Extraction of Scalable HiCuM Parameters and Verifikation ... · Extraction of Scalable HiCuM Parameters and Verifikation ... San Diego, San Diego, CA 92093 . ... Ick lim 2 lim 0 1

5/12/2006 © 2006 IBM Corporation

IBM Systems and Technology Group

Workshop On Compact Modeling, Nanotech 2006

Extraction of Scalable HiCuM Parameters and Verifikation for Advanced SiGe HBTs

Ramana M. Malladi, Kim M. Newton, Michael Schroter*, VukBorich†

Susan L.Sweeney, Jay Rascoe, Sonal Venkatadri, JianYang†, Steve Chen† IBM Semiconductor Research & Development Center, Essex Junction,VT 05452

*University of California, San Diego, San Diego, CA 92093

© 2006 IBM Corporation2 5/12/2006Workshop On Compact Modeling, Nanotech 2006


OutlineMotivation

Modeling Methodology

Parameter Extraction ProcedureResults

– DC

– Small Signal

– High-Frequency Noise

– Distortion

Summary



200 GHz SiGe HBT Technology: Applications in the W-band

• 60 GHz ISM band

• Wireless PAN

• Large Available Bandwidth (57-64 GHz)

• Potential for high data rates: 200Mb/s to > 1Gb/s

• Point to Point links

• 77 GHz Automotive Radar Sensors

SiGe−Chip

Radiation

I/Q

I/Q

PLL

QFN−Package

Package Mold

Package Pin

Tx/Rx Flip−Antenna

Wirebond Pad Wirebond

Filter Structure

C4−Balls

Substra

te

Underfill

90

MIX

MIX

LNA

I−signal

Q−signal

VCO

90PA

MIX

MIX

Q−signal

I−signal

VCOIBM’s Vision for a Fully Integrated Packaged

Low-cost SiGe MMW Transceiver



STI

SiGe Base

C CB BE

SIC

Structure of an Advanced SiGe HBT

Emitter

Deep Trenchx

E B C

Ge

x

• Ge ramp and base thickness help in Speed

• SIC tailoring for Breakdown

J.S.Rieh, Proc. of IEEE



Expected Scope of Model Scalability

p-cell

Performance Type (HP, MP, HB)

Emitter Width (w)

Emitter Length (l)

Emitter Stripes (n)

Multiplicity (m)

Contact Config

(CBE,CEB,..)

Reliability Options



0

80

160

240

1.E-05 1.E-04 1.E-03 1.E-02

Ic (A)

fT(GHz)

0

10

20

30

1.E-04 1.E-03 1.E-02 1.E-01

Ic (A)

fT (G

Hz)

0

20

40

60

1.E-05 1.E-04 1.E-03 1.E-02

Ic (A)

Variation of fT - Ic with Vcb

• Depending on the collector profile, fT may increase or decrease with Vcb

• Difficult to model this effect by Spice Gummel- Poon

HP MP HB



0

1

2

3

4

0 2 4 6

Vce (V)

Ic (m

A)

0.78

0.82

0.86

0.90

0 2 4 6

Vce (V)

Vbe

(V)

• The BVCEO spans a wide range in a given technology

• The effect of self-heating becomes severe in advanced HBTs

• Quasi-saturation modeling critical for applications using HB devices (eg. PAs)

Avalanche and Self-heating Regimes



Terminal Resistances 4

Capacitances (BE, BC, CS C-V ) 23

Saturation Currents 22

(Fwd and Inverse, Ideal and non-ideal)

Low current transit time 3

High current transit time 10

Self-heating 2

Avalanche 2

Substrate Transistor/Substrate Network 7

Noise and NQS 5

Temperature coefficients 14

HICUM (High-Current Model) Parameters for Extraction# of parameters



Methodology

Test Structures: Transistors of various geometries

Specialized Test Structures

Measurements: DC, S-parameters (110GHz)

Model: Hicum 2.1

Parameter Extraction: Matlab

Simulations: Spectre (Cadence)

• Scaling Equations are implemented in Spectre language in the model file



⎟⎠⎞

⎜⎝⎛

+=

⎟⎠⎞

⎜⎝⎛

+

−=

−=

+=

22121

22121.2

12

)12(2

)12()11(

YYrealRsub

YYimagf

Ccs

fYimagCbc

fYimagYimagCbe

π

π

π

Junction Capacitance Extraction Flow

C B E B C

Cold-S: (i) Vbe=0, vary Vcb (ii) Vcb=0, vary Vbe Total Capacitancefrom Cold-S (Cj)

(On Modeling Transistors)

STI oxide Cap.(Large area Str)

OverlapCapacitance

(Large area Str)

Deembed and Split Cbei into Area & Peri Parts

Cj= CaA + CpP

Area PartC-V fit

Peri Part

C-V fit

C-V Fit



0.0

1.0

2.0

3.0

-3.0 -2.0 -1.0 0.0

Vbc (V)C

bc (f

F/Sq

um

)

0

25

50

75

100

-1 -0.5 0 0.5 1

Vbe(V)

Cbe

(fF)

HP

MP

HB0.52 μm

3 μm

18 μm

12 μm

)(55.0),(65.0),(95.0)()()(

ln.

