wkshop2007 noise llopis - CORE

2007 IEEE-MTT S. Workshop « noise in nonlinear circuits »

O. Llopis, L. Escotte, S. Gribaldo, C. Chambon

LAAS-CNRS, Toulouse University7 av. du Colonel Roche, 31077 Toulouse, FRANCE

e-mail : [email protected]

Nonlinear noise measurement of microwave amplifiers HF noise parameters and residual phase noise


Outline

Tools for theory to experiment comparison in nonlinear noise modelling

• Introduction : difficulty of nonlinear noise modelling

• Residual phase noise measurement of microwave amplifiers / transistors

• LF noise measurement under large signal operation

• Nonlinear noise figure measurement of microwave amplifiers / transistors

• Correlation between nonlinear HF noise and phase noise floor

• Example of an SiGe HBT phase noise modelling


Introduction :

usefulness and difficulty of noise modelling in nonlinear devices and circuits

1) Usefulness of nonlinear noise modelling

• Predicting phase noise in oscillators • Predicting phase noise in mixers, amplifiers, phase shifters…• Predicting the noise floor in the vicinity of an amplifier driven by a large signal

2) Difficulty of nonlinear noise modelling

• Correlation between noise sources at the signal harmonics frequencies• Modification of the equivalent circuit model at a given harmonic by the large

signal at a different harmonic frequency • Change of the bias point with the large signal• Nonlinear behaviour of the noise source itself


Physical noise source 1

Active Device



k2

k3

k1 Phase Noise


Active Device



k2

k3

k1 Phase Noise

Sφ = k12(Vgs,Vds) S1+ k2

2(Vgs,Vds) S2+k32(Vgs,Vds) S3

The intrinsic noise sources are related to the phase fluctuations (or to the extrinsic LF noise) through nonlinear coefficients

2

2

1 ⎟⎟⎠

⎞⎜⎜⎝

⎛+

=

c

IIKS

ωω

I2 = IDC2 or I = < I(t)2 > ?

The intrinsic noise sources are instantaneoustlydependent on the RF current

-> cyclostationary noise

Ex : G-R noise

Necessity of an efficient measurement tool to evaluate the modelaccuracy, at different RF pump power levels


The simplified feedback oscillator model

ΔV

Δφ tΔf

Δf = oscillator frequency noise

Δφt = transistor (open loop) phase noise

ΔV = equivalent input noise voltage (LF or HF)

t0t

Q2f

dfd

f φΔ=φφΔ

=Δ

dφdf = 2 Q

f

Example of phase to frequency conversion in a 4 GHz SiGe HBT oscillator

-200

-180

-160

-140

-120

-100

-80

-60

-40

-20

0

Phas

e no

ise

(dB

c/H

z)

30 dB/dec10 dB/dec

10 102 103 104 105 106 107 108

SiGe HBT DRO phase noise

SiGe HBT amplifier phase noise

fQo

2

1

Frequency (Hz)Q = 160

-200

-180

-160

-140

-120

-100

-80

-60

-40

-20

0

Phas

e no

ise

(dB

c/H

z)

30 dB/dec10 dB/dec

10 102 103 104 105 106 107 108

SiGe HBT DRO phase noise

SiGe HBT amplifier phase noise

fQo

2

1

Frequency (Hz)Q = 160


Residual phase noise measurement of low phase noise amplifiers and transistors

Advantages of open loop (or residual) phase noise studies versus oscillator phase noise studies :

SIMULATION

Faster for a driven circuitthan an autonomous circuit

Possibility to analyse the noiseconversion both in linear andnonlinear modes

MEASUREMENT

Better control of the parameters(no needs for a loop phase control)

Measurement from linear tononlinear mode, with a precise set of input power levels


Δ Δ Δ Δ Δ ΔV G K G K f G K A G K G Vs t AM mel ampli2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 22= + + + +φ φ φϕ πι ϕ( )

DUT noise Noise Floor

Simplified measurement bench

DUTAtt. Att.

φ

G

DUTAtt.Att. Att.

φφφ

G

fo, Δf, ΔA

Amplifier phase noise characterisation


Amplifier phase noise characterisation

Δ Δ Δ Δ Δ ΔV G K G K f G K A G K G Vs t AM mel ampli2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 22= + + + +φ φ φϕ πι ϕ( )

DUT noise Noise Floor

A B LFG

FFT analyser

.

