Content uploaded by Qazi Mashaal Khan
Author content
All content in this area was uploaded by Qazi Mashaal Khan on Aug 26, 2022
Content may be subject to copyright.
Akram Ramezani
Dept. EMC Engineering/ Electronic
Engineering
Melexis Technologies NV/ KU Leuven
Tessenderlo, Belgium/ Leuven,
Belgium
rrk@melexis.com
Qazi Mashaal Khan
Dept. Electrical Engineering and
Control Engineering
ESEO / INSA Rennes
Angers, France / Rennes, France
qazimashaal.khan@eseo.fr
Hugo Pues
Dept. EMC Engineering
Melexis Technologies NV
Tessenderlo, Belgium
hpu@melexis.com
Abstract— In this paper, a new TRL/TRM calibration
method is described and compared to an electronic calibration
module (ECal) method which is widely used in industry. This
method needs a specifically made de-embedding board but does
not require an expensive ECal or any special precision board-
level calibration devices. The method is applied to an automotive
sensor interface IC showing the new calibration method enables
us to conduct accurate on-board S-parameter measurements up
to 4 GHz whereas the other method becomes inaccurate above
500 MHz.
Keywords— VNA, S-parameter measurements, calibration,
thru-reflect-line (TRL), short-open-load-thru (SOLT)
I. INTRODUCTION
Characterization of electromagnetic systems by their
scattering parameters is the most suitable tool for full wave
analysis and EMC problem simulations [1]. By using a Vector
Network Analyzer (VNA), scattering parameters (S-
parameters) can be measured over frequency. However, as in
reality making perfect hardware is not possible, it is necessary
to do a measurement calibration before S-parameter
measurement.
Among a variety of calibration methods, the most
commonly known are SOLT [2] and TRL [3] which are based
on 8-term or 12-term error models. SOLT is an abbreviation
for Short, Open, Load and Thru. It measures one transmission
standard (T) and three reflection ones (SOL). TRL is an
abbreviation for Thru, Reflect, and Line and it measures two
transmission (T and L) and one reflection (R) standards to
determine the error coefficients. Different calibration
techniques such as TRM (Thru, Reflect, Match), LRL (Line,
Reflect, Line), LRM (Line, Reflect, Match) belong to the TRL
family. An overview is given in [4].
Each of these calibration techniques has its advantages and
disadvantages depending on frequency range and application.
While the math behind SOLT is simpler than the TRL, the
latter -
defined standards whereas a limited knowledge of the
calibration standards does not impact the accuracy of the TRL
method [5]. TRL standards only need to be representative and
repeatable without being precisely known [6].
This paper presents a new on-board calibration method
which is based on the TRM and TRL methods. In comparison
to an ECal method which basically is an automatic SOLT
calibration technique that enables to calibrate at the PCB
connector level, this is a manual method that enables to
calibrate at the level of the device under test (DUT) pins
resulting in more accuracy at higher frequencies. This method
is mainly useful for testing on-board DUTs such as SMD
components and PCB circuits. In this paper we used both
calibration techniques and did S-parameter measurements on
a DUT to compare the two methods. The measurement result
shows that the new calibration method is accurate up to 4 GHz
whereas the ECal result becomes inaccurate above 500 MHz.
On the other hand, a disadvantage of this method is that it
is a manual calibration and needs more time compared to the
electronic module. While there may be other TRL calibration
structures which can be as accurate as this one, this method
has proven to be very convenient in practice.
The structure of the paper is the following: in section II, a
description of the method and the theory behind it are
presented. In section III, the method is verified by measuring
the S-parameters of a test chip that was produced in a previous
R&D project (i.e. an automotive sensor interface chip that was
called TOLERAN TC1 [7]). Finally, in section IV we offer
some concluding remarks.
II. THEORY
A. ECal Calibration
The calibration of a VNA has a double function: setting
the reference planes of the test ports, and correcting the
systematic measurement errors of the test setup. Typically, the
VNA will be calibrated using either a traditional mechanical
calibration kit (Calkit) or an ECal module. When using a
Calkit, different connections to the test ports need to be made
for a single calibration, while a full calibration can be done
with a single connection to the ECal module. Therefore, an
ECal calibration is not only faster but also less susceptible to
operator errors [8].
The ECal operation is based on a SOLT calibration, which
can determine the error coefficients of a full 12 term error
model [9]. This device sets the reference planes at the end of
the test cables where they connect to the test board on which
the DUT is mounted. Hence the effect of the coaxial-to-PCB
The research leading to these results has received funding from the
the Marie Sklodowska-Curie grant agreement No 812.790 (MSCA-ETN
PETER).
