Land-bear-shark Update 3: Listening to Charles complain about everything

chorl

I was wrong.

The metal didn’t come “that week“. Or, for that matter, the next, when I was in Singapore and therefore couldn’t do anything about it anyways.  After it all finished, my metal was delayed by a solid month. What was the vendor’s excuse?

I looked up your order and it seems yours and a
couple other orders that were supposed to get
shipped fedex never got printed out and therefore
didn’t get shipped out.

Right. Just like that. After an email asking for an update on the order (to no reply), and 2 more asking for a refund with the last being a ship-it-express-or-negative-feedback ultimatum (still to no response), the metal for LBS finally showed up on the 20th of January, of course shipped ground the week before. I’m sorry, you’re not gaining back the negative feedback point. For future reference, never, ever buy from this guy on eBay. The prices are low, but I could have ordered from Speedy Metals three and a half times over in the same period. Or got it made straight through Simon like 5 times, and not had to deal with excuses and silence. Next time I get anything from eBay, I’m staying on the seller’s ass so hard it’ll cost me the point.

So that’s why there’s not a Melontank at the moment. With IAP having ended, I’m faced with less time to hammer away on it as a whole. Luckily, the 8 miles or snow or so that’s on the ground shouldn’t melt until… I don’t know, July at this rate, so maybe I’ll get to climb a little hill or something soon. Assuming I pick up the pace and actually design the hands-free control – who knows, the first run might be me holding a R/C transmitter.

Alright, so here goes. This is roughly the span of the last week or so, which is when I had to play the wake-up-during-the-daytime game to get things cut.

As usual, we begin with the waterjet puzzle.

Wait – there’s actually more of a story to this. I had to rush-order another plate of 1/4″ aluminum from Speedy Metals (who lived up their name) because I wrote off an entire plate after trusting the OMAX autorouter more than I should. I’ll take this moment to complain some more – namely the autorouter software is the least intelligent thing I’ve ever had the joy of experiencing. I’m not sure what’s so hard about finishing one part before moving onto another, but it seems to revel in making cross-stock express runs right over previously cut features. So, no matter how closely you watch it, inevitably it runs into itself and bumps the piece. As a result, the entire coordinate system of the part becomes shifted mid-cut, and pretty much everything is ruined.

After losing that plate, I taught myself how to manually route files pretty much on the spot. The backup plate was cut without incident as I removed completed parts out of the tank one after another.

Come on OMAX, stop making your interns write the path software already. One closed external profile at a time.

Now that I’ve complained and bitched enough, here’s an elaborate waterjet-cut box.

The plates were cut with a .003″ closer offset distance such that most of the tabs and slots just fell together. However, I seem to have neglected to do that on the first plate (before I destroyed everything), so the bottom side panel in this picture is a really stiff hammer fit.

If you’re also intending to make things fit together on the waterjet, I’d recommend either reducing the tool offset a few thousandths or making your slots a few thousandths bigger and tabs a few thousands smaller, just so they slide together easily. For 1/4″ aluminum, 0.003″ inwards (i.e. total width of any kerf is actually 0.006″ larger) seems to be great.

The box is complete.

Well, minus a whole ton of hardware. I ran out of 4-40 button-head screws, so for now, this will have to do. I also seem to have physically lost the 1/8″ plates I cut out for the triangular things at the end between last week and now.

Since this project has gotten to the point where it can hold its own parts (hey, it is a box), I threw together this poser-shot. I think it’s pretty reflective of what the final vehicle will look like.

course 6

None of my excessively dangerous vehicles are complete without some kind of controls headache, and Melonsharktankbeargryllspigweek is no different.

In updates prior, I clearly showed the Melontrollers being mounted on the control deck. The fact of the matter is, I don’t think they’ll actually make it into version 1. Melontroller is not yet stable enough for me to put it on anything less benign than RazEr…which has already seemingly ate one controller. For once, it wasn’t in a fiery explosion, but the Arduino Mini just…. died. My guess is that a transient made it past the regulator somehow and killed it. It was also never really running a motor that smoothly, nor did I have the current sensors even mounted on the board. There’s also a ton of changes I want to make to the design to make it more noise-tolerant and robust… including ditching the chopped and screwed linear regulator feeding the logic supply.

Pursuant to the delay of Melontroller, I’m probably going to end up using some mini-Kellys and a custom signal interface board. You know, some electronics I can actually handle. There’s several obsolete KBS36051 units hanging out in the Media Lab ever since we switched the model Citycar to DEC boards.

While they’re advertised as “50 amps”, they really limit the phase current to about 20 amps continuously. On a stock 80xx “melon” winding, which is like 2 and a half turns of 4/0 copper busbar, that does absolutely nothing. Even if LBS was able to get up to speed, the motors are just too fast to work well with the current speed reduction – 5:1 on a motor that’s rated for 170 RPM/V and running at 42 volts still gets me over 35mph.

I don’t particularly feel like dying that badly, so I’m electing to rewind the two Melons to something more reasonable. Like, say, 18 turns of double #20 wire, Y-terminated, instead of the stock 8 turns of…. something… that’s Delta terminated.

By some crafty math, this should get me a motor with a RPM/V rating of about 40 (actual math: 170 original Kv * ((8 / 18) turns ratio / sqrt(3) Y-termination factor) ) With the as-designed 5:1 gearing, this will actually be quite tame at 15 miles per hour. The gearing can then be fiddled with. With the Y-termination and higher turn count resulting in increased phase inductance, the Kellys ought to freak out alot less than they did when I tried to run Melonscooter with them.

All that’s left for me to do is to design the interface to toggle the Kelly throttle, reverse, and brake inputs remotely. This will, again, probably be accomplished using some flavor of Arduino.

