I was as excited as anyone when I found out they built practical, working BB-8s for the new Star Wars, but I'm not really "$150 toy" excited, and I can't really even justify spending $50 on a used Sphero to cut open, so I decided to build upon the work that's already been done here on Thingiverse. My goal was to build a BB-8 using only printed and hardware store parts that could roll along the floor and keep it's head upright, and spoiler alert: it finally works! https://www.youtube.com/watch?v=bYBuFEJ_PQc
I made some initial modifications to the existing parts, adding holes for 1.75mm filament in the ball halves so it could snapped open and closed (repeatedly) and in the head to make antennas.
The hardware store parts are 19mm ceramic disc magnets (same as the ones for Google Cardboard, left over from that build), 9 7/16" ball bearings (the biggest they carried), 2 8mm OD/3mm ID/4mm wide roller bearings from old spring-loaded extruders (and technically, a hobby store and not a hardware store originally), and 3 #4 x 1/2" screws.
I experimented with 3 different heights for the magnet within the central "stalk". At it's highest, the whole internal assembly just stuck to the head through the ball. At it's lowest, the magnet wasn't close enough to keep the head from just falling off. I've settled for a position right in the middle of those two, but it might be worth printing a version that puts the magnet a millimeter or two higher or lower. It's also possible that this magnets are just too strong for this purpose.
I then thought there might just be too much friction acting on the head, so I split it into two parts and made some room for a pair of roller bearings inside. The head moves better on the ball now, but you still can't roll the ball without it tipping right off.
In experiments with both versions of the head, I found that my internal mechanism seems to stay upright when tilted with no head around, but when I install the head it just gets tippy. I assumed that this means the head is moving the center of gravity too high, so my last idea was to cut open the bottom of the stalk and fill it with pennies. This seems to have only made the movement more sluggish. It may be time to actually do math or reevaluate my approach entirely
The state of BB at this point can be seen here: https://www.youtube.com/watch?v=2JGKrskvX3U
Version 2 replaces 6 of the 7/16" ball bearings with 5 3/4" ones, and incorporates a single #8 x 1" screw. It's more stable and stays upright better, but it still doesn't really "work". I think version 3 will be larger, to cram even more weight inside and maybe give the mechanism more leverage??
Version 3 was basically the same idea scaled up 50% to fit 12 ounces of fishing weight inside. I also added more detail to the head at this time. It didn't really work as well, so I never uploaded the .stls, but it is in the .blend if you really want to try it.
Version 4 kind of works!!!?!?!? I'd been watching James Bruton's full-scale BB-8 builds, not because i thought any of what he was doing would apply to this and just because he does good work, but it turns out the single-axis approach makes a lot of sense. Printing the ball as a "barrel" with two side caps means the internal mechanism can run on a smooth round track instead of running into the seam where the two halves used to join or the roughness at the higher layers. It doesn't balance as quickly as I would like and it doesn't roll at all on smooth surfaces like wood or tile, but it can be pushed or pulled along on carpet and it can be rolled down a reasonably steep ramp without falling over or the head falling off. https://www.youtube.com/watch?v=gLeMTFUNe4I
I've also added opening panels in the side caps in this version, the "pocket" on one side and an arm on the other.
The non-printed part list for this one is 12 1oz fishing weights, 4 19mm ceramic disc magnets, 4 3/4" ball bearings, and 2 #4 x 1/2" screws. I made spaces for some of the 7/16" ball bearings but found that the extra weight up high didn't help.
I think the next version will have a fixed axle at the center rolling on skate bearings.
Version 5 is good enough! It turns out even the cheapest eBay bearings roll more smoothly than ball bearings running in a 3D printed race that I sanded for maybe 15 minutes. Maybe it also helps to have a fixed, defined axis for everything to rotate around. it actually doesn't work nearly as well not that I've painted it, but I'm still satisfied.
The final non-printed parts list is:
10 1oz fishing weights (pictured now)
6 #4 x 1/2" screws
4 19mm ceramic disc magnets
2 608ZZ bearings
2 M8 25mm bolts
2 matching M8 nuts
The cost, even including all of the ball bearings I ended up not using in this final design, is probably less than $20. (If I put a value on the TIME I put into all the previous designs, maybe a bit higher)