General Disclaimer

One or more of the Following Statements may affect this Document

This document has been reproduced from the best copy furnished by the
organizational source. It is being released in the interest of making available as
much information as possible.

This document may contain data, which exceeds the sheet parameters. It was
furnished in this condition by the organizational source and is the best copy
available.

This document may contain tone-on-tone or color graphs, charts and/or pictures,
which have been reproduced in black and white.

This document is paginated as submitted by the original source.

Portions of this document are not fully legible due to the historical nature of some
of the material. However, it is the best reproduction available from the original
submission.

Produced by the NASA Center for Aerospace Information (CASI)

 tliAS1-Z!l-I-7.Et5) 111CISIC VA6iCIi rchif 177-12388
Ieilt IbflfGe 14I1Iti:Ei: SCli (CDS1111675 III
CCNSIL11IIIC3_ IbISA) i1 F EC ACt/Pf AC1
CSCI CSI unclas
G3/33 25100
I
ELECTR i C VEH I CLE POWER TRA I N I NSTRUMENTAT ION —

SOME CONSTRA I NTS AND CONS I DERAT IONS

James E. Triner and Irving G. Hansen

National Aeronautics and Space Administration
Lewis Research Center
Cleveland, Ohio 44135

Work performed for

ENERGY RESEARCH AND DEVELOPMENT ADMINISTRATION
Office of the Assistant Administrator for Conservation
Division of Transportation Energy Conservation
Under Interagency Agreement EC-77-A-31-1011

`f
Mr,Y 1:J'77

Technical Paper to be presented at International Electric RECEIVED

NASA S1i FACIUR
Vehicle Exposition sponsored by the Electric Vehicle Council INPUT BRANCH

Chicago, Illinois, Apri 26-29, 1977

s
 v

ELECTRIC VEHICLE POWER. TRAIN INSTRUMENTATION -

SOME CONSTRAINTS AND CONSIDERATIONS

by James E. Triner and Irving G. Hansen

National Aeronautics and Space Administration

Lewis Research Center
Cleveland, Ohio 44135

ABSTRACT

The application of pulse modulation control ( choppers) to DC motors

creates unique instrumentation problems. In particular, the high har-
un monic components contained in the current waveforms require frequency
Cq
^ response accommodations not normally considered in DC instrumentation.
W In addition to current sensing, accurate power measurement requires
not only adequate frequency response but must also address phase errors
caused by the finite bandwidths and component characteristics involved.
This paper will discuss the implications of these problems, and re-
port on the degree to which they have been solved at Lewis Research
Center.

INTRODUCTION

The DC series type motor has been used for many years for traction
type applications because it has desirable speed torque characteristics.
In early vehicles, power losses incurred with dissip « ive type regulators
were accepted as the penalty resulting from achieving the desired level of
control. With the development of high power semiconductor switches
[Ref. 1], the application of pulse width modulation techniques ( " chopping
substantially reduced the power system losses.

STAR category 33

i
 I

Pulse width modulation (PWM) controls the average voltage applied to

the motor by controlling the on time of the switch. The resulting average
current is controlled, but at the expense of high peak current drawn from
the source-or, high crest factor operation [Ref. 2 ].
. The effects of PWM on the rower system performance are the subjects
of other investigators [Refs. 3 and 4]. This paper documents measurement
errors resulting from the high crest factor current pulses of PWM and rec-
ommends instrumentation techniques to achieve the accuracies specified in
SAE XJ227(A) "electric vehicle test procedure" [Ref. 51 paragraph 3.4
which allows a maximum of f2% full scale error.
To understand the measurement errors incurred in the control and in-
strumentation of PWM controlled power systems it is necessary to define
the dynamic (AC) characteristics as well as the static (DC) characteristics
of the power train and the instrumentation sensors.
The DC characteristics are easily understood, and are sometimes re-
garded as the only input during system modeling and analysis. This i s mis-
leading since the AC characteristics hold the key to understanding the inherent
nonlinear characteristics of the total system. Questions such as overall effi-
ciency, system performance, input and output loading characteristics, and
system response can only be answered by accurate static/dynamic DC/AC
wideband modeling and analysis of the physical subsystem components.

