This is the unedited transcript for webinar:Advanced Motor Control Technologies.
To view this presentation on demand, click here.
Aimee Kalnoskas:
Hello and thank you everyone for attending part two of our two part webinar, Advanced Motor Control Technology: Meeting high efficiency, compact size, and low cost for high volume application. Brought to you by Design World Magazine and Drive Tech Inc., and sponsored by On Semi Conductor.
Today’s second segment of our series will move into additional advanced brushless motor control technology. If you missed part one, you can view it from an email link you received for part two. We’d like to thank our presenter Dr. Dal Ohm for preparing and presenting today’s webinar. I’m Aimee Kalnoskas, an editor with Design World and EE World online, and I’ll be your moderator today.
Now, let’s get to know our presenter if you missed it the first time. Dr. Dal Ohm has spent much of his industrial and academic career in applied R&D and product development of diverse AC and DC motor drives, motion control, grid connected inverters, and power electronics. His research area also includes system analysis, modelling simulation of dry systems.
He has successfully applied many advanced concepts and technologies into products for performance improvement and cost reduction. He is now president and principal engineer of Drivetech Inc.. Prior to his current position, he was with Cole Morgan Motion Technologies Group as technical director and program manager. He was also with Valdor Electric Company, Electrocraft Corporation, and LG Industrial Systems.
As an adjunct professor, he taught engineering courses at San Jose University and NPU. He received his PhD and MS degrees in electrical engineering from Texas A&M University. Dr. Ohm has led many motor drive short courses at major technical conferences and is currently a faculty member of FMMA Motor and Motion college courses. He is the inventor of several patents on motor control methods and grid connected renewable energy inverters, and the author of more than fifty technical articles, papers, and conference proceedings.
He is affiliated with IEEE, FMMA, and FAE. If you have questions for Dr. Ohm you can reach him at an email provided at the end of both parts of this presentation. And now, let’s get started. Dal, it’s all yours.
Dr. Dal Y. Ohm:
Now, as I mentioned before in AC motors people try to do the field oriented control. When using this field oriented control, most of the modern digital drives are using synchronously rotating reference frame. In this control, AC motor current is divided in two components. One is controlling torque producing current such as just like amateur current in DC motors. The other part of the AC motor current is field flux current component just like permanent magnet motor under water.
So for permanent magnet motors, this field flux current is zero normally because all this field flux is produced by the permanent magnet D cell. But for interior permanent magnet motors, this field flux current may be non-zero to control maximum total control or field weakening. This field flux current is equivalent to the magnetization current in induction motor.
So we want to control those two current components independently, and once then then the operation is very similar to separately excited DC motors. I’m going to go more in detail for this. Now there are some simple math that we have to understand. You don’t have to it fully, but try to understand conceptually.
This is three phase to two phase transform. And so I mentioned before, all poly-phase sinusoidal inputs will produce rotating flux. That means whether it is three phase or two phase, it doesn’t matter. It produces north and south pole which is sinusoidally distributed. So this equation says from A, B, C current, it produces equivalent alpha and beta phase current. So in A, B, C, frame, three phase system, A and B and C are now controlled in this frame. Alpha phase which is same as A frame and then beta frame is ninety degree behind it. In this two frame, two phase transformation, this is equivalent to two phase machine, and we want to translate or frame transform this A, B, C quantities in alpha, beta quantities first.
And then another frame transform is about the coordinate attached to the rotator, or rotating coordinate. In other words, X and Y axis in stationary coordinate is very fixed in space. But D axis and Q axis in this is attached to the rotator. So when rotator is rotating, or C theta is rotating, those coordinates are also rotating. In this frame transform from X, Y to D and Q axis transform, the conversion equations are a function of sinusoidal and inverse transform is possible.
By using this rotational coordinate two phase system, we can rewrite the dynamic equations of Brushless motors. In voltage model, in synchronous frame, we have Q axis and D axis voltage is a function of Q axis and D axis current. And there’s a back EMF term is here, and then those two are cross-coupling term. As you notice, this equation is very much similar to DC motor equation and this dynamic model of Brushless models can be extended to induction models as well.
Once we understand this two phase rotating coordinate dynamic model, we can have a synchronous current regulator mean that rather than A, B, C current is directly measured and compared with A, B, C current command and through a three PI regulator. First, out of this three phase current, we convert Clarke and Park transform so that now current is in D axis and Q axis coordinates. And we close the two PI regulator based on this D axis current command which is usually zero, and Q axis current command. And then on the output we inverse transform, TQ axis to the alpha beta, and also the inverse Clarke to a three phase, and then Parks with modulation is used to produce three phase voltages.
