Belt Driven Drive Trains

[Sh1ine]

Hi all. We have been running belt driven drive trains for about 5 seasons now. We run 15mm belt on a 24 tooth pulley inside of a 1.5 x 3 tube. (we have never had a belt break) I am interested in trying to fit our pulleys into a 2 x 2 tube, using a 24 tooth pulley would be cutting it really close, to the point where I don't think it is worth even prototyping. Has anyone used a VEX 18 tooth pulley with a 15 mm wide belt? If so how did it work? Did you have belt issues? If you have used something smaller than 24 tooth but larger than 18 I would be interested to know about that too. There is a Google Sheet that has some infomation on it, but it seems like it hasn't been undated for some time, so any recent insight would be helpful.


Thanks!

[asid61]

How close is close? If you measure your belt around the pulley and it's less than the tube, then you should be okay even if it's very close.
I know 192 did a 2x2 Belt-in-tube drivetrain in 2014.

[Aren Siekmeier]

Originally Posted by asid61

How close is close? If you measure your belt around the pulley and it's less than the tube, then you should be okay even if it's very close.
I know 192 did a 2x2 Belt-in-tube drivetrain in 2014.

See the spreadsheet he linked. 192 has traditionally run 22 tooth pulleys. 24 tooth pulleys have a 1.654" OD, which inside a 1.75" tube interior leaves no room for even a 0.1" drop. And you want some breathing room for the belt.

The numbers tell me 18 tooth pulleys are risky. You are definitely beyond the torque rating at stall, and would probably see ratcheting and serious wear under full forward to full reverse or other high load conditions. See page 18 of this document. 22-24 tooth pulleys are only ~50% better, but maybe that's enough. 192 seems to have success with 22.

This may also be of interest. That's a 2"x1.5" tube but it looks like the belts are only 9mm wide and the pulley looks like 24 teeth. Looks like they get away with this by clearing away the top and bottom wall of the tube. 1625 went on to field a similar drive (plus some sweet drop down "lobster" modules to drive sideways) for competition in 2011.

[AustinSchuh]

Originally Posted by asid61

How close is close? If you measure your belt around the pulley and it's less than the tube, then you should be okay even if it's very close.
I know 192 did a 2x2 Belt-in-tube drivetrain in 2014.

To be honest, if clearance is the only thing you are worrying about, then you have all your priorities wrong.

971 always starts with the following document: http://www.gates.com/~/media/files/g...nual.pdf?la=en Bookmark it and read it. It is awesome and we've found it pretty accurate over the years.

The biggest challenges in a belt drive system are belt life, strength, and proper tension. We've struggled most with proper tension over the years. We also try to design for 0 belt failures over the year, since the drivetrain is such a key part of the robot. Our year is 3 competitions, and 3 offseasons on one robot, and many many weekends of practice on the other robot.

Lets work through some of the numbers. I'm going to start with numbers from 971's drivetrain, which I know well, and then lets extrapolate.

For the last couple years, we have been running 24 tooth pulleys with 3.5" wheels. In 2012 and 2013, we ran GT2 belts, 9mm wide, and they wore out by the end of the year and needed to be replaced. In 2014, we ran GT3, 9mm wide, and when they weren't perfectly tensioned, they broke. Too little tension and the teeth ripped off. To much and we broke the tensile element. It was a fine line. To me, that meant that we were on the edge of what they are capable of. We've since moved to 15mm wide GT3 so we have some safety margin.

Lets pick GT2, 9mm wide as the 971 suggested torque without a safety factor.
GT2, 24 tooth, 9mm wide -> 75 in-lb of torque.

Stall torques for the other configurations we've used in the past:
GT3, 24 tooth, 9mm wide -> 86 in-lb of torque.
GT3, 24 tooth, 15mm wide -> 159 in-lb of torque.

