Is lossless really as good as wav? Options

[Sunhillow]

bhoar,
your dictionary forgot to explain the difference between inherent losslessness and losslessness with added computing

[AndyH-ha]

The question isn't difficult to understand but it appears some people manage it anyway. Think of it as analogous to writing to CD-R. The process can, and usually is, good enough that one can later extract a file that will prove to be bit for bit identical to the original source that was written to the CD-R. However, if the computer is too busy with other tasks at the same time it is writing to the CD-R, the data on the CD-R might end up with defects. In some cases that will mean the CD-R is unplayable, in others it means the music is mixed with occasional bad noise.

The inquiry, reworded, was 'does/can anything similar happen during FLAC playback? Might the data being decoded, or the data going to the DAC, be corrupted because total system load is too high?" The answer is yes, it is possible. The FLAC source isn't effected by this. The process of decoding might also remain pure, but the decoding might not be accomplished in real time. If there is any interruption in flow because of too much multi-tasking, what comes out your speakers might not be what you expected. The same thing can happen during recording from analogue sources, even though the computer that is capable of making perfect recordings. Clearly this is not a frequent problem for FLAC playback.

[JensRex]

What you are describing, is possible (though highly unlikely) for all data stored on a computer.

Edit: On second thought, I'm not sure it's possible at all (hardware failure aside). In the case of system overload during audio playback, you'll just get gaps in the decoded output. I guess you can describe that as loss if you want.

This post has been edited by JensRex: Nov 13 2006, 22:49

[greynol]

What you are describing, is possible (though highly unlikely) for all data stored on a computer.

(addressing the bolding which I added)
...for flac maybe, but not for Monkey's Audio.

I also note your edit.

[bhoar]

bhoar,
your dictionary forgot to explain the difference between inherent losslessness and losslessness with added computing

How would you convert to another format without added computing?

-brendan

[Firon]

Magic.

[Spare Tire]

I propose that we use the term lossless only for the encoding process, and use the term sonically identical for the decoding process.
Some people claim playing music from ram sound better than from HDD, eh, they are both from "lossless" sources. Whether that is true or not doesn't matter, just that it is possible for it to sound different, depending on the timing. And i don't care if people hear it or not, i know i wont.

This post has been edited by Spare Tire: Nov 13 2006, 23:40

[Tung]

You guys are all very entertaining!

Funny thread.

I'd add that FLAC, as a codec and format, is independent of the hardware.

FLAC files can be played while being decoded; FLAC files can be played after they're decoded (to WAV)..

While you can experience hardware problems or interruptions in the playback caused by a busy CPU (FLAC wouldn't keep it very busy, likely another process or processes), you could also experience the same problems while playing back a WAV file. Does that mean there is a problem with the WAV file?

These possible interruptions are independent of the FLAC or WAV files themselves.

One thing I love about FLAC is that it is designed to use very little processing in the decode process, and this encourages hardware player vendors to support the format.

Long live FLAC.

[Firon]

Some people claim playing music from ram sound better than from HDD, eh, they are both from "lossless" sources.

Now that's just a very stupid claim. It's not going to happen. If the HDD is bad, then your files aren't gonna decode properly. If the RAM is bad, it'd have be a pretty large amount of errors by RAM standards to affect the sound (and only if its to the decoded data). And if it happens before decoding, then again, you're going to have errors, not different sound. Plus, bad RAM tends to also cause all sorts of problems, so the odds of it only affecting sound are even LESS likely.

The odds of either one happening are quite low.

[Bourne]

ok, fellas some of you are getting me wrong... I'm not talking about able or not able to hear artifacts... I'm talking about knowledge of having or not artifacts. THINK of a lossy implementation that could set the bitrate of a sample up to 1411 kbps, while 32 kbps in the "white noise" space. Got it now?

Imagine if LAME could encode some samples up to 1411kbps, and not be limited to 320kbps. That is what I'm talking about when I say 'virtual lossless' and 'not technically having any kind of artifacts'... it's not about hearing it or not...

[Firon]

Yeah, um, that's called basically unrestricted VBR, which vorbis and AAC can do (well, to an extent..).
It's still not "virtual lossless". Stop using that term.

This post has been edited by Firon: Nov 14 2006, 03:02

[greynol]

That's funny, 'cause I was going to take the other side of it.

There are codecs that encode with no artifacts whether they can be heard or not heard, they are lossless codecs with bitrates that are variable. Some of the more common ones are flac, WavPack and Monkey's Audio.

There is no such thing as "virtual lossless" as there is no such thing as lossy but with no artifacts.

This post has been edited by greynol: Nov 14 2006, 02:46

[kwanbis]

What I was asking, was if the extra processing introduced any quality change.

If it did, i wouldn't be lossless.

[kjoonlee]

Let's say you have an old laptop that gets its cooling fans revved up under heavy processing load. (Do old laptops have that feature? Anyway....) If you play 3D-intensive games on it, the laptop will get hotter and start whirring louder.

Does it make your 3D game behave weirdly? No. (But if it overheats, your laptop may crash...) Heavy load doesn't affect calculation.

Same thing with lossless audio. Heavy load doesn't affect calculation, and lossless decompression is usually not a CPU-intensive process.

