

f

f )v,,t(SNRSNR s2

jADC Sample frequency

Measuring bandwidth

ClockjitterInternal jitter

Equiv. Inputnoise of ADC

Number of samples :

BPFBPFRF-input

f

: MMaster-

Oscillator

ADCs SNR,fDigital

I,Q-Detection

IFfI samples

Q samples

Averaging

or

Filtering

Frank Ludwig, DESY

Down-converter limitations from noise and non-linearity

Passive Mixer + FET: Active Gilbert-mixers:

+ High linearity

+ Low NF- Large LO drive needed

(additional phase noise)

- High LO/RF crosstalk

+ High conversion gain

+ Low LO drive needed

+ Low LO/RF crosstalk

- Normal NF

- Additional 1/f-noise

Compromise between noise and linearity :

Signal distortions :

- intermodulation effects (IP3)

- higher harmonics (IP2)

IP3

RFP

IFP

dB1,OUTP

Non-lin

earit

y

Noise

floor

Noise

IP2

Spurious FreeDynamic Range (SFDRout)

Dynamic Range (DRout)

Multi-channel detector board :

- Gilbert cell mixer- Linearization during beam pauses- Filtering of distortions

Filtering of distortions :

f

Filter

Signalinformation