Sensorless, halls, encoders, motor efficiency?

[bowlofsalad]

Hello,

Has anyone made any measurements in motor efficiency differences comparing running a motor sensorless, using hall sensors, and using various encoders?

 

[Russell]

Tests I did years ago showed no difference between running sensored or sensorless. Now that I have controllers which can run geared motors glitch free (the KU93 is my favorite) I run all three of my E-bikes sensorless.

Because I pedal a lot the thing which influences how much battery power I use is how much like a regular bike my E-bike is. That is anything a regular biker would pay attention to like weight, rolling resistance, aerodynamics and especially fit is what affects my battery power use. For example using my Jeep "trekking" bike as a baseline I tend to use 10-14 Wh/mi to average 18 mph. On the lighter and more efficient Nashbar flatbar bike I built last year I consistently used 25% less battery power over the same route at the same average speed. When riding my Townie I don't pedal as efficiently plus the bolt upright seating position catches more wind resulting in at least 25% more battery use than the Jeep. When riding over the same course with very little pedaling the differences are much smaller and all fall within the 16-20 Wh/mi range over a relatively easy 18 mile course at 18 mph.

For someone who rides fast and full throttle most of the time aerodynamics and choosing the correct motor for the job will be the primary factors which influence power consumption. It is also generally accepted that high power bikes should be run sensored.

-R

 

[bowlofsalad]

Tests I did years ago showed no difference between running sensored or sensorless.

Thanks for the reply. I did some searching, I wasn't able to find the results of these tests in your posts, did I miss them?

 

[Russell]

I didn't do any formal testing of sensored vs sensorless however I have kept a detailed log of my rides since 2005, several years before I electrified. Originally I used it to see what small changes to my bikes and how training affected my average speed. I continued this when I started E-biking to see what changes I made in components did. My conclusion over 6 years is that for a class of motor and controller, I concentrate on small geared motors, there is very little difference in efficiency from one setup to the next. For the type of riding I do, human powered w/motor assist, the overriding factors which affect my battery power use are; 1) the bike I choose to electrify, 2)the speed I ride at, and 3) my conditioning and power contribution.

If the type of riding you will be doing and the E-bike you'll be constructing is more of a moped or even motorcycle then the priorities change.

-R

 

[bowlofsalad]

there is very little difference in efficiency from one setup to the next.

But there is a difference, and that is what this thread is about, knowing, through scientific method, what that difference is.

You seem to be approaching this question with the mind set that I am seeking advice for electric bicycles or to be saved from something. My question primarily relates to the various options and ideas for designing motors and simply because I am incredibly curious among other things. Really, even if the difference was .1%, I'd still genuinely like to know.

A lot of why I ask this question is because I am thinking on how to perform these sorts of scientific tests myself. In part, I am skeptical and curious of many of the claims that groups make on their motor efficiencies, and the rest is, as I previously mentioned, knowing what gives what sort of values. But if someone has a deep knowledge on this subject and science already made with the will to share, I'd gladly pick up whatever I could.

This video should illustrate a technique for measuring motor efficiency. Though I don't really know if this is an optimum method or if there is something better.

[youtube]OUt9th2D-Pg[/youtube]

No matter what, I may end up performing these tests simply because it may determine more accurate controller settings. ASI and adaptto controllers seem to require a fair amount of configuration and I am sure that a mildly higher amount of efficiency when comparing settings would be an excellent indicator of what is good and what isn't.

 

[Jason3]

Hello,

Has anyone made any measurements in motor efficiency differences comparing running a motor sensorless, using hall sensors, and using various encoders?

Yes. The lower the eRPM, the more difference higher resolution feedback to the controller makes. As the revs rise, the difference shrinks markedly. I've seen measurable differences, but can't give you accurate numbers because I was not specifically looking at efficiency. It was complicated by the fact that none of the controllers I have can handle all three options yet (one manufacturer promised to support all three, but now says it's unlikely they will be able to). Some other points to note - inaccurately positioned hall sensors can potentially do more harm than good, and magnetic encoders don't like to be too close to the motor windings or magnets!

I know that it was not your question, but commutation methods made a bigger difference than feedback type in my case. Justin has posted some results of his experiments to quantify the difference between FOC and Trapezoidal commutation on his BAC500 thread, as you'll know. Incidentally, the BAC500 has the steadiest voltage output at constant (low) speed of any drive I've tried, if any conclusions can be drawn from that...

This question interests me too, any improvement in efficiency could be weighed against the cost and effort of adding feedback, and any drawbacks of controller choice that feedback might dictate.