Dynamic Characteristics

A simplified schematic for a DC/DC chopper is shown in Fig. 1(a). For

convenience, the series motor is assumed to be operating in a stable manner
with its armature current am) varying between Imax and Imin. When Im
is equal to Imin the switch is closed and Im begins to increase towards
Imax (Fig. 1(b)).
When the armature current reaches Imax the switch opens, and the
energy stored in the series field maintains the armature current through
the free wheeling diode. During this portion of the circle the terminal
diode (Fig. 1(c)).
This is the text book idealized description of chopper operation. It is
presented only to provide a foundation for the observations to follow:
(a) The current Ib flows from the source only during the time that the
switch is closed. It is to be noted that the rise and fall times of this current
are limited only by the external circuit parameters, not by the motor in-
ductance. In addition, brush shorting times, system storage, and various
stray effects all add high frequency components that must be measured in
order to define the actual input current. A typical input current would look
much like Fig. 1(d).. Due to battery droop and ER drops the voltage across
the chopper input appears as shown in Fig. 1(e).
(b) In addition to the case illustrated, in practice some choppers inter-
rupt the negative line to the motor. Such a system may have the negative
line connected to instrument ground, either by design or accidently by vir-
tue of degraded isolation. It is to be observed that this alternate connection
will present the instrument system with a common mode voltage at least
equal to the system supply voltage. This problem must be addressed in the
design of any instrument monitoring such a chopper system.

Instrumentation of a Chopper System

The two sets of fundamental parameters that are readily available to

analyze the chopper system are (1) input voltage and current, and (2) output
voltage and current. Present technology for wideband voltage measure-
ments present little problem in achieving accurate input/output voltage
measurements.
However, the instrumentation for accurate wideband current measure-
ments are not as readily available. DC sensor limitations such as range,
AC response, and nonlinear transfer characteristics, can often lead to
erroneous data. The results of utilizing this inaccurate data are incorrect
efficiencies, unexplainable system performance, unpredictable input/output
loading characteristics, and unreliable system response.
 4

Model of a Wideband Current Sensor

To aid in assessing the present state-of-the-art of wideband high cur-

rent measurements, an analysis will be shown for the equivalent circuit of
a high current sensor to determine the appropriate transfer functions. The
model can be reduced to a simple resistor RDC for the DC case, but is a
complex function of frequency under AC conditions. As shown below, the
model consists of a series-parallel network consisting of inductance L,
resistive elements RDC , which is independent of frequency, and R AC 9
which is a function of frequency due to skin effect, and capacitance C.

R RDC RAC

it C 12
(Input + _ (output
current eS current)

(Shunt voltage)

The following wideband properties can be derived for the above model:

DC + R AC + Jm L(1 - w 2 LC) - C(RAC + RDC)2

Z= R (1)
(1 - W 2 LC)2 + w2C2 (R DC + R AC) 2

w2L2 .1/2
IZI (RDC + RAC )2 +
(2)
(1 - W 2 LC) 2 + W2C2(R DC + RAC)2
 5

-1 w L(1 - w2 LC) - C(RDC + RAC)2

phase angle V =tan (3)
RDC + RAC

and

es
= Z s = current sensor transfer impedance (4)
i1

Note that as the frequency approaches zero, i. e. , w 0, the complex

DC/AC model reduces to

Z = RDC = RDC (5)

(1)2

1/2
Z I = ( RDC )2 = RDC
I 2 (6)
(1)

Phase angle q = tan -1 0 =0 (7)

RDC

I =RDC
ES (8)
1

Subsequently, voltage, current, and power relations can be represented

simply by applying Ohm's law:

ES
ES2
ES =I IIZDC I1= R PS =12R 1 DC =
R DC
 6

However, to apply these same equations for AC system parameters it is

I
clear that substantial error would be incurred due to frequency dependent
magnitude changes, and compounded further for power measurements due
to frequency dependent phase angle changes. In addition to the above fre-
quency dependent changes due to L and C, the effective AC resistance
('AC) of the model is also dependent on frequency due to skin effect. The
skin depth and the resistance per square in rationalized units, are [Ref. 4]:

1 /2
M
\\ITa µC

_ 1 SZ
RSQ Sv j
I A

where

C = 2.998x108 M
S

µ = 47TX10 -7 µr H
M

1 = 1. 72440 -8 -L q - M
a
µc

From these equations, the R AC term can be calculated for different types
of current probe configurations. For a rectangular configuration the
resistance/foot would be:

RAC AB RSQ B
A
and, for a cylindrical configuration:

RAC rrD RSQ

D
To make a comparison of commercially available wideband current
sensors, a study must be made of the magnitude and phase angle transfer
characteristics versus frequency. From these data one can determine the
major parameters such as frequency response, bandwidth, range, and per-
cent error for both current magnitude and power measurements.

RESULTS AND DISCUSSION

Comparison of Wideband Current Sensors

To determine the state-of-the-art of commercially available wideband

current sensors, four types were investigated and analyzed to determine
their applicability to the measurement of key electric vehicle power train
currents. The four types are the: (1) standard DC current shunt (e, g. ,
200 mV at 500 A), (2) narrowband clamp on DC current sensor (0 to
600 A range), (3) co-axial resistor current sensor, and (4) wideband
Hall effect current sensor.
The transfer characteristics of each of these sensors were measured
and are shown in Figs. 2 and 3. To analyze each of the sensor character -
istics, the equivalent model discussed earlier will be applied.
Standard DC current shunt - Equation (2) is used to determine the ac-
curacy of the shunt for current measurements only. For a typical standard
shunt, C << 1µF , L < 1µH, RDC < 10 m Q , and R AC = kV _f (0 ' RAC
RDC ). For these characteristic values, w 2 LC << 1 and w 2C 2 (RDC +
RAC )2 <'< 1, the magnitude equation is reduced to:

Z (_ (RDC + RAC)2 1 /2
+ w2L2
I

Plotting + Z I versus f with RDC = 0.0063n, L = 0.2µH, and RAC = kAfF

yields the plot shown in Fig. 2(a). There is good correlation between the
theoretical and measured impedance of the current sensor. The major non-
linearity in the magnitude of the probe is due largely to skin effect (RAC).
The useful bandwidth of this sensor for a 1% current error is 2 kHz. If this
sensor were to be used for power measurements , the effect of the phase
angle is shown by eq. (3). If W2 LC << 1 and C(RDC + RAC )2 << L the
phase angle equation is reduced to

cp = tan-1 WL
+
RDC RAC

Plotting cp versus f with the same values for L, R AC , and RDC yields
the plot shown in Fig. 2(a). There is a good correlation between the theo-
retical and measured phase angle co of the sensor. For power measure-
ments, the useful bandwidth for 1% error is 1 kHz.
Narrow band clamp on current sensor - The high frequency character-
istics of this sensor are deteriorated by the same effects as that of the con-
ventional sensor. However, the output of the basic current sensor is com-
pensated using an operational amplifier. Figure 2(b) shows the frequency
response of the sensor and compensated amplifier. The probe bandwidth is
1.3 kHz for a 1% error. The phase angle transfer characteristics of the
amplifier and the effects of the sensor itself can be seen in the cp versus f
plot (Fig. 2(b)) of the current sensor.. For power measurements the useful
bandwidth for 1% error is 225 Hz.
If this type of current sensor were applied to attain control information
from the electric vehicle power train. close attention would have to be paid
to maintain overall system stability due to the slim phase margin that this
type of sensor would provide.
Co-axial resistor current sensor - The transfer characteristics of this
type of current sensor (Fig. 3(a)) are almost identical to the DC case of the
current sensor model (eqs. 5-8). Both the magnitude and phase angle are
essentially independent of frequency up to 40 MHz (manufacturer's data).
 9
This result is consistent with the typical co-axial transmission line param.
eters where:

C = 2rrE F capacitance/unit length

ln(b) M
a

= 2ay Cl
C7 conductance/unit length
ln(b ) M
`a

In b H inductance/unit length
L EXT = -P—
2rr aM

LEXT is an external inductance and does not take into account any
flux within either conductor. At typical operating frequencies (<40 MHz)
the skin depth is so small that there is negligible flux within either con-
ductor and negligible interior inductance. These characteristics indicate
that this type of current sensor could be used as a standard for both analyz-
ing electric vehicle performance and calibration of other types of instru-
mentation. The accuracy of this sensor for both current and power meas-
urements can be << 1% for bandwidths up to 40 MHz.
Wideband Hall effect current sensor - From Fig. 2(b), it is clear that
the transfer characteristics of this sensor are almost identical to that of
the co-axial current sensor except for usable bandwidth. For current and
power measurements the useful bandwidth for <1% error is >20 kHz.

Fourier Analysis

For electric vehicle power trains employing some form of pulse width
modulation control, the ideal current waveforms can be represented as a
ratio of the on time to the total period T(TON + TOFF) as follows:
 10

14

t
Ton(DT) Toff (1 - D)T

Ton + Toff (T)

Since these pulse are continuous with tune they have a characteristic
for a constant load which appears as follows:

i(t)

By proper selection of time scale t, the input current pulses become even
functions of time:
 11

-T/2 -L)T/z T/2 t

U

The Fourier series of this PWM pulse train can be shown to have a
similar spectrum to that of the Fourier integral of a single pulse [Ref. 6].
This similarity is shown below:

DT/2
An = 1 e- j nt dt
Fourier series
An U

-1)JT2 , --00
ul)"f

o T'z t D sin -i I i 2
f (t) = 21u
nDT i h,ne spectrun,i

n=-x 2
-a-

DT/2
f;,urier
g(W) = 1 e-jwt di
T -DT/2
D T 2 i DT 2
DT
°` D sin w DT ,„
f(t) = 2 dwt dcv l r) l.
DT
-CC 2

to
 1 ^ -

12
Effect of bandwidth limitations - As shown in the previous section, the
inverse Fourier transform for the rectangular pulse is:
00

D sin w DT
f(t) = 2
ej wt dw
W DT
f_-O 2
Since the bandwidth of the measurement equipment is generally limited, the
limits of integration are restricted to w 0 << - thus,

w0
sin w DT
f(t) D 2 e, wt dw
W DT
2
WO

For typical electric vehicles applications the range of values for T

and D are:

0. 002 < T < 0.01 sec

0.01 < D < 1

For T = 0.006 sec and D = 0. 1 a quick inspection indicatFs that a band-

width of 3.33 kHz might be adequate to reproduce this pulse accurately.
However, this conclusion might be inaccurate unless supported by analysis
of the frequency spectrum. If the inverse Fourier integral were evaluated
for w 0 = 2rr • 3.3x103 the following result is obtained
E ^- w.

