Sei sulla pagina 1di 0

Digital Signal Processors Usher in Green Motors

by Kedar Godbole
Digital Control Systems,
Texas Instruments Incorporated
Electric motors account for over half of U.S. electricity consumption. Very often the
motor is over-sized for the application, to compensate for variations in the applied load,
leading to the motor operating at partial load. Uncontrolled induction motors do not
operate efficiently under light loads, leading to unnecessarily high power consumption.
Smart motor control technologies can boost operating efficiencies into the 90% range but
require computationally intensive algorithms. By running the algorithms in a matter of
microseconds and integrating the right mix of peripherals, DSPs that have been optimized
for motor control are ushering in a new era of green motors. They will also make it
possible to use permanent magnet motors that are inherently more efficient.
Motor Control Basics
Applying a 3-phase voltage to the stator windings of a 3-phase motor produces sinusoidal
currents that are 120 out of phase. These currents produce magnetic vectors that rotate at
the so-called synchronous frequency -- either 1800 or 3600 rpm for most motors. In an ac
induction motor this moving magnetic field induces currents in the rotor windings and
they in turn produce their own magnetic field. The rotors magnetic field can also be
created by a permanent magnet.
Regardless of how the rotors magnetic field is produced the interaction of the stator and
rotor magnetic fields produces torque that rotates the rotor and produces a useful
mechanical output from the machine. Torque is highest when those fields are orthogonal.
The angle between them varies considerably when traditional scalar control methods are
employed. If there is a change in the load then the system response is rather slow.
Fig. 1: Scalar Control Delivers Good Performance Only At Nominal Speed
Power efficiency varies because at nominal speed the voltage drop across the stator is
negligible. But at low speed a higher voltage than the V/f curve would suggest is needed
to overcome resistive losses. This is also associated with a large proportion of the flux
generating iron losses. At low speed the output is small and therefore the efficiency is
very low, causing a large proportion of the energy input into the machine to be wasted.
Field-Oriented Control
Field-oriented control (FOC) or vector control is one of the most effective methods of
controlling ac motors. FOC operates by measuring and adjusting the stator currents to
keep the angle between the stator and rotor fluxes as close to 90 as possible. Unlike
scalar control techniques, where the three-phase supply is varied only in magnitude and
frequency, vector techniques control the magnetic vectors generated inside the motor.
Vector control offers the ability to control the load angle and by keeping the fluxes
orthogonal it becomes possible to control the torque much more directly.
Vector control also makes optimal use of the energy in the motor: The fluxes being
optimally aligned produces the most torque for the energy input. Vector control needs
knowledge of the rotor flux position and this can be computed by advanced algorithms
from knowledge of the terminal currents and voltages, as long as there are sufficient
computational resources available.
Sinusoidal stator currents and voltages are measured using a feedback loop that samples
at around 20,000 times per second. To simplify the dynamic model, current and voltage
are transformed into the rotor domain. (The rotor domain is a co-ordinate system rotating
in step with the rotor flux.) These values are used to calculate the voltages that, when
applied to the stator windings, will correct the load angle to as close to 90 as practical.
From a mathematical perspective the obvious place to start is with the known sinusoidal
currents that can be measured in the stator. These currents can be represented as a single
complex stator current vector.
The instantaneous currents in the stator phases can be represented as i
a
, i
b
and i
c
,
corresponding to the a, b, and c system axes. It can be transformed into a two-coordinate
system that does not vary with time by using two transforms. First the Clarke transform
inputs the three-coordinate (a,b,c) vector and outputs a two-coordinate time variant vector
(!, "). The formulae that effect this transformation are:
i i
a S
=

i
b
i
a
i
S
3
2
3
1
+ =

To simplify the dynamic model the stator current representation must still be transformed
into a co-ordinate system that is synchronous with the rotor flux. This is accomplished by
the Park transform. Here, the reference coordinates are d (flux) and q (torque) and the
reference frame is with the d axis aligned with the rotor flux position, !.
The two equations defining the flux and torque components of the stator current vector
are:


