What do teams do to ensure they dont have problems with chain or belts on their drivetrains

I am wondering what design considerations teams make when using belts or chain on a drivetrain (particularly WCD)


  • how do you ensure the belts wont get destroyed?
  • how do you determine pulley size to ensure belt teeth don’t shred?
  • what causes chain to break / fall off of sprockets?
  • what causes sprocket teeth to break off?

610 did a robot showcase with Karthik on first canada live (FIRST Canada LIVE! - March 23rd, 2020 - YouTube)
They apparently had many problems with chain and fixed most of the problems by switching to belts so I am wondering what they might have done wrong with the chain and what they did right with the belts.
Summary of their drivetrain from the video:

  • 1x3 box tube
  • 6"? Colson wheels
  • Chain kept falling off
  • Several chains broke
  • Sprockets lost some teeth
  • Shafts bending the box tube
  • Bearing Failure
  • They switched to 2x3 box tube to fix the bending issue
  • They also switched to belts and i guess they were more reliable

If anyone knows more information regarding sprocket size, pulley size, chain and belt types, etc. that they used that might help.

I found their issues surprising because I see teams like 254 who run 1x2 box tube with 22t sprockets with 25 or 25H chain without much issue.

Note: this is not meant to be a discussion about which system is better but I am looking to solve the “my team tried system A and everything broke so we used system B and it solved our issues”. I want to know why the issues happened in the first place and how it could have been designed better.


Ancedotal experience from our team: The worst failures you’ll see on a drivetrain come from exceeding force/torque specs on parts.

Even if you do all your designs as perfectly and carefully as you can, you have very little control over the external forces that will get applied to your robot (by walls or other robots). Additionally, my thoughts align with Karthik’s mentality that “driving is most important”. Therefor, I’m a big believer in overdesigning the drivetrain before anything else.

Usually, this means using larger sprockets to spread the load over more teeth, moving up the belt->#25 chain -> #35 chain steps, using steel for heavy load-bearing components, or picking less “aggressive” ratios to reduce torque.

For both belt and chain, the biggest thing is to get your distance between the shafts correct - both in the initial design, as well as when you actually go to machine the holes. While there’s various tensioning solutions out there, I can’t say I like any of them.

Not 100% sure if I can meet this while still making my point. But, here’s my punchline: I’m not a good enough engineer to trust that I can make a belt solution not break on the field. We’ve seen too many examples of it going wrong in all sorts of ways. Therefor, we overdesign. We can feel warm and fuzzy that if we used #35 chain and steel sprockets, and something still broke, well, we probably have bigger problems than our drivetrain.

WRT the other things you talked about - The bearing failures, bent shafts, and bent tube smells to me like either an over-tensioning issue, or some really wonky set of loading (ex: not enough tension, chain slips and re-engages, wonky side loads ensue). But, 610 would have to be the ones to confirm what they actually had issues with.

It’s also interesting to note that they led off by talking about weight issues, then later mentioned the other issues more in passing. Perhaps they primarily made the move because of weight, but in the process just happened to “reset” the other issues they saw? Again guessing.

That, and, while I’m totally being an armchair engineer here, I’d be a tad bit concerned about the pocketing they did in the frame tube. Not to say it absolutely is wrong, but I just didn’t see if they had any analysis to prove it was not wrong.

1 Like

We have run belt, #35 chain, and #25H chain on our drivetrains without issue. The biggest advantage you can give yourself in guarding against problems are alignment (particularly with #25) and tension. We’ve used 15T sprockets for #25 chain without issue on drives, and provided you’ve properly tensioned and aligned your sprockets, your odds of breaking chain or jumping is very very low. Mostly because the forces put on chain is generally bound by the stall torque of your motor or friction with your wheels (though I’m sure there are extenuating situations in which something can get bound up in your chains that causes problems). I don’t think I’ve ever seen #35 chain break in an FRC competition in which it was used properly.