DUTDUT φφφ

XX

XX

A LF

A LF

RFAttAtt AttAtt

OL

AttAtt

Synthesised source with AM capabilitiesfor AM minimization

AM limiter

LF signal

Faraday’s shielding + battery bias

Low FM and AM noise source

Optimised measurement bench

Dual Φ detector

Delay balance and low phase noise source

AM limiter and low AM noise source

Low 1/f noise Si detector / amplifier+ cross-correlation detection


Advantage of the dual detector / cross-spectrum technique : effect of the averaging number on the 3.5 GHz noise floor

Amplifier phase noise characterisation : measurement capabilities and AM rejection problem

G. Cibiel et al., IEEE Trans. on UFFC, jan. 2002

Phase (or frequency) detection, at the mixer level versus mixer DC output

AM noise of different microwave sources

-200

-190

-180

-170

-160

-150

-140

-130

1 10 100 1000 10000 100000

Voie A100100010000

Offset frequency (Hz)

Phas

e N

oise

(dB

rad2 /H

z)

-200

-190

-180

-170

-160

-150

-140

-130

1 10 100 1000 10000 100000

Voie A100100010000


Phas

e N

oise

(dB

rad2 /H

z)

-70

-60

-50

-40

-30

-20

-10

0

10

-300 -200 -100 0 100 200 300

AM & FM detection [dBV] vs Mixers DC level [mV]

AM Detection

35 dB

FM Detection

-180

-170

-160

-150

-140

-130

-120

-110

1 10 100 1000 10000 100000

AM noise [dBc/Hz] vs frequency [Hz]

3,5 GHz DRO

3,5 GHz DRO and Amplitude fluctuation limiting device

Synthesizer


LF noise measurement under large signal operation

Complementary measurement to phase noise, to investigate experimentally the correlation between LF noise and phase noise

FFTanalyser

RF source :DRO or low AM noise synthesiser

Pin

Att

Measurement performedin a Faraday’s shield

Battery biasBattery bias

LNA

~ 20 μF

RcRb

Rload_RF

> 20 μF


Measurement example 1 : case of a PHEMT device

1,E+01 1,E+02 1,E+03 1,E+04

Frequency (Hz)

10-16

10-11

Inpu

t noi

se v

olta

ge sp

ectr

al d

ensi

ty (V

2 /Hz)

10-12

10-13

10-14

10-15

10 102 103 104 105

Frequency (Hz)

Pin

-170

-160

-150

-140

-130

-120

10 102 103 104 105

Frequency (Hz)

Phas

e N

oise

(dB

rad/

Hz)

Pin

Equivalent gate voltage LF noisespectral density versus Pin (from the linear regime and up to 4 dB compression)

Residual phase noise (dBrad/Hz)versus Pin

Llopis et al., IEEE-IMS 2001

Measured LF drain current noise divided by the measured LF transimpedance


1.E-16

1.E-15

1.E-14

1.E-13

1.E-12

1.E-11

10 102 103 104 105

Frequency (Hz)

Inpu

t Noi

se V

olta

ge S

pect

ral D

ensi

ty (V

2 /Hz)

Pin

Sv

Si

Cpg

Rg Cgd

Cpd

Rd

Cgs

IdsRi

Rs

G DSv

Si

Cpg

Rg Cgd

Cpd

Rd

Cgs

IdsRi

Rs

G DSv

Si

Cpg

Rg Cgd

Cpd

Rd

Cgs

IdsRi

Rs

G DSv

Si

Cpg

Rg Cgd

Cpd

Rd

Cgs

IdsRi

Rs

G D

-155

-145

-135

-125

-115

10Frequency (Hz)

Phas

e N

oise

(dB

rad/

Hz)

102 103 104 105

Drain(Reflection)

Gate(Reflection)

Gate-Drain(Transmission)

Pin = 4.3 dBm-155

-145

-135

-125

-115

10Frequency (Hz)

Phas

e N

oise

(dB

rad/

Hz)

102 103 104 105

Drain(Reflection)

Gate(Reflection)

Gate-Drain(Transmission)

Pin = 4.3 dBm

-170

-160

-150

-140

-130

-120

10 102 103 104 105

Frequency (Hz)

Phas

e N

oise

(dB

rad/

Hz)

Pin

-170

-160

-150

-140

-130

-120

10 102 103 104 105

Frequency (Hz)

Phas

e N

oise

(dB

rad/

Hz)

Pin

Transmission and reflection phase noise measurement (for phase fluctuations location)

