paper special section on analog circuits and related soc ... · 812 ieice trans. electron.,...

812IEICE TRANS. ELECTRON., VOL.E93–C, NO.6 JUNE 2010

PAPER Special Section on Analog Circuits and Related SoC Integration Technologies

A De-Embedding Method Using Different-Length TransmissionLines for mm-Wave CMOS Device Modeling

Naoki TAKAYAMA†a), Kota MATSUSHITA†b), Shogo ITO†c), Nonmembers,Ning LI†d), Student Member, Keigo BUNSEN†e), Nonmember, Kenichi OKADA†f),

and Akira MATSUZAWA†g), Members

SUMMARY This paper proposes a de-embedding method for on-chipS-parameter measurements at mm-wave frequency. The proposed methoduses only two transmission lines with different length. In the proposedmethod, a parasitic-component model extracted from two transmissionlines can be used for de-embedding for other-type DUTs like transistor,capacitor, inductor, etc. The experimental results show that the error incharacteristic impedance between the different-length transmission lines isless than 0.7% above 40 GHz. The extracted pad model is also shown.key words: de-embedding, S-parameter measurement, mm-wave, RFCMOS, transmission line

1. Introduction

Recently, the research on mm-wave CMOS circuits be-comes a hot topic [1], [2]. CMOS processes are employedfor developing mm-wave wireless systems because of thelow cost fabrication comparing with compound semicon-ductors [3]. In mm-wave circuits, even small parasitic ca-pacitors and inductors will considerably affect the total cir-cuit performance. Therefore, it is needed to build accu-rate models of the components including parasitic elements,such as transistors, capacitor, transmission line, and so on.

Before building models of them, the process, de-embedding, is needed to remove parasitic components frommeasurement data because measurement data include theparasitic components of the contact pads. A variety of de-embedding methods for on-chip measurement have beenproposed [4], [5]. There is one of the most common methodscalled open-short method [6]. The method, however, cannotremove parasitic components of the contact pads completelyin high frequency because it is difficult to fabricate an idealshort pattern.

Some de-embedding methods using only line patternswithout short pattern were proposed [7]–[9]. One of themuses a thru pattern [7]. However, the measurement resultof the thru pattern has a large uncertainty and is not re-liable because the value of the S-parameter is too small.

Manuscript received October 14, 2009.Manuscript revised January 20, 2010.†The authors are with the Department of Physical Electronics,

Tokyo Institute of Technology, Tokyo, 152-8552 Japan.a) E-mail: [email protected]) E-mail: [email protected]) E-mail: [email protected]) E-mail: [email protected]) E-mail: [email protected]) E-mail: [email protected]) E-mail: [email protected]

DOI: 10.1587/transele.E93.C.812

The length of thru pattern must be short because the mea-surement data is converted to a π-type lumped-componentcircuit. On the other hand, too short thru-pattern causeslarger measurement error. According to the probe manufac-turer (Cascade Microtech, Inc.), RF probes should be sep-arated more than 200 μm. Too close probes have larger in-terference, and it results in larger measurement error [10].Thus, the de-embedding result is also not reliable. The de-embedding method using only long line patterns is proposedin [8]. However, this method considers only the shunt par-asitic component of the contact pads and the series par-asitic component are ignored. Moreover, this method isonly applicable to de-embedding of a transmission line, notfor other test element groups. In addition, there is a de-embedding method using two transmission lines [9], andthe method is evaluated up to 110 GHz with experimentalresults using a 60 GHz power amplifier [11]. However, themethod in [9] has the limitation that one transmission linehas to be exactly twice as long as the other.

The new technique proposed in this paper can removenot only the shunt parasitic components of the pads but alsothe series one by using two transmission lines with differentlength [12]. In addition, this method can be applied to othertest element groups, such as transistors, capacitors, and soon, by building a model of the contact pads.

