I read an interview with Neil Young way back in the early days of
digital (Sound) He said words to the effect of
"No matter how frequent the sampling rate, the thing is off for as long
as it is on. The gaps are always equal to the sound so whatever you do
with digital, you lose half the music - and it may just be the half with
the soul in it!"
karl shah-jenner wrote:
Roy :
(did I send one in html with no edits? sorry!)
Karl,
I did look up nyquist on Wikipedia and it made no sense to me so I
just have to let the more technie people (like you) fight it out about the
limit of resolution. I will spend my time trying to generate random numbers
so I can win the lottery and be able to afford a Leica system.
Roy
hahaha, good luck with the lottery :)
Basically the nyquist limit is all about sampling frequency and digital.
In the digital world - binary - something is either zero or 1
if i'm sampling an analogue source at the simplest level, digital will say
it's either 'there' or 'not there'
I'm oversimplifying things a lot.. but it might help
ooh, found a good page:
http://www.evilrob.org/journal/archives/2008/04/09/nyquistshannon.html
the explanation is solid, and worth wrapping your mind around. The problem
with digital photography, sensor size and resolving power in applying the
concept of a nyquist limit is that it's usually only applied to a simpler
system - in photography where we are not just photographing verticals and
horizontals the sample rate must be higher to take into accound sampling on
the diagonals and here it falls to around THREE times the sample rate.
Effectively, to obtain 8Mp of accurate data, you'd need 3x8 = 24Mp sensor.
and that's ignoring bayer arrays and the fact a single pixels data is
interpolated from 3 filtered sensors.
really then it should be 3x24 Mp .. 72Mp
and that is the minimum LIMIT to obtaining an accurate 8Mp of resolved
information.
phew! Sounds iffy I know, but this mathematical limit can't be 'got
around'. Interpolation and clever computer guesswork in-camera in the
image processing side of things does a very good job of compensating for
the limitations but it's just guesswork.
if I have this
--------..-------
and i can only sample at this rate
I I I I I
I will only record
_ _ _ _ _
then the .. will miss being recorded. I need 2x the sample rate to see
those ..'s (well, one of them at least)
clear? as mud.. I know :(
k
