‚Noise characterization of the LLRF system‘.

Frank Ludwig

Content :

1 Introduction to noise 2 Noise characterization of the actual LLRF system 3 Conceptional improvements 4 Different sensors for phase- and amplitude detection 5 Outlook Frank Ludwig / 03.12.04

Introduction to noise 

Definition of the spectral density :

S U

(

f

) 

T

lim   1

T F

1

T

[

U

(

t

)] 2 ,

F

1

T

[

U

(

t

)]   

T



T U

(

t

)

e

 2

π ft

(The spectral density of a time limited signal vanishes)

dt



Mean square value of a signal :

δ U

2     

S U

(

f

)

df

(derived from the Parsevalschen theorem) 

Bandlimited noise through a filter function :

δ U

2     

S U

(

f

)

H

(

f

) 2

df



S U



f



H S U

( (

f f

)  ) 1    0 else

const

.

f

 

f

, 

Replacement circuits of devices:

Summation of noise sources : Real resistor :

δ U



f

 200

kHz



f

 20

kHz



f

 2

kHz



Noise sources :

t

Thermic noise white noise Surface effects, trapping of charges flicker-noise

δ U

1

δ U

2

δ U δ U



δ U

1 2 

δ U

2 2  2

C δ U

1

δ U

2 Frank Ludwig / 03.12.04

R = R (noisy)

S U

(

f

(noiseless)

δ U

)  4

k B TR

(

R S U

  1

nV

50  ,

T

/

Hz

 300

k

) Amplifier :

δ U

= (noisy) (from datasheet)

...

δ I

(noiseless) [Low-noise electronic design, Wiley 1993], [Halbleiter-Elektronik, Rauschen, Springer 1990]

Example: 1/f-noise from SIS-junction 

Intrinsically-Shunted-Junction model :



Correlation analysis in the time domain :

I

S I S

U δ I c δ R n C

( 0 )   0 .

715 Superposition of single electron lorentz-type spectra with different occupation time results in a 1/f-noise spectrum.

Localized states within the barrier cause 1/f-noise ?

C

( 0 )   1

δ I c δ R n

fluctuation of the critical current, resistance fluctuation Frank Ludwig / 03.12.04

f c

T=300K, <140pV/Hz 1/2, (up to know world record) T<77K, <20pV/Hz 1/2 , SQUID+FLL-electronics Aliasing effects ? Better measurement method

Example: 1/f-noise from SIS-junction 

Correlation analysis in the frequency domain:



New method :

Idea: Redistribution of noise power ?

C

( 0 )   0 .

702 Microscopic description! Conservation of noise power :

γ

(

f

)  4

S

12 (

f

2 )  (

S

1 (

f

) 

S

2 (

f

))

S

1 (

f

)

S

2 (

f

) Frank Ludwig / 03.12.04

+ + + Frequency dependence of the correlation coefficient Finite amplifier bandwidth

f c



f g

Amplifier noise can be considered

Noise characterization of the LLRF System (TTF2) 

RF digital feedback system (TTF2) :



+I,-I,+Q,-Q detection scheme :

Rotation of the LO-signal in four 90 o steps (-I,+Q) Im (+I,+Q) (-I,-Q) Re (+I,-Q) Phase modulation Bandwidth for transforming 250kHz squared pulses : 

f

 10

MHz

Required regulation bandwidth only : 

f

 1

MHz

Frank Ludwig / 03.12.04

Noise characterization of the LLRF System (TTF2) 

Stability requirements on phase and amplitude of the cavity field vector :

δφ

Amplitude stability : and linearity Phase stability :

δ A

 10  4

A δφ

 0 .

01 

A δ A δ U XFEL

 100

μ

(normalized to A=1V)

V



Noise measurement at input of an ADC :

δ U TTF

2  1 .

0

mV

 10 

δ U XFEL

voltage 2mV/div  rms-voltage noise :

δ U

+ + +  

f



S U

(

f

)

df Reduce the measuring bandwidth Low-noise design



Averaging, switched low-pass!