2. Conventional Methods

2.1 Open-Short De-Embedding

At first, we introduce Open-Short De-embedding notedabove [6]. This method builds the model of the contactpads and the ground plane, as shown in Fig. 1, from themeasurement data of an open pattern and a short pattern.The equivalent circuits of the patterns are shown in Fig. 2.The shunt parasitic components (YP1, YP2, YP3) are removedfrom the measurement data by subtracting the Y-parameter

Fig. 1 Equivalent circuit used in the open-short de-embedding.

Copyright c© 2010 The Institute of Electronics, Information and Communication Engineers

TAKAYAMA et al.: A DE-EMBEDDING METHOD USING DIFFERENT-LENGTH TRANSMISSION LINES FOR MM-WAVE CMOS DEVICE MODELING813

of the open pattern. As the same procedure, the shunt para-sitic components of a short pattern can be removed. There-fore, the matrix consisting of only series parasitic compo-nents (ZS1, ZS2, ZS3) is obtained. By subtracting this matrix,the matrix of the DUT is obtained.

However this method has an issue at high frequencieslike mm-wave frequency. It is impossible to fabricate idealopen and short patterns due to parasitic capacitance and in-ductance, and the non-ideality causes considerable degrada-tion in the de-embedding accuracy, especially at the mm-wave frequency.

2.2 Thru-Only De-Embedding

Next, we introduce the thru-only de-embedding [7]. Fig. 3shows the structure of a transmission line and the equivalentcircuit of the measurement data. This method uses only ashort thru pattern. Figure 4 shows the structure of a shortthru and the equivalent circuit. The π-type equivalent circuitis utilized to characterize the measurement data. The padmodel is built by separating the equivalent circuit into twosymmetric parts.

However, this method has an issue. The length of theshort thru has to be as short as possible because the equiv-alent circuit consists of lumped components. However, ifthe length is short, the distance between the probes is tooclose. Therefore, the measurement data become poorly-reproducible and unreliable.

Fig. 2 Equivalent circuits. (a) Open pattern. (b) Short pattern.

Fig. 3 Transmission line. (a) Structure. (b) Equivalent circuit.

Fig. 4 Short thru. (a) Structure. (b) Equivalent circuit.

3. Proposed Method

3.1 Multi-Line De-Embedding

As explained in the previous sections, the conventionalopen-short and thru-only methods are not so accurate atmm-wave frequencies, which is caused by the non-idealityof open- and short-patterns and the inaccuracy of theshort thru pattern. Thus, this paper proposes a novel de-embedding method using two long line patterns [12]. In thismethod, first the transmission line is characterized by us-ing the measurement data. The parasitic components of thecontact pads are calculated and the pad model is built. Thestructure of the contact pad and the equivalent circuit areshown in Fig. 5. The circuit consists of 4 components andthe values of them are frequency dependent. The pad modelbuilt by two transmission lines can be utilized to de-embedthe pad parasitics of other-type DUTs such as transistors,capacitors, inductors, etc. The procedure of the proposedmethod is as follows.

3.2 De-Embedding Procedure

Before building the model of the contact pads, the transmis-sion line has to be characterized. Measurement data of twotransmission lines with a length of �1 and �2, where �1 < �2,are used in this method. While the method in [9] requiresthe condition �2 = 2�1, the proposed method can utilize var-ious length of transmission lines. Separating the segments ofcontact pads from the intrinsic line, the T-parameter matrixof test structure �i, Tm

�i, can be expressed by the following

equation.

Tm�i= TPL · T�i · TPR, (1)

where

T�i represents the T-parameter of the intrinsic line �i,TPL represents the T-parameter of the left pad, andTPR represents the T-parameter of the right pad.

Subsequently TX1 is defined as multiplying Tm�2

by the in-verse of Tm

�1(Fig. 6).

TX1 = Tm�2· [Tm

�1]−1

= TPL · T�2 · TPR · T−1PR · T−1

�1· T−1

PL

Fig. 5 Contact pad model. (a) Structure. (b) Equivalent circuit.


Fig. 6 Multiplying the matrices of the transmission line.

Fig. 7 Canceling the pad parasitic ZP.

