| From: | Andrew Crowe |
| Date: | 5 Aug 2000 at 00:32:55 |
| Subject: | Re: Mpeg 3 playback. |
Hi Everybody,
> It does a pretty good job, but I think I'm right in saying that it
> also steals CPU cycles from you. Basically though, at the end of
the
> day, pseudo 14 bit sound from an 8 bit sound chip is not going to
be
> on a par with that from a 16 bit or greater soundcard.
Well, I dunno, my amiga sounds better then a lot of PC sound cards
I've listened to :)
>> By the way, are you sure that mp3's are saved as 16 bit, or any
>> number of bits for that matter? I'm not sure, but I have a feeling
>> it saves waveform data, not the actualy wave (like instead of
>> saving a wave like a sample, it saves it as a frequency and volume
>> or sommin.) If that's true, then in theory a decoder could decode
>> at an infinite bit res.
>
> I don't pretend to understand the intricacies of how an mp3 data
> stream is built up. But since, in the case of CD audio anyway, it
is
> built from a 16 bit wave, and since it outputs a 16 bit waveform,
> how it actually stores that waveform is immaterial.
Nope, a CD stores the wave as it is, like a line of points plotted
on a curve (say -127...-100...-50...0...50...100...127...100...50 etc
for a sine wave). Now, I'm not sure exactly, but an MP3 stores it as
a formula to make the wave, so it would be like Sin X. Therefore you
could expand the mp3 by plotting the inbetween points.
Technically you could smooth the raw data too, but it would loose
accuracy.
> As to infinite
> bit resolution, what would be the point??
There wouldn't be any really. I doubt even an audio engineer could
tell the difference between a 16bit and inf. bit sample
> Thus, if you decoded a 16 bit sample at a higher bitrate, then
> surely you'd just "pick up holes", i.e., at some points, you'd be
> decoding white noise, because there'd be no signal there? Besides,
> if the pundits are correct (and to be honest I don't think they
are)
> only the audio signal between 20Hz and 20KHz is important, because
> they are the audible limits that the human ear imposes. Thus 44100
> KHz is about as high a samplerate as you need, and this is
> adequately provided for by a 16 bit sample rate.
You're mixing rate and sample resolution. The rate is how many
values are used to make the sample per seccond. A CD uses 44100
values/sec to create sound. The bit res is how big those values can
be.
> Whats more important, at least so it seems to me, is the bitrate of
> the mp3, i.e., whether it is 96, 128, 192, etc, KBPS. This, I
think,
> is the rate at which the mp3 encoding is sampling the waveform of
> the wave sample. So that the mp3 isn't actually a 16 bit sample, so
> much as a bit rate, i.e., sampled so many times a second (kpbs)?
The
> greater the kbps is, the bigger the sample will also be. An
infinite
> bit rate would generate a massive file size.
>
> Ah well, you can all tell me where I'm wrong now :)
Well, I dont know the details myself, but the mp3 bitrate has
nothing to do with the sample frequency/bit rate, its just how much
data per seccond the mp3's formula for the sound can take up. The
higher this bitrate, the more complex the formula for the sound can
be, so the better the mp3 will sound.
And anyway, as the mp3 is encoded as formulas, I'm guessing that
you can increate the bit resolution just by using more decimal places
when you work the formulas out. It wouldn't affect the sound quality
though, only make the decoder take up more cpu, which is probably why
decoders only do 16 bit, even on 32/64 bit sound cards.
See ya :)
Quote carefully and read all ADMIN:README mails