European Union from any liability. Project website: http://etn-peter.eu/.
978-1-6654-2380-9/22/$31.00 © 2022 IEEE EMC Compo 2021112
2021 13th International Workshop on the Electromagnetic Compatibility of Integrated Circuits (EMC Compo) | 978-1-6654-2380-9/22/$31.00 ©2022 IEEE | DOI: 10.1109/EMCCOMPO52133.2022.9758617
Authorized licensed use limited to: University of York. Downloaded on August 20,2022 at 16:17:28 UTC from IEEE Xplore. Restrictions apply.
transitions and the PCB itself are not taken into account and
will affect the measurements. Hence, we only correct for
transmission loss and phase errors of the cables but not for
PCB traces and mismatch at the SMA connectors. Moreover,
as the frequency increases, this error increases too. Therefore,
there is a need for de-embedment of the PCB effects but no
simple and good solution is available [10]. Using the port
extension feature of the VNA can correct for the electrical
length between the SMA connectors and the pins of the DUT
when we use ECal for calibration, but there is still inaccuracy
in measurement results due to mismatch errors of the SMA
connectors and miscrostrip lines.
B. Proposed Calibration Method
The proposed method is a combination of the TRM and
TRL calibration techniques and is based on True, Reflect,
Match and Line calibration standards. It fully corrects 10 error
coefficients of 12-term error model [9], two other error terms
are due to leakage that we did not consider in this study. At
higher frequencies, TRL calibration is used which is the more
accurate approach since it does not need lumped models and
precise calibration standards. But at lower frequencies, using
TRL is impractical because of the long length of lines [11,12].
The alternative approach at lower frequencies is TRM, which
applies a match standard instead of the line standard used in
TRL. This perfect match can be considered as an infinitely
long line, which is also the limitation of the TRM method as
determining accurately the load impedance is very difficult at
higher frequencies [13,14]. In this study, we use the
advantages of both methods where we use the TRM method
for lower frequencies up to 500 MHz, and the TRL calibration
at higher frequencies. This calibration requires a specially
made de-embedding board called calibration board which will
be explained in detail in the next section.
This new calibration method is a manual technique and
takes more time than using an ECal. The advantage is that it
is more accurate as it sets the reference planes at the
component terminals and additional de-embedment is not
required. This technique is also convenient to accurately
measure an SMD component or an on-board structure. Then
the reference planes can be located at the on-board terminals
of the component-under-test.
III. METHODOLOGY & MEASUREMENT RESULTS
Measurements were performed using an advanced four-
port Keysight E5080A, with frequency range from 9 kHz to 9
GHz and a set of short precision test cables with guaranteed
phase and amplitude stability. A four-port ECal N4431B,
9kHz to 13.5GHz was also used. On the calibration board, 8
SMA through-hole PCB-mounting connectors were used and
FR4 with the dielectric constant of 4.3 was the material of the
PCB. The calibration quality is validated for both TRL/TRM
method and ECal. This evaluation of the two methods was
performed before the TOLERAN TC1 measurements.
For TRL/TRM calibration board, the primary length
which is the length of traces from the SMA connector to the
reference plane was set to 37mm, the trace impedance was
tracks
on the surface layer are microstrip lines. To cover the
frequency up to 6 GHz, on the calibration board one thru, one
open (Reflect), 2 lines (called 1 and 2) and one load were
considered. See Fig.1.
Thru length is twice the primary length and it has connect-
Fig. 1. TRL/TRM calibration board
ors at both ends. For the Reflect standard, an open is used.
Each trace of line1 and 2 has a specific offset length compared
to thru. The offset length of two lines is controlled at least 20
-30 degree phase margin at the determined boundary
frequencies. Thru and open can cover the whole frequency
range 0-6 GHz, line 1 with an offset length of 20mm
referenced to thru is used for frequency range of 0.5 GHz to 3
GHz and line 2 with an offset length of 7 mm compared to
thru can cover frequencies between 3 to 6 GHz. The Load
covers the low frequency region below 500 MHz which
behaves as an infinitely long Line. For the Load, one precision
towards the end of the
primary length for termination.
A. Evaluation of TRL/TRM Method
For comparison purposes, we performed both ECal and
TRL/TRM calibrations and measured S-parameters for a test
board similar to the calibration one which had two extra traces,
one short and one line with 20 mm shorter length compared to
thru called line 3.
Thru measurement data which were acquired after
TRL/TRM method and ECal calibrations are shown in Fig.
