Non-Linear Diode Rectifier Analysis

for Multi-Tone Wireless Power Harvesting

Ana López-Yela Zoya Popović
Alberto López-Yela Department of Electrical, Computer, and Energy Engineering
Daniel Segovia-Vargas University of Colorado
Department of Signal Theory and Communications Boulder, Colorado, USA
University Carlos III of Madrid [email protected]
Leganés, Spain
[email protected], [email protected],
[email protected]

Abstract—This paper presents an analysis of the non-linear

performance of a zero-bias Schottky diode under multi-tone exci-
tation for wireless power harvesting applications. At low incident
power levels, it has been shown that multi-tone inputs increase
RF-DC conversion efficiency. Here we extend the theoretical
analysis to include a more practical diode model which includes
series resistance essential for determining efficiency. We show
that the more complex theory approximates the real diode IV
curve more accurately at low input power levels than previous
models which neglect the series resistance.
Index Terms—Non-linear model, zero-bias Schottky diode,
Taylor expansion, multi-tone signals, energy harvesting.

I. I NTRODUCTION
In RF energy harvesting, the available incident power
densities are low (less than µW/cm2 [1]) and the power is Fig. 1. RF wireless power harvesting scenario with multiple input signals
radiated from generally unknown sources which can cover with unequal frequency spacing and of unequal powers.
a broad bandwidth and can be arbitrarily modulated. The
relative position between power transmitters and receivers can
A. Background
vary over time requiring non-directional antennas and power
reception circuits that can maintain efficiency over variable Diode performance in the forward mode is governed by the
power levels. DC IV relationship
In a low-power time-varying environment, the received V
( nVj )
power by a rectenna is possibly not even sufficient to always I = IS (e T − 1) (1)
turn on a rectifying device such as a Schottky diode. This
where IS is the saturation current, Vj is the built-in junction
limits the usefulness of RF energy harvesting. Early work [2]
voltage, n the ideality factor and VT = kT /e the thermal
has shown experimental results on a wideband rectenna array
voltage. As seen in the simple diode equivalent circuit, Fig. 2,
with two arbitrarily spaced signals of varying power level and
if V is the voltage drop at the diode terminals, V 6= Vj , but
frequency spacing in the 1-8 GHz range. With 10,000 different
rather
power-frequency combinations for the two tones, it was shown
that in all cases the rectified power increases relative to the
Vj = V − IRs (2)
sum of the same two individual signals. This is due to the
nonlinear diode characteristics. Two operating regions can be distinguished as a function
Subsequent work with high peak-to-average power ratio of the voltage level. Since the motivation of this work is
modulated signals [3], chaotic signals [4] and multi-tone energy harvesting, only the low voltage quadratic region is
signals [5], [6] have shown similar effects with limited the- of interest. In the quadratic region, the IV curve is dom-
oretical treatment. The work presented in this paper focuses inated by the exponential term, while in the linear region,
on a more comprehensive analysis of the rectifying diode IV the series resistance dominates at larger values of V . In [5]
nonlinear characteristic, leading to improved understanding of and [6] a Taylor expansion is used to approximate the IV
the rectification process for multi-tone signals present in a curve. The resistance Rs is neglected in order to obtain a
wireless powering scenario shown in Fig.1. simple theoretical relation for the DC term. Two cases were
 input RF voltage, the diode switches between ON and OFF
states, and for an accurate mode RS needs to be taken into
account since it affects the slope of the IV curve. Since
nonlinear effects, such as rectification, are due to the quadratic
region, the slope of the curve will critically affect the RF-DC
