Chain Tensioning

[2348humanplayer]

Really, it depends on what kind of drivetrain you have. Our team had separate methods.

If the wheels didn't need to worry about being bent (as in, they can just be a straight line) we used the simple method of just getting the chain, wrapping it around the two sprockets, and seeing which link to break. We broke the right one, and used the master link to join them. Simple.

However, there may be a time when you can't make the chains go straight to the wheels. We had this issue in 2010, when we needed our robot to be able to traverse the bumps, but a straight link for the chains would grind on the bumps, causing damage to the field and the robot. To fix this issue, we basically had the chains go across the robot, parallel to the frame, before bending down to the sprockets. However, if we did this method, measuring the chain the old way would be incredibly difficult. That is why we made tensioning devices to help keep them taut. What we did is we made a mount to go under the robot, putting two of the hard white plastic spools on them. They were attached to the underside of our robot by two bolts (we always use extruded aluminum for our robots, we just put a couple of cut bolts into the groove for placement, held in place by lock nuts and washers). if we ever needed to replace chains, we simply released the tensioners by sliding them towards the center, loosening the tension so it was easy to remove the longer chain. When replacing it, we made the chain long enough where it would at least reach the two and be able to have enough give. We strung them over the tensioning spools, then slid them out until the chain was taut, and bolted it into place. The wheels were still able to turn with our direct drive, and the chains were well out of the way.

But, in all honesty, that's a very odd circumstance, like going over tall obstacles or having tiny wheels and big sprockets. The easiest thing to do is just dead reckon it by making the chain the right size and measuring the links needed. If anything, a slightly loose chain is better than an overly tight one in any case.

[ToddF]

This summer we are experimenting with non-adjustable wheel positions, and chains inside our side rails.

The first rail we made had nominal C2C distances. While it is functional, the chains sag enough to touch the inside of the rail tube, making the drive assembly noisy when running. For the second tube, we took a WAG (embarrassing to admit, for engineers), and increased the C2C distance about .008". This resulted in saggy chains again. After that, we did what we should have done from the start, and actually measured the C2C distance of tensioned chains, instead of just guessing.

We made this test setup, first using the chain from our experimental drive train, with a nominal C2C distance of 10.75". Note that the sprockets, bearings and shafting are all 1/2" hex type, all from VEXpro. There are slip fit tolerances between these parts that require the C2C distance of the bearing pockets to be greater than nominal to achieve a tight chain. Also note that the neither the hex holes in the bearings nor the hex holes in the sprockets are truly concentric. This leads to visible wobble in the OD of the sprockets, when they spin. This causes the chain tension to vary, and is a source of vibration and cyclic fatigue to the overall drivetrain.



To test whether any C2C variations were related to the length of the chain, we also tested a nominal 4" C2C chain setup.



The chain tension was set by anchoring one of the bearing blocks, pulling the chain "finger tight" and locking down the other bearing block. We then spun the chain by hand to observe that it could run free, and hand checked the chain tension the same way we do in the pits, to be sure it was tensioned comparably to our competition drive trains. Then we removed the chains, sprockets, shafts, and upper bearings. The upper bearings have a tight slip fit. To remove them, it was necessary to insert the end of an axle shaft and wiggle it around, working the bearing loose. This left the two bearing blocks with the bearing pockets exposed.



We then used the milling machine edge finder and the dimensional readout to find the C2C distances (we actually measured to the left sides of both bearing pockets).

Here are the results:


We concluded that the increase in C2C distance was due to the tolerance stackup of the parts, and not tightly related to the length of the chains. We plan to incorporate a .019" to .020" delta to our nominal C2C distances in furute designs.

[Michael Hill]

Originally Posted by ToddF

This summer we are experimenting with non-adjustable wheel positions, and chains inside our side rails.

The first rail we made had nominal C2C distances. While it is functional, the chains sag enough to touch the inside of the rail tube, making the drive assembly noisy when running. For the second tube, we took a WAG (embarrassing to admit, for engineers), and increased the C2C distance about .008". This resulted in saggy chains again. After that, we did what we should have done from the start, and actually measured the C2C distance of tensioned chains, instead of just guessing.

We made this test setup, first using the chain from our experimental drive train, with a nominal C2C distance of 10.75". Note that the sprockets, bearings and shafting are all 1/2" hex type, all from VEXpro. There are slip fit tolerances between these parts that require the C2C distance of the bearing pockets to be greater than nominal to achieve a tight chain. Also note that the neither the hex holes in the bearings nor the hex holes in the sprockets are truly concentric. This leads to visible wobble in the OD of the sprockets, when they spin. This causes the chain tension to vary, and is a source of vibration and cyclic fatigue to the overall drivetrain.



To test whether any C2C variations were related to the length of the chain, we also tested a nominal 4" C2C chain setup.