So without this Park and inverse Park transform, we are having simply PI regulator for two axis, both D and Q axis. And then after this Park and Clarke transform and inverse transform, this word is a stationary word, A, B, C phases, which is related to the stationary motor and drive. Now the insider control, we are considering in rotating two axis coordinate, and use two PI regulators to control. This is called synchronous current regulator and used in most of the [inaudible 00:10:43] controlled, high performance model control.
One of the advantage of this synchronous regulator is that zero status state phase error with PI control. So in traditional phase current regulator with the three PI loops, as the speed goes up high and the excitation or current frequency goes high, and then there’s a limit that because of this phase delay coming from the rotating speed. But with this synchronous frame controller, that phase delay is completely removed because inherently this synchronous current regulator has internal model dynamics in it and then there’s no phase delay. This is one big advantage of using a synchronized current regulator.
In the model, as you saw it in the previous dynamic model, there are some cross coupling compensation. Let’s take a look at those. Yeah this one. Those terms, these omega LDID and then omega LQIQ are the cross coupling terms because D axis variable is effecting Q axis, and Q axis variable is effecting D axis. This omega ramda N is the back EMF term. By adding this disturbance as a feed fort, then the proponents of the drive are getting better with very quick integrative response. So the overall with disturbance before the technique will improve the system performance.
Now speaking of this control performance, the tuning is one of the important steps in designing and producing the model control. A lot of times people use this conventional, experimental tuning, but once we know the dynamic model of the motor, then we can do some analytic tuning. Here’s the two kinds of analytic tuning that I’m going to introduce. One is called cancellation tuning for the desired bandwidth of omega V. So this is the basic model equation. Inertia, friction of time constant, and torque constant of the motor. An here is PI regulator. So we want to have a speed command and actually a speed compared and go through a PI regulator. Those closed loop dynamic equation or transfer function and then setting that closed loop transfer function poles to be omega V which is the desired bandwidth, then we will end up with this proportional gain to be this and integral gain to be this.
On the other hand, pole placement tuning is another type of tuning. This one uses, not try to cancel or pose in zeros, but try to make this second or assisting poles to be in critically damper the poles. So in this case the desired KP and omega I is calculated by these two equations. So it is up to the application that whether we use cancellation tuning or pole place tuning, but regardless of which one to choose, you have to do some fine tuning to make the system perform well for application. One other thing is, with this pole placement tuning, you may get a higher bandwidth. The problem is during the robustness. On other words, of the models for control parameters are changing such as model resistance or load, pole placement tuning tends to have some unstable condition may occur during worst case parameters. From that sense, the cancellation tuning of the previous slide is more robust compared to this pole placement tuning. But it is entirely up to the application to choose which one.
Now, the discussion on previous two slides, the idea system with all SI units but in a lot of implementation of model control with micro controllers, those units, kind of a [inaudible 00:17:24] unit scale is used. In other words, just like when you are measuring the current then you have to use A to D converter to use the current. The maximum value representable from the A to D conveniently it is considered one, and all the current that you are measuring is pressure and number. So is the speed. Speed, it may not be measured through the A to D converter, especially if it is from like an encoder, but you also set a maximum value which is bigger than the practical maximum operational speed of the motor to be one, and all this measured speed would be fractional.
So in this [inaudible 00:18:27] unit system, the gain has to be changed because in this diagram assuming that everything is in SI unit, speed in terms of radian per sec and current in terms of amp, then your analysis shows KP as ten and omega I is some numbers. But now we are dealing with a PI regulator which is different unit. So in this conversion, let’s say omega max is the maximum speed that is considered unity, and omega max prime is simply one. And then here I max is the maximum current that makes unity per unit system one, and then I max prime is here one. And then the KP prime and omega I prime is the tuning game that is used inside the PI regulator which is [inaudible 00:19:55] implemented.
From the outside point of view, whether the system is in SI units or the system is in [inaudible 00:20:09], those two block diagrams should be identical. By equating this block diagram equation and those equations identical, we can compute easily the kP prime and omega I prime based on this [inaudible 00:20:32] unit system value. So this conversion has to be done to use it inside the software control to make the system performs as in the SI unit.
As I mentioned, for high performance systems we have to close the current loop, and closing the current loop is also, we have to use the analytical tuning. This is the electrical equation of the model. Mainly, this is L of R dynamics. We are also adding a PI regulator. So this closed loop control, in this case cancellation tuning gains are this and pole placement placed tuning gains are this. Very similar to the [inaudible 00:21:45] control. Similar equations can be used for gain translation for [inaudible 00:21:54] unit systems.
