‘v . I .' . - u .. ‘. :.;.'r, 3"- 3 31.... 3..-..333 .Ifl." _' ”1'23 .‘ 3.3.!” 1g“ .3_333 II.-‘"I-3'I.;. ,3: 3‘ Orv—'3‘.
.3 "‘ """I.||I’III._ I'"‘ ui'Ir-‘II' I'II'.",.1.I IIIJ,’ 1..., I _. 3g; . '-.".J ‘ v. .
13.“. :.;f -.1 -.: IIIj'I VIII-1mm "'."= ;~'I: 223‘. II 33.3 4:; IEIIHIII' I?“ I I I,
-. III”, duh-3 '_.‘-’IIMI'u' «II:;.:III ."' 'IluIJIII .3. '.I.3I:g I...;:|’fI3I,,3.,34;'1';II‘.I r .- 333 I3 _3- '31 ..'. ..

" IIII ‘ 37.3 3.3‘ .'II' I" I. f '.‘ I ' I: I' I'3’I3 ."' ‘ ' ' . (.‘IMI’ u" ‘.’h .ji': 3. 'I' ..I

III ”I ”I I I..-I"..I' "I"1"I.=II“..iIII... qu' :II. l"L‘"" ..!I g .“ ‘E- F

.- ~.'-.3I 1... 333333333 . 33.333“ 333 ..

IH'IIII. IIIIi .‘ "' "‘ . _‘;‘_3I33 -' “III-PHLIv'I'I'II" 93133 I333: “3'3 3.7-I'I 3'
II 1' .I .c

- IOI A-
IN. 'I‘

‘ “’_EI‘) 'I-~+ Vfi {13“13‘3‘3‘Ifh é... fitnr. 133?;sz I‘.
35.»... 2...}. r... " '_.'.3 ii. .33- 33!:
','I ... 3

v ‘E' l" "
-.' am... .3- I133
“TI“

   
  
  
  

Etv_3r

    
   

.' 4 |Q1ir I; ::.'.3' 3'3, ..3
92“qu II" III; '-:3- - II

     

      
  
   
    
     
   

 
 
 

      
   

AMI?” I333;

 

MI...“

II J
3.»: K33 “‘I‘III3u3OI3: III|II3
3 .I!’I'._; ‘I. 121: [II ' v 11'.-
-:',l'_-III "rII'ILH‘I F.13’3"I'I'I33:3§F3I 333333
333: I I333 .333 III 3,333.11. 4&4}? 333333.313: ._.<,
7333? :{3 "- 7'

 

  

   
  
     
   
  

I .' I Lind]??? s; 3‘? o

:9': "- I!" 4‘3 ' '
'33I.:I: H'I. r 333(1303
‘4' 31.3..

 
 
 
 

I"

  
  

I, .
“$4.251...”
.I'fl ‘53,.

  

«was.»

.

154' "

("f .

'G‘ll

'i

W04. ;' ‘h‘

Ev.

1‘

3 4 :vamm

     
    
 
 
 
 
  
    

3 .-
«1.3.1333. . :

Ila

       
  

    

 

    

         
    
  

             
  

«.3._ I311! I; ';3‘£3(; .
. . #3 3 . II ., “‘nagt'lgII.’
. I ".'!‘- 51.13”. ‘ ' '1_I'EI';..; {M' I
. ' 3;} . ' ' ”I .I" - 'nfiIIII." ' 3 IIIIIII 13313.-
- :‘I'i-Ti E... J M‘ 'I‘I LIN“. 1°“- MI! :95
-,-.'} . '. , ”If - I '..‘é.-.I:O‘~ds,' fjé 3 5.31 “1-“ '
' '3‘ ‘ '- ' H '3". 1-H: “I." I"? 1 I
. 3». ' " ' ‘ $135, I [:_D_ 'nI
5' ‘ . - c: '. ' I 97-" I” II "
-.-. ' v: .5 ' 'I' .
z 3;... . I. -..=.I ‘ .‘ I.“
I :3: In 5 g 'i' 3 3 33 3 . 3.1“” " ”I
3" «*7 ~— ..~ %-':- =. .. 3'.
. . __ : - - ;_".*~'~'I 'e.‘

‘9. I
333.

n
.2

    
    
 
 
 
 
  
  
   
 

       

  
 
 

      
     
   
 
 
  

-' 7 Hf: ' .I!. 3.13
I... -‘ 'I " ".1 "' .- . ounl‘. i'. ...I
. Ii'~ II I? ~' "F": ‘I'. .. .. .'= ui' II.”- '.' ;. . I~-"-'
"‘ I I I H. ‘ . I“, - .5 . ‘ ’u
I -.-:‘:'. ’- 33‘" .~.-:.!;;. '- '~...‘.'oh-.r _II~.' . IL'. -':
'. 'v' 4' w". " "‘.. I ."I ~. .‘- i ' " "’ " i"?
I I433. 1‘ _‘ 3 3 . .. .
l 'quto W In"), I.
' ‘W'L: I'sz'I "'I‘L. u
I: I3... . ,,_ .333I3 Om} II
I ' “In.“ I
I. IF... 3 -
I .II3II'I . II | 33: 5:" 3
I .33’l||33 :l3'f";-3 '7" <II33“ “iii... '4‘3-313
I

I ;;v_J3 ;‘.l‘33331|_

   

1:3;
'3

 

  
 
    
 

  

    

gums" ' ' '33...
I..." II..." ...:3 '33: “.I ..,-.,.II

II: .I

 

...'— ’~..

new. llllllllllllllllllllllllllllllllllllllllllllllllllllllllll

31293 01581 9034

LIBRARY
MIchlgan State
University

 

 

 

 

This is to certify that the

thesis entitled

A BRUSHLESS DIRECT CURRENT MOTOR DRIVE
WITHOUT ROTOR POSITION SENSORS

presented by
Michael Paul Copeland
has been accepted towards fulfillment

of the requirements for

Master's degree in Electrical Eng.

 

 

Major professor

Date I M07! 9}

0-7639 MS U is an Affirmative Action/Equal Opportunity Institution

PLACE II RETURN BOX to remove We checkout from your record.
TO AVOID FINES return on or bdore dete due.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.l ICE—7
L__J-L_J

 

 

 

 

 

 

 

|__J-
T—T—T—j

MSU le An Afflrmetlve ActlorVEquel Opportunlty lnetltulon

A BRUSHLESS DIRECT CURRENT MOTOR DRIVE WITHOUT ROTOR POSITION
SENSORS

By

Michael Paul Copeland

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of '

MASTER OF SCIENCE

Department of Electrical Engineering

1 997

ABSTRACT

A BRUSHLESS DIRECT CURRENT MOTOR DRIVE WITHOUT ROTOR
POSITION SENSORS

By

Michael Paul Copeland

A drive which enables sensorless operation of brushless direct cmrent motors with
trapezoidal flux distributions is proposed. The proposed drive improves upon an older
version but is implemented using difl‘erent hardware and a microcontroller for greater
flexibility. The microcontroller allows easy implementation of more sophisticated control
schemes which result in extending the speed range of the motor. Experimental results
show that the proposed operating scheme without modification allows the motor to be
operated at speedsas lowas 55 rpmandhigherthanZSOO rpmwithreasonable dynamic

performance. More sophisticated control schemes can only improve these results.

ACKNOWLEDGMENTS

The work contained herein would not have been possible without the guidance of Dr.
Elias Strangas and the support of the students of the Electrical Machines Lab at Michigan

State University.

iii

TABLE OF CONTENTS

LIST OF FIGURES ............................................................................................................ vi
CHAPTER 1
INTRODUCTION ............................................................................................................... 1
Structure and Operation of BLDC Motors ............................................................... 1
Reasons to Eliminate Rotor Position Sensors .......................................................... 7
Challenges of Sensorless Operation ......................................................................... 8
CHAPTER 2
METHODS OF SENSORLESS OPERATION ................................................................. 10
Estimator Based Position Detection ....................................................................... lO
Discrete Position Detection .................................................................................... 11
Zero Crossing Detection Method ............................................................... 12
Ogasawara and Akagi’s Method ................................................................ 15
Disadvantages Of Ogasawara and Akagi’s Method ............................................... 23
Difficulty in Maximizing Motor Operating Speed Range ......................... 24
Problems with Dynamic Performance ....................................................... 27
CHAPTER 3
MODIFICATIONS TO OGASAWARA AND AKAGI’S METHOD .............................. 28
Proposed Improvements ......................................................................................... 29
Proposed Sensorless Operating Scheme ................................................................ 33
CHAPTER 4
EXPERIMENTAL SETUP ................................................................................................ 37
Chopping Selector Circuit ...................................................................................... 37
Detection Circuit .................................................................................................... 39
Inverter ................................................................................................................... 42
Brushless DC Motor .............................................................................................. 43
Microcontroller ...................................................................................................... 43

iv

CHAPTER 5 .
DESCRIPTION OF EXPERIMENTS AND RESULTS ................................................... 49

Determination of Constants ................................................................................... 49
Tuning and Validation of the Detection Circuit ........................................ 50
Determination of Kc ................................................................................... 54

Validation of Chopping Reduction and Chopping Frequency Adjustment Scheme
.................................................................................................................... 57

CHAPTER 6

FUTURE IMPROVEMENTS ........................................................................................... 59
Dynamic Load Compensation ................................................................................ 6O
Elimination Of Chopping ....................................................................................... 62

CHAPTER 7

CONCLUSIONS AND COMMENTS .............................................................................. 63

APPENDIX A .................................................................................................... , ............... 65

BIBLIOGRAPHY .............................................................................................................. 74

LIST OF FIGURES

Figure 1.1 - BLDC Motor .................................................................................................... 3
Figure 1.2- Developed Torque ............................................................................................ 4
Figure 1 .3 BLDC Motor with Hall-Effect Sensors ............................................................ 5
Figure 2.1- Line-to-Neunal Back EMF and Corresponding Switching Events ................ 13
Figure 2. 2- Inverter and Motor with A+ and B- Conducting ............................................ 15
Figure 2.3 - Inverter and Motor with D13+ and A+ Conducting .......................................... 17
Figure 2.4 - Line-to-Neutral Back EMF and Switching Events Including Chopping ....... 19
Figure 2.5 - Detection Circuit Using Floating DC Supplies .............................................. 20
Figure 2.6 - Motor Cross Section During Initialization of Starting Procedure .................. 22
Figure 2.7 - Motor Cross Section During Second Phase of Starting Procedure ................ 23
Figure 2.8 - Detection Error at High Speeds ...................................................................... 25
Figure 2.9 - Detection Error at Low Speeds ...................................................................... 26
Figure 3.1 - Detection Circuit Implemented with Current Transducers ............................ 30
F igm'e 3.2 - Waveforms of the Current Detection Circuit ................................................. 31

