A Generalized DQ Impedance Model of Three

Phase Diode Rectifier

Qin Lei, Sisheng Liang, Fang. Z. Peng Miaosen Shen, Vladimir Blasko
Electrical and Computer Engineering Department Systems Department
Michigan State University, United Technologies Research Center
East Lansing, MI, USA East Hartford, CT, USA
[email protected] [email protected]

Abstract: Three-phase ac systems can be transferred into dq controllers will be used, line-commutated rectifiers are
axis, where stability analysis becomes much simpler. In favored over PWM rectifiers as front end for these motor
order to apply the stability criterion in AC system, an controllers due to the stringent airborne system
equivalent DQ admittance model for the three-phase diode requirements for reliability and efficiency. In these
rectifier has been developed in this paper, based on the ABC
systems, dynamic interactions between the rectifiers and
impedance model in paper [1]. Compared to the literature
work, this model has two better features: (1) it is valid in a the source can lead to system instability.
wide frequency range, which is from below fundamental The impedance of the line-commutated rectifier or
frequency to high number times of fundamental frequency; PWM rectifier is classified into two categories: one is the
(2) it considers the effect of the ac line impedance and output output dc impedance and one is the input ac impedance.
dc impedance. At different circuit parameters, including the The topologies also have two major kinds, one is PWM
extreme cases, the analytical modeling results and the rectifier and another is diode rectifier. For the input
numerical simulation results present excellent agreement, impedance models of PWM rectifier, the d-q frame
which demonstrates the accuracy of the analytical model. impedance model based on the full order circuit model and
the reduced order circuit model were reported in several
I. INTRODUCTION papers [7-12]. For the stability of diode rectifiers, most
As power electronics technology gets more and more existing works have been focusing on output impedance
involved in the power processing in power system, it modeling of diode rectifiers at the dc terminals [13-15],
introduces the opportunity and requirement of applying but it is not applicable to the complex ac source in practice.
closed-loop control. With this involvement, the stability of The ac impedance calculated in paper [16] only considers
the power system becomes more complicated. While the multiple integer times of fundamental frequency, thus can
larger national power grid can tolerate many of the not be used for stability analysis below the fundamental
behaviors of the power electronics converters like the frequency. The model proposed in paper [17-20]
negative incremental impedance characteristics, as their concentrates on the frequency below fundamental
effect is small compared to the size of the system, many frequency, including both positive-sequence and
interconnected smaller system can not, like the aircraft, negative-sequence ac impedance model in frequency
ships, hybrid-electric vehicles, and small islands with domain. It derived the Fourier transformation for
small power processing plants. The stability of perturbated switching function. However, it is based on the
independent component can not ensure the stability of the assumptions that the input ac line impedance is equal to
system. Therefore, the stability evaluation for the interface zero and the output of diode rectifier is a constant dc
of ac system becomes more and more important now, and current source. These assumptions make the ac input
also is considered as a key criterion in the system design. impedance almost constant at low frequency range, which
The traditional ac stability analysis [1-6] requires the is not true in real case. The conclusion in paper [17] that a
source impedance and the load impedance at the interface reasonable amount of line commutation and source voltage
to employ the Nyquist criterion on the ratio between them. distortion will not cause significant changes in the input
The major difficulty for the impedance model derivation is impedance of diode rectifiers considers the commutation
in the impedance for nonlinear device in ac system, since angle, but not with the real ac line impedance. Also a big
there is no dc operation point. The most common inductance is still adopted for the dc link. In the practical
nonlinear loads in ac power systems are those with a case, the dc side inductance sometimes is much smaller
rectification front end, including both line-commutated than ac side, even equal to zero. So the model of paper [17]
and pulse-width-modulated (PWM) rectifiers. in frequency below fundamental frequency can not be
Line-commutated rectifiers are still widely used today due widely applied. In addition, the model can not attain
to its advantages of low cost, high voltage high power and satisfied accuracy in high frequency range, when the input
high reliability. Examples of such applications include impedance presents a pure inductor characteristic. The
computer power supplies (low cost), high-voltage dc model also loses its effectiveness at frequency that is not
transmission (high voltage and power), and aircraft power integer fraction or integer times of fundamental frequency.
systems (high reliability). Taking one as an example, in In order to solve the aforementioned problems, a
the future more-electric aircraft (MEA) where large motor

