Now that I’ve seen belt drive in action on a couple of “city bikes” in town, I can’t seem to let this one go. On an upright bike, belt drive is smooth, reliable, light, and silent. On a recumbent it has the potential to solve what for many is their bane: a long, greasy, noisy, unwieldy chain. This post is my attempt at taking a closer look to see what might be preventing the adoption of belt drive for recumbents and, specifically, ‘bent trikes. I’ll discuss the issues I see, many unique to ‘bents. Then I’ll see if I can figure out what it would take to adapt my current trike to belt drive.
I doubt there’s much market for a single speed recumbent. Without an internally geared hub that’s what you’d have because belts can’t tolerate the lateral misalignment and effective length changes inherent in external gears. So the combination of recumbent and belt drive necessarily means IGH. For me that’s a good thing, I’ll probably go IGH even without belt drive if the Alfine 11 works out. But in a segment that already seems slow to adopt IGH, this requirement might be another reason for a trike manufacturer not to go there.
And then there’s gear range. In my gearing post it was complicated enough figuring out how to get the range I wanted. It’s even harder with belt drive because the choices for IGH and cog sizes are far narrower. Well, it’s actually easier, but the odds of a successful outcome are less. For example my goal of combining a Schlumpf Speed Drive with the Alfine 11 appears to be a non-starter. Gates is the only belt manufacturer that I know to have product in current production for bikes, and their system requires a relatively high tension to prevent teeth skipping in the cogs. Schlumpf states that such high tension is not compatible with their drives because it causes excessive bearing wear. I don’t know if the Schlumpf and Alfine have drastically different bearing designs, Shimano is just ignoring the problem and so the Alfine may wear poorly with the Gates system, or if Schlumpf is just trying to differentiate their own belt drive system that requires less tension.
What to do? One solution is to use the Schlumpf belt drive instead of Gates. Schlumpf states that their drives are compatible with Alfine when using a chain, I don’t know why this wouldn’t also be true with a belt. And it seems they would have to support Shimano IGH if they want their system to be successful. Unfortunately I don’t see much evidence that the Schlumpf belt drive system is much more than an idea at this point. Another possibility is to use a Rohloff hub which provides more range than the Alfine-11 and is compatible with the Gates system. For me, I’d seriously consider living without high gearing if it meant I could add belt drive to my Sprint Di2 trike.
Adjustable Leg Length
Perhaps the biggest challenge with ‘bent belt drive is that in a typical recumbent, the rider’s leg length is directly tied to chain (drive line) length. This is annoying with a chain but nearly a show stopper with a belt. Tension is critical in a belt drive system and they can’t easily take up slack introduced by adjusting the boom length. Common techniques like spring tensioners and chain guides are expressly called out by Gates as not compatible with their system.
All is not lost. If you look at an ICE trike, for example, their three-piece frame design provides an obvious way to take up some amount of slack. The rear section of these trikes slides into the cruciform section and is secured with a couple of clamp bolts. There is a certain acceptable range for how far in you can slide these pieces so that the seat mounts properly. In my experience this range is fairly wide and could be made wider with different seat brackets. As it is, it’s certainly wide enough to provide the 12 mm Gates specifies (minus 10 to provide enough slack to install or remove the belt and plus 2 to tension the belt) without resorting to horizontal dropouts like you see on many upright frames. I think with some attention, or maybe even as it is, this feature alone could account for a fair amount of adjustability in leg length. Note that I’d invest in the quick release option for the clamp bolts because changing the rear tire would require cracking these. And there might be need of a more sophisticated way to make fine changes in belt tension.
Trice HDQ (Q with HD Adventure Cruciform)
Without Seat and Rack
The above sounds promising but seems unlikely to provide a very high range of adjustability. Another technique is to provide some fore-aft adjustability in the seat, like on my wife’s TerraTrike Cruiser. In theory this could completely solve the leg length problem but I think it comes at a fairly high cost—from a handling stability standpoint I don’t think you want to be moving the center of mass around too much. I certainly notice a change in handling when I do this on the Cruiser and it’s perhaps telling that few trike designs seem to do this—most seem to favor the “infinitely adjustable boom.” But maybe a combination of some range of fixed boom lengths (again, like on the Cruiser), seat adjustment, and fine-grained boom adjustment could limit the number of belt lengths required to cover all rider sizes from nearly infinite (i.e. custom) to a manageable number of pre-defined lengths.
