2008 IEEE INTERNATIONAL RF AND MICROWAVE CONFERENCE PROCEEDINGS 2008 IEEE INTERNATIONAL RF AND MICROWAVE CONFERENCE PROCEEDINGS 2008 IEEE INTERNATIONAL RF AND MICROWAVE CONFERENCE PROCEEDINGS 2008 IEEE INTERNATIONAL RF AND MICROWAVE CONFERENCE PROCEEDINGS December 2December 2December 2December 2----4, 2008, Kuala Lumpur, MALAYSIA4, 2008, Kuala Lumpur, MALAYSIA4, 2008, Kuala Lumpur, MALAYSIA4, 2008, Kuala Lumpur, MALAYSIA

978-1-4244-2867-0/08/$25.00 ©2008 IEEE

R F

M 08

A Hybrid OPEN Model Technique for Accurate Device Characterization

Wen-Lin Chen

1,2, Sheng-Fuh Chang

1, Kun-Ming Chen

2, Kuo-Hsiang Liao

2, Guo-Wei Huang

2,

Jen-Chung Chang3

1 Department of Electrical Engineering, the Department of Communications Engineering, National Chung Cheng University,

No.168, University Rd., Min-Hsiung Chia-Yi, Taiwan, R.O.C.. 2 National Nano Device Laboratories, No. 26, Prosperity Road I, Science-based Industrial Park, Hsinchu, Taiwan, R.O.C..

3 United Microelectronics Corporation.

wlchen@ndl.org.tw, ieesfc@ccu.edu.tw, kmchen@ndl.org.tw.

Abstract - A hybrid OPEN model, which consists of

equivalent circuit model and behavior transmission

line (TL) model, is proposed to accurately extract

device characteristics. The proposed method requires

only one OPEN dummy structure so the silicon

occupation area is efficiently reduced. The other

benefit is it avoids the errors, which is occurred while

performing the classical de-embedding procedure. A

Ka-band CPW filter is fabricated in 130 nm RF CMOS

technique to prove this method is efficient up to 67

GHz.

Keywords: model, transmission line, Ka-band, coplanar

waveguide (CPW), CMOS.

1. Introduction

Demand for more precise and reliable device

models in describing the RF behavior has lead to

numerous de-embedding techniques. Most classical

de-embedding techniques rely on the equivalent lump

circuit model and base on the cascade concept [1]-[3].

The measured s-parameters of a device-under-test

(DUT) will be transferred to y-parameters and

subtracted by the y-parameters of the open dummy [1].

Afterward the subtracted data will be transferred to z-

parameter and subtracted by the z-parameter of the

interconnections. This correction is only valid when

assuming the interconnecting transmission line is

modeled as a lossy inductance. Consequently, this de-

embedding technique is limited for high frequency

applications due to the inconsiderate equivalent circuit

model.

In order to improve the accuracy, the three and four

steps method were proposed. The three steps de-

embedding method uses “open” ”thru” “short1” and

“short2” dummy structures to extract the device

parameters [2]. Furthermore, the extra “simple open”

and “simple short” dummy structures are required for

the four steps de-embedding technique [3]. Although

the accurate results are obtained, it accompanies with

the increase of wafer occupation areas. The open-line

method was proposed in [4] and it efficiently reduces

the number of dummy structures. However, this

method requires a set of accurate technological process

parameters (metal and oxide thickness, permittivity

etc.) for electromagnetic (EM) simulation.

Unfortunately, the process parameters are frequency

dependent such that some modeling errors may be

induced, especially for the wideband application.

Moreover, the effect of the discontinuity between the

probing pads and the metal interconnections is

neglected.

In order to avoid the de-embedding errors, a novel

device characterization method by combining the

dummy model with the device model is proposed. The

effect of probing contact, pad capacitance,

discontinuity and the interconnecting transmission line

are well modeled such that the proposed method is

accurate and applicable to wideband applications. In

this paper, the CPW TL was characterized from 100

MHz up to 67 GHz by using the proposed method. In

order to validate the proposed method, a Ka-band filter

Fig. 1 Schematic diagram representing the device

combined with the proposed hybrid OPEN model.

