In all my research mechanical -> electrical -> mechanical double conversion is significantly less efficient than a direct mechanical drivetrain. It seems to come down to about 80% for the best electrical vs >95% for mechanical [1][2].
Also there is a weight issue the generator is at least as heavy as the motor and thats a big problem on bikes.
Trains do it because they need precise traction control and have little concern for the weight. They sacrifice running efficiency for it but everything else makes them very efficient at running (low rolling resistance and low aerodynamic drag).
Also the unsprung weight of hub motors causes various issues as well especially if you have suspension for ride comfort. Most electric bikes are moving to mid drives anyway which is going to have a chain or belt.
Your 80% figure seems rather spurious, coming I think from extrapolating a gross simplification of some 2009 figures for much bigger motors. My vague understanding is that smaller motors are hard to get as efficient. Some things I was looking at last year suggested things keep improving and that in excess of 90% was in theory quite attainable now for such motor sizes, though in practice you probably won’t find it on the market. But I did find components that should combine to something like 85–87%. (I don’t think I wrote my findings down, or if I did it was on a device I don’t have access to at present. But a quick search shows https://outriderusa.com/pages/electric-bike-efficiency claiming a motor efficiency of 93%, which if paired with a generator of similar efficiency would yield 86.5%, assuming no transmission loss which is of course unreasonable, and assuming the efficiency stays the same at lower power levels, which is likely screamingly wrong.)
>95% for chain systems assumes good components excellently-maintained. In practice chain drives are generally somewhat less efficient, commonly more like 85–90% is what I think I read, with poorly-maintained ones often more like 65–75%. This is where an electrical drivetrain could really shine: it’ll require roughly no maintenance while maintaining its not-best-in-class-but-still-pretty-good efficiency.
These figures are all rather nebulous, and get muddied a lot further once you’re putting a battery into the mix too, trying to take the power from your feet and augment it with another 200W of power from a battery (but preferably without feeding your foot-power through the battery). Then I go from being an amateur that knows just enough to be dangerous to a rank ignoramus.
For my purposes, an electric drivetrain would probably be markedly less efficient, but if it can be enough more convenient, it could be worthwhile anyway. I also have a pet scheme in mind where you simulate 100% efficiency by adding a little more power from the battery to compensate for known losses. I feel that could fit within the intent of legislation, if only that legislation allowed electric drivetrains in the first place.
Also there is a weight issue the generator is at least as heavy as the motor and thats a big problem on bikes.
Trains do it because they need precise traction control and have little concern for the weight. They sacrifice running efficiency for it but everything else makes them very efficient at running (low rolling resistance and low aerodynamic drag).
Also the unsprung weight of hub motors causes various issues as well especially if you have suspension for ride comfort. Most electric bikes are moving to mid drives anyway which is going to have a chain or belt.
1. https://pages.jh.edu/news_info/news/home99/aug99/bike.html
2. http://large.stanford.edu/courses/2010/ph240/veltman2/