Kimmo 

I sullied the weapon today with flat bars and yucky plastic canti brake levers, not to mention the crapola plastic Shimano friction thumbies (two lefts).

But this temporary insult was merely to verify that this will shift under load:



As you can see I had to lift the DA derailer a few mm higher to clear the 50t. Works, but I might find a better fit with something else... or maybe I could dremel the cage a bit; the nickel-plated ally cage isn't long for this world anyway...

I probably could've left the drops on it if I was prepared to go the hack on my left Doubletap; the overall throw was pretty close to right, and I suspect the trim position could've been shifted to line up with the 50t... but never mind that.

This whole idea is enabled by my DIY Di2 project; there's no way I'd bother to shift this manually. Concerning the DIY Di2 thing, watch this space (I'll put the link in my tag, which dynamically updates)... I've bought nearly everything I can buy and I don't need to make much, and I have some preliminary code I can test the system with. Initially it'll be a sequential manual (ie, 'sequential' in that I won't have to worry about what ring/cog combo I'm picking, it's just up/down), and then when that's sorted I'll introduce a speed sensor for the rear wheel and a fine-grained speed and position sensor for the crank, and make it automatic, with the buttons selecting a desired cadence, basically like NuVinci control except digital and without the CVT. The half-step goes a long way to addressing the shortcomings of available ratios with a derailer setup (this physical limitation is a result of the packaging restrictions forcing us to use few rings / many cogs instead of the other way round, and the fact that a given tooth pitch becomes a larger fraction of the circumference as circumference diminishes, causing the line on the graph to curve upwards, when we'd prefer it curved the other way, due to the exponential nature of drag)...



The blue numbers (which include the green) are the ratios of the half-step triple, the red numbers are a 53/39 on the same cassette for comparison, and the green is the half-step setup short-shifting.
The X-axis on the graph is the gear number (not that that's a thing in the bike world - yet) and the Y-axis is the chainring/cog (multiply by tyre circumference for gear-inches or whathaveyou).

Looking at a standard double, I saw a fair bit of room for jamming another ring in there, and even without doing horrible things to the chainline. The big rings are about 5.5mm c-c, about the same as 6s cogs. The 52t is outboard by pretty much its own thickness... I started filing the chain-catching pin down, and then realised it was redundant so just punched it out (there isn't room for the chain between the ring and the crank arm). The 38t is inboard 2.4mm. Rather than go to the horrific expense of buying spacers, and particularly since anything over 2.5mm tended to be the wrong sort, I bought a length of 14mm rod for a few bucks and got a mate to turn some up for me on his mini lathe. We could only scrape up a 9.5mm drillbit and I had my heart set on 9.9mm (nice and snug), but it turned out I have a rat-tail file that was perfect to finish the job.

Now the packaging job on the front of the drivetrain is starting to look as impressive as the other end... moreso on a beefy carbon frame with little room to spare.



If it fits on this, it should fit on most things I bet. One caveat is there was no way it was gonna work with my RED Yaw FD, but that's understandable.... so anyway, the specs:

Inner ring: Origin8 10-hole 110/130BCD 38t (the same thing is available with some other names on it, eg Action SS)
Spaced: 2.4mm inboard from spider

Middle ring: 110mm 50t (for a 36t small ring; neither love nor money will get you one for a 38t I bet... seems to shift pretty well though)
Spaced: 3.0mm outboard from 38t (front face of the ring pretty much in line with the spider's front mating face)

Outer ring: 130mm 52t (a plain old-skool ring, since the chain won't see any shift-assist, not that there's any designed for a 50-52t shift anyway)
Spaced: 2.9mm outboard from spider

There's one set of standard chainring bolts holding the 50t on the 38t, and the bolts going through the 38t and 52t rings need to be 13mm between the heads. Such a thing is generally only available as a MTB bash guard bolt in 15mm sets of four... (if you're lucky you might a 15mm set of five, for a Birdy folder or something) or there's the Thorn SS 15mm chainring sleeves in sets of five, available from SJS Cycles. I asked about the last set of red bashguard bolts Moove had on their site, and a nice bloke called Ronan threw in a fifth bolt for free : ) Many filings later, I had 13mm bolts...

There'd be some room to tighten this up a tad; although the gap between the 38t and the 50t is 0.5mm narrower than usual and they're inboard by a chainring's thickness, there's also a smaller diameter difference and the chain can reach the 8th cog on the 38t without catching the 50t, when I only need it to reach the seventh.