sin cos
i i i
S S Sd
+ =


cos sin
i i i
S S Sq
+ =
The most interesting aspect of these equations is that their values depend on the position
of the rotor flux (!). If the rotor flux position is known, or can be accurately estimated,
the two components become controllable levels, rather than varying sinusoidally as the
motor rotates. These are now fed in to a dynamic model of the induction motor. This
dynamic model estimates the rotor flux position based on knowledge of the stator
currents and voltages, in addition to the motor parameters and by controlling these
components it is possible to align the stator flux orthogonal with respect to the rotor flux.
This provides direct control over the torque.
In order to feed the results back into the stator, however, voltage values must be
calculated and applied to the stator phases. This is accomplished simply by using the
inverse Park transformation.
Fig. 2: Clarke And Park Transforms Are The Core Of A Vector Control System
A block diagram for a field-oriented control system for an ac induction motor is shown in
Fig. 2. All the blocks within the dashed line are inside the DSP and are implemented in
software. Current in only two stator phases has to be measured since the motor is star
connected and the third phase current is known because the sum of the three currents is
zero.
Rotor Flux Position
As described earlier, use of the Clarke and Park transforms bring the stator currents into
the rotor domain. Doing this allows the control system to determine the voltages that
should be supplied to the stator to maximize torque under dynamically-changing loads.
Both the Park and the inverse Park transforms need to know the rotor flux position. This
is accomplished in the flux estimation block, which executes an algorithm using the same
stator current measurements that served as inputs to the Clarke transform as well as the
corresponding stator voltages.
The speed regulator block controls the generated torque to match the load torque, without
which the system speed would tend to ramp if the motor is producing too much torque,
and slow down if it is producing too little torque (assuming a constant load torque.)
Estimating the correct motor speed and supplying that estimate to the overall calculation
helps control the speed to the desired value. Software-implemented PI (proportional and
integral) regulators regulate the torque producing and magnetizing components of the
(fictitious) stator current vector.
After the estimated feedback values are used by the inverse Park transform, the stator
current values -- still in the d (flux) and q (torque) coordinate system -- are converted
back to the a, b, c coordinates so each phase can be individually powered by the
appropriate feedback voltages. A PWM block -- typically built into the same DSP chip
(see Fig. 3) -- is used to generate the actual stimulation voltages.
Fig. 3: Typical Ac Induction Motor Control System Using Low-Cost DSP-Based
Microcontroller -- Sensor Can Be Removed For Sensorless Operation
FOC Advantages
The vector control scheme has inherently better torque characteristics, and significant
potential for better energy efficiency than scalar control.
Another advantage that is often not immediately obvious is the ability to size motors
according to the actual torque requirements of the application, rather than oversizing.
This is more important than it sounds: When a motor is properly sized it is operating at a
higher percentage of its rated load, and has better efficiency off the bat. So, the better
torque characteristics make it possible to choose a smaller motor that is more efficient,
and cheaper to buy as well.
Vector control also paves the way for more extensive use of permanent-magnet
synchronous motors (PMSMs). To run vector control for such a motor also needs
knowledge of the rotor position. Mechanical sensors on the rotor shaft can provide this
information but add cost. Also, sensors are often the weak link in a harsh environment. A
sensorless vector controlled system solves these problems. PMSMs are inherently more
efficient because they do not have the conduction losses associated with induced rotor
currents. They also offer better mechanical characteristics such as lower inertia, lower
torque ripple, and quieter operation, and are usually smaller for an equivalent mechanical
power output. Permanent magnet motors are also becoming cheaper as the manufacturing
technology improves and volumes increase.
Vector control of torque eliminates the poor speed and torque dynamics of scalar-
controlled motors. More precise control makes it possible to have simpler mechanical
components that are cheaper. That can lead to cost savings for consumers, particularly in
Europe, where appliance makers have realized that developing an FOC system would (1)
reduce noise and (2) allow apartment dwellers and owners of small homes to run the
appliance at night when electrical costs are lower.
Computational requirements place this control method beyond the reach of all but the
most powerful microcontrollers if all of the required algorithmic functions are executed
in software.
With the position information being almost mandatory for permanent magnet motors, a
microcontroller-based system could, however, take a different approach to determining
the rotor position by using either Hall Effect sensors, or position encoders. This approach
would add the cost of the sensors and would also suffer from the reliability issues
inherent in running wires from the sensors to the controller. If the motor is hermetically
sealed, as is the practice in home refrigerators, wiring to the controller would create still
another cost issue. A powerful computational engine can alleviate this difficulty by
estimating the position based on measurement of motor terminal variables.
In many countries, power factor correction has become a critical issue. Converting
