repeatability issues with roller chain

[prcdslnc13]

Im pretty much finished with my machine build now, but im having issues with repeatability and backlash in my long axis that uses roller chain. I pretty much used the drive system for my Y axis that Patrick from Buildyourcnc.com uses for the blacktoe. I dont see how he says theres "little to no backlash" as im seeing all the way up to .010" in postion variation on my dial indicator just going back and forth in .010" movements in EMC2. (Yes you read that right, moving in .010" increments and up to that much in variation both short and over travel)

You guys have any ideas? Its a 9 tooth sprocket with 25pitch chain.

[CarveOne]

Im pretty much finished with my machine build now, but im having issues with repeatability and backlash in my long axis that uses roller chain. I pretty much used the drive system for my Y axis that Patrick from Buildyourcnc.com uses for the blacktoe. I dont see how he says theres "little to no backlash" as im seeing all the way up to .010" in postion variation on my dial indicator just going back and forth in .010" movements in EMC2. (Yes you read that right, moving in .010" increments and up to that much in variation both short and over travel)

You guys have any ideas? Its a 9 tooth sprocket with 25pitch chain.

My opinion only, with no mechanical design engineering background to back it up:

If there is a droop in the chain you may never get it tight enough to take it all out, and it will be putting a lot of stress on your sprocket shafts. You may be able to add enough idler sprockets (top and bottom chain sections) to keep the chain well supported so that there is not so much tension required, but, if the chain links run bottomed out in the sprocket grooves, it will never be very close to a "zero back-lash" system.

Getting the perfect long chain setup is why I think it has generally been discouraged around here in applications where precision is required. Short chain lengths fair better but there are better ways to do this. Usually in the form of ACME rods or ball screws.

CarveOne

[macgemby]

Hi
I had similar problems with a test of this chain system, and eventully discarded the use of it, luckily it was only a test of the concept on one slider, and did not put too much money into it, I was getting way more play, then what you describe I came across some good info why this is not good here Can I use a Roller Chain and Sprocket drive ? - Rockcliff CNC Machine Forum

[prcdslnc13]

I have some extra timing pulleys laying around, so Im going to order some timing belting from Mcmaster Carr. I want to try running that and see if that makes my repeatability better before I look at running a R&P setup for it. Anyone else done that?

[CarveOne]

Long belting of any kind, whether chain or timing belts will have similar issues. ACME rods and ball screws whip if they are not straight or are rotated to fast for the length involved. R&P have their own issues, but they are about the only solution for really long travels. Grunblau is using rotating lead nuts to get around the lead screw whipping issues. You may want to explore his build log in this forum to see how that has worked out for him. Pretty much anything much over 48" length of travel is going to have some type of problem that will have to be dealt with.

CarveOne

[prcdslnc13]

the difference between the belts and the chains is that the belts are in constant contact with the pulley, with little to no slop in the teeth. With the sprocket and chain theres slop between the teeth of the sprocket and the rollers themselves. If the sprocket had involutes that would help correct this but the one I have doesn't. Thats why Im thinking the belting would work as a better solution.

Even though I am stretching over 60" There is no slack in my chain, so i dont think thats my issue. I saw the issues that slack caused in my experimentation.

I was able to get alot of my backlash out through the BL comp in EMC2, But im still kind of iffy on the configuration of it. Hopefully Ill get some time to work on it tonight.

[JD_Mortal]

Have you done a break-in on the chain yet? (Run the full length like 1000+ times, fast and slow, on all axis. Tighten slack and re-grease it, and then run again, etc...)

If you do not have sufficient "solid tension", you may get undesired play. Also, just the fact that the chain-gears are not, "Pressed-fit" will translate in play. I assume you used normal sprockets, which have a key-way and a tension-screw. (Thus, the screw-side is 0.01mm off-center, as are the bearing-play values, and are the tooth-gaps and link-gaps.)

With over 100 links, that is a LOT of play to contend with. Even just from simple heat-expansion in the chain.

With time, and a little wear, your play will become more exaggerated, but it will also become more consistent, except for the off-center tension-screw issue. (That can be temporarily reduced/resolved by using a thin coat of enamel on the shaft you are attaching to, if the gap is visible. Ultimately, you want to get press-fitted sprockets. For the least possible off-center accuracy.)

If you have a tension-sprocket that is not solid, eg it is springy or moves, that is the first thing to fix.

Also, you may have inner-burrs on every segment of the "stamped chain links", which will initially wear, and cause undesired play, until they all break-off, or wear away. They cause the chain to be shorter than normal, due to the added thickness from the raised-edges. Once worn, the links rest better against the pins and the teeth.

Design-wise... The closer you get to direct-drive, the more potential error you will have. That is the only benefit to slower speeds, higher ratio-gearing, and higher RPM setups. With a 1/10 precision, that 0.01 inch error becomes ~0.001 with a more solid drive system. (However, with chains, you can use a 1/2 and 1/4 pulley setup, to increase torque and precison, at the direct loss of speed.) Your best bet, if a chain is needed, is to use a triple-wide chain, similar to those used in industrial drives. The error is a lot less, the wear more balanced, and tension is easier to control without stressing the motor or drive-system.

[Aryantes]

I started out with the same build from buildyourcnc.com. 25 pitch, 9 tooth sprocket drive. 2' x 5' cutting area.

