"David L. Johnson" <
[email protected]> wrote in message news:<
[email protected]>...
> On Wed, 10 Dec 2003 20:24:09 +0000, Carl Fogel wrote:
>
> > You measured a front-sprocket hop of about 0.3mm and a rear-cog hop of about 0.1mm on your fixed-
> > gear bike.
>
> I just looked at my own bike, which has a Dura-Ace flip-flop hub, Shimano cheapie cranks with a
> 42tooth chainring, a 16t old British cog from about 1970, and a 15t Campy cog from about that same
> vintage.
>
> The chainring had a high spot (measured as Sheldon suggested, by the chain, but since it was over
> several revolutions any chain irregularities would have averaged out), of less than a mm. Let's
> call it 0.5mm
>
> The cheap British cog had no noticeable out-of roundness, but the Campy cog was about as far out
> of round as the chainring. So, with the two we are talking 1mm --- radius.
>
> Note that I am talking about the tolerance that most people have for wheel roundness. Any further
> out of round than that, and on a wheel it would be noticable. As I spun the crank/wheel, I could
> see and feel the out-of roundness, just as I would with a rim. Were a rim this out of round, I
> would keep working on it.
>
> What would you expect the chain to do under those conditions? Let's presume that, for the sake of
> argument, half of the sprocket is round with one radius, the other half with the larger radius. I
> know that is not the situation, but modeling non-concentric circles will be more of a headache and
> will not really increase the accuracy.
>
> That results in about a 3mm difference in chain length on the two half-circles, 1.5mm difference
> on the chainring, and about the same on the cog. When the top is under tension, that 3mm would be
> all on the bottom run of the chain. On a 450mm run of chain (taught), that 3mm would change the
> chain from a straight line, or nearly so, to a graph with a considerable dip to it. The shape of
> the chain is the solution to the hanging cable problem, just like the utility wires on a rural
> road. But if you grabbed it and pulled down in the middle, so that the chain was again straight
> with an angle at your finger, it would pull down x mm. There is a triangle, with base 225mm (the
> straight path), hypoteneuse
> (226.5mm), and height x formed by the straight-path versus pulled-down path of the chain to the
> mid-point. By the Pythagorean theorem, x = sqrt(226.5^2 - 225^2), which is a whopping 26mm.
>
> So, if under these conditions you adjusted the chain to be taut at the tightest position, it would
> sag over 2.5cm at the loosest.
>
> Now, my back-of-the-envelope calculation makes a worst-case scenario, and also doesn't let the
> chain just sag, it pulls it tight. But certainly half that much sag might well occur based just on
> these measurements. That is certainly enough to explain what we see.
> >
> > This suggests a maximum variation in the distance between the teeth grabbing the top chain run
> > of about 0.04mm, which rounds up to about 0.016 inches.
>
> I don't get that bit. It's not a variation in the distance between the teeth, it's that there
> would be fractionally more teeth on one side than another, measured by using a line through the
> center of the bottom bracket as the dividing line. I think my 1.5mm figure is more common than
> your .04. The slack/loose difference comes from how much chain is wrapped around the sprocket -- a
> half circumference of the slightly off circle. If one half is longer than the other, as would
> happen in an out-of-round chainring, that would change how much chain it takes to get half-way
> around. That difference in chain length is what you see.
>
> If you presume instead that both sprockets are perfect circles, but the axles are not concentric
> by that 0.5mm, then instead the length of the spans between sprockets would change by as much as
> 1mm. That would give a 2mm extra length of chain at the loosest point, which would look pretty
> much the same.
>
> > Elsewhere, Sheldon Brown has suggested that, with good-quality modern parts like the ones that
> > he used today to build a fixed-gear bike, the chain-tension variation vanishes.
>
> I don't know whether you consider Dura-Ace or Campagnolo to be good quality, but they work pretty
> well. The only thing that might be better now is if you consider a CNC machined chainring, which
> might be a truer circle and have the bolt circle very close to concentric with the teeth. But CNC
> machined chainrings are not the best for other reasons, and we still have to deal with out-of
> round crank spiders and attachment variation (the largest source of error for the chainring).
>
> >
> > Try not to be too snippy if I've goofed a calculation. I really am curious about this matter
> > because I can't make it add up. Possibly someone who remembers geometry
>
> I remember a bit of it.
>
> > will step in and explain that the thickness of four sheets of paper really can make a practical
> > difference in chain tension.
>
> Your analysis that led to the 4-sheets of paper is off, in any event.
Dear David,
I'm not sure that my analysis is off.
I think that you are assuming that the cog-hop is the result of a distorted sprocket and then doing
some admirable calculations.
But this assumption seems to be mistaken.
The amount that the tips of the gear teeth (or the crucial curve just below them) varies from an
ideal circle is what's being measured.
This variation partly reflects any distortion in the sprocket's circularity, but also any
distortion in:
a) how the sprocket is centered on the pedal arm
b) how the pedal is centered on the spline
c) how the straight the spline is
d) how circular the cone face of the spline is
e) how accurately machined the balls are
f) how smoothly machined the cups are
g) how parallel the two cups are to each other
I expect that these are all pretty darned good, but each minor inaccuracy is magnified from as we
progress outward toward the tips of the gear teeth.
If the tips of your sprocket are hopping 0.5mm as you turn it, I suspect that a significant amount
of the the hop is not the sprocket, but everything from its attachment point on down to the frame.
That is, your sprocket removed from the bike and measured just by itself seems unlikely to show
0.5mm of variation. So while I expect that your calculations make sense, I don't think that they
actually apply here.
I think that you're working on how a lumpy sprocket mounted on a perfect shaft would affect chain
tension and correctly concluding that X amount of lumpiness creates >X increase in chain-tension
variation.
I'm assuming that a much less lumpy sprocket that's mounted a little off center on a somewhat
eccentric. If that's the case, then X amount of hop may be a better rough-and-ready measurement.
Carl Fogel