So by using this analytic calculation, one thing you know is, you know this desired bandwidth. If you do not know some of the parameters and you find the bandwidth, then you can also calculate the unknown parameters of the model. Typically it’s not in the current loop. These are mostly in the [inaudible 00:22:37] loop. A lot of times the inertia value is cannot be measured or unknown. In that case, we can manipulate with our assigned closed loop bandwidth and then compare with the actual response. And then we can back calculate the unknown inertia.
Now in this, one of the objects here to make it low cost over a system for high volume applications, sensors. Feedback sensors are an important part of the overall cost. So using the low cost sensors or using the sensorless control is one of the objectives and very attractive for control point of view. Three hall sensors in three phase control is commonly used, lowest cost sensors. And with linear interpolation, we can use three whole sensors but in the end we get continuously moving angles. Recently there are magnetic, rotary, hall position sensors. An example of manufacturers would be ANS from Austria and Avago in US. These proved demand on hall position sensors will end up with very high resolution feedback signal and commonly used in many local applications as well. Of course it is more expensive than hall sensors and also more expensive than sensorless control.
Well sensorless control is very attractive, and even though you are not using sensorless sensor, you have to estimate the flux angle and velocity to make a good high performance model drive. Popular model technologies include back EMF detection. This is one big advantage if six step models because in six step control, one of the phases are not energized at sort an angle, and during that time we can measure the back EMF voltage and then that is used for to get the rotor position. If it is not measurable, such as in the sixth step in sinusoidal operation, we have to use the model based observer. Observer is another thing named for the estimator and it tries to estimate the rotor position based on model current and voltages. We’ll discuss about this in the next slide.
Another popular technology is called carrier injection method. It uses some kind of a five hundred to several kilohertz, signal is injected on top of the control signal and then we detect the, based on this carrier signal injection, we want to extract the position information. We’ll go with a little bit in detail shortly. So as for this local sensors, or sensorless control, the performance should be limited accuracy and the accuracy is bandwidth limited. In terms of real high resolution sensors, this limitation is almost infinity by selecting a very high cost, a high resolution feedback. But with this sensorless control or limited feedback devices, we are sacrificing some accuracy and bandwidth. Especially with the sensorless control, this model based observer, especially this, is estimating the model back EMF but during the initial start up or near zero speed, there is no back EMF so it’s a challenge. To make the precise model-based observer, we also have to consider some of the non-linearities in the model drive system. Contributing extra-voltage drops, such as deck time and semi-conductor drop.
So in this sensorless control, the selection of the algorithm and tuning is the key to success. This diagram explains some of the Carrier Injection Method. The white block inside the dotted box is the normal model drive functions assuming that the angle, feedback angle, is normal. So on top of this motor control block, by using current regulations and synchronous current regulation, we are injecting a carrier voltage about five hundred hertz to two kilohertz injected. And then, when the current is measured, this current is first low pass filtered, and then go back to the model control or PI regulator. We also have to have a band press filters to filter out the signal of this carrier injection frequency.
And then we do the heterodyning and calculate the aero-angle and then Luenberger Observer is used. And the end result is we are estimating the rotor position. That estimated rotor position is going back to the model control. In this Carrier Injection Method, operation looks good but carrier injection, voltage injection from five hundred to two kilohertz is within audible range and you can hear the carrier signals noise from the drive. This is one of the big disadvantages of this Carrier Injection Method.
In sensorless control … oh this is not the right, yeah. This is the observer, block diagram of the observer. Observer first internally we have a motor model. There’s a real motor. If voltage is injected into the real motor, then current comes out. This is a motor model when the same voltage is injected, then this will compute to the dynamic motor model and state of current comes out. We are comparing this model current and the real current, and then compared it and applied a regulator, and then his regulator output is put in as a feedback.
One of the objectives that we want to find out in this observer is we want to find out the back EMF, which is not measurable in semi-[inaudible 00:32:29] motor drives. So when there’s a difference in back EMF from motor model and the real motors, than obviously the current will be different and this difference will be fed back to compensate for the difference. So once this observer is operating well, then this back EMF from this observer and back EMF of the real motor should be very close. That’s the objective.
So once we have this model back EMF … by the way, in this block diagram it is not just a one loop but it’s a factor of two loops. The Q axis and D axis, or more precisely alpha and beta axis, which is in stationary frame. So once we estimated the back EMF of the model, then we [inaudible 00:33:42] it and calculate the arc tangent, and we are estimating the rotor position, beta S.
