Thursday, January 28, 2010

EAM: Single-Winding Back EMF Test

I finished up the sheet metal can of axial motor fluxage:

It's four layers thick, squeezed down from 12" ID to 9" ID by painstakingly tightening the hose clamps, flexing the can a bit to get it to slip, then tightening again, for about three hours. That's after I used up an entire cutoff wheel to cut it out of the large roll. Don't worry, I cut the ends of the clamps off before I ran it. This isn't part of the real motor, it's just a hack to get a more realistic result from the single-winding test. See this post for a better explanation.

Now, I'll admit, I kinda over-hyped this experiment. Although I still wouldn't want to be anywhere near it, this was a low-power real load is being applied to either motor. So, although it looks "terrifyingly epic," the kart motor had no trouble at all getting it up to speed with just a bit of throttle. In "low gear," that means roughly 1,500 RPM. The vibration was just about as bad as I expected. (Remember, it only has one stator segment in place...) But, nothing broke. Here's a video. If you look carefully, you can actually see the entire test rig flexing.

But it's clearly generating a trapezoidal back EMF, and it got above the minimum speed I set for a valid result (1,000RPM). This is more so that the kinetic energy of the rotor overwhelms the push-pull of the magnetic forces, giving a smooth speed. Smooth here is a relative term.

The peak of the trapezoidal back EMF is proportional to motor speed, and the constant of proportionality in this case turns out to be 5.7 Volts per 1,000 RPM. Without the flux jacket, it is 5.2 Volts per 1,000 RPM. I expected a bigger difference, but maybe there are return paths I don't know about that are doing the job of the flux jacket already.

More importantly, how does this compare to the prediction? Well...the shape is right. And it's within a factor of two, so I probably didn't really mess up the calculations. But, plain-old math assuming a 1 Tesla airgap magnetic field predicts 9.7 Volts per 1,000RPM with this geometry and number of turns. A linearized simulation with FEMM predicts 9.6 Volts per 1,000RPM.

I have no idea if adding in the rest of the segments will make up the 40% difference, or some of it. Let's say it didn't, for now. Since back EMF tells you most of what you need to know about a motor, these results are very useful. Extrapolating these results, which are for a single winding on a single segment, to the whole motor gives the following as a rough performance estimate:
Operating Voltage: 96V-144V
Top Speed: 4,000RPM @ 144V
Torque Constant: 0.33Nm/A
Peak Torque: 66Nm @ 200A
Peak Power (3-minute): 20kW @ 3,000RPM and 200A
Continuous Power: 10kW @ 3,000RPM and 100A
That puts it squarely in the category of "small EV motor." It's the next step up from the kart motor, I guess. It could run well on inexpensive Kelly brushless controllers. Doubled-up, it would make a nice rear axle motor set, although I'd wind it for direct drive: lower speed, more torque. The catch: you have to build it! And in low quantities, the stator segments are still pretty expensive to produce. Please somebody steal this design, fix it up a bit if you like, and make one like it in large enough quantities to be under $2,000. I'll buy it! Seriously. Let me know when you do. Until then I'll be figuring out how the heck I might assemble the whole thing...


  1. I love this motor! I think it is so cool that I could practically build it, if not economically. Speaking of economics, how much do the stator laminations cost? I took a look at the web site, but couldn't find anything regarding prices. And speaking of the stator segments, how did you reshape to get the trapazoidal shape? One of the things that I did read on the Protolam site was that post lamination machining can change the properties of the material. Maybe this is another reason why the experimental didn't match the simulation?
    Last question: regarding the direct drive you referenced, how would you change the stator windings to set the motor up for direct drive?

    Anyway, really cool and I hope you get a chance to put in a few more segments.

  2. Thanks! I'm not sure how much I would change for a direct-drive motor. Maybe I would make it a little wider or larger diameter, to get more torque per amp. If it displaces the need for a differential, the extra space might be justified.

    The trapezoidal shape comes from the straight edges of magnets and stator segments, as well as the concentrated distribution of coils. If the magnets or stator segments were skewed, or if the coils were distributed or overlapped, the shape would get more sinusoidal.

    I don't want to give away Proto's pricing, but I can tell you it's very reasonable and set up for short-run prototyping, even single motors.

  3. I'm going to have to give Protolam a call.... I want to build one of these.

    As far the the trapezoidal shape, I wasn't clear what I was asking. I was talking about the actual shape of the "H" cores. In the initial pictures of the cores, the end shape of the core is a rectangle however, when you installed it onto the stator, it is shaped like the magnets. I was curious how you did that.

    On another front, did you consider a thrust bearing around the outside of the rotor to maintain the airgap or is that just an extra complication?

    Finally, if I built one (or 4) of these, I would be looking for direct drive, meaning a motor that would be maximized for 1500 rpm, max torque. Suggestions?

  4. Gotcha. Actually, I sheared them all down to progressively smaller sizes myself. Not really the fastest way, if you need to do 18 segments. But if you only need to do one, it turns out to be faster than setting up a machining jig.

    The rotor is stiff enough to maintain the air gap, even without the outer can or spacers. Even under the full force of the magnets, it would only deflect a couple thousanths of an inch at the edge. That's assuming everything is properly aligned and loaded, which so far has not been easy to achieve. I'm debating the idea of using "outrigger" bearings on each side to help. Then, each rotor would have its own full set of radial and thrust bearings.

    For direct drive: The number of turns would probably change, but that's the most flexible design constraint so I wouldn't start there. Other than that, you might consider a slightly wider motor, a larger diameter, or more poles. Any or all of those could increase the torque per amp and lower the speed per volt (with a given number of turns), which is what you want for direct drive.

    I would LOVE to see four of these! But definitely try one first before you spend all the money on an unproven design. :)