TransmissionPhase noise

ReflectionPhase noise

Model proposed (2001 IEEE IMS)

Computed extrinsic gate noise Computed phase noise


Residual phase noise of an SiGe HBT amplifier at 3.5 GHz (LPNT 32)Three types of base bias networks are compared : high impedance bridge, capacitive filtering and low impedance

The low impedance bias cancels the effect of the Sibe low frequency noise source

-180

-170

-160

-150

-140

-130

-120

1 10 100 1000 10000 100000

Open loop phase noise [dBrad/Hz] vs Frequency [Hz]

HIHIHI + C=1HI + C=1μμFF

HI + C=20HI + C=20μμFF

LILI

Measurement example 2 : case of SiGe HBTsinvestigations on the effect of the base bias network on the phase noise of a SiGe HBT

G. Cibiel et al., IEEE Trans. on UFFC, jan. 2004


Measurement example 2 : case of SiGe HBTsInvestigations on the phase noise of SiGe HBTs versus the microwave input power

Case of an SiGe HBT (LPNT32)loaded onto 50 Ω at 3.5 GHz

from linear regime up to 1dB compression

-180

-170

-160

-150

-140

-130

-120

-110

1 10 100 1000 10000 100000


Res

idua

l Pha

se N

oise

(dB

rad²

/Hz)

-180

-170

-160

-150

-140

-130

-120

-110

1 10 100 1000 10000 100000


Phas

e N

oise

(dB

rad²

/Hz)

Pin-20 dBm-15 dBm-10 dBm

-5 dBm0 dBm

Pin-20 dBm-15 dBm-10 dBm

-5 dBm0 dBm

Case of an SiGe HBT (BFP620)loaded onto 50 Ω at 3.5 GHz

from linear regime up to 1dB compression

S. Gribaldo et al., 2005 ICNF and EFTF Conferences


Phase noise in microwave SiGe HBTs : discussion on the results versus Pin

1/f phase noise is necessarily a conversion noise, and is

related to the device LF noise

White phase noise can result either from a conversion

process or be an additive noise

LF-noise

Amplifier phase- noise

Oscillator phase- noise

f

HF noise (additive)

o

L

f2 Q

fc f

f

fc

fc’

fc’

fc

LF Converted noise (multiplicative)

LF Converted noise

-10 dB/dec

-10 dB/dec

-30 dB/dec-20 dB/dec-20 dB/dec

LF-noise

f

HF noise (additive)

o

L

f2 Q

fc f

f

fc

HF noise

fc’

fc’

fc

LF noise (multiplicative)

LF Converted noise

-10 dB/dec

-10 dB/dec


LF-noise

Amplifier phase- noise

Oscillator phase- noise

f

HF noise (additive)

o

L

f2 Q

o

L

f2 Q

fc f

f

fc

fc’

fc’

fc

LF Converted noise (multiplicative)

LF Converted noise

-10 dB/dec

-10 dB/dec


LF-noise

f

HF noise (additive)

o

L

f2 Q

o

L

f2 Q

fc f

f

fc

HF noise

fc’

fc’

fc

LF noise (multiplicative)

LF Converted noise

-10 dB/dec

-10 dB/dec


LF Noise Source

HF Noise Source

nonlinearmodulation

X

additive

+ RF


-180

-170

-160

-150

-140

-130

1 10 100 1000 10000 100000Frequency [Hz]

Res

idua

l pha

se n

oise

[dB

rad/

Hz]

-178

-174

-170

-166

-162

-10 -5 0 5 10

Pin [dBm]Re

sidu

al p

hase

noi

se (

dBra

d2 /H

z)

@100 kHz

1 dB/dB slope, at least when the device is not too much driven into compressionSignature of an additive phase noise

indut

add PTkFS 0=Φ

Phase noise in microwave SiGe HBTs :phase noise floor analysis –> HF noise contribution to the phase noise

Linear noise figure

Nonlinear noise figure

Cibiel et al., IEEE Trans. on MTT, jan. 2004


Question : is there a different way to evaluate the far from carrier phase noise floor and to optimise its level versus the device RF load ?

yes, investigations on this noise contribution can be performed using a (modified) noise parameter measurement set-up

Nonlinear noise parameters measurement set-up• Pump frequency (synthesiser) : 10 GHz• Noise measurement : 4 GHz


Noise parameters of two microwave amplifierslow noise 2 – 22 GHz amplifier (Miteq)low phase noise 2 – 6 GHz amplifier (AML)