I could never get the roller chain to function perfectly.

I mean, I was able to cut MDF / Plywood with pretty good accuracy but repeatability/precision was a different animal.

As soon as cutting forces were involved, I would experience play in the chain.

I had a chain tensioner that I put pretty tight to remove slack in the chain but I noticed that, for instance, instead of moving the gantry 0.03 inches, sometimes it would stay still and stretch the chain instead.

This didn't happen all the time, but there was definitely cogging and other issues involved. Tightening super tight had even worse consequences and I feared that I would shear off the shafts on the motor as one member did at buildyourcnc.

Circles were not perfect. Squares not perfect. I have a hundred MDF test pieces cut all over my basement now haha.

5' chain of 25 pitch is long I think. If I had a 1' x 2' machine with very light cutting forces I think it would be working just fine. I had the problem on both the long axis and short axis though.

I gave up the chain and recently purchased R&P from FineLineAutomation / CNC Router Parts.

I haven't finished the build yet though.

[harryn]

At least one of the problems that comes up in chain and belt drive arrangements is that the typical design has used under rated sized parts. I did a calculation based on belt stretch for a wood router, and it is pushing toward 4-6 in wide belts to make the stretch a "non issue". If you are imagining that a 1 inch wide belt is sufficient to be happy - forget it.

The chain size that you mention is pretty light (based on the ratings of the similar stuff from McMaster-Carr's site) Nonetheless, as pointed out, there are a lot of things moving around on a chain drive. The concept was invented to move power, not really precision positioning. I imagine that it is possible to make chain drives into precision positioning devices, but the off the shelf stuff is not usually designed that way.

[prcdslnc13]

Is there any chance you can show me where you got the calculation you used for that? I understand the small belts can be an issue, but some of the big lathes (think 48" cut table) in the shop I work in are belt driven with 2-3" wide belts with perfect repeatability.
Id like to know how to do the math anyway though

I messed with everything over the last few nights, and played with the backlash comp in EMC2 and got it repeatable within about .001" but I could physically watch the chain stretch on my indicator haha. Ill be looking into a better solution, Ill probably try belts just because its a quick easy conversion. But Eventually I plan to move to a rack and pinion. This has turned out to a good informative thread

[JD_Mortal]

Can you post a photo of the problem area?

I suspect that there is a burr on your track, or a flattened/scuffed bearing on your glide system... Possibly, also an issue with too wide a span with inadequate support.

If your width of the gantry is too wide, and your guide rollers form a square (Actually is a circle with four points.), it may require some structure modifications to make it up to your standards.

I am assuming that you obviously have higher standards than the creator of the original plans, or some other sacrifices were made in the building process.

How is your gantry weight, with the spindle loaded? Does it move easily when you detach the drive? Try loading it by hand, in a similar direction that shows the most error. Does it still move freely, or do the guides expand or compress?

You may just have to use dual-drives on the gantry. One chain per side, controlled by two motors or joined to one motor.

Also, if your chain hangs freely, over a great span... That slack will first have to be absorbed before there is sufficient pulling power to the mounted location. If that is the issue, it will be less noticeable as the gantry is closest to the center of the machine. This will be due to the semi-supported chain having two half distances to the tensioned rollers, as opposed to having one full length spanning the entire length of the machine. The only thing that will help in that case, is a push-away chain support. (Like a long bowed brace underneath, supporting the length of the chain while the gantry is at the ends. The gantry would push the bow away as it glides, but still support the rest of the chain length which the bow still touches.)

#25 chain is commonly stretchy, unless it is hardened steel. Most #25 chains are just low-grade steel. They use them in all the china-scooters and small bikes, and project robots, where the span is not greater than 2-feet. They also have poorly aligned joints, which contributes to the uneven left-right slipping/stretching as the chain is tensioned.

You say it is unrepeatable... suggesting that each time the inaccuracy is not the same. Are you absolutely sure it is unrepeatable, and that it is also not consistent? Something that backlash settings in the program might reduce to a more desired level of inaccuracy?

[hemsworthlad]

Hi,
Unless you already have the belts and pulleys I wouldn't waste your time and money with long belts, Go straight to R&P or screws.
If your not chasing high rapids then cheap chinese ballscrews are good if not too long say over 80" .
If you choose an high enough pitch and Dia screw then the speeds can be kept lower and still have decent rapids without whip. I use 60" x 20mm x5mm pitch Dia BS geared 1/2 and I dont get whip at 400Ipm which is the point my motors run out of steam.
I have had them as high 600Ipm with out whip when geared 1/3 but this is too stress full on the small pitch screw and would cause excessive wear reducing overall life and I just dont need high rapids.
A 10mm pitch screw geared 1/2 would double this and still be in the same region as R&P in resolution terms. Then the other added benefit of gearing which I do using short timing belts is you can gear 1/1 for higher resolution and torque if needed.

R&P rules for long length's, BS rules for precision and effiecency but only in the short 60-80" region. BS are far easier to setup and align and when done pritty much hassle free and free from lots of issue's that can come with R&P like crap getting in between teeth and jumping etc all things which need considering if you want a hassle free machine.!

For ME personally and MY cutting needs BS wins every time but in your world it could be very differant and suck big time if too long.? . . . . Now with rotating nuts it could be an all new ball game.?

My point dont count BS out for long length's if you need high precision/repeatabilty.