Machining precision required to make custom gearboxes

[Cog]

My team is currently making a (simple) custom gearbox in place of the COTS gearboxes we were planning to use that ran out of stock.

My question is: Do you really need thou precision to make decent gearboxes? Because after looking at some other threads, it seems like even measurements as small as .003" can make a significant difference in the performance of a gearbox.

[Cory]

I would say you should be able to relatively easily hold .002" on position of hole centers on a decent quality manual mill with a somewhat experienced operator. That would be good enough for a custom gearbox.

[JesseK]

What about oblong holes? Our DIY gantry-style CNC consistently cuts oblong bearing holes that are exactly correct in the 'X' direction, but approximately 0.001-0.002 off in the 'Y' direction. That's about as close as they could get before needing to move on to other things.

As the guy CAD'ing our first-ever custom (non-drivetrain) gearboxes this year, is there anything I can do to make this a non-issue? The concern for us is gearbox longevity in ~120 matches between March & November.

[mman1506]

Originally Posted by JesseK

What about oblong holes? Our DIY gantry-style CNC consistently cuts oblong bearing holes that are exactly correct in the 'X' direction, but approximately 0.001-0.002 off in the 'Y' direction. That's about as close as they could get before needing to move on to other things.

As the guy CAD'ing our first-ever custom (non-drivetrain) gearboxes this year, is there anything I can do to make this a non-issue? The concern for us is gearbox longevity in ~120 matches between March & November.

Your probably going to be fine, the 20DP gears commonly used in FIRST are purposely larger than necessary so they can handle being abused in low precision gearboxes and high sudden torque situations. Usually the failures of gears in FIRST are either due to gears being too small and cracking at propagated stress points of the hex shaft or flexing and allowing the gear mesh to become far too loose (the VEX Ball shifter 3rd stage seems to have this issue). In the scheme of things the amount of time FIRST robots run is very short and wearing your gears to the point of failure is very unusual unless something is really wrong like the mesh being too tight. This why its common to add .002" to the C-C mesh of gears in FIRST.

[asid61]

Originally Posted by mman1506

Your probably going to be fine, the 24DP gears commonly used in FIRST are purposely larger than necessary so they can handle being abused in low precision gearboxes and high sudden torque situations. Usually the failures of gears in FIRST are either due to gears being too small and cracking at propagated stress points of the hex shaft or flexing and allowing the gear mesh to become far too loose (the VEX Ball shifter 3rd stage seems to have this issue). In the scheme of things the amount of time FIRST robots run is very short and wearing your gears to the point of failure is very unusual unless something is really wrong like the mesh being too tight. This why its common to add .002" to the C-C mesh of gears in FIRST.

Quick correction: gears in FRC are 20DP (which is actually a very odd pitch for some reason, most sites only sell 16p or 24p) and are actually way undersized. If you compared the tooth of a Vex gear to that of a Boston Gear gear, you'll notice that the Vex teeth are slightly thinner and have large sections of the root of the tooth taken out.

I second that you need .002" precision at the minimum. Personally I prefer .001". If you do not have a mill, please don't attempt this mid-season- there is almost a saturation of gearboxes available from WCP, Vex, and Andymark.

[protoserge]

Originally Posted by JesseK

What about oblong holes? Our DIY gantry-style CNC consistently cuts oblong bearing holes that are exactly correct in the 'X' direction, but approximately 0.001-0.002 off in the 'Y' direction. That's about as close as they could get before needing to move on to other things.

As the guy CAD'ing our first-ever custom (non-drivetrain) gearboxes this year, is there anything I can do to make this a non-issue? The concern for us is gearbox longevity in ~120 matches between March & November.

Do you have backlash or is it just lead error? Can you program backlash compensation into the machine configuration?

[JesseK]

Originally Posted by protoserge

Do you have backlash or is it just lead error? Can you program backlash compensation into the machine configuration?

I do not know, but I will ask those questions. They spent about 6 hours re-calibrating a new bit Saturday, but since I was knuckles-deep in drive train gearboxes I don't know the specifics.