-40 -30 -20 -10 00,0

0,5

1,0

1,5

2,0

2,5

3,0

3,5

4,0

4,5

5,0M

inim

um n

oise

figu

re (d

B)

Input power (dBm)-40 -30 -20 -10 0

0

10

20

30

40

50

60

Equ

ival

ent n

oise

resi

stan

ce (o

hms)

Input power (dBm)

-40 -30 -20 -10 0

-60

-50

-40

-30

-20

-10

0

10

Opt

imal

noi

se re

flect

ion

coef

ficie

nt (p

hase

°)

Input power (dBm)

-40 -30 -20 -10 00,0

0,2

0,4

0,6

0,8

1,0

Opt

imal

noi

se re

flect

ion

coef

ficie

nt (m

odul

e)

Input power (dBm)


Correlation of noise figure data with residual phase noise floor data at 4 GHz

Due to the phase noise bench frequency offset limitation (100 kHz, due to the FFT analyser), the phase noise floor has been extracted using a fit with a simple model (1/f contribution + noise floor)

Noise figure extraction based on noise parameters measurement (lines) and on 50 Ω residual phase noise measurement

100 101 102 103 104 105 106 107 108

-170

-160

-150

-140

-130

-120

Pin = -10 dBmPin = -5 dBmPin = 0 dBm

Pin = -16 dBm

Res

idua

l pha

se n

oise

(dB

rad²

/ H

z)

Frequency (Hz)

Pin = -26 dBm

100 101 102 103 104 105 106 107 108

-170

-160

-150

-140

-130

-120

Pin = -10 dBmPin = -15 dBmP

in = -20 dBm

Pin = -25 dBm

Res

idua

l pha

se n

oise

(dB

rad²

/ H

z)

Frequency (Hz)

Pin = -30 dBm

-40 -30 -20 -10 00

1

2

3

4

5

Noi

se fi

gure

(dB

)Input power (dBm)


Case of a microwave transistor : SiGe HBT BFY 420 (Infineon)

Minimum noise figure

Equivalent noise resistance

Characterisation up to 3 dB compression at different bias current

To be presented ICNF 2007, C. Chambon et al.


BFY 420 SiGe HBT

Optimum noise reflection coefficient

The change of this parameter with power, on this transistor, is not too strong.

However, there is probably a trade-off to find between the optimum load for HF noise and the optimum load for 1/f noise conversion (case of a phase noise application)


Use of residual phase noise data to model the 1/f noise contribution to phase noise

Case of an SiGe HBT – BFP 620 (Infineon)

-25 -20 -15 -10 -5 0 5 10

-30

-25

-20

-15

-10

-5

0

5

10

P out (d

Bm)

Pin (dBm)

Measured first harmonic Simulated first harmonic Measured second harmonic Simulated second harmonic Measured third harmonic Simulated third harmonic

1rst step : extraction of an accurate nonlinear model of the device

3.5 GHz

Rem : this transistor model will be presented at the 2007 ICNF, TokyoS. Gribaldo et al.


E

CB

IceSIbeS

VeS

E

CB

IceSIceSIbeSIbeS

VeS

2nd step : choice of an LF noise model (intrinsic transistor), not too complex in order to be able to include, as a third step, the nonlinear behaviour


LF noise extrinsic noise sources measurement at different RF power levelsBFP620 loaded onto 50 Ω @ 3.5 GHz from linear condition to 1dB compression

1 10 100 1000 100001E-23

1E-22

1E-21

1E-20

1E-19

Inpu

t LF

nois

e S I B

E (A²/H

z)

Frequency (Hz)

PIN=-20 dBm PIN=-10 dBm PIN=0 dBm

Pin

11

1

Pin( )1bk = a *(e - )1 c

22

2

Pin( )1bk =a *(e - )2 c

S (P )=(S +k (P )*S )*(1+k (P ))V in V 1/f 1 in V floor 2 in

10 100 1000 100001E-19

1E-18

1E-17

1E-16

1E-15

1E-14

Nonlinear LF noise model Nonlinear LF measurement

Inpu

t Low

Fre

quen

cy n

oise

(V2 /H

z)

Frequency (Hz)

evidence of the independence versus the RF power of the base current noise source

evidence of a strong nonlinear behaviour of the equivalent base-emitter voltage noise source

