De-Embedding for Coupled Three-Port Devices

Yuandong Guo1, Bo Pu2, DongHyun Kim3, and Jun Fan4

Electromagnetic Compatibility Laboratory

2022 Asia-Pacific International Symposium on Electromagnetic Compatibility (APEMC) | 978-1-6654-1671-9/22/$31.00 ©2022 IEEE | DOI: 10.1109/APEMC53576.2022.9888499

Missouri University of Science and Technology

Rolla, Missouri, USA
1
ydggdd, 2bpdbh, 3dkim, 4jfan@mst.edu

Abstract—In many applications, the device under test (DUT) explained in Section II, which is extended for the de-
is embedded into a test setup. Various de-embedding techniques embedding of uncoupled and coupled three-port devices in
have been proposed to expose the real electrical behaviors of a Section III and IV, respectively. The proposed de-embedding
DUT, e.g., the traditional thru-reflect-line and short-open-load- algorithm is verified through the measurement-based studies
thru algorithms, where the T-matrix and its inverse form are performed on a vehicular three-phase braking system and given
adopted in the mathematical process. In the fields of in Section V. Section VI concludes this paper.
radiofrequency and electromagnetic compatibility, a DUT may
have three coupled ports, and the symmetry in the associated S-
matrix breaks down, because the numbers of entry and exist II. DE-EMBEDDING FOR ONE-PORT DEVICES
ports are not equal, which results in a non-square T-matrix based
upon the definitions. Given that the inverse expression of a non- A. A Two-Port Network and Its S-matrix
square matrix does not exist, the conventional de-embedding Given a common two-port network shown in Fig. 1, the
methods are not applicable for a coupled three-port network. In relationship between the incident waves ai and reflected waves
this paper, a de-embedding algorithm which is feasible for bi are correlated in the S-matrix expressed by
coupled three-port devices is proposed and verified through the
measurement-based studies. The de-embedding technique may  b1   S11 S12   a1 
also be applied on devices with more than three ports.  =   (1)
 b2   S 21 S 22   a2 
Keywords—coupling; de-embedding; S-parameter; T-matrix;
test fixture; three-phase motor; AC bus bar. Expanding the matrices in (1) yields
b1 = S11a1 + S12 a2 (2a)
I. INTRODUCTION
In many radiofrequency and electromagnetic compatibility b2 = S 21a1 + S 22 a2 (2b)
scenarios, a multi-port device under test (DUT) is embedded
Solving for a2 and b2 gives
into a test setup. In order to characterize the electrical
performances of the DUT, de-embedding is generally needed −1 −1
a2 = − [ S21 ] ⋅ S11 ⋅ a1 + [ S21 ] ⋅ b1 (3a)
during the post-processing phase.
The traditional de-embedding techniques [1][2] employ the b2 = ( S12 − S22 ⋅ ( S21 )−1 ⋅ S11 ) ⋅ a1 + S22 ⋅ ( S21 ) −1 ⋅ b1 (3b)
T-matrix and its inverse form of both the DUT and test fixture.
A network with even number of ports is symmetric, since it has
B. De-Embedding for One-Port Devices
the same number of entrance and exit ports. However, this
symmetry may break down for networks with odd number of Assuming a one-port network illustrated in Fig. 2, the
ports. In this case, their T-matrices are not square, and the “Total” is outlined by the dashed box, and the corresponding S-
inverse matrices do not exist [3]. Therefore, the conventional matrix, in which both a1 and b1 are included, is measurable. If
de-embedding techniques, e.g., thru-reflection-line [1], short- the “Fixture” can be characterized through its S-matrix, the
open-load-thru [2], and 2X-Thru algorithms [4][5], are only incident and reflected waves of the DUT, namely, b2 and a2,
feasible for networks with even number of ports. can be obtained through (3), since
In this paper, we present a de-embedding approach, which b2 = aDUT (4a)
can be applied to coupled three-port devices. This paper
extends the application of the de-embedding algorithm a2 = bDUT (4b)
documented in [6] by illustrating the mathematical derivations
and measurement-based validation. The proposed technique The S-parameter of the one-port DUT is:
can be further developed and applied to generic devices with bDUT a
S DUT = = 2 .
odd number of ports, which will be reported in future aDUT b2
publications.
The de-embedding method for one-port devices is briefly III. DE-EMBEDDING FOR UNCOUPLED THREE-PORT DEVICES
For the uncoupled three-port network sketched in Fig. 3,
This work is supported in part by the National Science Foundation under
Grant No. IIP-1916535.

