I'm having a Great Big Cargo Bike Sale! Only three ex-demo Riese & Muller cargo bikes left at big discounts.

Belt Drive Theory

Belt drives have been around on bikes for a while, after several false starts – they’re now very common on electric bikes, cargo bikes, and I’ve been fitting them to Bromptons for years. I get quite a few questions about belt setup, especially on the Brompton, so this article lays out how belt drives work.

Fundamentally, belts and chains both work on the same principle – the crank is turned by your legs, this rotates a toothed cog which pulls on a matching flexible transmission element, which then pulls on the teeth of a cog in the back wheel, which rotates a hub gear or the wheel directly. If you think about it, the top part of the transmission element is doing all the work – all the bottom part has to do is get back to the rear cog without dragging on the ground. The whole transmission element is continuously rotating, but it’s very helpful to think of it as these two top and bottom halves.

An Aside About Wear: Chains and belts wear differently. When a chain wears, each pin and roller loses a tiny bit of metal, so the effective pitch of the chain – the distance between each link – increases. We often say the chain “stretches”, though the metal itself isn’t stretching. With belts, the teeth do also wear, but critically the kevlar strands holding them together don’t stretch, so the effective pitch of the belt stays the same.
This affects how the system eventually fails. With a chain, the increase in pitch mean that the load is no longer spread over most of the teeth of the cog, it’s concentrated on one or two – the cogs also wear which means this takes longer to happen, but eventually the chain will “skip”, the load on one tooth forces the chain to hop over the tooth.
With a belt system, the pitch doesn’t increase, so the load is always shared over most of the teeth of the cog. The whole cog wears equally, until it no longer has rounded tooth profiles, it has sharper edges that eventually cause the belt to wear. With chain drive, the chain wear is the critical factor – with belt drive, it’s cog wear that is critical. This is partly why belts last a lot longer than chains.

So the important bit is to prevent the chain or belt from skipping – hopping over the teeth at the back. (The front cog is much larger, so much less likely to skip). This really comes down to how the bottom chain or belt is guided onto the cog. In a simple single-speed chain drive, there are no guides – the cogs are simply positioned (usually with track dropouts) so there’s not enough slack in the chain to allow it to skip. In a modern derailleur chain system, the top pulley of the derailleur acts to guide each link of the chain onto a tooth of the cog, and prevents it skipping.

In a belt system, it’s a bit more critical, because belts are more prone to skipping – the teeth are rounded and a bit squidgy, so they want to skip more easily. So the traditional way of preventing skipping was to set the tension of the belt (equivalent to setting the distance between the cogs) high enough that the belt couldn’t skip, but not so high that it causes excessive drag or wear.

Sometimes, even with the best setup, it’s impossible to prevent skipping just by adjusting belt tension – if the rear cog is too small, it can skip, even if tension is perfect. So a “snubber” is used – a simple roller that, like the top cog of a derailleur, guides the belt onto the cog and prevents it from hopping over a tooth. This is needed for Rohloff gearing, for instance, because Rohloff hubs need quite small cogs.

An Aside About Efficiency: If you pick up a belt and a chain, it seems simple – the chain wants to fold itself up with almost no resistance, the belt needs a bit of effort to squeeze it into a smaller loop. You’d think it would make sense that a belt is less efficient than a chain – the chain just folds around the back cog, the belt needs a bit of force to do it. But that’s to ignore the effect of tension – a chain under tension, like all the force from your pedalling, no longer wants to bend easily. The effect is so big that, with a derailleur system, there’s three times as much drag on the small cogs compared to the big cogs – the chain has to bend more to get around the small cog.. So efficiency is a bit more complicated – in reality, there probably isn’t much difference between the average belt and the average chain.

So, having a frame with some way of precisely setting the chainstay length (the distance between front and back cogs) works well – you can fine-tune the belt tension precisely, and there are fancy tools to do that. The problem arises when you have frames where the chainstay length changes – full-suspension bikes, or folding bikes. For them, there’s no way to set it up so the belt tension is perfect for all situations. We’re back to the situation of a derailleur bike, and need a tensioner. This setup is now common, often with a tensioner near the front, and the idea is that the tensioner keeps the belt at about the perfect tension so it won’t skip (though a snubber might be needed as before).

A full suspension bike with a front tensioner and a rear snubber

An Aside About Reverse Bends: For a long time, Gates said that you simply should never bend a belt backwards. This is still a common comment I get about belt drive Bromptons. But a quick look at the above picture shows that you can’t have a tensioner without bending the belt backwards, so a while ago Gates relaxed the rule, and said that it is fine to do as long as the pulley diameter is large than 45mm (quite a few manufacturers ignore this recommendation as well).

But that raises a question: If the tensioner is keeping the belt from dragging on the ground, and the snubber is stopping the belt from skipping, then is belt tension really all that critical? We’re talking here about static tension of course – the tension in the belt when it’s on the work stand. In use, the tension is very different because of the pedalling forces.

This leads to where I’ve got to with the Brompton belt setup – there’s a snubber at the back that stops the belt from skipping, and the tensioner just takes up the slack to compensate for the changing chainstay length as the rear triangle moves. If you test it with a traditional belt tension measuring tool, the tension seems low, but that’s because it doesn’t have to be high to prevent skipping.

I still have sliding dropouts on the frame, but these aren’t for setting belt tension, they’re just to take out most of the slack. The snubber I use is a normal Brompton rear roller – a use it isn’t designed for , but it’s cheap, lasts a long time, and it is easy to find spares. The tensioner spring is stronger than normal, to compensate for the force needed to make the belt bend around the floating pulley when the bike folds.