1

Louay Degachi and Fadhel M. Ghannouchi, Senior Member, IEEE

Systematic and Rigorous Extraction Method of
HBT Small-Signal Model Parameters


Abstract— This paper presents a systematic and rigorous

analytical parameter-extraction method for a heterojunction
bipolar transistor (HBT) small-signal equivalent-circuit
model. The proposed method relies exclusively on S-parameter
measurements. Exact closed-form equations are used for the
direct extraction of circuit elements. The method is
characterized by its simplicity and ease of implementation. It
is applied to predict the small-signal characteristics of
transistors from different foundries. Excellent agreement
between modeled and measured S-parameters is observed up
to 20 GHz.

Index Terms— Heterojunction bipolar transistor, small-

signal equivalent-circuit model, parameter extraction.

Fig. 1. Adopted HBT small-signal equivalent-circuit model.

I. INTRODUCTION

A key issue in HBT modeling is the availability of an given. In section III, the extraction method is described in
accurate and systematic model-parameter extraction detail. Section IV presents the extraction results. Section V
procedure. In the last decade, several analytical methods concludes the paper.
were proposed for the parameter extraction of HBT small-
signal models [1]-[8]. Most of these methods make use of
approximations when deriving appropriate equations II. MODEL DESCRIPTION
allowing for the direct determination of model parameters. The adopted HBT small-signal equivalent-circuit model
Of particular interest is the method developed by is shown in fig. 1. Parasitic capacitances (Cpbe, Cpce and Cpbc)
Bousnina et al. [5], where small-signal hybrid-π model and inductances (Le, Lb and Lc) are extracted from S-
parameters are extracted rigorously and no approximations parameter measurements when the transistor is biased
are used. However, certain shortcomings are noticed. On respectively in the cut-off and in the open-collector
the one hand, the derivation of the intrinsic base-emitter conditions [5]. Rb, Rc and Re are the series resistances and
resistance makes use of the formula Rbe = nKT/qIbe they are determined from fly-back measurements. All these
involving parameters such as junction temperature and parasitic elements are de-embedded from S-parameter
junction emission coefficient, and thus requiring a prior dc- measurements following the procedure described by
parameter extraction. On the other hand, the base-emitter Dambrine et al. [9]. After de-embedding, the circuit is
capacitance is derived as the solution of a 2 nd degree reduced to the one shown in fig. 2. Cbe is the intrinsic base-
equation. Such an equation has generally two roots, and emitter capacitance. Cbc and Ccx are the intrinsic and
additional arguments are necessary to eliminate one of extrinsic base-collector capacitances, respectively. Rbe and
them. Rbc are the intrinsic base-emitter and base-collector
This paper presents an improved analytical extraction resistances, respectively. Generally Rbc has a high value and
method, compared to [5]. The proposed method is rigorous, therefore its effect might be neglected. Nevertheless, we
simple and relies exclusively on S-parameter include it in our model for sake of completeness. Gm and
measurements. Therefore, it is very suited for automation. Gmo are the small-signal and dc transconductances,
In section II, a brief description of the adopted model is respectively. Rbi is the intrinsic base resistance. τ is the
delay time.

III. PARAMETER EXTRACTION PROCEDURE

First, the well-known T-П transformation is applied as

L. Degachi and F. M. Ghannouchi are with the Poly-GRAMES

Research Center, Electrical Engineering Department, École Polytechnique
de Montréal, C.P. 6079, succ. Centre-ville, Montréal, Qc, Canada H3C
3A7 (e-mail: [email protected])
 2

Fig. 4. Final circuit after T-П transformation.

Fig. 2. HBT small-signal equivalent-circuit model after de-embedding.

Fig. 3. T-П transformation.

where
described in fig. 3. The parameters Z1, Z2 and Z3 are given
by: (10)

(1) (11)

(2) It follows that:

(3) (12)
where
(13)
(4)
(14)
(5)

(15)
(6)

At this point the circuit parameters are determined

(7) analytically as follows.

