Field Weakening

General topics and discussions about the VESC and its development.
ColinK
Posts: 1
Joined: 11 Apr 2017, 00:39
Location: Austin, TX USA

Re: Field Weakening

Postby ColinK » 14 Apr 2017, 15:18

“However, in general however, field weakening should be avoided if possible. A motor will run significantly more efficient by increasing the bus voltage to produce torque at high speed compared to doing field weakening. If the bus voltage cannot be increased a motor with higher kv should be considered if possible. Field weakening should be the last thing to use. I might implement it in my ebike though since I already have the motor and don't feel like rewinding it (but my ebike is actually fast enough).”

I expect Benjamin is referring to RC type motors (high flux, low inductance, no saliency). These motors don’t tend to be very amenable to field weakening operation (FW). But then consider that Toyota engineers its hybrids/EV’s to be in FW state for most of their speed range. As does the rest of the PM traction motor industry. There are several advantages to doing so and I expect personal EV’s like ebikes and eskates will benefit soon. It’s a new frontier for us. Vesc developments may help usher in that era and I’m pleased Benjamin is working on it.

*FW is not merely a trick to compensate for mismatched winding, gearing, and voltage, but rather a feature to craft a design around*

Let’s look at the example of an RC plane. The propeller speed is independent of the aircraft speed and the propeller has low inertia. It can quickly be accelerated to the peak power point of the motor and the plane will accelerate (or climb) quickly while the prop remains at this speed of maximum power output. In that sense, the plane has maximum power available over different airspeeds.

On a PEV, we’d like maximum power to be available over a wide speed range too; not just at a single speed (thus having better utilization of motor, battery, and controller). A system designed for FW will have a wide constant power speed range (CPSR), a typical goal for traction motors. It’s the alternative to having multispeed gearboxes that are needed to maximally exploit motors that peak and fade quickly.

Have a look at this photo from a Parker brochure for its IPM motors, as might be used in a Brammo racing motorcycle. Notice the huge high efficiency region:
http://i.imgur.com/VCuvpQk.png

What are the characteristics of a motor the make it well suited to FW? It’s a balancing act between saliency ratio and normalized flux linkage. For RC type motors that lack saliency, a normalized flux linkage of .707 results in a design with wide CPSR. My understanding is that most RC motors exceed this, but it’s possible to hit this mark with external inductors. I encourage you to start with this paper to learn more:
http://moscow.sci-hub.io/38c192253ac721 ... ng1994.pdf

Another advantage of saliency is that it enables certain zero-velocity motor position estimation methods, which should help with sensorless high-torque starts. My hope is that Vesc motor detection (BLDC tool) will be able to estimate both Ld and Lq (and also be able to map them into a lookup table to compensate for iron saturation).
When a motor has a salient rotor, the ideal current vector will have D-axis current prior to entering FW. The vector is determined by max torque per amp (MTPA) protocol.

Benjamin, in regards to your thoughts on how to go about FW control, I encourage you to take a look at page 41(30) of the following. Several methods are listed. It also goes thru MTPA on page 36(25) if you aren’t familiar and wish to learn.
http://epublications.marquette.edu/cgi/ ... heses_open

It seems your approach to FW is good for motors that have finite FW capability. Motors that lie to the left of the optimal IPM line (Soong paper) will eventually won’t accept rated current, regardless of angle. A third control mode will be necessary in that case.

Disclaimer- my engineering expertise isn’t in motor design, control, or power electronics. But I’ve been fascinated and doing a lot of reading.

A scenario that concerns me is what would happen if power got interrupted during FW operation (BMS stopped supply, for instance). Would the back-EMF fry the controller without Id to suppress it? My knowledge of power electronics is still too thin. Passenger car traction motors are driven by IGBTs. We hobbyists use lower voltages, which allow for Mosfets. Would mosfets be more at risk? “pf26” may have been referring to a similar situation.

Irreversible demagnetization is a risk with FW. It’d be great for a controller to have some protections to stay within the safe operating envelope. Hot magnets are more susceptible. The controller may be able to estimate magnet temperature from flux linkage estimate. Magnet flux decreases with temperature for NbFeB but not so much for SmCo (but SmCo is less susceptible to demagnetization so might not need to be monitored).

Keep up your amazing work!

elux
Posts: 10
Joined: 15 Apr 2017, 00:33
Location: Detroit

Re: Field Weakening

Postby elux » 15 Apr 2017, 00:55

I have been watching the developments of this project for almost a year. This is the first post I've seen that demonstrates an understanding of field weakening. I work in the electric vehicle industry and I tinker with projects like electric skateboards in my spare time. I have a VESC powered board and I've built a small dyno for testing motors. I am waiting for the VESC 6 and more importantly, the new firmware and VESC-tools so I can start contributing to this project. An efficient means for control in the field weakening region, failure mode handling, induction machine control, and wound field synchronous machine control are all possible features that I could help contribute toward.