For grins, lets look at some other numbers. VexPro uses the HTD tooth profile.
HTD, 24 tooth, 9mm wide -> 36.9 in-lb of torque
HTD, 24 tooth, 15mm wide -> 70 in-lb of torque

You can run GT2/GT3 belts on HTD pulleys, and you will get performance somewhere between the two (Gates won't give you numbers on it, but will tell you that it is supported and is better than pure HTD). Food for thought.

So, lets analyze the proposed configuration

HTD, 18 tooth, 15mm wide -> 49 in-lb of torque

My opinion is that if you were to put that on one of 971's robots, you'd destroy the belt pretty fast. I don't have enough info from your original post to run those numbers. Wheel-size is another variable that I've been ignoring. Bigger wheels will require stronger belts.

Some more numbers to get you thinking:
HTD, 22 tooth, 15mm wide -> 62 in-lb of torque
GT3, 18 tooth, 15mm wide -> 101 in-lb of torque

(If you've already run these numbers, then this post will help others understand the tradeoffs we go through in belt selection).

[Joe G.]

1687 and 5400 have been running 30t GT2 with 15mm belts inside of tubes for the past three years with no failures. It's overkill and we know it, but we like zero failure systems. Our resource set does not enable us to take the meticulous tensioning steps that a team running right on the edge like 971 does, so we like the insurance. When we were doing research for our switch to belts, I believe we saw some pretty spectacular failures in 18t systems.

If clearance is your concern, perhaps consider an alternate form of structure. We do full sheet metal chassis builds with a multi-piece tube which contains our belts, and one of the many reasons that we love it is that we can make our channels any size we want, instead of being limited by commonly avaliable extrusion profiles. The tube height is often a pretty ugly decimal, but it doesn't matter for our fabrication process.

[Sh1ine]

Thank you for your feed back. If clearance were the only thing I was worried about I would just go with the 18 tooth pulley and not bother to ask how they would work here.

In my opinion, if you have the machining capabilities, belts are superior to chain. Proper tension is not an issue if you use Center to Center calculators and have the manufacturing capabilities to meet those tolerances. We have a HAAS mill at school and each rail is machined from one piece of aluminum tube. Using our construction technique we are able to keep our belt centers within a few thousands if not better. I think many teams assume chain needs a tensioner so a belts must also. For the purpose of our application adding tension adjustment just adds to the chance that the belts will fail, try Center to Center and you will never go back.

Thank you for running the calculations and directing me to the Gates manual. I need to go through that. I am not an engineer, calculations are not my strong point, but it is something that I need to work on. I know that we are able to run a 15mm HTD belt with a 24 tooth gear for an entire season, plus with out an issue. (In fact we have used one set of belts two seasons in a row without a break) According to your calculations it seems like I could go to a 18 tooth pulley if I went with a GT3 system, which is good food for thought. A 21 tooth pulley would also be an option for us, we like to stick with teeth in multiples of three so that we can hold them in a 3 jaw chuck.

On different years we run different size wheels. Typically we use a 6" Colson wheel, but when we have a flat field we use a 4" wheel. With a 3" tube we get about 1/2" of clearance under the robot. I would love to be able to use a 4" wheel more often, so I want the team to develop a prototype this fall. We have always used a single speed gearbox, in 2014 we had a 6 CIM drive train with 4" wheels that traveled at about 13 f/s. Last year we used a Mecanum drive, so that does not count for this conversation. I will take a look through our old CAD to find a good representation of our drive train tube. It will shed some light to the design.

Again, thank you for your feed back. It is very valuable.

Originally Posted by AustinSchuh

To be honest, if clearance is the only thing you are worrying about, then you have all your priorities wrong.

971 always starts with the following document: http://www.gates.com/~/media/files/g...nual.pdf?la=en Bookmark it and read it. It is awesome and we've found it pretty accurate over the years.