The final circuit is shown in fig. 4 where

A. Determination of Rbi /Rbc and Rbi x Cbc
From (1) and (3), it can be written that
(8)

(16)
(9)

or equivalently
The parameters Z1, Z3 and Z4 are related to Y-parameters as
follows [5]:
(17)
 3

(18)

Rbi /Rbc and Rbi x Cbc are determined from the least-squares
lines fitting to real(Z1/Z3) and imag(Z1/Z3), respectively,
considered as functions of the angular frequency ω.
Equations (17) and (18) are illustrated in figs. 5 and 6.

B. Determination of Rbe x Cbe

Using relations (1) and (4)-(7), Z1 can be written as:

Fig. 5. Plot of real (Z1/Z3) vs. frequency

or equivalently

(19)

where

(20)

(21)

Fig. 6. Plot of imag (Z1/Z3) vs. frequency

(22)

(26)
From (19), it can be deduced that
A and B are determined from the least-squares line fitting to
F1, considered as function of ω2. Equation (23) is illustrated
in fig. 7. Then, from (24) and (25), Tbe is determined as

or equivalently,

(23)
C. Determination of R and R x T
where From (19), it can be written that

(24)
or equivalently

(25) (27)
(28)

R and R x T are determined from the least-squares lines

 4

Fig. 8. Plot of real (F2) vs. frequency

Fig. 7. Plot of F1 vs. frequency

fitting to real(F2) and imag(F2), respectively, considered as

functions of the ω. Equations (27) and (28) are illustrated in
figs. 8 and 9.

D. Determination of Rbe and Rbi

From equations (20) and (21), the following linear
system can be written:

Fig. 9. Plot of imag (F2) vs. frequency

or in a matrix form
From (10), it can be written that

(30)

This gives directly Rbe and Rbi. (31)

E. Determination of Rbc, Cbc and Cbe Then, Ccx, Gmo and τ are determined from the least-squares
Once Rbe and Rbi are determined, it becomes lines fitting respectively to imag(1/Z4 – 1/Z2), |X/B| and
straightforward to calculate Rbc, Cbc and Cbe respectively [–phase(X/(Gmo B))]. Equations (29), (30) and (31) are
from Rbi/Rbc, Rbi x Cbc and Rbe x Cbe (= Tbe). illustrated in figs. 10, 11 and 12, respectively.

F. Determination of Ccx and Gmo and τ IV. RESULTS AND DISCUSSION

Once the values of the intrinsic elements are calculated,
In order to validate and assess the accuracy of the extraction
the parameters Z2 and B are determined from equations (2)
procedure, three transistors from two different foundries
and (11), respectively. From (8) and (9), it can be written
were investigated. Measurements were performed with a
that
microwave probing system and a vector network analyzer
(VNA) over the frequency range 1-20 GHz. The extraction
procedure was implemented as a Matlab program. The first
(29)
transistor was an AlGaAs/GaAs HBT with a 2 x 20 μm 2
emitter area. Figs. 13 and 14 show comparisons between
 5

S(2,1) / 15

S(1,2) X 25

S(3,4)*25

S(1,2)*25
S(4,3)/15

S(2,1)/15
S(3,3)*2

S(1,1)*2
S(4,4)

S(2,2)
S(2,2)

Fig. 10. Plot of imag (1/Z4 – 1/Z2) vs. frequency

S(1,1) X 2

Fig. 13. Measured (o) and simulated (-) S-parameters for a 2 x 20 μm 2

freq
emitter-area HBT (Vce(1.000GHz to 20.00GHz)
= 3 V, Ic = 18.728 mA, Ib = 240 μA)

S(2,1) / 15

S(1,2) X 25
S(3,4)*25

S(1,2)*25
S(4,3)/15

S(2,1)/15
S(3,3)*2

S(1,1)*2
S(4,4)

S(2,2)

Fig. 11. Plot of mag (X/B) vs. frequency

S(2,2)

S(1,1) X 2

Fig. 14. Measured

freq (o) and simulated
(1.000GHz to (-) S-parameters for a 2 x 20 μm 2
20.00GHz)
emitter-area HBT (Vce = 2 V, Ic = 19.302 mA, Ib = 240 μA)