Jason

 

[bowlofsalad]

Hello,

Has anyone made any measurements in motor efficiency differences comparing running a motor sensorless, using hall sensors, and using various encoders?

It was complicated by the fact that none of the controllers I have can handle all three options yet (one manufacturer promised to support all three, but now says it's unlikely they will be able to).

Thanks for the reply, Jason3.

I've wondered about this notion greatly, if an optical encoder can be used with any controller hall sensors can be, sounds highly unlikely.

Yes. The lower the eRPM, the more difference higher resolution feedback to the controller makes. As the revs rise, the difference shrinks markedly. I've seen measurable differences, but can't give you accurate numbers because I was not specifically looking at efficiency. It was complicated by the fact that none of the controllers I have can handle all three options yet (one manufacturer promised to support all three, but now says it's unlikely they will be able to).

I am likely to try performing these sorts of tests using 2 or 3 different motors (geared and direct drive) making similar observations.

This question interests me too, any improvement in efficiency could be weighed against the cost and effort of adding feedback, and any drawbacks of controller choice that feedback might dictate.

Jason

This is largely the question I am trying to answer through learning about how various ideas impact motor efficiency. If it cost an addition 50 dollars per motor to gain .5% and in efficiency using some type of encoder and needing a specialized controller where instead thinner lamination would gain considerably more overall motor efficiency, pros and cons to weigh and so on.

 

[Russell]

OK if you want to squeeze every little bit of efficiency out of your system then you start with the low hanging fruit - losses due to resistances. First get a controller with IRFB3077 or 4110 FETS. Next use some good high C-rate batteries. Then use large gauge wire for all connections.

A direct drive motor is going to have a higher efficiency than a geared motor. If the motor will be running all of the time it is the logical choice for maximum efficiency. If you're in a hilly area the DD motor offers regen which will boost system efficiency. The less the motor is used the more the benefit of a lighter geared motor becomes since it freewheels. Either way picking the correct speed motor is very important. A too fast wound motor for application will have lousy efficiency.

These are the things I would most pay attention to in order maximize the electrical system efficiency before looking at more esoteric parameters. To run proper experiments on how much difference each item makes requires controlled conditions, preferably a dynamometer. When I built my first E-bike I used an indoor trainer to serve this function. I dialed in the resistance on the trainer which best simulated level road cruising then read power into the controller via a Watt's-Up meter and read power out from the trainer's display.



As I mentioned previously you can throw all the electrical system efficiency quickly out the window by installing it on one bike vs another. Of course that's another experiment dealing with aerodynamics, rolling resistance, mechanical resistance and rider power. Addressing these parameters will have far more impact on power consumption than anything on the electrical side.

-R

 

[bowlofsalad]

OK if you want to squeeze every little bit of efficiency out of your system then you start with the low hanging fruit

I have considered all of these things you have mentioned in a likely obsessive level of depth, but they don't really relate to the subject this thread is about. The subject this thread is about is comparing various types of motor feedback systems and how they influence motor efficiency. I fully appreciate your comments and ideas relating to things like drag, I ride primarily recumbent bicycles for this very reason, but I much prefer that forum threads stay on subject.

You seem to have a lot to share on the subject of watt hours used per distance traveled, perhaps you should start a thread that acts as a guide to or discussion of that subject.

 

[Russell]

OK, OK I'll try to stay on the subject Something interesting you might want to pursue is changing the timing of the hall signals. I know it can be used to increase motor speed but I wonder what it does to motor efficiency. There is someone on the for sale thread selling a device that changes timing on the fly. Also some controllers can be programmed to advance timing. My KU93 controllers actually run 3% faster in sensorless mode compared to sensored mode so the calculated hall signals must be advanced slightly over the signals from the motor.

-R

 

[speedmd]

This whole topic of wave forms, pulse widths / frequencies and timing has some interesting possibilities /results. Lots to explore beyond the simple/gross efficiency question. Harmonics and how a controller best deals with them is another whole area to explore. I need to learn much much more :?

 

[Burtie]

Hello,

Has anyone made any measurements in motor efficiency differences comparing running a motor sensorless, using hall sensors, and using various encoders?

Via PM...
... I'd love to read your thoughts on the subject..

IMHO It is not easy to generalise and say that one method is more efficient than another
Trapezoidal controllers we commonly use, need rotor angle feedback resolution of 60 Edegrees.
The more complex FOC (sine) controllers require a very much finer angle resolution to operate effectively.
What works well for one commutation algorithm may not suit another.


