after a few more extended and “spirited” runs at 120V, me and a couple other folks couldn’t help but notice the sweet smell of toasty old motor insulation. The motor does get rather warm after longer runs, and will stink up the hallway as its insulation offgasses. It was still working fine, but in the spirit of the impending teardown and reassembly, I popped the endcaps off to have a look inside.
As a refresher, here are the
Treadmill Motor Specs
Kv: 20 RPM/VNominal Voltage: 90V
Rated Power: 1.5 HP / 1.12 kW @ 2600 RPM
Weight: 12 lb / 5.4 kg
The rated power is just what’s stamped on the tin. I haven’t yet done the obvious thing, and metered this vehicle to see how much current it draws when flooring it from a standstill, or to get an idea for overall efficiency. Perhaps it’s time for a garage run. Since we’re over the 60V limit of the Turnigy 130A wattmeter and power analyzer, Bayley suggested using a kill-a-watt.
A pair of 12″ long 1/4-12 screws clamp the two endplates together. The end of one was well ground over from sliding around on the night of the first snowfall. I cleaned up the threads with a small file and installed a fresh nut. I really should chop off that protruding bit of motor shaft, as the whole thing does stick out quite a bit.
The commutator looks healthy, no major scarring. In the background is a nice big wad of PC-7 or similar epoxy gunk globbed over the windings, presumably used to balance the rotor.
Here’s a brush. This motor seems to be in inches, and the brush was about 1″ x 1/4″. I did observe some pitting on the trailing edge of both brushes.
The wire leading to the brush block was silvered, indicating that it had gotten rather toasty at some point. I’m not sure how hard this motor was run before it was pressed (clamped) into brems-duty.
We’ve got plenty of brush left, so no worries there.
As warm as the motor gets, neither the batteries nor the controller seem to care. And as much as I’d like to label the motor the weak link, it seems to be holding up fine in spite of the abuse. Murmurings of “brushless bremschopper” dot the MITERSphere, but if that means building a three phase bridge of massive Infineon IGBT modules (that’s 6 bricks), we may need to move on to a bigger vehicle. Anyway, we’ll have to thoroughly work this motor to its limits first.
Power Brick Abuse
Since we’re now at 120V DC (sweet jesus, be careful with that) pack voltage, we’re in the range where it’s valid to use power supply bricks as cheap DC/DC converters. These easily identifiable black plastic bricks are often found in techno-trash piles and dumpsters, and come in a variety of useful output voltages such as 5, 12, 18, 32.. etc. I and others have verified that many will work just fine off of DC. They usually “turn on” around 80-100VDC, and I’ve found one that turns on at about 60V. Usually the ones that will work on DC are the ones that go straight into a rectifier into a switching converter, instead of stepping down through a bulky 60Hz transformer first.
for more information, check my experimental guide: Will This Wall Wart Work On DC? (brought to you by MS paint)
The easy (and potentially destructive) way to test is to apply DC voltage to see if it pops or not. If the brick has got a transformer right on the input, it’ll pop because the transformer primary looks like a short circuit to DC voltage. MITERS has Sorenson, a crufty old SCR based power supply that will do up to 200V at 20A, which is very convenient for testing these bricks because it has a built in ammeter. The ammeter makes it easy to see if the brick is drawing more current than it should be.
I’m using a 12V output brick to power both the arduino-logic-gate-drive-board and a big old contactor. It claims it can do a cool 5A continuous. To use it, I made an awful adapter which should never be made:
At least the contacts are mostly covered and you can’t plug it right into the wall. “The wall” in this case, has been replaced by a whopping 36S4P of Quality American LiFePO4 Buttery.
