Judging sharpness is very subjective.
Having an objective way to measure relative sharpness without sophisticated equipment is quite valuable.
The
technique I will describe and demonstrate depends on the behavior of
point spread and is distantly related to Modulus Transfer Function
(MTF).
The precise mathematical foundation is beyond the scope of this thread and is not important to the application of the technique.To conduct measurements it is best to use a high contrast black and white image.
The image should have numerous sharp edges but most lines should not be so thin as to exceed the resolution of the camera.
(We are primarily measuring sharpness, not resolution.)
I used a photocopier test chart mounted on a piece of foamcore. Keeping lighting constant take a series of images of the test chart.
Vary only one parameter in your experiment.
I varied the distance from the film plane of the camera to the target. In
PhotoShop (or your program of choice) measure the standard deviation of
the luminance channel of a selected area that is consistent from image
to image.
The ideal selection has about equal amounts of black and white.
I chose a set of resolution bars on my test chart. The higher the standard deviation; the sharper the image. Here's my experiment:
I used my D70 and 50mm f/1.8 @f/1.8 and focused using AF at a certain distance (my secret for now).
Then
I set the camera to MF and without changing the focus took 31 images at
1mm intervals from 440mm to 470mm distance to the target. Here are the results of the entire test range.
 (I am actually surprised at how smooth the data turned out.) But does it measure what we're after?
Here are the results zoomed in on the center of the test range.
And directly below are 100% crops from the corresponding test targets.
 I think the answer is definitely "yes", this measure works.
BTW,
the crop was choosen because each horizonal black/white lines is about
one pixel; so the camera and lens do seem to resolve at the maximum
expected resoltion.
Also, the crops don't look very white due to my halogen illuminant and slight underexposure. One way to use this result is to tune our choice of the Circle Of Confusion (COC) value used in computing Depth Of Field (DOF).
If you're happy with 450-459mm then .033mm is for you.
Prefer 451-458mm? Then .025mm COC is what you want.
Only satisfied with 452-457mm, then you need to go down to a COC of .015mm Of course this test is wide open at f/1.8.
Tests stopped down should prove interesting too.
Perhaps I'll see improved sharpness at the edges of the computed DOF. So far I have tested three different lenses at a variety of distances and the technique holds up.
BTW, which distance do you think the AF sensor picked?
(Hint, it was between 447 and 462mm) 
Bill
Visit me, info and galleries at: Nikon Photos and Information | 
Printer-friendly copy
NIKON PRODUCTS FORUMS 
bclaff">
bclaff">
