Science on the efficiency of chain drives begins on page 3 of this publication:
http://www.ihpva.org/HParchive/PDF/hp50-2000.pdf
Measured efficiencies of several derailleur and hub gear drivetrains begins on page 3 of this publication:
http://www.ihpva.org/HParchive/PDF/hp52-2001.pdf
I'm glad to see a lot more data than what was in the tiny Bicycling article a few years ago. This is so much more enlightening.
The second one above was not my favorite of the two articles, but it still shows a lot of work went into the tests. They mentioned more than once that they didn't have time to take very complete data, and it shows. It would have been easy to use the computer for not just data acquisition but also for control, to vary the load, the speeds, and the gears, so it could take more complete data in far less time. I do this kind of thing frequently in my work, although with electronics instead of mechanical.
I wish they had connected the ergometer wheel directly to the hub instead of going through another chain drive which had, as they admitted, an unknown amount of loss. They did a preliminary test, but not at all the wheel speeds and torques involved in the subsequent drive tests. Both this report and the other one say that the speed and force affects the efficiency significantly. The unknown quantity still allowed relative comparison of drive types to each other, but the true efficiency of a given system could only be estimated without knowing that missing variable.
They used only 75rpm, which is unrealistically low for riders who will be most interested in this material, ie, the ones who want to go faster. As for the power levels, I only dip to the 80W level if I'm riding with someone who's slow (like my wife). I will be most interested in the efficiency numbers when I'm producing 150-900W. Interestingly, both types of systems get more efficient at higher wattages; but they assumed that the rider of a geared hub won't often produce more than 150W, so they didn't test the derailleur systems at the levels many of us are interested in either.
They tested the derailleur system, at least the Shimano 27-speed, apparently with an Ultegra rear derailleur (or its MTB equivalent), which has those pulley dust seals that quickly get stiff and don't want to turn. (The top one is especially bad.) After trying various kinds of lubes, I went to the ball-bearing type to save several watts. Cheaper Shimano derailleurs without those dust seals are better in that respect. Then they made the mistake of sorting gears instead of leaving them as 1-9 being granny ring, 10-18 being middle ring, and 19-27 being big ring. It took more time to figure out which combination they're really talking about for a particular result, and whether it was even one that a skilled cyclist would avoid (e.g. the tiny ring and the smallest cogs). Few road riders will be using a 22-32-42 crankset either. 30-39-52 or 30-42-52 is the common set, and will make for better efficiencies.
I was a little surprised that the geared-hub systems generally came out only a couple of percentage points worse than the derailleur systems. I'm sure the spread would have been greater with better pulleys giving a greater advantage to the derailleur systems. The only non-Shimano-27 derailleur systems they tried were the Browning, which no one here would consider serious contenders for someone interested in performance and efficiency. It would have been so much more appropriate to try typical road set-ups from Shimano, SRAM, and Campy. I don't think it will ever be possible for a geared-hub system's efficiency to meet or exceed that of the derailleur system, since the geared hub does not get rid of the chain's losses. The geared hubs also give bigger step sizes between gears, which will make the human engine less efficient than he would be with the derailleur system and our common 12-25 cassettes.
http://www.ihpva.org/HParchive/PDF/hp50-2000.pdf
Excellent article, although it does not cover internally geared hubs at all. I was a little surprised at how little a bad chain line affected the efficiency (even though it is definitely harder on the equipment). It was interesting to note also that the type of lube had almost no effect on efficiency either. They did say at the end that friction only accounts for a small part of the loss-- which is very puzzling. Where is the rest going? Perhaps the chain pounding the teeth as it lands on them to engage? (That would be constant for a given speed, correlating with the fact that efficiency increased as power increased.) This seems to be the only hole in this particular article-- the fact that they did not look into this further after raising the question.