[FrankJ]

If you don't have a CNC or a manual mill, you can do this manually. With the typical FRC gears and a simple gear box. you can lay out and drill holes using a machinist scale, scribe, prick punch, a drill press, and a little OCCD. Clamp & drill the side plates together. This takes a lot of care, but is very doable.

[Cory]

Originally Posted by FrankJ

If you don't have a CNC or a manual mill, you can do this manually. With the typical FRC gears and a simple gear box. you can lay out and drill holes using a machinist scale, scribe, prick punch, a drill press, and a little OCCD. Clamp & drill the side plates together. This takes a lot of care, but is very doable.

While what you describe certainly could be done by someone with very steady hands, good eyesight, and great attention to detail, I would argue that any team which only has a drill press shouldn't be considering building custom gearboxes given the plethora of great COTS options available.

Much more valuable to expend that effort on other areas of the robot.

[InFlight]

Are there any other teams in your area that have access to a CNC mill or water-jet? Most teams are happy to support other nearby teams with machining or needed parts. FIRST is really a community, we only compete on the field.

[s-neff]

Originally Posted by FrankJ

A little OCCD

A lot of OCD and a quality drill press. Even with the plates clamped together, our drill press isn't straight/rigid enough for me to be comfortable with 841 trying this [with aluminum gears].

Originally Posted by Cory

Much more valuable to expend that effort on other areas of the robot.

Yes. Highly recommend finding the extra ~$100-200 to buy-not-build.

Originally Posted by JesseK

...oblong holes...

Consider milling undersize and finishing with a 1.1250" reamer - biting the bullet and getting a reamer has allowed 841 to use our not-super-rigid benchtop tools to rough out bearing holes at the right location (on say a WCD rail or mechanism arm) quickly and a little undersized, and then finish the holes to a really nice bearing fit.

[protoserge]

Originally Posted by s-neff

Consider milling undersize and finishing with a 1.1250" reamer - biting the bullet and getting a reamer has allowed 841 to use our not-super-rigid benchtop tools to rough out bearing holes at the right location (on say a WCD rail or mechanism arm) quickly and a little undersized, and then finish the holes to a really nice bearing fit.

That is how I have done this in the past to deal with our Taig cnc. I grabbed a 1.124 reamer off eBay that I'd like to try.

I would venture to think that there is backlash in the router that can be corrected with software (LinuxCNC/Mach/etc.), but I haven't seen their setup.

[JesseK]

I believe the calibration issue we're seeing with the x/y is in the machine code processor. Things can also get a little messed up if there is debris in the rack and pinion drives. Just haven't had an opportunity to tease it out yet.

We do have a reamer, but have had mixed results when using it on oblong holes. Using a micrometer on a sample after under-sizing still shows a position margin of error of around +/- 0.001 in a dimension.

In any event, since the gears arrive tomorrow we'll likely just make the gearbox and report back in November. If we see signs of premature wear during our planned 50-climb test, we'll probably get steel gears and just brute force it. It's in an easy-to-replace spot and we're building a spare as it is.

[protoserge]

Originally Posted by JesseK

We do have a reamer, but have had mixed results when using it on oblong holes. Using a micrometer on a sample after under-sizing still shows a position margin of error of around +/- 0.001 in a dimension.

On a DIY setup, that's pretty respectable. You could try running a spring/multiple finishing passes to see if that helps remove hole eccentricity.

[FrankJ]

Originally Posted by Cory

While what you describe certainly could be done by someone with very steady hands, good eyesight, and great attention to detail, I would argue that any team which only has a drill press shouldn't be considering building custom gearboxes given the plethora of great COTS options available.

Much more valuable to expend that effort on other areas of the robot.

I agree that you are better of buying the solution when one is available. The COTs choices have grown tremendously in the last few years. But one stage is fairly easy with 20 DPI gears. More than 2 stages becomes really difficult. I always take stated accuracies with a grain of salt.

In terms of drill presses. We are fortunate to have a heavy 1/2 press with good bearings. But I get as accurate results by using a 1/4 in Ryobi and either a short 1/8" bit or center bit.