= TPL · T�1 · T−1PL (2)

T-parameter of this matrix TX1 is transformed to the Y-parameter matrix YX1. A parallel combination of YX1 anda port-swapped version of itself, Swap(YX1), is defined asYX2. Thus, we can cancel the effect of the pad parasitic ZP

as shown in Fig. 7.

YX2 = YX1 + Swap(YX1)

=

[YX1_11 YX1_12

YX1_21 YX1_22

]+

[YX1_22 YX1_21

YX1_12 YX1_11

](3)

YX3 is defined as the Y-parameter matrix of the intrin-sic TL connected with the pad series parasitic component ZS

and the negative parameter of itself −ZS . Fig. 8 shows theequivalent circuit of YX3. Assuming the structure of the in-trinsic TL is perfectly symmetric, the Y-matrix of the in-trinsic TL which length is �21(= �2 − �1), YT L_�21 , can beexpressed by Eq. (4). By using Eq. (6), YX3 is approxi-mated by Eq. (5). ytoz() is a transformation function fromY-parameter to Z-parameter.

YT L �21 =

[YT L_1 �21 YT L 2 �21

YT L_2 �21 YT L 1 �21

](4)

YX3 = ztoy(ytoz(YT L �21 ) +

[ZS 00 −ZS

])

�[

YT L 1 �21 − ZS (YT L 1 �212 − YT L 2 �21

2)YT L 2 �21

YT L 2 �21

YT L 1 �21 + +ZS (YT L 1 �212 − YT L 2 �21

2)

](5)

The following assumption is used.

Z2S �

1YT L 1 �21

2 − YT L 2 �212

(6)

Therefore, YX2 can be obtained as the twofold ofYT L �21 as explained in Eq. (7). By the procedure asmentioned above, the Y-parameter matrix of the intrinsicTL YT L �21 is obtained.

YX2 = YX3 + Swap(YX3)

Fig. 8 The equivalent circuit of YX3.

=

[2YT L 1 �21 2YT L 2 �21

2YT L 2 �21 2YT L 1 �21

]= 2YT L �21 (7)

YT L �1 can be obtained from YT L �21 by using Eq. (8).

YT L �1 = f toy((yto f (YT L �21 ))n) (8)

�1 = n�21 (9)

This method mentioned above is the same as [8]. Forde-embedding of other DUTs such as transistors, capaci-tors, inductors, etc, the pad parasitics of ZS and ZP are de-rived, which are frequency-dependent parameters. The Y-parameter matrix of the measurement data, Ym

T L �1, of the

TL is expressed by YT L �1 in Eq. (10). ZP is a shunt com-ponent at both ports, so it is derived from the Y-parametermatrix.

YmT L �1

= ztoy

(ytoz(YT L �1 ) +

[ZS 00 ZS

])

+

[ 1ZP

00 1

ZP

]=

⎡⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎣ZS + Z′1�Z′2

+1

ZP

Z′2�Z′2

Z′2�Z′2

ZS + Z′1�Z′2

+1

ZP

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎦(10)

�Z′2 = (ZS + Z′1)2 − Z′22

Z′1 =YT L 1 �1

YT L 1 �12 − YT L 2 �1

2, Z′2 =

YT L 2 �1

YT L 1 �12 − YT L 2 �1

2

The Y-parameter matrix YX4 is defined as expressed inEq. (11). If Eq. (13) is satisfied, the shunt parasitic ZP isobtained by adding Y(1,1) and Y(1,2) of YX4 as expressedin Eq. (12).

YX4 = YmT L �1− YT L �1 (11)

YX4 11 + YX4 12 � 1ZP

(12)∣∣∣∣∣∣ 1ZP · (YT L 1 �1 + YT L 2 �1 )

·(1 +

1ZS · (YT L 1 �1 + YT L 2 �1 )

)∣∣∣∣∣∣ � 1 (13)

To use the approximations, Eqs. (6) and (13), the para-sitic components in the contact pad should be small as com-pared with the transmission line.