As for belts, things get a little more dicey. Check out Gates’s guide on power ratings. It varies quite a bit depending on if you’re using HTD, GT2, or GT3 belt. Since belts are generally used in higher RPM applications, they rate their torques based on speeds, but in a drivetrain, speeds are relatively low (compared to things like shooters). Additionally, in FRC, we tend to push the envelope a bit because Gates wants to ensure long life of their belts in industrial use, so they tend to conservatively rate them. So we typically get away with fewer teeth on sprockets and such because the lifetime of the belts is much shorter in FRC than industry.

One other thing, based solely on intuition, I prefer to use older, more worn, chain if I can. Because chain will loosen over time, it will get longer, so in my mind, it makes sense to use chain that has already loosened so we don’t have to tension as frequently in the beginning. Not sure what other peoples’ opinion are on that.

1 Like

So how do teams ensure the belts have the proper amount of tension? Not enough tension will cause it to slip and I’ve heard that too much tension can also be bad. I envision using a clamping bearing block for ease of manufacturing. This could also help know when a fixed center to center is correct.

Definitely start with this.



Again, be sure you are careful that cam doesn’t slip or move during competition.


Since timing belts don’t stretch, if you have a CNC machine (edit: or you’re very good with a manual milling machine), you can design the perfect center-to-center distance in CAD to give the belt the perfect tension after assembly.

If you don’t have a CNC machine, the clamping bearing blocks with WCP cams are a good option.

Use the KOP, and you won’t have problems.


I came here to say the exact thing. So yeah, do that.

1 Like

We used WCP bearing blocks pretty successfully for the past few years, but definitely using a CNC and seating bearings is the better choice if you have access to one. If you’re looking for a good chain supplier, take a look at nitrochain.com (link to #25 chain). Their chain is “pre-stretched” and seems to change “less” as the drivetrain wears in.

A huge issue we ran into when using the bearing blocks was the axles sometimes shifting as the robot drove around. Chain tension has a huge influence on this behavior, so keep a careful eye on your cams and sprockets. We found that the threads in the bearing blocks sometimes become loose over time, so we ended up drilling out the threads in the blocks and installing slightly longer bolts and nylock nuts to prevent this from happening:


One thing I learned the hard way is that it doesn’t matter what you design if you can’t implement it properly. This especially applies to chains and belts.

Story Time:

My senior year of high school I made a roller claw using belts. While not ground breaking, the design was sufficient. Then I built it using rulers, bandsaws, and drill presses (probably not center punching it first). I ended up with a mechanism that sheered the sides off it’s pulleys every event and destroyed its bushings within two matches. It worked, sure, but it wasn’t especially good.

Fast forward to the summer: my friends are redesigning their robot for an offseason. It comes up in conversation that they don’t cut their drivetrain rails on a bandsaw nor measure using a ruler. They used at best a CNC machine and at worst a mill and calipers.

I refer to these tools as “precision machining.”

Many teams who don’t have precision machining likely don’t realize that it’s an issue. Chains fall off because the holes weren’t spaced properly for tension. The belt is too snug or walks to the left because the axles aren’t parallel.

Not having precision is fine - teams get by without it, but the key is knowing it’s a limitation you have and working around it (ie bearing blocks, calipers).

I’ve seen teams who’ve built the same robot (conceptually) as a top team but the difference is construction: manufacturing tolerances in thousandths vs eighths.

There are a lot of tricks and tips others will have, but the most direct way to reduce drivetrain problems is using properly applied precision machining*.

Lo and behold it occurred to me: I was unable to make identical sides for the claw. The roller claw shafts were not parallel, causing the pulleys to be slightly askew from each other, causing the belt to walk, causing the belt to rub against the thin pulley wall, causing the wall to eventually fail.

*Clarification: Teams can and do get along fine without precision machining… but knowing how (and why) to compensate for that starts with understanding that it is missing.


Great topic!

Kinda points to the larger question of “where does engineering knowledge come from?”

It might be worth perusing this spreadsheet of 30 design anecdotes that @AdamHeard collected in this 2012 thread.

We don’t have west coast but we did have belt drive last year

We basically took Andymark’s KOP chassis and use the same wheel spacing and pullies for that. It turned out really well for us

1 Like

That’s great ingenuity! Along those lines, I have long thought that a COTS supplier should make belt-and-pulley spaced gussets (similar to how these Vex gussets can be used to set exact gear spacing between shafts).

