See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/3604540

Speed measurement using rotary encoders for high performance AC drives

Conference Paper · October 1994

DOI: 10.1109/IECON.1994.397844 · Source: IEEE Xplore

CITATIONS READS

28 4,962

3 authors, including:

Fernando Briz J.A. Cancelas

University of Oviedo University of Oviedo
222 PUBLICATIONS   4,986 CITATIONS    15 PUBLICATIONS   67 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

“Silicon Carbide Power Technology for Energy Efficiente Devices (SPEED)”, Ref. FP7-NMP3-LA-2013-604057 (Ref. Uniovi: UE-14-SPEED-604057) View project

Design of advanced control and modulation strategies for losses minimization of AC drives. View project

All content following this page was uploaded by Fernando Briz on 07 April 2015.

The user has requested enhancement of the downloaded file.

 Speed Measumment Using Rotary Encoders
for High Performance ac Drives

F. Briz', J. A. Cancelas", A. Diez

Departamento de Ingenieria E16ctrica. Electrbnica, de Computadores y Sistemas (DIEECS)

-
E.T.S. Ingenieros Industriales e Ingenieros Informhticos Gijon
-
Universidad de Oviedo Spain

k-mail: [email protected]
e-mail: [email protected]

-
A b s t d Most of high pelfonnance speed conbol of electrical lost, absolute position can be obtained with an incremental encoder.
motors, like ac Vector Conhol, use mtay encoders. The accuracy The main adventages of incremental encoders respect abolute ones,
obtaining the mtor speed fmm encoder signals is an essential is a smaller number of tracks, and indepentent of its precission, just
mquimment to achieve a good dynamic msponse, but it usually as a constructive simplicity. So this paper will be centered in this
implies a delay in the speed conbol loop, which can cause stability kind of encoder.
problems. The output signals of an incremental encoder are usually two
In this paper we will expose the altemotives for getting the speed square waves with an electrical gap of 90'. The phase shift sign into
from an encoder, highlighting their advantages and disadvantages, and these signals allows us to know the direction of rotation. It is also
mtifying them by means of simulation We will also show some easy to quadruple the number of pulses per revolution (PPR), by
e x p e h n b l lesulb obtained with an induction motor contmlled in detecting the edges of the encoder signals, as we can see in the
Efeld Cooniinaks using the pmposed methods. Figure 1.
In order to achieve a high performance control system, it would
INTRODUCTION be desirable to have an encoder with as many PPR as possible.
Nevertheless, this will be limited by the higher rotor speed, due to the
Most high performance speed control of electrical motors use maximum frequency of the encoder electronics, just as by
rotary encoders. Their first goal in the ac Vectrol Control Systems constructive limitations, mainly in the radial grating.
is to get the rotor position to implement field orientation, but, as it is There are two basic ways to get the rotor speed by means of an
already present, it is usually used to estimate velocity for the torque encoder: counting the encoder pulses for a known time, what is a
control loop. frequencimeter. or measuring the width of a pulse, wath is a
Rotary encoders can be absolute or incremental. The first one periodimeter. We will discuss the advantages and disadvantages of
supplies a binary representation of the absolute position of the both methods, along with some practical considerations.
encoder. An increment of its precission will imply an increment in It is also possible to improve the behaviour of the control system
the number of tracks, and therefore in the number of wires. The using together the encoder signal processing, other techniques such
incremental encoder supplies a square signal (supposed it includes as speed observers. They are usually based on the motor-load
digitizing electronics) as it rotates, with a known ratio pulseshad. So, mechanical ecuation [l], [2], [3], [4], [SI. The main parameter in
by counting pulses we can know the angular displacement from the these observers is the mechanical inertia, when it is unkown or
previous position, but no the absolute position. However, incremental variable, it may be necesary to include some identification schemme
encoders usually includes a zero signal. It is an aditional channel to get it [4], [SI. We will not by that matter in this paper.
which supplies a pulse in a fixed position of the encoder. By finding
this signal in the begining of the control, and supposing no pulse is SPEED MEASUREMENT
USING A FREQUENCIMETER
The simplest way to get the rotor speed is counting the number
of pulses from the encoder p, for a fixed time ,T,which is the speed
(torque) loop period (Figure 1). Being a the corresponding angle of
each encoder pulse, the rotated angle in a period will be:

PPR'Q
.j .j I I. 1. 1 I I I I I I ! ! I I L therefore the rotor speed will be
. . ia-
. . n*a
:

m
I
T m Wr= aJT (2)