A note on the ratios selected: these were chosen of course with my own preference in mind (I'm a light guy on fairly flat terrain) and in consideration of what's commonly available, but IIRC that nice flat curve I was able to obtain on the graph was a happy accident - as I recall (it was a few months ago now) there was nothing else in the generally obtainable range that gave as flat a spread. I'd be delighted if anybody feels like firing up Excel and proving me wrong, though.

A chunk more top-end range could be had by going 11s and using an 11t (bah) or going the other way with a 25t...

Roboshifting allows the indexing to be pretty arbitrary, so maybe you could do a 12s setup using unmodified big cogs on a 10 or 11s carrier, with the other 8 or 9 being separate cogs spaced closer...

And the fact the FD here is hardly optimal can be compensated for by integrating overshifts into the code, and mitigated by phasing the shift with the crank angle. Front and rear shifts will be timed to coincide, using cadence and chain position info, by controlling the delay on front shifts following rear shifts.

Retro Grouch 

At some point gearing ratios can become too closely spaced. I'm thinking that half step 10-speed is it.

fietsbob

Yea Half steps were 80's 6~7 speed .. my Touring bike shipped as a 48-44-28, back then ..


But I'm glad you found something to keep you Busy for a while.


obviously with the 130bcd all you can manage is a 38t.. to have it fit.

I'd put K-edge chain catcher on there to keep from dropping your chain.

Dropped chains make ugly things happen, on Carbon chainstays.

davidad

My tourer is set up with a 20-40-44 front and an 11-13-16-19-23-28-34 cassette. 18 usable distinct gears.

HillRider

Wow! The concept of swatting a fly with a howitzer comes to mind. I've heard of agonizing over gear spacing but this is really overkill.

ThermionicScott 

The beauty of this system is that once he's done the agonizing up-front, the computer can shift sequentially through all of the combinations. I've sometimes wondered if this would have been the preferred direction if half-step had held on a little longer.

Kimmo 

Look at what the industry is doing to chase diminishing returns of efficiency. This here, this is the last big chunk of low-hanging fruit available. Mark my words - this will come to pass as standard stuff.

If you think I'm being a little grandiose, consider that I predicted compact road frames in the 80s; I've always been ahead of the curve.

I know what that sounds like, but for some reason this field of endeavour seems to be mostly populated with folks who have a hard time imagining that the design of the bike isn't finished yet.

I'm not one of em.

We shall see... I've never felt like my gears were too close; only that I don't have enough of them. I suspect anyone who time trials, or rides lazily (those most interested in efficiency) would agree. I'm both aggressive and lazy, so I have a particular interest in efficiency... It doesn't matter how many speeds; that's only about range. As long as we're talking half-inch pitch, the only way to get any closer than say, a corncob 6s, is to half-step. Or use abnormally large rings and cogs.

This range suits my riding. But the only obstacle to doing this on a 110 crank is finding a 130mm 50t and 110mm 52t, pretty doable. Oh, and identifying the best tooth count to use... fun with Excel. Oh yeah, the bigger difference between small and middle rings would also mean leas clearance to reach the 8th cog, which might be a dealbreaker, since you're talking less overlap on the cassette...

As for the chain catcher, I could put my RED one back on, using the SRAM FD's trick internally-threaded mounting bolt, but I think it's surplus to requirements - there's no room for the chain between rings and frame.



Yeah, but it looks like a pretty rough spread to me, and you're not avoiding the problem I'm addressing here, ie bigger gaps between gears at higher speed.

Your curve shares pretty much the same angle as 53/39x12-23 once you're in your tenth gear...



Also, you're talking some pretty fancy fingerwork to shift that sequentially. That chain would be dancing all over the place.

Exactly. Except that I don't experience any agony dealing with ratios; I seem to understand that stuff better than most (I'm a whiz at stoichiometry). What's giving me a hard time is figuring out how to code interrupts in C++... plenty of agony there.

Keep an eye on pro time trial rigs and tri bikes over the next few years... probably sooner, now that I've pointed out how to test the idea with mostly standard parts...

fietsbob

Good Luck, then, Suck can take out enough carbon(or aluminum,etc.), to make room for the chain,
when the forces kick in..



For others ..
conventional, sort of .. TA triple-isers replace the double's inner ring .. offer a 74bcd to go inside ..

but could require a BB length change which you cant do with the spindle fixed to the crankarm

like current trends have pushed..

davidad

If you look my spread is 10% on the two big rings and the grandpa gear is a bailout for steep hills and a load.

