Hub Motor Design Simulations

But how would you design a motor with a more linear torque curve? If u did it would therefore have a more linear efficiency curve no? I’ve never heard of such a thing in motor design. Gaining more torque through motor design, sure, but it will still follow the same slopped performance curve when graphed. Can u please show me two graphs with one having a more linear torque curve and explain why

We tested different voltages and they did have an effect on efficiency. Low KV seams to like less Amps and higher voltages. Technocratically it should not have a huge impact within a certain range, agreed. Anyway, we always wanted to go for 12S, so most test were done at 45V and therefore I state the results we got together with the voltage we used. Roughly 88% efficiency at peak. When you read 95% efficiency somewhere, be sceptic! 75-85% is realistic for hobby size BLDC Outrunners. Torque is linear and nothing special - did I say we made torque linear? I said Torque is linear. Anyway, I don’t want to debate our Motor or any other motor. I stated that before: We simply assume the perfect motor to debate that issue. The “my motor is better than your motor debate” is not helpful in that case. All I want to say: Most of these Motors are quite good already and if you wind them accordingly to application they are performing quite good. There is an optimum RPM you should run them at in order to stay efficient. That is what Benjamin stated in his article. You can’t increase torque and lower RPM beyond a certain point without getting huge losses (square relation). If that was possible we would not see gearboxes in electromechanical machines.

More torque needed @ low RPM @ high efficiency: Bigger motor required, or gearbox. End of the Benjamin Story.

Basically my point is: Physics set limits and it has nothing to do with the design of the motor if you stay in the same size class. Some motors are better, other worse, magic won’t happen unless someone finds a major design tweak allowing sort of magnetic motor gearing or whatever. People spend their lives on motor improvements. Its not that easy.

Don’t get me wrong, the idea of Hubs is fascinating and you can ride them somehow if you can accept the facts above. Hubs are not yet what customers believe they are. They will find out. But: There is a market and someone will serve that. I really respect Jacob Bloy’s and your work on Hubs! Good job, go on, never give up.



The trouble is people want to hit sales and make products more than pioneer a science.

Here are two example virtual motors I toyed with. If these motors were being used under 1000rpm. Torque would be linear. At this point to make the torque flatter at higher RPMs, then the motor does have to be bigger.

I don’t understand how you can have a linear torque as speed is increasing. They are said to be inversely proportional as with increased speed the bemf increases and limits the current that can be put to the motor. It seems like those examples of a motor are stifled in some way and they should have greater torque at the slowest speeds on the graph. What characteristics of a motor would equate to a linear torque curve, meaning level? I still am not a believer yet. Convince me please!

I’m also not a believer in higher voltage making a more efficient motor if it’s wound appropriately for the voltage and load. @trampa. I’m sure I’m just sounding like a contrarian but I’ve seen too much evidence saying otherwise.

If the torque at 1rpm is below 0Nm, the graphs tend to be upwards sloping.

It’s annoying to get it right. Also level for a given frame of reference.

Don’t understand what ur saying How can a motor have a linear torque curve inherent to the motor? Never seen that other than through the controller limitation. I’d heard, as I just wrote above and makes sense, the back emf will reduce current at increased rpm. This is universal as far as I understand

I’ve stressed this a few times.

It’s a frame of reference issue. If I graph the height of a falling object, but zoom in to 0-100Ns, the change in high will look insignificant. It’s not representative of the whole picture.

If I take the tangent of a curve. It’s representative of the curves behavior at and around that point but not necessarily of the whole curve. If you build a motor that is efficient a a billion rpm, and you sample 0-4000 RPM, the massive slope over a billion could be virtually flat during the sampled range.

These graphs visualize data and performance. I set the window to 1-4000 because it’s the range I am most concerned with. If I set it to 1-10,000 the torque will become negative which is useless and not helpful for our purposes. I am including the relevant data here. Not the extraneous points.

Maybe you’re talking about a different graph or seeing it differently. Looking at the second graph you posted the most efficient speed would be roughly 15000 rpm, and assuming maybe 85mm wheel roughly about 15 mph and a possible top speed for a hub motor. In fact it doesn’t seem a small window on the bigger picture, this graph shows the motor spinning up to 3500, way beyond practical, without an increase in voltage somehow. This as it shows here would be a dream motor where torque doesn’t decrease with speed but instead is continuous all the way from 0 till the most efficient rpm 80%. There isn’t a power graphing but would like to see. Full torque at full speed I’ve never seen.

Hummie, you haven’t simulated nearly as many motors as I have. When you spend three months fiddling with the numbers to get SVM and DTC results. You’ll see a lot. I’ve kept the results I care about. The results I dint care about I didn’t bother to render as a PDF.

I have a life and I would appreciate it if you accepted my word that these graphs are a SMALL picture of a HUGE number of simulated results.

If you feel so compelled, I will generate a shitty graph for a motor with a peak torque at 8000 rpm that would have a flat torque at 0-4000rpm that’s below 0 Nm.

I’m just trying to talk about motors. I’m questionng how you can design a motor to have a linear torque curve as in the graph you posted These guys half way down the page lay claim to it and show a graph:

and they give the impression it’s due to the motor. I still have it unanswered for sure though as what they say is ambiguous and i think it would be the esc doing it and that’s what makes sense to me. I don’t think you can design a motor to have linear torque.

Oh I see. Yeah I think it’s nonsense. They’re ESC and motor are probably very well tuned to each other

Just claiming their motor is doing everything is probably advertising

And here I can page him with a snap @Mellow. Pretty awesome.
How’s the torque stay level as in your ad?