conversion efficiency for multi-tone signals.
To include the series resistance, Eqs. (1) and (2) are com-
bined to give:
( V −IRs )
I = Is (e nVT − 1) (3)
Fig. 2. Simple equivalent circuit of a diode with a voltage-variable capacitor, As already explained, previous studies approximated this
Cj (Vj ), and series resistance RS . I(Vj ) is a dependent current source given curve through a Taylor expansion centered at the origin,
by Eq.(1).
neglecting the Rs parameter. The DC term was obtained after
filtering out the RF and baseband components [5] where 4
considered: (1) the input x(t) is a signal with ∆ω << ω0 , tones were taken into account, repeated here for completeness:
where ∆ω is the spacing between tones and (2) harmonically- 4
X
spaced multi-sine signals. In both cases, the DC term is shown x(t) = A cos(ωn t + φn ) (4)
to be higher for the multi-tone signal as compared to individual n=1
CW signals with the same average power. For case (2), due
The value of the DC component is obtained as:
to the harmonic relationship, some terms will further increase
the DC component, and reduce the ripple across the DC load, 4A2 k2 21A4 k4 3A4 k4
IDC = + + cos(2φ3 − φ2 − φ4 )
thus increasing efficiency. 2 2 2
In [3] a rectifier scheme is proposed to mitigate the ripple 3A4 k4 (5)
+ cos(−2φ2 + φ1 + φ3 )
when ∆ω is small, but no theoretical approach is presented. In 2
[7], multi-sine signals are used to extend the reading range of + 3A4 k4 cos(φ1 − φ2 − φ3 + φ4 ).
an RFID tag by increasing rectification efficiency. In [8], the
where kn are the Taylor series expansion coefficients. In the
effect of the bandwidth and number of tones is investigated
case when the frequencies are not harmonically spaced, this
but the analytical approximations are not stated. A CW signal
value reduces to:
in [9] and [4] is compared in terms of conversion efficiency
to three signals with different PAPRs: a chaotic signal, white 4A2 k2 21A4 k4
IDC = + . (6)
noise and an OFDM LTE FDD signal, however no theoretical 2 2
analysis is presented. When the series resistance is included, the Taylor expansion
In this paper, a mathematical expression for the DC IV cannot be computed as above since the current is in the expo-
curve that includes a series resistance is developed. The model nent. Therefore, we invert the current-voltage dependence:
is applied to a Skyworks SMS7630 Schottky diode. It is shown
that the model has a small relative error compared to the 1 I
V = Rs I + log(1 + ), (7)
nonlinear diode model provided by the manufacturer at low δ Is
input voltages. For simultaneous low input multi-tone signals, where δ = nV1T . Then, we apply the Wright Omega function
the model can give an accurate prediction of the expected defined as a function of a variable x as follows [10]:
rectified DC output current.
ω(x) = {x ∈ R | x = ω(x) + log ω(x)}, (8)
II. A NALYTICAL A PPROXIMATION OF D IODE IV C URVE
where the nth derivative of this function can be expressed as:
WITH S ERIES R ESISTANCE
an,n−1 · ω n−1 (x) + ... + an,2 · ω 2 (x) + ω(x)
Different input tones result in different DC output across a ω (n) (x) = ,
(1 + ω(x))2n−1
nonlinear diode, depending on the input amplitudes, frequency (9)
spacings and relative phases. The diode non-linearity was and where the coefficients satisfy the recursive relation:
studied in [5] using Eq.(1) with Vj = V to see how an N -tone
low input voltage can produce a higher DC output than a single an,n−k = (n − k)an−1,n−k − (n + k − 2)an−1,n−1−k (10)
tone with the same total power. This equation describes an
for k = 1, . . . , n − 2, being an,1 = 1, an,n = 0 for n ≥ 2
ideal diode model with no parasitics and no series resistance.
and a1,1 = 1. Now the diode current can be written in terms
When a more realistic intrinsic diode model is considered,
of the Wright Omega function and its derivatives used for the
it is necessary to take into account the voltage drop across
Taylor expansion:
the series resistance in Fig.2. The series resistance can be
neglected only for very low input voltages below threshold,
   
