Has anyone tried the steel-reinforced polyurethane timing belts? If so, how do they compare to the rubber ones?
It's a question of what you want to use the belt for.
All Belts are subject to stress as they run around the motor and idlers and gears and bend. They will get eaten as they are subject to friction against parts, they will stretch as they are subject to tension. All this applies some sort of stress or another on the belt. Anything that is subject to stress wears. And as it wears, it will fail. The question is just: What is your life expectancy? The lifetime expectancy is again dependant on how the belt is used, so usage dictates lifetime of the belt.
Each belt type has an application it is designed for. A bending radius that is to be kept at, a gear it is to be used with, a load it is expected to move, a tension it is expected to uphold, and a lifetime it is expected to serve.
Let's take a short look at some rough ideas what
a soft rubber belt without additions
This is usually a bad choice for printers, but can occasionally be in very cheap kits. This kind of belt has applications. It is for really low loads, it can work tight bends for a long time and provides superb self-tensioning with just a very short tensioning device. On the other side, most rubbers don't take heat well, stretches a lot, wear fast and can only move a light load. If it rubs against a standing object, it squeaks horribly.
a medium hard, somewhat elastic belt reinforced with fibers
The standard belt we often get with printer kits. the fibers lessen the material's stretching ability while offering still some self-tensioning (depending on the fibers). These properties allow to move medium loads yet harder material demands slightly larger bend diameters (not too much, but measurable!). When the piece gets friction against a standing surface the sound will be not that horrible, but the harder rubber might start to shave off, break teeth and destroy the belt after some time. The typical reinforcing is done with cloth, usually cotton, so there are some limits to how much they can take.
reinforcing with aramid fibers
Some fibers are better than others for our applications because of how they stretch and reinforce it. Aramid fibers, for example, are better than cloth. Some consider them as impossible to wear out with the tension in 3D Printing applications alone, but that is not entirely the case. They still stretch under larger loads or on longer lengths enough that it can be a problem (or become one). They also can still be shredded with friction (and heat). For a professional machine, the expected lifetime is quite nice. They are though, but only as good as the rubber used and if properly installed. There are also generally two types: long fibers and short fibers. Long fibers have a somewhat even stretching behavior, while short fibers have totally different stretching behavior. The latter type is looked at in scholarly articles like Yin, Zhou, Lou et al.
a hard belt reinforced with fibers
Let's take... Uhm... a toothed belt from a car. I believe some of these are aramid-reinforced, old ones are on cloth cores, most modern ones I know are steel reinforced. It is somewhat hard. By design, it is made to withstand a strong tension at high speeds. It needs somewhat big bending radii for the hard rubber, but that is easily given by the gears and the large teeth don't break easily. Over the lifetime it will slowly stretch due to the temperature it will encounter, till at some point it gets too loose or looses too many teeth as the parameters don't match up with the teeth anymore and they get ground away. It is made to deliver high torque and could move - if something similar was used on a printer-like machine (like a large CNC), a heavy load for a long time, but the machine needs to be made accordingly.
steel reinforced belt
Steel is going stiffer. Steel reinforcements usually get the stretching parameter to a bare minimum at the cost of increasing their minimum bend and added weight. They can be used to move very heavy loads without the belt starting to slack as it simply can't stretch to give slack on the underside of the belt loop when forces are applied. Steel belts are pretty much Heavy Duty. If you can work with the bending radii belonging to them (as said, usually a little bigger than a similar fiber-reinforced one), you could make use of these properties, for example in a CNC Router with a full spindle or when moving very heavy tool heads.
And then there is the super heavy-duty carbon fiber. One can barely stretch it (steel is less stretchy), making these belts super tough in the stretching compartment and granting an extremely long life under most conditions. In comparison to steel, they can work much tighter radii. Usually, they are coupled with urethanes to get a high wear resistance against abrasion. The biggest downside: their price.
Silicone is soft but high temperature resistant but dislikes abrasive forces.
Urethanes are not the best to handle the heat, but they can handle abrasive forces. They also don't create dust in the same way as other rubbers.
The most interesting feature of neoprene belts is their reduced noise under work.
Belts come in several formulations. This page from McMaster-Carr lists several types of belts. The main materials (rubbers) are Neoprene and urethane, with fiberglass, Kevlar, and steel reinforcement. I would suggest spending some time looking at these, comparing the specs, and basing your choice on the needs of your application.
I used 1/4" wide MXL series cut-to-length in my machine. I have yet to see a wear or stretch problem and haven't yet done the high-speed photography to look for dynamic stretch. IMO, I have a bigger problem with the length of belt resonating like a guitar string than I do with the stretch of the belt. The compliance needed for resonance can come not only from the belt but also from the mechanical system it is mounted to. In some cases, a slightly stretchy belt could dampen oscillations that would otherwise result from impulse forces being coupled into the printer frame.
I have recently changed the standard belts on my Prusa MK3 for the aramid fiber reinforced E3D neoprene belts. Those belts are really tough and hard to cut even with a quality side cutter! They have a much smoother surface and a lot more aramid fibers than the standard belts.
Prior to the change, the belt on the X axis was twisting when changing directions, now it is running straight and smooth. Surface quality has improved a little, the ripples are smaller and more regular. Overall a small but noticeable improvement
P.S. I am using belt tensioners on both x and y axis.
I do not believe the standard rubber/plastic belts have any significant stretching over time, nor do they stretch under drive motor force during acceleration (and in any case, extrusion takes place mostly under steady-velocity conditions).
While I suppose it's possible a steel-reinforced belt might have a longer lifetime, replacing a belt is quick and cheap, so why bother?
[incorporating comments] My suspicion is that standard belts, e.g., as supplied with Prusa-clone kits will outlive us mere mortals. They can move an entire print bed+heater, so a larger print head is not an issue.
The usual problem with any belt material is ensuring there's no significant slack (which leads to backlash/hysteresis), and changing the material won't help or hurt tensioning control systems.