Updated Aug 10, 2020 including parts list and model of axle assembly. Note that I have put 150-200 miles on this board at this point.
Updated Sep 8, 2020: Fixing a 5W power leak through the footpad and removing the power meter boosted efficiency almost 80% to 10 wh/km; very noticable improvement in the ride, especially at the lower voltages! Also the rear top plate has been replaced to match the front so the wheel no longer looks quite the same. With these changes it's starting to feel good for short sprints on easy off-road (gravel, dense grass, bark mulch, etc.) The power meter was very useful so I plan to add it back and measure the result in hopes of finding a better way to display the data.
Updated Sep 10, 2010: Over 500 miles on this board. (451 recorded miles, but many rides not recorded.)
During COVID-19 quarantine in 2020 I decided to build myself a one-wheeled skateboard, aka self-balancing scooter, aka funwheel. The machine itself is not my ultimate goal, it is only one step towards my goal, which I will not reveal here. When I started, I had no CAD skill, and no desire to overcome what I felt was a large learning curve to get that CAD skill. Instead, I wanted to hone the ability to build a working electric powered personal transporation machine, and a funwheel seemed like a great way to do that. It was! Objective achieved! This build was a complete in-place build, the construction started by mounting the motor into the frame with nothing more than a pair of matched holes, and proceeded to iterate until completion. It was not the most efficient way to get a funwheel, but my goal was to learn what problems I would need to overcome, and that certainly happened. As a follow-up to building the machine, I decided it would be a good idea to learn CAD after all, and so I retroactively modeled my build using Autodesk Fusion 360. The result of those renders are what you see here.
Would I do it twice like this? Overall yes, the result has been a very ridable, very fun machine. Generally this was a fun build, and I wouldn't change anything if someone set out with the same goals I had. I learned a lot and so will you. However, for a great Funwheel, if there's one thing I would do differently, it would be to have a motor mount that attaches to the frame, rather than mounting the motor right into the frame. This is a tough call because having the motor mounting straight into the frame is so simple and easy, but complicates tire changes. Also, I would use different batteries. The Samsung 25R was great for learning, and the 20A per cell output rating (60A combined in this 3p configuration) means that I can lose a cell (or two!) and still limp home. Normally that would not be desirable, because you lose capacity for this, but I had some pack challenges along the way and arguably the safety of the Samsung 25R saved my board and my home from fire due to the challenges I encountered. At one point early in the build there was a twisting motion in the board that was being transmitted into the battery pack, causing a pack fracture that eventually resulted in a non-catastrophic meltdown; the 25Rs allowed the board to limp on. Good? Bad? Hard to say. However, with battery pack experience and a bit more clearance for the pack itself (so that the pack does not experience any twisting) a better choice would be Samsung 30Q, or Sony VT6A, among others. More power density allows more range. A 60A continuous discharge is really much more than is needed, I feel that even 30A continuous discharge rating would be adequate. Stiffening of the bays with a battery box, wrapping the packs with a hard shell such as fiberglass, strengthening the pack with epoxy like Tesla does it, building a stiffer frame from the start, using thinner material for the bay floors, or increasing the thickness of the bays slightly would all relieve this pack twisting problem. I ended up increasing stiffness by using a wider board, stronger fasteners with much more holding power, and adding 1/16" of clearance by putting washers between the top plate and the frame. Early on I overdischarged my packs due to a failure to correctly configure my voltage cut-off and I ended up having a small fire that ocurred as I was putting the cells on to charge. I learned my lesson, and now I do not discharge below 3.0V. Some would consider even that to be too low.
Here is a video of me riding it. Here is another video, later revision of the build. It weighs 10.7kg/23lbs and I get about 10Wh/km, or 16Wh/mi (I weigh about 200lbs). The initial build was only 20Wh/km, due to inefficient footpad wiring. It isn't as capable as I want yet, but with some motor tuning and an additional voltage boost (up to 14s/50.4V) it should get there. Currently speed is not an issue, I am easily attaining 20mp/h (30km/h) on flat ground. Steep hills are a challenge when the battery is below 40% because of a low-voltage glitch that causes motor ticks (note! This was later fixed by decreasing motor observer gain); this is not a real issue for general ridability. Additional enhancements will round out the board's usability:
There is a spreadsheet with various tuning options. The Trampa IMU appears to be pre-filtered, and so due to bad interactions between the Trampa filter and the Balance filter, I needed to drop the accelerometer confidence decay down to zero. However, I was late in tuning a good value for Mahoney Kp, and so tuning here is still ongoing. The board rides really well, but improvements are still possible especially for handling rough road texture (like on a boardwalk or sidewalks with tightly spaced relief cracks.) * VESC Tool & firmware: 2.06
From PHUB188 vendor page, with annotations.
Pay attention to this graph! The shape of the power/efficiency curve can be felt in riding. If you wish to avoid nosedives, getting familiar with this curve would be good.
Early on I solved the problem of the motor turning in the axle mounting hole by putting an ultra-thin 10mm collet wrench to work holding the axle. This eventually stopped working when the wrench started to round out the axle. The axle metal is quite soft, so I decided that drilling through it and the mounting nut together was a good solution. Using a guiding jig and successively larger drill bits, I was able to grow the diameter of the hole in-place to eventually get to a nice tight-fitting and perfectly aligned hole for a hitch pin. This has worked really well and remains the way on my board today.
The fang wheel is not as depicted. Instead, I found this wheel at the local home depot, named a "Richelieu Hardware 1-9/16 in. 20 kg General-Duty Rubber Rigid Casters":
Early wiring pic:
Later wiring pic; both packs are fused, the main power cut was moved from the front to under the aluminum angle, main wiring simplified, footpad wiring is in, among other changes:
At the time I first rode a Onewheel, my wheel was not well tuned and I experienced a big revelation in how good the ride could be. The Pint ride was luxurious, confident, and so powerful. I am still in awe of how smooth the road to grass transition is. And the stability when sprinting off road is just phenomenal. I was inspired to see if my board's tune could get closer to that Pint experience.
A round of tuning did get to a much better ride. My Funwheel is a blast, but it is not as strong as that Pint that I tried. This remains now a community challenge, to see if we can match and perhaps even exceed the power and smoothness of the Future Motion product. I do think that a Future Motion motor mated to a really solid VESC would be a fun machine. Probably not quite as buttery smooth as the Future Motion controller, for now. There will probably be a niche for people who want to breathe new life into old FM boards, experiment with larger or different motors, or really take Onewheel customization to the next level with trick footpads and other experiements in board control.
I'm happy to talk about my project. Hit me up at firstname.lastname@example.org !