Been also doing some work on redesigning the double stage drive for the inrunners, and I can definitely compact it down from the original. Did a lot of engineering math behind it, so I’m quite confident it will test well. More on that once it’s finalized. I do still need to finish my involute internal gear generator which is quite tedious, then model what gear widths fit for the preferred hanger width. Then input that into my Mathcad calculator to finalize the gear modulus of stage 2.
6 Likes
My main goals for the redesign of the double stage drives were to #1 reduce size as much as possible without sacrificing any reliability or longevity, #2 apply engineering math to optimize the design, and give a decent estimation on strength.
I really wanted to play around with internal gears like these:
As I’m set on using SLM already, I thought I might as well take full advantage of the technology’s ability to customize and implement stuff sane people normally wouldn’t. Internal gears like this can be used to shrink the casing of a multistage drive… However they are quite hard to implement. For one, a gear generator doesn’t exist for them, I’ll have to make my own… The equations are freely available, but more complex than regular gears. I just don’t even want to get started on implementing that until I pick up the secondary monitor on facebook marketplace this weekend.
So back to the sketches… The first sketch I shared preliminarily… Highlighted is the pitch circle of the big intermediate gear. I forgot about the truck! Yikes. I guess that’s what implementing complex assemblies on sketches at 2am can occassionally result in.
So how about moving the internal gear to the intermediate stage? Should still offer similar benefits in terms of overall drive size.
This meant the two intermediate gears are no longer just printed onto each other - the simple way would require supports, and I am not confident in a load bearing SLM part printed on supports. So figured out a way to bolt them together. All looks good, and this acts as a bearing hub as well for 608 bearings, perfect.
Here’s sideview of what I have in CAD for now - gears are just represented in pitch circles for the time being.
Next step is figuring out what’s the max face width I can get away with for the gears, then design the casing, taking good care around the floating shaft. The CAD is all parametric so moving the finalized mod back into fusion shouldn’t mean too much work 
Preliminarily it’s looking like first stage will be around mod 1.3 and second stage around 2.6; but this might be lowered a bit if I can fit slightly wider gears. Once I’ll have the face widths finalized I’ll share the math on which I based my numbers. TLDR I estimated the material’s capacity for infinite cycle life, then I matched the gear geometry so that the stress approximately equals what’s allowed.
Overall, this gear drive redesign is turning into a more complex project than I initially anticipated. For now, working on it in “quiet times” pretty much only when I can’t run the CNC and don’t have the energy for Simulink simulations of FOC…
11 Likes
SD card wasn’t the fix for CNC gremlins. Didn’t realize initally that the control board’s PSU had isolated output, so now I grounded it’s output and hopefully that’s the fix Cutting right now, I’ll see soon if it continues to work. Also lowered the current of the stepper drivers to make sure the gremlins are not temperature related. The drivers were pretty cool but the motors are toasty all the time even at 7/16 setting. Now I’m trying 5. That said the motors don’t have temperature feedback, so it shouldn’t have been the cause 
In other news, I ordered 120 RS50 cells for this board. Cells should arrive around late June, I’ll get by until then with lipos. Every liion cell that’s any good is preorder only now in Europe, which kinda sucks.
For the time being this will definitely be 20S6P. However, the full current will actually have to be “limited” even with 0.3 copper interconnects. I’ve ordered 12 extra cells to throw together a 2S6P for testing purposes, and 8 extra cells to throw together a 2S4P for testing purposes - namely I want to test the copper’s temperature rise with high amp pulses. From my preliminary calculations it’s looking like 0.3 copper at 50mm width is pretty far from 100A per cell in 6P.
I say “limited” in quotation marks, because that assumes that I (or anyone on an esk8) can actually draw that much for any significant time. I can’t. Probably noone can. Nevertheless. Actual test data is always cool to see.
4 Likes
Running that kind of power, can you tell if the battery is sagging a few laps in?
1 Like
I’m hoping with the new pack I won’t be able to notice any sag. With the 21s4p p42a pack I could feel sag from around 4-4.05V per cell
2 Likes
I bring this up, as I have a fraction of the power, but sometimes I add my 7s2p 178wh feeding a ‘600 watt’ CCCV voltage booster ‘portacharger/range extender/ Antisag device’ to the deck of my board set to just below my 10s2p Tenpower 50XG max charge voltage, feeding upto 8 amps, and it is very noticeable that the throttle is far spicier and the board is quicker and faster, until The portacharger is depleted.
If I am in the mood to roll as fast as I can, then strapping the portacharger/antisag device to the deck, plugged into my charge port from the start is the way to go.
Even if I strap it on at 50% charged, the throttle feels like battery voltage is still in the 4v+ range instead of 3.6v. I have been continually surprised every time I use it, that it is hard to not use it, although I can feel the extra weight regarding nimbility as My board only weighs 12.25kg/27 lbs. and the portacharger is 1.5 more Lbs
I was just wondering if 4x the battery and having 14x more power, whether a far more capable Antisag device could make your last lap potentially as first as the first.
CNC gremlins weren’t the grounding / PSU either. I’ve been told to check the couplers maybe one of them isn’t tight enough. Hopefully that’ll be it…
While I’m at that, I took the bearing assembly of the ballscrew apart to take some measurements and hopefully find some replacement components.
This is where the crunchiness comes from on the other axis, this axis was easier to take apart though.
Just for my own future reference, 60mm dia 16mm width bearing block, 20mm inner 23.5mm outer bearing dimension. Has a sleeve inside which is 20mm dia 20mm long.
35mm outer 15mm inner, 12mm width.