978-1-4799-0336-8/13/$31.00 ©2013 IEEE 3340

generalized model for the input ac impedance of three ∞
phase diode rectifier has been derived theoretically in Sa = ∑ ( Asn cos(nθ ) + Bsn sin(nθ ))
paper [21] which has been demonstrated by computer n =1
simulation. The input ac line impedance, output ⎛ 2 3(−1)l sin n(u + φ ) + sin nφ ⎞
⎜ Asn = ⎟
impedance and load impedance in the rectifier are just ⎜ π n ⎟
considered as parameters in the model. The final input ac ⎜ l ⎟
⎜ Bsn = 2 3(−1) cos n(u + φ ) + cos nφ ⎟
impedance is a function of these parameters. At different ⎜ π n ⎟
groups of parameters, including some extreme cases, the ⎜ n = 6l ± 1 (l = 0,1, 2,...n > 0) ⎟
calculated impedance from the model matches with the ⎜ ⎟
⎜ ⎟
simulated impedance from the real circuit very well. This ⎝ ⎠
is an important feature to differentiate this one with the The meaning of each variable is:
model in the literature work [17-20]. The model in u : overlap angle in the diode rectifier commutation
literature work assumes the input of diode rectifier is a φ : firing angle, which is defined as the phase difference
three phase ideal voltage source and the output is an ideal between the starting point of the commutation and the
current source, which leads to a big difference between the actual line voltage zero point; the details of the above two
impedance from the model and the real values. The new variables can be refereed to Fig. 2 in paper [21];
model in paper [21] is validated with high accuracy within Asn1 : the nth(n=6l+1) coefficient of the cosine part of the
a wide frequency band from multiple division of the
switching function;
fundamental frequency to large multiples of fundamental
frequency. A pure time domain mapping method plus the Bsn1 : the nth(n=6l+1) coefficient of the sine part of the
Fourier series analysis are the major focus during the switching function;
modeling. This paper is an extend work to paper [21]. In Asn 2 : the nth(n=6l-1) coefficient of the cosine part of the
order to apply the impedance model in the stability switching function;
analysis, the d-q impedance model is derived in this paper Bsn 2 the nth(n=6l-1) coefficient of the sine part of the
based on the abc impedance model proposed in paper [21]. switching function;
This d-q impedance switching model of the diode rectifier
Z o(ωx ) : the equivalent dc impedance of the ac side
has not been reported in literature work. Compared to
average d-q impedance model, it can describe the impedance: Z o(ω x ) =(2-3u/2π )(R s +jω x X s ) , where u is
impedance characteristics far beyond fundamental
frequency, and it also can monitor the impedance below the overlap angle and Rs and X s is the ac line
fundamental frequency more accurately than the average impedance;
model. It inherits all the benefits of the aforementioned Z (ωx ) : the circuit impedance on dc side, including the
abc impedance model. The circuit diagram that has been
output inductor, capacitor and load:
investigated is shown in Fig.1, and the derived impedance
model equation is shown in eq. (1). Z o(ω x ) =jX l + jX f ( jX d ⋅ Rd ) / ( jX d + Rd )
I d : average dc current of the diode rectifier at which the
small signal analysis is taken place;
E : the per unit amplitude of the ac side source voltage;
va
vb Ad (mω ) : is defined as the Fourier coefficient of the cos
Vdc
part of output dc link voltage; the voltage and the
vc coefficients are defined as:
3 3E ( −1) k cos( m + 1)(u − φ ) + cos( m + 1)φ
Ad (mω ) = {
2π m +1
Fig.1. Diode rectifier with small signal perturbation injected
cos( m − 1)(u − φ ) + cos(m − 1)φ
The derived impedance at ωh is: − }
m −1
1
Z a (ωh )= (1) 3I d Rs ( −1) k sin m(u − φ ) + sin mφ
⎛⎛ 3 ⎞ ⎞ +
⎜⎜ ∞ ( Asn12 + Bsn12 ) ⎟ ⎟ π m
⎜⎜ ∑ Z
4
+ Z
⎟ ⎟
3 3E ( −1) k sin( m + 1)(u − φ ) − sin( m + 1)φ (2)
⎜ ⎜ l =0,n =6l +1 o ( nω −ωh ) ( nω −ωh ) ⎟ ⎟
Bd ( mω ) = {
⎜⎜ ⎟ ⎟ 2π m +1
⎜⎜ 3 2 2 ⎟ ⎟
j* ⎜ ⎜ ∞ ( Asn 2 + B sn 2 ) sin(m − 1)(u − φ ) − sin( m − 1)φ
⎟ ⎟
⎜⎜ + ∑ Z 4
+ Z ⎟ ⎟
−
m −1
}
⎜ ⎝ l =1,n =6l −1 o ( nω + ω h ) ( nω +ω )
h ⎠ ⎟
⎜ ⎟ 3I d Rs (−1) k cos m(u − φ ) − cos mφ
⎜ 1 3 3 1 ∞ ⎛ Ad ( mω ) j 3kπ ⎞ ⎟ −
⎜ d E π 2π E ∑ ⎜ Z
+ I + ⎜ ⋅ e ⎟
⎟⎟ π m
⎝ k =1 ⎝ ( mω ) ⎠⎠ m = 6k (k = 1, 2, 3....)