2.a. For S12 magnitude, deviation from 0dB for the
calibration board is -0.0105dB up to 6GHz and phase
deviation from 0 degree is within 0.3 degree up to 4.5GHz, at
the worst case it reaches to 5.1618 degree for frequencies near
6GHz. But for ECal S12 magnitude deviation is -11.6733dB
and phase diagram shows different oscillations between -180
and +180 degrees. Also for the calibration board figures show
nearly perfect phase linearity for frequencies up to 6GHz
while for ECal some nonlinearities can be seen at different
frequency sections.
Fig.2. (b, c) show measurement results for line 1 and 3. As
the length offset for two lines referenced to thru is the same,
we expect to have symmetrical phase and magnitude for S12.
Plots prove perfect symmetry in phases and magnitude for 2
lines when we used calibration board. In contrast, there is no
symmetry in phase or magnitude when calibration was done
by ECal.
In Fig. 2.d, the measurement result for short is shown.
Here also the difference between 2 methods are clear. While
the deviation from 0 dB is -0.1919dB by calibration board it
is -10.5801 dB for ECal one. Also in phase, the nearly perfect
short circuit can be seen up to 4GHz, but for ECal an obvious
113
Authorized licensed use limited to: University of York. Downloaded on August 20,2022 at 16:17:28 UTC from IEEE Xplore. Restrictions apply.
(a)
(c)
(b)
(d)
Fig. 2. S-parameter measurement results versus frequency for the test board calibrated by TRL/TRM calibration method and ECal. S12 magnitude calibrated by
TRL/TRM method , S12 magnitude calibrated by ECal , S12 phase calibrated by TRL/TRM method , S12 phase calibrated by ECal (a) Thru (b)
line 1 (c) line 3 d) short
nonlinearity can be seen at different frequency sections, even
though we can see short circuit behavior in the phase plot.
B. Measurement Comparison
TOLERAN TC1 which is a sensor interface IC used in the
TOLERAN project [7] was measured using the two different
calibration methods. In this case, there was a need for
designing an additional board to measure S-parameters. In
fabrication, we used the same PCB purchase order as
calibration board and the same 4-layer SMA connectors. No
on-board supply and monitoring networks were used, supply
and monitoring were done via the integrated bias tees of the
VNA.
On S-parameter test board, all injection lines have the
same length as the primary length of the calibration board.
Here it is 37mm. This length is measured between the center
of each SMA connector and the corresponding center of the
solder pad of the IC pin-under test. Fig. 3 shows TOLERAN
TC1 schematic and external components for applying this IC
in S-parameter measurement. This device is a mixed signal
sensor interface IC which converts small changes in resistive
Wheatstone bridge to large output voltage variations.
According to application information of this IC, four
3.9kresistors were used as bridge resistors which provided
a basic protection but no decoupling capacitors were used in
this measurement. Nominal supply voltage and current were
5V and 9.1mA during this test. Also, nominal voltage of Vbrg
and output pins were 3.083v and 2.434v respectively, as well.
The injected power was set to 10dBm at the VNA to have
linear S-parameters. It was also possible to increase the power
to 15dBm where the S-parameter results are still in linear
region and did not change during the measurement.
As it is shown in the picture, port 1, 2, 3 referred to supply,
output and bridge voltage called Vbrg, respectively. Fig. 4 (a,
b, c, d) clearly proves using different calibration methods
affects the result of S-parameter measurements. Comparison
between S11, S22 and S33 measurement calibrated by
TRL/TRM method shows that most of the incident RF energy
is reflected from the output pin. However, the Vbrg pin is the
most susceptible one and chip can be disturbed by it very
much. This means that more protection for this pin may be
needed when an application PCB is designed.
On the other hand, by applying ECal calibration the
magnitudes and phases of the reflection parameters, S11, S22
and S33 are different from the former ones. While it was clear
Pin3 is the most susceptible one with the TRL/TRM
calibration method, it is not obvious which pin has the worst
case by ECal calibration. In this case measurement results
show the weakest reflection parameter is S22 with the
magnitude equal to -26.4284dB at 4.136GHz while the
weakest one by the TRL/TRM calibration method is S33 at
3.322GHZ with -18.07 dB magnitude. Fig. 4.c also shows a
big difference in phase of S33 between two methods and
specially at higher frequencies.
Looking at the Fig. 4.d again confirms the importance of
using this method in our calibrations which shows the results
for the transmission parameter between bridge pin and supply
one.
Fig. 3. Application schematic for TOLERAN TC1 in S-parameter
measurements
114
Authorized licensed use limited to: University of York. Downloaded on August 20,2022 at 16:17:28 UTC from IEEE Xplore. Restrictions apply.