Competing, non-complementary requirements and design limitations. Finding a middle ground that works is just engineering. But I think it’s safe to say that with belt drive, your ‘bent is going to be more personal than ever.
Two holes in frame and four in bracket provide eight seat positions. Boom extends only in pre-defined chunks, using extenders.
Bent Drive Line
Most ‘bents require a bend in the drive line to get around the frame, some more than one. The traditional way to do this is with idlers on the drive side and either idlers or a chain guide (ICE uses a chain tube) on the return side. Things are more complicated with belt drive—chain tubes can’t be used for the return line because the belt is under relatively high tension. The typical side-by-side pair of idlers can’t be used because the belt won’t tolerate the lateral misalignment required. Finally, one of the idlers required to “bend” the belt is on the back, smooth side of the belt and effectively bends it “backward.” I’ve not seen anything in the Gates documentation that addresses this and I haven’t seen an example of doing this with the Gates system. I suspect they might take issue with it, even though the belts in my car’s engine do this sort of thing all over the place. The closest thing I’ve seen with Gates is the “snubber” used to increase belt wrap on the rear sprocket of Rohloff hubs. The snubber may be doing something similar to the the I-1 idler in my “Single Belt” diagram below. That is, bending the belt backward on the smooth side. But it’s doing it on the return side, not the drive side like my drawing. And the Gates documentation says the snubber shouldn’t be in contact with the belt, so I guess I don’t understand when it comes into play. Maybe there’s a lot of slack on the return side of the belt?
Following is a possible way to route a single belt on my HDQ using the Gates belt drive system. For the sake of this discussion, pretend the rear is not suspended. The I-1 idler would have a smooth surface, much like the stock ICE idler. I imagine the I-2 idler would need to be a Gates cog to avoid damaging the teeth of the belt. The smallest cog I see on the Gates site is 22 teeth which is the diameter I show here. One of the idlers could be adjustable perpendicular to the belt to provide fine-grained control of belt tension.
If the idlers in the single belt solution are a show stopper, another option is two belts with a mid drive. The mid drive is a pair of connected cogs at a location on the frame with line-of-sight to both the front and rear cogs. This solves the problem of “bending” the belt but adds two new problems: The mid drive will see greater force than the idlers in the single belt solution and so will have to be made more substantial. And the primary way of compensating for changes in belt length due to leg length adjustments in the front boom, by moving the rear frame relative to the cruciform, is eliminated. A mid-drive design moves the drive line at the bottom bracket away from the optimal position but I think this can be compensated by a longer spindle length.
With my trike, there are at least a couple possible locations for a mid drive. In the first diagram it keeps the belt below the cross tube and out of the way of the seat and handlebars. The location in the second diagram has the advantage of being mounted directly on the frame tube, which should be easier to fabricate, but greatly complicates the cockpit area by running a belt through it. At least with a tensioned belt, it won’t be flopping around and shedding oil. Still, the best bet seems to be a single belt with idlers and no mid drive.
Two Belts, Mid Drive Under Cross Tube
Two Belts, Mid Drive Over Cross Tube
The Gates documentation dedicates an entire section to the importance of frame stiffness. Recumbents may be particularly vulnerable because the frame is so long and the boom is often cantilevered out from the main cruciform. Flex can cause the belt tension to decrease, leading to skipping and possible derailment. Gates offers engineering consultation to frame designers for this and other design issues. This points to another reality of belt drive: because of the unique requirements placed on frames, belt manufacturers really only want to talk to bona fide bike manufacturers. The rest of us have to rely on the bike manufacturers whenever they decide to bring belt drive to market, or whatever information the belt drive manufacturers release directly to consumers. Which in the case of Gates is quite a lot, actually.