Here’s a quick overview of the current configuration of electrical subsystems:
The Arduino Micro R3 that previously commanded Brems-Troller has this awkward startup routine where it jiggles all the pins around and fades an LED. Unfortunately, this meant turning the big brick on all the way for a second or two during bootup. My workaround was to just wait a few seconds after plugging in the Dedicated Logic LiPo (DLL) before plugging in the big pack that feeds the motor. That works fine for me, but I can imagine
some dumb froshling any reasonable person trying to plug it in and having it forcibly slam into their friend or a wall or something. The microcontroller on that flavor of Arduino is the Atmel 32u4, and while I’m sure there’s some way to make it stop doing that awkward bootup routine, it made much more sense to replace it with the Venerable And Trustworthy Chinese Arduino Nano Knockoff. It has an identical pinout (minus four serial pins or something – not used on this board anyway) and uses the Atmel 328p instead of the 32u4. The code ported over just fine.
Much Improved Buttery Storage, dropped handlebars, dropped seat
The lower handlebars and seat were implemented to get the rider… lower. I did this to reduce the distance one might fall in the event of lowsiding, reasoning that if you’re going to fall, being lower to the ground is probably safer.
An unfortunate consequence of this is that it shortens the rider’s profile and makes them less visible on the road. I think it also makes the vehicle look somewhat less silly, which means it may be a change in the wrong direction. As a result, the next set of handlebars will almost certainly be ape hangers.
The battery tray is the only legitimate upgrade in this category. It is designed to fit the three 40v packs in the triangle of the frame. I had to make the two lower packs straddle the down tube in order to make room for the upper pack. I used a blowtorch to heat and bend some 5/16″ “strengthened” acrylic sheet into the shape that you see here. Acrylic sheets are usually quite brittle, so I avoid them, but this stuff is much tougher to break. It is a snug fit over the down tube, and it’s held in by screws to the M5 braze-on (braze in?) water bottle holder bosses. Big velcro straps hold the packs snugly in place.
it’s arguable that with the addition of more going power, there is a need for more stopping power. Well maybe that’s less arguable here in Cambridgetown with cars and curbs and stuff. That one bicycle rim brake up front is getting a bit sketchy for my liking at 30 mph. In response, I rustled a brake disc and caliper out of a pile of parts in N52, presumably leftovers from 2.00gokart.
Now, every time I pull a vehicle to a stop from 30, 60 whatever, I think “dang, that’s a crap ton of energy I’m dumping right there!” On a regular bicycle, yeah, whatever. The weight is only yours and the bike’s. It’s not asking that much of your legs to accelerate it again. But on a bike with pounds and pounds of motor+battery or engine+gas tank, it takes proportionally more energy to accelerate it again.
Regen has always been out of the question on this vehicle due to the rear wheel freewheel existing. I suspect that locking up that freewheel and redesigning the transmission and controller to be regen capable would not result in significant efficiency gain under normal driving conditions. Why? Because the motor has pretty significant friction, or “zero output power draw.” Leaving the motor connected for the whole time instead of letting it freewheel would be like driving around with the brakes pulled in a bit. Freewheeling reduces the rolling resistance to just rolling resistance, not backdriving the whole transmission.
If I ever built some big(ger) derpy electric vehicle, I’d give it the ability to regen, and also a manual clutch. For city driving where speeds are fairly low and there’s a good deal of stop-n-start, it makes sense to floor it, cruise (pull the clutch in), then regen to soak back up some of that cruising energy when coming to a stop.
For now, we just need more stop. We’ll think about soaking up extra juice later.
First up was machining a metrick’d out disc adapter. The stock scooter freewheel/hub assembly that I’m using has a dinky band brake drum that threads on through some M20something x 1 threads. I yanked it off with a vise and some Elbow Grease. I turned a backer for the disc out of a circular chunk of aluminum plate. The MITERS Clausing 6918 lathe is in inches, so I did the internal threads with the threading gears set at 26 TPI, close enough to the 25.4 needed for the real (metric) deal.
I use vise parallels to get disc shaped workpieces like this parallel with the face of the chuck.
