478 A. E. CONRADY ON RESOLUTION WITH 



This immediately raises the question as to what resolving 



power is obtainable with objectives of greater N.A. than l/3rd of 



that of the condenser. Fig. 2 supplies the answer. Again the 



full circle represents the aperture of the objective, the dotted 



circle that of the condenser. The best possible result will 



evidently be secured from light just entering the extreme margin 



of the condenser aperture as at a, with a structure of such 



fineness that the second spectrum only just gets into the 



farther margin of the objective at c. The first spectrum will 



be midway between a and c, at b, well within the aperture 



of the objective, thus showing the reduced resolving power. 



Now, we can at once see from the diagram that the distance 



from a to c is half the aperture of the objective plus half 



the aperture of the condenser; and as the distance from b to c, 



which measures the resolving power, is half of a-c, we obtain 



the second result : 



(2) If a dark-ground illuminator has an aperture less than 

 three times that of the objective, then the limit of resolving 

 power of the combination is measured by one quarter of the sum 

 of the numerical apertures of illuminator and objective. 



Thus if we again work out a limiting case, we may take the 

 limit of N.A. for the condenser at 1*40, that for the objective 

 at rOO, sum 2-40, and l/4th'of this or 0*60 will give us the N.A.. 

 of an objective which would have the same resolving power with 

 extremely oblique direct light. In other words, under these 

 conditions we should only realise 60 per cent, of the extreme 

 resolving power of our objective. 



There is another important result to be deduced from our 

 diagrams. Take fig. 1, and assume that the condenser is fitted 

 with a " wheel-diaphragm " approximately equal to its full aper- 

 ture, so that the dotted circle represents the narrow ring of 

 light which would pass through. The limit of resolving power 

 w^ould still be realised. But supposing we substituted an object 

 a little coarser than the limit. The spectra would be closer 