[Rotareneg]

Ok, if you've got a bad quality motherboard with bad power filtering and a bad sound card that also has poor power filtering, it's possible that the electrical interference generated by the processor (and hard drive, and everything else in the computer for that fact) could generate audible noise at the analog output of the soundcard. But if that's the case, doing anything with the computer is going to cause problems. And of course the actual digital data will not be affected (otherwise the errors would corrupt all data the computer is processing and cause it to crash,) only the analog output from the D/A converters and/or amplifiers on the sound card.

[kjoonlee]

There is no such thing as "virtual lossless" as there is no such thing as lossy but with no artifacts.

Depends on how you define artifact. If you define it as "perceptible difference," then you have "transparency."

[greynol]

That's fine, and it brings me back to post #14 of this thread.

EDIT: ...and note the possible definition that you've introduced, "perceptible difference" will not be the same from person to person (that is to say the difference, not the definition). Bourne, OTOH is talking about artifacts as if they aren't based on subjective listening which was the precise context of my response.

So, can you convince Bourne that what he's asking for already exists or if it doesn't already exist, enlighten those of us (me in particular) as to what he's talking about?

This post has been edited by greynol: Nov 14 2006, 04:35

[Firon]

Depends on how you define artifact. If you define it as "perceptible difference," then you have "transparency."

Is it even theoretically possible for a lossy algorithm to never have perceptible differences (within the human range of hearing)?

This post has been edited by Firon: Nov 14 2006, 05:02

[Rotareneg]

Sure, but the problem is making an efficient lossy algorithm that doesn't require significant resources to decode.

[Spare Tire]

ok, fellas some of you are getting me wrong... I'm not talking about able or not able to hear artifacts... I'm talking about knowledge of having or not artifacts. THINK of a lossy implementation that could set the bitrate of a sample up to 1411 kbps, while 32 kbps in the "white noise" space. Got it now?

Imagine if LAME could encode some samples up to 1411kbps, and not be limited to 320kbps. That is what I'm talking about when I say 'virtual lossless' and 'not technically having any kind of artifacts'... it's not about hearing it or not...

Kinda like saying the sum of an infinite number of fourier series tend towards the exact approximation of any function. If you add an infinite number of different codecs.... Eh.

[Lynch]

Let's say you have an old laptop that gets its cooling fans revved up under heavy processing load. (Do old laptops have that feature? Anyway....) If you play 3D-intensive games on it, the laptop will get hotter and start whirring louder.

Does it make your 3D game behave weirdly? No. (But if it overheats, your laptop may crash...) Heavy load doesn't affect calculation.

Same thing with lossless audio. Heavy load doesn't affect calculation, and lossless decompression is usually not a CPU-intensive process.

If I understand Countryman's question, I'd disagree. Playing a 3D game on an old laptop would be a totally different experience compared to playing it on a faster desktop in terms of fps. But, perhaps he does mean problems with calculation. In that case I'd agree.

Seeing the amount of discussion of what was actually the question, I'm interested to see countryman's reaction to the answers sofar.

This post has been edited by Lynch: Nov 14 2006, 06:26

[kjoonlee]

Is it even theoretically possible for a lossy algorithm to never have perceptible differences (within the human range of hearing)?

Yes, upsampling to 192 kHz.

If I understand Countryman's question, I'd disagree. Playing a 3D game on an old laptop would be a totally different experience compared to playing it on a faster desktop in terms of fps.

OK, it isn't a perfect analogy. Nevertheless, consider a game that would run just as well on both systems, but one that would make a laptop whirr.

[IgorC]

It's 02:26 and I can't sleep after reading this topic. I just was thinking in my bed about infinity of human bluntness (stupidness). I think human bluntness has no limits and less has dimensions (2d 3d ?).
Somebody can be wrong double-infinitely.

Bourne 5 mins ago you didn't know that FLAC is vbr and at the same time introduce a new therms like 'virtual' lossless. Even young man who can think logically know that ' each txt page can be compressed differently' depends of complexity. Speaking roughly the same way each second of wav has different complexity -> different compress ratio -> VBR.

ok, fellas some of you are getting me wrong... I'm not talking about able or not able to hear artifacts... I'm talking about knowledge of having or not artifacts.

Does such kind of knowledge is possible to introduce to any program language? Until now PC hasn't a status of a human brain!
From this point of view your statement is infinitively wrong.

But let's just think that PC has a status of human brain. But artifact can exist or not ... for different people.
So how PC will take decision?
Including this impossibility you're already wrong twice infinitively.


There is no human who can change the informatic (mathematic ... call it whatever) concept of LOSSLESS.
LOSSLESS = no data loss = bit identical

And what the huck is 'virtual'?
Virtual = not real
Vırtual lossless = not real lossless = not bit identical information = lossy


Imagine if we could have sex virtually. There wouldn't neither one man live in planet.

It is discrete.

[WaxCyl]

A lossless algorithm can have some losses in some circumstances such as when the lossless codec is not fully lossless in which case the partially lossless algorithm should be redefined as "semi-lossless", although this may not always be the case in some circumstances such as a "fully lossless" codec, which by definition does not have any losses at all....
I hope this clears it up.

[Bourne]

ok, i stop calling it lossless, virtual hell, the hell... etc...

i am calling out for a new lossy codec, that has unrestricted VBR up to 1411kbps. Ok? I do not accept codecs that can do that thing only "to an extent" at all...

This post has been edited by Bourne: Nov 14 2006, 08:10