Accurate measurement of radius is critical for high tolerance lens manufacturing but not usually so demanding for the primary mirror of amateur telescopes. Some methods in fact are rather crude such as the method of using feeler gages with a straight edge to measure the sag of a mirror. This method is quite suitable for a telescope maker who hasn't yet built his telescope tube and doesn't require a highly accurate focal length. On the other extreme optical systems often require tight tolerances in radius. These tolerances are usually expressed as two numbers for power and irregularity for the number of fringes (in green light normally) as measured against a test plate. For example 10/2 means a radius deviation showing no more than 10 fringes and the fringes must not deviate more than 2 fringes from a circle. These tolerances were sometimes as loose as 30 fringes in power or as tight as 5 fringes or even less of power. I have a commercially made spherometer with an accuracy of about .1% but I found it often wasn't good enough. I will describe here the method I used to measure more accurately than with even this commercial spherometer.
Balls are much more affordable, I have 6 mm ruby balls on my spherometer and they won't scratch a lens easily but cost about 11 dollars each from Edmunds (there's probably a cheaper source... ball bearings will work and are cheaper but aren't as wear resistant). I glue these with epoxy to a ¼-20 stainless steel machine bolt and trim the diameter on a lathe (not really necessary). The threaded shank allows me to raise or lower the balls for convex or concave curves. This bolt screws into an aluminum base and is secured with a nut after making all three balls the same distance from the base. Some have used pulleys as spherometer bases which is a neat idea.
I find it easier and quicker to calibrate it another way (not requiring micrometers) using a reference radius. Simply put I use a reference radius part and solve for the ball feet radius. Since we know the ball diameter, the radius of our radius reference and the sag measurement we can solve for the feet radius using the following formula:
Feet radius = square root ((radius-ball radius)^2 - ((radius-ball radius) - sag)^2
A simple way to calculate this formula is to use an excel spreadsheet (which can be downloaded here). To use the spread sheet you simply:
Be sure to follow the sign convention of concave curves having negative (-) radius and negative sags, vice versa for convex curves. Otherwise you get a math error.
This method depends on having an accurate reference radius. They are easy to come by in optical shops but an amateur can often find a chipped or not so good mirror at a swap meet to use for a reference. A short focal length is best to use. For the greatest accuracy the spherometer feet should be spaced as big as possible but you have to be careful not to exceed the measurement range of your indicator. I'm always changing the feet radius on my spherometer for small or large parts or for steep curves and find this method a lot more practical than measuring the feet radius with mics. The only thing you must do is to be sure the leg height the same.