Prior to the modern era, coin dies were manufactured individually by hand by artisans known as celators. In demanding times, such as the Crisis of the Third Century, dies were still used even when they became very worn or even when they cracked. The die that was on the anvil side, usually the reverse (back), tended to wear out first. Since the metals that the dies were made out of weren't as hard as modern steel, the flans (the blank pieces of metal), needed to be heated prior to striking. On some Roman provincial coins, the tongs used to move the heated flan sometimed left permanent center indentations on the finished coins.
The process of making dies to strike coins in today's mint has quite a few steps. First, an artist creates a large plaster model of the coin. The plaster model is then coated with rubber. The rubber mold is then used to make an epoxy galvano. All of this takes place on a scale of around eight inches. Next, a Janvier reducing lathe takes several days to reduce the image onto a steel master hub in a process that has not changed in over a hundred years. The master hub is then tempered to make it hard. A small number of master dies (incuse) are then made from the master hub. These are then used to make working hubs. The working hubs are then used to make working dies. With each step, the number goes up. The working dies are then used to strike coins. All dies are incuse, and all hubs look like the coin being struck (with the devices raised.)
The final step of course is that the dies are used to strike images onto the planchet so that it becomes a coin.
Of course, mistakes can happen at any stage of this manufacturing process, and these mistakes are something that certain collectors look for. Coin errors that occur on the die are generally more desirable than errors made at the time of the strike. For example, a doubled die, where a date or another device appears twice slightly offset, is often a highly desired error. Strike errors are generally unique, whereas all coins struck with an error die will have the same characteristic. This makes them more easily collectible. The most famous doubled die in the past hundred years is the 1955 doubled die Lincoln cent. These trade for hundreds of dollars because the error can easily be seen by a casual observer. Many doubled die errors require at least a jewler's loop (if not a healthy imagination) to be seen. Doubling can occur at the hub stage as well. Some more recent errors are hub doubled. Most famously, there is a 1995 doubled die cent that is hub doubled.
Since coin production in the United States has exceeded 20 billion coins in some recent years, this means that a lot of dies must be manufactured as well.
On the edge of the US dime, quarter and half dollar, and many world coins there are ridges called reeds. Some older US coins, and many world coins have other designs on the edge of the coin. Sometimes these are simple designs like vines, more complex bar patterns or perhaps a phrase. These kinds of designs are imparted into the coin through a third die called a collar. The collar is the final size of the coin, and the planchet expands to fill the collar when struck. When the collar is missing, it results in a type of error called a broadstrike. A broadstruck coin is generally a bit flatter and quite a bit bigger around than the regular non-error coin of the same denomination. Modern die production
The Third Side of the Coin