Gerry Hull

i've been working with Shimano hyperglide cassettes over the last year to see if, by prudent modifications to some of the tooth profiles (i.e. changes that do not involve the load bearing areas or compromise the overall strength or durability of tooth) the shifting could be improved. Hyperglide is a brilliant piece of engineering; i've probably been through my weight in discarded cassettes, just trying to deduce how it worked. As a cyclist, though, I was curious to see if something could be done to eliminate the unpredictability of chain engagement when upshifting, and accomplish downshifting with a little less fuss and clatter. I have finally come down to a few standardized modifications that someone who knows what they're doing can perform in about an hour and a half with very basic tools. The results are worth noting. and since the hyperglide tooth sequence is the same on all cassettes from a 7 speed acera to a 10 speed dura ace, it requires only knowledge of a few basic principles to be able to work with any one of them.

It does take rather a while to explain this stuff, though, and i thought that before getting into the thick of it I would first ask if anyone is interested first.

Gerry

NightShift

I'm interested.

Kimmo 

Sure it sounds interesting... but the odds of you intuitively improving on CAD aren't great...

Not everyone will listen with an open mind, and most are prolly gonna want to see video of the difference.

shelbyfv

Yes, I'm interested in learning how they work.

Gerry Hull

Lord it was anything but intuitive! It was trial and error of the most tedious, expensive, and time consuming kind: garbage cans full of discarded cogsets and a constant pit in my stomach from knowing I should be out training instead of spinning a crank on a stand trying to micromanage chain behavior. I'll begin to explain how the system works when I have a bit of time to dedicate- i'll also need to figure out how to upload a few drawings so that what I did can be duplicated. That way nothing has to proved to anyone- one can try it oneself and decide that either the refinements work or they don't. And please bear in mind that the improvements are very modest- many riders might not even notice; though cyclists who are highly sensitized to how well things are working certainly would-
Gerry

2_i 

I have been putting together custom cassettes out of cogs that were either taken out of different cassettes, including different speeds and different manufacturers. The profiles teeth then obviously do not line up. My solution with the involved inaccuracies in shifting was to change the rear derailleur from Shimano to Sram 1:1 system. The problems of shifting, associated with the teeth profiles and other, have virtually evaporated. However, I am interested in any tweak that improves the shifting, whether with Sram RD or not.

spike420211

Ranmping profiles are common to all brands of freewheel and cassete cogs nowadays.
What gets me is that if you look closely, there are sharp edges where the chain sits, usually in the gulleys,
often along the upper part of the tooth itself.
I often wondered what would result if those edges were slightly radiused/blended,
but, alas, I have no f***ing Dremel tool to test out my theory.

Gerry Hull

2-i, that is a topic close to my heart. Yes, there is another tweak you can perform. Youve already discovered the most important, which is to use a derailleur that has zero float in its tension pulley and is therefore more capable of muscling the chain when it doesn't really want to go. The same thing can be accomplished with a shimano derailleur if you simply remove all of the float from the upper jockey wheel.

I had a special grind for making mismatched cogs work together and luckily it is a hell of a lot easier to perform than what I was about to describe in the main topic.

I wont bore you with how it works, I will just tell you how it's done and you can try it yourself. All you need is a drill with a rotary sander attachment and some 120 or 150 grit paper.

What you want to create is something of a 'shark tooth' profile, so that the front of the tooth is slightly lower than the back. You can grind an awful lot of material off of a tooth without affecting its durability so long as you dont remove any from the corner at the top of the load bearing portion of the gulley. When i say 'front' of the tooth, I mean the part that meets the chain
first but is non-load bearing when tension is applied to the chain. if you are looking a a cog from the outboard side as if it were a clock, the leading edge of the tooth at noon points towards the front.

Take your sanding/grinding implement of choice and grind each tooth at about a 20 degree angle; ive used angles up to 40 degrees, on 8, 9, and dime-thin 10 speed cogs and have never suffered a broken tooth, but a severe grind like that isn't necessary. All you want to do is get that leading edge low enough so that it no longer has any of the chain-deflecting chamfering that is present on the outboard side of most of the teeth, a nice sharp chain-grabbing edge that will do the pick up job with no hesitation.

For a decent visual example, take a look at the cassettes offered by Action-Tec. That is about what you want to end up with. A couple of years after i started doing this i learned that they not only employed a shark-tooth design but also utilize a zero-float pulley to get the job done.

2_i 

Gerry Hull, is there any chance of you having a worked out cog lying around and taking a photo? One thing that worries me is that unevenly worked over teeth would accelerate chain wear. Otherwise, regarding the derailleur, would it be then enough to replace the upper pulley in Shimano with Sram or any aftermarket one? The latter, in my memory, often lack play. This morning, after just few month with the Sram RD, I cracked its apparently flimsy alu cage. Thus, I am getting to a starting point with my rear shifting. This is just my quick reaction - I will need a longer time to chew over all the points that you are raising.

operator

Did you look at the 7900 series cassette and chains?