13

ec

f(t)
t
f0 = 3.33 kHz

Therefore, a pulse similar to the original pulse, with the exception of

slight rounding of the corners, is transmitted. However, if D = 0.01
the corresponding frequency response would have to be 33.3 kHz. Since,
the equipment would have a limited bandwidth of 3.33 kHz the effect on the
pulse transmitted would be:

VO
r0 g(^)

r.0 g0cj,' t dy'

^
1.
1
N ir -
t(t
J ^0
i I
t
f kHz

f0 = 3.33 kHz

Thus, the rectangular pulse has degenerated as a result of limited

bandwidth.
Similar results were reported by J. B. Folsom for the measurement
of "true -rms" signals [ Ref. 71. The error is plotted as a function of
bandwidth/fundamental frequency and crest factor. Crest factor is defined
as the ratio of the peak value to the rms value of the waveform. Similar
 14
results to those obtained by the inverse Fourier transform with limited
bandwidth are shown in Fig. 4 (1. e. , as DT — 0 bandwidth -- oo for
% error = constant).
This is an important result to take into account especially for meas-
uring the true characteristics of electric vehicle power trains operating
under PWM conditions. The instrumentation bandwidth must be adequate
to insure accurate data acquisition over the entire operating range of the
power train controller.
Based on the above investigations the co-axial type resistor will be
used as the Lewis Research Center standard in the checking and calibration
of other current sensors. Calibration of current sensors will comprise
both pulse calibration and gain and phase measurement as a function of
frequency.

Power Measurements

Figure 5 is a photograph of the terminal voltage measured on an

EV-106 traction battery under pulse discharging. Note that the battery
voltage, which is often times taken at its average value for power cal-
culations, sir ,.ificantly droops during the "on time. " Or stated in another
way the actual power delivered is less than would be indicated by an aver-
age voltage measurement.
The actual power drawn from the battery is the instantaneous product
of the voltage and current.
Compliance with the electric vehicle test procedure SAE XJ227(A)
[Ref. 5] requires a measurement error of less than t2%. At first glance
a power measurement having this accuracy does not appear to be too dif-
ficult to implement. However, many phenomena are involved, any one of
which may easily cause errors well in excess of the allowance. Some of
the more obvious of these are:
 15
(1) Envelope distortion: This is directly analogous to the phase shift
encountered in conventional single frequency AC circuitry, and indeed is
caused by phase shifts in the various frequency components involved. Any
distortion or delay of the waveforms will result in serious errors. These
distortions may be due to limited bandwidths and phase response of the sen-
sors as discussed earlier. Limited bandwidths in signal conditioning asso-
ciated with either the voltage or current measurement (adi area of particu-
lar concern is the voltage isolation amplifier), or e •-an quadrature compo-
nents magnetically induced in signal lines will create significant distortions
and phase shifts.
(2)Scale factors in the multiplier: Output offsets, thermal drifts, and
transfer linearities are typically specified relative to the full scale product.
A traction motor will draw peak armature currents many times larger than
its typical constant speed current. The result is a power measurement
having a full scale value much larger than the nominal power measured, {
with a consequent degrading of what would initially appear to be an accurate
measurement.
A final observation on bandwidths is in order. While the voltage and
current components involved in the power measurement have high frequency
components, power by definition is always an average. Therefore, the out-
put of the multiplier may have a relative limited bandwidth with no loss in
accuracy.
A laboratory type wideband wattmeter using electronic multipliers has
been evaluated at the Lewis Research Center. Power measurements were
verified to be accurate within A. 5% of full scale for all voltage and current
components with a bandpass of DC to greater than 20 kHz. This wattmeter
has been chosen for electric vehicle service testing and is presently being
modified for vehicle service.

J
 16

CONCLUSIONS

The application of pulse width modulation control (choppers) to DC

motors creates unique instrumentation problems. It has been shown that
high harmonic components in current waveforms require frequency re-
sponse accommodations not normally considered in DC instrumentation.
A mathematical equivalent circuit of a current sensor was developed.
It was experimentally verified that the model accurately represents the
response to higher frequency components. Standard shunts were found to
have the greatest error (greater than 2% at a few kHz and greater than
100% at 20 kHz) while the co-axial resistor specified was accurate to within
±0.25% from DC to 40 MHz [Ref. 81. Based on these results, Lewis Re-
search Center plans to use the co-axial resistor as a calibration stand-
ard. Electric vehicle current sensors will be of the wideband Hall effect
type, calibrated under pulse conditions against the co-axial standard
resistor.
Conventional power measurements utilizing standard current shunts
and averaging type meters will result in errors easily an order of mag-
nitude in excess of the f2% required by the SAE test procedure. The ab-
solute value of error will vary as a function of the pulse repetition rate
and duty cycle. Also, errors in power measurements will occur not only
from current sensing errors, but also are affected by phase and band-
width limitations of the voltage measurement, resulting in improper
definition of what constitutes the "input power" to the vehicle system.
 17