The biggest challenges in a belt drive system are belt life, strength, and proper tension. We've struggled most with proper tension over the years. We also try to design for 0 belt failures over the year, since the drivetrain is such a key part of the robot. Our year is 3 competitions, and 3 offseasons on one robot, and many many weekends of practice on the other robot.

Lets work through some of the numbers. I'm going to start with numbers from 971's drivetrain, which I know well, and then lets extrapolate.

For the last couple years, we have been running 24 tooth pulleys with 3.5" wheels. In 2012 and 2013, we ran GT2 belts, 9mm wide, and they wore out by the end of the year and needed to be replaced. In 2014, we ran GT3, 9mm wide, and when they weren't perfectly tensioned, they broke. Too little tension and the teeth ripped off. To much and we broke the tensile element. It was a fine line. To me, that meant that we were on the edge of what they are capable of. We've since moved to 15mm wide GT3 so we have some safety margin.

Lets pick GT2, 9mm wide as the 971 suggested torque without a safety factor.
GT2, 24 tooth, 9mm wide -> 75 in-lb of torque.

Stall torques for the other configurations we've used in the past:
GT3, 24 tooth, 9mm wide -> 86 in-lb of torque.
GT3, 24 tooth, 15mm wide -> 159 in-lb of torque.

For grins, lets look at some other numbers. VexPro uses the HTD tooth profile.
HTD, 24 tooth, 9mm wide -> 36.9 in-lb of torque
HTD, 24 tooth, 15mm wide -> 70 in-lb of torque

You can run GT2/GT3 belts on HTD pulleys, and you will get performance somewhere between the two (Gates won't give you numbers on it, but will tell you that it is supported and is better than pure HTD). Food for thought.

So, lets analyze the proposed configuration

HTD, 18 tooth, 15mm wide -> 49 in-lb of torque

My opinion is that if you were to put that on one of 971's robots, you'd destroy the belt pretty fast. I don't have enough info from your original post to run those numbers. Wheel-size is another variable that I've been ignoring. Bigger wheels will require stronger belts.

Some more numbers to get you thinking:
HTD, 22 tooth, 15mm wide -> 62 in-lb of torque
GT3, 18 tooth, 15mm wide -> 101 in-lb of torque

(If you've already run these numbers, then this post will help others understand the tradeoffs we go through in belt selection).

[Thad House]

Originally Posted by AustinSchuh

To be honest, if clearance is the only thing you are worrying about, then you have all your priorities wrong.

971 always starts with the following document: http://www.gates.com/~/media/files/g...nual.pdf?la=en Bookmark it and read it. It is awesome and we've found it pretty accurate over the years.

The biggest challenges in a belt drive system are belt life, strength, and proper tension. We've struggled most with proper tension over the years. We also try to design for 0 belt failures over the year, since the drivetrain is such a key part of the robot. Our year is 3 competitions, and 3 offseasons on one robot, and many many weekends of practice on the other robot.

Lets work through some of the numbers. I'm going to start with numbers from 971's drivetrain, which I know well, and then lets extrapolate.

For the last couple years, we have been running 24 tooth pulleys with 3.5" wheels. In 2012 and 2013, we ran GT2 belts, 9mm wide, and they wore out by the end of the year and needed to be replaced. In 2014, we ran GT3, 9mm wide, and when they weren't perfectly tensioned, they broke. Too little tension and the teeth ripped off. To much and we broke the tensile element. It was a fine line. To me, that meant that we were on the edge of what they are capable of. We've since moved to 15mm wide GT3 so we have some safety margin.

Lets pick GT2, 9mm wide as the 971 suggested torque without a safety factor.
GT2, 24 tooth, 9mm wide -> 75 in-lb of torque.

Stall torques for the other configurations we've used in the past:
GT3, 24 tooth, 9mm wide -> 86 in-lb of torque.
GT3, 24 tooth, 15mm wide -> 159 in-lb of torque.