Next, we evaluate the pad series parasitic compo-nent ZS . By subtracting the shunt parasitics from the mea-surement data of the TL, we can obtain YX5 that is the pa-rameter of the intrinsic TL with the series parasitic compo-nents as expressed in Eq. (14).

YX5 = YmT L −

⎡⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎣1

ZP0

01

ZP

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎦ (14)


Here, ZS is derived from YX5. It is assumed thatYX5 expresses a reciprocal and uniform transmission line.The resistance and the inductance per length are defined asRip( f ) and Lip( f ), respectively. These value can be calcu-lated from YX5 [13], and the values are frequency depen-dent. As shown in Fig. 9, the total series resistance Ri( f )and inductance Li( f ) can be expressed by multiplying theline length �i as follows.

Li( f ) = �iLip( f ) (15)

Ri( f ) = �iRip( f ) (16)

The following assumption is used (See Appendix).

|γ�i| � 1, (17)

where γ is the propagation constant of the transmission line.This procedure is implemented by using two transmis-

sion lines with different length. The values of the calculatedinductances are plotted on a graph as shown in Fig. 10(a).Then the line connecting these two points is drawn. The in-tercept of the line, L0( f ), is the inductance of the series para-sitic components of the contact pad because Li( f ) is the onlycomponent of the pad when the line length is zero. L0( f ) canbe calculated by Eq. (18).

L0( f ) =�2L1( f ) − �1L2( f )

�2 − �1 (18)

In this case, the measurement data of two transmissionlines are used. However, by using growing number of trans-mission lines in different length, the more accurate value ofL0( f ) can be obtained. In that case, the offset value can bederived by the method of least squares.

The same procedure is applied to the resistance. Thegraph of the resistance is shown in Fig. 10(b). R0( f ) can becalculated by Eq. (19).

Fig. 9 The resistance and inductance of YX5.

Fig. 10 The graph of total series components. (a) Inductances. (b)Resistances.

R0( f ) =�2R1( f ) − �1R2( f )

�2 − �1 (19)

The series parasitic components of the pad, ZS , is de-fined by the following equation.

ZS =12· (R0( f ) + jωL0( f )) (20)

The parasitic components, ZS and ZP, can be obtainedby Eq. (20) and Eq. (12). This pad model can be utilized forthe de-embedding of other-DUTs.

4. Experimental Results

The evaluations of the conventional and proposed meth-ods are applied by using the measurement data of transmis-sion lines with different length. Twelve-metal-layer 65 nmCMOS process is used. The structure of transmission linesis a coplanar strip line with a bottom ground as shown inFig. 11. In this experiment, the lengths of the transmissionlines are 200 μm and 400 μm. Because the probe manufac-turer (Cascade Microtech, Inc.) recommends the distancebetween probes should be more than 200 μm. Besides, ifthe length is long as compared with the wave length, theλ/4 resonance occurs and it has effect on the measurementdata. The structure and size of the contact pad are shownin Fig. 12. The size of signal pad is minimized as possible,40 μm × 60 μm, to reduce parasitic inductance and capaci-tance. Therefore, the pad model can be assumed as lumpedparameters as shown in Fig. 5. Fig. 13 shows the chip mi-crograph of the transmission lines with the contact pads.

Figure 14 shows the calculated results by the proposedand conventional methods. After de-embedding, the char-acteristic impedance Z0 is calculated from S-parameter byEq. (21).

Z20 = Z2

nom

(1 + S 11)2 − S 212

(1 − S 11)2 − S 212(21)

where Znom represents the normalized impedance, and itis 50Ω in this case. Figure 14(a) shows characteristicimpedances of 200 μm-long and 400 μm-long transmissionlines, which are de-embedded by the open-short method.Figure 14(b) shows a result by the thru-only method, andFig. 14(c) shows a result by the proposed method.

The characteristic impedance should be equal to each

Fig. 11 The structure of TL.


Fig. 12 The structure of pad.

Fig. 13 The chip micrograph of TL.

