Dave Rosser said:

I realise I should have given the depth of focus formula in terms of subject distance (note this formula strictly falls apart if you try to put infinity in as the subject distance but if you pretend infinity/infinity is 1 and not indeterminate it works.)

d=cDf/(D-F)

where

d=depth of focus in either direction

c=limiting circle of confusion

D=distance to subject

f=f-number of lens aperture

F=focal length of lens.

I have had futher thoughts on this and with a subject at infinity the formula reduces to d=cf, i.e. the depth of focus is proportional to the circle of confusion, this leads to some interesting conclusions.

The same autofocus mechanism is used on the F6, D2Hs and D2X but the circle of confusion requirements are different.

From experience over many years with film you are pushing things if you go beyond 12.5x enlargement from 35mm, taking the accepted 0.25mm as the point at which normal sighted person cannot distinguish between a point and a disc then the circle of confusion you need for the F6 is .02mm. (0.25/12.5)

For the D2Hs and D2X let us assume that there is no point, due to aliasing concerns etc., in having a circle of confusion less than 2 pixles in diameter. For the D2H this gives a circle of confusion of 23.3/1232 (23.3mm wide sensor 2464 pixles accross) which equals .019mm while for the D2X the circle of confusion is 23.7/2144 which equals .011mm.

From this it can be seen that the tolerence in setting up the focusing mechanism in the D2X must be approximately half that used in the F6 and D2H if similar results in focusing accuracy are to be achieved.

With the number of complaints about back focus problems on D2X I wonder if Nikon have fully taken this decreased set up tolerance into account on their production line.