It’s possible that frame flex alone may be the real deal killer for belt drive on recumbents. I can actually see my boom flex when I crank on the non-drive (left) side. Combined with a hard right turn a belt might be in trouble. An interesting future investigation would be to quantify the flex on my frame and compare with the Gates specs.
Even the best rear suspension design runs the risk of belt shrink or growth, causing fluctuations in belt tension. Common ways of managing it, like spring loaded tensioners, are specifically not recommended by Gates (at least). Somewhat surprisingly, Gates doesn’t rule out suspension with their belt drive system but does relegate the design to direct consultation with them.
I don’t have a problem with this—I’m leaning away from suspension on my next trike anyway. But companies like ICE that have made full suspension a big part of their offering might not be too keen on a new technology that complicates that story.
Unlike a chain, a belt can’t be “broken” to get it around the chain or seat stay tube of typical frame designs. This makes belt drive incompatible with these frames. A common solution is to provide a coupler like the S&S coupler in the tube that requires the belt to pass through it. Frame manufactures provide the coupler as part of a belt-ready frame design, I doubt it would work to retrofit a coupler into an existing frame (but I haven’t looked into it).
My current trike has a more elegant, if accidental, solution. Ignoring for the moment that it was designed to provide rear suspension, the arched rear section of Trice Q era trikes is inherently belt friendly. No chain stay to break. Suspension aside, I have to admit I was a little sad when I learned that ICE was going back to a conventional braced rear section on the new (and current) line of trikes. The arch was unique and I thought looked pretty cool. Function combined with aesthetically pleasing form—that’s engineering gold.
Unfortunately the shortest belt in my calculations, either of the mid drive configurations, is longer than the longest belt currently offered by Gates. About 1650 mm vs. 1452 mm. The single belt configuration requires around 3349 mm, more than double the biggest size available. I’m not counting the 2000 mm tandem belt offered by Gates because it has different pitch.
I don’t know if there is a design limit for belt length or if there just isn’t any demand for longer belts, but this would seem to shoot down hopes of conversion to belt drive any time soon, at least with the Gates system.
Update: Gates has a line of industrial timing belts called PolyChain that run as long as 4480 mm. Certain versions of these belts are constructed nearly identically to the Carbon Drive belts and seem to be their very recent predecessors. At least one person is running this system for the synch drive on their tandem. With this information I’d say it’s not likely a design limitation that’s preventing the development of longer belts. Also note that on a quick glance it seems it may be possible to satisfy both the single belt and mid-drive configurations with the lengths available with PolyChain. Might be worth some further investigation.
Belt Design and Availability
I suspect the biggest reason we don’t have belt drive on production ‘bents today boils down to the difficulty and poor economics of producing large, custom belt lengths. Typical recumbent designs require this to cover various leg lengths. But my understanding of the Gates manufacturing process is that the belts are formed as a complete loop from a contiguous mold rather than cut to a specific length from a spool of belt and then spliced into a loop. There are undoubtedly ways around this, I talked about a few, but this constraint combined with low production volumes in one of the smallest niches in the universe of human powered vehicles might make adoption of belt drive in recumbent trikes very slow indeed.
Belt Drive for my Trike?
For this exercise I made some simplifying assumptions. First, I limited it to what I could do with the trike in my garage, a 2008 ICE Q with a 2014 ICE Adventure HD cruciform. It turns out this trike design has some advantages when it comes to belt drive:
- The arched rear frame design allows a belt to be installed without needing a break in the frame,
- The modular three-piece frame design provides an easy way to take up slack when installing the belt or making small adjustments in leg length.
- ICE puts about a 6mm right offset in the rear dropouts to allow the rear wheel to be built with the rim centered between the flanges, eliminating the need to dish the wheel. This partially compensates for about a 4mm left offset introduced by the Di2 version of the Alfine-11 hub. The result is very close to my current 45mm drive line.