Fig. 14 The calculated result. (a) Open-Short. (b) Thru-only. (c)Proposed.

other even if it is calculated from different-length transmis-sion lines. However, the open-short method and thru-onlymethod have large error in characteristic impedance, whichis caused by the error in the de-embedding calculation. Inaddition, Z0 calculated by the thru-only method has the vari-ation caused by the measurement error of the thru pattern,

Fig. 15 The values of the transmission line per length. (a) Inductance.(b) Resistance. (c) Capacitance. (d) Conductance.

which is caused by the too short thru pattern, 40 μm.Using the proposed method, the characteristic imped-

ances of the transmission lines agree with each other asshown in Fig. 14(c). In the thru-only method, the error be-tween characteristic impedances is 1∼4% above 40 GHz.Meanwhile, in the proposed method, the error is less than0.7%. This result demonstrates the accuracy of the proposedde-embedding method.

The values per length of the transmission lines calcu-lated from YT L �21 are shown in Fig. 15. Figure 16 showsthe pad parasitics calculated by the proposed method. Fig-ures 16(a)(c) show the series inductance and the shunt ca-pacitance, and they are almost constant with frequency. Theinductance is about 13 pH, and the capacitance is about20 fF. Figures 16(b)(d) show the series resistance and theshunt conductance, and they increase with frequency be-cause of skin effect and frequency-dependence of dielectricloss. The value of R0 is unstable around 60 GHz. The rea-son is that the measurement of resistance is difficult at highfrequency. However the inductance and the capacitance aredominant at high frequencies, so the error in R0 is not soimportant.

The extracted values shown in Fig. 16 are reasonableaccording to the actual structure of the contact pad, and ap-proximate values can be calculated as follows. By calculat-ing from the value of intrinsic transmission line, the induc-tance of 30 μm-length transmission line is about 9 pH. Thehalf length of the contact pad is 30 μm, so we can expectthat its inductance is nearly 9 pH. If the probes are put oncenter of the contact pads, the DC resistance of the contactpad is about 0.1Ω. It is nearly equal to the value of R0.Calculating from process parameters, the capacitance of the


Fig. 16 The pad parasitics calculated by the proposed method. (a) Se-ries inductance. (b) Series resistance. (c) Shunt capacitance. (d) Shuntconductance.

Fig. 17 The left terms in (a) Eq. (22) and (b) Eq. (23).

contact pad is about 11 fF. There are dummy metals underthe contact pad, so 19 fF of C0 is a reasonable value. Thetangent-delta values calculated from the transmission lineand the pad model are 0.04 and 0.08, respectively.

In this way, the pad model can be built and be also ap-plied for de-embedding other DUTs, e.g., transistors, capac-itors, etc.

Next, the effect of the approximation, Eqs. (6) and (13),in the proposed method is discussed. Equations (6) and (13)can be converted to Eqs. (22) and (23), respectively. Thereal and imaginary parts in the left terms in Eqs. (22) and(23) are shown in Fig. 17.∣∣∣∣∣∣Z

2S

Z20

∣∣∣∣∣∣� 1 (22)

∣∣∣∣∣∣ZS · ZP · (YT L 1 �1 + YT L 2 �1 )2

1 + ZS · (YT L 1 �1 + YT L 2 �1 )

∣∣∣∣∣∣ � 1 (23)

The absolute values of the real and imaginary parts areless than 0.02 among the entire frequency. Thus, the approx-imation is valid in this case.

5. Conclusion

A new de-embedding method using two long lines is pro-posed. Using this method, the accurate de-embedding is re-alized, and it can be applied to transistors, capacitors, etcas well as transmission lines. The proposed de-embeddingmethod is accurate even at mm-wave frequencies because itdoes not utilize the open- and short-patterns and the shortthru pattern used in the conventional methods. The conven-tional de-embedding patterns cause the de-embedding errorat mm-wave frequencies. The de-embedding of the trans-mission lines is demonstrated. In the experimental results,the error in characteristic impedance between the different-length transmission lines is less than 0.7% above 40 GHz forthe proposed method while two conventional methods havelarger error.

