Yeah, sorry I didn’t see your reply.
Boosted and Mellow probably use DTC, not SVM. That’s the secret tbh. The VESC can be programmed for DTC, but it would be a significant firmware rewrite. Plus it works great for external motors so I think we just need a hub centric ESC
Nice stuff. Sorry to be coy. We’d love to be able to contribute, but want to hit the market first before spilling (too much) of our guts.
Isn’t Svm a variant of dtc…? Don’t you mean dsc? Why is it so much better? Are you winding the motor to use more of the winding >66% all the time?
You guys already hit the market what are you talking about lol
Nice Job @anon94428844. I Think before going that deep into detail, you should assume a perfect motor and a perfect electrical system. When you do that, you dig down to the physics behind BLDC motors. And you can stop to debate different motor designs. The “my design is better than yours” discussions don’t help to understand the matter.
We just assume the designer did everything right, he managed to build a nearly perfect motor for the size of a 90-100mm Skate wheel.
Benjamin Vedder did some nice explanations on that issue: http://vedder.se/2014/10/chosing-the-right-bldc-motor-and-battery-setup-for-an-electric-skateboard/
So that this size motor wants to run at about 60K ERPM to be at peak efficiency for matters you can’t influence. You can’t cheat the Physics behind BLDC outrunners. Dividing 60K ERPM by 7 is 8570 RPM. Using a 83mm wheel, wanting to hit roughly 35 km/h you would need to run a 1:4 gearing. Using 5M belts you can achieve roughly 1:2.5, so your system is already a bit less efficient, running the motor slower than it should. Anyway, such a geared system could still be up to 85% efficient. Losses in the belt system are only 1-2%, so nothing that needs to be addressed. Now we look at 1:1 gearing and drastically reduce the RPM. We will end up with a motor that runs most of the time in a region of the efficiency curve that you don’t want the motor to run in. At best you are 50-60% efficient, since the efficiency curve usually has a steep drop!
What happens: A huge amount of your stored energy (inside the battery) will dissipate into heat, reducing range and causing problems for the motor (overheating)
What do you do: Lets use the hanger as a heat sink (works at least to cool the stator, not the magnets)
What does that tell you: You ride a super inefficient system. Heat = Losses
If your electrical system is staying cool, you are running a efficient system - If you face heat problem, something is funny.
I do understand that hub motors are interesting and everyone wants them. Please understand that they are at this stage a lot more inefficient and do have disadvantages. Its O.K. if you live in a rather flat and cool environment and if you can live with a lot less efficiency for the benefits of the stealthy look. If your tarmac is 70°C in summer + you face some hills, or want max. range, or you are going quite slow (20-30 km/h) or you need more urethane on the Hub (bad roads) it might be problematic to ride hubs. Understanding the Pros and Cons is important. I am curious if the simulations will go hand in hand with the measured results.
Frank
I think the variable load and speed people do while riding negates a lot of the value of doing much optimizing. There’s compromises made. we all know that the most efficient way to run the motor is 80% of the no-load but no one ever stays at that speed for long, and if you also want a high top speed you’re forced to have a large Rpm range and greater inefficiency.
With the way the vesc is getting more efficient and able To take more amps it seems more practical to make a high kv motor and then if someone wants to run it more efficiently they can lower the voltage.
I’ll get you number and shapes Ken when the motors come in soon though. I’m interested in what you come up with and have been looking to do simulations but know that a lot of the variables are set in stone for me (stator and back iron shape and magnet thickness) and all that I have adjustable is the magnet strength and temp ability and kv. And in my experience the biggest factor has been what the manufacturer decided they can wind with. All I’m left to do now…all I feel is worthwhile doing now…is repeated asking the manufacturer to jam as much wire in as they can. Even maybe multi strand would be better if they can get more copper in. It’s largely copper losses with these slow turning hub motors…more copper less Resistance and a better motor. From the little I’ve found magnetic saturation and the increased hysteresis that comes with it is at flux densities way high and not as much of a loss as I’d thought maybe possible. But I’d like to see and can get you what steel the stator I’m using is made of. Known as Kawasaki steel.
What I really think would be an improvement which no one is doing is halbach arrays for the magnets!! I hope to do that as soon as I can I think losses in the belt-pulley system are much more than just a couple percent @Trillium. Pushing the board I can feel a lot more resistance. If a hub motor and pulley set-up are compared with both operating at 80% of the no-load I bet the hub will by far be more efficient. Not getting up to speed but once there. And there’s other benefits in a hub motor of course which I’d say out-value possible losses in efficiency.
I don’t disagree with you about efficiency and losses. I agree that hub motors in their current setup dont work all that well. Essentially I am moving from SVM motor control to DTC to have better information about the thermodynamics before I wind a motor.
A motor used in SVM based on the simulations with no load tends to have a sharp torque curve. The strandcount/total wire gauge is low and the number of turns is high. You don’t need 100% fill to get peak efficiency with SVM. Essentially, Jacob and I scratch our heads at the SVM simulation data cause it doesn’t always make logical sense.
Using DTC control, a motor is wound to maximum capacity with the thickest gauge and maximum number of turns. The thickness of the gauge does correlate to the speed you’ll get and the number of turns are also related to the toque. If you assume 30% grade, 100kg, 100mm wheel you’ll want to be on the safe side and make at least 10-15Nm of torque per wheel. Not only does DTC follows all rules of motor logic making it easier to fiddle with the numbers, but the simulated thermal profiles with and without load are promising.
