home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!news.claremont.edu!nntp-server.caltech.edu!andrey
- From: andrey@cco.caltech.edu (Andre T. Yew)
- Newsgroups: rec.audio
- Subject: Re: CD Sound Quality
- Date: 26 Dec 1992 17:56:33 GMT
- Organization: California Institute of Technology, Pasadena
- Lines: 47
- Message-ID: <1hi6chINNnom@gap.caltech.edu>
- References: <1992Dec22.090725.11365@leland.Stanford.EDU> <7490273@hpfcso.FC.HP.COM> <vanz.02bd@tragula.equinox.gen.nz>
- NNTP-Posting-Host: punisher.caltech.edu
-
- vanz@tragula.equinox.gen.nz (Martin Nieuwelaar) writes:
-
- >Thanks also to Adre Yew for the text on the sampling theorem.
- >Everything seems straight-forward up to the part at the end
- >to do with the inverse transforms. If I have this right, what
- >you do is point sample the signal, do a Fourier transform,
- >remove all components above half your sampling rate, and do
- >an inverse Fourier transform.
- >This is then convolved with the inverse transform of the box filter. ???
-
- Right so far. I guess I should have made this clearer.
- The only reason we're Fourier transforming (FTing) and
- inverse-FTing back is because many times, especially in
- things like this, FTing a signal and the operations we're
- performing on it gives a clearer view of what's going on.
-
- Now, if I only said, "No, you need to interpolate the
- points with a sinc, not a line", then who knows where we might
- have been. Since you clearly understand what a lowpass
- filter does (sorry if I seemed demeaning of your intelligence),
- I'll say this instead: the box filter is a low pass filter.
- If you multiply it against the FT of your signal, you let
- only the lower frequencies pass through and you wipe out
- all the higher frequencies. Now, to see what the effect of
- lowpass-filtering our signal has in the time domain, ie.
- what it does to our sample points, we perform an inverse-FT
- to get back to the time domain from the frequency domain.
-
- Just a note of caution here: all those other people
- who say you get rid of your higher frequencies in a square
- wave by lowpass filtering are correct. But, you can't
- just use a normal box filter to do that, you have to use
- a different filter (in fact it turns out to be something
- that looks like a sinc function in frequency domain, except
- it's not all real, and it gets chopped off after a certain
- frequency). This is called a zeroth-order hold (it makes
- your samples look like stairsteps), and point sampling
- could be called infinite-order hold. It's a slightly
- convoluted (no pun intended :) exercise to show that as
- you increase the order of your interpolation, your
- interpolation shape approaches a sinc, ie. point sampling.
- Just more party trivia for you.
-
- --Andre
-
- --
- Andre Yew andrey@cco.caltech.edu (131.215.139.2)
-