Lossless not lossless?

Topic: Lossless not lossless? (Read 10517 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.

Lossless not lossless?

Reply #25 – 2003-09-01 07:57:40

Quote

Doesn't one of the lossless codecs use Joint Stereo if the user so desires? In that unique case, then, lossless wouldn't be quite lossless, not would it? Or would the Joint Stereo work differently that Lame-MP3's mid-side/left-right joint stereo does?

Yes, a lot of lossless codecs use some kind of joint stereo. The simplest form is indeed just Mid/Side stereo. But this does not make them a bit lossy.

-Eugene

Lossless not lossless?

Reply #26 – 2003-09-01 08:09:04

The difference between "pure stereo" lossless and "joint stereo" lossless is similar to the difference between compressing two WAVs in RAR non-solid mode and compressing the same two files in RAR solid mode. (this is a much simplified explaination though, lossless audio encoders are far more optimized to utilize corelation between channels).
I really don't understand where some of you get those ideas about "lossless not being lossless". If one of those codecs was actually altering the sound data at that big bitrate, it would become absolutely useless, because lossy codecs can achieve transparency at 2x-3x lower bitrates. Lossless codecs are meant to store (archive) original unaltered sound data, rather than cut file size while preserving the way it sounds.

Lossless not lossless?

Reply #27 – 2003-09-01 10:52:02

[span style='font-size:21pt;line-height:100%']lossless = no loss = identical to the original![/span]

"Just An UnCool Cat" = no knowledge = worth it for the entertainment value and pure exasperation!

Cheers,
David.

P.S. let's just put another link to that thread:
Are CDs lossless? including a nice definition of lossless.

Lossless not lossless?

Reply #28 – 2003-09-05 02:40:27

CRC32 tests show that the originals remain exactly the the same when compressed and then uncompressed. Full stop.

Lossless not lossless?

Reply #29 – 2003-09-05 06:07:04

CRC add unnecessary uncertainty. A 32-bit CRC can match on different files once out of about 4 billion (4e9) times. Actually, that's probably wrong, but my CRC knowledge isn't that great. You have the original, the encoded version and a decoded version. You can compare the files byte for byte for 100% certainty (or within the error rates of your hardware at least).

But yes, lossless == no loss == the guy is an idiot.

Lossless not lossless?

Reply #30 – 2003-09-05 06:12:32

Quote

CRC add unnecessary uncertainty. A 32-bit CRC can match on different files once out of about 4 billion (4e9) times. Actually, that's probably wrong, but my CRC knowledge isn't that great.

I was thinking that this must be the case. After all if CRCs were absolutely perfect you should be able to play back the file from just the CRC.

Lossless not lossless?

Reply #31 – 2003-09-05 06:21:10

Quote

Quote
CRC add unnecessary uncertainty. A 32-bit CRC can match on different files once out of about 4 billion (4e9) times. Actually, that's probably wrong, but my CRC knowledge isn't that great.

I was thinking that this must be the case. After all if CRCs were absolutely perfect you should be able to play back the file from just the CRC.

HUH?!

Lossless not lossless?

Reply #32 – 2003-09-05 09:14:10

Quote

Quote
CRC add unnecessary uncertainty. A 32-bit CRC can match on different files once out of about 4 billion (4e9) times. Actually, that's probably wrong, but my CRC knowledge isn't that great.

I was thinking that this must be the case. After all if CRCs were absolutely perfect you should be able to play back the file from just the CRC.

Meh ?
CRC32 can't store entire information from some file (after all its ONLY 4 bytes), it's just a number that it always the same for two identical files but extremely unlikely to be the same for two different files.
Probability of finding two different files having same CRC32 is low enough to use it for comparing files in real life.
You can't restore original data from CRC32, just like you can't guess numbers if you only know their sum.

Lossless not lossless?

Reply #33 – 2003-09-05 09:29:37

Quote

I was thinking that this must be the case. After all if CRCs were absolutely perfect you should be able to play back the file from just the CRC.

IMO this is correct: Absolutely perfect means there are no 2 files of a given length that result in the same checksum. This is only possible if the checksum is of the same size as the original file, e.g. the "checksum" would be a copy of the file.

Lossless not lossless?

Reply #34 – 2003-09-05 09:51:25

Quote

"checksum" would be a copy of the file.

Which wouldn't serve its main purpose anymore.

Lossless not lossless?

Reply #35 – 2003-09-05 11:29:35

The other very funny thing in that thread (which I'll post now since this thread came back up) is that he explains how he heard differences between the original and lossless files.