The main disadvantageof this method is that, in each period, the

Figure 1.-Encoder output signals. measured angle a, will be always an integer multiple of a,which
Speed measurement with a frequencimeter. produces a quantization error in the estimated speed, with a maximum

0-7803-1328-3/94$03.0001994 IEEE
538
 , . .... ... ... ...
. .. ... ... ... .. . .
(00 ...........? - ' .
....4................ i..........,......... +.........i..........,..........
. . . . .
... " .. . ..
!
..
..
: .. " ..
....
"
....
.'
.... .,. . ... . .
... ............................................................................
..
..................
.,. ..
,..
.
..,. .. .
..... .,... ..... . .
... ... .
,
.. .. . . , . .
0 " " " " '

....... ..................... .....................

....... .....................

. . . . . . . . .
om am
0 M
aecon&
MI M1 om M Mu
A U

Figure 2.- Accuracy of the frequencimeter. Figure 4.- Speed measurement using
a frequencimeter. T=l ms.

.lo. i J
* ............................. +..................... i ..................... J ...................... :....... *' ;
, ....... ..................... c ..................... i...ura........... ..................... i....... , .............................. J- .................................................................. i ......._

4 -............................. +........................................... i..................... i

* .......i.....................
M 04u
;......................

Figure 3.- Speed measurement using

*
1 ..................... ;

LOI
...................... i

am
.......

Figure 5. Speed measurement with a

a frequencimeter. T=0.2 ms. second order, 500 Hz IIR filter.

value of d.
Obviously, this error will decrease as a decreases, that is, as the feedback deadtime T,which could affect the system stability.
number PPR increases, and as the period T increases. It is shown in We also can improve the quantization problem by digital
Figure 2, with the real and estimated speeds, obtained through filtering. This will produce an attenuation in the estimated velocity
simulation, when the motor brakes linealy from 157 radk (1 5 0 0 rpm) ripple, but will also introduce a lag in the transients which could be
to zero, with T=O.O2ms, and with an I500 PPR encoder, with critical. In Figure 5 we can see the same process as in Figure 2, but
quadruple frequency. We see that the absolute error does not depend using a second order, 500 Ht IIR (recursive) filter. The drastic
on the speed. Nevertheless its effects on the control system get worse reduction of noise is accompanied by an increment of the phase
as the speed decreases, and the vibration in the rotor takes a higher during the transients.
value in relation to the average velocity. We can conclude that a frequencimeter shows quantization
Figure 3 shows a simulation with the motor rotating first with problems, which makes them not very suitable at lower speeds. These
a constant speed of 6.28 rad/s (60 rpm), where we can evaluate its could be solved by using digital filters, or increasing the period of the
response in the steady state at low velocity, and then with an torque controller (which at last is a kind of moving average filter).
oscillatory (sinusoidal) one with a frequency of 120 Hz, where we Nevertheless, both solutions introduce a delay obtaining the velocity,
can extract some conclusions related to its trasient response. It shows which can degrade the control system perfotman~e~. In the former
the real speed in each sampling time T,and the measured speed. We simulations, with sinusoidal speed, the phase delay between the
see that at lower speeds there is a high quantization error, even in the hndamental component in the estimated speed, and the real speed,
steady state, which will produce disturbances in the torque controller, was slight with a sampling period T=0.2 ms, w20° with the digital
and as result of this, a system with a lot of noise and vibration. filter, and w35O with T=l ms.
This is especially critical in AC Vector Control schemmes. Its main advantages are that it is easy to implement,with little
Because fast torque controllers and high current loop bandwidth are hardware and a short processing time, providing good behaviour at
needed, noise in the measured speed goes trough the control loops, higher speeds.
producing a torque ripple. So, it is necesary to use soft torque
controllers, which of course will limit the system dynamic response. SPEED MEASUREMENT
USING A PERIODIMETER
Figure 4 shows the same process with a sampling period T=Ims.
which reduces appreciably the speed error. However, there is an If we use a periodimeter, we will get the speed not by counting
increment in the phase shift into real and measured speed due to the pulses from the encoder, but by counting a high frequency (HF) clock

539
 PHASE A JAL
!+F I

! n-t i
Figure 6.-Speed measurement with a periodimeter.

n4h
- 1 .. : .. : ..: : .. : .. : .. . ...
...
...
...
...
.....
.
...
.....
.
..
...
:

.....
.
:
..
...
.aped:...........i.......... J........... ;........... i...........j .
...
:.
.
..
!.
..
...
:.
.
I -- Figure 8.- Encoder output signals
........... . ......... .. ......... ............ <............:..
_i i i during an inversion.
. . . .