Correlation methods S U



f

time 100ns/div ACC5, Probe DCW, AN-36 Superposition of all noise contributions :

δ U

2

DWC



δ U

2

IQ



δ U

2

MO



δ U

2

extern

 ...

 100

μ V

Frank Ludwig / 03.12.04

Frank Ludwig / 03.12.04

Noise characterization of the LLRF System (TTF2)

Where comes the noise from ?

Noise characterization of the LLRF System (TTF2) 

Noise from sensor (down-converter) :

δ U DWC

 2 .

0 

δ U XFEL P RF

 [  40

dBm

,  10

dBm

],  70

dB linearity S U

, 

S U

,

AMP f RF v f LO

Frank Ludwig / 03.12.04

S U

,

DWC

(

S U

,  

S U

,

AMP

)

v

2 

S U

,

DWC S U

,   4 .

5

nV

/

S U

,

AMP

 7

nV

/

S U

,

DWC

 70

nV

/

Hz Hz Hz

,

v

 8 .

5

Frank Ludwig / 03.12.04

Noise characterization of the LLRF System (TTF2) 

Noise from IQ-driver modul :

?

LSB jumping from 16-Bit DAC, power supply

δ U IQ

 3 .

5 

δ U XFEL - Merge fiberlink+DAC+VM, - Merge DWC+ADC+fiberlink - Low-noise design down to 10mHz for long term stability!

Noise characterization of the LLRF System (TTF2) 

Noise conversion at the down-converter :



Noise from sensor (LO-signal) :

S

 ,

RF

 0

γ

(

f

)  0

S

 ,

LO

(

f

)

S U

,

ZF

(

f

)

S

 ,

ZF

(

f

) 

S

  2 ,

RF γ

(

f

( )

f

) 

S

 ,

LO

(

f

)

S

 ,

RF

(

f

)

S

 ,

LO

(

f

) Frank Ludwig / 03.12.04

Assumption: Mixer acts only as a phase detector :

S U

,

ZF

(

f

)  2

K

2 

S

 ,

LO

(

f

) ,

S

 ,

ZF K

   30

S

 ,

LO mV

/ 

Noise characterization of the LLRF System (TTF2) 

Noise conversion over the LO-Signal at down-converter from master-oszillator :

Frank Ludwig / 03.12.04

δ U MO

 10 

δ U XFEL

Frank Ludwig / 03.12.04

Noise characterization of the LLRF System (TTF2)

How can we improve the signal-to-noise ratio ?

Conceptional improvements 

Properties of the RF digital feedback system (81MHz-CW) :

f LO f RF f IF

Measuring bandwidth : 

f

 1

MHz

Jitter conversion : 

t



t

  10

fs



f RF



T f IF



T

  160

fs

+ + + -

Suppresion of higher harmonics and disturbancies using narrow bandpass filter.

Changing of the bandwidth using averaging of the ADC.

No noise from IQ-driver and no additional ,uncorrelated effects‘.

Precise synchronization of ADC-clocks, averaging over time jitters.

May be limited by the effective ADC resolution.

Frank Ludwig / 03.12.04

Requirements on synchronization : 250kHz (TTF2) : uncritical 81MHz : normal Direct sampling : critical



Ideal mixer :

f RF f IF f LO



f RF

 Real mixer properties 

f RF



Mixing using a non-linear characteristic:

y



a

0 

a

1

x



a

2

x

2 

a

3

x

3  ...

y

All combination frequencies :

f k

 

ν f RF



μ f LO

,

ν

,

μ

 0 , 1 , 2 , 3 ...



f LO

 

f LO



f RF



f LO



f IF f IF f RF



f LO x

- Intermodulation effects (IP2,3) - 2nd harmonics

x

(

t U RF

)  (

t

)

U LO

(

t

)

U RF

(  

RF U

ˆ

LO t

) 

U LO

cos( cos(

ω RF ω LO t t

(

t

)  