978-1-6654-1671-9/22/$31.00 ©2022 IEEE 742

Authorized licensed use limited to: XIDIAN UNIVERSITY. Downloaded on March 14,2024 at 09:24:47 UTC from IEEE Xplore. Restrictions apply.
Fig. 1. A representation of common two-port networks with incident and
reflected waves.

Fig. 2. A one-port device with both DUT and “Fixture” exhibited.

Fig. 4. A coupled three-port network.

matrix, where the couplings are indicated by the non-diagonal

elements as suggested in (7).

b1'   S11' ' S11' S12' ' S1' 2 S13' ' S13'   a1' 
  S S12' S13'
 
b1   11' S11 S12 S13   a1 
b '   S ' ' S 21' S 2' 2' S 2' 2 S 2'3' S 2'3   a2' 
 2  =  21   (7)
b2   S 21' S 21 S 22' S 22 S 23' S 23   a2 
 '   
b3   S31' ' S31' S3'2' S3' 2 S3'3' S3'3   a3' 
b   S ' S31 S32

S33   a3 
 3   31 S32' S33'
Fig. 3. An uncoupled three-port network. The dashed circle mimics the
boundary of the “Total” structure, whose S-matrix can be acquired through Re-writing the S-matrix in (7) with six equations and
measurements. solving for the unprimed parameters, ai and bi, where i = 1, 2,
the unprimed coefficients, ai and bi, denote the incident and 3, yields
reflected waves of each fixture at the port adjacent to the DUT,
a1 = X1b1' + X 2b2' + X 3b3' + Y1a1' + Y2 a2' + Y3a3' (8a)
respectively. The primed parameters, ai’ and bi’, represent the
incident and reflected waves at the measurable ports of the
a2 = W1b1' + W2b2' + W3b3' + U1a1' + U 2 a2' + U 3a3' (8b)
Total structure, whose boundary is mimicked by the dashed
circle in Fig. 3.
a3 = J1b1' + J 2b2' + J 3b3' + K1a1' + K 2 a2' + K3a3' (8c)
It is evident that Fixture i, where i = 1, 2, 3, is a two-port
network. Similar to (3a) and (3b), ai and bi can be readily b1 = ( S11 X 1 + S12W1 + S13 J1 )b1' + ( S11 X 2 + S12W2 + S13 J 2 )b2'
depicted by ai’ and bi’ after expanding the associated S-matrix +( S11 X 3 + S12W3 + S13 J 3 )b3' + ( S11' + S11Y1 + S12U1 + S13 K1 )a1' (8d)
[6]:
' '
+( S12' + S11Y2 + S12U 2 + S13 K 2 ) a + ( S13' + S11Y3 + S12U 3 + S13 K 3 )a
2 3
ai = −[ Si'i ]−1 ⋅ Si'i' ⋅ ai' + [ Si'i ]−1 ⋅ bi' = A ⋅ ai' + B ⋅ bi' (5a)
b2 = ( S 21 X 1 + S 22W1 + S 23 J1 )b1' + ( S 21 X 2 + S 22W2 + S 23 J 2 )b2'
bi = ( Sii' − Sii ⋅ ( Si'i ) −1 ⋅ Si'i' ) ⋅ ai' + Sii ⋅ ( Si'i ) −1 ⋅ bi' = C ⋅ ai' + D ⋅ bi' (5b) + ( S 21 X 3 + S 22W3 + S 23 J 3 )b3' + ( S 21' + S 21Y1 + S 22U1 + S 23 K1 ) a1' (8e)
' '
Since the three ports are uncoupled, the matrices A, B, C, + ( S 22' + S 21Y2 + S 22U 2 + S 23 K 2 )a + ( S 23' + S 21Y3 + S 22U 3 + S 23 K 3 ) a
2 3
and D only contain diagonal coefficients. The S-matrix of the
DUT is derived as b2 = ( S31 X 1 + S32W1 + S33 J1 )b1' + ( S31 X 2 + S32W2 + S33 J 2 )b2'
+( S31 X 3 + S32W3 + S33 J 3 )b3' + ( S31' + S31Y1 + S32U1 + S33 K1 )a1' (8f)
[bDUT ] = ( A + B ⋅  S ' ) ⋅ ( C + D ⋅  S ' ) −1
S DUT = [ ] [ ]  [ ] [ ]  (6) +( S32' + S31Y2 + S32U 2 + S33 K 2 )a2' + ( S33' + S31Y3 + S32U 3 + S33 K3 )a3'
[ aDUT ]
where, S’ is the S-matrix of total and available through where the mathematical expressions of the coefficients Xi, Yi,
measurements. Wi, Ui, Ji, and Ki, where i = 1, 2, 3, are given in the Appendix.
Equations (8a) ~ (8f) can be manifested in the matrix form
IV. DE-EMBEDDING FOR COUPLED THREE-PORT DEVICES shown as
The general S-matrix expression manifesting the physical
relationship between the incident and reflected waves for the
coupled three-port device exhibited in Fig. 4 contains a 6×6