Figure 3.3 - Proposed System for Sensorless Operation .................................................... 33
Figure 3.4 - Simple Method of Commutation Time Prediction ......................................... 34
Figure 3.5 - Preferred Method of Determining Commutation Time .................................. 35
Figure 4.1 - Chopping Selector Circuit .............................................................................. 38
Figure 4.2 - Detection Circuit ............................................................................................ 40
Figure 4.3 - Flow Chart of Main Program and ITC ISR ...................... 47
Figure 4.4 - Flow Chart of Output Compare ISR ............................................................. 48
Figure 5.1 - A+ When the Motor is Operating Properly .................................................... 51
Figure 5.2 - A+ When Detection Circuit is Unbalanced .................................................... 52
Figure 5.3 - Phase Current and Its Detection ..................................................................... 53
Figure 5.4 - Phase Current When Kc is too Large .............................................................. 54
Figure 5.5 - Phase Current When Kc is too Small ............................................................. 55

Figure 5.6 - Phase Current For Proper Kc .......................................................................... 55
Figure 5.7 - Low Speed Error Correction .......................................................................... 56
Figure 61 - Algorithm for Dynamic Load Compensation ................................................ 61

vi

Chapter 1

INTRODUCTION

Brushless direct current (BLDC) motors are used in a wide variety of applications
due to their ruggedness, torque to volume ratio, and linear control characteristics. These
motors require rotor position sensors for reliable operation. The possibility of eliminating
position sensors has been investigated for some time. This thesis describes the theory
behind several rotor position detection methods and the implementation of one method in
more flexible hardware. Also discussed are some of the possible enhancements which

could be implemented with this new hardware.
1.1 STRUCTURE AND OPERATION OF BLDC MOTORS

BLDC motors have permanent magnets fixed to the rotor and stator windings
which are either sinusoidally distributed or rectangular. The method of operation is very
different for the two different types of motors. If the stator windings are sinusoidally
distributed, then as the rotor spins it will induce a sinusoidal voltage in the stator. If the
stator has rectangular windings the induced voltage (back emf) will be trapezoidal. The

nearly trapezoidal shape of the back emf is a property that will be exploited in this thesis.

2
This thesis concentrates on three phase wye connected BLDC motors that have

trapezoidal induced voltages and permanent magnets that cover the entire rotor
circumference. Throughout this thesis, BLDC motors with sinusoidally distributed back
emf will be called sinusoidal BLDC and motors with trapezoidal back emf and rotor
magnets that cover the entire rotor circumference will be called trapezoidal BLDC.
BLDC motors with rotor magnets that do not cover the entire rotor circumference are not
discussed here.

Figure 1.1 shows the wye connection of the three motor phases and a cross section
of a two pole motor. The outer ring of the motor represents the stator, and the inner circle
represents the permanent magnet rotor. Each of the three stator windings, labeled A, B,
and C (one for each phase), cover 120° of the stator. The stator is wormd so that when
the section of the winding labeled B carries current that directed out of the paper, the
section labeled B’ carries current that is directed into the paper.

During normal operation of a trapezoidal BLDC motor, only two of the three
motor phases conduct at any given time (sinusoidal BLDC operate with all three phases
conducting). This means that if, for example, phases B and C are conducting, current

must enter the motor through phase B and leave through phase C (or visa versa).

 

 

Phase Phase
A B
Neutral
Phase
C
Wye Connection Motor Cross Section

 

 

 

Figure 1.1 - BLDC Motor

To achieve the maximum torque the rotor poles should be aligned with the stator
windings that are conducting current, as shown in Figure 1.2. The torque is then caused
by the Lorentz force (B x i) that occurs when the current of the stator interacts with the
flux density of the rotor magnets, as shown in Figure 1.2. Figure 1.2 shows the cross
section of the motor when phases B and C are conducting and phase A is not conducting.
Current out of the paper is indicated by the small circles that contain dots, while current
into the paper is represented by the small circles that contain crosses. The developed

torque is represented by the arrows in the motor air gap.

 

 

 

 

 

Figure 1.2 - Developed Torque

Notice that as the rotor turns in the direction of the developed torque, the
maximum torque will continue to be developed for the next sixty degrees of rotation.
After sixty degrees, the rotor will be in the position indicated by Figure 1.3. At this
point, if the motor is to continue developing maximum torque, phase B must stop
conducting and phase A must start conducting (as shown in Figure 1.3). This event is

referred to as a commutation.

 

 

 

 

 

 

 

Figure 1.3 - BLDC Motor with Hall-Efl‘ect Sensors

Since the commutation depends on the position of the rotor, an encoder or Hall-
effect sensors (sensors which indicate which rotor pole is closest) can be used to
determine the proper time for the commutation. Figme 1.3 also shows where Hall-efl‘ect
sensors are normally placed in the stator.

The motor is generally fed by a three phase inverter which consists of the usual
six power transistors each equipped with a body or free-wheeling diode for protection as
shown in Figure l ..4 The DC voltage that 18 applied to the motor is usually increased or
decreased by simply chopping (turning off ) one of the inverter transistors at a fixed
frequency with a varying duty cycle. Speed or torque is then controlled by increasing or
decreasing the duty cycle'which respectively increases or decreases the applied DC

voltage.

 

 

 

 

 

 

 

Phase
2 - _ A
A B+ C+
VDC 25 25 25 Phase
__ B
A-

 

 

 

 

 

 

 

 

B- c- E
.T if 52 .3.

Figure 1.4 - Inverter with BLDC Motor Load

 

 

 

It should be noted that for the remainder of this thesis positive current will be
defined as current directed into the motor (or out of the inverter) and negative current is
current that is directed out of the motor (or into the inverter). So, for example, if it is
desired that phase A carry positive current and phase B carry negative current, the
transistors labeled “A+” and “B- “ must be forced to conduct. This will result 1n Vnc
being applied to the motor terminals. If only 25% of the value of Vnc is desired, then
either transistor A+, or B- (or both) must be turned off (chopped) 75% of the time. This

implies that a 25% duty cycle should be used for chopping.

7
1.2 REASONS TO ELIMINATE ROTOR POSITION SENSORS

There are essentially two “universal” incentives for eliminating rotor position
sensors. The first incentive is that rotor position sensors are a very common source of
failure in a BLDC motor. Therefore, their elimination would extend the useful life of the
motor. The other reason to remove rotor position sensors is that it requires extra steps in
the manufacturing process to add sensors to the motor. Therefore BLDC motors can be
manufactured less expensively if the sensors are removed. Although removing the
sensors requires extra control circuiu'y, this circuitry is external to the motor, and is easily
mass produced and less expensive to replace in the event of a failure.

In addition to these ‘fimiversal” benefits of removing rotor position sensors, there
are several specific applications in which sensors are not desired. As described in [1],
BLDC motors are often used in compressors and must be enclosed by a hermetic alloy.
This requires seals around all of the wires that connect the motor to the controller.
Removing the rotor position sensors will significantly decrease the number of wires
going to the motor and thus reduce the number of hermetic seals.

Sensorless BLDC motors are also desirable in environments where extra wires
effect the electro-magnetic interference (EMI) properties of the system. Sensorless
motors are very attractive for automotive applications since the EMI properties of an
automobile can be significantly effected by the number and length of wires that are

contained in the vehicle.

8
1.3 CHALLENGES OF SENSORLESS OPERATION

All sensorless operating schemes for BLDC motors rely on either detecting,
calculating or estimating the voltage that is induced in the stator by the moving rotor.

The rotor position can then be determined since the induced voltage has either a
trapezoidal or sinusoidal shape (with respect to the rotor position). This detection or
calculation is done by examining the phase currents and voltages. The problem is that the
back emf is very small at low rotor speeds and thus difficult to calculate. Of come there
is no back emf when the rotor is not moving. This makes starting the motor a challenge.
Most sensorless operating schemes require the motor to be operated in an open loop mode
until the rotor spins fast enough for the back emf to become detectable.

Open 100p starting procedures have several draw backs. The most obvious of
these is that the motor has no way of controlling acceleration or torque unless the load is
known. This makes open loop schemes susceptible to stalls and/or large torque ripple. If
the open loop starting procedure is too long, then the number of applications where the
sensorless operating method will be useful can become limited.

Another drawback to any open loop starting scheme is that the first time any
phase coil is energized, the rotor will move either forward or backward. The direction of
initial movement is determined by the initial rotor position which is unknown prior to
starting. Even if the open loop starting scheme lasts only one commutation, the rotor will
always have the possibility of momentarily moving backwards. This too may limit the

number of applications where an open loop starting procedure can be used.

9
Yet another problem with open loop starting schemes is that they can be very

inefficient since there is no way to ensure that the rotor poles are aligned properly with
the energized coils. If the proper stator windings are not energized at the proper times,
the rotor will draw more current than is needed. Since an open loop starting procedure is
required, a challenge of any sensorless operating scheme is to make that procedure as
short as possible.

Maximizing the motors speed range is also difficult for sensorless drives. As
mentioned before, the induced voltage is very difficult to calculate at low rotor speeds.
As will be shown in chapter 2, high rotor speeds can also present problems for sensorless

drives.

Chapter 2

METHODS OF SENSORLESS OPERATION

There are essentially twomethods of controlling a sensorless BLDC motor. Each
of these methods have several variations that are based on similar principles. One
method that is used on sinusoidal BLDC motors involves the use of an estimator to
determine the rotor position. Estimator techniques use the phase voltages and currents to
determine the rotor position. This method is discussed in section 2.1 and is included only
for completeness. The other method which can only be applied to trapezoidal BLDC
motors involves detecting when the rotor passes a known position. The most popular
variation of this method is discussed in section 2.2.1. Section 2.2.2 describes another
variation, which is the focus of this thesis. The disadvantages of this second variation are
discussed in detail in section 2.3. This will provide the motivation for the proposed

modifications which will be detailed in chapter 3.

2.1 ESTIMATOR BASED POSITION DETECTION

There are several estimator based techniques [2] for controlling brushless DC

machines, however most of these techniques are only applicable to motors with

10

. 11
sinusoidally distributed windings. The sinusoidal flux and induced voltage eliminate

many of the discontinuities found in trapezoidal machines and make this type of motor
well suited for coordinate transformations. Generally currents, voltages and sometimes
fluxes are transformed into two-phase quantities. Usually the two-phase quantities are
referred to the rotor frame of reference which decreases their time rate of change. The
transformation to the rotor flame of reference is done with an estimate of the rotor
position which generally contains some error. An observer is then used to drive this error
to zero.

