Search This Blog

Sunday, July 25, 2010

Lots of Snapping

Let's just get down to business. Time to wind a stator. I haven't changed the electrical design of the motor since the Design Review . Each tooth gets 15 coils of 17AWG wire, for a total of 270 coils per motor (about 18 meters of copper). My winding pattern was AaABbBCcCAaABbBCcC (google motor winding tables).

1st phase wound. Note the careful packing - it was the only way I could fit 15 turns on each tooth.

It took approximately 5 hours to wind the whole motor. Not bad, considering the gauge and my lack of experience. I have wound motors before, but they were those tiny little brushed kits with like, 28AWG wire and 3 teeth. Anyways: 
You can see the permanent marker (smudges) that I used to mark the teeth numbers.

Success! No shorts and I didn't wind a phase backwards! As I mentioned earlier, these GoBrushless (actually Scorpion, and actually actually Chinese) stators have a unsatisfactory epoxy coating that can result in shorted phases. I could tell that the extra epoxy I added to thin/chipped parts of the coat really helped. If you use these stators, I strongly suggest doing this. Moving on:

See the gap? I had these custom arc magnets made 1mm-ish too thin to allow for gaps between the magnets (it turns out that complete magnetic filling actually hurts performance). The problem is, these magnets really like to snap together, so I had to figure out some way to space them. And not only space them, but space them the exact right amount...hmmm. While I figured that out, I realized I needed some super glue and micro-balloons (already had JB weld) to glue the magnets with, so I put in an order to towerhobbies. While that was coming in, I finished these: 
Sorry for the bad pic quality, I'm still using a cellphone camera.
Much more badass with the MOSFETs.

Also during that time, I took a few days off to try to find a water-jet for the battery box sides. The Aero/Astro one will be down until at least Tuesday, and I'd have to attend an orientation during a time I can't make in order to use the HobbyShop one (and pay a membership fee...). Yeah, this sucked. It put me behind schedule on the battery box. I could have used the Edgerton/Archie laser to cut a template, which I could use to mark my aluminum, which I'd cut out on the band saw and spend an hour with the belt sander...but it turns out that I still couldn't use the laser for various computer glitchy reasons. *Sigh... oh well. So I was out a water-jet and a laser. 

Luckily the towerhobbies order came in. I found this, which allows you to print out a diagram of your magnet placement. Of course, I spent an hour trying to get a 1:1 copy to come out of the stupid Athena printers, but I eventually got one close enough.

So I figured out the placement issues, but what about the magnet-snap issues? Super glue and a smart placing pattern to the rescue. I placed/glued every other magnet first. The super glue is great because it wicks right in underneath the magnets. Credit goes to Charles for giving me that idea.

Anyways, I still had the problem of putting in the other magnets in such a fashion that they wouldn't snap together (the nice diagram I printed out couldn't help me here). Through trial and error, I eventually figured out exactly how much spacing I needed between magnets...

...and it turns out this thin wire was perfect. After I got the middle magnet to this point, I would carefully remove one of the wires, put in a drop of super glue, let it set for a few seconds, then do the same for the other wire. It worked great:

Now I just needed to fill the gaps between the magnets with epoxy to make sure they don't pop out. This is where the microballoons come in.

My setup.
You can mix microballoons (or silica, or a number of other things) with epoxy to make it thicker (or give it other properties). In this case, I mixed some with JB weld until I got a consistency that wouldn't run. Then I went about cramming it between magnets as best I could. The alcohol was for clean up (it does a nice job of dissolving/cleaning up epoxy); I soaked a couple paper towels and ran them around the inside of the magnet ring to wipe up any excess epoxy. You can also use MEK, Acetone, and various other chemicals to clean up epoxy. Anyways, results:

It worked great! The dark patches between the magnets in the above pictures are the iron/steel filings from the JB Weld (JB Weld actually has metal in it). The magnets were strong enough to partially suck them out of solution! While that dried, I finished up wiring the stator:

Note the small twist on the bottom left: I terminated this motor in "Wye", as opposed to "delta".

Oh, so THAT's what the through-holes in the CAD drawings of his hubs are for...Yes, instead of running the wires through the middle of the hub (which is filled by axle in my case), or running them under a bearing via a flat spot in the hub (which sucks because you end up with 3/4 of a bearing surface), I ran the wires like this. But what's the 4th hole for? You'll see.

I took a break to work on some of the mundane stuff, like making alignment pins:

Yes, I could have bought them pre-made, but that costs a fortune. Instead, I bought a couple rods of precision ground 1/8", O1 tool steel, cut them down with a dremel, and then beveled the ends. I made these all about a 1/4" to long, but oh well. 