For grins, lets look at some other numbers. VexPro uses the HTD tooth profile.
HTD, 24 tooth, 9mm wide -> 36.9 in-lb of torque
HTD, 24 tooth, 15mm wide -> 70 in-lb of torque

You can run GT2/GT3 belts on HTD pulleys, and you will get performance somewhere between the two (Gates won't give you numbers on it, but will tell you that it is supported and is better than pure HTD). Food for thought.

So, lets analyze the proposed configuration

HTD, 18 tooth, 15mm wide -> 49 in-lb of torque

My opinion is that if you were to put that on one of 971's robots, you'd destroy the belt pretty fast. I don't have enough info from your original post to run those numbers. Wheel-size is another variable that I've been ignoring. Bigger wheels will require stronger belts.

Some more numbers to get you thinking:
HTD, 22 tooth, 15mm wide -> 62 in-lb of torque
GT3, 18 tooth, 15mm wide -> 101 in-lb of torque

(If you've already run these numbers, then this post will help others understand the tradeoffs we go through in belt selection).

Are those numbers actually in in-lb? Every time I've done the calculations, I get somewhere in the range of 480 in-lb of stall torque for a 3 CIM gearbox at 10 ft/s. A CIM by itself has 21 in-lb of stall torque. We've calculated in the past that for a 24T belt, that equates to a pulling force on the belts of about 700 lbs. Am I just doing my numbers wrong, or are those numbers actually supposed to be in ft-lb?

[Aren Siekmeier]

Originally Posted by Thad House

Are those numbers actually in in-lb? Every time I've done the calculations, I get somewhere in the range of 480 in-lb of stall torque for a 3 CIM gearbox at 10 ft/s. A CIM by itself has 21 in-lb of stall torque. We've calculated in the past that for a 24T belt, that equates to a pulling force on the belts of about 700 lbs. Am I just doing my numbers wrong, or are those numbers actually supposed to be in ft-lb?

They are in in-lb.

You aren't going to see 480 in-lb per drive side, the wheels will slip first. 150 lb / 2 * CoF * wheel radius is at most 450 in-lb with 6" wheels and a phenomenal CoF of 2.0. A more reasonable max is 200 in-lb (4" wheel, CoF of 1.3). You'd have to lock the wheel to get more than that.

[Thad House]

Originally Posted by Aren Siekmeier

They are in in-lb.

You aren't going to see 480 in-lb per drive side, the wheels will slip first. 150 lb / 2 * CoF * wheel radius is at most 450 in-lb with 6" wheels and a phenomenal CoF of 2.0. A more reasonable max is 200 in-lb (4" wheel, CoF of 1.3). You'd have to lock the wheel to get more than that.

Even still, that is 4x greater then what the numbers Austin posted say belts can take. If a 15mm 24T pulley can only take 70 in-lb, teams would be snapping them all the time even with 2 CIM drives.

[GeeTwo]

Originally Posted by Thad House

Even still, that is 4x greater then what the numbers Austin posted say belts can take. If a 15mm 24T pulley can only take 70 in-lb, teams would be snapping them all the time even with 2 CIM drives.

First of all, exceeding the torques specified in the Gates guidelines will not cause the belt to break, but to jump teeth.

Secondly, the rated stall torque of a CIM is not achievable on an FRC robot. If you were to stall a CIM with 12V applied at the motor, you would draw 133A, which would probably trip the 40A breaker before you jumped two teeth. More realistically, the stall resistance of the CIM is about 12V/133A 91mΩ. The internal resistance of the battery is about 12mΩ, plus about another 2mΩ for the battery leads and power distribution panel, and another 8mΩ per branch for wiring and the motor controller. For two CIMs and a 13.6V battery charge, I calculate that the voltage on the CIMs would be about 9.75V, which would reduce current and torque by nearly 20%, and still trip the breaker quickly. A more reasonable current draw for a few seconds is 60A per CIM, which would give you less than half the stall torque. When you start stacking more CIMs on the drive train, the limit becomes the 120A breaker, or drawing the battery voltage down below roboRIO brownout.

