It’s legi…. wait… nevermind.

chorl1 Comment

There seems to be a disturbing trend in all my recent projects of everything progressing smoothly and going right, then the whole project grinding to a halt at the last possible second before completion and being wholly unrecoverable without spending an additional hefty sum of money. Case in point: TB4.5-SP1 for Motorama had essentially half the robot become crippled the day before I had to leave for the tournament.

New case in point: Snuffles Reloaded half-existed for about 30 seconds before the motor controller failed. The failure is undiagnosed at the moment, but probably revolves around blowing a voltage regulator which then shot the controller logic with 35 volts.

Here’s some build pics from over the past few weeks, and the (95%) complete vehicle, along with the half-existance video.

As the mechanicals were nearing completion, I started playing with the electronics. The R/C airplane controller requires a 1 to 2 millisecond long pulse spaced every 20 milliseconds, which is a hobby industry standard for control signals. However, the electric bike throttle I was going to use puts out an analog voltage from 0.8 to 4.2 volts. This is, again, a different industry standard control signal.

I decided to use a microcontroller to translate between the two instead of the awesome ghettomongous discrete-part boardamathinger that I built for the first scooter. A bit of messing around with blinkenlichten on a MITERS STK500 and I was in business.

After playing with the analog-digital converter, I made the output LEDs act as a throttle display meter of sorts, for kicks. Full throttle, all LEDs lit, and they go out in sequence as I let off the throttle. I might actually implement this using a LED bargraph chip in the future.

Next up was assembling the eight giant lithium-polymer cells without killing myself or burning down buildings.

To solder epic batteries, you need an epic soldering iron. Unfortunately, I couldn’t find the epic soldering iron I used at the Media Lab to assemble A123 cells, which are even more epic than my 4AH lipos. So if you do not have an epic soldering iron, you need an epic soldering tip.

A scavenged rod of copper and some lathe time later, and I had an epic soldering tip which fit a Radioshack 40 watt soldering iron, which I do have. The whole thing is 2.5″ long and is .400″ in diameter at the larger half.

The brass one was for testing and practice, since a rod of copper is actually pretty pricey these days to just fuck around with.

The bottom battery pack completed after some careful iron maneuvering. The epic tip made the whole operation touch-and-go, exactly what you want for soldering batteries. You never want to park the iron on a battery cell for more than a second or two.

This was a bit of a risky operation since I was laying the cells face-to-face to conserve wiring volume. Normally, batteries like this are stacked and the cell tabs folded over one another. One wrong move with the massive solid copper tip and I was probably looking at replacing a cell.

After each joint was made (and its balancer lead installed), it was covered in electrical tape. When all the joints were completed, I slammed the whole thing in a tube of giant heatshrink and parked the heat gun over it.

Giant heatshrink should be a primary structural fastener. When it starts tightening down, everything inside sort of scrambles for the lowest volume configuration, and the end result is a very neat package of parts. The pack was embedded into its mount with some also primary-structural double-sided tape.

This pack constitutes cells number 5 through 8. The balancer lead is a standard 3-pin R/C servo plug, which serves the interconnects between cells 5 and 6, 6 and 7, and 7 and 8.

And the test assembly. The connectors fit into the LASER-cut acrylic endcaps as they should, and some CA glue retains them.

Building the internal electronics bay was more interesting, since I had to fit batteries, large power wiring, a controller, a switch, the charging port, and all associated connectors inside. I cut out a rectangular piece of aluminum as the substrate (way to go, conductive mounting surface?). Originally, it was going to be flanged and shaped with a sheet metal machine to accommodate the parts, but I decided to not get fancy and just mount everything with double-sided tape or epoxy.

The same procedure of soldering and wrapping was done to the 4 internal cells (#1 through #4), with the exception that their balancer leads went straight to the Convenient DB-9 Connector of Cell Balancingâ„¢. An 8 cell pack requires 9 pins to be fully tapped. Guess what has 9 pins?

Three of the DB9 pins went to the rear of this pack, where they met with an R/C servo pigtail which connected the back half of the pack to the balancing port.  Two large power wires also connected to the bottom battery pack, one of which is the 0v  (ground) line, and the other an interconnect between cell 4 and 5.

After everything was assembled, it was time for a test run. Verdict: It moves.

Some shoving and… it fits! To install the thing, I had to take off the folding hinge, cram the assembly through to the back side about halfway, insert the folding hinge nut plate, then slide it in the rest of the way. The nutplate sat snugly above the controller, but not enough as to pinch wiring. This precision engineering part of the build came out (went in?) great.

Before I fitted the internal electronics, I threw together a source of 5 volts for the motor controller. It is an opto-isolated controller, and so needs separate logic and power rails. This was a simple 7805 regulator jammed in the empty space between the controller and switch.

Here’s an assembled-on-the-table test run video (.MOV, 4.8 megs) using the creepy custom throttle interface device. The mechanical noise is from the completely unbolted and unclamped motor and chassis resonating on the table.

Unfortunately, this regulator would ultimately cost me all the work for the past few nights and a good bit of money. Protip: A linear regulator cannot drop such a huge percentage of its input voltage and output any appreciable current. I was most likely hitting the top end rating of the 7805, around 35 volts, and expecting it to output a solid 5 volts with at least 100 or more milliamps of current. Shortly after the video was taken, some things went pop.

My best guess is that the controller logic board was hit with the full 30+ volts of the battery. A switching regulator, or even two stages of linear regulators (inefficient, but hey), even mounted externally, would have prevented this disaster.

Anyways, here’s a pic of the almost-running vehicle.