DeviceDeviceDeviceDevice

MMMM8888

MMMM7777

MMMM6666

MMMM1111

MMMM5555

SubstrateSubstrateSubstrateSubstrate

RRRRCCCC LLLLCCCC

CCCCpadpadpadpad

EquivalentEquivalentEquivalentEquivalent

Circuit ModelCircuit ModelCircuit ModelCircuit Model

BehaviorBehaviorBehaviorBehavior

TL ModelTL ModelTL ModelTL Model

CCCCdisdisdisdis

AAAA

Hybrid OPEN modelHybrid OPEN modelHybrid OPEN modelHybrid OPEN model

AAAA

RRRRviaviaviavia

LLLLviaviaviavia

BBBB

BBBB

206

is designed using the extracted CPW TLs model. The

method has been validated in the case of 130-nm RF

CMOS technology and is proved to be efficient up to

67 GHz.

2. Methodology

This section introduces the procedure of the

proposed device characterization method as follows.

1) Generate the hybrid OPEN model: The OPEN

dummy is modeled by using the equivalent circuit

networks and the behavior TL model, as shown in

Fig. 1. The construction of the equivalent circuit

networks depends on the layout structure and

application frequency. In this study, the effect of

probing contact, pad capacitance and

discontinuity are described with the lump

elements. On the other side, the interconnecting

TL is modeled by using the TLINP behavior

model, as shown in Fig. 2, provided by ADS

simulator.

2) Extract parameters for the hybrid OPEN model:

The initial values of the equivalent circuit model

and the TLINP model are extracted based on EM

simulation results. After fine-tuning these initial

values, an accurate hybrid OPEN model can be

constructed. Fig. 3 shows the result by comparing

the measurement results of the OPEN dummy to

the simulation data of the hybrid OPEN model.

3) Construct a model for device: Model selection

depends on the type of device under

characterization. For example, an equivalent

small-signal circuit model is used to describe the

behavior of MOSFET. The initial value of the

extrinsic and intrinsic parameters of the small-

signal model can be estimated by using the “Y-

parameter” method proposed in [6]. In this work,

the CPW TL was chosen as the device under

characterization and it was characterized by using

the TLINP model. The initial value for the TLINP

model was extracted based on the EM simulation

results

4) Combine the device model with the hybrid OPEN

model: Because the accurate hybrid OPEN model

is developed in step 2, only the parameters of the

device model can be fine-tuned. The optimization

Frequency (GHz)

0 10 20 30 40 50 60 70

-1.0

-0.8

-0.6

-0.4

-0.2

0.0

0.2

Real

Image

S11

Frequency (GHz)

0 10 20 30 40 50 60 70

-1.2

-0.8

-0.4

0.0

0.4

0.8

1.2

Image

Real

S21

(a)

(b)

Figure 4: S-parameters modeling results versus

measurements up to 67 GHz for a CPW TL.

Z

L

A

TanD

K

:

:

:

:

:

characteristic impedance

physical length

attenuation

dielectric loss tangent

effective dielectric constant

Figure 2: TLINP model for CPW TLs.

meas.

simu.

Figure 3: The prediction and measurement s-

parameters for the OPEN dummy.

207

was terminated until the simulated s-parameters of

the combined model agreed with the measured s-

parameters. In this study, the characteristic

impedance Z0, effective dielectric constant K,

attenuation coefficient A and loss tangent TanD

were fine-tuned. Fig. 4 depicts the good

agreement between simulation results and

measurement data.

5) Remove the hybrid OPEN model: In accordance

with the above steps, an accurate device model

can be obtained after removing the hybrid OPEN

model.

3. Results and Discussion

In this work, the top metal (M8) and the via-array

connected metals (M1-M8) were used to generate the

signal pad and the ground pad, separately. According

to this layout structure, a shunt capacitance Cpad was

used to model the coupling effect of the signal pad to

the ground pads and the dielectric substrate. Not only

the capacitance Cpad was considered, but also the

contact and discontinuity effects were well defined.