Finally, I believe that the numbers in Gates' document are applicable for exact C-C calculated pulley spacing. I suspect that many teams with adjustable tension in their belt setups over-tension them. This will shorten the lifespan of the belt (from thousands of hours to hundreds or even tens), but increase the torque that a given size sprocket can deliver to (or receive from) the belt by providing a controlling tension on the "return side" of the belt. With enough tension in the belt, the limiting factor really does become the tensile strength of the belt and sheaves.

[Richard Wallace]

See test data that Ether and I measured and reported here a few years ago.

Stall current or very close to it will be conducted for a few tenths of a second.

[GeeTwo]

Originally Posted by Richard Wallace

See test data that Ether and I measured and reported here a few years ago.

Stall current or very close to it will be conducted for a few tenths of a second.

Richard (and Russ),

Thanks for the link! I may have to update my previous comment to three or four jumping teeth rather than one or two. In any case, current even approaching stall values was momentary at best, and wheels were obviously free to rotate and provide some back-EMF from the motors within a few tenths of a second. Was this conducted with a single CIM behind a 40A breaker, itself behind a 120A breaker? If so, do you have any data on what would happen as you added more load to create a 2, 4, or 6 CIM drivetrain?

[AustinSchuh]

Originally Posted by Sh1ine

Proper tension is not an issue if you use Center to Center calculators and have the manufacturing capabilities to meet those tolerances.

See below.

Originally Posted by Sh1ine

Thank you for running the calculations and directing me to the Gates manual. I need to go through that. I am not an engineer, calculations are not my strong point, but it is something that I need to work on. I know that we are able to run a 15mm HTD belt with a 24 tooth gear for an entire season, plus with out an issue. (In fact we have used one set of belts two seasons in a row without a break)

You don't need to be an engineer to understand this stuff (though it does help with some of the language and what they care about). We have students every year who become belt experts and work through the calcs.

Knowing that a 24 tooth pulley, 15mm wide, HTD works for your specific application is good data. Make sure that is from a year with heavy D. In the end, the design of your drivetrain, the aggressiveness of your drivers, and the duration of your season all play into whether or not a specific belt design will be good enough for you. What works for us may not work for others.

Make sure to scale the torques accordingly with wheel size changes. Bigger wheels will require higher torques to drive.

Originally Posted by Thad House

Even still, that is 4x greater then what the numbers Austin posted say belts can take. If a 15mm 24T pulley can only take 70 in-lb, teams would be snapping them all the time even with 2 CIM drives.

Those numbers are in in-lb. If they look scary, you are right. We were hesitant to try belts for that reason, and only after talking with a Gates rep and 987 were we willing to even try it. We are really on the edge. Wheel traction and robot weight transfer are all key parts of the equation. We make sure to drive our center wheel with a gear, and then spread the torque to the outer wheels from there with belts. This reduces the torque requirements some. See below for more analysis.

Originally Posted by GeeTwo

First of all, exceeding the torques specified in the Gates guidelines will not cause the belt to break, but to jump teeth.

That's not fully true. The Gates guidelines are for 'infinite life', or close to it. You can exceed that torque for short periods of time, but you will dramatically shorten the lifetime of your belt. The exact failure mode of course all depends on the tension of the belts. Too little tension (for the desired load), and you'll see belts jump. To much tension and you risk breaking the tensile elements.

Not to digress too much, but you can tell a lot about how a belt is performing by what happens when it fails. There are 2 classic failure modes. 1) The teeth fall off the belt. This means that you were under tensioned. Each time the belt jumps teeth, it fatigues the teeth. This eventually causes the teeth to de-laminate from the tensile elements and fall off. 2) The 'valleys' in the belt, the space between the teeth, gets crushed and the fibers break. This is a cleaner break. This means that the belt is over-tensioned. 3) I've actually seen failures where both happened at once. This tends to mean that you are just putting too much power through the belt, and can't tension it enough.