V
when the diode is off. In the quadratic region, with low I = Is α · ω −γ −1 , (11)
β
where α, β and γ are given by: Next we use the multinomial theorem to expand the sum of
exponentials, which results in:
1 1
α= , β= , and γ = log(α) − δRs Is . N
δRs Is δ X
k
ejωr t + e−jωr t

Substituting the Taylor series of Wright Omega function in Eq. r=1
(10) yields: X  k
 PN (19)
= ejωt m=1 (pm −p−m )nm
" ∞   k # p−N , . . . , pN
X 1
(k) V0 V − V0 p−N +···
I = IS α ω −γ −1 (12) +pN =k
k! β β
k=0 where p−N + ... + pN = k means the partitions in which k
If we consider V0 = 0 for the unbiased case used in wireless can be divided into 2N non-negative integers and ω = 2πT .
power harvesting, we get: We will consider the general case for wireless power
" ∞ # harvesting, where the incident signal frequencies ωr (e.g.
X ω (k) (−γ) signals f1 , f2 and f3 in Fig.1) are non-harmonically spaced
k
I = IS α V −1 . (13) and satisfy:
k!β k
k=0
N
X
For simplicity, we write this formula in the form of a power (pm − p−m )nm = 0, p−m = pm , ∀m (20)
series: m=1
X∞
I= ck V k , (14) Notice that this condition
R T onlyk holds for a certain subset of
k=0 even partitions, hence 0 V (t) dt = 0 if k is odd.
where the coefficients ck are:
IV. A NALYSIS E XAMPLE
c0 =Is [αω(−γ) − 1] As an example of the derived analytical method, we choose
ω (k) (−γ) (15) a zero-bias Schottky diode manufactured by Skyworks (SMS
ck =Is α , k = 1, 2, . . . 7630) for which a nonlinear Spice model is provided. The
k!β k
saturation current Is = 5 µA, n = 1.05 and resistance
With the above formulation, multi-tone input signals can be Rs = 20 Ω from the Spice model are then used to perform
applied and the output current analytically calculated. the presented analysis, and the results are plotted in Fig.3.

III. M ULTI -T ONE E XCITATION A NALYSIS

10-4 IV curve SMS7630
10
Now that the analytical expression for the nonlinear IV Rs=0
characteristic is developed, we apply multiple RF tones. We 9 Real
are interested in finding the DC component, which for a Approx
8
periodic signal is computed as the first coefficient of its Fourier
series: 7

1 T
Z
IDC = I(V (t))dt, (16) 6
Current(A)

T 0
5
where T is the period of the signal. Let us consider the
4
following multi-tone signal:
3
N N
X A X jωr t
+ e−jωr t ,


V (t) = A cos(ωr t) = e (17) 2
r=1
2 r=1
1
2π
with ωr = Tr for r = 1, 2, . . . , N . The period of this signal 0
will be the least common multiplier of the periods of every 0 0.05 0.1 0.15
Voltage (V)
tone: T = LCM(T1 , T2 , . . . , TN ), this is, T = n1 T1 =
n2 T2 = · · · = nN TN with nr ∈ N. By substituting Eq. (14)
Fig. 3. Comparison of the approximated analytical IV curve with the one
in (16), we have: obtained from the manufacturer and for a model with no series resistance.
∞ Z T
1 X A comparison is shown between the real IV curve (red solid
IDC = ck V (t)k dt
T 0 line), the proposed analytical model (yellow dashed line) and
k=0
#k (18) the model when no series resistance is included (blue dotted
∞ Z "N
1 X Ak T X jωr t line). Note that the radius of convergence of the Taylor series
+ e−jωr t