743

Authorized licensed use limited to: XIDIAN UNIVERSITY. Downloaded on March 14,2024 at 09:24:47 UTC from IEEE Xplore. Restrictions apply.
 M 16   a1 
'
b1   M 11 M 12 M 13 M 14 M 15
b   M  ' 
 2   21 M 22 M 23 M 24 M 25 M 26   a2 
b3   M 31 M 32 M 33 M 34 M 35 M 36   a3' 
 =   (9)
 a1   M 41 M 42 M 43 M 44 M 45 M 46  b1' 
a  M  
M 52 M 53 M 54 M 55 M 56  b2' 
 2   51 
 a3   M 61M 62 M 63 M 64 M 65 M 66  b ' 
 3
where the elements in the matrix [M] can be readily identified.
For the sake of simplicity, equation (9) can also be written as
(a)
b1   a1' 
b   '
 2  a2 
b3   M D M C   a3' 
 = (10)
 a1   M B M A  b1' 
a   '
 2 b2 
 a3   '
b3 
The sub-matrices MA, MB, MC, and MD are suggested by the (b)
dashed lines in (9). Fig. 5. (a) The top view of the PMSM under study, (b) the associated 3D
model.
Assuming the total structure is characterized by [S’], the S-
matrix of the DUT is determined through
[ S DUT ] = ([ M A ] + [ M B ][ S ' ]) ⋅ ([ M C ] + [ M D ][ S ' ])−1 (11)

V. MEASUREMENT-BASED VALIDATION
The device is a three-phase permanent magnet synchronous
motor (PMSM), which is utilized in a highly integrated
vehicular braking system [7], whose top view is demonstrated
in Fig. 5(a). Three ports are defined between each phase and
PMSM’s metallic housing for the VNA measurements, and the
resulting S-matrix is adopted to reproduce the electrical
characteristics in the frequency range of 100 kHz to 120 MHz. Fig. 6. The definitions of each ports and measurement configurations for the
AC bus bar.
The accurate SPICE model of the PMSM is constructed with
the three-phase AC bus bar incorporated, and the methodology
is documented in [8]. VI. CONCLUSION
In the 3D model illustrated in Fig. 5(b), where the multi- With the limitations of using T-matrix for de-embedding
layer printed circuit board (PCB) is about 1 mm above the AC explained, a de-embedding methodology suitable for coupled
bus bar but omitted in the figure due to confidentiality reasons, three-port devices is presented in this paper, which is
the parasitic capacitances from PCB nets and heatsinks to each developed based upon fundamental S-parameter concept and
bus bar are estimated in ANSYS Q3D [9]. It is found that the verified through a measurement-based case. The proposed de-
AC bus bar is tightly coupled to the PCB and heatsinks in the embedding algorithm can be further expanded for generic
frequency range of interest, which means it has to be utilized in devices, which will be reported and verified in future
the full-wave simulation model together with the PCB and de- publications.
embedded from the measured PMSM’s S-matrix for the
electromagnetic interference study. APPENDIX
The AC bus bar is cut at the junction point with the three- The expressions of the coefficients Xi, Yi, Wi, Ui, Ji, and Ki,
phase motor, and Fig. 6 shows the definitions of each port and where i = 1, 2, 3, are given as
the measurement configurations. The measurement is S2'3 S3'2 − S3'3 S2'2
conducted using Agilent E5071C network analyzer. The S- X1 = A.1
S3'2 ( S 2'3 S11' − S13' S21' ) + S12' ( S33' S21' − S2'3 S31' ) + S2'2 ( S13' S31' − S3'3 S11' )
matrix of the three-phase motor is also measured, which is
regarded as the baseline data. S3'3 S12' − S13' S3'2
X2 = A.2
With the proposed de-embedding approach implemented, S3'2 ( S 2'3 S11' − S13' S 21' ) + S1'2 ( S3'3 S 21' − S 2'3 S31' ) + S 2'2 ( S13' S31' − S3'3 S11' )
the good agreements shown in Fig. 7 verify the algorithm.