Stator currents cannot be used to calculate rotor position unless they are in some
way effected by the back emf. Therefore the motor must be driven to some minimum
speed before any estimation based technique may be used to control the motor.

Estimator techniques for controlling BLDC motors may require sophisticated
control hardware including either a microcontroller or digital signal processor.
Estimation based control offers the advantage of fast torque/speed control since the phase
currents and voltages are monitored at a very fast rate which is independent of the rotor
speed. Unfortunately there has been little or no work done on applying advanced control

techniques such as estimation and adaptation to trapezoidal BLDC motors.
2.2 DISCRETE POSITION DETECTION

Trapezoidal BLDC motors operate with only two phases conducting at a time.

This leaves the open phase available for detecting the rotor position. The rotor position is

l2
detected only a fixed number of times per revolution, so it is desirable to have the

detection occur long enough before the next commutation so that the delays associated
with commutation can be accounted for. But if the detection occurs to soon before the
next commutation a dynamic load may change the rotor speed and cause the commutation
to occur at the incorrect time.

Section 2.2.1 describes a method of rotor position detection for trapezoidal BLDC
motors that is called zero crossing detection. There are several implementations of zero
crossing detection as described in [l] and [3], however the basic principles are the same.
General Electric (GE) has used this method in many of their products that contain
compressors. Another implementation of zero crossing detection, as proposed by Iizuka
et. al. in the work of [3], has been used in air conditioner imits since the early 1980’s.

Section 2.2.2 describes another method for positiOn detection that was proposed
by Ogasawara and Akagi [4] in 1991. This method is the basis for the rest of this thesis.
As will be seen in chapter 3 there are several modifications which can be made to make

Ogasawara’s method more effective.

2.2.1 ' ZERO CROSSING DETECTION METHOD

Zero crossing detection is a widely used sensorless operating method for
trapezoidal BLDC motors. This method relies on the nearly trapezoidal shape of the line-
to-neutral induced voltages. Figure 2.1 shows the shapes of the line-to-neutral induced

voltages and the switching pattern that corresponds to proper commutation.

13

 

  

F

   

 

:rjl

 

 

 

 

 

 

 

 

 

 

Figure 2.1 - Line-to-Neutral Back EMF and Corresponding Switching Events

The Shaded Area of F igme 2.1 indicates the period in which phase A is connected
to the negative DC rail, phase B is connected to the positive DC rail and phase C is open
circuited (only transistors A+ and B- are conducting). Notice that the next commutation
occurs thirty electrical degrees after the line-to-neutral voltage VCN crosses the zero
voltage level. So if the line-to—neutral induced voltages can be measmed, then the proper

commutation time can be determined.

14
Since phase C is open circuited, there is no voltage drop across the resistance or

inductance of that phase. This means that the induced voltage in phase C is present at the
motor terminal. If the neutral voltage is known then VCN can be easily determined.

The neutral voltage can be determined easily fiom the equivalent circuit as one
half of the sum of the terminal voltages of the conducting phases minus one half of the
sum of the induced voltages of the conducting phases. From Figure 2.1 it can be seen
that in the area of interest the sum of the induced voltages in the conducting phases
(phase A and B in this example) is zero. Therefore the neutral voltage is one half of the
sum of the terminal voltages of the conducting phases. This is the method that GE in [1]
uses to determine the neutral voltage. In the work of [3], the neutral voltage is
determined by examining the neutral of a wye connected resistive load that is fed by the
voltage at the motor terminals.

Afier the neutral voltage is determined, a comparator can be used to indicate the
instant that the induced voltage in phase C crosses zero. A motor operating with zero A
crossing detection can be controlled very similarly to a motor with Hall-effect sensors
except that the delay of thirty degrees from the point where the detection occurs to the
point when the next commutation occurs must be compensated for.

Zero crossing detection can work at low speeds since the induced voltage only
needs to be detected and not actually measured. The starting procedure in [3] requires
that the motor reach a minimum speed of 200 rpm before the closed loop control is
initiated. Unlike the method proposed in the following section, zero crossing detection

does not require the DC rail voltage to be chopped.

l 5
2.2.2 OGASAWARA AND AKAGI’S METHOD

In 1991 Ogasawara and Akagi [4] proposed an alternate method of rotor position
detection. This method is not based on terminal voltage measurements, but on the
conducting state of a free-wheeling diode in the open phase of the inverter. Consider the
equivalent circuit of the motor and inverter when transistors A+ and B- are conducting as
shown in Figure 2.2. In Figure 2.2 the solid lines of the inverter represent the sections
that are conducting current. The equivalent circuit of the motor including the winding
resistance, inductance and induced voltages are also shown in Figure 2.2. The winding
resistance and inductance are usually the same for each phase so RA = RB = RC = R and

LA = LB = LC = L. The induced voltages VAN, VBN and VCN can be seen in Figure 2.1.

 

 

 

 

 

 

 

 

 

Phase C

 

 

 

Figure 2.2 - Inverter and Motor with A+ and B- Conducting

l 6
Since no current is present in phase C, only the neutral voltage and the terminal

voltage are needed to determine VCN. The neutral voltage referenced to the negative DC

rail can be expressed as shown below:

di
‘iv =IDC "65 ‘R “LE-Km
=VCE +R'i+L"—“VBN
:
V _1
N 7013c av "th

where VCE is the voltage drop across the transistor during conduction and i is the
current through phases A and B. Figure 2.1 shows that when transistors A+ and B- are
conducting, the induced voltages V AN and VBN are equal in magnitude and opposite in
sign. This implies that the VN is simply one half of the DC rail voltage.

Now, if the rail voltage is chopped by turning off transistor B- for a short time, the
inductance of the motor will cause current to flow through the free-wheeling diode D3,.
until the current stops. Figure 2.3 shows the equivalent circuit when transistor B- is not
conducting and cmrent flows through diode D3... Again the solid lines on the inverter

represent the parts that conduct current.

17

 

 

 

 

 

 

 

 

 

 

 

 

Figure 2.3 - Inverter and Motor with DB+ and A+ Conducting

The neutral voltage referenced to the negative DC rail can now be derived as

shown below:

, di
’iv =‘bc TVCE 4445-52111
, di
=lbc+lb+R"+L;-‘13N

where VD is the forward biased voltage drop across a flee-wheeling diode. As

mentioned above, the induced voltages in phase A and B are approximately equal in

l 8
magnitude and opposite in sign in the period of interest (see the shaded portion of Figure

2.1). Therefore during the short period of time in which DB+ conducts the neutral voltage

can be expressed as shown below:

Notice that the voltage at the motor terminal of phase C is VN + VCN referenced to
the negative DC rail. If the terminal voltage becomes greater than Vnc + VD, then the
freewheeling diode Dc+ will start to conduct The derivation below shows the line-to-

neutral induced voltage that causes the diode Dc+ to conduct.

2}
VC’NZTITVTVDCTVD
:
V > V V 1 1
CN ‘_DC + DC '5’?) ”i3 +EVCE
2

Typically VD and Vcs are between one half and one volt. So as VCN goes slightly
positive, the free wheeling diode Dc+ will start to conduct. Similarly to zero crossing
detection as described in section 2.2.1 conduction occurs for the first time approximately

thirty electrical degrees before the next commutation. It is important to note that the zero

1 9
crossing of the back emf causes the free wheeling diode to conduct but that the zero

crossing itself is not detected.

Notice that if the analysis above is performed for the case when transistors A+ and
C- are conducting, then the transistor A+ will need to be chopped. This means that after
360 electrical degrees each transistor will go through a state where it is conducting for
sixty degrees and chopped for sixty degrees. Figure 2.4 shows the induced line-to-neutral
voltages and correspbnding switching states. The shaded area of the switching states

indicate that the transistor is being chopped.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 2.4 - Line-to-Neutral Back EMF and Switching Events Including Chopping

20
To detect the conducting state of the fieewheeling diodes another diode and

resistor can be placed in anti-parallel with each free-wheeling diode. The voltage drop
across the extra diode can be measured by a comparator as described in [4]. Figure 2.5
shows the structure of the detection circuit pr0posed by [4]. Note that one of these

circuits must be connected to each phase of the inverter.

 

 

 

 

 

 

 

 

 

Bias Voltage
“1
Detection
Free- 3 Diode
Wheelin
Diode E
To Motor Phase
Free- #5 %
ngcllmg SEDetection
l e I Diode I '
1
Bias Voltage

 

Figure 2.5 - Detection Circuit Using Floating DC Supplies

In the work of [4] the control of the motor is done very simply using a multiplexer
to select which comparator output to use and a digital phase shifter to implement the
thirty electrical degrees of delay that is needed for the next commutation signal.

The method proposed by Ogasawara and Akagi works at very low speeds (45 rpm

in [4]) since the diode conduction occurs whenever the induced voltage is greater than the

21
average of the forward voltage drop across the fi'eewheeling diode and the conducting

voltage drop across the power transistor (typically one half to one volt).

Since the minimum detection speed is so small, a very simple starting procedure
can be employed. As with any sensorless motor operating scheme the first time any two
phases are energized the rotor will move either forward or backward. If the phases are
energized long enough, the rotor will come to rest at a position where there is zero total
torque on the rotor. It should be noted that for a two pole machine there are two
equilibrium points for the rotor but only one stable equilibrium point. Therefore the rotor
will come to rest in only one position (assuming that fiiction is insignificant). Figure 2.6
shows a cross-section of the motor when two predetermined phases are energized and the
rotor has reached the stable equilibrium point. The circles in the stator indicate the
direction of current flow. The arrows in the air gap indicate the direction of the torque

caused by the Lorentz force (Bxi).

22

 

 

 

 

 

 

 

Figure 2.6 - Motor Cross Section During Initialization of Starting Procedure

If the next commutation then advances the switching sequence by 120 electrical
degrees (two steps) the rotor will accelerate at a constant rate for sixty degrees (assuming
constant load). The acceleration is constant because the current is constant and the flux
fi'om the magnet pole will be constant (therefore the torque will be constant) for the next
sixty degrees: The rate of acceleration is proportional to the current through the
conducting phases. Figure 2.7 shows the motor cross-section after this commutation.
Since the minimum detection speed is so low, the next commutation can be performed
automatically by the diode conduction detection circuit and the control hardware. So the
starting scheme consists of an initialization and only one additional open loop

commutation.