Some COTS gussets do just happen to work well enough for belt spacing. For example, in the 84t 20dp spacing of the 217-4183 gusset, you could do:

  • 30:18 ratio with a 46t 5mm pitch belt
  • 36:24 ratio with a 52t 5mm pitch belt
  • 42:18 ratio with a 53t 5mm pitch belt



Can’t speak to belts, but for chain I’ll echo what people have said above: make sure your side-to-side alignment is good, and that your spacing is proper.
I always add 0.012" to #35 runs and 0.018" to #25 runs, as long as they’re longer than ~3" C-C. I’ll do half that on shorter runs. I don’t think I’ve had a drive chain pop off the sprocket since 2016.

Anecdote: In 2018, our chain was so loose you could hear it slapping the bellypan every time we moved. However, the chain never jumped off the sprockets because we had the side-to-side alignment perfect, and even though it sounded bad, we weren’t off by all that much - it was just a very long run of chain. We had bearings in the drive tube and had the sprocket hubs flush against the bearing, ensuring that our spacing was the same between drive sprockets.

If you don’t have precision tooling (mill/CNC), go with chain and use cams to tension. Chain is more forgiving than belt on spacing. Be wary of versablocks going crooked like juchong posted above.

In terms of strength, the flowchart looks like this:
9mm wide 5mm HTD < #25 chain < 15mm wide 5mm HTD < #35 chain
GT2 and GT3 belts are significantly stronger than HTD. I believe 9mm wide GT3 5mm pitch belt is stronger than #25 chain.


How is the chain adder (.018 or .012) that we seem to have accepted a constant number and not a function of the number of links?

I understand that its been field tested and the same numbers have worked fine for me, just wondering what I’m missing here…


It’s both a constant and a function. IIRC, part of it is slop in bearings and sprocket bores, and another part of it is chain wear. Most people say to “test it yourself” which is vaguely annoying because I know of few teams that have actually done this, and fewer that have released data.

The 0.018" adder is actually based off drivetrain distance using hex bearings (2363 released this data in 2015 if I remember right?), which have more slop than thunderhex bearings. I’ve found that the chain wears in fast enough that the chain quickly becomes a good fit, and can even become loose on very long runs. Any run that’s longer than a drivetrain (think elevators and such) probably deserves a tensioner.
I haven’t used a ton of #35 chain outside of arms, where I like to tension it anyway, but it’s very difficult to push #35 chain off a sprocket.

Chains will derail from sprockets will they aren’t properly tensioned and/or the sprockets are not co-planar. The advent of COTS shaft spacers and 3D printed spacers has made the later concern far less common than it used to be, since you can get consistently sized spacers to ensure your sprockets are the same distance away from your bearing. In the older days, if teams didn’t have a lathe (or didn’t have the time to lathe) for spacers, they would cut them by hand and could sometimes end up with mis-aligned sprockets resulting in increased odds of throwing a chain. There are many different ways to tension chains, but for WCD’s in particular the advent of the VersaBlock and WCP cam tensioner have made maintaining proper chain tension relatively easy.

Chains will break when overloaded beyond their rating. Generally speaking both 25H and 35 chain will be more than adequate for all but the most extreme cases in FRC drivetrain loads and should not snap (8" wheels can push beyond 25H, especially in games with shock loads like 2016). Running a traction limited drivetrain can also help prevent chains from experiencing excessive forces.

The teeth on aluminum (or steel) sprockets (either 25 or 35) really shouldn’t break off in FRC. I’m surprised to hear this happened to 610 in your post. I know some teams have played with 3D printing sprockets recently, but I don’t know what types of results they’ve achieved and what levels of load they’re rated for.

Also, AVOID HALF LINKS AND MASTER LINKS WHENEVER POSSIBLE. Get a chain tool like these (25 35) instead of using master links or offset links.


For what it’s worth, 254 used two different offsets in 2017, when their drive gearboxes weren’t centered on their 6-wheel WCD:

1 Like

Link to writeup

However, since this time, we’ve joined Brando’s anti-adder camp. We always either (a) use a tensioning system on chain runs or (b) follow the measurement procedure described in the post to get the center-to-center distance for the hardware actually being used.