, which is based on the switching function of diode II. DIODE RECTIFIER DQ ADMITTANCE DERIVATION
rectifier:
A. The method to calculate DQ admittance

3341
 The reason to derive the dq impedance of the diode i (ω ) iq (ωh )
rectifier is to do the stability analysis in dq system. The Ydd (ωh ) = d h |v =0 , Yqd (ωh ) = |v =0
vd (ωh ) q vd (ωh ) q
theorem of Middlebrook’s stability analysis using loop (4)
gain is based on the dc system instead of ac. But the id (ωh ) iq (ωh )
voltage and current in ac system can be transformed to dc Ydq (ωh ) = |v =0 , Yqq (ωh ) = |v =0
vq (ωh ) d vq (ωh ) d
variables at a d-q frame which rotates at the same
frequency as the fundamental frequency. In that d-q frame, Since the impedance model for diode rectifier is already
the operation point at fundamental frequency becomes dc obtained in ABC domain, the method to calculate DQ
operation point. In order to derive the impedance at admittance here is to transfer the voltage perturbation in
different frequencies in d-q frame, similar to the dc system DQ domain to ABC domain, and then transfer the current
impedance derivation method, a series of perturbations at in ABC domain back to DQ domain. Make
different frequencies can be injected to the d-q axis, and Δvd = Eh sin(ωh t ), Δvq = 0 for Ydd and Yqd calculation,
the impedance is equal to the voltage perturbation divided and make Δvd = 0, Δvq = Eh sin(ωh t ) for Yqq and
by the current generated by this perturbation at the
perturbation frequency. The source dc voltage does not Ydq calculation. The detailed block diagram for the
generate any current at frequency other than 0, therefore proposed method is shown in Fig.3.
the current at the perturbation frequency is all generated Δv (ω )
id d h Δva Δia
by the perturbation voltage. Thus the key is to calculate Δvd (ωh )
the current component at the perturbation frequency. Two Δvb Δib
methods can be adopted to realize the goal. One is to Δvq (ωh ) Δvq (ωh )
iq
establish the equivalent circuit model in d-q frame; Δvc Δic
another is to transfer the perturbation into abc frame, get
the abc current and later on transfer this current back into
d-q frame. In order to give the same accurate current if the Ydd (ωh )
Δia
first method is used, the d-q equivalent circuit model Δid Δid (ωh ) Yqd (ωh )
needs to be switching model, not the average model. Δib Ydq (ωh )
Because the average model has two problems: first, the Δiq Δiq (ωh )
Yqq (ωh )
impedance derived from average model can only be Δic
effective under the switching frequency; second, the d-q
Fig.2. DQ admittance model for diode rectifier
axis has coupling with each other which makes the
impedance calculation difficult. For the diode rectifier, the
switching frequency is the fundamental frequency, thus the B. DQ to ABC transformation for perturbation voltage
first condition becomes that the impedance model can only The perturbation in d-q axis is in the form of a
be effective until fundamental frequency. However, the single frequency ac voltage. Its amplitude is far smaller
impedance at frequency higher than fundamental affects than the fundamental voltage amplitude. 2% is used here
the system stability significantly since the phase margin as the ratio between the perturbation voltage amplitude
usually decreases when the frequency increases. Based on and fundamental voltage. A single frequency voltage
the above analysis, the second method is adopted here. injection will becomes multiple frequencies after it is
And it follows three steps: (1) transform the perturbance in transformed into abc frame. The transformation matrix
d-q frame into abc frame using the dq-abc transformation; from DQ to ABC at fundamental frequency ω is defined
(2) derive the certain part of the current in abc frame as M dq _ abc (ω ) . In order to derive the four admittances in
which is generated by the perturbance and also will
generate the component at the perturbance frequency after equation (3), two forms of injections are adopted here,
it is transformed back into d-q frame; (3) transform the which is shown in equation (4).
current into the abc frame, do the fft analysis for the If the d-q perturbation is Δvd = Eh sin(ωh t ), Δvq = 0 ,
current in d and q axis respectively to extract the assuming the fundamental frequency is ω , the
component at perturbation frequency and then calculate corresponding perturbation in a, b and c axis are:
the impedance by voltage and current division. For the ⎡ ⎤
calculation convenience, the impedance can be changed to ⎢ − sin(ω + ωh )t − sin(ω − ωh )t ⎥
⎡ Δva ⎤ ⎢ ⎥
admittance. To be highlighted, the small signal voltage ⎢
Δv
⎥
=
1
E ⎢ − sin((ω + ω )t − 2π ) − sin((ω − ω )t − 2π ) ⎥ (5)
perturbation needs to be applied on d axis and q axis ⎢ b⎥ 6
h⎢ h
3
h
3 ⎥
⎢⎣ Δvc ⎥⎦ ⎢
separately, in which way the four admittance Ydd, Ydq,
⎢ − sin((ω + ωh )t + 2π 2π ⎥⎥
Yqd and Yqq can all be calculated. ) − sin((ω − ωh )t + )
⎣⎢ 3 3 ⎦⎥
The definition of the dq admittance
If the d-q perturbation is Δvd = 0, Δvq = Eh sin(ωh t ) , the
Ydd , Yqd , Ydq , Yqq is:
corresponding perturbation in a, b and c axis are:
⎡id (ωh ) ⎤ ⎡Ydd (ωh ) Yqd (ωh ) ⎤ ⎡vd (ωh ) ⎤
⎢ ⎥=⎢ ⎥⎢ ⎥ (3)
⎣⎢iq (ωh ) ⎦⎥ ⎢⎣Ydq (ωh ) Yqq (ωh ) ⎥⎦ ⎣⎢vq (ωh ) ⎦⎥
Thus the equations to calculate DQ admittance are:

3342
 ⎡ ⎤
⎢cos(ω + ωh )t + cos(ω − ωh )t ⎥
⎡ Δva ⎤ ⎢ ⎥
⎢
Δv
⎥
==
1
E ⎢cos((ω + ω )t − 2π ) + cos((ω − ω )t − 2π ) ⎥
⎢ b⎥ 6
h⎢ h
3
h
3 ⎥
⎢⎣ Δvc ⎥⎦ ⎢
⎢cos((ω + ωh )t + 2π 2π ⎥⎥
) + cos((ω − ωh )t + )
⎢⎣ 3 3 ⎥⎦
(6)
It can be seen: a ωh perturbation in d-q frame will be
transferred into one perturbation at ( ω +ωh ) and one
perturbation at ( ω -ωh ) respectively by dq-abc
transformation at fundamental frequency ω . For each of
the two frequencies, the corresponding perturbation
current can be derived. However, whether the total current
generated by total two perturbations is equal to the sum of
the two currents generated by each perturbation, depends
on whether the superposition rule is satisfied in diode
rectifier circuit. If it is satisfied, the three phase impedance
Fig.3. At Eh sin(ω + ωh ) and Eh sin(ω − ωh ) injection (a)
model of the diode rectifier which has been derived before,
can be used here to obtain the current at each perturbation. current ω + ωh term real part (b) current ω + ωh term imaginary part
(blue: double injection’s result; red: sum of single injection’s results)
C. Superposition verification
Diode rectifier circuit is a nonlinear circuit. However,
for small signal analysis, it is linearized by doing FFT on
the non-linear square waveform switching function, thus
superposition should be valid for any given operating
point. . In addition, the superposition rule is verified to
be effective for diode rectifier system through simulation.
There are three steps for the superposition rule verification.
The first is to find the target frequency of the abc current
that contributes to the corresponding frequency in the d-q
axis. The target frequencies that will finally generate the
ωh component in d-q frame are （ ω + ωh） and
（ ω -ωh） .The second is to inject the voltage perturbation
separately and derive the aforementioned current at target
frequency for each injection. A perturbation voltage of
Eh sin(ω + ωh ) is injected into the ac side, at which case
the current at ω + ωh and ω − ωh can be obtained. Fig.4. At Eh sin(ω + ωh ) and Eh sin(ω − ωh ) injection (a)
Assume Eh =0.02 here. Also a perturbation voltage of current ω − ωh term real part (b) current ω − ωh term imaginary part
Eh sin(ω − ωh ) is injected into the ac side, at which case (blue: double injection’s result; red: sum of single injection’s results)

the current at ω + ωh and ω − ωh can be obtained. The According to the results of Fig.4 and Fig.5, the
third is to inject the total voltage perturbation at one time superposition rule in this small signal injection case is
and extract the target frequency component from the verified. Therefore, the current components at frequency
resultant current. So a perturbation voltage of ( ω +ωh ) and ( ω -ωh ) that are generated by the
Eh (sin(ω + ωh ) + sin(ω − ωh )) is injected into the ac side,
perturbation of combined ( ω +ωh ) and ( ω -ωh ) voltage
at which case the current at ω + ωh and ω − ωh can also can be derived separately for each perturbation first and
be obtained. In conclusion, if the summation of current then be added together.
ω + ωh term from the first and second injection is equal
D. Derivation of ABC current at perturbation frequency
to the current ω + ωh term in the third injection, and if
The abc current under abc voltage perturbation can be
the same thing happens for ω − ωh frequency, the
derived according to the model introduced in eq.(1). The
superposition rule is verified. first important thing in deriving the abc current is to define
which are the interested frequency. The target is to derive
the ωh term in d-q current, thus the interested frequency
in abc domain are (ω + ωh ) and (ω − ωh ) , since the
dq-abc transformation matrix only contains fundamental

