I Built an E-bike Motor From Scratch

追加: 2026-05-20

[00:00:00]Can I make a useful electric motor? See, while I have built half a dozen electric motors on this channel before, then this is a toy demonstration.

[00:00:12]This one barely moved an RC boat around.

[00:00:16]This one would melt if I pushed more than a few amps through it.

[00:00:21]Self-destructed after 10 seconds of full throttle.

[00:00:25]Never really worked properly. and after modifications was significantly underpowered with tons of inefficiencies.

[00:00:33]So, you see, I haven't actually managed to build a useful motor before. So, this time I set myself a goal of building a motor that could power an ebike.

[00:00:44]Furthermore, I wanted this motor to be built out of as many off-the-shelf components as possible. I decided that I would make an axial flux permanent magnet synchronous motor. axial flux because it is the architecture I've worked with the most and BMS also known as BLDC motors because it's the most commonly used architecture in ebike motors which allows me to use an offthe-shelf electronic speed controller for actually driving the motor. I will also forgo any experimental design choices like my involute stator design.

[00:01:20]Now I have explained what exactly differentiates axial flux motors from radial flux motors as well as how BLC motors work in multiple of my previous videos links in the description. So I will only go over it quickly here. So the difference between axial flux and radial flux motors is the orientation of the electromagnets and the magnets in relation to the spin axis. And the defining characteristic of BMSA motors is the way they are controlled and the presence of permanent magnets in the motor as opposed to an inductive rotor.

[00:01:57]Got all that? Good. Now we can get on with the actual design challenge of this motor. The first things I decided on were the power output and voltage of the motor. I chose 500 W and 40 volts as those are fairly common ebike motor specifications and the power output is relatively easy to work with even in a DIY setting. Since power is equal to volt* amps, then it means my motor would have to be able to handle 12 1/2 amps of continuous current. Next, let's take a look at these two equations. Notice how they are suspiciously similar. Well, this is not just a coincidence and has a direct equivalence to our problem space.

[00:02:42]The rotational speed of an electric motor is directly proportional to the motor voltage and the torque to the motor current. Now, in our ebike motor scenario, we are more interested in having a high torque rather than high speed since the tires of a bicycle actually spin fairly slowly even at high velocities. In addition, we are interested in having the extra oomph to get going from a standstill. So, one simple way to increase the torque of a motor is to increase the number of wire turns on the electromagnets, increasing what's called ampear turns. Something to note and keep in mind here is that since we are keeping our power output the same but increasing the torque of the motor, then the speed of the motor must necessarily go down. At this point, one could ask, why couldn't we just keep adding more and more turns until we get the torque and speed we want? Well, unfortunately or fortunately, we live in a physical reality and not an infinite abstract space. And as a result, at one point, we'd simply run out of physical space to fit more turns. So, naturally, we just reduce the size of the wire, right? But that means that the resistance of the wire goes up dramatically, increasing power loss to resistive heating, which could get so bad that the motor melts down almost instantly. For example, for my motor, I used a table to decide that I shouldn't go below roughly AWG14 sized enamel copper wire. Luckily, there are other ways to increase torque in motors. For example, a respectable increase in torque can be achieved by using iron or electrical steel cores in the electromagnets. However, adding in these cores does add in the problem of magnetic saturation.

[00:04:40]Essentially, it sets a maximum amperage and winding wire turn count on the electromagnet core before the core becomes inefficient.

[00:04:49]Fortunately, there is one more way that we can increase the torque of the motor.

[00:04:54]just add more electromagnet cores. This basically acts like adding more teeth onto this electrical gear. When before a single electrical rotation would cause the motor to do one full turn, then adding a set of electromagnets, one for each phase, makes the motor do only half a revolution.

[00:05:16]Add another set, of course, it does a third of a turn, and so on. Since our power is still fixed but the speed has reduced then it means that the torque must have gone up. Now there are many things to consider like the cogging torque and winding factor that would tell you what number of electromagnets and magnets to have on your motor. But I simply solved this problem by looking up a good combination of slots to poles from a chart. half based on an educated guess and half on vibes. I decided that 18 slots to 16 poles is a good combination to go for. Consequently, this combination does give me a very certain kind of winding diagram I have to follow for which I once again just use an online calculator. In summary, designing a motor is a kind of balancing act between the power voltage, amperage, torque, speed, size, weight, and so on.

[00:06:16]The application dictates the parameters.

[00:06:20]At this point, a good engineer would use a whole bunch of formula to further close in on the set of parameters for further optimization.

[00:06:29]But I'm not a good engineer. In fact, I'm not an engineer at all. So instead, I just threw a whole bunch of compute at the problem.

[00:06:40]This is FEMM, a free finite element analysis tool for electromagnetto statics, heat flow, and current flow problems. While it is fairly easy to learn and use, it does have one problem, namely that it is a 2D solver. This is perfectly fine in the case of radial flux motors. But in axial flux motors, the size of the electromagnet cores and magnets is dependent on the radial distance, which means I'll have to do some interpolation. And even then, it's going to be only an approximation. Regardless, here is my motor. More specifically, this is an approximation of a radial slice of the motor. Here's the magnets.

