Diode Model Generation for Simulation of Harmonic Distortion

Jennifer Schütt Jens Werner Ayk Hilbrink

Nexperia Germany GmbH Jade University Nexperia Germany GmbH
of Applied Sciences
Stresemannallee 101 Friedrich-Paffrath-Str. 101 Stresemannallee 101
22529 Hamburg 26389 Wilhelmshaven 22529 Hamburg
Germany Germany Germany
jennifer.schuett@nexperia.com jens.werner@Jade-HS.de ayk.hilbrink@nexperia.com

KEYWORDS past had only limited attention on the frequency

SPICE equivalent device model, electrostatic dis-
charge, ESD protection, harmonic distortion
ABSTRACT
This paper presents the stepwise SPICE model
generation for Transient Voltage Suppression
(TVS) diodes allowing to represent the 2nd and 3rd
harmonic distortion behavior in circuit based
simulations. The model is based on semiconductor
physics and generated using measurement data of
the device. The generation of this model is based on
the I/V characteristics, the behavior of the
capacitance versus diode voltage variations, and
tuning at one single input power level at the
fundamental frequency as well as for the 2nd and 3rd
harmonics. The simulation tool used for the device
modelling is Keysight Avanced Design System
(ADS).
INTRODUCTION
Harmonic Distortion (HD) is a well-known
phenomenon in radio frequency (RF) applications
like amplifiers and receivers. It is part of standard
device characterization to look at the frequency
spectrum at the device output. Trends in the market
for TVS diodes show rising requests for diodes with
optimized HD behavior. Frequencies of interest are
mainly found in wireless applications like GSM
(900 MHz, 1800 MHz) and WiFi (2.4 GHz,
5 GHz), UMTS (3G) or LTE (4G).
In general, TVS Diodes are made for overshoot
protection of supply lines as well as for data lines of
interfaces, e.g. like USB, DVI or HDMI. The
development of TVS diodes for supply lines in the
behavior. To predict the result of design
improvements of an existing product, the easiest
way is to model the device first and simulate
potential improvements. In that manner, a new field
of modelling emerged. This paper presents the first
steps of this modelling and simulation process.
HARMONIC DISTORTION
According to (Pozar 2011) and (Maas 1988) the HD
phenomena can be described by an input voltage vi
and an output voltage vo which results according to
the Taylor series in equation (1):
𝑣𝑣0 = 𝑎𝑎0 + 𝑎𝑎1 𝑣𝑣𝑖𝑖 + 𝑎𝑎2 𝑣𝑣𝑖𝑖2 + 𝑎𝑎2 𝑣𝑣𝑖𝑖3 + ⋯ (1)
Considering vi being a single frequency sinusoid:
𝑣𝑣𝑖𝑖 = 𝑉𝑉0 cos(𝜔𝜔0 𝑡𝑡) (2)
The equations (1) and (2) lead after some trans-
formation to:
1
𝑣𝑣0 = �𝑎𝑎0 + 𝑎𝑎2 𝑉𝑉02 �
2
3
+ �𝑎𝑎1 𝑉𝑉0 + 𝑎𝑎3 𝑉𝑉03 � cos 𝜔𝜔0 𝑡𝑡
4 (3)
1 2
+ 𝑎𝑎2 𝑉𝑉0 cos 2 𝜔𝜔0 𝑡𝑡
2
1
+ 𝑎𝑎3 𝑉𝑉03 cos 3 𝜔𝜔0 𝑡𝑡 + ⋯
4
The harmonics in general are related to non-linear
dependencies, e.g. the quadratic, cubic and higher
terms in equation (1). From equation (3) the
generation of second harmonic waves caused by a
non-linear device can be identified as:

Proceedings 32nd European Conference on Modelling and

Simulation ©ECMS Lars Nolle, Alexandra Burger,
Christoph Tholen, Jens Werner, Jens Wellhausen (Editors)
ISBN: 978-0-9932440-6-3/ ISBN: 978-0-9932440-7-0 (CD)

