orange and dx4, thanks for the responses. Here's the thing, I'm wanting to improve the shifting precision of this:


The short story: The guide pulley is a bit far from the cogs. A larger pulley would move the chain closer. Those plates were modeled after the Rally plates so the Soma cage should be no different.

With a cage like this the G pulley moves up and down, mimicking the action of a slant parallelogram. It's even better because if the arm lengths are correct it can track cogs of any size whereas the slant parallelogram is set to a fixed angle. Its flaw is that you can choose a chain length to give the ideal pulley-cog distance for one chainring, but when you shift the front that distance changes. So you have to pick the chain length for the small ring and let the big ring suffer. However if the rings aren't too different (mine are 47/42) then the effect isn't too bad.

All that begin said, the real problem with this cage is that, as Grand Bois pointed out, it doesn't work very well. The reason is that the ratio of arm lengths should be about 3:1 (T:G), but these are more like 5:1. So the G pulley doesn't really move up and down enough. Unless I used (i.e. made) different plates the best I can do is move the pulley as close as possible to the biggest cog, which means it will be closer to all of them. Hence, the larger pulley.