Gerry Hull

hey, you can transfer the sram pulley to the shimano derailleur (the upper tension pulley is the important one here) or, grind down the little steel bushing in the shimano pulley to eliminate float (which is kind of tricky), or, performance makes a replacement pulley that is easy to modify to eliminate float. The problem with the latter is the little cartridge ball bearing is often not fitted very tightly and I usually end up knocking them out and putting them back in with some red loctite applied so they stay put.

I've seen no accelerated chain wear yet, so long as i do not take away any of the load bearing surface. Take a look at an old cog and you can see precisely where the chain pulls against the tooth. The grinding is done only on the
"passive" side of the tooth.

I made a huge error in my previous post about which way the leading, or passive, edge of the tooth faced in my example of the clock. I had it backwards. Hadnt had my coffee yet.

A much better visual, if you are looking at the outboard face of the cog, they should look like a bunch of little shark fins swimming in a clockwise motion.

i have been trying to think of a better way to describe how much to cut and what angle.

if you are looking at the tall part of a cog tooth, the front and back edges are virtually squared, almost 90 degree angles. After the cut, one side of the tall part will now be an acute angle, about 70 degrees. It will be the same height as it was before. The other side will now be an obtuse angle, roughly 110 degrees. That will be where you did most of the cutting.

I never used a protractor when doing any of this, though I've a good eye for angles. Its pretty easy to get the teeth all looking the same with a little practice. it takes about an hour, works really well for this particular application of mixing cogs, and it also looks cool when youre done.

i havent seen seen a 7900- one thing i have noticed, though, is that the highest grade cassettes have tooth profiles that are very close in appearance to what I did in the 'refinement' exercise with mostly medium grade cassettes. One day i saw an xt cassette and thought, hell, if id just bought one of these it would have cut my
work in half. True, I only did what had partly already been done, but I did it based on my own conclusions and observations and that made me feel like i halfway know what i'm doing.
Gerry

2_i 

Thanks for this clarification. I was puzzled at first as it sounded like trying to accelerate the aging of a cassette. The ActionTec photo linked to below also helps. The only thing that scares me is the number of cuts that need to be applied.

So you work both on the front leading edge of the tooth as well as the rear to get the fin shape. However, you do not touch the front load-bearing corner. The front is just touched around the top of the tooth and most material is removed from the back. The width of the top will get reduced then though. Correct?

The problem with higher-end cassettes is that the larger cogs come on a spider and are thus useless as far as mixing is concerned. Personally I have no regard for the choice of cogs in the commercial sets which emphasize racing and getting out the last bit of performance at the highest speeds while pushing air drag. What does this has to do with everyday riding?

The Action Tec cogs can be seen here. Why don't they make them out of carbon steel but only out of the generally flimsy titanium?? We are not fighting here for the mass-to-load ratio of a rocket that will carry a Mars rover.

Jeff Wills

Somewhere in the mid-'80's, Shimano produced a Dura-Ace cassette with pronounced "shark-fin" shaped teeth. I think this was 1985-ish... at the end of the AX era, before indexing took over. I wish I could find photos.

Gerry Hull

Jeff, I looked everywhere forna photo of that old Dura ace cassette! I found a copy of Shimano's 1985 dealer catalogue on Sheldon Brown's site but couldn't get a close enough look.

2-I, I think you've got it. Really, the Action-tec sample photo is a very close likeness to what I usually ended up with. Their 'shark fins' are a little more steeply angled than what I did- I didn't remove quite as much metal as would be required produce that, but it's an excellent model to go from. Observe the smallest cog in the set. See how ithe teeth appear to be chamfered on the outboard side? And yet the leading edge of the tooth is cut down to where the chamfer is eliminated. That is precisely what you want. It's what makes the cog nice and grab- happy. It doesn't take long. Sitting on the sofa with my reading glasses and my cheap ryobi drill with a 5 inch rotary sander attachment I could buzz my way through nine cogs in about 45 min. The hardest part was not wanting to leave the sofa to go change the sandpaper.

I've ridden nothing but mixed cogsets modified in this fashion for the last three years. I've used 7,8, 9 and ten speed cassettes, both sram and shimano (though i did get slightly better results when i didnt mix the two). I've used Alivio, Tiagra, 105 and Deore derailleurs with equal success; they worked equally well so long as the pulley float was eliminated. I've ridden in 2 and 300 mile events without once missing a shift or having to wait for the chain to catch. Only recently did I begin examining the hyperglide system and trying to refine it, using the standard cog combinations, rather than overriding it completely, which is what we are doing here. And i guarantee that one day you will encounter a guy blowing up his legs on an upgrade because he can't get his two thousand dollar gruppo to shift into low gear.
And yes, spidered cogs are anathema to me because I can't work on them. The one year i rode a ten-speed system I used nine speed cogs that i had to shave down from .075 thick to .065. I never want to do that again.
And by the way- this is quite important and i should have mentioned it earlier- Regardless of cog spacing, I used only nine or ten speed chain with this set up. Ten, most of the time, simply because it runs almost silently. Silent is good. 8 speed chain I found to be simply too grabby and oversensitive to shifter input.
Of the hundreds of ideas that I have tried, very few work. This one does. Good luck.
Gerry