Message 53/55
Date: 14-Aug-02 @ 04:44 PM -
RE: Virus Low End Good?
uh... try googling for "sample rate" and "bit depth".
the # of bits used to represent X doesn't directly affect the range of the values of X -- just the NUMBER of unique values X can have.
8-96.000 Mhz is meaningless -- that's governed by sampling frequency (nyquist or whatever) and NOT by bitdepth. you could have 1bit 8-96Mhz sound. (btw... 96 _M_hz?! maybe 96k)
imagine 3 points connected by two lines: 1-2-3 (the points are labelled as numbers)
let's say each of the lines is 1m long -- the whole "snake" is 2m long. so, if we measure from 1 to 2, it's 1m, and from 2 to 3, it's 1m -- the whole thing has a range of 0 (at point 1) to 2m (at point 3).
now let's say each line is 1000000m long -- wow! suddenly our range is HUGE! the whole thing is 2000000m long!!!... but... we didn't add any points/lines! how is this possible?!
easy -- we decreased the resolution. before, our "granularity" was 1m; now it's 1000000m. but the RANGE of possible values increased a LOT.
try this in CoolEdit -- take a 24bit sample and bitreduce it to 8 bits. notice 2 things:
1)the peaks and valleys are just as high -- the vertical range of the sample shouldn't have changed much. thus the range of amplitudes described by the bits is unchanged.
2)the wave looks more "jagged" -- it visibly steps from one sample point to the next. this is cause we now only have 256 possible amplitude values (height).
now try reducing it to 1bit -- you get rectangular waves. but guess what -- the range of frequencies is UNCHANGED. only the _granularity_/resolution changes.
aaaaaanyway -- bitdepth is WAY different from sampling rate. if i have an 8bit number, i can represent 256 unique values with it. but that doesn't tell me the RANGE of values -- i could represent {0,1,2,..,255} or {0,100,200,...25500} or {0,0.1,0.2,..,25.5}... it's arbitrary.
raigan