Port 1
Diode1
Diode2
Figure 1: Configuration of a Bi-directional TVS
Protection Device.
1
𝑎𝑎 𝑉𝑉 2 cos 2 𝜔𝜔0 𝑡𝑡 (4)
2 2 0
And the third harmonic wave is
1 𝐶𝐶𝐽𝐽 =
𝑎𝑎 𝑉𝑉 3 cos 3 𝜔𝜔0 𝑡𝑡 (5)
4 3 0
In this paper harmonic distortion of higher orders
than three will be neglected. This is reasonable since
for higher orders of HD components the magnitude
of those components is usually decreasing for
increasing order. Nevertheless, an extension of the
modelling process to consider higher harmonic
orders is straight-forward. In wireless applications
the second harmonic and the third harmonic waves
are of rather high importance: E.g. for any GSM-
850 mobile phone the specified uplink spectrum is
824 – 849 MHz (3GPP, 2005). Therefore any
unwanted harmonic signal would appear in the
spectrum at 1648 – 1698 MHz for 2 ∙ 𝜔𝜔0 and at
2472 – 2547 MHz for 3 ∙ 𝜔𝜔0 . The latter would
partly fall in the 2.4 GHz WiFi spectrum and might
cause a degradation. In the same way the 2nd
harmonic of a GSM-900 system might interfere
with the communication of a DCS-1800 system.
THE PROTECTION DIODE
The purpose of the protection diode (TVS) is to take
over the current in case of electrostatic discharge
(ESD). In the simplest case, one diode with low
ohmic resistance is used. According to the I/V curve
it is conductive above a certain voltage. For a silicon
diode that just means around 0.7 V. Inside an
application, the ESD protecting device is assumed
to be inactive until an ESD event occurs. This
requirement is usually also in line with the need of
low leakage currents. Putting these requirements
together it can result in a configuration with two
diodes connected in series where the anodes are
connected to the same node (Figure 1). In this case
one diode is always in reverse direction and makes
the combination inactive until the breakdown
voltage is reached.
A semiconductor device usually comes along with a
capacitive behavior caused by the junction
capacitance which is in general voltage dependent
(Reisch, 2005). The dependency of a junction
capacitance CJ versus the applied voltage V can be
seen in equation (6) where CJ0 is known as the
junction capacitance at 0 V, VJ is the junction
potential and M is known as grading coefficient.
The grading coefficient reflects the abruptness of
the PN junction.
𝐶𝐶𝐽𝐽0
𝑉𝑉 𝑀𝑀 (6)
�1 − �
𝑉𝑉𝐽𝐽
This dependency can be measured and used for the
model alignment.
THE DIODE MODEL
In general, a SPICE model (Nagel, 1973) of a diode
consists of several parameters related to the physical
behavior of the diode. An exemplary definition of a
diode model can be seen in Figure 2.
Figure 2: Diode model in ADS with a dedicated
set of electrical and physical parameters.
The represented parameters in this model are the
saturation current (Is), the ohmic resistance (Rs), the
linear capacitance (Cd) the zero-bias junction
capacitance (Cjo), the junction potential (Vj), the (Keysight E4980A) which offers precise
reverse breakdown voltage (Bv), the emission measurements of the capacitance value with an
coefficient (N), and the grading coefficient (M). As accuracy of +/- 0.05% and bias level up to 40 V. For
it can be found in (Reisch 2005) the extraction of the diode to be modelled a range from -4 V to 4 V
was used. The test frequency for the capacitance
30,0
20,0 measurement was 1 MHz. The device under test
(DUT) was probed with needles.
10,0
0,0
-10,0
measured power (dBm)

-20,0
-30,0
-40,0
-50,0
-60,0 power @ 1xf0
Harmonic Distortion
-70,0
-80,0
power @ 2xf0
The measurement setup for the harmonic distortion
-90,0
-100,0
power @ 3xf0
itself consists of a signal generator, attenuators and
-110,0 a DC block to protect the input of the spectrum
analyzer, a band pass filter for suppression of
-120,0
-20,0 -15,0 -10,0 -5,0 0,0 5,0 10,0 15,0 20,0 25,0
forward power @ f0 into DUT (dBm)
unwanted signals and a high pass filter to remove
the strong fundamental wave from the harmonic
Figure 3: Measured data of Fundamental Wave measurements in order to avoid saturation of the
and Harmonic Distortion versus Input Power. spectrum analyzer. In particular the 6 dB and 20 dB

Figure 4: Measurement setup for HD measurement of GSM-900 fundamental frequencies (8xx – 9xx MHz).

the parameters Is, N and Rs can be done by using attenuators ensure a proper 50 Ohm termination to
three points of an I/V curve. In reverse direction, the both sides of the DUT position since the band pass
breakdown voltage (Bv) and the related current (Ibv) and high pass filter have a reactive behavior outside
can be fitted along the measured I/V curve. To get their passband frequencies.
the values Cd and Cjo the model needs to be fitted
along an C(V) curve (Reisch 2004, p.75). The forward power applied to the DUT is measured
by a directional coupler and a power meter. The
spectrum analyzer is finally used to measure the
level of the harmonic signals for varying levels of
MEASUREMENTS the fundamental wave. The DUT itself was soldered
Voltage dependency of capacitance on a co-planar wave guide. The complete
The capacitance variation versus the applied diode measurement setup is depicted in Figure 4. The
voltage is measured using a precision LCR Meter analyzed fundamental frequency was 837 MHz,
 4,4E-12 Capacitance over voltage
4,2E-12 The measured C(V) curve was loaded into the
simulation setup. For the first step both diodes were
Csimulated
4,0E-12
Cmeasured modelled with an identical diode model DiodeM1.
Capacitance (F)