The way the simulation data churns out, SVM driven motors are just harder to get thermal calculations before manufacturing. DTC can be calculated thermally and with a load. The goal is to keep the temperatures under 200C for up to three hours at max. At this point, a 6352mm wheel goes from 23C to 200C under max torque and max rpm in a about 60 minutes (For normal riding varied flats and 30% grades, you can go 2.5 hours without needing to cool down, and on a total flat you won’t pass 80*C). That’s not bad, it’s not great though for people who put pedal to the metal. I figure a thermistor can tell people who do ride crazy hills all day that they need a few minutes to cool. I am using Seismic trucks (the G1 spring truck) because of the light aluminum and structural stregeic cutouts. Way better heat sink than any solid truck, even if it squeaks like there’s no tomorrow.
On manufacturing I just want to be clear - you can wind a stator with 12 AWG wire or 12 AWG worth of 36 AWG. The thermal properties are barely different in simulations, but it’s a ton more wire to buy and if you’re winding by hand it is an unfortunate business.
Hub motors are different from the ground up in what were asking from the BLDC. We should approach them not by adapting them to existing tech, by developing around its needs. What does it need to have both torque and speed? The answer is not SVM at this point. The VESC 6.0 schematic includes more current sense resistors so you’ll have a higher accuracy with FOC. Additionally it overall design is promising for the SVM driven hub motors. What is left is the firmware to take advantage of these hardware improvements. Benjamin believes that the 6.0s FOC will be as coordinated and powerful as DTC, but well see. It’s worth building anyway.
I read a paper that seemed to indicate Svm was generally superior to DTC now that processors are so fast and cheap. I don’t understand how you would get much gains from DTC that could’ve just be easily offset by stator geometry / cooling optimization in the first place. I think the only real gains with control are going to 6 phase or greater system. I would run a 6 phase in a heart beat if there were proven controllers (and small, like half size of vesc) available for it. I think 6 phase will be a later version of Trillium once the mechanicals are proven.
Depends on the paper you read and the specific application studied. DTC can be used with a lot of different motors, BLDC is one of them and put runner design is different from inrunner.
The reality is that the motors are going to have to get a bit bigger if we want efficiency as Frank said earlier. By how much? Depends really. The question for size is determining what sacrifice I can live with.
On my to do list is to build a dynamometer and get a steady test bench setup (@jacobbloy). It’ll make development a lot easier.
I have built one and it was not cheap. Let me know if I can help I’m in LA.
Medford Massachusetts.
How much did it cost and what kind of scale are we talking? car/motorcycle/bike?
I spent around $3k. half machining half instrumentation. I have it set up for up to 50Nm. Use a huge motorcycle hub for varying load. It was mostly intended for simulating rides with a truck and two motors, so currently it is only set up to measure one motor at a time. I could make it do both motors but just seems wasteful
Why do people want 200N of thrust per wheel? That’s insane. I don’t even think boosted is doing that with their geared system. Is there documentation somewhere on that? I was thinking of buying one just to benchmark, but they are having issues lately
A lot of customers want MTB being 50+km fast and able to climb massive hills and they should have a super good range. I tell them they should balance the system, using these tiny motors. It’s still a one gear system. If you aim at an efficient system, you need to find the right balance. Average Speed should match 80% of no load RPM. With Hubs that’s difficult because they start to feel well at 60+km/h. Average speed is realistically 18-22km/h. The gap between average speed and top performance speed is to big.
Belt systems are 98% efficient. The reason why you can’t push that good is simple. You push against a efficiently setup generator. The motor spins faster when you push (2.5x), creating more resistance. In many cases people do also over tighten the belt. Freewheeling pulleys would be nice…
I don’t think controllers will do the magic. Its really a motor problem. Maybe thinking outside the box will bring something up.
Frank
Consider 200N of thust as the max and that we want the ESC to keep the user around the 50-75% mark. The ability to push 200N should only be unlocked when the speed is not maxed at 25mph and the current draw is high (aka, accelerating up a terrible hill).
I’d rather over design this than underdesign it. We are all using off the shelf parts (the VESC 4.12 and VESC-X both are essentially off the shelf at this point).
Measure twice, cut once.
And yeah, currebt hubs aren’t there yet. But that doesn’t mean we can’t bring it there. It will take a ground up approach and it will take time.
Edit, Tampa beat me to the point about why 200N. 80% target.
I’ll definately finally do some comparitive testing of pulley vs hub and energy needed. You make exceptions for why such resistance and say it’s only a 2% loss at the belt but somehow in practice a coasting hubmotor will go over twice as far in my experience. Same electrical cogging issues with both set-ups but none the less much further. So much assuming and guessing and speculation related to efficiency, a lot of it from me, and the real world will tell us soon and I’ll get out there and report back. Maybe I can get @Jinra to come out again and we can do it this time. I’ll be rolling in a couple days.
Belts are efficient. 98%. You just spin the motor faster when pushing a geard system. Its a more efficient generator in that case.
Frank
Assuming 98 percent efficiency, maybe it’s not taking into account the gear ratio and extra rotation necessary. With a 3:1 maybe it’s 3x that 2% loss. Maybe. None the less it feels a lot more and I want to compare real numbers
Personally, when I ride my Boosted, I get to top speed and then I coast it out till I’m at 13mph and then go back up and repeat. I get about 9 miles on the V1 Dual now at Pro mode instead of the 6 I used to get riding at full throttle the whole time. Commuting time is a little longer than driving when you’re not at 22mph the whole time, but the range is worth it.
I just saw an article that I can’t find about a guy who put his Tesla into neutral all the time to ride/coast and he gets a crazy high range per charge.
Hubs can be great for riders with patience who are commuting. If you want to gun it all the time, then gears might be best for now. Hopefully the tables will turn and we can get the best of both worlds.
Main problem with belts are damage, that’s my reason to start developing a hub, what’s worth to have a better efficiency if I have to change belts every few rides?
This is my particular situation, “good” roads but full of little rocks
For those interested, there’s a software called Emetor, free and online, nowhere as good as motorsolve but it’s free