744

Authorized licensed use limited to: XIDIAN UNIVERSITY. Downloaded on March 14,2024 at 09:24:47 UTC from IEEE Xplore. Restrictions apply.
 S11' ( S3'3 S 2'2' − S 2'3 S3'2' ) + S31' ( S 2'3 S12' ' − S13' S 2'2' ) + S 21' ( S13' S3'2' − S 21' S1'2' )
U2 = A.11
S3'2 ( S 2'3 S11' − S13' S 21' ) + S1'2 ( S3'3 S 21' − S 2'3 S31' ) + S 2'2 ( S13' S31' − S3'3 S11' )

S11' ( S3'3 S2'3' − S 2'3 S3'3' ) + S31' ( S2'3 S13' ' − S13' S 2'3' ) + S 21' ( S13' S3'3' − S 21' S13' ' )
U3 = A.12
S3'2 ( S 2'3 S11' − S13' S21' ) + S1'2 ( S3'3 S 21' − S2'3 S31' ) + S2'2 ( S13' S31' − S3'3 S11' )

S 2'2 S31' − S3'2 S 21'

J1 = A.13
S3'2 ( S2'3 S11' − S13' S21' ) + S1'2 ( S3'3 S21' − S 2'3 S31' ) + S 2'2 ( S13' S31' − S3'3 S11' )

Magnitude of S11 Magnitude of S12 S3'2 S11' − S31' S12'

J2 = A.14
S3'2 ( S 2'3 S11' − S13' S 21' ) + S1'2 ( S3'3 S 21' − S2'3 S31' ) + S2'2 ( S13' S31' − S3'3 S11' )

S 21' S1'2 − S11' S 2'2

J3 = A.15
S3'2 ( S2'3 S11' − S13' S21' ) + S1'2 ( S3'3 S21' − S 2'3 S31' ) + S 2'2 ( S13' S31' − S3'3 S11' )

S12' ( S31' S 21' ' − S 21' S31' ' ) + S3'2 ( S 21' S11' ' − S11' S 21' ' ) + S 2'2 ( S11' S31' ' − S31' S11' ' )
K1 = A.16
S3'2 ( S 2'3 S11' − S13' S21' ) + S12' ( S33' S 21' − S 2'3 S31' ) + S 2'2 ( S13' S31' − S33' S11' )

S1'2 ( S31' S 2'2' − S21' S31' ' ) + S3'2 ( S 21' S1'2' − S11' S21' ' ) + S2'2 ( S11' S3'2' − S31' S1'2' )
Magnitude of S22 K2 = A.17
Magnitude of S13 S3'2 ( S2'3 S11' − S13' S 21' ) + S1'2 ( S3'3 S21' − S2'3 S31' ) + S2'2 ( S13' S31' − S3'3 S11' )

S1'2 ( S31' S2'3' − S21' S3'3' ) + S3'2 ( S 21' S13' ' − S11' S2'3' ) + S2'2 ( S11' S3'3' − S31' S13' ' )
K3 = A.18
S3'2 ( S 2'3 S11' − S13' S 21' ) + S1'2 ( S3'3 S 21' − S 2'3 S31' ) + S 2'2 ( S13' S31' − S3'3 S11' )