3,8E-12
3,6E-12
3,4E-12
-4,0 -3,0 -2,0 -1,0
Voltage (V)
Figure 5: Simulated and Measured Capacitance
versus Applied Diode Voltage at 1 MHz.
resulting in the second harmonic wave with
1.674 GHz and the third harmonic at 2.511 GHz.
The input power was varied from -26 dBm up to
+20 dBm. The measured data is plotted versus the
available forward power of the fundamental wave
(f0) at the DUT (Figure 3).
SIMULATION
The simulation was performed in three steps. First a
draft fitting of the I/V curve was done. That means
the breakdown voltage was adjusted according to
the measurement data. This ensures that no high
To get a good matching, a minimum of three diode
parameters and an additional capacitance
(C_extern) – which models the metal-oxide-silicon
capacitance – is needed. The external capacitance
0,0 1,0 2,0 3,0 4,0 was finally tuned to C_extern=0.8 pF. For the diode
a linear capacitance of Cd=1.85 pF, a zero-bias
junction capacitance of Cjo=5.05 pF and a junction
potential of Vj=1.308 V was found in order to match
the simulation with the measured data. The result
can be seen in Figure 5.
Harmonic Distortion
The complete test bench is shown in Figure 6. For
the particular analysis of the system the same
fundamental frequency of RFfreq= 837 MHz was
chosen as for the measurements. In order to
calibrate the harmonic distortion behavior the ratio
between Cd and C_extern was tuned. As this
changes also the C over V curve, this step needs to
be done iteratively until C(V) matches this specific

RF Output
RF Input R9 Vdut Vout
R12
P_1Tone DC_Block1 DC_Block2
PORT1 -6dB -20dB Term
Num=1 R11 Term1
Z=50 Ohm R10 R8 R13 Num=2
P=dbmtow(RFpower) DUT Z=Zload
Freq=RFfreq

HARMONIC BALANCE DUT

I__21
HarmonicBalance
HB1
Freq[1]=RFfreq
Order[1]=5
SavedEquationName[1]="RFfreq"
SavedEquationName[2]="RFpower"
SavedEquationName[3]="Zload"

Figure 6: Simulation Setup for HD “Measurements”

current flows through the diodes. Than the model measurement data as well as the measurement and
was tuned in order to achieve the C(V) behavior. simulation of the power of the fundamental
The third step was the tuning of the harmonic frequency. Cj itself varies with voltage V (Reisch,
distortion behavior: The model was tuned to match 2005) as a function of the grading coefficient M.
the second and third harmonics for a single level of This parameter was used to tune the 3rd harmonic
the fundamental wave of 20 dBm. distortion in order to match the measurement data.
To tune the 2nd harmonic distortion a minimal
P3
asymmetry of the grading coefficient was useful. Diode_Model
Num=3

Which results in two different diode models. In DIODEM1

Cd=C_di pF
Var
Eqn VAR
VAR4

Figure 8 the “Two-Diode-Model” is shown in Cjo=Cjo pF

Vj=1.308 V DIODE5
C_extern=0.79
C_di=1.85
Model=DIODEM1 Grad=0.5
detail. Finally, a matching for the HD behavior for
M=Grad
Bv=23 V G_diff=2.048e-3
Cjo=5.05
an input power level of +20 dBm was achieved. C6
Figure 7 compares the simulation results with the Diode_Model
DIODEM2 DIODE6
C=C_extern pF

Model=DIODEM2
measurement data for a variation of the forward Cd=C_di pF
Cjo=Cjo pF
Vj=1.308 V
power level at the fundamental frequency. In part a) M=Grad-G_diff P4
Num=4
Bv=23 V
the matching of the fundamental wave is shown. In
Figure 8: Final Diode Model
a) Fundamental Frequency
30
Otuput Power with DUT (dBm)

20
CONCLUSION
10

0 A diode model was generated, that allows to

-10
simulate the harmonic distortion of a device. It can
-20

-30
f_0_meas
be used for optimization of future devices by
f_0_sim
-40 changing the semiconductor processing and layout
structure towards an optimized design as an
-35,0 -25,0 -15,0 -5,0 5,0 15,0 25,0
Input Power at DUT (dBm)

outcome of the simulation. It was also shown that

b) 2nd Harmonic Frequency
even the simplest form of a SPICE based diode
Otuput Power with DUT (dBm)