Acknowledgement

This work was supported by MIC, NEDO, STARC, VLSIDesign and Education Center (VDEC), the University of To-kyo in collaboration with Cadence Design Systems and Agi-lent Technologies Japan, Ltd. and Fujitsu Laboratories Ltd.,Japan.

References

[1] A. Parsa and B. Razavi, “A 60 GHz CMOS receiver using a 30 GHzLO,” IEEE International Solid-State Circuits Conference Digest ofTechnical Papers, pp.190–191.606, Feb. 2008.

[2] J.F. Buckwalter and J. Kim, “A 26 dB-gain 100 GHz Si/SiGe cas-caded constructive-wave amplifier,” IEEE International Solid-StateCircuits Conference Digest of Technical Papers, pp.488–489, Feb.2009.

[3] Y. Hamada, M. Tanomura, M. Ito, and K. Maruhashi, “A high gain77 GHz power amplifier operating at 0.7V based on 90 nm CMOStechnology,” IEEE Microwave and Wireless Components Letters,vol.19, no.5, pp.329–331, May 2009.

[4] P.J.V. Wijnen, H.R. Claessen, and E.A. Wolsheimer, “A newstraightforward calibration and correction procedure for “on-wafer”high frequency s-parameter measurements (45 MHz-18 GHz),”Proc. Bipolar/BiCMOS Circuits and Technology Meeting, pp.70–73, Sept. 1987.

[5] L.F. Tiemeijer, R.J. Havens, A.B.M. Jansman, and Y. Bouttement,“Comparison of the “pad-open-short” and “open-short-load” deem-bedding techniques for accurate on-wafer RF characterization ofhigh-quality passives,” IEEE Trans. Microw. Theory Tech., vol.53,no.2, pp.723–729, Feb. 2005.

[6] M. Koolen, J. Geelen, and M. Versleijen, “An improved de-embedding technique for on-wafer high-frequencycharacterization,”Proc. Bipolar/BiCMOS Circuits and Technology Meeting, pp.188–191, Sept. 1991.

[7] H. Ito and K. Masu, “A simple through only de embedding methodfor on wafer s parameter measurements up to 110 GHz,” IEEE MTT-S Int. Microw. Symp. Dig., pp.383–386, June 2008.

[8] A.M. Mangan, S.P. Voinigescu, M.T. Yang, and M. Tazlauanu, “De-embedding transmission line measurements for accurate modelingof IC designs,” IEEE Trans. Electron Devices, vol.53, no.2, pp.235–241, Feb. 2006.

[9] J. Song, F. Ling, G. Flynn, W. Blood, and E. Demircan, “A de-embedding technique for interconnects,” Electrical Performance of


Electronic Packaging, pp.129–132, Oct. 2001.[10] L. Hayden, “Calibration errors when neglecting crosstalk,” Proc.

ARFTG Microwave Measurements Conference, Dec. 2005.[11] N. Li, K. Matsushita, N. Takayama, S. Ito, K. Okada, and A.

Matsuzawa, “Evaluation of a multi-line de-embedding techniqueup to 110 GHz for millimeter-wave CMOS circuit design,” IEICETrans. Fundamentals, vol.E93-A, no.2, pp.431–439, Feb. 2010.

[12] N. Takayama, K. Matsushita, S. Ito, N. Li, K. Bunsen, K. Okada, andA. Matsuzawa, “A multi-line de-embedding technique for mm-waveCMOS circuits,” Proc. IEEE Asia Pacific Microwave Conference,Dec. 2009.

[13] W.R. Eisenstadt and Y. Eo, “S-parameter-based IC interconnecttransmission line characterization,” IEEE Trans. Components andManufacturing Technology, vol.15, no.4, pp.483–490, 1992.

Appendix

In this appendix, it is shown that the series parasiticimpedance ZS can be calculated from the series impedanceper length Z0γ.