The discontinuity effect between the probing pad and

interconnecting TL was modeled using a capacitance

Cdis [5]. It is noted that each equivalent sub-circuit in

the hybrid model is expandable, and therefore the

proposed model is valid for high frequency and

wideband applications. In order to accurately capture

the behavior of the interconnecting TL, the TLINP

behavior model was adopted instead of the

inconsiderate equivalent circuit model. Using TLINP

model saves the simulation time and makes the

optimization easier that is very helpful for device

modeling and circuit design. In this case, two kinds of

CPW TLs (Spacing between center conductor and

ground plane, S= 2 um; center conductor width, W=

10 um and S= 8 um; W= 10 um) were designed as the

device under characterization. In order to substantiate

the extracted device model, a Ka band band-pass filter

was implemented in 130 nm RF CMOS technology, as

shown in Fig. 5. The CPW TL (S= 2 um, W= 10) is

used to generate the open stub line and the

interconnecting lines. Because bends and cross-

junctions are not specially modeled, the straight CPW

TLs are adopted for the circuit design. The measured

and simulated performances of the bandpass filter are

plotted in Fig. 6. Despite the cross-junction effect is

not well defined, the prediction performances shifts

slightly when compared to the measurement results.

Thus, the accurate prediction results reveal the

proposed method is effective in high frequency device

characterization. In addition, the hybrid model is directly connected to the I/O ports of devices such that

this method is also expected for the multi-ports device

characterization.

4. Conclusions

An accurate characterization procedure for on-

wafer device has been presented based on the proposed

hybrid OPEN model. Only the OPEN dummy is

required for device characterization such that the

silicon occupation area is efficiently reduced. This

method combines the TL behavior model with the

expandable equivalent circuit model, thus the

Frequency (GHz)

0 10 20 30 40 50 60 70

S parameters (dB)

-40

-30

-20

-10

0

meas.

simu.

S11 S22

S21

Frequency (GHz)

0 10 20 30 40 50 60 70

Group delay (ps)

-20

0

20

40

60

meas.

simu.

(a)

Fig. 6 Simulation and measurement results of a Ka-

band CPW filter (a) s-parameters, (b) group delay.

(b)

Figure 5: Schematic of a Ka-band CPW filter.

Hybrid OPEN

Model

Hybrid OPEN

Model

S= 2 um

W =10 um

S= 8 um

W =10 umL =311 umL =311 um

L =107 um L =120 um L =113 um

L =311 um L =311 um

208

prediction of the circuit performance can be efficient

and accurate. Compared to the classical methods, the

most difference is made in the interconnecting TL

model. Hence, we believe that the TL model

dominates the accuracy of the de-embedding methods.

A Ka-band CPW filter fabricated using 130 nm RF

CMOS technology has been verified up to 67 GHz.

The agreement of the simulation and the measurement

results demonstrates the proposed method is well

suited for the high frequency device characterization.

References

[1] M. C. A. M. Koolen, J. A. M. Geelen, and M. P. J.

G. Versleijen, “An improved de-embedding

technique for on-wafer high-frequency

characterization,” Proceedings of the 1991 IEEE

Bipolar Circuits Technology Meeting,

Minneapolis, Minnesota, USA, September 1991,

pp. 188–191.

[2] H. Cho and D. E. Burk, “A three-step method for

the de-embedding of high-frequency S-parameter

measurements,” IEEE Trans. Electron Devices,

vol. 38, no. 6, pp. 1371–1371, Jun. 1991.

[3] T. E. Kolding, “A four-step method for de-

embedding gigahertz on-wafer CMOS

measurements,” IEEE Trans. Electron Devices,

vol. 47, no. 4, pp. 734–740, Apr. 2000.

[4] Andrei, C., Gloria, D., Danneville, F., Dambrine,

G.,“Efficient De-Embedding Technique for 110-

GHz Deep-Channel-MOSFET Characterization,”

IEEE Microwave and Wireless Components

Letters, vol. 17, no. 4, pp. 301–303, April 2007.

[5] R. N. Simons and G. E. Ponchak, “Modeling of

some coplanar waveguide discontinuities,” IEEE

Trans. Microw. Theory Tech., vol. 36, no.12, pp.

1796–1803, Dec. 1988.

[6] S. C. Wang, G. W. Huang, “A Practical Method to

extract Extrinsic Parameters for the Silicon

MOSFET Small-signal Model,” Workshop on

Compact Modeling 2004.

209