I spent some time playing with some numbers lately and it's got me confused - anyone want to make sense of this? I was considering real resolution of film and how it compared to digital.. mainly looking at things like lens resolution in lines per mm, film resolution in lines per mm, total effective optical resolution and how this relates to digital. I got onto this as someone said it was pointless scanning film with high ppi scanners as the data produced was 'junk' data - stuff that contributed nothing to the image. It was suggested that a 4000 ppi scanner offered no benefit above a 1200 ppi scanner for high quality 35mm film.. Now I have had it explained to me that a close approximation of the limit of an optical system is based on the equation (lens res x film res) / (lens res + film res) If I pop in the figure for tech pan as 100 lpmm (it can go as high as 300 so I understand) and pick an arbritrary lens res value of 80 lpmm for a 35mm lens. 100x80 / 100+80 = 44.4 lpmm So I have a piece of film 24mm x 36mm so that's 1065 vertical lines x 1598.4 horizontal lines = 1,703,255 total 'blocks' of resolution. lets call it 1.7 Mega-'blocks' of resolution. If I call these line blocks 'pixels' (blocks of data) , that's a resolving power equating to a mere 1.7Mp (!) Sure the image has lots of detail and looks great, but the actual *resolving power* in line blocks, pixels, whatever is pretty small. OK, so I went an looked at some digital sensors to see what they could resolve in real terms.. some CCD/CMOS 's have as little as 70% of the surface covered in photosensors, but I decided to ignore that - also forget demosaicing and the rest of the digital stuff for the moment - I'm just trying to get my head around the actual resolving power of a 3Mp Coolpix 995 camera for comparison to the 35mm film example above. looking at the coolpix 995 for a moment we have a sensor that's 7.2x5.3mm and yields a 2048x1536 pixel image. that's around 285 lpmm resolution. To eliminate further complications I'll also forget the sensor resolution limit should also consider the "Nyquist" limit, which reduces the actual resolution substantially (it's calculated as only 145 lpmm for this sensor) - agin let's forget this for a moment. let's assume the lens has a res of around 80 lpmm again so (AxB) / (A+B) = approximately 62 lpmm total optical resolution. (niquist limit calcuations in place, the actual resolution is really limited to 51 lines per mm, but I said we'd forget that..) If there is an optical resolution of 62 lpmm then (62 lpmm x7.2mm) x (62 lpmm x 5.3mm) = ( 446.4 total lines x 3428.6 total lines) = 0.14 Mega'blocks' (Mp again?) of real resolution The D30 Canons 27.7 x 15.1mm sensor yields 2160 x 1440 pixels or 95.4 pixels per mm Assume a lens of 80 lpmm res again, that means an overall res of 43.5 lpmm. total resolution possible then is 43.5x22.7 x 43.5x15.1 = 0.64 Mega'blocks' or megapixels jumping up the quality lines somewhat, let's look at the Canon EOS-1Ds Mark II with a maximum resolution of 4992 x 3328 pixels covering a 24x36mm area. 4992/36 = 138.7 pixels (lines) per mm 3328/24 = 138.7 pixels (lines) per mm again using the same 80lpmm lens, we have a total maximum limit to resolution (again not considering the Nyquist limit) of 51 lines per mm 51 lpmm x 24 = 1224 lines vetical resolution 51 lpmm x 36 = 1836 lines horizontal resolution 1836 x 1224 = 2,247,264 mega'blocks' or 2.2 Mp in real terms, including the Nuiquist limit in the calcs, the resolution is actually half this or 1.1 Mp of real resolving power in the optical system, still less than that of the good ol' tech pan. If one can assume then that a stated 16Mp camera has only half the actual resolving power of the tech-pan, could one not be excused from concluding the tech pan shot is equal to a 32Mp image? Would not one then conclude a 5000 ppi scanner producing a 5000x7500 pixel (37.5Mp) scan is actually gathering legitmate data? karl