U
.., ..............................

i d
.

m
" ...............i................... ^. ................................ .-...................
0 on OI
wculaa
Figure 7.- Accuracy of the periodimeter.
U 001
", ...............
4
."."
....................
............... ...................
<
~

, ...............?..........
i
3

...
i
T...
. ....................
............... ................
>. . . ~

............... ................... .....................................

............... :................... .....................................
i
............
...................
............ ^.......

..........
*.................
L .................

............................. .....
for an encoder pulse, that is, measuring an encoder pulse period. It U ma U U U
r c m b '
is shown in Figure 6.
Let 4 be the corresponding angle to the higher part of an
encoder pulse (half of its period), t the period of the HF clock, and Figure 9.- Speed measurement by a periodimeter
n the number of HF pulses arrived at the HF counter. Then, the time without inversion detection system.
spent on an encoder pulse, T,, is:
frecuency, and PHASE A controls the HF counter, we see that at the
T, = n. t (3) encoder pulse end, the counter has a wrong value, because about half
of the HF pulses arrived on each direction of rotation. If the rotor has
The rotor speed can be obtained as: a constant acceleration, as shown in the illustration, at the encoder
pulse end, the real speed in case 1 will be greater than in case 2,
because there has been a greater time since the rotor crossed the zero
speed line. However, the HF counter in case 1 will be greater than in
As with the frequencimeter, here also there is a quantization case 2, and applying the equation 2, we obtain a greater speed in the
error because of the time T,,which will be always an integer multiple second case. We can conclude that if there is an inversion while the
of t With a fixed HFclock frequency, this error will increase as n HF counter is counting, its final value will be a random number and,
decreases, that is, as the encoder rotates faster, but it provides as we can see in the Figure 9, it will produce a wrong result, which
accurate measurements at lower speeds. Because the variable with can give rise to inadmissible disturbances in the control system.
quantization noise is now in the denominator, its effects are much If we use this method, it will be necessary to have some
more important than in the frequencimeter. This is clearly shown in additional system to detect the inversions. We can check the rotating
Figure 7, with the same braking process as in Figure 2. We can see direction in the beginning and at the end of the encoder pulse, that is,
that the absolute error depends on velocity, increasing its precision as while the HF clock has been amving at the HF counter. If it is not
velocity falls, but becoming useless at high speeds. An additional the same, there has been an inversion, and equation 2 becomes
advantage at low speed is that, as it only needs one pulse to get the erroneous. If the control system is not going to work with zero speed
speed, the delay is minimal, which is very important at lower speeds. or in position control, a very easy solution is to make an
It also possible to use quadruple frequency. Of course, this extrapolation using the previous velocity. assuming that this kind of
reduces the precision, and therefore the maximum velocity where this control system works with constant torque, and therefore with
method is suitable, but also, and this is much more important, it constant acceleration for the transients.
reduces the delay obtaining the velocity when the motor works at Figure 10 shows the measured speed with a periodimeter, taking
low speed. W=O when an inversion is detected during the HF clock counting. We
Nevertheless,this method produces stability problems in the zero can see that, at lower speeds, it provides precise results in the steady
speed cross. Figure 8 shows an inversion with two possible situations state, and a good response with variable speed, which allows us to
in the encoder output signals. Supposing that there is no quadruple use strong torque controllers without noise or vibration problems [6].

540
 .............................. ~ ....................

........................... ......................................... ........................ ..-

UI ooy am MI
&ndS
Figure 10.- Speed measurement by a periodimeter Figure 12.- Measured speed by a fmquencimeter
with inversion detection system. and by a periodimeter. W 4 . 2 Hz.