Benjamin, please respond to rew in the VESC 6 thread. We'd like to have an update on your progress. Something like, "the beta testers uncovered several issues and they are now testing a patch that addresses some of the most critical issues". There is so much potential in this project, but we can't make progress without an initial release. Trampa has already started selling the VESC 6 hardware, but there still isn't any indication when the new firmware for it is going to be released.

Also, when you update your new VESC website, make sure you include a clear means of donating. Some of us remember what is was like financially when we were getting our PhD.

benjamin
Site Admin
Posts: 280
Joined: 15 Dec 2015, 08:38
Location: Sweden
Contact:

Re: Field Weakening

Postby benjamin » 18 Apr 2017, 09:03

Thanks for the feedback, I'm glad to hear that there is so much interest in getting involved in the project.

I just wrote a long reply in the VESC6 thread, and there are also some comments about the situation with the source code right now at the reply. After finishing the test bench and updates I mentioned and the shipping of VESC6 has started, I will shift my focus towards documenting the source code on the website and writing contribution guidelines so that everyone who is interested can get involved.

Regarding MTPA of salient motors, that is also on my todo list and I have some papers with a bit of information about it. I have one paper that describes a formula for calculating the D axis current from the Q axis current for MTPA from the difference between ld and lq, and that will probably be the first version of MTPA on the VESC. I also found some information on including saliency in the observer that I might give a try. Measuring ld and lq fully automatically is not easy with the approach I'm using now as it only uses 6 distinct vectors, but what could be done is measuring inductance in only one vector and turning the motor slowly by hand while doing that. The max and min inductance for the revolution would represent ld and lq. A problem with the inductance measurement is that the absolute value is not too accurate (maybe 10 to 20 percent), and the difference used for MTPA seems to depend on the absolute value. A different approach for getting that difference right and the best MTPA would be to run a motor with a propeller using speed control while adjusting the value. Then the value for when the magnitude if id and iq is at minimum for a given speed is the best value to use for MTPA.

Regarding using saliency for low-speed operation under high torque, that is also on my todo list. Signal injection can be used for that, and only the relative inductance difference matters meaning that the absolute accuracy in not an issue. A problem with using this approach is that the estimation can be 180 degrees off if the initial position is not known since the estimation goes two revolutions per electrical revolution as there is no difference between a magnet north pole and south pole, but there are ways of dealing with that. One example is to solve the initial ambiguity by measuring the rate of saturation, which is affected by the magnet poles.

Regarding what I was referring to with field weakening is simply increasing the maximum speed of hobby PMSM motors by pushing high D axis current for cancelling out the rotor flux, and that is a lot more inefficient than increasing the bus voltage for increasing the speed as far as I can tell because 1. the extra RMS current cause by the D axis current will contribute to heat and 2. the lower rotor flux seen by the stator will result in less torque per amp, requiring even more current for generating the same torque. It has nothing to do with adjusting the D axis current for MTPA or for induction motor operation. Maybe there are some other effects that I'm not considering for high speed operation, but I'm quite sure that what I'm thinking is valid for hobby motors. If that is not the case I will give field weakening higher priority and work on it sooner.

I have to go now, but I will write a bit about my plans for funding the project later.

elux
Posts: 10
Joined: 15 Apr 2017, 00:33
Location: Detroit

Re: Field Weakening

Postby elux » 19 Apr 2017, 22:06

Benjamin, thank you so much for your reply, both here and in the VESC 6 thread. That was what I was hoping for. I especially appreciate this quote, "... I will shift my focus towards documenting the source code on the website and writing contribution guidelines so that everyone who is interested can get involved." Documentation is never the fun part of any project.

Regarding characterization of motors with saliency, I think this is the understatement of the year, "Measuring ld and lq fully automatically is not easy...". The D-axis inductance, the Q-axis inductance and the permanent magnet flux linkage are all non-linear function of the stator current (id and iq) and they vary with rotor position. Having said that, your suggestion of using fixed DC current in one vector and rotating the rotor slowly ("...what could be done is measuring inductance in only one vector and turning the motor slowly by hand while doing that. The max and min inductance for the revolution would represent ld and lq.") will probably produce adequate results for stable control if you use multiple current levels.

I think the idea of using a propeller might only work on a limited class of motors with narrow constant power speed range.

Hopefully, this feature will take a back seat for now since it's complex problem for a class of motors that are not readily available in the retail market. People like me love this idea and I think it will hold great promise for the future, but I think you have your priorities right with the direction you are going for near term.

RSR
Posts: 10
Joined: 08 Mar 2016, 16:58

Re: Field Weakening

Postby RSR » 23 Apr 2017, 19:57

Fault handling in field weakening: Motors with high ratio of inductance to flux linkage will, when short circuited at moderate or high RPM, enter a state of natural field weakening without exceeding max current.