The optical sensor boards that I produced, generated a 3 bit gray code, the same as that usually produced by embedded hall sensors. So they could drive trap controllers directly.
To work with the FOC controller, I wrote software to use linear interpolation with this 3 bit signal. This worked well when running, but was not ideal when starting from 0 rpm.
The use of a higher resolution encoder would have been an advantage here.


I did not collect any hard data to substantiate this, but during bench testing, my perception was: the smoothest, quietest, lowest current draw for given speed, (highest efficiency ??) was obtained using sensorless commutation with the FOC Controller, (the sensors were still needed for start up).
During road testing, there was no perceptible difference in efficiency or operation, between sensored or sensorless modes.


When comparing the performance of the optical sensor boards to that of a good hall setup, I could not see any significant difference. Perhaps just a bit less timing jitter with the optics (when viewed on a 'scope), which may work better with some controllers?
The main operating advantage of the optics was that they were immune to interference from stator magnetic fields, which was a major issue with some of my experimental setups.


Some things to consider when choosing sensing methods for commutation:

1) Many digital sensors (including hall and optics) use hysteresis to reduce problems with noise and false switching.
This means that optimum placement for forward and reverse operation will differ slightly.

2) Switching latency (delay) . This will result in timing signals becoming retarded as speed rises.

3) Max frequency of operation.

4) Environment -heat ?

5) Does it work at 0 or slow speed ?

6) what is the interface to the controller/ what resolution is needed


Loads of factors to juggle

Burtie

 

[Jason3]

OK, OK I'll try to stay on the subject Something interesting you might want to pursue is changing the timing of the hall signals. I know it can be used to increase motor speed but I wonder what it does to motor efficiency. There is someone on the for sale thread selling a device that changes timing on the fly. Also some controllers can be programmed to advance timing. My KU93 controllers actually run 3% faster in sensorless mode compared to sensored mode so the calculated hall signals must be advanced slightly over the signals from the motor.

-R

A FOC capable controller will automatically optimise the timing.

Jason

 

[speedmd]

Thanks Burtie for the detailed post.

What I don't yet understand is why a typical ebike motor is limited in rpms. With a certain voltage it turns x no load rpms. Is it reaching a best balance point between the losses (iron and copper) and BEMF, or is the controller not able to advance timing, raise the frequency enough or not worth it in terms of efficiency or potential power delivery. With the few I play with, it just seems like it is most all controller related and we are still in the product developments infancy.

 

[bowlofsalad]

The more complex FOC (sine) controllers require a very much finer angle resolution to operate effectively.

Thanks for the reply Burtie! I am so curious about what this means, can you elaborate?

Another way I might write your sentence(doubtful I understand what you mean) is field oriented control is less efficient than it can be when ran sensorlessly or using hall sensors (low resolution), which makes sense and is plenty of the reason why I made this thread. How much less efficient is a pretty important question in my mind.

I did not collect any hard data to substantiate this, but during bench testing, my perception was: the smoothest, quietest, lowest current draw for given speed, (highest efficiency ??) was obtained using sensorless commutation with the FOC Controller, (the sensors were still needed for start up).
During road testing, there was no perceptible difference in efficiency or operation, between sensored or sensorless modes.

It makes sense based on what Jason said that as erpm increases that sensors become far less important and their effect in motor efficiency because much more subtle, but at lower erpm, during things like climbing and acceleration I imagine there is a difference between sensored and sensorless, probably a measurable one much like what can be seen https://www.endless-sphere.com/forums/viewtopic.php?f=31&t=65031&start=75#p981872 here.

 

[Lebowski]

The more complex FOC (sine) controllers require a very much finer angle resolution to operate effectively.

Another way I might write your sentence(doubtful I understand what you mean) is field oriented control is less efficient than it can be when ran sensorlessly or using hall sensors (low resolution), which makes sense and is plenty of the reason why I made this thread. How much less efficient is a pretty important question in my mind. .

I have to correct you here, FOC can only be properly done sensorless and only when sensorless will give you the best efficiency.

 

[bowlofsalad]

The more complex FOC (sine) controllers require a very much finer angle resolution to operate effectively.

Another way I might write your sentence(doubtful I understand what you mean) is field oriented control is less efficient than it can be when ran sensorlessly or using hall sensors (low resolution), which makes sense and is plenty of the reason why I made this thread. How much less efficient is a pretty important question in my mind. .