3343
frequency ω . The second important thing is which Case 1: If Δvd = Eh cos(ωh t ), Δvq = 0 , the perturbations
voltage perturbation component can generate this in abc axis are shown in eq. (5).
component. Since the switching function contains the
Thus the current at frequency ω + ωh and ω − ωh are:
frequency of multiple times of fundamental frequency (the
multiple here is equal to either 6l+1 or 6l-1), the voltage at Δia (ω + ωh ) = Δia C (ω + ωh ) + Δia 2C (ω − ωh ) (11)
frequency of (ω + ωh ) and (ω − ωh ) will both contribute Δia (ω − ωh ) = Δia C (ω − ωh ) + Δia 2C (ω + ωh ) (12)
to the current at ( ω + ωh ) or the current at (ω − ωh ) . For , where Δia C is the current generated by the cosine
the perturbation in (5), the cos(ω + ωh )t term can result in perturbation and by the same frequency perturbation;
Δia C is the current generated by the cosine perturbation
both (ω + ωh ) and (ω − ωh ) components in the current,
but by the cross-coupling frequency.
as well as the cos(ω − ωh )t term. The third point is whether Case 2: If Δvd = 0, Δvq = Eh cos(ωh t ) , the perturbation
the source voltage generates the corresponding frequency in abc axis are shown in eq. (6).
term in current. The answer is no here because the source
Thus the current at frequency ω + ωh and ω − ωh are:
voltage and the switching function both only contains the
multiple times of fundamental frequency. Δia (ω + ωh ) = −Δia S (ω + ωh ) − Δia 2 S (ω − ωh ) (13)
Thus the final current is formed of two parts: one is the Δia (ω − ωh ) = −Δia S (ω − ωh ) − Δia 2S (ω + ωh ) (14)
(ω ± ωh ) term generated by (ω ± ωh ) voltage; another is The meaning of Δia S here is that the current is generated
the cross-coupling term: the ( ω ± ωh ) current generated by sine voltage perturbation
by (ω ∓ ωh ) voltage. E. Derivation of DQ current at perturbation frequency
There are four key equations related to this: The final goal is to get id (ωh ) and iq (ωh ) , which is
Δi (ω ± ωh ) = f ( Eh cos(ω ± ωh ))
id , iq at ωh frequency. The DQ current at ωh can be
Δi (ω ± ωh ) = f ( Eh sin(ω ± ωh ))
(7) obtained by multiplying the abc current by the
Δi (ω ± ωh ) = f ( Eh cos(ω ∓ ωh )) transformation matrix M abc _ dq . At d axis injection,
Δi (ω ± ωh ) = f ( Eh sin(ω ∓ ωh )) assume:
The difference between the first and the second, or Δida (ω + ωh ) = A1 + jA2
between the third and the fourth is that one is the cosine Δidb (ω + ωh ) = C1 + jC2 (15)
injection and another is sine injection. This is equivalent to
Δidc (ω + ωh ) = E1 + jE2
shift the phase angle of the perturbation by 90 degrees.
The first and the second equations can be directly derived and assume:
from the eq. (1), where i (ω x ) = Eh (ω x ) / Z (ω x ) . The Δida (ω − ωh ) = B1 + jB2
phase b and c holds the relationship with phase a: Δidb (ω − ωh ) = D1 + jD2 (16)
j 2π − j 2π Δidc (ω − ωh ) = F1 + jF2
Δib (ωh ) = Δia (ωh ) ⋅ e 3 , Δic (ωh ) = Δia (ωh ) ⋅ e 3 (8)
All the coefficients in the above equations can be
The cosine and sine injections hold the relationship: derived from equation (1) and (10), as analyzed in section
Δia _ cos (ωh ) = j ⋅ Δia _ sin (ωh ) (9) D. Thus through abc to dq transformation, the dq current at
The third and fourth equations in (7) needs to be derived ωh can be expressed as a function of these variables. The
separately. For example for phase A, if the injection q axis current generated by d axis perturbation injection is:
is sin(ω + ωh ) , the (ω − ωh ) term is: ⎡⎛ 1 1 1 1
⎢⎜ A1(ωh ) + B1(ωh ) − C1(ωh ) − D1(ωh ) +
3
C2 (ωh )
⎞ ⎤
⎟ ⎥
⎡ ⎡ 3 ⎤
⎢⎜ 2 2 4 4 4 ⎟ ⎥
(− Asn1 (φ , l ) Bsn 2 (φ , l ) ⎢⎜ 3 1 1 3 3 ⎟ ⎥
1 ⎢⎢ N ⎢ 4 ⎥ ⎢⎜ + D2 (ωh ) − E1(ωh ) − F1(ωh ) − E2 (ωh ) − F2 (ωh ) ⎟ ⎥
Δia _ sin (ω − ωh ) = − Eh ∑ ⎢ ⎥ 2 ⎢⎝ 4 4 4 4 4 ⎠ ⎥
6 ⎢l =1 ⎢ + 3 A (φ , l ) B (φ , l )) * ( 1
+
1 ⎥ Δiq _ d (ωh ) =
⎢ ⎢ 4 sn 2 sn1 )
Z (6lω − ωh ) Z (6lω + ωh ) ⎥⎦ 3 ⎢ ⎛ A (ω ) ⎞⎥
⎣ ⎣ ⎢ ⎜ 2 h − 3 C1(ωh ) − 1 C2 (ωh ) + 3 E1(ωh ) − 1 E2 (ωh ) ⎟ ⎥
⎢ ⎜ 2 4 4 4 4 ⎟⎥
⎡ ⎡ 3 3 ⎤⎤ ⎢+ j ⎜ ⎟⎥
⎢ N ⎢( − 4 Asn1 (φ , l ) Asn2 (φ , l ) − 4 Bsn1 (φ , l ) Bsn 2 (φ , l )) ⎥ ⎥ ⎢ ⎜ − 1 B (ω ) + 3 D (ω ) + 1 D (ω ) − 3 F (ω ) + 1 F (ω ) ⎟ ⎥
+ j ∑⎢
⎢ ⎥⎥ ⎣⎢ ⎝ 2
2 h
4 1 h
4 2 h
4 1 h
4 2 h
⎠ ⎦⎥
⎢l =1 ⎢*( 1 1 ⎥⎥
⎢ ⎢ Z (6lω − ω ) + Z (6lω + ω ) ) ⎥⎥ (17)
⎣ ⎣ h h ⎦⎦
⎡3 3 3 ⎤
+ ⎢ Asn1 (φ ,0) Bsn1 (φ , 0) + j ( Bsn12 (φ , 0) − Asn12 (φ , 0)) ⎥
1 ⎤ The d axis current generated by d axis perturbation
⎥
⎣2 4 4 ⎦ Z (−ωh ) ⎦ injection is:
(10)
If the injection is sin(ω − ωh ) , the (ω + ωh ) term ⎡ ⎛− 3
⎢ ⎜ C1 (ωh ) +
3 1 1 1 ⎞ ⎤
E1 (ωh ) + A2 (ωh ) − C2 (ωh ) − E2 (ωh ) ⎟ ⎥
⎢− ⎜ 4 4 2 4 4 ⎟ ⎥
generated by the injection is to replace the ωh by ⎢ ⎜ 3 3 1 1 1 ⎟ ⎥
⎢ ⎜− D1(ωh ) + F1(ωh ) + B2 (ωh ) − D2 (ωh ) − F2 (ωh ) ⎟ ⎥
−ωh in Δia 2 S (ωh ) . The case for cosine injection, as well Δid _ d (ωh ) =
2⎢ ⎝ 4 4 2 4 4 ⎠ ⎥
3 ⎢ ⎛ − A (ω ) 1 ⎞⎥
as the cases for phase b and c can be evaluated from the ⎢ ⎜ 1 h + C1 (ωh ) + 1 E1 (ωh ) − 3 C2 (ωh ) + 3 E2 (ωh ) ⎟ ⎥
⎢ ⎜ 2 4 4 4 4 ⎟⎥
− j
above case. ⎢ ⎜ ⎟⎥
⎢ ⎜ + 1 B (ω ) − 1 D (ω ) − 1 F (ω ) + 3 D (ω ) − 3 F (ω ) ⎟ ⎥
In summarize, according to superposition principle, ⎣⎢ ⎝ 2
1 h
4
1 h
4
1 h
4
2 h
4
2 h
⎠ ⎦⎥