Quote

i've been there, done it on an experimental basis.. and whilst for 99% of people the A/B difference i found was indistinguishable, it was just enough (on the most marginal basis) that no benefit of file size reduction was worth that compromise for me.

The idea of this guy switching between two identical files, and thinking to himself "ooo, yup, I can hear a difference there" is so funny!

And the serious point is that in confirms the placebo effect yet again: he heard a difference because he expected to hear a difference. Even though he was just listening to the same thing.

blind testing people - it's essential.

Unless you're a pure subjectivist, when you'll say that he must have heard a real difference, therefore it was down to the slightly different HDD velocity when reading different sectors of the disc, giving rise to small but perceptable changes in the current drawn from the PSU, which affected the sound card output in some way.

Cheers,
David.

Lossless not lossless?

Reply #36 – 2003-09-05 12:57:01

Quote

This really isn't a big deal, and certainly nothing worth bashing anybody about, but sometimes we all need to step back and realize that it takes a little patience to educate the public about audio encoding, and there are a lot of ways for a person to become misinformed. I'm guessing this guy will figure it out eventually, just like the rest of us did...

It might not be a big deal, but it's certainly something you'd want to stop while in the making. If you have a group of people thinking he is 'the knowledgable one', they will go around thinking that lossless is just a facsimile and could end up hard to convince otherwise because they were told this from a 'knowledgable' source. Misinformation is damaging I guess I'm saying?

Lossless not lossless?

Reply #37 – 2003-09-05 13:16:40

Quote

"I can ABX a 1 from a 1"

Lossless not lossless?

Reply #38 – 2003-09-05 13:47:57

Differenciam, I am hurt by your implications of the caliber of posters at that board. I post at those forums on an occasional basis. I'm not implying I'm some highly knowledgeable entity but neither do I pretend to be on every topic that surfaces on said board.

Uncool Kat has posted here several times and I believe he is no longer posting on the Iriver forums due to the anonymous user harassment. He seems to just be unconvinced about the validity of lossless mainly due to his experience with lossy formats and their inherent nature of destroying bits which DO affect sound quality.

Lossless not lossless?

Reply #39 – 2003-09-05 13:59:07

While we on lossless discussion again, I'd like to ask a tangible question:

Of course all of the people who ever looked at WMP 9 noticed that lossless option in the Music Copy tab, adds the word Mathematicaly to it. Now a thread on another board I visit (thread now deleted due to extensive trolling) there was a person claiming that the reason for word Mathematicaly to be there is that if someone decided to sue MS by saying just like the guy in the thread above, that Lossless has lost something dispite advertised "loss-less-ness", MS present a Mathematical proof that their lossless is in fact lossless. They didn't release this proof 'couse that would disclose the way the compression algorithm works, and whatever public opinion about business practices of MS is, they do have good mathematicians working for them (some with Field Medals), is that true.

Plus if you check out Flac docs, you'll see that a lot (if not all) of lossless compression is based on some works by several guys, which could imply that MS using those works also to some extent. Now if MS got a mathematical proof, woldn't it suggest that if done correctly other compressors would also fall under same proff (with adjustments) since they all derive from the same source?

Lossless not lossless?

Reply #40 – 2003-09-05 14:07:33

Quote

He seems to just be unconvinced about the validity of lossless mainly due to his experience with lossy formats and their inherent nature of destroying bits which DO affect sound quality.

This is possible but, to be honest, I think (from reading the whole thread) that he is just trolling. He is not even *considering* the fact that he *might* be wrong, even after a bunch of people tried to tell him that the files were bit identical. Plus, as Pio2000 pointed out, why would anyone use lossless, and put up with the huge files if it wasn't actually lossless? It just doesn't make any sense.

Lossless not lossless?

Reply #41 – 2003-09-05 14:28:43

Quote

...why would anyone use lossless, and put up with the huge files if it wasn't actually lossless? It just doesn't make any sense.

I do agree with your point in general, but you going over board with that last one. There are Millions of people smoking, eating fatty foods, and taking drugs, if it's bad why do they do it? It just doesn't make any sense.

Lossless not lossless?

