Noise Parameters of FET’s: Measurement, Modeling and … Noise Parameters of FET’s: Measurement, Modeling and Use in Amplifier Design ... S-Band 2-4 GHz (N/A) ... • Receiver

Noise Parameters of FET’s: Measurement, Modeling and Use in

Amplifier Design

Marian W. PospieszalskiCentral Development Laboratory

Outline • State-of-the-art • Noise temperature measurement methods employed at

CDL• Review of the sources of error • Some properties of signal and noise models of FETs• Method of measurement of noise parameters at cryogenic

temperatures• Optimal noise bias of a FET – lessons for future

improvements• Effects observed but not understood • Final observations

2RadioNet FP7-Workshop,Chalmers Technical University, June 2009

Mmin Prediction (1991) and State of the Art (2009)


Mmin Prediction (1991) and State of the Art (2009)


VLA/EVLATRx versus Frequency

TRx = m∙F + b ; m = 0.5ºK/ GHz ; b = 8ºK

EVLA Project Book - TRx Requirements (Band Center)

Band L S C X Ku K Ka Q

TRx 14 15 16 20 25 34 40 48

0

5

10

15

20

25

30

35

40

45

50

0 5 10 15 20 25 30 35 40 45 50

Rec

eive

r Tem

pera

ture

(K)

Frequency (GHz)

Simple Noise Model

L-Band 1-2 GHz (Interim)

S-Band 2-4 GHz (N/A)

C-Band 4-8 GHz

X-Band 8-12 GHz (VLA)

Ku-Band 12-18 GHz (N/A)

K-Band 18-26.5 GHz

Ka-Band 26-40 GHz

Q-Band 40-50 GHz

L#32(i)

C#20

X#01(t)

K#28

A#07

Q#18

S#01

R. HaywardNRAO, Socorro, NM

Cold Attenuator Measurement Method

Comp. Contr.

To ADC

Main source of error: Uncertainty in calibration of ENR


Cold Attenuator Noise Measurements

7

Amp.

Att.

SSair line

RadioNet FP7-Workshop,Chalmers Technical University, June 2009

0

5

10

15

20

4 6 8 10 12 14 16 18 20

Frequency (GHz)

Noi

se T

empe

ratu

re

(K)

0510152025303540

Gai

n (d

B)

u46 Noise u47 Noise u46 Gain u47 Gain

Noise and Gain of 5-20 GHz Amplifier at Ta=15 K


Hot-Cold Load Measurement Method

Sources of error:1) Uncertainty in calibration of Th

2) Change in impedance of “hot” and “cold” state


Noise Measurements of Waveguide Amplifiers


0.00

5.00

10.00

15.00

20.00

25.00

18.00 20.00 22.00 24.00 26.00

Noi

se T

empe

ratu

re (K

)

Frequency (GHz)

043 117 118 125 126 127

128 129 130 131 132 133

GBT K-Band Array Amplifiers at 19 K

Noise Performance of Q-Band Amplifiers


Sources of Error

• Uncertainty in calibration of Th and Tc

• Change in impedance of “hot” and “cold” state • Receiver nonlinearity• Receiver calibration• Receiver stability• Finite integration time for a given IF bandwidth


Common Noise Representations of 2-Ports

2

optRgRoptZgZ

oNTminT2

optZgZgRng

oTminTnT−

+=−+=

⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜

⎝

⎛

⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜

⎝

⎛

Γ−Γ−

Γ−Γ+=

2g1

2opt1

2optg

oNT4minTnT

oZoptZoZoptZ

opt +−

=Γ ngoptRN=

oNT4minT ≤

where

For All Linear Noisy Two-Ports:


Other Properties of Noise Parameters

( ) .}ngnRρReN{20

TminT +=

Hence, 2TNT4

1min

0 ≤≤

if and only if Re(ρ) ≥ 0 and correlation matrix is Hermitian and non-negative definite.

It holds for generally accepted noise equivalent circuits of both FETs and HBTs.

Simplest Noise Equivalent Circuit of a FET

fgs

rg

T42

gse Δ= ΔfdsgdkT4 2

dsi =ΔfgsqI22gsi =


Simplest Noise Equivalent Circuit of Intrinsic BT

ΔfgsqI22bi = ΔfbrdkT4

2be = ΔfcqI22

ci =

RadioNet Workshop, June 22 -23 2009, Chalmers University of Technology, Gothenburg, Sweden 17



RadioNet FP7-Workshop,Chalmers Technical University, June 2009 20

Noise Parameters of FHR01X HEMT Chip(1989)

21

It was easier to measure devices accurately 20 years ago!!!