This means that safe fault handling at high RPM can be achieved by short circuiting the winding (by low side or high side null vector) -- i.e. this will lead to a low braking power (P_brake = M_brake * w_mech approx. equal to I^2*R, where I = Lambda/Ld). M_brake will increase when the motor slows down while P_brake remains almost constant until reaching rather low speed.

elux
Posts: 10
Joined: 15 Apr 2017, 00:33
Location: Detroit

Re: Field Weakening

Postby elux » 23 Apr 2017, 20:19

RSR is absolutely right about using three phase short as a default condition during a high speed failure with motors that have lots of inductance relative to the permanent magnet flux linkage. As speed goes to infinity the short circuit current asymptotes to I_ss = Lamda_m/Ld where Lamda_m is the permanent magnet flux linkage. In practice, the short circuit current approaches this value within 1% at about base speed of the machine so you don't need to be spinning very fast.

Unfortunately, the short circuit current for most of the outrunning surface mount PM motors used for hobby applications is well above the max current for the motor so it's not a good idea to use three phase short as a response to a fault condition unless the VESC is connected to a high saliency motor.

RSR
Posts: 10
Joined: 08 Mar 2016, 16:58

Re: Field Weakening

Postby RSR » 23 Apr 2017, 22:58

I think RC outrunners often have Lambda/Ld = 2 to 3 times higher than their max continuous current rating. This brings the question; can/should the fail safe state be configurable based on measured flux/inductance? or automatic, e.g. shorting until reaching a max current?

Field weakening can, if done properly, bring high power to a wider speed range -- without additional cost of hardware, safety issues or noise. The main benefit of field weakening is AFAIK reduced cost of inverter, cables and connectors -- but not for all applications; it does not help if one only needs high power at a narrow speed range, nor does it have purpose when cost is of little concern.

One do not necessarily need a motor with high saliency to use field weakening -- just proper ratio of inductance to flux linkage.

BTW of using saliency to detect rotor angle; it may improve performance at combination of low speed and high torque, but note that saliency will change with current (due to magnetic saturation) -- this can decrease accuracy of calculated angle when at high torque, unless one maps the inductance values also at high current.

madcowswe
Posts: 14
Joined: 08 Jan 2016, 15:37
Location: San Jose
Contact:

Re: Field Weakening

Postby madcowswe » 07 Sep 2017, 20:58

Just to reinforce what people have been saying about efficiency: I think operating in FW makes sense even for hobby brushless motors.

I did some calculations on the Turnigy SK3 4250-350kv.
In the below plot I have fixed the bus voltage at 22V, and the current to 64A (and hence fixed copper losses), and I plot the torque and power. In this model Ld=Lq, so the MTPA is very simply along the q axis. Note that I don't think Ld=Lq is strictly ture for the SK3's, but that is what was used in this model. Also I should note that magnetic saturation was not modelled.

Nevertheless, you can see that the highest mechanical power coming out of the motor is above base speed (so in FW region), and since this is for a fixed copper loss, it implies this is highest efficiency. Of course we are neglecting iron losses, which would indeed skew this a bit, but the copper losses are usually dominant for these motors, and so I think this is true in general.

I can redo these calculations including an iron loss model if you guys are interested?

Cheers, Oskar

Image

markorman
Posts: 14
Joined: 25 May 2016, 11:35

Re: Field Weakening

Postby markorman » 08 Sep 2017, 06:18

Did you take into account how much Id current you need to achieve higher speeds? Did you also account in this current when calculating efficiency?

ThierryGTLTS
Posts: 80
Joined: 09 Aug 2017, 11:10

Re: Field Weakening

Postby ThierryGTLTS » 08 Sep 2017, 09:05

madcowswe wrote:Just to reinforce what people have been saying about efficiency: I think operating in FW makes sense even for hobby brushless motors.

I did some calculations on the Turnigy SK3 4250-350kv.
In the below plot I have fixed the bus voltage at 22V, and the current to 64A (and hence fixed copper losses), and I plot the torque and power. In this model Ld=Lq, so the MTPA is very simply along the q axis. Note that I don't think Ld=Lq is strictly ture for the SK3's, but that is what was used in this model. Also I should note that magnetic saturation was not modelled.

Nevertheless, you can see that the highest mechanical power coming out of the motor is above base speed (so in FW region), and since this is for a fixed copper loss, it implies this is highest efficiency. Of course we are neglecting iron losses, which would indeed skew this a bit, but the copper losses are usually dominant for these motors, and so I think this is true in general.

I can redo these calculations including an iron loss model if you guys are interested?

Cheers, Oskar



We are very interrested for several reasons.

But you certainly know that guys of Bimoz (CH) claim they have a motor that can deliver 50Nm @ only 607-70 RPM for an e-bike.

Of course the magnet count is high an the ERPM higher, but I'm curious to know if they use field weakening in their design.

Have a Nice Day.

Thierry


Return to “General”

Who is online

Users browsing this forum: No registered users and 1 guest