TABLE I
Extracted small-signal parameters for a 2 x 20 μm2 emITTER-AREA HBT
Vce = 2 V, Ib = 240 μA, Vce = 3 V, Ib = 240 μA,
Circuit element
Ic = 19.302 mA Ic = 18.728 mA
Rbi [ohm] 1.97 2.49
Fig. 12. Plot of [- phase (X/B.Gmo)] vs. frequency Rbe [ohm] 115.87 129.32
4
Rbc [ohm] 3.39 x 10 4.5 x 104
Cbe [fF] 2849 3058
measured and simulated S-parameters at the bias point Vce = Cbc [fF] 6.335 5.514
3 V, Ic = 18.728 mA and Ib = 240 μA, and at the bias point Ccx [fF] 52.28 44.47
Gmo [S] 0.755 0.701
Vce = 2 V, Ic = 19.302 mA and Ib = 240 μA, respectively.
τ [ps] 1.71 1.62
Table I shows the extracted values of the different circuit
elements. The second transistor was a GaInP/GaAs HBT
with a 1x10 μm2 emitter area. Figs. 15 and 16 show at the bias point Vce = 3 V, Ic = 5 mA and Ib =24.72 μA,
comparisons between measured and simulated S-parameters respectively. Table II depicts the extracted values of the
at the bias point Vce = 2 V, Ic = 5 mA and Ib = 24.5 μA, and different elements. The third transistor was a GaInP/GaAs
 6

TABLE II
Extracted small-signal parameters for a 1 x 10 μm2 EMITTER-AREA HBT
Vce = 2 V, Ic = 5 mA, Vce = 3 V, Ic = 5 mA, Ib
Circuit element
Ib = 24.50 μA = 24.72 μA
Rbi [ohm] 13.5 13.2
Rbe [ohm] 1509.8 1562.4
Rbc [ohm] 3.67 x 105 2.94 x 105
Cbe [fF] 358 428.6
S(2,1)
S(1,2) x 200 Cbc [fF] 3.368 2.969
Ccx [fF] 6.724 6.705
S(1,2)*200

Gmo [S] 0.1393 0.1383

S(2,2)*10

S(1,1)*5
S(2,1)

τ [ps] 1.41 1.11

S(2,2) x 10

S(1,1) x 5

S(2,1) / 40

Fig. 15. Measured (o) and simulated (-) S-parameters for a 1 x 10 μm 2 S(1,2) x 25
S(3,4)*25

S(1,2)*25
freq(V(500.0MHz to 20.00GHz)
S(4,3)/40

emitter-area HBT ce = 2 V, Ic = 5 mA, Ib = 24.5 μA)S(2,1)/40

S(4,4)

S(3,3)
S(2,2)

S(1,1)

S(1,1) S(2,2)

S(2,1)
S(1,2) x 200
S(1,2)*200
S(2,2)*10

S(1,1)*5
S(2,1)

Fig. 17. Measured (o) and simulated (-) S-parameters for a 1 x 100 μm 2
freq
emitter-area HBT (Vce (1.000GHz to 20.00GHz)
= 2 V, Ic = 30 mA, Ib = 215.742 μA)

S(2,2) x 10

S(1,1) x 5

S(2,1) / 40
S(1,2) x 25
2
Fig. 16. Measured (o) and simulated (-) S-parameters for a 1 x 10 μm
S(3,4)*25

S(1,2)*25
S(4,3)/40

S(2,1)/40

freq((V(500.0MHz
emitter-area HBT ce = 3 V, Ic =5 to
mA,20.00GHz)
Ib = 24.72 μA)
S(4,4)

S(3,3)
S(2,2)

S(1,1)

HBT with a 1x100 μm2 emitter area. Figs. 17 and 18 show

comparisons between measured and simulated S-parameters
S(2,2)
at the bias point Vce = 2 V, Ic = 30 mA and Ib = 215.742 μA, S(1,1)
and at the bias point Vce = 3 V, Ic = 50 mA and Ib = 355.760
μA, respectively. Table III depicts the extracted values of
the different elements. Excellent agreement was obtained
over the selected range of frequencies. The deviations
occurring in S12 at high frequencies (figs. 15-18) are
attributable to measurement problems. As expected the
value of Rbc is very high. The evolutions with bias of Rbe and Fig. 18. Measured (o) and simulated (-) S-parameters for a 1 x 100 μm 2
freq
emitter-area HBT ((Vce(1.000GHz to 20.00GHz)
= 3 V, Ic =50 mA, Ib = 355.760 μA)
Gmo, for the 1x10 transistor, are given in figs. 19 and 20,
respectively. Both parameters show smooth behaviors
which further validate the proposed extraction technique.
 7

