[Madison]

There is no reason why this sort of design cannot work. It would require that you gather data about what amount of belt tension is required, either experimentally or mathematically, as well as belt lengthening over time.

Where within a drivetrain did you anticipate placing this? Early in the reduction, you're transmitting far less torque along the belt and thus require less tension to keep everything spinning together, but by placing it later in the reduction -- or even as an output stage -- you can use larger materials and things, mitigating the need for precision.

Can anyone share some anecdotal information about what sort of tension might be required to keep the belts and pulleys in synchronization? Someone who works with cars may have a good idea.