The chain tension was set by anchoring one of the bearing blocks, pulling the chain "finger tight" and locking down the other bearing block. We then spun the chain by hand to observe that it could run free, and hand checked the chain tension the same way we do in the pits, to be sure it was tensioned comparably to our competition drive trains. Then we removed the chains, sprockets, shafts, and upper bearings. The upper bearings have a tight slip fit. To remove them, it was necessary to insert the end of an axle shaft and wiggle it around, working the bearing loose. This left the two bearing blocks with the bearing pockets exposed.



We then used the milling machine edge finder and the dimensional readout to find the C2C distances (we actually measured to the left sides of both bearing pockets).

Here are the results:


We concluded that the increase in C2C distance was due to the tolerance stackup of the parts, and not tightly related to the length of the chains. We plan to incorporate a .019" to .020" delta to our nominal C2C distances in furute designs.

This is some awesome information. I'm having a difficult time telling from the picture, is that #35 chain or #25 chain? (My guess is #35). Would you be able to test if switching to the other size would result in any more or less delta? My instict says to add more for #25 chain (given the same C2C) because there are simply more chain links and the tolerance will stack up even more.

Also, I'd be curious on the additional Run-out the hex broaching causes to the bearings.

[notmattlythgoe]

Originally Posted by Michael Hill

This is some awesome information. I'm having a difficult time telling from the picture, is that #35 chain or #25 chain? (My guess is #35). Would you be able to test if switching to the other size would result in any more or less delta? My instict says to add more for #25 chain (given the same C2C) because there are simply more chain links and the tolerance will stack up even more.

Also, I'd be curious on the additional Run-out the hex broaching causes to the bearings.

They actually did it with #25 chain.

[ToddF]

Originally Posted by Michael Hill

My instinct says to add more for #25 chain (given the same C2C) because there are simply more chain links and the tolerance will stack up even more.

Also, I'd be curious on the additional Run-out the hex broaching causes to the bearings.

As Matt said, this was done with #25 chain.

We believe the tolerance stack up is due to the following fits:
-sprocket bore to shaft
-bearing bore to shaft

The bearing fit into the bearing blocks was a tight slip/light press, with no observable play.

We tested two lengths of chain specifically to see if the chain itself was a contributor. If it were a major contributor, we would expect the delta values to be somewhat related to the lengths of the chain. Instead, the delta values for the two cases were nearly identical (.001" is probably within the margin of error of our measurements). We concluded that the chain was not a source of the delta.

Another possible contributor could be that the sprocket pitch diameters are undersized. It would take some thinking, and maybe a different test setup to check this.

[asid61]

Personally, regardless of belt or chain, I would use tensioners.
Cams are good, but they need to be cadded before you add them into a robot. You cannot just throw them on without considerable effort, depending on how your drivetrian looks.
Screw tensioners are really nice, as they use a solid bearing block and the tensioning is very fine, but they do add a pound or two to the drivetrain depending on their design. I think this would be my preferred method of tensioning chains.

Belts: Too wide for my liking. Width on a belt is several times more than the width on a chain, which decreases the amount of electronics space you have. Plus, if by some fluke a belt breaks, we would have to take apart a gearbox to access the pulleys. I would rather have 4 5-minute chain snaps during competition than 1 15-minute belt break.

Admittedly I hate (or better, despise) working with master links, but 221's chain attachment tool is great. Just make sure you design with enough space in the chassis to use it.

[Oblarg]

Originally Posted by asid61

Personally, regardless of belt or chain, I would use tensioners.

Tensioners on a belt are decidedly unnecessary, in my experience, and will do nothing but add additional friction.

[asid61]

Originally Posted by Oblarg

Tensioners on a belt are decidedly unnecessary, in my experience, and will do nothing but add additional friction.

Friction, but only if you use on-belt tensioners. Plus, that would be pretty low anyway depending on the setup.
I'm thinking more of cams or screw tesneioners. Even better would be to use 192 method that they used on the gearbox: screw holes that were slightly farther or closer to the other side of the belt, from -50 to +50 thousandths. It's a really clever system.

[mrnoble]

Has anyone used the nautilus cams from WCP? We are planning on incorporating them into one of our off-season drive bases this fall as the chain tensioner. I guess I don't understand what the big deal is about using tensioners anyway; why would that be less of a problem in a design than doing c2c calculations? I've worked with teams that have done both, with both chain and belt, and it seems to me that the advantage would tilt toward adjustability.

[Lil' Lavery]

Originally Posted by Oblarg

You could always consider using belts. Much less headache that way, and they're more efficient/make less noise, to boot!

If anything, c-c distance is much more critical with belts than it is with chain.

[Oblarg]

Originally Posted by Lil' Lavery

If anything, c-c distance is much more critical with belts than it is with chain.

Yes, but they're far more forgiving of minor alignment slop, and you don't have to worry about integer numbers of links, and they don't stretch.

I've done four drives with belt in my time in FRC, and I think four or five with chain. I've seen at least an order of magnitude more problems with the latter.