Be patient. shimano has a patent for a 14 speed cassette. Or you could get a Sram three speed hub with an 11 speed cassette for 99 speeds.

Kimmo 

Yeah well, I sure hope not, that'd be ugly. I'll bear the possibility in mind...

Come to think of it, not too many folks could say their downtube is at risk of making a chain sandwich with the rings...

Patience is for those without the means to innovate for themselves.

And I'm not interested in more cogs; more cogs and not enough rings is what got us into this steepening-ratio mess in the first place. More cogs isn't going to solve anything unless it's accompanied by a shorter pitch.

And since the whole point of the exercise here is greater efficiency, IGHs need not apply.

rydabent

IMO the old half step with its double shifting went out with 10 speeds that we had before the middle 80s. With the 8 and 9 speed rear clusters and more, shifting has taken on a whole new aspect. Most of us just look at triples with 8 or 9 speed rear clusters as ------------up hill, level, and down hill. After picking the proper front chainring for the road at hand, we then just shift across the rear cluster for slight variations in the road. We no longer need to be slaves to half step with its double shifting. Life is now simpler, dont complicate it if you dont have to.

Kimmo 

Before you can hope to make the slightest dent in my enthusiasm for this concept, how about demonstrating that you grasp the mathematical and physical constraints that I'm railing against here.

Do you spend much time trying to go fast in street clothes, while trying not to raise a sweat?

More cogs has not given us ratios as close as those of us who fully grasp the concept of gear ratios might reasonably desire.

The straight line of a corn cob cassette is actually a steepening curve, when you understand the ratios. And when you grasp the physics, you realise the curve is in the wrong direction.

In the end the combination of competition and marketing will unearth this fact for all, to take someone else's* word for.


* ie not some random dude on the internet extolling folks to figure it out for themselves, but glossy brochures and slick webpages and videos from trusted brand names.

Willbird

Seems to me you are more or less building a CVT (constant velocity transmission)....which is a very hot topic in the automotive world, some of the same factors apply.....being able to optimize efficiency. Technology allows progress on ideas that have been around for as long as there have been machines.

Memory tells me there was some work being done at one time on a chainring that could alter it's pitch dia on the fly ??

Bill

Kimmo 

Well, it's getting a little further towards a CVT, but as my detractors will tell you, a human can be reasonably efficient over a range of RPM, and it's a pretty safe bet that it's impossible to engineer a CVT for a bike that doesn't have mechanical losses far in excess of any gains it makes...

So my detractors and I just differ over the range of RPM we're prepared to tolerate... +/- 1.75RPM over 30km/h sounds pretty good to me.



RPM vs km/h, red: 53/39x12-23, blue: my setup

One way to think about the graph - how tall is the next step in a staircase, when the higher you go, the harder it is to climb.

You kind of have to get on top of each gear as you shift up; I don't want to shift further, the higher I shift.

Willbird

Well the beauty of it is (to me anyway) The electromechanics are the "hard" part, the user can decide what their goals are :-). The thing could be tied into GPS so all the rider has to do is thrash in the drops and is always in the gear they choose in advance for any section of a pre determined route :-).......and you could even have a "trim" that allows to back off or become more aggressive in that pre determined gearing.

I think it is a slick idea :-). You might even get an endorsement from the Amish because it no doubt runs on batteries :-)

Bill

Darth Lefty 

I like it. Tinkering is fun.

Wanderer

I always thought the perfect triple would be a 50, 40, 30................... but that is really interesting..................

Maelochs

I just ride my bike.

HillRider

And wait 5-1/2 years to comment.

Ronsonic 

He was busy riding his bike.

Kimmo 

It just occurred to me, I never said why this project died in the arse.

I got to the point of having written arduino code to run the servos, and halfway figured out how the servos were gonna actuate my derailers, but the fatal snag was front shifting was always gonna suck until chainrings and FD cage had been specifically CADed,

I reckon that's all this would take; if only a Di2 FD cage was replaceable and the system was hackable, it could be an upgrade kit.

oldbobcat

Half-step was a popular racing and sport bike setup in the '50s and '60s. I think 5-speed freewheels and rear derailleurs with limited range had something to do with that. When I started racing, in '73, the guys in my club were all buying 48- and 49-t rings. I fell in line. One or two years later we were all going back to 40s and 42s with single-step clusters plus bail-out cogs in the rear.

What I remember is dropping the chain somewhere short of the small ring on front downshifts, many times. This was delightful with a bunch of riders breathing down your neck. Many times. I think the wider range gave the chain more momentum and slack on the drop to the smaller ring.