Kitmotter Display Stand

chorl

A wild KITMOTTER appears.

I received the metal stock, bearings, and magnets for Kitmotter a few days ago, and so have been spending time hogging the 2D fabrication equipment when I’m awake during the day before the shop closes.

I’m building something a little special for Kitmotter 0001 (to follow the 4-digit versioning scheme I’ve seen in some Processing-derived software), something which should keep people from asking the question “Hey Charles, what actually powers your scooter?”. At least, it would provide an illustration to save me 1,000 words per RPM.


oh snap, it’s a motor on a stick.

It’s cute and simple. The Kitmotter is mounted on a box that will hold a controller and a knob. The knob is turned by a curious onlooker, and the motor spins. Not very informative, but it was something I had in mind since the days of FVM when I wanted to use LEDs to indicate the state of the motor and the direction of current in each phase. Because the kitmotter has transparent sides, what operating mechanics there are in it are presented in clear view.

The materials used in the construction are some 6mm and 3mm acrylic for the endcaps and some 1/4″ steel plate scrap for the rotor segments. Oh, and a Xerox Phaser 4400 stator. The rotor plates were waterjet-cut, and the endcaps laser-cut, on-campus as per standard operating procedure (but the unspoiled can go spam Simon for the same services).

Out of the same sheet of 6mm acrylic comes the stand itself. I think I’m going to start removing the paper coating from raw plates before cutting now, because it’s such a pain to take the stuff off afterwards.

Really quickly whipping the box together.

Yeah, this is definitely something I built. You can tell because of the regularly spaced T-nuts and lack of attention such details as “How many spacers did I need again?”. The answer is 2, but I cut only one. Go figure.

The solution was to 3d-print the equivalent double spacer using MaB.

Moving onto the motor itself, the first order of business was to assemble the rotor using the gratuitously oversized magnets. They’re from Applied Magnetics, which seems to be the least expensive place for gratuitously oversized magnets there is so far; at least for common fractional inch sizes. For everything else, there’s Mastercard Supermagnet George.

I must say I’m a fan of the waterjet-cut magnet retainers, since gluing these in was SO EASY AND PAINLESS compared to the martial arts tournament that was putting the DNIR together. The presence of so many magnets in a circle keep the rings aligned with eachother, and the roughened and uneven finish on the inside also helps with magnet bonding.

Great for making sure your motor never sheds a magnet, not great when you discover you somehow managed to put a magnet in backwards. I had to chisel it out and scrape off the remains with a punch.

My usual order of fluff-filled epoxy was used to bond the magnets.

The fully assembled rotor, on a makeshift shaft for concentricity validation (result: Not Very). I guess the X-Y tolerances of the machines sort of add together on assemblies like this. It’s not bad, but there’s a perceivable wobble to the bearings if spun by hand that’s not really visually detectable. More care in assembling the bearing would also fix it, I suppose. And also not spacing your bearing bores with Kapton tape because you thought the laser kerf was slightly smaller than it actually turned out to be.

Here’s a bit of a leap of faith. I ended up deciding to make the machined-shaft version, so I carved the shaft out of 0.750″ Delrin aluminum steel because I only had the raw stock size in steel. I even cut the axial pin hole to retain the stator, but then decided to just epoxy the whole thing.

The stator was wound 26 (+/- 1 ) turns and Y-terminated. On the behest of Shane, I tried terminating the windings in a different fashion. Whereas normally I’d just twist the three wires together and solder them in a giant ball and shove the ball into one of the gaps between windings, this method just runs a loop of wire to connect the three ends together. Same effect, and I think it gives a better symmetry to the motor despite it being a bit more manual work.

The phases themselves are terminated using some 16 gauge speaker wire (because it’s clear!). Here’s a mostly assembled motor. As the design showed, there’s plenty of airgap. Way too much for my tastes, but the ginormous deth-magnets should make up for it.

Once I mashed the last endcap on and bolted the whole thing together, it was time to use the official MITERS Brushless Motor Jogging Utility melonscooter to give a quick powered spin.

Result? Ignoring the severely unbalanced rotor which needed holding down of the box in order to not hover off the table, the whole thing is very smooth and quiet. My only real complaint was the massive sealed greased ball bearings adding substantial drag to the system. The idle current draw was around 2.2 amps just because of those damn bearings – at Melonscooter’s 40 volt happy discharge plateau, that’s easily 60 watts or more being dumped into the two bearings. RazEr’s original motor only drew 1.25 amps no-load. The DNIR, which is a motor 3 times as long and with something like 60% of the internal resistance (due to its longer windings being put into parallel), still only draws a bit under 2 amps no-load.

The result was that after a few minutes of fun with the throttle, the shaft was too hot to hold on to. That probably isn’t too good for the acrylic. Maybe it will run itself in with more operation.

A check with the oscilloscope confirms the RPMs per Volt as (334 electrical Hz / (7 electrical Hz / mechanical Hz)) * 60 sec. / 40 volts = about 70 RPM/V or so. In metricland, this is 0.14 Nm/A (or V*s). Note that this is with sensorless commutation using a controller which is probably playing mind games with the timing. A more reliable measurement can be had using sensored commutation.  Still, not bad for something thrown together effectively in 4 hours or so, like 2 of which were winding.

What’s next for Kitmotter? I’d need to take it apart again to wire up the sensors, then shove the controller and it into a box and call it done. The controller will probably be the shady e-bike controller that I got in Singapore.

Oh, and for total kicks, I made a 3d printed version of the endcap:

This fits the 1″ bearing version with the 3DP two-piece shaft. Kitmotter 0002 will probably be an experiment to verify that design.