φ RF φ LO

) ) P(RF) P(LO) P(IF) NF IP3 1dB MS11 PS11 MS22 PS22 MS33 PS33 Gain deg dB RF to IF deg isol IF RF dB iso LO RF dB iso LO IF dB IF(min) MHz dBm dBm dBm dB dBm dBm dB deg dB deg dB

down LT5522 -7 -5 13,2 25 10,8

up down down down down down LT5521 LT5526 MC1502 CDB-9050 DBM-182 MBA-15L -15 -10 -5 -10 -5 -7 12,5 24,2 11 -5 12,3 16,5 5 7 7,5 12 -5 15 -3 7 8,5 0

down HMJ7 -10 21 8,5 34 23 down down down HMJ7-1 IAM-92516 AD8343 -10 -10 -12 21 -3 -10 10,5 34 23 12,5 27 9 14,1 16,5 2,8 -0,4 50 49 0,1

-0,5 38 59 10 0,5 55 55 0,1 6 30 25 0 6 33 35 30 -7,5 25 20 0 -6,5 14 0

-8,5 24 24 -8,5 24 30 -5,5 34 56 0 7,1 54 0



Limits on the linearity

P IF P

1

dB

1db compression point Noise problems Crosstalk, isolation, leakage problems Noise floor

P RF

Frank Ludwig / 03.12.04

Low-level sensors for 250KHz or 81MHz

f RF



‘Parallel‘ connection of Gilbert-cell-mixers and using low-noise amplifiers :

+ -

Noise reduktion of about 4-8, high integratino, low crosstalk.

Small signals, active noisy mixer.

S U

,  /

N S U

,

AMP v S U

,

DWC f LO

(

S U

,  /

N



S U

,

AMP

)

v

2 

S U

,

DWC S U

,  /

N

 1

nV

/

S U

,

AMP

 0 .

8

nV

/

Hz

,

v

 8 .

5 ,

N Hz S U

,

DWC

 10

nV

/

Hz

 20

δφ

 0 .

001  ( 

f

 1

MHz

) Frank Ludwig / 03.12.04

f RF f LO f CLK

Frank Ludwig / 03.12.04

Pre-Averaging for the 250kHz concept

50 0 -50 Seperate phase and amplitude detectors 

Seperate phase and amplitude detectors using hybrids: Drifts of measurement setup and M.O.

150 100 

T

 70

fs

-100 0 500 1000 2000 1500

time [s]

( 60

μ V

 1

ps

, 

f

Drift of Amplifier  1

MHz

, Drift of phasedetector

f RF

 81

MHz

) Drift of Master Oscillator 2500

S U

,

RF

 80

MHz

(

f

 10

kHz

) 

S U

,

RF

 1 .

3

GHz

(

K

  5 .

5

mV f

/   10

kHz

)  10

nV

20

nV

/ /

Hz Hz

3000 Frank Ludwig / 03.12.04

High-level sensors for 250KHz or 81MHz 

Increase the signal by using high-level GaAs JFET-ring-mixer.

+ -

High noise reduction, low-noise passive mixer, high signal level.

RF-packaging, crosstalk, isolation and matching problems.

f RF

 1300

MHz

Amplitude noise reduction by using a limiter.

 20

dBm

0

dBm f LO

 1219

MHz f CLK

 36

MHz

- Clock jitter averages with 1 / - Gain a factor of 3-5 from bandwith reduction

N

1

μ s

Frank Ludwig / 03.12.04

Unsolved problems 

Consistency check +I-I=0, +Q-Q=0 ?:



Shielding, cable effects and rf-packaging, gun pulses :



Disturbancies from high-level gun pulses:

Frank Ludwig / 03.12.04

Outlook

R&D :

- Decrease phase noise of master-oscillator!

- Phase- and amplitude measurement for the injector - Test separate phase- und amplitude detectors including hybrids and ultra low-noise amplifiers - InP-based HEMTS, p-HEMT as mixers, RSFQ-logic, intermodulation effects in SIS devices - Measurement of phase- and amplitude using optical reference instead MO ? - Low-cost phase- und amplitude detectors for mass production - 81MHz-CW or 250kHz (TTF2) or a combination - Coupling of the MO and MLO - Correlation measurement of the short and long-term stability between different modules

Thanks for your attention!

Frank Ludwig / 03.12.04