Re: podcasts?

On 7/13/05 5:31 AM, "Curt Fischer" typed:

> Sean wrote:
>> On 7/12/05 2:47 PM, "Curt Fischer" typed:
>>
>>
>>> Have you ever tried a side-by-side comparison? The difference is
>>> noticeable, although perhaps not with a 196 kbps mp3, which is what the
>>> guy was talking before my top quoted response above.
>>>
>>> My understanding of the mp3 encoding process is not detailed, but I
>>> thought that more or less, Fourier transforms of the source are taken as
>>> functions of time (which involves inevitable frequency cutoff due to the
>>> finite time domain over which the FT is taken). Next, peaks in the
>>> frequency domain for a given time point are "sharpened" by removing
>>> lower-intensity frequencies present around the peak frequency. This
>>> works because of some sort of masking phenomenon with human hearing that
>>> says that if you have a say 50 dB sound at 440 Hz then no one can hear a
>>> 30dB sound at 441 Hz. This sharpening provides most of the
>>> compression, right, but is obviously a lossy process, right?
>>
>>
>> Of course I agree that it is "lossy." It's just not a simple frequency
>> filter or other straight-ahead waveform alteration. It's based on some
>> high-falutin' psychoacoustic stuff having to do with eliminating irrelevant
>> information.
>> There's a good explanation in Sound on Sound:
>> http://www.soundonsound.com/sos/may00/articles/mp3.htm
>
> Thanks for an interesting link.
>
> I don't know what you mean by straight-ahead "waveform alteration" that
> you keep mentioning. What is it and how would it help compress audio
> data? Also, I had a few details wrong but the gist spot on about the
> way the compression works: elements of the Fourier (or discrete cosine)
> spectrum are eliminated based on the psychoacoustic criterion you mentioned.
>
> Maybe I will have to repeat the side-by-side comparison on a high kbps
> mp3, but I don't have my good headphones with me here in Japan.

By "straight-ahead waveform alteration" I meant, but probably did not
convey, the simple hacking off of higher frequencies and lower frequencies
or cutting the sampling frequency in half or halving the bit depth of the
information. That kind of thing would treat the whole audio file as a single
thing and completely screw up the sound.
As you saw in that article, MP3 compression is a much more sophisticated
process.
As I said before, I utterly agree with you that MP3 is lossy. And it could
be that having frequencies above around 20khz filtered affects the tone or
colouration of a cymbal crash or snare hit in a way that you can perceive.
If so, I wish I had your ears. Mine are shot.

.

Relevant Pages

  • Re: OT Ipods
    ... >> There is absolutely nothing inherent to Fraunhauffer .mp3 compression ... >> the file size without affecting the sound. ... I won't argue that tube amps aren't the bomb. ...
    (rec.sport.golf)
  • Re: OT Ipods
    ... >>> Analogue tape always had a distinct limitation to how well it could record ... >>> There is absolutely nothing inherent to Fraunhauffer .mp3 compression ... >>> the file size without affecting the sound. ...
    (rec.sport.golf)
  • Re: OT: Mp3 quality Final
    ... Many of us enjoy the pursuit of experimenting and building sound ... The problems you see with MP3 are not the 44khz sampling rate. ... The problem is how the compression is done. ... hearing range, which affects those with extended hearing ranges as well as ...
    (rec.boats)
  • Re: podcasts?
    ... noticeable, although perhaps not with a 196 kbps mp3, which is what the guy was talking before my top quoted response above. ... My understanding of the mp3 encoding process is not detailed, ... 30dB sound at 441 Hz. ... way the compression works: elements of the Fourier ...
    (sci.lang.japan)
  • Re: podcasts?
    ... kbps MP3 and the original? ... side-by-side comparison IME it is easy to tell the mp3 from the uncompressed source by listening to high end sounds like e.g. the cymbals or snare. ... Human beings generally can't hear frequencies above that, ... quality of the sound. ...
    (sci.lang.japan)