Then I got the laser working!!

Cutting the hall-effect sensor board.
Cutting some spacers.
Finished product: laser cut acrylic spacers and hall-board.
The three lines etched in the hall-board are the angle lines for hall-sensor placement. Also, see the notch in the hall-board? The next thing to do was to drill a cross hole in the bottom of the hall-board for a screw that would tighten it onto the hub. Much easier said than done:

So I went back, redrew the part with thicker edges, and tried cutting another one:

Can you tell I'm a noob with the laser cutter? This is what happens when you pick the wrong color mapping (this laser works by assigning colors to an AutoCad file that translate into various cutting powers). 

Take 2: 

Much better, I even got the screw in without breaking it. 

Epoxied hall effect sensors.

Time for some sensored motor theory: To figure out the angle needed between hall effect sensors, you can follow this tutorial , or you can just do this: (360/# poles) / 3. Credit goes to Shane for that one. That get's you the number of mechanical degrees between two sensors, 12 in my case. Apparently, you can also just place them between the stator teeth. It's all mostly equivalent electrically (I'll let you work out the geometry), just differently mechanically.

Now I had all the major components made to start final assembly!

Everything laid out.
Of course I break off the hall sensors when wiring them...

RE-epoxying the hall sensors.

Oh, so THAT's what the 4th hole in the hub is for...sensor wires. I just fed the wires through, jammed the board up against the coils, tightened the screw...and SNAP. Crap. See the crack in the above pic? I officially disliked acrylic at this point (hating it came a few minutes later).

Lowering in the stator.
I used a vise and a drill press to slowly and carefully lower the stator into the rotor. I didn't want them snapping together and destroying each other. Next, I put the snap ring bearing retainers on, put the bearings on, put the spacers on, and put the hub caps on. Then after hammering in 10 alignment pins and putting in 9 screws:
TA-DA! The 10th screw is missing because it started to cross-thread for some unknown reason. 

I really hate acrylic; every single hole in the spacers cracked, either from drilling them out (I didn't trust the laser to cut perfect holes, so I undersized them), or from assembly. I'm just going to have to suck it up and make polycarbonate spacers and hall-boards. Another thing to note from this pic: the screws are too long...or I didn't thread enough of the holes in the steel...either way, I'm getting shorter screws. Oh, and as noted previously, the alignment pins are too long. I knew I would have to rebuild the motor anyways, so I didn't bother grinding them shorter. I also didn't bother with the o-ring seals.

The 1/16" polycarb inner-side bearing retainer plate thing isn't on yet (I haven't made them yet...need water-jet or 5 hours with a mill).

Despite all the minor problems, it turns incredibly smoothly.
Time for testing! Shane helped me with this part.

Here's the mess when we thought we could use a radio control Tx/Rx.
The drill batteries supplied 36V, the motor controller is a Turnigy 100A sensorless (no sensors for this test), and the scope is for measuring pulse width, i.e. speed. We were originally going to use a radio control setup for control, but it turns out all of the R/C stuff at the Edgerton Center is broken (and it's not user error...I've been driving/flying R/C stuff for years). Time to make a bigger mess:

The power supply is providing 5V for the R/C input and the function generator is providing the would-be-receiver output (pulse) signal. Changing the duty cycle is equivalent to throttle. Go Shane for figuring all that out.

It worked! Yay! Here are some specs:

2 phase resistance: 0.269Ohms (very consistent over all combinations ...within 1%)
kV: 50.5 rpm/volt
kT: 0.189 Nm/A

So at 10A per motor, I'm looking at a total of 8 Nm of torque, which is plenty. FEMM calculated about 2.3Nm per motor...not bad. This is the crazy part: at 48V, with 4in diameter wheels, this board will be going close to 30mph! Uhh...maybe I should rename it "DeathBoard".

Now I just have to make 3.5 more hub motors, including a rebuild of this one. Oh boy. LOTS of good news, though! I have most of the parts made for the other motors already, all of the parts I designed worked perfectly (I only had materials issues), and my torque estimates were right on the money. In other words, I think I did a good job engineering the motors. The only thing off target was my speed goal...I was aiming for about 15mph. I'm still not sure how I was off by a factor of 2.

Here's another pic:
See the ploycarb flakes in the bearing pocket? That's from the bearing race rubbing. I'll add that to the list of stuff to fix.
Next week: More motor building and (hopefully) battery box construction.


  1. I ordered a brand new $23 TX/RX from Hobby King so we shouldn't have to use three separate piece of expensive electronic equipment to generate and measure a valid servo pulse anymore...