domestic ac power into dc, so that all three phases can be balanced, involves rectifiers,
which produce spiky currents and a variety of harmonics. The processor of choice should
have the additional processing power to implement this algorithm.
Reduced Manufacturing Costs
An oversized motor is larger, requiring more materials to build and, as a result, is more
expensive to buy. Better control resulting in properly sized motors enables manufacturers
to buy lower-priced motors as well.
With better control it is much simpler to achieve various speed and torque profiles. Often
times mechanical components can be simplified, or eliminated, resulting in additional
cost savings to the manufacturer.
Also, an inefficient motor wastes a larger proportion of the power input. This means that
the heating effects of the wasted power and the associated reliability implications must be
dealt with.
DSP Selection And Software Development
Vector-controlled motors and the algorithms that implement them are sufficiently
specialized to require very specific characteristics in the DSPs that control them.
Just having the MIPS (millions of instructions per second) and single cycle MACs
(multiply and accumulate operations) mean far less without the proper complement of
peripherals.
In addition it is essential to examine the quality of MIPS offered by a particular
processor. Just offering single cycle MAC features, without the bus architecture to enable
sustained single cycle MACs, is less than useful.
The FOC algorithms, with the associated Clarke and Park transforms, for example, are
point-by-point algorithms that process one set of data samples at a time but must execute
each data set in 50 s. So a DSP architecture set up to take in numerous samples, working
on them as a group, is not an ideal choice either. What is needed is an architecture that
blends efficient mathematical processing, control operations, and the right peripheral
mix.
Since the control algorithms exist in software, storing them in Flash memory for rapid
retrieval is essential as is an abundant supply of on-chip Flash. Texas Instruments
TMS320F2812 and 2810 DSPs, for example, integrate 250 kbyte of Flash and, with Flash
Acceleration technology, program execution out of Flash ranges from 110 to 120 MIPS.
Another highly desirable DSP characteristic is the ability to handle microcontroller-like
instruction sets because several of the other algorithms found in association with the FOC
algorithm require powerful and efficient bit manipulation and effective use of branching.
On-chip PWM generators, fast event managers and a fast 12-bit ADC and CAN bus
support should also be integrated into the DSP. System integration has another benefit, of
course, which is a lower component cost than that which would be incurred if the
peripherals were added in the form of standard parts or even ASICs.
Software is a critical aspect of any DSP development and an extensive library of motor
control applications speeds development time and eases the process. Texas Instruments,
for example, has assembled a powerful software suite to support FOC control.
Conclusion
In a world where not just environmental but geopolitical issues are pushing efficient
energy use towards the top of the agenda of many nations, electric motors that consume a
large portion of electricity produced are a prime candidate for energy savings. Motors in
industrial applications can run much more efficiently.
Motors in clothes washers and dryers, dishwashers, and other white goods that have
dynamically-changing loads tend to be very inefficient. This is because they typically use
microcontrollers in their control loops that cannot handle the computational intensity
required for precise control.
In some countries, energy-efficiency issues are almost matched in importance by lifestyle
issues. In Europe and Asia, to cite two examples, small homes and apartments typically
mean clothes washers are in close proximity to bedrooms and living rooms.
Noise generated by oversized and poorly controlled motors is annoying. As a result white
goods appliance operation is frequently scheduled at the local electric utilitys peak load
times such as the afternoon.
Quieter motors that become an option with advanced motor control are making quiet
appliances possible, so that the appliances can be used overnight without disturbing the
owners sleep. This practice has the additional advantage of off-peak operation for the
electric utility and reduces the total generating capacity requirement. In some instances
utilities also offer reduced pricing to encourage off-peak use.
DSP technology offers a direct path to reduced costs because it delivers the required
computational power for accurate flux or position estimation without requiring sensors
that increase overall system cost.
A hardware/software combination of DSPs can be optimized to run specific algorithms
required by field-oriented control. Success is critically dependent, however, on the DSP
architecture -- processing one set of values at a time -- and the right peripherals. In
particular motor-control-optimized DSPs must be capable of handling microcontroller-
like instructions. A library of motor-control algorithms such as those offered by Texas
Instruments also speeds the design cycle.
For more information about Texas Instruments TMS320C2000 Control Optimized DSPs
follow the link http://www.ti.com/smartmc
About The Author
Kedar Godbole joined Texas Instruments in 1998 after he earned his MS degree in
Embedded Systems and Control from the New Jersey Institute of Technology. Kedar's
focus is on high-performance control and using DSP-based techniques to control flexible
structures and he is currently with the Digital Control Systems group at TI. His
responsibilities include driving the Digital Motor Control Software Program and
enhanced software reusability. He was awarded his BE in Electronics (1995) at The
University of Pune, India

Potrebbero piacerti anche