-70 model covers the non-linear effects that lead to

-95
harmonic distortion. This model can be used to do a
simulation in form of a design of experiments
-120
(DOE). That means by varying all available
H2_meas
-145
H2_sim component parameters the influence of each
-170
-35,0 -25,0 -15,0 -5,0 5,0 15,0 25,0
parameter variation can be evaluated. This allows to
Input Power at DUT (dBm) identify layout and process improvements. For
example the ratio of Cj, C_extern and Cd can be
c) 3rd Harmonic Frequency
-20 optimized. Future investigations will lead to
Otuput Power with DUT (dBm)

-40
-60
modelling of higher harmonics, as well as in more
-80 complex semiconductors. As silicon controlled
-100
rectifiers (SCR) as well as open base transistors are
-120
-140 H3_meas commonly used instead of TVS diodes, it would be
-160 H3_sim of interest to adapt this SPICE model approach for
these devices as well.
-180
-35,0 -25,0 -15,0 -5,0 5,0 15,0 25,0
Input Power at DUT (dBm)

Figure 7: Harmonic Power versus Input Power. REFERENCES

Measurements and Simulation in Comparison: a)
f0: power of fundamental wave; b) H2: power of Keysight. (2018) ADS - Advanced Design System.
2nd harmonic wave; c) H3: power of 3rd harmonic [2018-01-30]. [Online]. Available:
wave http://www.keysight.com/find/eesof-ads
part b) the second and in part c) the third Kories, R., Schmidt-Walter, H. “Taschenbuch der
harmonic wave is shown. For forward power Elektrotechnik“, 5. Korrigierte Auflage, 2003, 220.
level above -5 dBm a very good matching has Pozar, D. M. 2011. “Mircrowave Engineering. 4th”
been achieved. The deviation for HD2 and HD3 Edition. John Wiley & Sons, Inc. 511-51x
visible for level below -5 dBm is caused by the
noise floor of the measurement setup: In fact Maas, Stephen A. “Nonlinear microwave and RF
the low level of the harmonic signals is circuits”, 2nd ed., Artech House microwave library,
masked by the noise of the spectrum analyzer. 1988 (reprinted in 1997)
3GPP (3rd Generation Partnership Project), Ayk Hilbrink was born in Stade,
Technical Specification Group GSM/EDGE, Radio Germany in 1992. He studied
Access Network; Radio transmission and reception, electrical engineering at Jade
3GPP TS 05.05 V8.20.0 (2005-11) University of Applied Sciences,
Wilhelmshaven and received the
L. W. Nagel and D. Pederson, “Spice (simulation bachelor degree in 2016 and the
program with integrated circuit emphasis),” EECS Master of Science in 2018. His
Department, University of California, Berkeley, bachelor thesis “Generierung von Spice-Modellen
Tech. Rep. UCB/ERL M382, Apr 1973. [Online]. in ADS” was written in cooperation with NXP
Available: Semiconductors and covers the field of device
http://www2.eecs.berkeley.edu/Pubs/TechRpts/197 modelling in the RF range. His expertise in RF
3/22871.html applications led him to his Master thesis at Nexperia
about deembedding. Since 2017 Ayk is working for
Reisch, M. „Halbleiterbauelemente“, Springer, Nexperia as characterization engineer with main
2005, 65-87 focus on RF and EMI measurements.
Jennifer Schütt was born in Heide,
Germany in 1981. She received a
degree (Dipl.-Ing.) in electrical
engineering from Technical
University of Hamburg (TUHH)
in 2009. Since then she is with
NXP Semiconductors, Hamburg,
developing EMI-Filter and ESD protection devices.
Since 2017 she started working for Nexperia within
the same working area. Main field of expertise is in
device modelling, EM simulation, device physics
and project management. Currently she is working
on common-mode-filter designs with integrated
ESD protection for ultra-fast differential data lines.
In the field of simulation, especially for ADS related
topics, she organizes regular cross team meetings
for Nexperia engineers located in Hamburg.

Jens Werner was born in Cologne,

Germany in 1969. He received the
Dipl.-Ing. and Dr.-Ing. Degrees in
electrical engineering from the
Technical University of Braun-
schweig, Braunschweig, in 1996
and 2002, respectively. In 1996 he
was working with Aerodata AG as a Flight
Inspection Engineer on calibration of airborne
antennas. From 1996 to 2001, he was a Research
Assistant at the Institute of Electromagnetic
Compatibility, TU Braunschweig. His main
research interests were measurement techniques and
representation of guided and radiated
electromagnetic fields. In 2001 he joined the
Innovation Centre of Philips Semiconductors
Germany GmbH in Hamburg, (since 2006 NXP
Semiconductors). In March 2014, he became a
Professor at Jade University of Applied Sciences,
Wilhelmshaven, Germany. He is responsible for the
RF, Wireless and EMC laboratory.