How Chains Wear
Different types of chains
Motorcycles and bicycles both use what is called a 'bushing' chain. Bicycles over five speeds must use a 'bushingless' chain, due to the great angular displacement between cogs. Motorcycles tend to add o-rings within the chain parts for keeping dirt out and lube in the inner workings of the chain.
The 'Stretch' Myth
Bicycle roller chains do not "strech", they wear. Each pin is surrounded by a bushing that develops some play when it wears. A chain is said to be worn out and replaced on it's own when it's average length (pin to pin) has increased by .5%. So for a half inch pitch chain, 9 links should measure 4.500 " when new and 4.523" when it is spent. If the chain's average length has increased by 1% (4.545") then the entire drivetrain gearing must be replaced. You can measure from the side plates as long as you add the increase in diameter. An artical on this subject is HERE, HERE, HERE and HERE. I am still working on the problem that worn rollers really do effect performance and wear on the system. Expect this discription to be modified once I fully understand what is going on.
Strech of the materials in the chain is impossible given the relatively small forces that a motorcycle or bicycle can generate. The wear takes place where the pin from one link meets the bushing (or plates) of the next link. Interestingly, Since on every other link two pins are shared by one outer plate, The chain changes length only between every other link.
I do not read German. If i did, I would explain THIS document from Rohloff. It is the most comprehesive study of chain wear that I encountered while researching this topic. It obviosly covers every base, but I can not read it. please contact me if you can.
Measuring Chain Elongation
To measure chain elongation, measure as many pins as you can in a straight line and off the sprockets while gently applying a load to the chain. (length)/((the number of pin measured - 1) * Theoretical pitch of the chain))
Bicycle Chain Limits
Replace a bicycle chain after it has elongated 0.5 %.
A Shimano tool measures 0.4% for replacement. They also recomend cycling 4 chains over the life of a cog and ring set.
Wipperman suggests that a chain has exceeded it's wear limit when the distance between 10 rollers measures 107.4mm. (4.228"), measuring using calipers angainst the inside of the first and tenth roller. Although this measure combines measurement of chain elongtion and roller wear, it is worthy of consideration since is is coming from a manufacturer. Theoretically this measure should be 106.5mm (4.194") without any wear in a perfect system. This means that either the roller IDs have worn .028", the chain has elongated .62%, or a combination of the two. We of course know that it is a combination of the two. Say Wipperman had decided that 0.5% was the limit of elongation, then the roller wear limit would be about .006". Note on the chart below that Shimano Dura-Ace exceeds this by a factor of 2. If I understood german, I could clear this up by reading THIS publication by Wipperman. (something about 3%?)
By comparison, I have a Park CC-1 Chain Checker Tool. At a reading of 3.25 where the blue zone changes to red, the measurement is 4.741", this of course based on 10 pins. This means that either the roller IDs have worn .037", the chain has elongated .74%, or a combination of the two. Again, say park had decided that 0.5% was the limit of elongation, then the roller wear limit would be about .012". An interesting note on the park tool and the numbers labeled on the face. They are marked in even segments of the dial. A student of trigonometry would realize right away that this must be a meaningless scale. This tool works from of the motion of a cam to guage distance change. If the dial had any real meaning it would be labled sinusoidal rather than evenly. It is not, casting doubt as to whether the designer actually understood what they were doing. The maximum reading on this tool is 9, where the combined elongation and roller wear is 0.148", well out of any usable range. The Park CC-2 is essentially the same type of tool as the CC-1 but designed to be easier to read and more accurate. Supposedly it reads from 0.25% to 1.00% wear, but that is impossible for this type of tool, leading to some speculation as to the validity of it's use. Funny as well, the designer has still to study sinusoidal motion.
The Park CC-3 is designed soley to measure chain elongation rather than including roller wear. It will quickly tell you if you have passed 0.75% and 1.00% elongation. This seems fine, if those are your only target points. At least it is only measureing one parameter. Their statement says that after 1.00% the chain should be replaced. I belive the rest of the drivetrain replaced at this point as well. This tool does ignore the concept of roller wear, which is fine, but I do belive that should be addressed as well as elongation.