REFERENCES
[1] L. Ye et al., "Thru-Reflect-Line Calibration Technique: Error Analysis
for Characteristic Impedance Variations in the Line Standards," in IEEE
Transactions on Electromagnetic Compatibility, vol. 59, no. 3, pp. 779-
788, June 2017.
Magnitude of S23 Magnitude of S33 [2] H. Cho and D. E. Burk, "A three-step method for the de-embedding of
Fig. 7. The comparisons between the de-embedded and baseline data, which high-frequency S-parameter measurements," in IEEE Transactions on
verify the proposed de-embedding approach. Electron Devices, vol. 38, no. 6, pp. 1371-1375, June 1991.
[3] J. Frei, X. -D. Cai and S. Muller, "Multiport $S$ -Parameter and $T$ -
Parameter Conversion With Symmetry Extension," in IEEE
Transactions on Microwave Theory and Techniques, vol. 56, no. 11, pp.
S13' S2'2 − S2'3 S1'2 2493-2504, Nov. 2008.
X3 = A.3
S3'2 ( S2'3 S11' − S13' S21' ) + S12' ( S3'3 S21' − S 2'3 S31' ) + S 2'2 ( S13' S31' − S3'3 S11' ) [4] B. Chen et al., "Differential S-Parameter De-embedding for 8-Port
Network," 2018 IEEE Symposium on Electromagnetic Compatibility,
S1'2 ( S2'3 S31' ' − S3'3 S 21' ' ) + S3'2 ( S13' S 21' ' − S 2'3 S11' ' ) + S 2'2 ( S3'3 S11' ' − S13' S31' ' ) Signal Integrity and Power Integrity (EMC, SI & PI), 2018, pp. 52-56.
Y1 = A.4
S3'2 ( S 2'3 S11' − S13' S 21' ) + S1'2 ( S3'3 S 21' − S 2'3 S31' ) + S2'2 ( S13' S31' − S3'3 S11' ) [5] B. Chen, J. He, Y. Guo, S. Pan, X. Ye and J. Fan, "Multi-Ports
([$2^{n}$) 2×-Thru De-Embedding: Theory, Validation, and Mode
S12' ( S 2'3 S3'2' − S3'3 S 2'2' ) + S3'2 ( S13' S 2'2' − S 2'3 S12' ' ) + S 2'2 ( S3'3 S12' ' − S13' S3'2' ) Conversion Characterization," in IEEE Transactions on Electromagnetic
Y2 = A.5 Compatibility, vol. 61, no. 4, pp. 1261-1270, Aug. 2019.
S3'2 ( S 2'3 S11' − S13' S 21' ) + S12' ( S3'3 S 21' − S 2'3 S31' ) + S2'2 ( S13' S31' − S33' S11' )
[6] B. Pu, J. Kim, and W. Nah, “A de-embedding technique of a three-port
network with two ports coupled,” in Journal of Electromagnetic
S12' ( S2'3 S3'3' − S3'3 S2'2' ) + S3'2 ( S13' S2'3' − S2'3 S13' ' ) + S2'2 ( S3'3 S13' ' − S13' S3'3' )
Y3 = A.6 Engineering and Science, vol. 15, no.4, pp. 258-265, Oct.2015.
S3'2 ( S2'3 S11' − S13' S21' ) + S1'2 ( S3'3 S21' − S 2'3 S31' ) + S2'2 ( S13' S31' − S3'3 S11' ) [7] Y. Guo et al., "Ground Bridge Effect on Reduction of Conducted
Emission from Three-Phase Motor Drive System," 2019 International
S2'3 S31' − S3'3 S21' Symposium on Electromagnetic Compatibility - EMC EUROPE, 2019,
W1 = A.7
S3'2 ( S2'3 S11' − S13' S21' ) + S12' ( S3'3 S21' − S 2'3 S31' ) + S 2'2 ( S13' S31' − S3'3 S11' ) pp. 854-859.
[8] Y. Guo et al., "High-Frequency Modeling of Permanent Magnet
S13' S31' − S3'3 S11' Synchronous Motor Considering Internal Imbalances," 2021 Asia-
W2 = A.8 Pacific International Symposium on Electromagnetic Compatibility
S3'2 ( S2'3 S11' − S13' S21' ) + S12' ( S3'3 S21' − S 2'3 S31' ) + S2'2 ( S13' S31' − S3'3 S11' )
(APEMC), 2021, pp. 1-3.
S2'3 S11' − S13' S21' [9] C. Che, H. Zhao, Y. Guo, J. Hu and H. Kim, "Investigation of
W3 = A.9 Segmentation Method for Enhancing High Frequency Simulation
S3'2 ( S 2'3 S11' − S13' S 21' ) + S12' ( S3'3 S 21' − S 2'3 S31' ) + S2'2 ( S13' S31' − S3'3 S11' ) Accuracy of Q3D Extractor," 2019 IEEE International Conference on
Computational Electromagnetics (ICCEM), 2019, pp. 1-3.
S11' ( S3'3 S 21' ' − S 2'3 S31' ' ) + S31' ( S 2'3 S11' ' − S13' S 21' ' ) + S 21' ( S13' S31' ' − S 21' S11' ' )
U1 = A.10
S3'2 ( S 2'3 S11' − S13' S 21' ) + S12' ( S33' S 21' − S 2'3 S31' ) + S 2'2 ( S13' S31' − S33' S11' )

745

Authorized licensed use limited to: XIDIAN UNIVERSITY. Downloaded on March 14,2024 at 09:24:47 UTC from IEEE Xplore. Restrictions apply.