23

 

 

 

 

 

 

Figure 2.7 - Motor Cross Section During Second Phase of Starting Procedure

This discrete rotor position detection method allows very low operating speeds
and a simple, short starting procedure. However, there are several disadvantages to this
method which make it less attractive than zero crossing detection. If these disadvantages
can be overcome, the simple starting procedure and low operating speeds will make this

method superior to zero crossing detection.

2.3 DISADVANT AGES OF OGASAWARA AND AKAGI’S METHOD

The disadvantages of the discrete rotor position detection method of section 2.2.2

can be grouped into three categories. The first category is disadvantages associated with

24
the implementation of the detection circuit. The detection circuit described in section

2.2.2 requires six isolated voltage supplies, which are difficult to implement in a manor
which is both inexpensive and durable.

The second category is disadvantages associated with the operating range of the
motor. Proper explanation of this problem is the subject of section 2.3.1 . The third
category is disadvantages associated with dynamic performance, and is discussed in

section 2.3.2.
2.3.1 DIFFICULTY IN MAXIMIZING MOTOR OPERATING SPEED RANGE

The sensorless operating scheme described in section 2.2.2 detects the rotor
position only the instant that the inverter chaps. Therefore the time at which conduction
of the free-wheeling diode is detected may be as much as one chopping period after the
zero crossing of the back emf occurred. This means that when the rotor speed becomes
comparable to the chopping fiequency the time when the freewheeling diode conducts
may be become significantly different fiorn the time when the induced voltage crosses

zero as shown in Figure 2.8.

25

 

 

 

 

 

 

VAN
O V
\
\ \ .
\ .
.\'
..... ' \ h _ — — _. — — —
. Error
Pomt of zero Detection of
crossing \ /Diode Conduction
A+ FU‘UTLI—LI'LI L__

 

 

 

 

Figure 2.8 - Detection Error at High Speeds

If high rotor speeds are desired, the chopping fiequency must be comparably high.
In [4] the chopping is performed at a fixed frequency regardless of rotor speed. Chopping
at high fiequencies causes an increase in switching losses at all rotor speeds. A simple
way to lessen the degree of this problem is described in chapter 3.

In addition to the error that occurs at high speeds, an additional error occurs at low
speeds as well. This error becomes significant when the induced voltage becomes
comparable with the minimum voltage needed to cause the free wheeling diode to

conduct. For example, Figure 2.9 shows a motor that has a peak-to-peak induced voltage

26
of Kv [Volts/rpm]. If the average voltage drop of the inverter free wheeling diodes and

conducting transistors is Kc [Volts], the induced voltage must become greater than or
equal to Kc for the free wheeling diode to conduct. Figure 2.9 shows the case when the
rotor speed is low enough that the induced voltage is comparable to the voltage that is

needed for the free wheeling diode to conduct.

 

 

 

    

 

60° (Electrical)
VAN
1 :
ACK” x rpm) 3 Error 3
O V
Kc

--------- -------------------- d-----—-------------

 

 

 

 

Figure 2.9 - Detection Error at Low Speeds

For the machine to operate at its minimum possible speed, the digital phase shifter
described in [4] should account for this error. The phase shifter in [4] does attempt to
account for this error, but it does so by simply commanding the commutation to occur as
soon as the conduction is sensed whenever the rotor speed is below some predetermined
value. This is clearly not an optimal solution if the motor is intended to operate at low

speeds for any significant amount of time.

27
2.3.2 PROBLEMS WITH DYNAMIC PERFORMANCE

The DC rail voltage must be chopped so that the diode will conduct. Aside from
increased switching losses, this results in other dynamic performance losses. Since there
is always chopping, the maximum DC voltage can never be applied to the motor. This
results in a lower maximum speed and a lower maximum acceleration for any load.

Similarly to zero crossing detection, the rotor position is detected thirty electrical
degrees before the next commutation. The delay is large enough that a very dynamic load
could change the rotor speed enough to cause a misprediction of the commutation time.
The digital phase shifter with programmable delay used in [4] was made to account for
linear accelerations, but it will always mispredict the commutation instant when the
acceleration begins and when the acceleration ends. If the magnitude of this
misprediction is large enough, the motor will not run smoothly and the phase current may
become exceptionally large. If the load disturbance causes the rotor to slow down, the
next commutation may occur so early that the motor will not be able to deliver enough
torque to complete the rotation and the motor will stall.

With the exception of the problems discussed in the previous paragraph, all of the
disadvantages to this method of sensorless operation can be overcome or lessened by the
modifications proposed in chapter 3. These modifications may also lead to the eventual
elimination of all of the disadvantages to this method (including those mentioned in the

previous paragraph) as discussed in chapter 6.

Chapter 3

MODIFICATIONS TO OGASAWARA AND AKAGI’S METHOD

In Ogasawara and Akagi’s method, the conducting state of the a flee-wheeling
diode in the inverter is used to determine the rotor position. As mentioned before this
scheme allows the motor to operate at very low speeds. In fact the operating speed is so
low that the motor can be started by moving the rotor to some initial position and then
providing only one additional open loop commutation. However this method has several
disadvantages as described in section 2.3.

Section 3.1 is dedicated to proposed solutions to many of the problems presented
previously. In this section a new detection circuit is proposed that eliminates the need for
six isolated voltage supplies (see Figme 2.5). In addition to this the digital phase shifter
is replaced by a microcontroller which can be used eliminate the errors that occur at high
and low rotor speeds as described in section 2.3. The microcontroller can also chop in a
more selective manor so that some of the dynamic problems mentioned before can also be
lessened. The addition of the microcontroller and its increased abilities make it possible
to go even further to enhance the dynamic performance of the motor. This will be

discussed in chapter 6, and has yet to be implemented.

28

29
3.1 PROPOSED IMPROVEMENTS

Since the conduction of current in the free-wheeling diode corresponds to current
in the “open” phase of the inverter, the phase current can be sensed instead of the voltage
drop across the diode. Phase current detection can be done by adding a very small
resistor in each of the motor phases. This is already done many applications to
implement overcurrent detection. Or current can be sensed using current transducers.
Inverters for automotive and other applications may already contain three current
transducers so there should be little or no cost for this implementation.

The use of current transducers provides isolation between the phases and the rest
of the detection logic and makes it possible to implement current detection without the
use of isolated voltage supplies. Figure 3.1 shows the proposed current detection circuit.
Notice that the bias voltages for the comparators can be implemented with batteries or
more conveniently with Operational amplifiers since they reference a common ground.
One of these circuits should be used for each motor phase. Ideally the bias voltage
supplies for all six comparators could be implemented with only two operational

amplifiers. The value of the bias voltage depends on the current transducer and

comparator.

3O

 

Phase Wire Current
From Transducer Phase Wire
Inverter To Motor
Positive
Current

 

 

 

Measurement >£sent gmt
Current | '-

.. Detected
Measurement g .

 

 

 

 

 

 

 

Resistor >—
f— Negative
__ T Cm'rent
g: Present

 

 

 

Figure 3.1 - Detection Circuit Implemented with Current Transducers

When the motor is running with the proper chopping scheme the input and output

of the detection circuit will look like the waveforms in Figure 3.2.

31

 

Current Due To
Free-Wheeling
Diode

 

 

Phase W v
Current

 

 

 

 

 

 

 

 

 

 

 

 

Positive

Cm'rent

Present H '1
Negative

Current

Present H

Phase
Current
Detected

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 3.2 - Waveforms of the Cm'rent Detection Circuit

Of course the output of the detection circuit should only be monitored at the
proper time. Therefore there must be additional logic to gate the proper circuit output to
the controller.

To eliminate the errors that occur at high rotor speeds (see section 2.3.1), the
microcontroller can adjust the chopping frequency as the rotor speed changes. This
means that the chopping frequency does not need to be tmnecessarily high at low rotor
speeds. This will result in lower switching losses and still allow high rotor speeds. The
chopping frequency can be adjusted continuously as rotor speed changes, or one of only

two or three predetermined chopping frequencies can be selected as the rotor speed

32
reaches calibrated thresholds. Unfortunately the maximum rotor speed will still be

limited by the maximum chopping fiequency of the inverter.
The error in detecting the zero crossing at low rotor speeds (see section 2.3.1 and
Figure 2.9) can also be corrected by the microcontroller. The value of the error shown in

Figure 2.9 can be calculated using the following formula:

_ 59(3) _1_
Error — IO Kv P rpmz [seconds] (1)

where P is the number of poles the machine contains, and Kc and K, are the
minimum induced voltage for detection and the motor voltage constant respectively (see
Figure 2.9). Of course an alternate expression for this equation can be derived for more
convenient calculations inside the microcontroller.

Since chopping is required to detect the rotor position, the motor cannot achieve
maximum speed or maximum acceleration. This effect can be lessened if the chopping is
stopped just after current is detected in the open phase until the next commutation begins.
This will result in approximately 50% less chopping when the motor is commanded to
operate at its maximum voltage. Although this does not eliminate the problem it is
certainly a step in the right direction. This irregular chopping scheme may result in
torque oscillations. The experimental results (see chapter 5) discuss this potential

problem.

33
3.2 PROPOSED SEN SORLESS OPERATING SCHEME

The system proposed for operating the motor without rotor position sensors is

represented in Figure 3.3.

 

Chopping

Selector Inverter
Circuit

   
 
  

Micro-
controller

  
   
 

Detection
Circuit

 

 

 

Figure 3.3 - Proposed System for Sensorless Operation

The detection circuit is similar to that of Figure 3.1 except that three of the
circuits will be required (one for each phase) and extra logic is needed so that the
microcontroller can select which circuit output to read at the proper time. The chopping
selector circuit determines which transistor to chop. This function could be performed by
the microcontroller, but since the circuit is almost entirely combinational logic, it can be
easily implemented externally.

The microcontroller performs all of the calculations needed to determine the

commutation time. It must first perform the simple starting procedure, then determine

34
which phase current to monitor. Afier the current is detected in the open phase, the

microcontroller can use any of a variety of methods to predict the time for the next
commutation. The simplest prediction scheme is one in which the microcontroller
determines the time at which the current detection occurs (“B” in Figure 3.4) and
compares that time to the time of the last commutation (“A” in Figure 3.4). The change
in time (“AT” in Figure 3.4) will then be added to the detection time to determine the

time for the next commutation as shown in Figure 3.4.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Current Due To
Free-Wheeling
Diode
Phase V V
Current
. u A
Phase 0.,
Current
Detected 'A|'1|\ \ \
A AT B B + AT

 

 

 

Figure 3.4 - Simple Method of Commutation Time Prediction

35
Alternately, the microcontroller can compare the time at which current is detected

in the open phase (“B” in Figure 3.5) with the last time current was detected in the last
open phase (“A” in Figure 3.5) and add one half of the difference (“AT” in Figure 3.5) to

the most recent time that current was detected as shown in Figure 3.5.