Equivalent input voltage noise SVbe

Equivalent input current noise SIbe


SiGe HBT @ 1 dB compression point

HBT Gummel-poon model, using a nonlinear external LF voltage noise source

First approach used : RF power dependent external voltage noise source

10 100 1000 10000 100000

-170

-165

-160

-155

-150

-145

-140

-135

-130

Res

idua

l Pha

se N

oise

(dB

rad²

/Hz)

Frequency (Hz)

Measurement data Simulation data

Works very well to simulate the HBT phase noise behaviour, when the RF load is close to 50 Ω(conditions in which the power dependent model is extracted)

But what would happens if the RF load changes in the final circuit design ?

need for an intrinsic description of the nonlinearity of the SVbenoise source


VsVe

Vbe1 TermTerm

Z=50Num=

P_1TonePORT1

Freq=f GHzP=polar(dbmtow (pe),0)Z=50 OhmNum=1

reseau_polarX4potar=9.5

CollecteurB ase

LL11

R=L=400.0 nH

IsolatorSMLISO13

Z2=Z1=Isolat=100. dBVSWR1=1.Loss1=0. dB

CC15C=100.0 pF

IsolatorSMLISO12

Z2=Z1=Isolat=100. dBVSWR1=1.Loss1=0. dB

CC14C=100.0 pF

T_model_LPTN32X6

mod="pi"input_pow er=peIB=100e-6

C

E

B

I_ProbeI_Probe2

LL12

R=L=400.0 nH

I_ProbeI_Probe1

Agilent ADS software modelling

Sib

V e

sources_bruitX9

Plancher+

Plancher-Sice-

Emetteur

Base

Collecteur

Sibe-Sibe+

Sice+

Sibc-

Sibc+

jonction_BC+BEX4

E

C

BB'

capa_BEX2

EB

gene_courant_CEX3

E

B

C

capa_BCX1C

B

RR57

=sqrt(Is/Ikf)r=sqrt(Is/Ikr)t=0.125/(1-vbc/Vaf-vbe/Var)*(1+teta*exp_soft(vbe/(2*ut))+tetar*exp_soft(vbc/(2*ut)))

capa_BC_distribX5C

B

RR4R=Rc/4

R=L=Lci/1.1125

HBT nonlinear model

Nonlinear noise source defined versus instantaneous RF collector courant value

added to the non linear modelthanks to SDDP element in Agilent ADS

Base emitter current noise source and emitter voltage noise source are also included within

the SDDP, but as constant noise sources

Second approach used : nonlinear collector current noise source


Good agreement found with the nonlinear noise approach

Further work :

Investigations on the nonlinear effect on the extrinsic SVbe (difficult to simulate)

Correlation to the device physical parametersComparison between measurements and modelling for a BFP620 with Ic=10mA and three different LF bias impedances

Comparison between measurements and modelling for a BFP620 with Ic=10mA and -20dBm<Pin<0 dBm

1 10 100 1000 10000 100000-180

-170

-160

-150

-140

-130

-120

-110

Res

idua

l Pha

se N

oise

(dBr

ad²/H

z)

Frequency (Hz)

High Impedance With 20 uF capacitance Low impedance

1 10 100 1000 10000 100000-180

-170

-160

-150

-140

-130

-120

Res

idua

l Pha

se N

oise

(dBr

ad²/H

z)

Frequency (Hz)

Second approach used : nonlinear collector current noise source

Model validation

3.5 GHz phase noise at three different input power levels

3.5 GHz phase noise with different LF base impedances


9.6 GHz two stages low phase noise amplifier

SiGe HBT + Si BJT

8.2 dB Power Gain -163 dBrad2/Hz @ 10 kHz

Application : oscillation on a sapphire resonator

Application example : X band low phase noise amplifier

Simulation : intrinsic nonlinear noise source

10 100 1000 10000 100000-170

-160

-150

-140

-130

Res

idua

l pha

se n

oise

(dB

rad²

/Hz)

Frequency (Hz)


10 100 1000 10000 100000-165

-160

-155

-150

-145

-140

-135

-130

-125

Our amplifier A 8-12 GHz AML amplifier

Res

idua

l pha

se n

oise

(dBr

ad²/H

z)

Frequency (Hz)

Comparison with a low phase noise commercially available amplifier

Our X band amplifier :

G = 8.2 dBPower consumption = 2V, 20 mADesigned to oscillate on a resonator with 6 dB coupling

AML X band amplifier :

G = 24 dBPower consumption = 15V, 600 mADesigned to amplify low level low phase noise signals


Thank you for your attention

Documents

wkshop2007 noise llopis - CORE