[magnets]

Originally Posted by Oblarg

Yes, but they're far more forgiving of minor alignment slop, and you don't have to worry about integer numbers of links, and they don't stretch.
I've done four drives with belt in my time in FRC, and I think four or five with chain. I've seen at least an order of magnitude more problems with the latter.

I disagree with all three of your claims.

1. You can run chain considerably looser than exact C-C (because of alignment slop), and it will work, though it will be noisy/sloppy. It will continue to transmit torque until it is loose enough for the chain to slide off of a tooth, which requires the chain to be quite loose. Belts will slip if there is slop or if the center to center is to small, and snap if they are too tight. Chain is much more tolerant of C-C misalignment.

2. I don't understand this comment. Belts require an integer number of grooves/teeth, and chain does too. I don't see how this can be an advantage of either.

3. Belts stretch, but admittedly, less than chains.

I have been in charge of drive chain on the past 6 robots I've been involved with. We have never had a single issue. We've never broken a chain, as they're rated for considerably larger loads than belts, we never thrown a chain on a competition robot, as we always cover them and ensure sprockets are aligned, and we've never had any other failures. We haven't always had access to great machining resources, and we've never had a problem.

I have been in charge of one belt drive, so I realize my experience with them is limited, but they're thicker, they slip, and we had a belt snap. So we're going back to chains.

On the original topic, I find that putting in a floating idler sprocket, like this (http://www.team228.org/gallery/106/s...998-37b1e.jpg), is the way to go for chain. If you want it tighter, move it closer to the smaller sprocket, and if you want it looser, move it away. It's easy, cheap, light and adjustable.

[Oblarg]

Originally Posted by magnets

I disagree with all three of your claims.

1. You can run chain considerably looser than exact C-C (because of alignment slop), and it will work, though it will be noisy/sloppy. It will continue to transmit torque until it is loose enough for the chain to slide off of a tooth, which requires the chain to be quite loose. Belts will slip if there is slop or if the center to center is to small, and snap if they are too tight. Chain is much more tolerant of C-C misalignment.

I have never, ever, ever seen a HTD belt come off of a sprocket, no matter how hastily or imprecisely-assembled. I've used quite a few HTD belts. I've seen sprockets buck chain regularly for my entire experience in FRC. So I think we'll just have to disagree here.

2. I don't understand this comment. Belts require an integer number of grooves/teeth, and chain does too. I don't see how this can be an advantage of either.

You're correct here, I meant to edit that out of my post but you responded first.

3. Belts stretch, but admittedly, less than chains.

Not by any appreciable amount in any drive I've worked with.

I have been in charge of drive chain on the past 6 robots I've been involved with. We have never had a single issue. We've never broken a chain, as they're rated for considerably larger loads than belts, we never thrown a chain on a competition robot, as we always cover them and ensure sprockets are aligned, and we've never had any other failures. We haven't always had access to great machining resources, and we've never had a problem.

I have been in charge of one belt drive, so I realize my experience with them is limited, but they're thicker, they slip, and we had a belt snap. So we're going back to chains.

On the original topic, I find that putting in a floating idler sprocket, like this (http://www.team228.org/gallery/106/s...998-37b1e.jpg), is the way to go for chain. If you want it tighter, move it closer to the smaller sprocket, and if you want it looser, move it away. It's easy, cheap, light and adjustable.

It seems we've had different experiences. It'd be nice to know what we've been doing differently, since I've never ratcheted a belt nor have had one snap in competition, but have thrown plenty of chains (due to misalignment, bad masterlinks, and other causes).

[BBray_T1296]

Chains take 2 minutes to fix, unless your design has them hidden away stupidly.

Even an easily accessible belt requires significant work to replace, in comparison. WCD makes this easier, but still a major PITA.

TL;DR, I would rather repair the same problem in 2 minutes 5 times, than a 10 minute fix once.

[Oblarg]

Originally Posted by BBray_T1296

Chains take 2 minutes to fix, unless your design has them hidden away stupidly.

Even an easily accessible belt requires significant work to replace, in comparison. WCD makes this easier, but still a major PITA.

TL;DR, I would rather repair the same problem in 2 minutes 5 times, than a 10 minute fix once.

FWIW, I'm pretty sure even if I were to weight failures by repair time my experienced rate of chain failures would be about ten times greater than belt failures. Maybe I've been (un?)lucky.

Originally Posted by Jared

Our team has never used belts in a drive system, but we've always used chains. Our experience is that chain issues are mainly caused by stuff getting stuck in the chains or the sprockets, or the sprockets being bent or misaligned (located incorrectly axially on the shaft, not C-C being wrong).

This has been my experience, as well, though I've had a few come off due to them becoming exceptionally loose if they weren't tight to begin with and had time to stretch.

Chains are also absolutely terrifying at 5,000 rpm, and belts aren't great for higher torque.

This is also certainly true.