At 12:55 AM +0000 11/18/00, atk@xxxxxxxxxx wrote:
Ok -- I'm trying to figure out digital resolution for 'decent' 35mm. Lets
assume each pixel is high fidelity. A good film/lens combo gives you 50 lpm
(you might be able to get 100 -- but 50 is pretty good for every day usage).
You need two rows per 'lpm' so 100*35*24*100 = 8400000 or 8 meg pixel (around
24 meg bytes). According to these calculations an 8meg pixel camera should be
on par with typical high (but not optimal) 35mm system. I thought the number
was much higher -- what is wrong with my calculations ?
Two things are wrong. First, there's the notion that 50 lpm is really
adequate (it might be, but that's pushing it). Second is that you're
reckoning without alignment issues. Film grain is much finer than 50 lpm,
and is randomly dithered, so it doesn't matter where your lines pairs
hit the film plane, as long as they're focused. With 2x sampling on pixels,
in contrast, all you need is for the line pair to hit the sensor grid on
a half-pixel boundary to get nothing at all. You need something like another
factor of two to get randomly positioned lines and edges right. (You can
see this even on high-end scanners where weird things happen in parts of
an image depending on how grain lines up with scan dots.)
The other thing, which is not quite relevant to what you said, is that
the optics of your hypothetical camera have to be up to the task as well;
the simplistic calculations I've made seem to show that a sensor pitch of
3-4 microns and focal length of less than 10mm doesn't really cut it.
paul
--
Paul Wallich pw@xxxxxxxxx
< This message was delivered via the Olympus Mailing List >
< For questions, mailto:owner-olympus@xxxxxxxxxxxxxxx >
< Web Page: http://Zuiko.sls.bc.ca/swright/olympuslist.html >
|