3344
 (18) III. SIMULATION MODEL AND VERIFICATION
At d axis injection, the equations are similar, and the terms The rectifier circuit can be simulated in a numerical
are named as Δiq _ q(ωh ) and Δid _ q (ωh ) . Thus the dq simulation program to determine its small-signal DQ
admittance can be calculated according to eq. (4): impedance. To do so, a perturbation at a given frequency
iq _ d iq _ q f h is injected to the input voltage at dq axis, and the
i _d i _q
Ydd = d ; Yqd = ; Yqq = ; Ydq = d (19) circuit is simulated for a sufficiently long time to make it
Eh Eh Eh Eh
reach steady state. The (ω + ωh ) and ( ω − ωh ) component
The same method can be used to derive the negative
sequence d-q admittance/impedance.

Δvd

Δvq ia (ω + ωh ) Δvd
i (ω ) Ydd
Moving Average ib (ω + ωh ) abc / dq d h Δvq
at ω + ωh ic (ω + ωh ) at ω id (ωh ) Yqd
iq (ωh )
ia (ω − ωh ) Ydq
id (ωh ) iq (ωh )
Moving Average ib (ω − ωh ) abc / dq Yqq
at ω − ωh ic (ω − ωh ) at ω iq (ωh )

Fig.5. Simulation model of DQ admittance derivation of diode rectifier

in the input current can be extracted and then be 6

transformed into dq frame to do spectrum analysis. The 5
input impedance is calculated by dividing the perturbated
4
voltage by the perturbated current at the perturbation Ydq_Sim
frequency f h according to eq.(4).
Ydq

3 Ydq_Cal

TABLE I. PARAMETERS IN PU IN THREE PHASE DIODE RECTIFIER SYSTEM 2

Parameter E Rs Xs XL Xf Xd Rd
1
Case I 0.816 0.006 0.005 0.05 0.3 0 1
Case II 0.816 0.006 0.02 0.05 0.3 0 1 0
Case III 0.816 0.006 0.02 0 0.3 0 1 0 500 1000 wh 1500 2000 2500
70
3
60
2
50
1
40 Ydd_Sim
Ydd