Hi, I put something together for you guys:

This would be a typical Dyno result: The 120 and 140 KV are real data from prototype motors we made in 2015, the 70KV Motor is a guess. I think a very friendly one, putting out 3Nm @ 30A @ 62.5%efficiency

Basically when you go for lower KV/RPM, the efficiency curve drops steeper. You can’t push as many amps at a given voltage + we deal with this issue: Copper losses are proportional to the square of the torque produced by the motor, and at low RPM and high load they are dominant.

If I now gear down our 140 KV Motor, putting out 3,5 Nm @ 50A @ 45V, using a 1:2.85 gearing, I will have roughly 10Nm at the wheel on a single motor setup. A twin drive could output nearly 20Nm! I would still be close to 80% efficient in that case. The Board is a beast taking any hill without even getting really warm. We managed to push 200Kg up a 12% hill with a single motor using a VESC4.12., 118 KV motor and 14/37 transmission.

Now we look at the Hub: 3Nm at 30A, only 62.5% efficient. A Twin would only output 6Nm under a lot of sweat. I would say 2Nm is more realistic if you don’t want to cook your coils and magnets.

Now some say: But the 15mm belt can’t handle that >> It can. If its rated 3.0 Nm for a given transmission, the 3.0 Nm applies to the smallest pulley within the system (weakest part of the chain). Assuming a 1:2.85 Gearing: 3 Nm x 2,85=8,55 Nm at the wheel. This is the safe side and usually the belts can handle 1.5x that. Otherwise you could not push 200Kg up a 12% incline on a single.

I still like the idea of hubs, I’m just very realistic. Its a motor problem that can be solved using a much bigger motor.



Hey Frank, When you talk about much bigger motors what size do you think would be adequate?

To big to fit inside a 100mm wheel.

I would love to see the dynos of some Hubmotors, so we are out of the guess range. Maybe 4-6Nm is achievable when accepting the massive low RPM high torque losses and managing to cool the system. Anyway, this is sort of funny because you don’t want to operate in that 50-60% efficiency range. lets assume you drop in temp and hall sensors. @105°C plus hall sensors will suffer and give up soon. Many parts of the motor don’t like the heat above 100°C over time. Heat is your enemy No1. Electric motors are so efficient, why wasting all that efficiency + get heat issues? Technically its the wrong approach to try to squees a competitive torque out of a 1:1 transmission without scaling the motor. Problems will occur!
When I understood Benjamins Videos correctly, VESC Tool will be able to calculate efficiency because all the relevant data is being gathered. So its just a matter of generating an output file. That would really help to understand the matter and everyone running Vest Tool would be able to see the performance of the system under strain.

@Hummie: Do you have some data to fill in? Torque, efficiency etc.

With regards to size: I’m not an expert but assume its a lot bigger in diameter and weighs like double to triple (sorry, square relation again).

I’m not against hubs! I like the idea, but currently I see to many issues to get really thrilled. Scaling would be needed. I could probably ride small hubs over here in Berlin. The town is flat like a pancake and my weight is low. I could go slower and my way to work is only 7Km. In Summer at 35°C, in combination with 70° hot tarmac + knowing my ride style I would pretty soon hit the limits. If you know the compromise, its fine. If the motor gets to hot, you should work towards a cooler system and not try to dissipate the heat (losses). A cure is always better than fighting the symptoms. But: Sometimes you have to accept a compromise.

In my eyes the compromise is the belt drive. Others see the compromise in the acceptance of losses.

In Germany we would say: Egg-laying-wool-milk-pigs don’t exist.



VESC with a few additional data could do that, the problem is that we have to measure mechanical output, if we knew precisely the drag coefficient, grade, wind speed, rolling resistance (including loss in bearings and belts) then we could do it, otherwise the value would be to far from reality

One way to do is to use a dynamometer to map the efficiency surface of a motor, like the picture from mellow i posted a while back, and store that on VESC or some microcontroller connected to it, them knowing the current on the motor and speed would could go on that data base and see what efficiency we are getting, would be a cool information to display on the remote to ride as efficiently as possible

@Mellow could you guys share what software suite have been used to develop your motor? i’m trying to find a cheaper alternative to MotorSolve, even though i think it does not exist

It’s not new turf and electric bikes are still using hubs and mid-drives, as their pulley setup is termed. Hubs are going strong still. Some have developed internal gearing but a lot haven’t and as this article states the gearing just got in the way of where there could’ve been more motor.

Motor simulations may reveal a lot but I’ve read of too many incidences of them not being accurate and they don’t take into account the losses of the gear system (I’ve read that the pulley and belt is maybe 3% but simply spinning a hub motor vs pulleyed wheel I have suspicion it’s much more). The real revealer is on the road and how many watthours spent to reach a distance at a certain speed and I haven’t seen a test yet! Next week I will make it happen. At least I can definately do the hub motor numbers and hopefully can find a pulley setup with a vesc to compare to. Anyone in San Francisco capable of doing the pulley board numbers? @Jinra

One mile at 15mph seems a good test but the variation in kv and gearing will make it unbalanced and have to see. I guess I could just lower voltage to get to 80% no load at that speed…although very rarely someone @trampa states different voltages will change efficiency performance

We could use the results to see if the computer simulations are accurate

I’m glad you also bring up hall sensor temp maxes. I sent Enertion a PM about this, but they didn’t seem too concerned. Only time will tell.

What seems like a really important thing to build and design as a community is a universal dynamometer we can all compare things on.

There are a few different projects going on to make this happen, but a community effort might be a great thing.