 

 

Phase A V
Current Mill"

 

 

 

 

 

 

 

A . —/¥
A g ‘ *—B+AT

Phase C V
Current '

 

 

 

 

 

 

 

 

 

 

 

 

X '1

AT = '/z(B-A)

Current
Detected in '—
Phase A

 

 

 

 

 

 

 

 

 

 

 

 

Figure 3.5 - Preferred Method of Determining Commutation Time

In addition to either of these prediction schemes, the microcontroller must also
compensate for the low speed errors associated with (1). At high speeds, the
microcontroller must also increase the chopping frequency. When the inverter is

chopping at maximum duty cycle, the microcontroller should also signal the chopping

36
selector circuit to stop chopping after current is detected in the open phase as described

before. The implementation is this system is described in Chapter 4.

Chapter 4

EXPERIMENTAL SETUP

An experiment was performed to test Ogasawara and Akagi’s method with the
proposed improvements. This chapter describes the hardware and software used in the
experiment. This chapter is organized into sections that correspond to the components of

the proposed system as shown in F igm'e 3.3.
4.1 CHOPPING SELECTOR CIRCUIT
The chopping selector circuit of Figure 3.3 was implemented by the circuit shown

in Figure 4.1. The circuit represented in Figure 4.1 is slightly redundant, and contains

some extra logic to protect the inverter.

37

38

 

FROM TO
MICRO- INVERTER

CONTROLLER

 
 
 
 
 
  
  
 
 
  
  
 

A+:
TPU CH 1
A-:
TPU CH 12

8+:
TPU CH 1
B-:
TPU CH 1
C+:
TPU CH9
C-:
TPU CH 8

PWM:
TPU CH 0

CHOP:
TPU CH 6

 

 

 

Figure 4.1 - Chopping Selector Circuit

The microcontroller simply indicates which phases should be energized, provides
the chopping waveform, and indicates whether or not chopping is to be allowed. The
selector circuit applies the chopping waveform to the appropriate phase. The D-type flip-
flopsareusedtoensurethatanychangeininverterstatewillbesynchronizedwitha
change in the PWM signal. This is done to ensure that a phase cannot be turned “on” and

“ofl” again in an extremely short amount of time for the protection of the inverter.

39
The exclusive-OR gates are used for protection against shoot-through. Shoot-

through is caused when any two transistors in a given phase (i.e. A+ and A-) are
conducting at the same time. This causes a large current flow through the transistors and
can destroy the inverter.

As indicated by the use of both AND and NAND gates in the final stage of the
circuit, the inverter allows the transistors that are connected to the positive rail (A+, B+,
and C+) to conduct with high inputs. However the transistors connected to the negative
rail (A-, B-, C-) are active low.

To disable chopping the microcontroller. simply sets the value of em high.
This is only done when the PWM is at its maximum duty cycle and after the current has

been detected in the open phase.

4.2 DETECTION CIRCUIT

The detection circuit fiom Figure 3.3 was implemented using a circuit similar to
that of Figure 3. 1 but with some slight modifications. The experimental circuit for all of

the phases rs shown m Figure 4. 2. The signal labels include the names of the

microcontroller pins that either provide or receive the signal.

40

 

 

 

 

   
   

 

IN PHASE A:
Port E Pin 3

 

 

 

 

 

 

 

 

 

 
 
  
  

CURRENT
IN PHASE B:
Port E Pin 2

 

 

 

 

 

 

 

 

   

CURRENT
IN PHASE C:
Port E Pin 1

 

 

 

 

 

 

 

Figure 4.2 - Detection Circuit

41
The operational amplifies used for both comparators and bias voltage were

number LF347 which contains four amplifiers per integrated circuit. Since these
Operational amplifiers were used with :5 V rails, the diodes and resistors were needed so
that a -5 V output would not be applied to the two-input OR gates. Although six
Operational amplifiers were used to generate the bias voltages for the comparators, fewer
amplifiers (as few as two) may be used to generate the bias voltages for the three phases.
Notice that the bias voltages are all positive. One would expect the bias voltages for the
comparators that detect negative phase current (represented by a negative voltage across
the measm'ing resistor) to be negative. However the internal bias of the Op-amps and
current transducers allow a small positive voltage to be used. The value of the bias
voltages are less than 30 mV.

The structure of this circuit provides some insight into the control algorithm
executed by the microcontroller. For example, if the microcontroller initiates a
commutation in which phases C and B are conducting and phase A is Open, then the
signal labeled “CURRENT IN PHASE A” is monitored to ensure that the commutation is
complete. After the current in phase A has decayed, the microcontroller sets the signal
labeled “CHECK PHASE A”. This allows the detection of any current in phase A to

cause an interrupt to indicate that the induced voltage has just crossed zero.

42
4.3 INVERTER

The inverter used for the experiment was built in the Electric Machines Lab at
Michigan State University. It consists Of a power transistor module and a driver board.
The transistor module (also known as a power module) used is an Intellimod
PM3 ORFSO60 manufactured by Powerex. This module contains the usual six IGBTs and
free-wheeling diodes along with a seventh power transistor which can be used to short the
DC rail. The forward voltage drop across a conducting flee-wheeling diode (V n) was
measured to be 0.53 V. The saturation voltage drop across the power transistors (V CE)
was measured to be 0.99 V. The average of these two voltages is 0.76 V. This is the
induced voltage need to cause a flee-wheeling diode to conduct (K).

The power module also contains protection circuitry which may be activated
when the transistors are too hot, conduct too much current, or are driven with improper
voltages. The driver board provides isolation between the high voltage of the inverter
and the TTL signals produced by the chopping selector circuit. It converts the six TTL
signals into the six driving signals for the power module.

It should be mentioned that shoot-through protection is provided by two separate
sources. First the power module has internal protection fiom shoot-through, however this
protection is not intended to be used on a normal basis. In fact repeated activation of the
shoot-through protection devices on the power module is considered to be misuse and can

lead to damage. The second layer of shoot-through protection comes from the chopping

43
selector circuit of Figure 4.1. As mentioned before, the exclusive-OR gates prevent two

transistors in the same phase from being accidentally turned on at the same time.

4.4 BRUSHLESS DC MOTOR

The motor used for the experiment was a four pole Bodine 3/ 8 HP trapezoidal
brushless DC motor (type 34B6BEBL, model #3309). The motor is rated at 2.7 A and
2500 rpm. The rotor contains rare earth permanent magnets which produce an induced
voltage of 42 mV/rpm peak-tO-peak (K,) in the stator. Since the flee-wheeling diode will
conduct when the magnitude of the induced voltage exceeds 0.76 V the theoretical
minimum speed of the motor is 36.19 rpm.

The motor was originally intended to be Operated using the three Hall-effect

sensors attached to the back Of the machine. These sensors were disconnected for the

experiments.

4.5 MICROCONTROLLER

The microcontroller used for the experiment was a Motorola 68332. This is a 32-
bit processor capable Of operating at 16.67 MHz The 68332 also contains a Time
Processor Unit (TPU) which was used to facilitate the PWM and timing functions
required to operate the motor. The processor contains two 8-bit parallel I/O ports (ports E

and F) and the TPU has 16 channels which may be used as I/O or to implement any of the

44
pre-programmed timing ftmctions. All of the microcontroller inputs except the signal

“CURRENT IN OPEN PHASE” (Figure 4.2) use the parallel I/O port B. All of the
microcontroller outputs are implemented using TPU channels as output pins since port E
has only eight pins (port F was intentionally not used to leave room for future expansion).

The TPU’s PWM function was used to facilitate chopping. This pre-programmed
firnction generates a PWM signal of period and duty cycle specified by two registers.
This makes it possible to control the motor voltage by simply increasing or decreasing the
value in the duty cycle register. The period register can be adjusted to increase or
decrease the chopping frequency which will enable the motor to Operate properly at high
and low speeds as described earlier.

The TPU’s Input Transition Counter (ITC) fimction was used to capture a timer
value when the signal labeled “CURRENT IN OPEN PHASE” (Figure 4.2) transitions
fiorn low to high. This will also cause an interrupt. The interrupt service routine (ISR)
for the ITC then calculates the time at which the next commutation should occm' (actually
just the difference between the current time and the next commutation time is calculated).

The first closed loop commutation time is calculated using the simple method
shown in Figure 3.4. After the first closed loop commutation, the algorithm shown in .
Figure 3.5 is used. The slightly more complicated algorithm Of Figure 3.5 is preferred
since it is based on actual current detection and not a previous calculation. The more
simplistic method of Figure 3.4 is used for the first closed loop commutation because it is
easy to initialize since the time of the last Open 100p commutation can be used for

initialization. Unfortunately an interrupt must be generated to capture the time of the last

45
open loop commutation. Therefore the first time the ITC ISR is executed the time of the

last Open loop commutation is captured. The second time the ITC ISR is executed the
first closed loop commutation time is calculated using the method of Figure 3.4. During
the third invocation (and during all subsequent invocations) of the ISR the commutation
time is calculated using the method shown in Figure 3.5.

At low speeds, the commutation time (calculated by the method described above)
must be corrected since the induced voltage is comparable to the voltage needed for the
flee-wheeling diode to conduct. Since all of the timing is done using a TPU timer,
equation (1) must be modified to give the error in timer counts. Since the motor has four
poles and the TPU timer is configured to run at 65.10 kHz, equation (1) can easily be

modified as shown.

4 K
Erro = EEC—(1536 - 10 - 6)AC2 [counts] (2)
V

where AC is the value of AT (fi'om Figures 3.4 and 3.5) expressed in 15.36 us
increments.

Multiplying by the constant, 15.36x10'6 be can accomplished by shifting the
binary value 16 places to the right, or taking only the high word of AC2 if it is stored as a
long word. This is not exact but produces less than 1% error. The values for Kc and K,
can be determined easily from the characteristics of the inverter and motor respectively.

However the value Of Kc determined from the inverter characteristics is only valid if the

46
detection circuit is able to detect infinitesimally small amounts of current. The actual

back emf that must be Obtained for the detection circuit to detect the rotor position on
time must be determined through experiment (see section 5.1).