REFERENCES

1. Dewan, S. B. ; and Straughen, A.: Power Semiconductor Circuits.

John Wiley and Sons, 1975.
2. Amato, C. J.: Latent Losses in 'Lectric Lizzies. IEEE Trans.
Ind. Gen. Appl. , vol. IGA-5, no. 5, Sept. /Oct. 1969, pp. 558-565.
3. Demerdash, N. A. 0.; and Hamilton, H. B.: Effect of Complex
Forms on Copper Losses in Large D.C. Motors. IEEE Conference
Record of Fifth Annual Meeting of Industry and General Applications
Group, Inst. Electr. Electron. Eng. , Inc. , 1970, pp. 77-81.
4. Reference Data for Radio Engineers. 4th ed. , International Telephone
and Telegraph Corp. , 1956.
5. Electric Vehicle Test Procedure. SAE Recommended Practice
XJ 227(A).
6. Skilling, H. H.: Electrical Engineering Circuits. Wiley and Sons,
1965.
7. Folsom, J. B.: That 'True -RMS" Meter: Will It Be True to You?
EDN, vol. 20, no. 20, Nov. 1975, pp. 91-95.
8. Silsbee, F. B.: Notes on the Design of Four-Terminal Resistance
Standards for Alternating Current. Bureau of Standards Journal of
Research, Vol, 4, January 1930.
 ^ ^ v
IC
C
CD t
N N
C

u
U
C
p v
Zy Z,^
N N
,
^.. . rr
G.' 15
U = U V
^ 2 N
CI

D ^ A
Z ^ Y
p ^ N
H
X00 N N
_
L
Q c C
R
6 C d
A L
G
._ u
1O C
GJ v
!V
L L
^ A
Q C

pp pp Cp p p8
OD ^O ..r s R OD'_'
PA t4 N fV . I

001VWa0N ' S O
03ZIIVWSON 'So beD '& 69P

C W
O ^
Fyn
W
a LSu
o
U
U
D
W d 0

N
N
_

Q2 CD Z

I ^ L
I I U
p^ ( I I C
E- YJ I I I _

t 1 ^ I ^

^ I
U Z d
I I Ii G WQ V
Q I I i C1 ^
p _ W O^ C
_^ _^,_
____ w a
W O
a.
I
lI -O^
w I -:
to I I
I I Z ^
..pZ i I ~ O e@ o+
 2.
e
1.

Js
Y

120

9
a 0
.60

(a) CO-AXIAL RESISTOR CURRENT SENSOR.

2.0
8
N
1.0 0 n
zz

1w
$' 60
v
6 0

- 100 200 400 600 1000 2000 4000 6000 10 000 20 000
f, Hz
01 WIDEBAND HALL EFFECT CURRENT SENSOR.
Figure 3. - Gain and phase characteristics of a co-axial resistor current sensor and a
wideband Hall effect current sensor.

10

fDT

z
n 0 o
sT'
YO O?
z
K
D D IT o
1 t— -4
z
O ^ \
W
01 1 J__
10 100 1000 10 000 100 000
BANDWIDTHIFUNDAMENTAL FREQUENCY
Figure 4. - Error versus crest factor and bandwidth for rectangular
pulse trains.
N
.r
c^^
1
W
Figure 5. - Battery Ell characteristics.

NASA-Lewis