000

0

0

HBVSPVHPVVbiHBnSPnHPn

nnkTVbi

nnn

p

n

=>>

⎟⎟⎠

⎞⎜⎜⎝

⎛=

Scaling of Junction Capacitances

zVbiVj

CC⎟⎠⎞⎜

⎝⎛ −

=/1

0

● Data, — Model0.12xL



Base Resistance Test Structures

C CB2 B1 B2

w

0

300

600

900

1200

0.0 0.1 0.2 0.3

W/dL

Rb

(Ohm

s)Extrinsic BaseResistance

Increasing Vbe

LWrr

IV

R bix

BBb Δ

+=Δ

=2

2,1B1B2

C

E



Bias Dependence of Base Resistance and

Extraction of Zero-bias Hole Charge in the Base

3000

3500

4000

4500

5000

-0.6 -0.3 0.0 0.3 0.6Vbe (V)

Rbi

(Ohm

/Sq)

250

300

350

Rbx

(Ohm

-um

)

010

QpQjciQjei

rbirbiR +

≈−≡′∫=2

1

1 x

xp

bi

pdxqr

μ

-0.2

0.0

0.2

0.4

-10 0 10

Qj (fC/squm)R

'

Qp0 ~ 60 fC/μm2



⎥⎦⎤

⎢⎣⎡ −=

VtV

VtV

QQIsi bcbe

ppTT expexp

/ 0 fcfcfbfefefT

rTfTjcijcijeijeippT

QhQQhQ

QQQhQhQQ

++=

++++= 0

1.E-12

1.E-08

1.E-04

1.E+00

0.4 0.7 1.0Vbe (V)

Ic, I

b (A

)

1.E-14

1.E-10

1.E-06

1.E-02

1.E+02

0.4 0.7 1.0

Vbe (V)Ie

, Ib,

Ix (A

)

Transfer Current: Forward and Reverse Gummels



0.E+00

2.E-03

4.E-03

0.E+00 1.E+03 2.E+03 3.E+03

1/Ic 1/

2.pi

.fT

0

100

200

1.E-04 1.E-03 1.E-02

Ic (A)

fT (G

Hz)

Extraction of Low Current Total Transit Time

jcT

jcjefT

CrercxIcVCC

f)()(

.21

++++=τπ

pCBCBBEEfecf τττττττ ++++==



( ) ⎟⎠⎞

⎜⎝⎛ −+−Δ+= 111000 c

c bfvlhf ττττ

Extraction of the Low Current Transit Time Components

0

2

4

6

-1.0 0.0 1.0 2.0 3.0Vcb (V)

● Data, — Fit

τ f0

(ps)

Collector τ0(ps) Δτ0h(ps) τbfvl(ps)

HP 0.66 -0.1 0.1

MP 2.2 2.6 2.73

HB 4.9 3.5 6.0

Δτ0h= { τbfvl = carrier jam component

BC SCR Component

Early Effect Component

c = cjci0/cjci

τ0 = predominantly neutral base (not f(V))

HP

HB

MP



Extraction of the High Current Transit Time Parameters

• Ick: critical current for the onset of high-current effects (base push-out)

• High-current part of the transit time:

• Ick is determined graphically0ff τττ −=Δ

⎟⎟

⎠

⎞

⎜⎜

⎝

⎛ −′+

⎟⎟

⎠

⎞

⎜⎜

⎝

⎛ ′+

′=

PT

ce

ceci

ce

VVV

VV

rVIck lim

2

lim

0

1

1

1

0

2

4

6

8

0.0 0.5 1.0 1.5 2.0

Ic/Ick

Δτ(s

)

0

4

8

12

0.0 0.5 1.0 1.5

Vce (V)

Ick

(mA

)

0

4E-12

8E-12

0.0E+00 8.0E-03 1.6E-02

Ic (A)

Δτ(p

s)

Vcb

Ick1 Ickn

● Data, — Fit



0.E+00

1.E-04

2.E-04

3.E-04

4.E-04

0.0 0.5 1.0 1.5 2.0

Vce (V)

Ic (A )

0.52 μm

0.E+00

1.E-03

2.E-03

3.E-03

4.E-03

0.0 0.5 1.0 1.5 2.0

Vce (V)

Ic (A)

3 μm

0.E+00

4.E-03

8.E-03

1.E-02

2.E-02

0.0 0.5 1.0 1.5 2.0

Vc e ( V)

I c ( A )

