Hi guys. I just found out the hard way that Shimano 11 speed road RDs use a different shift actuation ratio than their 7-10 speed ones. The old ratio was 1.7x(1.7mm of RD movement per mm of cable movement). With the 11 speed they changed that to 1.4x, so the old days of slapping on the cheap and cheerful RD are gone in that respect (my 11 sp RD has cage twist and I put on an old Sora i had hanging about).
Similarly for MTB the 9 speed mechs are 1.7x, but 10 speed are 1.2x and 11 speed are 1.1x.
I bought a SRAM apex RD from @riderx because he didn’t have a Shimano road 11 sp (or Tiagra 4700) RD lying about. But I didn’t put it on since it’s 1.3x so not quite 1.4x as Shimano. It is a backup for now.
So I decided to see if I could hack my Sora RD just to be 100% sure all my shifting problems are down to the RD (it was actually the chain jumping off the tension pulley even with new chain, cogset, and pulleys).
For the hack I recalled the Hubbub mod that changes the Shimano ratio from 1.7 to 1.6 to use with Campy ergopower (aka Shimergo http://www.cyclinguk.org/cyclists-library/components/transmission-gears/derailleur-gears/shimergo). So I took that idea and moved the clamp point even farther along so the cable does a hairpin rather than a 90 like in the hubbub.
On the work stand it shifts adequate in the small ring (hard to get on the big cog but no skips) and barely acceptable (skips) on the big ring. But it should be ok for me to be sure I need to invest in a new RD… or perhaps my old 11 sp RD can be salvaged?
Basically I used 2 clamp washers. The bottom one is in the normal spot and I use the tab as a guide for the cable. The cable is retained between the top and bottom washers. The top washer is rotated 180 vs the bottom one and the cable wraps around the hook and is secured under the tab.
Some similar ideas might keep you going if you have an old RD hanging around and can’t get to a bike shop. I did have to file my RD a bit so the hook would sit on the side it’s not supposed to.
Maybe the mechanical engineers here could compute if I am at all close to a 1.4x and if it’s closer than the Hubbub would be (it seemed to work better from my experiments).