Another simplifying assumption was to limit the exercise to a Gates belt drive with Alfine-11 Di2 rear hub. This turned out to be a huge simplification because Shimano/Gates currently only offer one choice for front and rear cog and belt type. CDX CenterTrack belt (as opposed to their CDC MudPort line of belts) and 50/28 tooth cogs.
The gear inch numbers for this setup should look familiar to anyone who read my gearing post: the 50/26 combination provides almost exactly the same range as the 32/18 combination I was looking at. This isn’t too surprising as it sets the gearing at the load limit specified by Shimano for their Alfine-11 hub.
Gates makes a lot of claims about the benefits of belt drive. One of them is fairly easy to evaluate right now: weight savings. Comparing a single belt Alfine-11 Di2 configuration with an equivalent chain setup, it looks like belt drive is about 200g lighter, before accounting for the weight associated with mounting a second idler. So maybe a wash, or a little heavier for belt drive. I can live with that.
|Chain Drive||Belt Drive|
|Front Cog||40g (32t – Deore)||120g (50t)|
|Rear Cog||20g (18t— total guess)||79g (26t)|
|Belt / Chain||630g (2.5 standard lengths DuraAce 10 spd)||245g (extrapolated)|
|Second Idler||n/a||50g (22t)|
Taking some rough measurements from my ICE HDQ trike and consulting the Gates documentation, I was able to produce the belt routing diagrams shown previously. Based on these drawings I came up with a belt length of around 3350 mm for my single belt configuration. As stated earlier, this is a long way from what Gates currently produces. So it looks like a number of things may dash my hopes of converting to belt drive anytime soon:
- Belt availability in a really long, specific length that might only work on a specific trike design/rider combination
- Idler in the single belt configuration that bends the belt backward
- Suspension on my current rear frame section
- No frame break in my likely future rear frame section
- Frame flex
Based on my (admittedly limited) experience with some belt-driven “city bikes” here in Boise, I’m convinced many of the wondrous benefits claimed by Gates are at least potentially valid. In their words (with the most flagrant marketing gibberish removed) these benefits include:
- Clean – No Lubrication means no greasy build up of dirt, keeping your hands, clothes, and children clean.
- Smooth – The Carbon Drive system eliminates the metal-to-metal contact of chains, delivering an unparalleled smoothness that you need to feel to believe.
- No Noise – The Carbon Drive belt is virtually silent in most riding conditions.
- Light Weight – the entire system, 2 sprockets and a belt, weighs less than just a typical chain.
- Life – Testing has shown the Carbon Drive belt to last more than twice the life of chains.
- Consistent Efficiency – the Carbon Drive belt provides consistent efficiency over the lifetime of the drive. In contrast, chain efficiency decreases with wear and dirt.
If true, some of these benefits, like Clean, Smooth, No Noise, and Life are potentially more of a benefit to a ‘bent than an upright due to the bent’s long and unwieldy chain. Note that their efficiency claims aren’t “more” or even “the same”, but “constant.” I’m curious what that translates to in actual use but would be willing to give it a try. And it looks like the weight will be about a wash compared with a chain.
Though I’m not too hopeful belt drive is in the cards for the Sprint Di2 build, I’m still not convinced the identified problems are more than tricky engineering challenges for a trike that is intentionally designed for belt drive. The next step to validate this claim would be talking to a Gates engineer, if possible. Here are some of the questions I’d ask:
- Is there a design limit to belt length? (Update: I now think the answer to this is likely no)
- Will the belt tolerate a roughly 30 degree backward bend around an idler on the drive side? What shape of idler would keep the belt on track?
- What are the requirements for making belt drive work with rear suspension?
- Has Gates ever tested a typical recumbent frame design (long with cantilevered boom) for stiffness?
- I’d probably save for another day the obvious marketing question: Will Gates ever see enough market for really long belts useful only for recumbents?
Belt drive for recumbent trikes is compelling enough and feasible enough that I’d imagine manufacturers have at least done some preliminary legwork and engaged with a belt manufacturer. I’d love to know how those conversations went. Considering how long belt drive has been commercially available, it may not be a good sign that I can’t seem to find even one recumbent manufacturer talking about it.