18 μm

Geometry Scaling of the Output Curves

W=0.12 μm W=0.12 μm W=0.12 μm


© 2005 IBM Corporation9 5/12/2006Workshop on Compact Modeling, Nanotech 2006

0

50

100

150

200

250

1.E-05 1.E-04 1.E-03 1.E-02 1.E-01Ic (A)

fT(GHz)

0

10

20

30

1.E+09 1.E+10 1.E+11 1.E+12

Frequency (GHz)

MA

G/M

SG (d

B)

fT-Vcb and MAG/MSG: Model vs Measurement

Vcb= -0.4, 0, 0.5

Vbe

● Data, — Model



0

50

100

150

200

250

1.E-05 1.E-04 1.E-03 1.E-02 1.E-01

Ic (A)

fT (G

Hz)

Geometry Scaling of fT-Ic

● Data, — Model

0.52 μm

3 μm

18 μm



Noise Parameter Simulations

High Frequency Noise Parameter Simulations

• Simulated by SP (with noise) engine in Spectre

2

,min optGGG

n YYGRFF −+=

Fmin : Minimum Noise Figure

Rn: Noise Resistance

YG,opt: The optimum admittance to get minimum noise (GG,opt + jBG,opt)



0

1

2

3

0.7 0.75 0.8 0.85 0.9 0.95

Vbe(V)

Fmin

(dB

)

0

40

80

120

0.7 0.75 0.8 0.85 0.9 0.95

Vbe (V)

Rn

(Ohm

s)

0

0.4

0.8

0.7 0.75 0.8 0.85 0.9 0.95

Vbe (V)

|Gop

t|

0

40

80

120

0.7 0.75 0.8 0.85 0.9 0.95

Vbe (V)

Phas

e(G

opt)

Noise Parameters: Measurement vs Model (at 18 GHz)

● Data, — Model



Linearity Measurements and Simulations

Measurements

• Measurements done on an ATN Load pull system

• Tuned for Maximum PAE/ Maximum Pout

• f1=5 GHz, f2 = 5.001 GHz

Simulations

• Harmonic Balance in RFDE (IC51)

• QPSS in Spectre (IC51)

• Simulated 2-tone distortion behavior

2f2-f1

f1 f2

3f2-2f1

• Linearity simulations provide a good check to the accuracy of

the model parameter set and model implementation in the simulator



pbii qq

qrr

Δ+=

0

00 fjcijeip qqqq +Δ+Δ=Δ, where

0qq p −≤Δ , leads to divide-by-zero errors

We addressed this in ADS2005 by limiting pqΔ

fjcijeip qqqq ++=Δ lim0qq p >Δ

,lim0qq p =Δ lim0qq p ≤Δ

when

when

Improved Rbi Implementation for Better Convergence-1

,0*lim0 qFq −=where .10 << F

A value of F=0.75 is found to provide a good trade-off between convergence and compatibility with the reference implementation.



[ ]22lim00lim00lim0

0

)(21 δ+−Δ+−Δ+=Δ

++=Δ

qqqqqq

qfqjciqjeiq

ppp

p

Hyperbolic smoothing is applied to the limiting function to eliminate derivative discontinuities at q0lim-

-

Improved Rbi Implementation for Better Convergence-2

Where δ controls the trade-off between smoothness and accuracy. A good choice for δ is d=1e-3*q0’ where q0’ is the parameter describing the temperature-independent part of q0. Derivative expressions are altered as follows:

dTqd

qdqd

dTdq

dqqd

dTqd p

p

ppp 0

0

lim0

lim0

..Δ

Δ

Δ+

Δ=

Δ,. 0

0 bei

p

p

p

bei

p

dVqd

dqqd

dVqd ΔΔ

=Δ



Inter Modulation Distortion: Measurement vs Model

0

10

20

30

40

-50 -30 -10 10

Pin (dBm)

Ic (m

A)

-0.1

0.3

0.6

0.9

1.2

Ib (m

A)

-40

-20

0

20

-50 -40 -30 -20 -10 0

Pin (dBm)

Pout

(dB

m)

Vbe

• Biased at Vbe = 0.86 V, Vce= 1.36V, corresponds to 35% of peak fTcurrent

• Self-biasing effect is well captured

• Bias dependence is well modeled

● Data, — Model

-120

-70

-20

30

-50 -30 -10 10Pin (dBm)

Pout

, IM

3, IM

5 (d

Bm

)



Summary

• Most of the important parameters can be extracted by simple

measurements and expressions and fitting routines

• Scaling of these parameters is technology and device layout

dependent

• Hicum 2.1 is able to describe the device characteristics

satisfactorily

Documents

Extraction of Scalable HiCuM Parameters and Verifikation ... · Extraction of Scalable HiCuM Parameters and Verifikation ... San Diego, San Diego, CA 92093 . ... Ick lim 2 lim 0 1