gloomyandy wrote:
Well using your favoured app 24x17 is still a valid gear (for the double case) with the setting at < 1.5 degrees so I can only assume it is less than 1.9 degrees?
sorry yes you are quite right, my mistake
You might expect that 36x24 would be a better combination than 46x32 based on the chainline, but this may not be the case when you actually test it. For instance in:
http://www.ihpva.org/HParchive/PDF/hp52-2001.pdf
using the 44x34 35" gear (which in this case is big/big) is 92.1% efficient but using the (closest equivalent in this test) 32x26 33.5" gear it is only 90% efficient despite having a much better chainline. So by using the bigger gear combination you get a gain of 2.1% and of course in this case (big/big) a double will actually offer a better chainline than the triple.
Triples may be a good solution for some people, but I'm not convinced that the chainline efficiency argument is really that compelling.
IIRC that data showed low efficiency for all tests including the middle chainring which was anomalous behaviour. I would expect that was because it was a heavily manipulated chainring (i.e. with lots of funny-shaped teeth) which hadn't been run in yet. You can draw some good conclusions from that data (better than they did, in fact) but comparisons of results from individual chainrings are not reliable indicators in this case.
This image
shows that if you run small chainring to the third sprocket in you are running a rather lossy gear by comparison with a larger chainring and sprocket, more favourably aligned. The losses due to variations in chainring/sprocket size are comparable with the losses that are caused by cross-chaining. As you downshift on the big ring, the cross-chaining losses start to overtake the benefits of larger sprockets. In the big/big combination you are losing over 2W in cross-chaining losses alone.
cheers