[Richard Kuschel]

Is there really such a thing as lossless Audio Data compression?

I have heard of Flac and Monkey, and I am certain there are others.

Do any of them do what is claimed ie. you get out exactly what you purt
in.

I realize that many compression schemes work on reduction of repeated
patt4erns, but I don't see how this could work with audio.

[pumpcat]

I have personally verified that FLAC, Monkey, and Shorten are
bit-for-bit accurate, and I would assume that other codecs like Apple
Lossless and WMA Lossless are as well.

There is a good technical explanation of the process on
http://flac.sourceforge.net

-Matt M

Richard Kuschel wrote:
Is there really such a thing as lossless Audio Data compression?

I have heard of Flac and Monkey, and I am certain there are others.

Do any of them do what is claimed ie. you get out exactly what you purt
in.

I realize that many compression schemes work on reduction of repeated
patt4erns, but I don't see how this could work with audio.

[anahata]

Richard Kuschel wrote:
Is there really such a thing as lossless Audio Data compression?
[...]
Do any of them do what is claimed ie. you get out exactly what you purt
in.

Yes, they all really do that. You don't get more than about 50%
compression (programme dependent).
It's much better than ZIPing a WAV file, though.

I realize that many compression schemes work on reduction of repeated
patt4erns, but I don't see how this could work with audio.

It doesn't look for repeated patterns. I think it works by differences
between consecutive samples, which have a predominance of small values
and therefore respond well to Huffman coding (variable word length with
shorter words for more frequent values). That's what I'd try first if I
was designing a lossles audio compressor, anyway.

--
Anahata
-+- http://www.treewind.co.uk
Home: 01638 720444 Mob: 07976 263827

[Laurence Payne]

On 21 Sep 2006 15:48:08 -0700, "Richard Kuschel"
wrote:

Is there really such a thing as lossless Audio Data compression?

I have heard of Flac and Monkey, and I am certain there are others.

Do any of them do what is claimed ie. you get out exactly what you purt
in.

I realize that many compression schemes work on reduction of repeated
patt4erns, but I don't see how this could work with audio.

Sure. There are plenty of lossless compression systems. ZIP, RAR...
And some optimised for audio. But you won't get the degree of
compression that we've come to expect from MP3.

[Scott Dorsey]

Richard Kuschel wrote:
Is there really such a thing as lossless Audio Data compression?

I have heard of Flac and Monkey, and I am certain there are others.

Do any of them do what is claimed ie. you get out exactly what you purt
in.

Sure.

I realize that many compression schemes work on reduction of repeated
patt4erns, but I don't see how this could work with audio.

It works surprisingly well with audio, all things considered. You don't
get a huge amount of compression because there isn't all that much redundancy
in the source material, but you'd get more than you'd expect. Do a google
search on FLAC... there is a bunch of propaganda on their encoding algorithm.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Julien Bernier]

---beside note----

anahata wrote:
It doesn't look for repeated patterns. I think it works by differences
between consecutive samples, which have a predominance of small values
and therefore respond well to Huffman coding (variable word length with
shorter words for more frequent values). That's what I'd try first if I
was designing a lossles audio compressor, anyway.

Hi, I'm a coder and I totally agree with anahata. Huffman is a "simple"
way to explain how monkey, winzip and any zip/unzip works. Except they
are MUCH more complicated than this, packed with algorithms to do their
job efficiently. my 2 cents...

[Richard Kuschel]

Julien Bernier wrote:
---beside note----

anahata wrote:
It doesn't look for repeated patterns. I think it works by differences
between consecutive samples, which have a predominance of small values
and therefore respond well to Huffman coding (variable word length with
shorter words for more frequent values). That's what I'd try first if I
was designing a lossles audio compressor, anyway.

Hi, I'm a coder and I totally agree with anahata. Huffman is a "simple"
way to explain how monkey, winzip and any zip/unzip works. Except they
are MUCH more complicated than this, packed with algorithms to do their
job efficiently. my 2 cents...


Thanks for the explanations

Ok, that makes a lot more sense. I once used a Sony system that had
96dB dynamic range but only about 24dB s/n ratio. It used 4 bits on
some kind of a sliding scale. It got geally loud and also very soft,
but you could always hear the grunge behind it.

Now that we have determined that these schemes exist , is anybody using
them or they really too much bother.

[Romeo Rondeau]

"Richard Kuschel" wrote in message
ups.com...
Is there really such a thing as lossless Audio Data compression?

I have heard of Flac and Monkey, and I am certain there are others.

Do any of them do what is claimed ie. you get out exactly what you purt
in.

I realize that many compression schemes work on reduction of repeated
patt4erns, but I don't see how this could work with audio.

What you're talking about is RLE (Run Length Encoding). It doesn't work on
audio files. What is done on audio files is Delta compression (encoding the
differences)

[Chevdo]

In article om,
says...