The Z-parameter of �i-length transmission line is de-fined as ZT L �i . ZT L �i is expressed as shown in Eq. (A· 1).

ZT L �i = ytoz(YT L �i )

=

⎡⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎣

YT L 1 �i

Y2T L 1 �i

− Y2T L 2 �i

−YT L 2 �i

Y2T L 1 �i

− Y2T L 2 �i

−YT L 2 �i

Y2T L 1 �i

− Y2T L 2 �i

YT L 1 �i

Y2T L 1 �i

− Y2T L 2 �i

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎦(A· 1)

=

[A BB A

], (A· 2)

where A and B are defined as follows.

A =YT L 1 �i

Y2T L 1 �i

− Y2T L 2 �i

=Z0 cosh(γ�i)

sinh(γ�i)(A· 3)

B =−YT L 2 �i

Y2T L 1 �i

− Y2T L 2 �i

=Z0

sinh(γ�i)(A· 4)

The characteristic impedance Z0 and the propagation con-stant γ are expressed by using A and B [13].

Z20 = A2 − B2 (A· 5)

eγ�i =AB+

√A2 − B2

B(A· 6)

In addition, the resistance Rip( f ) and the inductance Lip( f )per length are expressed by the following equations.

Lip( f ) =Im[Z0γ]

2π f(A· 7)

Rip( f ) = Re[Z0γ] (A· 8)

Next, the influence of the series parasitic impedanceZS is calculated. The total series impedance is expressed byZ0γ�i, and the difference by ZS is defined as �(Z0γ�i). Inthis case, the term A becomes A + ZS . The difference canbe expressed by the sum of the differences of Z0 and γ�i as

follows.

�(Z0γ�i)Z0γ�i

=�Z0

Z0+�(γ�i)γ�i

(A· 9)

Here, �Z0/Z0 can be obtained by the following equa-tions.

�Z0

Z0=

√(A + ZS )2 − B2 − √A2 − B2

Z0

=

√1 +

2AZS

Z20

+Z2

S

Z20

− 1 (A· 10)

� AZS

Z20

=ZS

Z0 tanh(γ�i)(A· 11)

where the following assumption is used.∣∣∣∣∣∣Z2S

Z20

∣∣∣∣∣∣ � 1 (A· 12)

Then, the difference of �(γ�i) is calculated as follows.

�(γ�i)γ�i

=ln X′ − ln X

ln X

=ln X′

X

ln X, (A· 13)

where X and X′ are defined as follows.

X = eγ�i =AB+

√A2 − B2

B(A· 14)

X′ = eγ′�i =

A + ZS

B+

√(A + ZS )2 − B2

B(A· 15)

The difference of �(γ�i)/(γ�i) can be calculated by the fol-lowing equations.

X′

X=

A + ZS +√

(A + ZS )2 − B2

A +√

A2 − B2

� 1 +ZS

Z0(A· 16)

�(γ�i)γ�i

=

lnX′

Xγ�i

(A· 17)

=

ln

(1 +

ZS

Z0

)γ�i

(A· 18)

� ZS

Z0γ�i(A· 19)

From Eqs. (A· 9), (A· 11) and (A· 19), the followingequation can be obtained.

�(Z0γ�i) = ZS

(1 +

γ�itanh(γ�i)

)(A· 20)

� 2ZS , (A· 21)

where the following assumption is used.


|γ�i| � 1. (A· 22)

Under the assumptions Eqs. (A· 12) and (A· 22), Eq. (A· 21)means that the series parasitic impedance ZS can be derivedfrom the product of the impedance per length Z0γ and theline length �i.

Naoki Takayama received the B.E. degreein Electrical and Electronic Engineering fromTokyo Institute of Technology, Tokyo, Japan, in2008. He is currently perusing the M.E. degreein Tokyo Institute of Technology, Tokyo, Japan.

Kota Matsushita received the B.E. degreein Electrical and Electronic Engineering fromTokyo Institute of Technology, Tokyo, Japan, in2009. He is studying toward the M.E. degreein Department of Physical Electronics, TokyoInstitute of Technology, Tokyo, Japan. His re-search interests include RF circuit design.