After the error is subtracted fiom the commutation time, the ITC ISR initializes
another TPU firnction called an Output Compare (OC). The OC causes an interrupt at the
calculated commutation time (from the ITC ISR). The OC interrupt service routine then
initiates the commutation. The commutation is complete when a low voltage is present
on the “CURRENT IN PHASE x” signal (x is the Open phase) for a predetermined
number of loops. The OC interrupt service routine then sets the proper “CHECK PHASE
x” signal (x is the open phase) so that current in the open phase can cause another ITC
interrupt. The OC ISR also adjusts the chopping frequency to allow high rotor speeds.
Finally, the ITC interrupt is enabled, which causes the process to repeat indefinitely.

Figure 4.3 shows a flow chart for the control algorithm executed by the main
program and ITC ISR. Figure 4.4 shows a flow chart for the OC ISR. Appendix A
contains the actual assembly code which implements these functions. It should be noted
that in the experiment, the motor spins in the Opposite direction than that implied by

Figures 2.1, 2.4, 2.6, 2.7, 2.8 and 3.5.

47

 

 

 

 

[initialize Microcontroller ]

I Disable Interrupts I

 

 

 

 

[ Turn Off Inverter l

Lgear all “CHECK PHASE” signals]

 

 

 

 

Initialize Rotor Position to:
A- & C+ on, Phase B Open

 

 

 

 

 

[Enable ITC Interrupts]

 

 

 

Advance Commutation by 120°:

 

 

 

 

 

 

 

 

    

 

 

 

 
 

 
   

 

 

 

 

 

 

 

 

 

 

 

 

A+ & B- on, Phase C Open
Set “CHECK PHASE A" (to Cause
Interrupt to get Initial Time)
M Second Time
1 Through?
Afier ISR has Executed Once Set
“CHECK PHASE C” (Initiate Closed
L°°P “mm” Shift Value Right
1 L
Endless Calculate Error
Loop
M Subtract Error
I Store Result in 10C Register 1
Enable 0C & Interrupts

 

 

 

 

 

 

 

 

 

 

 

 

@wm@

 

Figure 4.3 - Flow Chart Of Main Program and ITC ISR

 

48

 

 

L Disable Interrupts j

 

 

 

rInitiate CommutaticFl

 

  
    
  
 
  
  
 
    
  
   
  
 

“CURRENT IN
PHASE A” = 0

 

Set “CHECK
PHASE A” &
Clear “CHOP”

L__

 

 

 

 

 

YES

   

PHASE B” = O

    

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

f,
for N loops . Set “CHECK
PHASE B” &
Clear “CHOP
PHASE C” = 0
Set “CHECK
PHASE C” &
Clear “CHOP”
RPM > R YES Chop at High
9 Frequency
Chop at Low
Frequency
L
5
Enable ITC &
Interrupts

 

 

 

 

L
@m@

 

Figure 4.4 - Flow Chart of Output Compare ISR

 

Chapter 5

DESCRIPTION OF EXPERIMENTS AND RESULTS

Several experiments were run to determine the performance of the proposed
system. This chapter is organized by experiment. Section 5.1.1 describes the tuning and
validation of the detection circuit. Once the detection circuit was properly tuned, an
experiment was performed to determine the actual value of Kc used in equation (2), as
described in section 5.1.2. This value corresponds to the minimum speed at which the
motor can spin with optimal commutation. Section 5.2 describes experiments that were
performed to investigate the effects of changing the chopping scheme to lessen the

amount of chopping required and to allow for higher rotor speeds as proposed earlier.
5.1 DETERMINATION OF CONSTANT S

The constant K, which represents the minimum induced voltage at which current
can be detected in the Open phase, has a theoretical value of 0.79 V. The limited
sensitivity of the detection circuit results in a higher actual value for Kc. The actual value
of Kc was determined through a series of experiments. The first set of experiments was

designed to get the motor to operate optimally when the detection circuit is tuned to its

49

50
maximum operable sensitivity. After this was accomplished, a second set of experiments

was done, in which the value of Kc was determined by trial and error using the shape of
the phase currents to determine when the appropriate value was selected. Section 5.1.1

describes the first set of experiments and section 5.1.2 describes the second set.

5.1.1 TUNING AND VALIDATION OF THE DETECTION CIRCUIT

The value of Kc depends on how soon the detection circuit can detect current in
the open phase. If the detection circuit is tuned to its maximum sensitivity, the value of
Kc can be minimized, thus enabling the motor to operate at lower rotor speeds. The
sensitivity of the detection circuit depends heavily on characteristics of the current
transducers, the comparators, and the bias voltages. For a given cmrent transducer and
comparator, adjusting the bias voltages is the only way to adjust the sensitivity.
Unfortunately variations in the comparators and current transducers make it necessary to
individually tune the bias voltages to ensure that the detection circuit responds the same
for positive and negative current in each phase. The motor will not run optimally if the
current detector in one phase is more sensitive than the detector in the others, or if a
detector is more sensitive to positive current than negative current.

The first step in the timing process is to determine which comparator output
provides the least sensitivity. This comparator will be the limiting factor for the detection

circuit, and all of the other comparators must be tuned to match it. To determine the least

5 1
sensitive comparator we can examine the inverter input signals that are generated by the

chopping selector circuit while the motor is Operating with the proposed sensorless drive.
Consider the inverter input signal that controls the transistor A+ as shown in
Figure 5.1. When the motor is operating properly under steady state conditions the
transistor should be chopped for 60 electrical degrees and conduct for sixty electrical
degrees as shown in Figme 5.1 (see also Figure 2.4). The small spikes that occur when
A+ is not being chOpped are the product of noise from the microcontroller and can be

ignored.

 

 

 

 

 

 

Figure 5.1 - A+ When the Motor is Operating Properly

When the transistor A+ is chopped, the transistor B- is conducting and phase C is
open The microcontroller should be examining phase C for positive current (this will
indicate that the free wheeling diode is conducting). When the transistor A+ is not being
chopped, and transistor 0 is conducting (as in the right hand side of Figure 5.1), the

microcontroller should be examining phase B for negative current. If for some reason the

52 .
comparator which detects positive current in phase C is less sensitive than the comparator

that detects negative current in phase B, then the amount of time A+ is chopped will be

longer than the amount of time that A+ is conducting as shown in Figure 5.2.

 

 

 

W

 

 

 

Figure 5 .2 - A+ When Detection Circuit is Unbalanced

If the bias voltage of each comparator is adjusted to provide the minimum offset
(the minimum offset is the offset below which the motor stalls), then the maximum
amount of time that the inverter stays in any one state will correspond to the least
sensitive component of the detection circuit. The rest of the comparator offset voltages
can then be adjusted until the corresponding inverter state remains active for the same
amount of time as the state that corresponds to the least sensitive comparator. This will
ensure that the detection circuit is “balanced”.

The minimum offset voltages (for balanced operation) for the comparators that

detect positive current in phases A, B and C were determined to be 20.0, 15.3 and 23.8

mV respectively. The comparators that detect negative current in the motor phases were
unusual because they also required positive ofi'set voltages (negative voltages should be
expected for these comparators). The maximum values for the bias voltages (for

balanced operation) for the comparators that detect negative current in phases A, B, and C

53

were determined to be 5.9, 13.7 and 6.1 mV respectively.

Figure 5.3 shows the current in phase A along with the signal generated by the
portion of the detection circuit that indicates positive current in phase A. Notice that the

detection circuit detects current that is almost indistinguishable fiom noise when viewed

on an oscilloscope.

 

Phase
Current

Positive
Current
Detected

 

 

 

 

”LIL.

Current Detected 3min:

 

 

 

 

 

 

 

 

 

200 ms

{r

1.5A

 

Figure 5.3 - Phase Current and Its Detection

 

54
5.1.2 DETERMINATION OF Kc

After the detection circuit is set to its maximum balanced sensitivity, the
minimum value of Kc (and hence the minimum rotor speed for optimal operation) can be
determined by examining the phase currents when different values of Kc are used by the
microcontroller to determine the error correction term represented in equation (1). If the
experimental value of Kc is too large, then the low speed error will be too large as well.
This will cause the commutation to occur too soon. As a result of the early commutation,
the maximum induced voltage will not yet be present in the conducting phases. Since the
maximum induced voltage cannot “push back” against the applied voltage, the motor will
draw a larger current. As the rotor position increases and the poles are covered by the
conducting phases, the current will decrease back to its normal value. Figure 5.4 shows

the phase current of the motor when the value of Kc is chosen to be too large.

 

 

 

 

Figure 5.4 - Phase Current When Kc is too Large

55
If too small of a value is chosen for K, then the calculated commutation time will

be longer than the Optimal commutation time. This means that the rotor poles will pass

the energized phases and the phase current will increase as shown in Figure 5.5.

 

 

 

 

Figure 5.5 - Phase Current When Kc is too Small

When the proper value of Kc is chosen, the phase current should be flat (when the

phase is energized) as shown in Figure 5.6.

 

 

 

 

Figure 5.6 - Phase Current for Proper Kc

56
The actual value for Kc was determined by experiment to be 1.155 V. This

corresponds to a minimum rotor speed of 55 rpm.

After selecting the proper value for K, the low speed error correction term
represented by equation (2) was used to adjust the commutation time predicted by the
method of Figure 3.5. Figure 5.7 shows how the low speed correction term adjusts the
predicted commutation time. Note that the values are expressed in timer counts (1 count =

15.36 microseconds).

 

 

 

 

 

 

 

 

 

 

6000 . . .
+ + Predicted Commutation
5000- + —Adjusted Commutation
AC (Counts) 4000f +
+
3000- ++
+++

2000 ’ ++++++ l
1000 - M.

0 f . .

.0 50 100 150 200

Rotor Speed (RPM)

 

 

 

Figure 5.7 - Low Speed Error Correction

In addition to the correction factor produced by equation (2), an additional 20

counts are subtracted fiom the predicted commutation time to account for the delay

57
between the time the OC interrupt occurs and the time when the commutation actually

occurs. This delay is associated with the microcontroller interrupt service time and the

length Of the code that is executed before the commutation in the OC ISR.

5.2 VALIDATION OF CHOPPING REDUCTION AND CHOPPING FREQUENCY

ADJUSTMENT SCHEMES

As mentioned earlier, the proposed BLDC drive should stop chopping the inverter
rail voltage after current is detected in the open phase, if the chopper is running at its
maximum duty cycle. Attempts to validate this procedure and to quantify the actual
increase in obtainable voltage as well as torque ripple were not successful for the
following reasons:

It was observed that at high duty cycles, the detection circuit becomes less
sensitive which results in late commutations. It was Observed that at high duty cycles
(97%) and low rotor speeds (less than 600 rpm) the detection of current in the Open phase
occurs so late that it is impossible to initiate the next commutation on time. This is true
even when the commutation is commanded at the instant current is sensed in the open
phase. However this is not the case if the commutation is commanded the instant that
current is sensed in the Open phase at high rotor speeds (greater than 600 rpm). This is
fortunate since high duty cycles and low rotor speed are generally not allowed due to
restrictions on phase current, however high duty cycles and high rotor speeds may be
very common. Unfortunately, commanding commutation to occur as soon as current is

detected in the Open phase does not allow the microcontroller to stop chopping.

