Hacker News Reader

Comment by SavantIdiot | Hacker News Reader

Wow, really short section on gears. But there is so much to cover I can't blame it.

If you've ever dabbled in robotics, you've probably found it is an enormous step function from kits like Actobotix [1] to custom parts. It is infinitely easier to use kits or frameworks that have already taken a LOT of the guesswork out of gears, linkages, motors, compatibility, etc. Especially gears, they are an enormous pain in the ass to source online, and even when you find the right gears, you still need hubs and axles, and then you hit min volume issues, non retail sales, overseas... I even have access to a phenomenal machine shop, but what I don't have is a machinist. (There are a few that mill around (pun intended) but they are mostly introverted decent guys that don't really want to help a noob.)

[1] https://www.servocity.com/actobotics/

View on HN

In case anyone wanted an approachable introduction to machining your own gears, there's a This Old Tony video about it: https://www.youtube.com/watch?v=Q-XOM4E4RZQ

I spent 5 years making gears in a job shop.

I did a cursory look through the section on gears, it seemed adequate for someone who is doing a hobby project.

I'll watch this thread and answer any questions you might have.

Having spent a million dollars making gears, my ultimate conclusion is it’s the sort of thing you machine from brass gear stock unless you want to spend a million dollars.

For what 99.99% of makers are doing, Actobotix is fine. But if you want to make your own fancy differential or a hypercycloid linear translator, yeah, you're gonna have to do it yourself.

What dimensional scales, materials, turn-around, volume, and/or processing?

Breaking tools and getting G-code just right seem like the perpetual headaches of modern machinists. Other than occasionally being bitten by a bitchy blade-like chip begging to fillet your grizzled knuckles one last time before banishment from the bench to the backyard becycling bile of betal bits.

If electricity were cheap (or the electric meter rewindable), would multiple beam concurrent FDM be a consideration if the materials and metallurgical characteristics allowed it? I think at some point, it's easier to print complex shapes additively than machine them subtractively, and that needs to be part of structured cost decision-analysis for manufacturing.

I think if you want to make one or two custom gears, you just manually mill them. CNC involves the usual programming tradeoffs -- you spend a ton of time writing and debugging the program, and then you can run the program a million times to make a million parts. For making one part, it's not clear that it's worth the overhead -- just set up your stock on a dividing plate and use a gear cutting tool to cut the teeth. If when you're turning the handwheels the machine makes an unpleasant noise or it feels like you're going to break the tool, stop turning the handwheels and use your brain's built in AI algorithm to develop a new plan ;)

Usually a short run of a new part doesn't take more than half a day to set up on a CNC. If you are making thousands of parts to run unmanned you might have to add another day to it.

Watch AvEs vijeos.

Problems.

No problems. After they are fixed by the meat confusor.

They are solved.

Your tool in a vise.

What is 'brass gear stock' as opposed to normal brass?

gear stock is basically a gear that's really, really long so you can just cut off whatever width gear you need. Similar to bar stock, which is just rods of metal, but in gear shape.

Something like this: http://shop.sdp-si.com/catalog/?cid=p593

So you can decide on a module/PA, buy a few diameters then cut off the required face width you need and machine the bore as needed.

Would you know what type of approaches and machinery is needed for relatively loose tolerance / low load helical gears? If going the CNC route, do I need full 5 axis and specialized end mills?

Even with regular involute straight gears, there's a bunch of ways one could manufacture one, and I'm out of my depth here.

The industrial strength version: https://m.youtube.com/watch?v=EefFxEGVbWo

I made sprockets with an endmill and a simple macro on a C-axis lathe. Worked just fine since the tolerance on those is very forgiving.

In principle you could make a helical gear as a multi-start thread. I'll have to test it once I have the machine free an afternoon.

It's way easier to start with components you can readily source and design your system to use them than to make custom components. If your robot isn't going to space, the benefits of custom components will not be worth it.

Reply on news.ycombinator.com