The accuracy of the periodimeter is, more than the

frequencimeter, related to the accuracy of the encoder, which in
general depends on the errors in the radial grating, and the
eccentricity of the graduated disk to the bearing. To avoid these
effects, just as the problems due to the cuantification error of the HF
clock when a very precise speed is needed, period of more than one
pulse can be measured, assuming that the delay introduced in the
control loop has no effects on stability.[7]

EXPERIMENTAL
RESULTS
The proposed schemes have been verified on a 6 Kw induction
motor, with Field Coordinates Vector Control. We have used an
encoder with the same characteristicsas in simulations, that is, 1500
PPR and quadruple frequency, and a HF clock of 3.125 MHz in the
periodemeter. The Vector Control has been implemented on a DSP. Figure 13.- Mcasured speed by a frequencimeter
The frequencimeter is based on the chip THCT12024 of Texas and torque controller output.
Instruments, which includes a working mode with quadruple
frequency and direction discriminator. Though it is also
programmable to make pulse width measurement, in this mode it does
not include direction discriminator, and as an inversion detection
system had to be included, it was decided to implement the
frequencimeter using an intemal counter of the DSP, with a few
additional external electronics. A general block diagram is shown in
Figure 11. The method to measure the speed is selected as a function
of its absolute value. The limit has been fixed at 30 Hz, where both
of them provide a good response. To avoid problems when the motor
works near this limit, a hysteresis in the commutation speed has been
implemented. A sampling time T=lms has been stablished in the
frequencimeter. and therefore in the torque controller, whithout any

Figure 14.- Measured speed by a periodimeter

and torque controller output.

additional filtering, which provides a good relation between noise and

dynamic response until some tenths of rpm.
Figure 9 shows superimposed the measured speed by a
frequencimeter and by a perodimether, with W-1.25 rad/s (12 rpm).
With the established period, the frequencimeter read one or two
pulses from the encoder on each period, which means 10 or 20 rpm.
If the controllers are strong, the control system makes the motor work
Figure 11. Implemented scheme.

541
 "rebounding" between these two limits, with an inadmissible level of This project has been financied by the Comision Intemacional
noise and vibration. Nevertheless, as is shown in the figure, the de Ciencia y Tecnologia (CICYT), with reference TAP92-0753.
periodimeter gives an accurate value, which allows us to use strong
controllers without the mentioned problems.
Figure 13 shows the measured speed through a frequencimeter
with a sampling time T=lms. and the torque controller response
during an inversion from -628 to 6.28 radls (-60 to 60 rpm). We can
see the speed steps due to the quantization error, and the torque
controller perturbation, which is transmitted to the motor.
Figure 14 shows the same process using a periodimeter, where
we can see the improvement in the estimated speed.
Though the process shown in Figures 13 and 14 worked with the
same torque controllers. in the last one it may be changed in order to
achieve a better dynamic response, whilst if we made stronger the
first one, it would imply an important increase of noise and
vibrations.

CONCLUSIONS
An essential requirement in Vector Control Systems is to have
an accurate and fast measure of the velocity. However, these two
specifications are opposed. We have shown the main alternatives for
measuring the speed using an incremental encoder, based on the
measure of the frequency and the period of its pulses, including
simulations and experimental results, where were shown the
advantages and limitations of both methods. The final proposed
scheme uses the advantages of both frequencimeter and periodimeter,
obtaining a fast transient response, just as an accuracy measure in the
steady state over the whole speed range. As no special hardware is
required, and the procesing time involved is minimal, it is a very
interesting altemative.

REFERENCES

Robert D. Lorenz et al. (1988). "High-Resolution Velocity

Estimator for All-Digital, ac Servo Drives". IEEE [ A S Annual
Meeting, Pittsburgh. Also published on IEEE Trans. on Industry
Applications. Vo1.27. no 4. July/August 1991.

Yasuhiko DOTE. Ph.D. (1990). "Servomotor and Motion

Control Using Digital Signal Processors". Prentice Hall.

Yoichi HORI et al. (1991). "An Instantaneous Speed Obsever

for High Performance Control of DC Servomotor Using DSP
and Low Precision Shaft Encoder", EPE'91.

Kouetsu Fujita et al. (1992). "Instantaneus Speed Detection with

Parameter Identification for AC Servo Systems". IEEE IAS
Annual Meeting, Seattle. Also published on IEEE Trans. on
Industry Applications. Vol. 28, no 4. July/August 1992.

Yoichi HORI (1993). "Robust and Adaptive Control of a

Servomotor using Low Precision Shaft Encoder". IECON93.

Femando BRIZ et al. (1993). "Design of AC Drives with

Position and Speed Dynamic Control Using DSP". CIEP'93.

Li-Cheng ZAI et al. (1992). "An extended Kalman Filter

Approach to Rotor Time Constant Measurement in PWM
Induction Motor Drives". IEEE Trans. on Industry Applications.
Vo1.28, no 1.

542

View publication stats