If someone is a bearing expert, I’d very much appreciate some help identifying these as my own searches didn’t return very useful results…
EDIT: found it, AXNBT 15 60
I think an “antisag” device is just extra batteries with extra steps. I don’t think I’ll really feel sag on the new pack as I just don’t think 120 tabless cells can be pushed to the limit on an esk8. The cell to cell interconnects is the only worrying point really, but that’s mostly from a temperature point of view, not in regards to sag, when we are talking such high amperage (p=i^2*r, we are talking insane amps). Don’t think I can (or even want to) weld 0.5 copper 
1 Like
I’ve done some testing with the RS50 cells. I don’t ride as hard as you do since I don’t race but I still managed to get 50wh/mi despite being ~120lbs. With my 21s6p P45b I was getting 25-28 miles, my current 20s5p RS50 I’m getting at least 30 miles. The kicker is my smaller RS50 pack is getting more range in 35°F weather than my P45b in +75°F weather.
5 Likes
New servos, couplers, MQL coolant system, and a separate PSU for the control board’s MCU side is coming in 2-3 days. I am really hoping the solution to the gremlins is in these orders, given the 1200€ ish I just spent for the parts. The only other thing it could be is the control board, which just sounds very implausible. If this indeed fixes things, I think I could get the first CNC’d alu part finished by Friday.
Gonna sacrifice my Huawei board charger to be the PSU for the servos until I buy a separate PSU for them. Huge nema34 660W 72V JMC servos.
Also received the quote for replacement bearings for the X and Y axis.
So this adds an extra project to the list, modelling an adapter / bearing housing, that fits standard bearings. On the other hand I discovered maybe my preload wasn’t perfect and the bearings aren’t as bad as I thought originally, so I also have a torque wrench coming that covers the expected preload range. Hopefully by playing around with the preloads I can get the bearings to quiet down and get rid of the low speed vibrations on the X axis.
6 Likes
Holy fuck that’s big!
The long one has brakes that electromagnetically deactivate when powered so that a heavy Z axis on ballscrews stays where it is when the machine is off.
5 Likes
Via Smith Diagrams?
Is this an actual picture by you?
Which alloy, what magnification, and which acid was used? Looks really low in C and not hardened at all, in addition the grain size looks rather big considering this might be a more macroscopic magnification level.
If you need any help with selecting alloys and heat treatment, feel free to ask - BSc Material Science here.
1 Like
Its just a picture I found on google comparing the manufacturing methods, to show that lasercut walls are never perfectly perpendicular and that’s why the wear was looking a bit assymetrical in the beginning. The open gears on the board now are 4130 lasercut, no extra hardening. They still seem to be just fine after probably little over 50 track sessions on them. They’ll definitely survive until I get the new gear drives going.
I’ll get back to working on the new gear drive beginning of june, I have a few more urgent projects until then. I think I’m probably set as far as materials and manufacturing methods go, but thanks for the offer. I’d like to share my calculations sometime in a PM, I’d appreciate a quick look over it from a material sciencist POV.
For now I’m working on the CNC itself and then the race chassis - I’d like to get the chassis done before the next race, which is in two weeks time. Will be a very busy two weeks.
In my experience with lasercut steel the edges are quite hardend by the cutting process itself.
Would be cool to tweak lasercutting process parameters to get a quite deep hardening. I wonder if this could work out for the size and (kind of simple) geometry gears we use in esk8. Did we ever talk about this on the forum?
I’d agree that lasercutting does some hardening, the effect of it is not something I can quantify myself though. What I can say is that the lasercut gears I got are holding up pretty well over time. I’d prefer to have more engineering freedom and not run 2D gears though… At least for more complex drives like what I’m working on. I’m quite set to give SLM a try
Installed the new servos on the mill, had some trouble with a stripped grub screw which delayed things by basically a week. Still have to do the wiring for the servos. Then pray that this actually makes my issues go away. The motor swap eliminates most possible causes.
Next weekend is race in Pardubice. Could I finish the chassis till then? Maybe, but as I also have exams to learn for, sounds quite idealistic. Doesn’t mean I’m not working on this in the meantime though. I’ll be wiring up the servos asap then run some cuts to see if the issues actually disappeared. If they indeed disappear - the most likely scenario is that I’ll mill out the main chassis part and the footpads on time.
I have also printed out one of the swingarms. Obviously, not gonna be racing on 3d printed swingarms… But since the most likely scenario is that I can get the chassis done, but not the swingarms, this way I can still take the prototype with me to show it off, then finish it after my next exam on 2nd of june.
We will see how far I manage to get before the race.
12 Likes
Just picked up some lightly used Mojo D5 and Lecont LPM tires. Will be interesting to test both, I’ll try to find the time to have both sets mounted on rims for Pardubice. I think both should perform better than my brudgestones, we will see.
1 Like
How much is a set of tires?
Technically I bought 4 full sets (2 front and 2 rear per set) this way I have 4 fronts of mojo d5 and 4 fronts of lecont lpm. And 4 rear from each which I am not planning to use though it would be hilarious if I’d buy some rims for it 
One set of kart tires (2 rear 2 front) cost ~50€ used for around 20 laps. A set is ~200€ brand new, but you lose like half a sec per lap if they aren’t brand new, so kart racers with infinite money just change them all the time and sell them for cheap.
So I paid 200€ and have 2 skateboard worth of front tires plus 2 skateboard worth of rears which are just hilariously huge. 11x7-6. Is that Onewheel size or even wider?
I guess I could say 100€ to kit out a skateboard plus 4 onewheels, with the catch that I don’t have a onewheel yet and if I had I probably wouldn’t use kart slicks on it anyways?
2 Likes