Originally Posted by GeeTwo

Finally, I believe that the numbers in Gates' document are applicable for exact C-C calculated pulley spacing. I suspect that many teams with adjustable tension in their belt setups over-tension them. This will shorten the lifespan of the belt (from thousands of hours to hundreds or even tens), but increase the torque that a given size sprocket can deliver to (or receive from) the belt by providing a controlling tension on the "return side" of the belt. With enough tension in the belt, the limiting factor really does become the tensile strength of the belt and sheaves.

Gates does _not_ recommend exact C-C. They have some pretty choice words about it. Please read them. It will make you think twice.

Exact C-C only works if your application is forgiving. There are a bunch of manufacturing tolerances that come into play. Actual pulley diameter, pulley runout, and belt pitch tolerance are all key parts of the equation. The tolerance of the belt pitch can be enough for longer runs that some belts will run over-tensioned on exact C-C drives, and some belts will run under-tensioned. We haven't viewed it as worth the risk to do a study and see if we could get away with exact C-C on a drivetrain. For our manipulator gearboxes, we over-design things enough that we feel comfortable doing an exact C-C for some systems.

Read Page 67 for what Gates has to say about exact C-C. Their manual is very good.

Page 65 has the equations for determining the recommended static tension for a given operating point. It basically boils down to the ratio of the tension in the loaded portion of the belt to the unloaded portion of the belt at your desired operating point. You want something like a 6:1 ratio (check the equations to be sure). Nowhere in there does it say that one tension is best for everything.

There are 3 main ways to test belt tension that I know about. 1) pluck the belt and listen to the note it makes. You can compute the natural frequency of the belt and determine the desired tension. 2) apply a known force to the center of the span and measure the deflection. 3) buy a fancy tool. We actually use 1. It was fun to see the students all running around with tone generators on their phones plucking all the belts and trying to match the pitch. Everyone learned a lot, and the tools are cheap.

At one point, I got Gates to give me rough static break numbers for the tensile elements. The numbers are comparable with #25 chain for a 9mm wide belt, GT2. Food for thought. That is the reason why we can actually get away with the belts in a high peak load system like a DT.

If you actually respect the ratings, #25 chain is only good for a working load of 140 lbs. That works out to 122 in-lb for a 22 tooth sprocket.

Belts are pretty cool. We love their performance, damping, and sound characteristics. Our experiences and safety margins may be a bit higher than other teams due to the long and hard life of our robots. In 2014, we probably approached 100+ hours of runtime on each of our robots. Most of that was hard focused practice under heavy D. My goal is to be able to design systems for FRC that survive those lifetimes from the beginning without over-designing too much.

[Aren Siekmeier]

Originally Posted by Thad House

Even still, that is 4x greater then what the numbers Austin posted say belts can take. If a 15mm 24T pulley can only take 70 in-lb, teams would be snapping them all the time even with 2 CIM drives.

I'm suggesting even 200 in-lbs is probably very high (I said max). Remember it's also assuming ALL of the robot's weight and drive power is going through ONE belt (not two or three, or through a direct driven wheel as well). Typical loads are probably closer to 100 in-lb, with rare edge cases pushing it no higher than 200. 971's 24 tooth GT3 can handle that.

2 or 3 CIMs isn't the issue. Force on the transmitting belt maxes out when the wheel starts slipping, regardless of the transmission's stall torque.

And we have seen that 24 tooth HTD drives can and often do slip under full load. Many belt drives either use GT2/GT3 or use larger pulleys (30 or 36 tooth).

[jkelleyrtp]

Stemming from this, is there a maximum length before it is necessary to drive to another pulley and belt system? I could imagine that longer stretches of belts would be more likely to hop teeth and shorten the lifespan, but I see a few teams running belts through their tubing for all 3 wheels.