Reply #42 – 2003-09-05 14:38:55

Quote

While we on lossless discussion again, I'd like to ask a tangible question:

Of course all of the people who ever looked at WMP 9 noticed that lossless option in the Music Copy tab, adds the word Mathematicaly to it. Now a thread on another board I visit (thread now deleted due to extensive trolling) there was a person claiming that the reason for word Mathematicaly to be there is that if someone decided to sue MS by saying just like the guy in the thread above, that Lossless has lost something dispite advertised "loss-less-ness", MS present a Mathematical proof that their lossless is in fact lossless. They didn't release this proof 'couse that would disclose the way the compression algorithm works, and whatever public opinion about business practices of MS is, they do have good mathematicians working for them (some with Field Medals), is that true.

That just sounds like speculation.

The more mundane answer is probably that a lossy codec is "perceptually lossless" (at least to marketing people!) if it (supposedly) doesn't audibly alter the sound.

Saying something is "mathematically lossless" simply removes this possible confusion - the numbers you get out are the same as the numbers you put it.

Anyway, after saying that 128kbps is CD quality, these interfaces were always going to struggle to explain something that was better! "really, genuinely sounding even more like the original CD than the actual CD does - er...." etc. To cut through the cynicism this hype has caused, just add the word "mathematically", and everyone knows that there's no argument about this one!

Well, that's my guess anyway.

Cheers,
David.

Lossless not lossless?

Reply #43 – 2003-09-05 19:33:55

Quote

Of course all of the people who ever looked at WMP 9 noticed that lossless option in the Music Copy tab, adds the word Mathematicaly to it. Now a thread on another board I visit (thread now deleted due to extensive trolling) there was a person claiming that the reason for word Mathematicaly to be there is that if someone decided to sue MS by saying just like the guy in the thread above, that Lossless has lost something dispite advertised "loss-less-ness", MS present a Mathematical proof that their lossless is in fact lossless. They didn't release this proof 'couse that would disclose the way the compression algorithm works, and whatever public opinion about business practices of MS is, they do have good mathematicians working for them (some with Field Medals), is that true.

The proof would be pretty trivial. All lossless audio codecs I know of work essentially like this:

encode:
- take input f(t)
- magically conjure up g(t)
- store g and the difference(error) which is f(t)-g(t)

decode:
- compute g(t)
- add the error

Proof that it's lossless:

g(t) + [f(t) - g(t)] = f(t) + g(t) - g(t) = f(t)

QED

Notice it doesn't matter even what g(t) is, unless of course you care about the compression ratio

Maybe then someone would challenge the losslessness of the error-coding stage; for most of those, proofs are already available.

Josh

Lossless not lossless?

Reply #44 – 2003-09-05 22:12:41

Quote

The more mundane answer is probably that a lossy codec is "perceptually lossless" (at least to marketing people!) if it (supposedly) doesn't audibly alter the sound.

Arrrgh.... Using the term lossless in a sentence discussing a subjective concept (transparency).

Just kidding!

I know we've been through the debate about multiple concepts of "lossless", and we obviously agreed that the term has to be qualified for proper use...meaning
"Lossless compared to what source?"

I guess I'm just sensitive to seeing the word lossless used in any way but the absolute. [Forgive me...I'm standing at the gates again. ] I'd use the term tranparent to mean what you say with the phrase perceptually lossless, though I know (assume?) you mean the same thing.

Sorry...just being neurotic. I know you know far more than I do about this kind of thing...I just get very philosophical about the sanctity of absolute concepts (and terms). Once we start saying "perceptually lossless", and JoeAverageMusicFan hears this, how long before the word "perceptually" is dropped among the less informed, and people start going around saying, "Yeah, my MP3s are lossless."

Edit: And I know that it's the fault of marketing organizations misusing these terms, but all the more reason for us (and everyone) to permeate them with the word transparency. I'd much rather hear them say "128kbps is transparent" than even "320kbps is lossless", though even the former is not true for a lot of people.

[span style='font-size:7pt;line-height:100%']Edit: Clarification...[/span]

Lossless not lossless?

Reply #45 – 2003-09-05 23:24:02

Quote

Anyway, after saying that 128kbps is CD quality, these interfaces were always going to struggle to explain something that was better! "really, genuinely sounding even more like the original CD than the actual CD does - er...." etc. To cut through the cynicism this hype has caused, just add the word "mathematically", and everyone knows that there's no argument about this one!

umm... I don't understand how 128kbps and transparency came up. I always thought that Lossless means it can be decoded to original WAV, or played back in the encoded form as if it was original WAV (on the fly decoding I presume). And I thought that the MS lossless was just another kind of compression, with different compression ration, playback speed. So I don't understand the point you make about transaperency, how can it not be transparent, we aren't about WMA std 128kbps we talking lossless, arent' we?