Which the better rear mech for 8sp?

[Freddie]

Not that this should sway your decision any, but I also think it is somewhat ugly.

Aren't they all today? The mechs above look like a vision of the future straight out of Star Trek: The Next Generation.

As an aside, do any of these new mechs work better than an old mid 90s LX, when it comes to friction shifting (or in fact, indexed). Has there been much in the way of functional improvement in the last 20 years? One would hope so, given the aesthetic improvements .

[gregoryoftours]

One thing that I think is bit of an improvement in rear mechs is the change from the use of a torsion to a tension spring, less likely to get bunged and stiffened up, although one good thing about the old type is that you'd sometimes get variable spring settings on the torsion type. I do think that the new clutch mechanism on shimano mtb mechs is really good for offroad, although I don't know how it is in terms of reliability. I do think that older rear mechs were nicer quality in general.

[Brucey]

FWIW I think that shifting quality has improved as we have gone from 8s to 9s to 10s. I can't comment on 11s because I've not used it much yet. I think that in most cases these shift quality improvements have been at the expense of durability in the chains and sprockets, less so the rear mech.

However these improvements are at least as much because of design features as they are anything else, and those design features can in time transfer to lower groupets with fewer gears. So for example, today, you can buy a 6s freewheel and chain that shift slicker than any that I used back in the day, even with old-style mechs that were not noted for their shift quality back then. Provided it isn't worn, I'd expect a 20 year old 8s mech to shift pretty well even today, especially if it is running a good quality chain and cassette.

What has happened is that mechs have become more specific to a particular task; having been all as one, shift ratio-wise at the time of 9s, shimano mechs are gradually becoming fragmented, as a species, with many different shift ratios in evidence, and some (like the shadow mechs) that will only work best on a wide ratio freewheel.

Fundamentally a single pivot slant mech (most SRAM, Shimano Shadow, old SunTour etc) will work pretty well on a small range of freewheel types; basically the slant has to match the range of the cassette pretty closely.

Most modern mechs combine the dual pivot principle (shimano 'panta-servo' they used to call it at one point, even though simplex had done it decades earlier... ) with the slant parallelogram principle. The dual pivot (and the dual parallelogram eg Huret duopar) principle should allow the top pulley to track the sprockets closely. This occurs at the expense of more moving parts (that can wear) and furthermore for various reasons the dual pivot principle has never been implemented in a perfect way.

What do I mean by this? Well the idea is that the pivots are sprung so that they counteract one another, thus leaving the pulleys to 'see' the approach/exit chain angles and therefore guide the top pulley to a constant distance from the sprockets. However in order to achieve this both springs ought to have equal (or constant) force, and they cannot do so; the springs are finite in size; they cannot be constant force, and each works in opposition to the other, so they can only ever be equal at one point. In fact I'm not sure that they can even be made constant rate, they are so small in size. If the dual pivot principle worked properly, the slant parallelogram would be unnecessary.

There are secondary issues which have received attention in recent years, such as damping of chain slap, shift initiation threshold (*), wear, adjustment/alignment tolerance, wheel removal, parasitic losses, automation.... but none of these is anything other than subservient to the basic issues concerning guide pulley alignment, which is still an incompletely solved problem IMHO.

So, I believe that there is a yet-to-be-developed derailleur that is fundamentally better in its primary (rather than secondary) characteristics. Such a mech ought to be able to handle a corncob (straight through) cassette just as well as a very wide range cassette, with minimal, (or no) adjustments. I think it can be done!

cheers

[(*) by which I mean how much input is required to get the chain started on the next cog. Tooth and chain shapes help considerably, but the mech must extend as the chain starts on the next sprocket and then contract once the shift is complete. Because of the way they move, some dual pivot designs are noticeably slicker than others when a downshift is initiated..]

[Sweep]

I checked the spec on the link you kindly provided Brucey.

Yes, that mech you recommended will take a 22T front and 34 back.

Thanks again.

[MikeF]

I have a Deore LX RD-T661 which now appears to be obsolete - it was probable obsolescent when I bought it. It's running with an 8sp cassette with bar end shifters, and is fine at the moment. I have an RD-M391 which is similar, as a "spare" but this is intended for a mountain bike rather than trekking. I'm not sure what the difference would be in use. There is now the current RD-T4000 which seems to be current trekking one and is similar to the other two, but with resin plates? Like Sweep, the plethora of Shimano products also leaves me confused. How do you choose? What actually are the differences? Even looking through spec sheets it's not obvious.