ntom;418890 Wrote: 
> As stated there is no way to reverse compression such that you get back
> to the original signal.
If all you're interested in is reversing the effects of the final
"smash-the-levels" compression that was applied during mastering (and
will accept the compression applied during recording and mixdown), then
in principle if you know *all* the compression parameters (eg. threshold
level, compression ratio, attack and release times, and thers I've no
doubt forgotten about) it is possible to get back to the original
signal (albeit with added quantisation noise).

But of course you'll never find out what all those parameters were, so
the point is somewhat academic.

ntom;418890 Wrote: 
> Similarly there is no way to reverse clipping such that you get back to
> the original waveform.  However the claims for this software suggest
> that using 'heuristics' you rebuild the waveform so that it approaches
> the original, and should at least sound better than a badly clipped
> signal, which seems feasible though whether they have any scientific
> foundation for the heuristics I don't know (just extrapolate based on
> the rate of change of the waveform on either side on the clip....or am
> I being too simplistic here)?
When a waveform is clipped, 100% of the information is lost within the
clipped region. So yes, it's impossible to reconstruct it. But you can
make an (informed) guess, typically by interpolating with some kind of
spline curve. I suppose you might even do a frequency analysis of the
preceding and following sections and construct something more
sophisticated that includes a similar frequency spectrum.

Phil Leigh;418953 Wrote: 
> Having thought about this a bit more and chatted with my DSP buddies, I
> can see how this might be approached for brief (handful of
> samples)periods of "flat top" clipping.
Formally there is no way to know for sure that something is clipped. If
you happen to see a "flat top" waveform then you can deduce with fairly
high confidence that it *is* the result of clipping, but you can't know
for sure the original signal wasn't intended to be like that. So not
only is clipping restoration a heuristic process, so too is clipping
*detection*.

Interestingly, not all clipping will reveal itself as flat topped
waveforms. If something is clipped and subsequently passed through a
filter, it'll acquire a slope. I've seen examples of this on commercial
CDs.


-- 
cliveb

Transporter -> ATC SCM100A
------------------------------------------------------------------------
cliveb's Profile: http://forums.slimdevices.com/member.php?userid=348
View this thread: http://forums.slimdevices.com/showthread.php?t=57872

_______________________________________________
audiophiles mailing list
[email protected]
http://lists.slimdevices.com/mailman/listinfo/audiophiles

Reply via email to