TABLE III 0.35

Extracted small-signal parameters for a 1 x 100 μm2 EMITTER-AREA HBT Vce = 2 v
Vce = 2 V, Ic = 30 mA, Vce = 3 V, Ic = 50 mA, Ib 0.3 Vce = 3 v
Circuit element
Ib = 215.742 μA = 355.760 μA
Rbi [ohm] 2.05 1.88 0.25
Rbe [ohm] 149.31 78.209
Rbc [ohm] 2 x 105 8.04 x 104
0.2
Cbe [fF] 3559 5926

Gmo
Cbc [fF] 63.34 51.21
Ccx [fF] 40.03 28.51 0.15
Gmo [S] 1.0465 1.8396
τ [ps] 1.03 0.555 0.1

0.05
14000

0
12000 0 2000 4000 6000 8000 10000 12000
Vce = 2 v
Ic (uA)
Vce = 3 v
10000 Fig. 20. Plot of Gmo vs Ic for the 1x10 transistor

8000
Rbe (Ohm)

REFERENCES
6000
[1] B. Li, S. Parsad, “Basic expressions and approximations in small-
signal parameter extraction for HBT’s,” IEEE Trans. Microwave
4000 Theory and Techn., vol. 47, no. 5, pp. 534–539, May 1999.
[2] A. Ouslimani, J. Gaubert, H. Hafdallah, A. Birafane, P. Pouvil, H.
2000 Leier, “Direct extraction of linear HBT-model parameters using nine
analytical expression blocks,” IEEE Trans. Microwave Theory and
0
Techn., vol. 50, no. 1, pp. 218–221, Jan. 2002.
0 20 40 60 80 100 120 [3] H. C. Tseng, J. H. Chou, “An efficient analytical approach for
Ib (uA) extracting the emitter inductance of collector-up HBT’s,” IEEE
Fig. 19. Plot of Rbe vs Ib for the 1x10 transistor Trans. Electron Devices, vol. 51, no. 7, pp. 1200–1202, July 2004.
[4] H. C. Tseng, J. H. Chou, “A pure analytic method for direct
extraction of collector-up HBT’s small-signal parameters,” IEEE
Trans. Electron Devices, vol. 51, no. 12, pp. 1972–1977, Dec. 2004.
V. CONCLUSION [5] S. Bousnina, P. Mandeville, A. B. Kouki, R. Surridge, F. M.
Ghannouchi, "Direct parameter-extraction method for HBT small-
In this paper, a simple and systematic parameter-
signal model," IEEE Trans. Microwave Theory and Techn., vol. 50,
extraction method is presented. The main features of this no. 2, pp. 529–536, Feb. 2002.
method are the followings. [6] M. Rudolph, R. Doerner, P. Heymann “Direct extraction of HBT
1. The method relies exclusively on S-parameter equivalent-circuit elements,” IEEE Trans. Microwave Theory and
Techn., vol. 47, no. 1, pp. 82–84, Jan. 1999.
measurements. No DC parameter is required. [7] B. Sheinman, E. Wasige, M. Rudolph, R. Doerner, V. Sidorov, S.
2. The extraction of all elements is achieved Cohen, D. Ritter “A peeling algorithm for extraction of the HBT
rigorously by skillfully deriving a number of small-signal equivalent circuit,” IEEE Trans. Microwave Theory and
Techn., vol. 50, no. 12, pp. 2804–2810, Dec. 2002.
relations without employing any approximation. [8] D. Costa, W. U. Liu, J.S. Harris “Direct extraction of the
These features make the method very suited for automation. AlGaAs/GaAs heterojunction bipolar transistor small-signal
Excellent agreement is obtained between modeled and equivalent circuit ,” IEEE Trans. Electron Devices, vol. 38, no. 9, pp.
2018–2024, Sept. 1991.
measured S-parameters, for three transistors from two [9] G. Dambrine, A. Cappy, F. Heliodore, E. Playez, “A new method for
different foundries. The results indicate the accuracy and determining the FET small-signal equivalent circuit,” IEEE Trans.
consistency of this method. Microwave Theory and Techn., vol. 36, no. 7, pp. 1151–1159, July
1988.