58
It may be somewhat misleading to refer to high duty cycles, since it is the amount

of time which the transistor is chopped and not the percentage of conducting time which
causes the detection circuit to become less sensitive. Since the TPU timer was configured
so that a chopping fi'equency of 3 kHz was achieved by making the chopping period 34
counts long, the maximum duty cycle was 97% (33/34). If the chopping period were
longer, the maximum duty cycle would increase without further loss Of sensitivity by the

. detection circuit.

Several experiments were run to determine the maximum rotor speed when
chopping at 3 kHz (a typical chopping frequency). The experiments indicated that rotor
speeds above approximately 2300 rpm were difficult to achieve due to the high speed
errors indicated in Figure 2.8. This is the same as the maximum speed in [4]. To allow

higher speeds the microcontroller was programmed to change the chopping frequency to
I 6 kHz when the rotor speed exceeds approximately 700 rpm. This allowed rotor speeds
well above the motor rated speed of 2500 rpm. However, decreasing the chopping period
also lowers the maximum duty cycle. In this case the maximum duty cycle drops from
97% to 94%. This decrease in voltage will caused a loss in maximum rotor speed Of

approximately 28 rpm (from 1299 to 1271 rpm) for a 50V DC rail.

Chapter 6

FUTURE IMPROVEMENTS

The proposed system has several flaws which may limit the number of
applications for which it may be used. There are two basic limitations to the proposed
system. The first limitation is the system’s inability to compensate for acceleration. A
correction factor could have been added to account for linear accelerations as was done in
the original system, however there are many applications where linear acceleration are
not common. A simple method for providing compensation for all types of acceleration
is proposed in section 6.1.

The second limitation is that chopping is required to detect the rotor position. The
possibility of not chopping is discussed in section 6.2. Both of the proposed future
improvements use the computational abilities of the microcontroller in an attempt to
create a hybrid drive that uses both estimation and detection techniques to Operate the

motor.

59

60
6.1 DYNAMIC LOAD COMPENSATION

To account for accelerations, the microcontroller can be used to integrate the
velocity of the rotor to determine position, and hence the proper commutation instant.
The linear speed-torque and torque-current relationships of the trapezoidal brushless DC
motor make it easy to calculate the motor velocity if the applied voltage and current are
known.

The integrator must be initialized to some known position. Since the detection of
current in the open phase corresponds to a known rotor position (after taking into account
the low speed errors), the detection can be used to initialize the integrator. The integrator
can then indicate when the commutation should occur, thus compensating for any type Of
acceleration or deceleration. Figure 6.1 indicates how the microcontroller activities could
correspond to the detection of current in the open phase. Notice that the integration
occurs only fiom the time current is detected in the open phase until the next
commutation. NO calculations are needed after the commutation occurs until current is
detected in the next open phase.

As mentioned earlier, the detection circuit becomes less sensitive at high duty
cycles, and requires the commutation to be initiated as soon as current is detected in the

Open phase. Therefore no integration will be needed when Operating at high duty cycles.

61

 

 

Trigging
Event

 

 

Current
Detected in

Open Phase

 

 

:Time

 

Micro-
controller
Tasks

 

 

 

Calculate
Rotor
Speed

Through
Phase

Current

 

Initiate

Commutation

When

Indicated by
Calculated

Position

 

 

 

 

 

Calculate
Rotor
Position
From
Speed, Kc
and Kv

 

 

 

 

Begin
Integration
Of Speed
Using
Calculated
Initial
Position

 

 

Wait Until
Next Time
Current is

Detected in

the Open
Phase

 

 

 

Figure 6.1 - Algorithm for Dynamic Load Compensation

 

62
6.2 ELIMINATION OF CHOPPING

There are many applications where the ability to apply the full rail voltage is
desired. This is impossible for the proposed system since the full rail voltage can only be
applied when there is no chopping (this is equivalent to chopping with a 100% duty
cycle). Fortunately the full rail voltage is usually only commanded at high speeds due to
current limitations. At high rotor speeds the induced voltage has a profound effect on the
phase current. This makes it easier to calculate the rotor position via integration of the
velocity for long periods of time.

It is proposed that if the duty cycle is commanded to 100%, the last occurrence of
current in the Open phase (from before the duty cycle was 100%) should initiate and
initialize integration of the velocity (as described above), and that the integration continue
until a lower duty cycle is commanded.

For example, assume that the motor is operating at a low duty cycle, and
integrating the phase current as described earlier. The integration was initialized the last
time current was detected in the open phase. If the duty cycle is then commanded to
100%, instead of ceasing the integration alter the commutation, simply continue until the
duty cycle is lowered and current can be detected in the open phase again. Since the
induced voltage is large, the integration should provide a sufficiently accurate estimate of
the rotor speed. This is a calculation intensive procedure that represents the transition

fiom a sensorless detection drive to a sensorless estimation based drive.

Chapter 7

CONCLUSIONS AND COMMENTS

The experiments of Chapter 5 prove that the proposed sensorless driveallows the
motor to operate at a wider speed range than the original system proposed by Ogasawara
and Akagi. Experiments show that the motor can operate at speeds from 55 to over 2500
rpm efficiently. The proposed drive should be more efficient than the drive in [4] since it
provides much more accurate error correction at low rotor speeds. The proposed system
corrects for low speed error down to 55 rpm, whereas the original system initiated
commutation as soon as current is detected in the open phase at all speeds below 500
rpm.

The proposed system can achieve a maximum rotor speed that is more than 200
rpm faster than the original system due to the implementation Of a variable chopping
fiequency. However the increase in chopping fi'equency is generally accompanied by a
decrease in the maximum applicable voltage since the DC rail voltage must be chopped
for some fixed minimum amount of time.

The minimum speed of the motor with the proposed system is approximately 10
rpm higher than that Of the original system. However this should be expected since the

motor used in [4] had a higher voltage constant (Ky) and a lower conducting voltage drop

63

64
across the flee-wheeling diodes and power transistors of the inverter (Kc). The theoretical

minimum operating speed of the motor used with the proposed drive is approximately 36
rpm whereas the theoretical minimum Operating speed Of the motor used with the original
system was 22 rpm. The proposed system is able to come much closer to its theoretically
minimum speed than the original system. This indicates that the use of current
transducers and the proposed detection circuit is equally if not more sensitive than the
detection circuit of the original system (when the duty cycle is not tOO high).

Attempts to reduce the amount of time that the rail voltage was chopped were not
successful due to an unexpected decrease in the sensitivity Of the detection circuit at high
duty cycles. This loss Of sensitivity, however, does not interfere with the motor’s ability
to achieve high speeds, but makes the drive less efficient when operating at low speeds.

It should be noted that speed and torque control were not mentioned in this thesis.
The goal of the proposed drive was to allow the motor to Operate properly, no matter
what the commanded speed or torque. This means that any speed or torque control may
be implemented independently of the commutation scheme.

In summary, the two main weaknesses of the proposed drive are the need for
chopping, and the inability to compensate for a dynamic load. These weakness may be
overcome using the modifications suggested in the previous chapter. In nearly all other
respects the proposed drive obtains superior performance when compared to the original

system.

APPENDIX A

65

*************eeea***********************************************

* This Program is intended to drive a brusless D.C.

* motor without rotor position sensors.

* The software requires very specific hardware to

* interface with the drive

*****************************************************************
org $4000

SYNCR EQU $FFFA04 CPU CLOCK SYN. CONTROL

REG

TPUMCR EQU $FFFE00 TPU Module Control Register

TICR EQU $FFFEO8 TPU Interrupt Enable Register

CIER EQU SFFFEOA TPU Channel Interrupt Enable Reg

CISR EQU $FFFE20 TPU Channel Intrpt. Status Reg

CFSR3 EQU $FFFE12 TPU Channel Function Select Reg

HSQRI EQU $FFFE16 TPU Host Sequence Register 1

CHOPARO EQU SFFFFOO TPU Channel 0 Parameter O Reg

CHOPAR2 EQU $FFFFO4 TPU Channel 0 Parameter 2 Reg

CHOPAR3 EQU $FFFFO6 " " 3

CHIPARO EQU $FFFF10 TPU Channel 1 Parameter 0 Reg

CHlPARl EQU $FFFF12

CH1PAR2 EQU $FFFF14 TPU Channel 1 Parameter 2 Reg

FINTRAN EQU $FFFF18

CH2PARO EQU $FFFF20 TPU Channel 2 Parameter 0 Reg

CH2PAR1 EQU $FFFF22

CH2PAR2 EQU $FFFF24 TPU Channel 2 Parameter 2 Reg

CH2PAR3 EQU $FF FF 26

HSRRO EQU $FFFE18

HSRRI EQU $FFFE1A TPU Host Service Request Reg 1

CPRO EQU $FFFE1 C

CPRl EQU $FFFE1E TPU Channel Priority Reg 1

PEPAR EQU $FFFA17 PORT E PIN ASSIGNMENT REG.

DDRE EQU $FFFA15 PORT E DATA DIRECTION REG.