The ’empty weight’ is probably around 13 or 14 pounds. Yes, I had to remove the rear brake in order to pass the motor cables – perhaps it would have been better to route them externally. But who needs brakes anyway?!

Regardless, there’s 29.6 volts of 4AH lithium polymer cells, a (former?) 100 amp motor controller, and a very chunky brushless motor shoved into the space of a Razor scooter. I think it’s pretty damn awesome just for that.

I will need to get a new motor controller and devise a new solution to get a stable 5 volts out from the battery pack before the vehicle will run. Seeing as how this will easily cost over $100, it might have to wait a bit. Possibly a long time – we’ll have to see.

Here’s a closer shot of the undercarriage, which houses all the interesting bits.

In the mean time, finals! Bot on, folks, while I attend to these…uhh, pressing matters.

Pop Quiz 2 Update 2

chorl

The calm before the storm has begun.

After two weeks of hosing, I now have a week of reprieve before the wall of finals hits… hopefully not too hard.

So naturally I’ll take the opportunity to catch up on two weeks of building. First off is Pop Quiz, which I actually have a fair percentage of the parts for.

From United Hobbies are the replacement internal parts. After some idea juggling, I settled on some 460mAh packs rated for 15C discharge. This of course gets me an incredible 6 amps of maximum current, but given my intention to run everything in the bot mildly, should not get in the way.

These packs are 3S each, and I will remove one cell from each pack, then series the remaining to yield 4S x 460mAh.

The new 10A controller is unbelievably small – MUCH smaller than Pop Quiz’s current 10A controller by far. It has one SO-8 FET per leg of the motor driver bridge.

Still missing is the micro-receiver and drive ESCs. These parts are contingent on whether or not I can shove the old ones into the new design.

PQ2 drive motors! This is a side-by-side comparison of the stock Mabuchi slot car motor, and the neodymium-tuned one next to it (On the right). The “freewheeling noise” of the “enhanced” motor is substantially lower. According to my R/C meter, it draws an obnoxious .4 amps no-load, but oddly enough, runs smoother and with significantly less brush arc than the stocker. I wish I had some real equipment such that I can collect real data instead of jamming my finger into the rotor to see how easily it stops (Alot harder to stop the new motor, by the way!)

SDP-SI is out of stock on the exact size gear I need to fit the new drivetrain. Go figure – I’ll keep bugging them until they are restocked.

With those little details on hold, it was time to work on the new motor.

…but alas, disaster strikes even before starting.

After popping the stators out of the motors I intended for the bot, I discovered they are in fact 3mm thick, not 4mm like I had visually inspected (Note that these aren’t the “flat motors” from the last update, but rather some HDD motors I had bought in 2006)

This was quite bad, as I had sized the design for a 4mm tall stator. This isn’t to say 3mm won’t work, since small fits in place of large, but would cost me some power as well as style points.

What ensued was a night of terror levied upon a box of old hard drives sitting around MITERS. I went through one of every unique make to see if they had usable spindle motors. There were only 4 brands and models in total – the rest were duplicates.

(MITERS obtained this after some server cluster on campus dumped their old hardware)

I hate new stuff. I really do. Because they ALL LOOK AND FUNCTION ALIKE on the inside. Not a single 4mm, 12 pole motor was found out of four drives. The only 4mm stator was 9-poled and 20mm in outer diameter, totally unworkable. Of course, going through more was only going to be redundant, so I called it a night and just worked on the motor frame itself.

I have some older (1990s) hard drives hiding under my bed that might prove more useful. Old things tend to have bigger motors. Or I could just harvest some plates from one stator and transfer them over.

Result of the Night of Terror. The other MITERers were, of course, glad to act as parts vultures, stripping the discs, magnets, RW heads, and various random bearings and spacers. I wonder how many cents I can get if I recycle the casings.

Time to get down to business. Behold, the sketchiest sawing setup that has ever existed.

I needed a chunk off this 8 foot long, 1.6″ diameter steel pipe. There wasn’t really an easy way to handle it besides propping it up on the work table and using the abrasive saw.

Problem was, there wasn’t really a sawhorse or structure that could easily fit in the space provided. The solution was to prop it up with a wooden board and just hope nothing moves.

Nothing moved, but the effort proved fruitless as the inside of the steel pipe, when cleaned of rust, was around 1.3″ diameter. I need a 1.295″ ID on the magnet ring, and would rather trim outwards from a smaller pipe.

D’oh.

So the next thing to do was to start on the aluminum bits. Here’s the completed motor base. Or, rather, here’s a completed dummy motor base. It was a good practice run to get the quirks of manufacturing out, but ultimately there were some inaccuracies.

The MITERS lathe is well-known for being inconsistently inconsistent (consistently inconsistent would mean I’d be able to predict and offset appropriately to account for its behavior), and I had all sorts of weird things happen again, like the tailstock that points to any one of 5 constellations depending on where it is on the bed, how hard I crank the camlock, what bit I use, how I mount it in the chuck, how far out the tailstock ram is, and the phase of sunspots. And again, the toolpost proved troublesome – it’s awfully flexible for being a block of metal, and some times flexes slightly on cuts, which throws off the dimensions.

I might need to hit up one of the student shops for this application, which requires a bit of precision.

Assembled (with a stock ganked HDD stator). The stator is supposed be flush with the top of the stepped nub, but of course it’s 1.2mm too short to do so. This is actually not as bad as I make it out to be. The base, as it is, will make a good backup part, since there’s only a small amount of error on the OD and the depth of the center hole.

Overall, the test assembly confirms some of the hopes I had about fitting the motor can and maintaining space for the windings. There’s a fair amount of space to wind with.

More work to come!