Ydd_Cal 0 Angle(Ydq)_Sim
Ydq

30 0 500 1000 1500 2000 2500 Angle(Ydq)_Cal

-1
20
-2
10
-3
0
0 500 1000 1500 2000 2500 -4 wh
wh

3 Fig.7. Comparison between simulation and calculation results of (a)

Amplitude of Ydq (b) Phase of Ydq at case I
2
2.5
1
Angle(Ydd)

2
0 Angle(Ydd)_Sim
0 500 1000 1500 2000 2500 Angle(Ydd)_Cal
-1 1.5 Yqq_Sim
Yqq

Yqq_Cal
-2
1
-3
0.5
-4 wh
Fig.6. Comparison between simulation and calculation results of (a) 0
Amplitude of Ydd (b) Phase of Ydd at case I 0 500 1000 wh 1500 2000 2500

3345
 2.5
4
3
2
2
1 1.5 Yqq_Sim
0 Yqq_Cal

Yqq
-1 0 500 1000 1500 2000 2500 1
-2
-3 0.5
-4
f(Hz) 0
Fig.8. Comparison between simulation and calculation results of (a) 100 600 1100 wh 1600 2100 2600
Amplitude of Yqq (b) Phase of Yqq at case I
4
6
3.5
5
3
4
2.5
Yqd_Sim Yqd_Sim
Yqd

Yqd
3 Yqd_Cal 2 Yqd_Cal

2 1.5

1
1
0.5
0
0 500 1000 1500 2000 2500 0
wh
100 600 1100 wh 1600 2100 2600
Fig.10. Comparison between simulation and calculation results of
Amplitude of Ydd, Ydq, Yqq, Yqd of case II in table I
90
80
70
60
Ydd_Sim
50
Ydd

Ydd_Cal
40
30
20
10
Fig.9. Comparison between simulation and calculation results of (a)
0
Amplitude of Yqd(b) Phase of Yqd at case I
0 500 1000 wh 1500 2000 2500

7
50
45 6
40 5
35 Ydq_sim
4
Ydq

30 Ydq_Cal
Ydd_Sim 3
Ydd

25 Ydd_Cal
2
20
15 1
10 0
5 0 500 1000 wh 1500 2000 2500

0 3.5
100 600 1100 wh 1600 2100 2600
3
4
2.5
3.5
2 Yqq_Sim
3
1.5 Yqq_Cal
2.5
Ydq_Sim
1
Ydq

2 Ydq_Cal

1.5 0.5

1 0
0 500 1000 1500 2000 2500
0.5

0
100 600 1100 wh 1600 2100 2600

3346
 6 REFERENCES
[1] N. Sokal, “System oscillations from negative input resistance at
5
power input port of switching-mode regulator, amplifier, dc/dc
4 converter, or dc/ac inverter,” in Proceed. IEEE PESC’73, pp. 138–
Yqd_Sim 140, June 1973.
Yqd