(a)
(c)
(b)
(d)
Fig. 4. S-parameter measurement results versus frequency for TOLERAN_TC calibrated by TRL/TRM calibration method and ECal. S-parameter magnitude
calibrated by TRL/TRM method , S-parameter magnitude calibrated by ECal , S-parameter phase calibrated by TRL/TRM method , S-parameter
phase calibrated by ECal , (a) S11 (b) S22 (c) S33 d) S13
Here more important than magnitude is the difference
between phase measurement result. While by ECal we can see
different frequency sections resonance between 180 degree
and -180 degree, by the calibration board S13 phase is
between 89 degree and -153 degree. As the transmission
parameter can show the unintentional parasitic effects, the
importance of this inaccuracy is obvious although part of the
inaccuracy may be due to the difference in electrical length as
no port extension was applied to the reference planes set by
the ECal.
We also repeated the calibrations and measurements after
one month and the obtained results perfectly matched with the
first ones, which shows the repeatability of TRL/TRM
calibration method.
IV. CONCLUSION
Even though the ECal method is much more convenient
than the TRL/TRM calibration method, the accuracy should
not be sacrificed by the simplicity. In this paper, we first
evaluate this method by comparing the S-parameter results
calibrated using this method and ECal which clearly shows
the accuracy of TRL/TRM calibration approach. In another
experiment, the S-parameters of an automotive sensor
interface called TOLERAN TC1 measurement reveals the
importance of using an accurate calibration method. This
method is also cost effective compared to ECal which is an
expensive device.
The accuracy of the proposed method was nearly perfect
for frequencies up to 4GHz while ECal was only accurate up
to 500MHz. To design this calibration kit for frequencies
above 4GHz we need to take some other approaches which is
not in the scope of this paper.
Looking at Fig. 4. a, b we observed that the impedance of
the supply and output pins become inductive at higher
frequencies which can be due to package effects. Also Fig. 4.
C shows the bridge pin is no longer high impedance at higher
frequencies. These results show the need for behavioral EMC
models of ICs to accurately model the scattering parameters
of a device.
Therefore, because the value of the calibration board has
been demonstrated, the authors will apply this methodology
to behavioral IC models in further work.
This calibration approach should also be interesting for the
validation of EMC simulations, the characterization of the
EMC behavior of SMD passive components, the RF
characterization of other passive PCB structures, and the
development of new technologies which need to be
characterized.
REFERENCES
[1] An Analytical
Approach for the Estimation of the Far-Field Reduction Obtained by
Electromag. Compat., vol. 63, no. 5, pp. 1384-1394, 2021.
[2]
vol. 21, no. 5, pp. 6366, May 1978.
[3] -Reflect-line: An improved
technique for calibrating the dual six-port automatic network
987993, Dec. 1979.
[4]
Magazine, vol. 9, no. 3, pp. 8699, Jun. 2008..
[5] with Characterized
Calibration
5C-090.pdf.
[6] -
https://coppermountaintech.com/wp-content/uploads/2018/05/Design-
and-Fabrication-of-a-TRL-Calibration-Kit.pdf
115
Authorized licensed use limited to: University of York. Downloaded on August 20,2022 at 16:17:28 UTC from IEEE Xplore. Restrictions apply.
[7] TOLERAN _TC1_Datasheet_v1 (test chip produced by Melexis in the
TOLERAN R&D project that was sponsored by VLAIO, see
www.vlaio.be)
[8] ECal comparison with mechanical cal kits
https://edadocs.software.keysight.com/kkbopen/ecal-comparison-
with-mechanical-cal-kits-577602842.html
[9]
MTT22, pp. 457-458, Apr. 1974M .
[10]
a single-ended TRL calibration pattern to de-embed coupled
1722, 2009, pp. 197201..
[11] Orii et al., "On the length of THRU standard for TRL de-embedding
on Si substrate above 110 GHz," 2013 IEEE International Conference
on Microelectronic Test Structures (ICMTS), Osaka, Japan, 2013, pp.
81-86.
[12] D. F. Williams, R. B. Marks, and A. Davidson, "Comparison of on-
wafer calibrations," 38th ARFTG Conf., pp. 68-81, Dec. 1991.
[13] hru-match-reflect: One result of a rigorous
theory for de-
Eur. Microw. Conf., Sep. 1988, pp. 909914.
[14]
procedures for network analyzer self-
Microw. Theory Techn., vol. 39, no. 4, pp. 724731, Apr. 1991.
116
Authorized licensed use limited to: University of York. Downloaded on August 20,2022 at 16:17:28 UTC from IEEE Xplore. Restrictions apply.
