Before the chain can be installed, it needs to be cut to the correct length. For my build I'm using Campagnolo's Centaur groupset and hence the 10-speed, ultra-narrow Centaur chain. Prior to the installation I did a bit of research on chain sizing. Remarkably all bicycle chains come in the same pitch (the distance from one roller to the next) and this is set at 1/2" (half-inch). Given the bewildering variation in almost every other aspect of the drivetrain, resulting in incompatibility between components from different suppliers, I'm surprised that chain-pitch has remained so constant. Anyway, this means my 114 link Centaur chain is 57 inches long.
Replacing an existing chain is easy, you just count the links and make the new one the same size. You have to be careful not to just measure the length because a worn out chain will have stretched a bit and you might end up with an extra inch or two on the new chain. But with a new build you need a different method.
Given that I recently bought a Park Master Chain Tool (more below) I thought I'd start at Park's excellent pages aimed at the DIY mechanic. They outline two methods of sizing the chain for a new build. The first involves measuring the chain against the two largest cogs without threading it through the rear derailleur and then adding one inch (two links) to compensate. Their second method is similar but uses a calculator instead, in other words, you put the size (in teeth) of the largest chainring and sprocket into an equation along with the distance from bottom bracket to rear axle.
Sheldon Brown deals with chain length in a longer article on derailleur adjustment. He uses the same method, in essence, resulting in a chain 1 inch longer than the minimum required to run the large/large combination of chainring and sprocket. He does hint at a possible reason for this algorithm being the accepted norm: a chain that is too short will cause damage whereas a chain that is too long will only affect gears that you shouldn't be using anyway.
The accepted answer on StackExchange simply cites Sheldon Brown, but at the time of writing there is a comment from one of the moderators to the effect that you should allow 2 links (in this case meaning 2 inches I assume from the context) for Shimano.
While researching this topic I found another new source of information I didn't know about, there is a wiki book on Bicycles with a detailed section on Maintenance and Repair. Again, the conventional wisdom is big/big + 1 inch.
Campagnolo Method (pdf)
With all this general advice to be around the large/large method I was surprised when I unpacked my Centaur 10-speed chain and found a completely different method of chain length sizing. Whereas most people advocate a method which results in the shortest possible chain that can work (on the large/large combination) the Campy method is a small/small based method which results in the longest possible chain than can work without rubbing the derailleur.
The basic idea is to draw the chain tight enough to ensure a small gap at the point where the chain is at its closest approach. This is the part of the chain feeding in to the lower jockey wheel. When the derailleur is retracted to take up the slack the lower jockey wheel bends back and upwards reducing the gap to the part of the chain going around the upper jockey wheel to a minimum. The instructions suggest setting this gap at a maximum of 10-15mm.
To make it easier to follow this method I cut a small piece of card diagonally, 15mm at one end and 10mm at the other:
The result of using this method was that I needed to remove 3 inches (6 links) from the chain. That makes it a 108 link/54 inch chain, two inches longer than the chain predicted by the equation above. I think there are several reasons for this difference. Firstly, I could have stuck with the 34T chain ring that came with the crankset and that might have meant knocking out one more pair of links (taking me down to 53 inches) - which is within an inch of the length predicted by the equation. And anyway, perhaps these new Centaur 10-speed derailleurs require 2 extra inches, rather than the 1 inch given in the formula.
In all of my internet searches (on a range of topics) I quite often got results from the blogger "Sprinter Della Casa" and he is the only one I found who seems to espouse the method of using the longer chain method. I'm not sure about the rationale but I enjoyed the article: How To - Campy Chain Install, Illustrated.
The Mystery of the Missing Bearing
Installing these new chains always makes me a bit nervous. Each time the technology is updated the tools become more specialised and the margin for error is reduced. The Campy tool shown in the illustrated How-To above is obviously the right tool for the job but Park do make a more general purpose tool that might be more useful when working on other bikes too. Not to mention being cheaper.
Park clearly have a good reputation and other tools I have from them have been flawless but this time they let me down, at least initially. I had just started breaking the chain to remove the extra 3 inches when I heard an alarming and very audible popping sound. I examined the tool and the chain and I couldn't see any signs of damage or anything that could have made the noise so I continued pushing the pin through until the handle on the chain tool would go no further. At this point it was clear something was wrong because the roller had not been pushed all the way out of the link.
The design of the tool is for the pin to sit on a small bearing in the barrel of the tool. The pin and bearing are held in place with a retaining nut so that they can be changed should the pin become worn or break.
So I unscrewed the retaining nut intending to take the pin out for a closer look. The pin was stuck fast in the body of the tool. Now I figured that if the pin didn't protrude from the tool far enough to drive the roller out of the chain then the bearing that should be in there is missing. Sure enough, it came up short by what would have been the width of the bearing. I returned the tool for a warranty replacement and got a replacement from the supplier pretty quickly.
This time I checked that the pin and bearing were present. They were, and I noticed just how snuggly the pin sits in the tool, as you load the pin and bearing into the barrel you can feel the air trapped in the barrel pushing back against the pin. You have to push it slowly in, giving time for the air to escape around the pin. My theory is that the bearing was missing originally and that the final part of the barrel is an even tighter fit - as I turned the handle the pressure built up in the barrel until the air popped out, allowing the pin to jam itself in the barrel for good.