After drilling and tapping the hole pattern to accept the disc, I stuck it on the wheel assembly. The disc is held on to the backer with 6 M3 socket head cap screws, and the backer is screwed onto the hub in a right handed fashion. This means the disc brake only works in one direction. In the other it unscrews. I only had one choice since the freewheeling ratchet already dictated rotation direction of this hub assembly.
The observant may note that I changed the tire from a knobbly one to a smoother one. Again, thanks to 2.00gokart there is now an abundance of silly vehicle flotsam in the vicinity of MITERS. The knobbly 12×1.25″ tire was kind of noisy on the street in the same way that a huge old truck set up for mudding is noisy when it’s doing 65 on the freeway. Next time it snows/ices, I’ll probably be switching back to the knobbly one, or making a studded one.
I ended up having to thin the backer down by a few millimeters to get the disc to clear the left chainstay of the frame. I spun it and added cut up pieces of soda can as shims between the backer and the disc until the gyrating wonk of the disc with respect to the frame cleared up. After getting the whole thing to fit, I scrubbed the left dropout and neighboring regions paint-free with a wire wheel in anticipation of welding on a bracket for the caliper.
I shaped templates out of MDF until the caliper sat where I wanted it to, then cut out a piece of 3/16 steel to match one of the templates.
I firmly fixed the mounting bracket to the frame with a C-clamp, then attached the caliper to check the fit. I jiggled it until it was right, then headed upstairs to lay down some hot metal with Dane, who was working on his flying nimbus balancing hoverboard.
I dropped a big old loogie of filler rod on there with the TIG. Disgraceful, but at least it’s kind of aligned right. And You Know there’s nothing like a little bit of pink glitter nail polish to distract welding critics. It keeps the rust out too.
but that was only the start of the disc brake atrocities. I found that neither of the pads rubbed, but some radial misalignment of the disc did touch the armpit of the caliper in one spot. The bracket wasn’t conducive to remachining, so I took a little bit off the radius of the disc with the Trusty MITERS Harbor Freight Whangle Grinder. The pads still make full contact in the right spot, so it’s no big deal.
The Absurd Headlight consists of a $10 100W LED module, a 36V power brick, and an SPST switch. A similar setup was present in a previous video in the form of a flashlight made by Mike. Update your blog, Mike!
The LED module is of the COB variety, meaning it contains an array of LEDs in series and parallel. It’s probably supposed to be used with a nice current controlled supply, but the 36V 2500mA power brick seems to drive it just fine without either the brick or the LED module getting sad. Since it probably has a luminous efficiency of less than 10%, it’s basically a 100W power resistor that puts out useful light. To prevent it from overheating, I attached it to a CPU heatsink with some wood screws.
A 3D-printed thing holds the headlight to the handlebars.
Killswitch of Safety feat. Big Ole Contactor.
not much to report here. I printed out little boxes to hold the kill switch and to cover the contactor so there’s nothing about 120VDC out in public.
I struggle with code
While sleepily trying to upload some new, untested code, I initiated the not-yet-seen failure mode of Fully On. This led to an epic burnout and bremschopper was hurled across the room.
Fortunately, I hit the newly installed kill switch before the handlebars disappeared from my hands. hey, at least THE KILL SWITCH WORKS. No one was hurt, but the Arduino pin controlling the gate driver seemed to be dead. Very strange. The pin floated high after the incident, but since the UCC37321 gate driver has an inverted input, the gate of the IGBT was held low. I replaced the TQFP Atmega328p on another (totally dead) Arduino Micro and replaced the one with the single dead pin. I’m still not sure why only one pin was fried there, especially considering that the gate driver it controlled tested totally fine.
I gave up on rewriting/debugging my own code to perform a running mean for throttle smoothing after remembering that I had the library on another computer back at home. I grabbed that and got it all working. Future Testing was done with the rear wheel off the ground.
Here is the current state of the brems.
Mars lovingly donated a moped seat to the cause. It used to say PUCH on the back, but parts of the P and H are worn off so it looks more like FUCK.