= ck k e dt.
T 2 0 r=1 of the Wright Omega function is small and therefore holds for
k=0
 Maximum
V close to 0. If we consider the expansion to order 8 and V Case Amplitude (Vin )
Predicted Simulated
Relative
in the interval (0, 0.15), the maximum relative error will be IDC (µA) IDC (µA)
Error (%)
lower than 6%. (a) 4-tone 0.01 0.71034 0.71963 1.29
(a) 1-tone 0.02 0.693 0.702 1.28
We next apply 4 tones using an 8-th order Taylor expansion. (b) 4-tone 0.025 5.9567 6.046 1.477
The tones have equal amplitudes and are randomly spaced in (b) 1-tone 0.05 5.1123 5.187 1.44
the frequency domain corresponding to some standard wireless (c) 4-tone 0.03 9.9547 10.06 1.046
bands that would be appropriate for energy harvesting: 0.85, (c) 1-tone 0.06 8.0054 8.1286 1.51
1.85, 2.1 and 2.45 GHz. The predicted DC value of the output (d) 4-tone 0.04 25.30 24.82 1.93
(d) 1-tone 0.08 17.45 17.72 1.52
current obtained by the approximate theory is then compared
with AWR Microwave Office nonlinear simulations with the TABLE I
Spice diode model, Fig.4. R ELATIVE ERROR VS INPUT VOLTAGE FOR A 4 TONE EXCITATION

Predicted DC output current Vs Input Power

30
4-tone excitation IDC
7
1-tone excitation IDC
25

Predicted output DC Current ( A)

IDC

DC current improvement ( A)
6

20
5

15 4

3
10

5
1

Fig. 4. Diode spice model in AWR Microwave Office excited with a four-tone 0
signal with equal amplitudes at 0.85, 1.85, 2.1 and 2.45 GHz. 0 10 20 30 40 50 60 70
Input Power 10-4 W/

The results for several relevant scenarios are next presented. Fig. 5. Predicted DC output current versus normalized input power for single
In Table I, equal-amplitude tones and a single tone are applied tone and four tone excitation. DC current improvement for each case is also
plotted.
at four different amplitude cases, to compare the 4-tone case
to that of a single tone with the same total power. Let A be
the amplitude of the one-tone signal and B the amplitude for mance of the diode using the nonlinear mode is obtained as:
each of N tones so that B = √AN . In this specific example,
c2 c4
as N = 4, A = 2B. IDC = c0 + A2 2
· 8 + A4 4 · 168
2 2
In this table, the first column represents four different c6 c8
amplitude cases, (a) to (d). The second column shows the + A6 6 · 5120 + A8 8 · 190120. (21)
2 2
tone amplitude and the third one the predicted output DC where the ck coefficients are calculated using (15). A small
current value. The fourth column shows the DC term obtained relative error compared to harmonic balance nonlinear simula-
with the manufacturer-provided nonlinear model and harmonic tions is obtained with this approximation, which gives a more
balance simulations in Microwave office. Finally, the last intuitive and very fast estimate of multi-tone behavior. Thus,
column shows the small relative error between the two cases. we conclude that this expression is valid for approximating
The predicted value for the DC current is higher when using diode nonlinear performance under multi-tone signal excitation
a multi-tone signal when compared to a single CW signal with low input power levels, such as the ones present in energy
with the same total power. Again, the relative error between harvesting scenarios.
the approximate theory and harmonic balance simulations is
small. Fig.5. shows the predicted DC output current for a V. C ONCLUSION
single tone and a four-tone excitation at different input power This paper develops an analytical model for rectifier diode
levels (normalized to 1 Ω). The DC current improvement is performance under multi-tone low-power input signals. We
represented on the right axis, and as expected in all cases the include the diode series resistance and obtain a better approxi-
4-tone output is higher for the same total input power. mation of the diode IV curve. This new type of approximation
To summarize, the analytical formula that fits the perfor- is valid for low input voltage levels, due to the fact that the
radius of convergence of the Taylor series of Wright Omega
function is small. The tool can be useful for energy harvesting
simulations, where different randomly-spaced and dynamic
signals are incident on a rectenna element. Ongoing work
addresses experimental multi-tone validation, as well as a
theoretical approach to extending the model for larger voltages
in order to fully characterize diode performance.
It is worth mentioning that the case considered here pertains
to a resistive rectifier, i.e. the nonlinearity of the junction
capacitance is not used for rectification and the model follows
the Shockley diode equation (1). The nonlinear capacitance
can be included in the future in analogy to the classical
analysis of reactive mixers, see e.g. [11].

