Can mp3 quality be improved?
Doug McDonald wrote:
On 12/13/2011 1:20 PM, Dick Pierce wrote:
Strictly speaking, techniques such as dBx and Dolby A, Dolby
B and such, are all data-reduction compression techniques.
Their purpose is to attempt to fit as much of the "important"
data into a naroowed-bandwidth channel, be it a transmission
channel or a cassette tape. They all work on the smae principle:
they (physically) discard information which, in the eyes of the
designer, are deemed "insignificant."
I disagree with that. If the ideas behind any of those three systems
work perfectly, as they are supposed to, then they do not discard
information.
The recording media they are used with do of course cover up
information with noise, but those three systems do in fact reduce
the amount lost.
By working perfectly I mean that the encode and decode cycles
are the exact inverse of each other.
Then if that is true, the input should match exactly the output.
The simple fact is that some information in the original is
masked by the noise of the process and therefore lost.
Consider the simplest, dBx. It (the wideband one) is a simple
volume changing scheme, with the encode and decode systems being
feedback systems that are, if no sounds are out of frequency range
of the recording medium, exact inverses.
Yes, and in the process, information is lost.
Working properly, so that so that noting between the encoder and decoder
overloads, the only thing that happens is that the noise floor at the
output goes up and down. If the noise floor of the transmission medium
is much larger than half the dynamic range of the input signal, the
output should be essentially an exact copy of the input. If the
transmission system has a much larger noise than that,
then the the output will, at low levels, be a much better
copy of the input than if the encode-decode cycle were not used.
The broadband dynamic range of the coompact cassette is pretty
seriously limited, open reel tape less so, but in either case,
they can be significantly less, by not a small margin, then
the input signal. SImply consider the noise floor of a pair of
good studio microphones, equivalent to maybe on the order of
10-20 dB SPL, with signals in a concert hall situation easily
exceeding 90 dB SPL: that's a dynamic range that exceeds magnetic
media and FM broadcast: the result is loss of information.
--
+--------------------------------+
+ Dick Pierce |
+ Professional Audio Development |
+--------------------------------+
|