[00:07:28]Here's the back irons, the electromagnetic cores, the wires, and the rest is left as air because most plastics and resins have very similar magnetic characteristics to plain old air. By running the analysis on a few radial slices and then interpolating the results, I could get some approximate simulation results for my axial flux motor. The simulations allowed me to figure out at what point my motor cores would be pushed into saturation, how thick I should or could make the back iron, and most importantly of all, it allowed me to calculate the static torque of the motor.

[00:08:09]In the end, this is the motor design I came up with.

[00:08:15]Let's go through each of the assemblies one by one. Starting with the stator. To start with, I needed to make the iron cores for the electromagnets.

[00:08:25]So, I got my hands on some transformer core laminations.

[00:08:29]Now, generally speaking, transformer core laminations are not the best for use in motors. So, I ran some tests. I measured the heating level in the material based on magnetic saturation levels and switching frequency. But I found that for my use case, these transformer core laminations would work just fine. I started by shaping the laminations into the required wedge shapes with a hacksaw, which went as well as you'd think.

[00:09:00]So, I acquired a band saw.

[00:09:04]It's not that I couldn't do it all using a hacksaw and meat power, but I really did not fancy a repetitive strain injury this time around.

[00:09:14]So, I 3D printed some small forms to hold onto the laminations while I cut them, which went as well as you'd think.

[00:09:24]You know what insanity is? Well, it's not this because these more beefy 3D printed forms actually work the treat.

[00:09:34]So, here's a quick rundown of how you make an axial flux electromagnet core out of transformer laminations at home.

[00:09:43]First you pack the first form with laminations. Then fix the laminations in place.

[00:09:50]Align the cut.

[00:09:52]Cut.

[00:09:56]Release the laminations.

[00:10:01]Debur.

[00:10:04]Transfer the laminations to the next form.

[00:10:08]Fix the laminations in place. Align the cut. Cut.

[00:10:14]Release the eliminations.

[00:10:17]Deeper and transfer.

[00:10:20]Align the cut. Perform a super sketchy cut.

[00:10:27]Deeper. Wrap some captain tape around the laminations to keep them together as a core. And finally, sand or grind the cores to the correct final size.

[00:10:39]and then repeat for all 18 cores. Next, I needed to wind the cores. Instead of using AWG14 wire, I decided to use five parallel strands of AWG24 wire instead as it's a little easier to wind. For the winding, I used an old decrepit desktop lathe.

[00:11:02]With the help of Canoacrylic glue and silicone as adhesives, winding the cores was just a matter of not messing up the winding count and getting all the windings as consistent and tight as possible from core to core.

[00:11:21]Next, the cores need to be attached to something for which I decided to make a 3D printed chassis which went well. The first time I used ASA as the material for its heat and UV resistance, as this is the one part of the motor that will get quite toasty during operation, and I'd rather not have this happen while I'm going somewhere with considerable speed. Once printed, I placed all the cores in their corresponding slots based on the winding diagram and then soldered the required connections. You might notice that there is still a bunch of free space, and that's because I designed it to be filled with epoxy resin. I calculated out the approximate volume of resin I would need, mixed it well, and then deass it in this super jank DIY vacuum chamber.

[00:12:20]Pretty sure something is wrong with the vacuum bump because it doesn't seem to go beyond about 25 in of mercury of vacuum.

[00:12:29]Oh well, I'm sure it's good enough.

[00:12:33]I kept degassing the resin for about 20 minutes and then carefully poured it into the stator, trying to make sure I leave no air pockets anywhere.

[00:12:47]And then to finish off, I placed about 50ish kg of pressure on it.

[00:12:55]Over the next days, I could monitor the curing process by checking the consistency of the leftover resin. I think that for my first resin casting, the result is actually not bad. Oh, about those cables. Uh those are for the hall effect sensors which are used for determining the rotor position from a standstill. Usually they would be spaced out by 120° but my motor configuration also allows for a 60° separation. So I used that instead as it was more convenient. And that's the stator done.

[00:13:31]Next the rotor.

[00:13:33]First of all, this time I actually went ahead and ordered some custom wedge-shaped magnets from China because it's really difficult to find offthe-shelf magnets in the correct shape and size for axial flux motors. It was quite expensive, costing as much as a normal ebike motor would. So, don't do this kind of project if you think it's going to be cheaper. It's just not going to happen. These magnets need to be connected to one another through a back iron. The point of the back iron is to ever so slightly increase the magnetic strength of the magnets. Read up on magnetic permeability and magnetic reluctance if you want to learn why. It also acts as a solid superructure for the magnets to attach to, as well as shielding the other ferroagnetic components of the motor from the magnets. As the back iron, I used some cheap off-the-shelf e- scooter brake discs with as much material in them as possible. Ideally, I would have ordered some literal iron and cut out the needed shape on a CNC. Oh, yeah, the CNC. I have a CNC now.