FORTE EQU $FFFA13 PORT E DATA REGISTER

******Logic Signals For Inverter, Flags & Variables*****

*

STATO

A+A-B+B-C+C-

dc.w

%0000011010011010 First Inverter state A+ and B- on Coff

66

STAT] dc.w %0000011010101001 A+.C- Boff

STAT2 dc.w %0000101001101001 B+,C- Aoff

STAT3 dc.w %0000100101101010 B+,A- Coff

STAT4 dc.w %0000100110100110 C+,A- Boff

STATS dc.w %0000101010010110 C+,B- Aoff

STATZ dc.w %0000101010101010 All OFF

STATE dc.b $0 STATE OF INVERTER

F irsTlM dc.b 0 Flag to get initial time

LsTran dc.w 0 Last time spikes occured
SndTIM dc.b 0 Second time through ISRl (first comm)
ChOpMOde dc.b 0 Chopping Freq (0 = Low, 1 = HI)

********************************

*************Inten-upt VCCIOIS**********************

org $104
CHIVEC DC.L $10004 TPU Ch 1 Interrupt Service Routine location
CH2VEC DC.L $10204 TPU Ch 2 "

*****************************************************

**************** Initialization ************************

org $3000
move.w #$FFFF,D2
move.w #$CF80,SYNCR SET SYSTEM CLOCK TO l6MHz

*******SETUP PORT E AS INPUT PORT**********

move.b
move.b

#$00,PEPAR Set Port E to I/O function
#$00,DDRE Set Port E to INPUT

********SET UP TPU MAIN REGISTERS************

move.w #$608E,TPUMCR PRESCALE TCRl BY 8 SET IARB TO $E
move.w #$O640,TICR TPU INTRP LEV. 6 VECTOR AT 4*40
move.w #$8EA9,CF SR3 Ch 1 = ITC ChO=PWM ch2 =OC ch3=DIO
move.w #$0888,CFSR3-2 Ch 4,5 = DIO, CH6 = NO_ChOp

move.w #$8888,CFSR3-4 Ch 8,9,10,11 = DIO

move.w #$0088,CFSR3-6 ch 12,13 = DIO

******SET UP DIO CHANNELS (ENABLES and Inverter)*****

move.w #%0010101010000000,HSRR1

Set ENABLE A,B,C Low (disable)
Allow Chopping

67

move.w #%0011111111000000,CPR1 Set CH 3-6 & 8-13
move.w #%0000111111111111,CPRO to High Priority

* **********SET UP CHANNEL 1 (INPUT)**********
ori.w #$0004,HSQR1 Chl One Continuous mode (no links)
move.w #$0007,CH1PARO Chl Detect Either edge
move.w #$000E,CH1PAR1 No Start Link or Link chan cnt or BA

move.w #$0001,CH1PAR2 Max Count = 1 (count 1 transition)
move.w #$0004,HSRR1 Initialize ch] (ITC)
ori.w #$000C,CPR1 Set CH1 high priority
clr.w D0

WAIT] move.w HSRR1,DO WAIT TO MAKE SURE ITC INTTIALIZTION
andi.w #$OF,DO IS COMPLETE
bne WAITl

* **************end OfChl semp*****************

* ***********SET UP CHANNEL 0 (CHOPPER)********
move.w #$0091,CHOPARO SET CHANNEL ZERO

move.w #$1C,CHOPAR2 HIGH TIME (COUNTS)
move.w #$22,CHOPAR3 PERIOD (COUNTS)
move.w #$0002,HSRR1 Initialize ch0 (PWM)
ori.w #$OOO3,CPR1 Set ChO high priority
clr.w D0
WAITO move.w HSRR1,D0 WAIT TO MAKE SURE PWM INTTIALIZTION
andi.w #$OF,D0 IS COMPLETE ~
bne WAITO
III *****************end OfChO setup***********

*II!*********************************************************

* Turn Off all phases and wait about 6 sec.
************$**********************************************
lea STATZ,A0
move.w (A0),D0

move.w D0,HSRRO

clr.l D0
delay addi.l #1,D0
cmpi.l #$1FFFFF,D0
beq start
bra delay

68

************************II!********************************************

a********************************************************************

******* Set duty cycle to max safe voltage
start move.w #$15,CHOPAR2 High time (counts)

******* Initialize rotor position
clr.w DO
lea STAT3,A0
move.w (A0)+,DO

move.w D0,HSRRO

* *** * * * Wait for rotor to move to initial position
c1r.l D0
INIPOS addi.l #1 ,DO
cmpi.l #$FFFF,D0 delay time can be adjusted

bmi INIPOS
******* Start first commutation (state 5)
lea STAT5,A0
move.w (A0)+,D0

move.w D0,HSRRO

******* Get Initial Time (using Channel 1)

move.w #$0002,CIER ENABLE CHANNEL 1 INTERRUPTS
andi.w #$FOFF,sr ENABLE INTERRUPS
ori.w #$0500,sr "
move.w #$540,HSRR1 set all enables to trigger intrpt
* that gets initial time
getTIM move.b ’ FirsTIM,D0 Wait until Initail Time is stored

beq getTIM

69

******* Check for no current in Open phase (A) and set enable

NOtOFF btst.b #3,PORTE if commutation is done (no current in
bne NotOF F open phase) then
move.w #$40,HSRR1 set ENABLEA

*************************************************************

******* Initiate Closed Loop Control ***********
*************************************************************
move.w #$0002,CIER ENABLE CHANNEL 1 IN TERRUPTS
andi.w #$FOFF,sr ENABLE INTERRUPS
ori.w #$0500,sr "

III************************************************************

********e****************************************************

******** MAINPROGRAM *************

****************************#********************************

outloop bra outloop

III**********III***********#**************************************

****************************************************#*********

*************Interrupt Service Routine For Channel 1*********

org $10004

andi.w #$FFFD,CIER DISABLE CHANNEL 1 INTERRUPTS
move.w CISR,D6 .. Clear Channel 1 IRQ

andi.w #$FFFD,CISR

move.w #%0000101010000000,HSRR1 TURN OFF ENABLEA,B,C

* Stop Chopping if at max duty cycle
move.w CHOPAR3,D5

sub.w CHOPAR2,D5
cmpi.w #7,D5
bhi CHOP

move.w #%0001000000000000,HSRR1 Stop Chopping

CHOP

NotFrst

ROLLOVER

CHKsnd

NotScnd

move.b
bne
move.w
move.b
bra

move.w
move.w
move.w

cmp.w
bhi

sub.w
bra

clr.l
move.w

sub.w
add.w

move.b
bne
move.b
bra

move.w
lsr.w

7O

FirsTIM,D0
NotFrst

F INTRAN,LsTran
#$FF,FirsTIM
END2

FINTRAN,D3
LsTran,D4
D3,D7

SAVE COUNTER VALUES

D3 ,D4
ROLLOVER

D4,D3
CHKsnd

Calculate Offset time

D0
#$FFFF,DO
D4,DO
D0,D3

SndTIM,D0
NotScnd
#$FF,SndTIM
SETINT

D3,D2
#1 ,D3

STOR FOR USE IN OC ISR

******* Adjust for errors *****

clr.l
clr.l
move.w
mulu.w
mulu.l
lsr.l
lsr.l
sub.w
sub.w
clr.w

bpl
clr.w

D0
D1
D3,D1
D3,D1
#22,D1 (15 is theoretical min)
#8,Dl shift to divide
#8,D1 by 2"-16
D1 ,D3
#20,D3
D3 Use this line
at max duty cycle
only!
SETINT
D3

71

SETINT move.w F INTRAN,LsTran

* ***********SET UP CHANNEL2OUTPUT DELAY********
move.w #$OO83,CH2PARO SET CHANNEL TWO
move.w D3,CH2PAR1 OFFSET (WRITTEN BY CH1 ISR)
move.w #$0018,CH2PAR2 REFERENCE ADDRESS FROM CH1 INPUT TIME

ori.w #$OOOO,HSQR1 Schedule Interrupt
move.w #$0010,HSRR1 Init
ori.w #$OO30,CPR1 Set Ch2 high priority
clr.w D0
WAIT2 move.w HSRRI ,DO WAIT TO MAKE SURE OC INITIALIZTION
andi.w #$0030,D0 IS COMPLETE
bne WAIT2
* *****************End 0fCh2 Setup!!!********************
ori.w #$0004,CIER Enable Channel 2 Interrupts
rte

******************End ofChmnel1ISR*********************

*****************CHANNEL 2 ISR******************************

org $10204
andi.w #$FFFB,CIER DISABLE CHANNEL 2 INTERRUPTS

move.w CISR,D6 Clear Channel 2 IRQ
andi.w #$FFFB,CISR

cmpa.l #STAT5,A0
bhi NXTREV
bra COMM

NXTREV lea STATO,A0
move.b #00,STATE

COMM move.w (A0)+,D0
move.w D0,HSRRO

* *****Wait For Commutation to complete******** **
clr.l D0

AOFF
AOFF 1

BOFF
BOFF]

COFF
COFF]

END2

move.b
beq
cmpi.b
beq

bls
cmpi.b
beq

bls

bhi

move.b
btst.b
bne
subi.b
bne

move.w

bra
move.b
bne

subi.b
brie

move.w

bra
move.b
bne

subi.b
bne

move.w

bra

72

Determine which phase should be
Open

STATE,DO
COFF
#2,D0
AOFF
BOFF
#4,D0
BOFF
COFF
AOFF

#60,D5
#3,PORTE
AOFF

#1 ,DS

AOFF]
#$2040,HSRR1
END2

if commutation is done (no current in
Open phase) then

set ENABLEA Start Chopping

#60,D5
btst.b #2,PORTE
BOFF
#1 ,DS
BOFF]
#$2100,HSRR1
END2

set ENABLEB Start Chopping

#60,D5
btst.b #1,PORTE
COFF
#1 ,D5
COFF]
#$2400,HSRR1

END2 set ENABLEC Start Chopping

******** Adjust Chopping Frequency to account for speed **.***

clr.l
move.b
bne
cmpi.w
bhi

move.w

151.1
131.1

D0
ChOpMode,D0
HiMOde
#446,D2

Cont
CHOPAR2,DO
#8,D0

#7,DO

Check to see if rpm > 700

move.w #$1 1 ,DO
movel D0,CHOPAR2
move.b #1 ,ChOpMOde
HiMOde cmpi.w #446,D2
bls Cont
move.w CHOPAR2,D0
151.1 #8,D0
lsl.1 #8,D0
131.1 #1 ,D0
move.w #$22,D0
move] D0,CHOPAR2
move.b #0,ChopMode
Cont addi.b #1,STATE
ori.w #$0002,CIER
rte

if rrnp > 700 go to hi mode

Check to see if rpm > 700

if rpm < 700 go to low mode

Enable Channel 1 Interrupts

****************end OfChZ ISR*****************************

BIBLIOGRAPHY

[1]

[2]

[3]

[4]

74

G. C. Yomg and R Kiefer, “Electronically Commutated Motor for Driving a
Compressor,” United States Patent, Patent Number 5,491,978, February 20, 1996

Nobuyuki Matsui, “Sensorless PM Brushless DC Motor Drives,” IEEE
Transactions on Industrial Electronics, vol. 43, pp. 300-308, April 1996

K Iizuka, H. Uzuhashi, M. Kano, T. Endo and K. Mohri, “Microcomputer Control
for Sensorless Brushless Motor,” IEEE Transactions on Industry Applications,
vol. IA-21, pp. 595-601, June 1985.

Satoshi Ogasawara and Hirofurni Akagi, “An Approach to Position Sensorless
Drive for Brushless dc Motors,” IEEE Transactions on Industry Applications, vol.
27, pp. 928-933, September 1991

 

"Illllilllllllllilli