Shogo Ito received the B.E. degree in Elec-trical and Electronic Engineering from TokyoInstitute of Technology, Tokyo, Japan, in 2008.He is studying toward the M.E. degree in De-partment of Physical Electronics, Tokyo Insti-tute of Technology, Tokyo, Japan. His researchinterests include RF circuit design.

Ning Li received the B.S. degree in Elec-tronics Engineering and the M.S. degree inPhysical Electronics from Xi’an Jiaotong Uni-versity, China in 1999 and 2002. In 2002she joined Department of Electronics and Infor-mation Engineering, Xi’an Jiaotong University,Xi’an, China. Currently she is studying for herPh.D. degree in Tokyo Institute of Technology,Tokyo, Japan.

Keigo Bunsen received the B.E. degree inElectrical and Electronic Engineering from To-kyo Institute of Technology, Tokyo, Japan, in2009. He is studying toward the M.E. degreein Department of Physical Electronics, TokyoInstitute of Technology, Tokyo, Japan. His re-search interests include RF circuit design.

Kenichi Okada received the B.E., M.E.and Ph.D. degrees in Communications andComputer Engineering from Kyoto University,Kyoto, Japan, in 1998, 2000, and 2003, re-spectively. From 2000 to 2003, he was a Re-search Fellow of the Japan Society for the Pro-motion of Science in Kyoto University. From2003 to 2007, he worked as an Assistant Profes-sor at Precision and Intelligence Laboratory, To-kyo Institute of Technology, Yokohama, Japan.Since 2007, he has been an Associate Professor

at Department of Physical Electronics, Tokyo Institute of Technology, To-kyo, Japan. He has authored or co-authored more than 150 journal andconference papers. His current research interests include reconfigurableRF CMOS circuits for cognitive radios, 60 GHz CMOS RF frontends, andlow-voltage RF circuits. He is a member of IEEE, the Information Pro-cessing Society of Japan (IPSJ), and the Japan Society of Applied Physics(JSAP).

Akira Matsuzawa received B.S., M.S.,and Ph.D. degrees in electronics engineeringfrom Tohoku University, Sendai, Japan, in 1976,1978, and 1997 respectively. In 1978, he joinedMatsushita Electric Industrial Co. Ltd. Sincethen, he has been working on research anddevelopment of analog and Mixed Signal LSItechnologies; ultra-high speed ADCs, intelligentCMOS sensors, RF CMOS circuits, digital read-channel technologies for DVD systems, ultra-high speed interface technologies for metal and

optical fibers, a boundary scan technology, and CAD technology. He wasalso responsible for the development of low power LSI technology, ASIClibraries, analog CMOS devices, SOI devices. From 1997 to 2003, he wasa general manager in advanced LSI technology development center. OnApril 2003, he joined Tokyo Institute of Technology and he is a profes-sor on physical electronics. Currently he is researching in mixed signaltechnologies; CMOS wireless transceiver, RF CMOS circuit design, dataconverters, and organic EL drivers. He served the guest editor in chief forspecial issue on analog LSI technology of IEICE transactions on electronicsin 1992, 1997, and 2003, the vice-program chairman for International Con-ference on Solid State Devices and Materials (SSDM) in 1999 and 2000,the Co-Chairman for Low Power Electronics Workshop in 1995, a memberof program committee for analog technology in ISSCC and the guest editorfor special issues of IEEE Transactions on Electron Devices. He has pub-lished 26 technical journal papers and 46 international conference papers.He is co-author of 8 books. He holds 34 registered Japan patents and 65 USand EPC patents. In April 2003 he joined, as an Professor, the Departmentof Physical electronics at Titech. He received the IR100 award in 1983,the R&D100 award and the remarkable invention award in 1994, and theISSCC evening panel award in 2003 and 2005. He is an IEEE Fellow since2002.

paper special section on analog circuits and related soc ... · 812 ieice trans. electron.,...

Documents