Now that we have determined that these schemes exist , is anybody using
them or they really too much bother.


I use wavpack sometimes, if I've got something that is bigger than will
normally fit on a CD, but will fit if wavpack is employed. So far that's only
happened once, and I'd probably use it more often if DVD media weren't so
cheap. I know the brother of the author of wavpack, and he says his brother is
obsessed with always trying to eek out just a bit tighter lossless compression,
and is always trying to figure out ways to keep wavpack ahead of the
competition. I'm not sure it is ahead of all its competition anymore, though,
and WinRAR with 'media file' optimization seems to pack nearly as tight as
wavpack.

[Paul Stamler]

"Richard Kuschel" wrote in message
ps.com...

Ok, that makes a lot more sense. I once used a Sony system that had
96dB dynamic range but only about 24dB s/n ratio. It used 4 bits on
some kind of a sliding scale. It got geally loud and also very soft,
but you could always hear the grunge behind it.

Now that we have determined that these schemes exist , is anybody using
them or they really too much bother.

If you buy tracks from the Smithsonian/Folkways website (most of the
Folkways catalog is available at $0.99/song), you have the option of
downloading in .mp3 or flac.

Peace,
Paul

[anahata]

Richard Kuschel wrote:
Ok, that makes a lot more sense. I once used a Sony system that had
96dB dynamic range but only about 24dB s/n ratio. It used 4 bits on
some kind of a sliding scale. It got geally loud and also very soft,
but you could always hear the grunge behind it.

That's certainly not lossless compression. It's a crude floating point
system. A similar technique is used, but on many narrow frequency bands,
in MP3 encoding.

Now that we have determined that these schemes exist , is anybody using
them or they really too much bother.

I considered it as a way of archiving an album I had produced on CDR;
FLAC compression of the audio left more than enough space to put all the
booklet and tray card graphics on the same CD as the audio.

--
Anahata
-+- http://www.treewind.co.uk
Home: 01638 720444 Mob: 07976 263827

[Scott Dorsey]

Richard Kuschel wrote:

Ok, that makes a lot more sense. I once used a Sony system that had
96dB dynamic range but only about 24dB s/n ratio. It used 4 bits on
some kind of a sliding scale. It got geally loud and also very soft,
but you could always hear the grunge behind it.

That isn't compression at all, that's just a nonlinear scaling. You can
think of that as sort of like putting dbx Type I around a digital system;
the scale is not linear and there is now a logarithmic relationship
between the digital value and the analogue voltage.

Now that we have determined that these schemes exist , is anybody using
them or they really too much bother.

DVD-A and SACD both have lossless compression systems built into them.
How many folks are using DVD-A and SACD, though....
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Richard Crowley]

"Scott Dorsey" wrote ...
Richard Kuschel wrote:

Ok, that makes a lot more sense. I once used a Sony system that had
96dB dynamic range but only about 24dB s/n ratio. It used 4 bits on
some kind of a sliding scale. It got geally loud and also very soft,
but you could always hear the grunge behind it.

That isn't compression at all, that's just a nonlinear scaling. You
can
think of that as sort of like putting dbx Type I around a digital
system;
the scale is not linear and there is now a logarithmic relationship
between the digital value and the analogue voltage.

Sounds like a marketingspeak description of what we
would call in normal English: "floating-point".

RC [Sat, 23 Sep 2006 06:32:50 -0700]:
Sounds like a marketingspeak description of what we
would call in normal English: "floating-point".

Sounds more like adaptive-differential to me, as in,
your old ADPCM. No way it would have anything to do
with floating point, not with only four bits.

http://en.wikipedia.org/wiki/Pulse-code_modulation

:
An ADPCM algorithm is used to map a series of 8 bit
PCM samples into a series of 4 bit ADPCM samples.
In this way, the capacity of the line is doubled.
The technique is detailed in the G.726 standard.
:

--
40th Floor - Software @ http://40th.com/
iPlay : the ultimate audio player for mobiles
mp3,mp4,m4a,aac,ogg,wma,flac,wav, play+record
parametric eq, xfeed, reverb; all on a mobile

[Scott Dorsey]

Richard Crowley wrote:
"Scott Dorsey" wrote ...
Richard Kuschel wrote:

Ok, that makes a lot more sense. I once used a Sony system that had
96dB dynamic range but only about 24dB s/n ratio. It used 4 bits on
some kind of a sliding scale. It got geally loud and also very soft,
but you could always hear the grunge behind it.

That isn't compression at all, that's just a nonlinear scaling. You
can
think of that as sort of like putting dbx Type I around a digital
system;
the scale is not linear and there is now a logarithmic relationship
between the digital value and the analogue voltage.

Sounds like a marketingspeak description of what we
would call in normal English: "floating-point".

No, it's not floating point. There is no seperate exponent and mantissa
being stored, just one scalar.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."