3 Yqd_Cal [2] R. Middlebrook, “Input filter considerations in design and application

of switching regulators,” in Proceed. IEEE IAS ’76, pp. 366−382,
2
1976.
1 [3] S. Hiti, V. Vlatokovic, D. Boroyevich, and F. Lee, “A new control
algorithm for three-phase PWM buck rectifier with input
0 displacement factor compensation,” IEEE Trans. on Pow. Elec., vol.
0 500 1000 wh 1500 2000 2500 2, no. 2, pp.173−180, Mar. 1994.
Fig.11. Comparison between simulation and calculation results of [4] V. Vlatkovic, D. Boroyevich, and F. Lee, “Input filter design for
Amplitude of Ydd, Ydq, Yqq, Yqd of Case III in Table I power factor correction circuits,” IEEE Trans. on Pow. Elec., vol. 11,
no. 1, pp. 199−205, Jan. 1996.
The simulation model is shown in Fig.6. By swiping [5] M. Belkhayat. Stability Criteria for AC Power Systems with
the perturbation frequency and repeat it in a few times, the Regulated Loads. Ph.D. thesis, Purdue University, Dec. 1997.
[6] R. Burgos, D. Boroyevich, F. Wang, K. Karimi, and G. Francis, “On
impedance amplitude and phase curves from very low the ac stability of high power factor three-phase rectifiers,” in
frequency(1/20 of fundamental frequency) to enough high Proceed. IEEE ECCE’10, pp. 2047–2054, Atlanta, GA, Sep. 2010.
frequency (like 20 times of the fundamental frequency) [7] J.Weimer, “Electrical power technology for the more electric
can be obtained. aircraft,” in Proc. Conf. Proc. IEEE (DASC 1993), pp. 445–450.
[8] K. J. Karimi, “The role of power electronics in more-electric airplanes
The simulation and calculation are completed on a (MEA),” presentation at the IEEE Workshop Comput. Power
per unit value basis. The base values of the three phase Electron., Troy, NY, Jul. 2006.
diode rectifier system are: Vb=400V, Ib=50A, Rb=8ohm [9] F. J. Chivite-Zabalza, A. J. Fenrens, and D. R. Trainer, “Analysis and
with a fundamental frequency of 400Hz. Three groups of practical evaluation of an 18-pulse rectifier for aerospace
applications,” in Proc. IEEE (PEMD 2004), vol. 1, pp. 338–343.
parameters have been exammed, which are listed in table I. [10]A. Uan-Zo-Li, R. P. Burgos, F. Lacauz, A. Roshan, F. Wang, and D.
The feature of these three groups of parameters are: in case Boroyevich, “Analysis of new step-up and step-down direct
I, the ac side inductance is 1/10 of the dc side; compared symmetric 18-pulse topologies for aircraft autotransformer-rectifier
to case I, case II has a much bigger ac side inductance units,” in Proc.2005 IEEE Ind. Appl. Soc. Annu. Meeting,
pp.1142–1148.
which is 0.02; case III is an extreme case with a 0 dc side [11]R. D. Middlebrook, “Input filter consideration in design and
inductance. The reason to choose case I is that this case is application of switching regulators,” in Proc. IEEE Ind. Appl. Soc.
the most common case for line-commutated rectifier, Annu. Meeting, 1976, pp. 366–382.
which is also the closest case with the literature study [12]M. Belkhayat, Stability Criteria for AC Power Systems with
Regulated Loads, Ph.D. thesis, Purdue Univ., Washington, DC, 1997.
which assumes the ac side is an ideal voltage source and [13] Uan-Zo-li, A., Burgos, R. P., Lacaux, F., Wang, F., and Boroyevich,
the dc side is an ideal current source. Case II changes the D. Assessment of multi-phase converter average models for stability
ratio between the ac impedance and the dc impedance; if studies using a quasi-stationary small-signal technique.
the impedance matches in this case, it reveals that the [14] Alt, J. T., Sudhoff, S. D., and Ladd, B. E. Analysis and average-value
modeling of an inductorless synchronous machine load commutated
impedance model is adapted to different ac and dc line converter system. IEEE transactions on Energy Conversion, 14,
impedance, which also means the conventional 1(Mar. 1999), 37-43.
assumptions are broken by this model. Case III is an [15] Sudhoff, S. D., Corzine, K. A., Hegner, H. J., and Delisle, D. E.
extreme case but also exits in the practical industry; if the Transient and dynamic average-value modeling of synchronous
machine fed load-commutated converters. IEEE transactions on
impedance model is accurate at this case, that means it can Energy Conversion, 11, 3(Sept. 1996), 508-514.
tolerant almost all the cases between the extreme and the [16] Arrillaga, J., Smith, B. C., Watson, N. R., and Wood, A. R. Power
normal ones. Simulation results reveal that the amplitude system Harmonic Analysis. New York: Wiley, 1997.
of the perturbation voltage does not affect the input [17]Jian Sun, Kamiar J. Karimi, “Small-signal input impedance modeling
of line-frequency rectifiers,” IEEE Transactions on Aerospace and
impedance. So a 2% perturbation voltage with initial phase Electronic Systems, vol. 44, no. 4, October 2008.
angle equal to the fundamental voltage source is adopted [18]Jian Sun, Zhonghui Bing, and Karimi, “Input impedance modeling of
here. Simulation results and calculating results are multipulse rectifiers by harmonic linearization,” IEEE Transactions
compared in Fig. 7-11 for three cases. The agreement on Power Electronics, vol. 24, no. 12, Dec. 2009.
[19]Zhonghui Bing, Karimi, “Input impedance modeling and analysis of
between the two sets of responses can be observed, thus line-commutated rectifiers,” IEEE Transactions on Power
confirms the accuracy of the analytical models. Electronics, vol. 24, no. 10, October 2009.
[20]Zhonghui Bing and Jian Sun, “Frequency-domain modeling of
IV. CONCLUSION multipulse converters by double-fourier series method,” IEEE
Transactions on Power Electronics, vol. 26, no. 12, Dec. 2011.
A time domain Fourier series analysis and [21]Qin Lei; Miaosen Shen; Blasko, V.; Peng, F.Z., "A generalized input
time-domain mapping method have been used to derive impedance model of three phase diode rectifier,"Applied Power
the DQ impedance model of three-phase diode rectifier in Electronics Conference and Exposition (APEC), 2013 Twenty-Eighth
this paper, which has been validated through circuit Annual IEEE , vol., no., pp.2655,2661, 17-21 March 2013
simulation. The good agreement between analytical model
and simulation results demonstrates the accuracy of the
impedance model. This work could also be extended to
multi-phase rectifiers and phase-controlled rectifiers.

3347