A Method of Noise Parameters Measurement (1986)

22

nT

⎟⎟⎠

⎞⎜⎜⎝

⎛++⎟⎟

⎠

⎞⎜⎜⎝

⎛−=

s

opt

opt

sn

RR

RRNN

TT

TT 2

0

min

0


A Method of Noise Parameters Measurement (1986)

23

( )∑=

−=l

i

cni

mni TTE

1

2


24

Noise Parameters of FET: Approximation

For:dsgsd

g

t g r1

TT

ff <<

dds

ggstopt T g

T r

ff R ≅

o

dds2

tn T

T gff g ⎟⎠

⎞⎜⎝

⎛=

ggsddst

min T r T g ff 2 T ≅ 2

TNT4min

o ≅

Broadband Matching Approximation

⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜

⎝

⎛

⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜

⎝

⎛

Γ−Γ−

Γ−Γ+=

2g1

2opt1

2optg

oNT4minTnT

At some input plane of reference A :

minAmin TT = NNA = 0A

g =Γ

2Aopt

2Aopt

minAn

1

1TT

Γ−

Γ+=


Optimal Noise Bias of InP HFET at 18 K (1993)

.1x50μm

Optimal bias minimizes the value of: ( )m

D

t

dsddsds g

IfgT

I,Vf ≈≈


EVLA Ka Band Amplifier

Cryogenic Tmin at 8.4 GHz and DC Pinch–off Charactersitcs of GE HFET’s (1987)


Example of “poor pinch off” InP HFET (1993)

05

101520253035404550

0 2 4 6 8 10Ids (mA)

gm (m

S)

Dev. A Dev. B

05

101520253035404550

0 2 4 6 8 10Ids (mA)

gm (m

S)

Dev. A Dev. B

Ta=297 K Ta=18 K

Dev. A Dev. B

Tmin(297 K) Tmin(18 K)

148134

1221

at 40 GHz


Comparison of GE/MM, HRL and TRW HFET’s

Comparison of gm at 18 K (Vds=.8 V)

05

101520253035404550

0 1 2 3 4 5Ids (mA)

gm (m

S)

Dev. A Dev. B HRL TRW 041

.1x50μm HFET’s


Device Scaling: Gate Width

31

dds

gttopt T g

T r

ff R ≅gtdds

tmin T r T g

ff 2 T ≅

( )m

D

t

dsddsds g

IfgT

I,Vf ≈≈

Width R opt

Tmin

Tmin

in principle

in practice

sggst rrrr ++=


Device Scaling: Gate Length

32

gtddst

min T r T g ff 2 T ≅

ftLength gr dst

)C(C2gf

gdgs

mt +≅

π

Tmin

Ambient temp. depends on channel structure andcurrent Id but mostly not onambient temperature

Td Tg ≅

For every wafer structure there must exist a lower limit on Tmin upon further device scalingDependence of Td on device structure and propertiesof electron transport in the channel is not known


“Hot electron” noise


34

To Cool or Not to Cool

(15K)T5(77K)T nn ×≈

(15K)T10(297K)T nn ×≈

2). For well behaved cryo-devices the rule of thumb for amplifiers are:

1). No cryogenic performance can be predictedfrom room temperature performance

No change in bias upon cooling

Optimal bias at each temperature

NGST/JPL CRYO3 WAFER


Cryo3 Wafer gm vs Vgs

36

J.Shell,IPN Progress Report 42-169, JPL, May 2007



37




38


Double sweepstaken in fastsuccession


Cryo3 041 Wafer I-V Characteristics

39



Cryo3 Questions Still Open:

• 300 μm wide InP HFET’s do not behave as expected from scaling (this applies to all the discrete wafersevaluated at NRAO)

• 200 μm from cryo3 wafer exhibit a very strong dependence of noise on drain voltage at L, S and C bands

• 80 and 60μm wide devices from cryo3 4044-041 waferexhibit (sometimes) dc instability which seems to be relatedto device layout


4_8 GHz AMPLIFIER AT 15 K

05

10152025303540

3 4 5 6 7 8 9

FREQUENCY (GHz)

NO

ISE

TEM

PER

ATU

RE

(K)

20

25

30

35

40

GA

IN (d

B)

NOISE, Ids=6.5 mA NOISE, Ids=8.1 mAGAIN, Ids=6.5 mA GAIN, Ids=8.1 mA


Final Observations

• Only three wafer runs of InP discrete devices (NRAO/HRL, WMAP/HRL, NGST/JPL cryo3) have been used in construction of great majority of amplifiers for radio astronomy (VLA/EVLA,VLBA, GBT, ALMA band6, CBI, SZ-Array, WMAP, Planck LFI, VSA, AMI, MPI, JPL/DSN and others)

• No single wafer devices have ever been fully understood• There has been no significant progress in the low noise performance

of cryogenic FET’s for the past 16 years; Are we approaching the limits?

• Amplifier noise temperature is no longer the dominant component of the system noise for radio astronomy instruments with cryogenic receivers


Documents

Noise Parameters of FET’s: Measurement, Modeling and … Noise Parameters of FET’s: Measurement, Modeling and Use in Amplifier Design ... S-Band 2-4 GHz (N/A) ... • Receiver