I have to correct you here, FOC can only be properly done sensorless and only when sensorless will give you the best efficiency.

Any chance you've made some dynamometer or motor efficiency graphs for comparison? I would literally pay to see them.

 

[Russell]

Any chance you've made some dynamometer or motor efficiency graphs for comparison? I would literally pay to see them.

Have you read Justin's thread about his FOC controller testing?

http://endless-sphere.com/forums/viewtopic.php?f=31&t=65031&start=50#p981766

-R

 

[speedmd]

Thinking that "controllers" deserve their own category on ES and should not be mixed into sales, tech or motor groups. The part is a major component and shared across most all of the other categories and deserves its own resources, educational /reading links, test results, supplier listings and the like as done on the motor page. Now it is scattered throughout the site in various sub categories where it is a bit difficult to pull all together to get up to speed. Where to start? :? For me, I am searching for a most simple explanation /understanding of how a sensorless FOC operates. Timing and wave forms are more straight forward to understand, but what makes for feedback/ control settings for max efficiency over a broadened rpm range is a topic that is most widely interesting as the technology develops.

 

[Burtie]

The more complex FOC (sine) controllers require a very much finer angle resolution to operate effectively.

Another way I might write your sentence(doubtful I understand what you mean) is field oriented control is less efficient than it can be when ran sensorlessly or using hall sensors (low resolution), which makes sense and is plenty of the reason why I made this thread. How much less efficient is a pretty important question in my mind. .

I have to correct you here, FOC can only be properly done sensorless and only when sensorless will give you the best efficiency.

As I understand it:
When trying to make phase currents accurately match a desired formula in relation to rotor angle (example - the sine of an angle) it is better to use many data reference points rather than few, as this will give a more accurate target to aim for. This is why the rotor angle feedback resolution is important with FOC.

Surely, whatever the mechanism is that provides this rotor angle data is not so important.
What is important, is that the data is accurate and of sufficient resolution.
Doubtless there are good and bad implementations of both sensored and sensorless systems that provide this data.

@Lebowski - Are you favoring sensorless because it has the potential to give better resolution than any encoder?

 

[Lebowski]

Only the backemf itself can tell you what it's phase is and therefore give you the correct phase for supplying current (into the motor)...

 

[speedmd]

Thanks Lebowski

Starting to understand better. You hitting a moving magnetic target with another timed electro magnet. Sort of like a baseball except in attraction.

Back emf adds another dimension effecting balance /timing as does the green ball.

 

[Lebowski]

It is actually a lot simpler than that.

Take a voltage source. If you want to take power out of this, you drawn current out from the voltage source.
If you want to put power into the voltage source you stuff current into the voltage source. Like charging and discharging a battery.
From an electronic point of view you don't care about all the chemistry behind it, it's just a matter of pulling power out
or putting it in.

Now, a rotating motor is also a voltage source, albeit an AC source with a sinusoidal waveform. This source is reffered to as the backemf.
Again, to put power into the motor you stuff current into this sinusoidal voltage source. If you put a constant current in you're not
applying power as the voltage is positive and negative as it goes through its sinewave shape.
For stuffing maximum power into the backemf voltage source the current must also be sinusoidal and have the exact same phase as the
backemf voltage source.
From an electronics point of view you don't care about magnets and attraction and things like that, all you care about is you have a sinusoidal voltage
source and you want to put power into this by stuffing in a sinusoidal current. How this is then transformed into mechanical power, as a controller builder you don't give a r@ts @ss about this.

Now, what the halls do is they tell you roughly what the phase is of the sinusoidal voltage source, info you need to determine the phase for the current
you need to apply to put power in...

So, this is where my statement comes from, the only correct information comes from the backemf voltage source itself, and therefore sensorless is best
when it comes to efficiency.

 

[speedmd]

Thanks Lebowski for the simplification. Got it, we are just timing the input to match. Most times it is much simpler than it originally seems. Now, do we always assume the back EMF is a pure sine wave or care about matching it's shape for efficiency? Would think with inductance variations we see some slight variations in wave forms. On some controllers when going to 120% speed settings are turning to square waves and doing other unexplained things. Is this just a barbaric way of stuffing more current in regardless of the possible efficiency hit?

 

[bowlofsalad]

Only the backemf itself can tell you what it's phase is and therefore give you the correct phase for supplying current (into the motor)...

I doubt I understand what this means. Are you saying that it's possible to always, with perfect accuracy, to read rotor angle with zero sensors and that sensors are simply used to compensate for some sort of inadequacy?