ACKNOWLEDGMENTS
Ana Loṕez-Yela’s work is supported by the Spanish Min-
istry of Education, Culture and Sport under a predoctoral con-
tract for University Teacher Training (FPU). Alberto Loṕez-
Yela acknowledges CAIMAN project (TEC2017-86921-C2-2-
R). Z. Popović acknowledges the Santander Chair of Excel-
lence at Carlos III University in Madrid.
R EFERENCES
[1] M. Piuela, P. D. Mitcheson and S. Lucyszyn, ”Ambient RF Energy Har-
vesting in Urban and Semi-Urban Environments,” in IEEE Transactions
on Microwave Theory and Techniques, vol. 61, no. 7, pp. 2715-2726,
July 2013.
[2] J. A. Hagerty, F. B. Helmbrecht, W. H. McCalpin, R. Zane, Z. Popovic,
“Recycling ambient microwave energy with broad-band rectenna arrays,”
in IEEE Transactions on Microwave Theory and Techniques, vol. 52,
no. 3, pp. 1014-1024, March 2004.
[3] Z. Liu, Z. Zhong and Y. Guo, ”Intermodulation harvesting rectifier
design for high efficiency multi-sine wireless power transfer,” 2016 IEEE
MTT-S International Microwave Symposium (IMS), San Francisco, CA,
2016, pp. 1-3.
[4] A. Collado and A. Georgiadis, ”Improving wireless power transmission
efficiency using chaotic waveforms,” 2012 IEEE/MTT-S International
Microwave Symposium Digest, Montreal, QC, 2012, pp. 1-3.
[5] A. S. Boaventura and N. B. Carvalho, ”Maximizing DC power in
energy harvesting circuits using multisine excitation,” 2011 IEEE MTT-
S International Microwave Symposium, Baltimore, MD, 2011, pp. 1-4.
[6] D. Belo and N. B. Carvalho, ”Harmonic spaced multisines for efficient
wireless power transmission,” 2015 IEEE Wireless Power Transfer
Conference (WPTC), Boulder, CO, 2015, pp. 1-4.
[7] A. S. Boaventura and Nuno Borges Carvalho, ”Enhanced front-end to
extend reading range of commercial RFID readers using efficient mul-
tisine signals,” 2012 IEEE/MTT-S International Microwave Symposium
Digest, Montreal, QC, 2012, pp. 1-3.
[8] N. Pan, A. S. Boaventura, M. Rajabi, D. Schreurs, N. B. Carvalho and S.
Pollin, ”Amplitude and frequency analysis of multi-sine wireless power
transfer,” 2015 Integrated Nonlinear Microwave and Millimetre-wave
Circuits Workshop (INMMiC), Taormina, 2015, pp. 1-3.
[9] A. Collado and A. Georgiadis, ”Optimal Waveforms for Efficient Wire-
less Power Transmission,” in IEEE Microwave and Wireless Compo-
nents Letters, vol. 24, no. 5, pp. 354-356, May 2014.
[10] Corless, R. M. and D. J. Jeffrey. The Wright omega Function. Artifi-
cial Intelligence, Automated Reasoning, and Symbolic Computation (J.
Calmet, B. Benhamou, O. Caprotti, L. Henocque, and V. Sorge, eds.).
Berlin: Springer-Verlag, 2002, pp. 76-89.
[11] D. N. Held and A. R. Kerr, ”Conversion Loss and Noise of Microwave
and Millimeterwave Mixers: Part 1 - Theory,” in IEEE Transactions on
Microwave Theory and Techniques, vol. 26, no. 2, pp. 49-55, Feb. 1978.