[00:14:45]I actually even used it for one part of the rotor assembly, the rotor connector.

[00:14:51]As the FMM simulation shows, this motor actually has two rotors that will have to somehow be coupled together. And to do that, I simply decided to use a hunk of aluminum.

[00:15:04]Now, perhaps using a CNC mill goes a little against the spirit of simple DIY maker content, but to be fair, I could have simply used drills and files to make it. Anyways, I also used a drag scale and a chunk of electrical steel to figure out the relative magnetic strength of the magnets so I could better balance the magnetic pull of the rotor. Then I used a 3D printed aligner in order to make sure that the magnets on the rotors would be placed exactly in the right positions. I cleaned the surfaces and then used some cyanoacrylic glue to give that slight extra stickage to the magnets. Usually the magnetic attraction to the back iron is more than enough to keep the magnet seated properly, but I just wanted to make sure. Making sure to alternate between the polarity of the magnets, I eventually constructed a rotor.

[00:16:01]And then I repeated the process for the second side of the rotor as well.

[00:16:07]After completing the rotor, I also produced the various other bits and pieces like the rest of the chassis and some bearing retainers.

[00:16:19]Finally, the last problem to solve was to figure out how to couple the rotor and the shaft of the motor. Now, I could have made a spline shaft or perform a shrink fit, but I figured it would be overkill when I could just use an expansion sleeve. So, that's exactly what I did. This tiny thing posts a 19 Newton meter torque transmission capability. And after testing, I actually found that it's even more than that.

[00:16:48]Now, you might say that 20 Newton me is not enough for an ebike. And you'd be right as most ebikes have two to four times more torque. But from the simulations done, I already knew that I would have to do at least four times mechanical reduction as well. So the 20 Newton meters is going to be just fine.

[00:17:09]Once I had produced all the bits and pieces of the motor, it was time to assemble it all.

[00:17:29]Everything went together without much trouble until it was finally time to remove the last wedges separating the rotors and the stator.

[00:17:45]Disaster.

[00:17:47]See, because there is electrical steel inside the stator, then the magnets on the rotors are really eager to hang out with the electrical steel around 1,300 Ntons of force eager. I was a bit naive and hoped that since there's a rotor on both sides of the stator, then the forces would mostly cancel out. But the problem is that real life is messy and one of the stators will always have more pull. And as one side moves closer, the other side moves further, meaning the imbalance of forces is in a positive feedback loop. And despite the chassis being 7 mm thick, 100% infill PTG, it was still no match against the magnets. I was afraid that this could happen as earlier on I had actually ran a different simulation to see how much the back iron would bend under the load of the magnets. And already there I found that 2 1/2 mm of steel would bend by 0.1 mm. So instead of figuring out a way to make the chassis more rigid, I decided that I would simply ditch the idea of having irra in the stator. From the simulations, I knew that this would hurt the performance of the motor at low speeds, but I was really just hoping that it would be fine. So, I wound yet another 18 electromagnet course with actually a WG14 wire this time.

[00:19:41]I once again placed them accordingly in the new stator, soldered the connections, mixed up another batch of resin, degassed it, poured it, held it under pressure for a few days, and eventually I had myself a brand new air core stator instead. This time the assembly of the motor went completely flawlessly as there was nothing for the magnets to want to attach to.

[00:20:17]And so with the motor done, there was nothing more left to do but to hook it up to an ESC and test it. I started with a known good ESC meant for RC cars. And well, apart from having the wrong throttle configuration, the motor spun up without much trouble and worked fairly nicely.

[00:20:55]I then tried it with the cheap ESC I got that is specifically meant for ebikes and ecooters.

[00:21:09]And while it of course worked, it didn't sound all that good.

[00:21:25]And the unloaded power it used was honestly kind of insane. But the fact that the higher quality and much smaller RC car ESC used only a fraction of the power at a given speed suggested to me that the poor performance was probably the fault of the ebike ESC. Guess that's what I get for trying to cheap out.

[00:21:47]There is one other thing that most ESC's do that this motor probably doesn't like. Namely, most simple ESC's actually generate the trapezoidal three-phase waveform for the motor. However, since my motor has quite a bit of inertia in the rotor, then the difference between the trapezoidal and the ideal sine wave waveform of the motor might be fighting the ESC quite severely. instead.

[00:22:13]Ideally, the motor would be controlled by pure sinosodal waveforms instead.

[00:22:18]Lucky for us, these kinds of ESC's do exist. For example, the Vesque lineup of speed controllers, which use something called field oriented control, which honestly just means that the waveforms are sine waves and you can control the sine waves with greater precision. This relatively cheap flips variant has been recommended to me before, so I decided to get one. And the difference is kind of insane.

[00:22:58]So much quieter.

[00:23:03]So much more efficient.

[00:23:13]So much more control.

[00:23:17]It's all looking up after all. Maybe the motor is going to end up being useful.

[00:23:24]But to know that, you'll have to wait until the next video where I modify a standard road bike to use this motor. If you're interested in watching that video, then I'm sure you know what to do. Until next time, stay curious.