It is applied to two phase stationary quantities, this current and voltage. We want to estimate the back EMF on alpha and beta frames. There are two different strategies in regulation, or observer commutation. One is the Luenberger observer which is a linear regulator. Say like A and K which is the slope so that observer bandwidth is much much greater than ten times the modern model. Another method id called the Sliding Mode Control. The idea of Sliding Mode Control is like a bang bang control, is just change the sign of the output and without any slope. So in this sliding mode control the response could be faster, but the alpha is chattering and we have to have a very heavy [inaudible 00:35:22]. A lot of times in practice, we combine those Luenberger observer and Sliding Model observer. Combine those two and then you use them so that during this linear region at small errors, we use Luenberger observer, and when errors peak we use sliding mode control. When you are calculating arc tangent, often people use face, lock, [inaudible 00:35:58] as they’re directly calculating arc tangent.
So when motor is up to a certain speed, the previous observer base sensor this control, whats where. But the problem with it is start up strategies. First thing, most popular, the start-u[ strategy is Open-Looped Forced Start. Just like step motors, we are injecting a fixed amount of current with the ramping speed command. In this case we have to consider the dynamics of the motor itself so that when fixed amount of current is injected and KT is the co-constant of the motor, this is inertia and this is low torque.
When we are doing the full stop, then this acceleration rate or the ramping rate of the speed and this amount of current has to be in such a way that … start of current and acceleration rate should be in such a way that this starting current is slightly higher than worst case inertia with low torque so that when it is, startup can be made okay regardless of inertia and low torque. If it is too high, than the current is, the startup often fails, if these conditions are not met. Meaning they deviate too much from the ideal case. So based on these mechanical dynamics of the motor, you have to select the amount of current control, current command, and acceleration rate.
If the motor has big inductance difference, such as IPM motors, than we can utilize this inductance difference to detect the angle. Even though it’s not precise, we can roughly know the angle and that’ll help initially how to start. [inaudible 00:39:35] occasions of the motor, in [inaudible 00:39:39] V minus IR drop will be L times the increase of the current. So by measuring the rise time after voltage induction, we can estimate the inductance. And then if you know the inductance at which angle the inductance is high, at which angle the inductance is low, we kind of sort of know the angle of the rotor and then we startup based on this angle.
Another similar way is use the carrier injection which also requires higher inductance difference and then carrier injection method can be used. So in some high volume models, the carrier injection startup is used. It may just produce some audible noise but only during startup, and after that it’ll move to the observer based sensorless control.
Lastly, the power factor correction as I mentioned before is another concern from the regular [inaudible 00:41:14] issues. If the input AC voltage is rectified with [inaudible 00:41:23] as you can see the current wave form is very peaky. Power factor is defined as a average power output divided by RMS voltage and current. So this power factor can be divided into two sections. One is cosine V which is conventional steady state power factor which we call it displacement power factor and is a cosine V. The difference between current phase and voltage phase.
Another of this part is called distortion factor so that IS is the magnitude of the R current and IS1 is the magnitude of the fundamental current, and the difference between those two or the ratio of these two is called distortion factor. So mainly because it is due to the harmonics. So to improve the power factor we have to have the displacement factor is close to one, also the distortion factor has to be close to one. Sorry about that. I’m not used to this handling this. Okay. And then IEEE 519 which talks about this power factor correction, it recommends to less than five percent. This is known mostly for very high power motors. But in a low power high volume application, we may have to consider power factor correction soon. For this power factor correction, instead of using bridge rectifier, people use this dedicated power factor correction converter along with necessarily gate drives and power [inaudible 00:44:09] can be used. Recently several micro-controllers are developed which can do this single phase power factor correction more nicely with a processor so that that processor can be used for a power factor correction as well as motor control.
In conclusion, this high efficiency motor reduces the size and total operating cost. So we want to move this permanent magnet brushless motors. With this permanent magnet motors, sensorless or local speed where control reduces the cost. We have to have a proper control algorithm and also optimally tuned drive which result in precise commutation and improve efficiency and produces maximum torque while achieving good dynamics. So if we implemented this, selection of micro-controller, gate drops and power switches for low cost and compact design is necessary.
I conclude my presentation for now. Thank you very much for attention.
Aimee Kalnoskas:
And thank you Dal for an especially detailed that’s relevant and technically interesting in depth review of advanced motor control technology. And while it was quite thorough, [inaudible 00:46:04] content in your presentation and perhaps have some questions for you. You can send questions to Dr. Ohm at ohm@drivetechinc.com. That’s ohm@drivetechinc.com. And thank you very much for coming to our EE World online seminar.
Dr. Dal Y. Ohm:
Thank you.
Aimee Kalnoskas:
Have a great day. Thank you Dal. Great presentation. Lots of terrific information in both parts of this series. And for our audience remember if you did miss part one you can view it from the email. There will be a link in there that you received for today’s webinar. Thanks for joining us!
Leave a Reply
You must be logged in to post a comment.