Alright, from my thread on whether or not WAV files are lossless, for all intents and purposes, with regards to ripping direct from a CD, it seems like, yes, WAV is lossless.

So going back up the chain from a lossless audio format, so far all data is exact, from lossless audio file <-> wav.

Given that... are CDs lossless? In recording a CD, there has to be a conversion from analog to digital. At the same time, a human voice and instruments, I believe, are not limited to samples per second. So in theory, unless I am wrong (which I will admit that I definately could be), to have a lossless audio CD, the result would have to be an infinite bit rate and an infinite sample rate.

This would mean because CDs are at 16/44.1k (I think?), there is quite a bit of data that is lost. Unless you define lossy, as compression.

So... is a CD lossy?

And if not, how far up the chain do you have to go to get to the point of loss, if there is loss at any point at all? (Chain = Lossless audio format -> Artist)

::sigh::

and so a can of worms is opened...

if you are asking whether an increase in the sampling rate or bitrate, for example that employed by a SACD, will result in less loss (better quality), the answer is no, from what i have read on this forum. it will not result in better quality. its like frame-rates in video games: once you're over about 60 frames per second, any more is a waste.

have i actually learned anything from reading this forum? or is this just a bunch of BS?

The artist himself is lossy.

Why do you think a particular recording may require multiple takes?

answer 1) every A/D and D/A conversion is "lossy". the better the converter the better the sound.

answer 2) a loss is only important if someone can hear the loss. as far I know, 16 bit should be sufficient to be transparent.

can I expand the original question, please? thanx

what format is original audio for Movies stored in? In the booklet for T2:Extended Edition (2003) it says that when sound is converted to DVD format the audio is reduced in size by 98%. somehow I don't think original was wav

The data on Audio CDs is not compressed, so the terms "lossy" and "lossless" do not apply.

i wouldn't call it loss. cds are just made in a certain standard. i don't think that you can call this a quality loss because the only thing that has no quality loss is when you're listening to an instrument live and unplugged. this has infinite resolution.

What a worthy and useful thread. >_<

1: Lossy / lossless refers to compression, and thus has nothing to do with CDs
2: Many modern recordings are completely digital, from the microphone to the CD. Therefore analog to digital is irrelevant.
3: No. You cannot hear the sound "degredation" on a CD. Hell, most people can't even seem to tell the difference between 128 bit MP3 and CD. But CDs contain more data than the air can even carry. There is no loss, at least not a loss perceptible to humans, or detectible on modern audio equipment.

The "original" digital medium for audio is always PCM.

This part is not necessarily accurate. Nearly all sounds we hear are not electric (even 'electric guitars'). In order to capture that sound digitally, it has to be sampled (which is the fundamental concept of digitization). True lossless would have a sampling rate 2 times higher than the highest harmonics created by a sound.

Even then, don't forget that air is one of the ULTIMATE lossy compressors. So if we're being excessively pedantic here, there is loss between the sound source and the transducer of the microphone.

The point is that if you want to be absolutely technical about things, there is loss well before it makes it to the final media. Do you ever hear this? Again, technically yes, depending on your reference. One statement I've always wanted to make with ABXing is that no setup (or VERY few) can pass an ABX from a live performance. (Not that it's really been attempted).

Also, this was more to try and draw some parameters from the original question, how far back in the recording/processing mix before you get to lossless (essentially).

I don't even know where to start

Without getting all cranky, I would suggest you go out and do some reading before starting any more "profound" threads like this one.

Any recording is lossy of course, since it does not record everything happening in the room at the time of the original performance. Audio is lossy because you loose the smell and feel of the room. Stereo is lossy because it puts all the subtle acoustic directivity information into only one of two channels. The digital recording or mastering process is lossy because, at a basic level, it ultimately has to quantize things into 1s and 0s. Lots of other little losses of course also come into play.

Obviously (well, at least to me) we are not concerned with that on this page. We are concerned with "psychoacoustic audio compression". Please for godsakes read up on some audio basics.

Jebus, you are just a hair too slow

Is there any actual point to this, Mr. ? Or are you trying to out-l33t somebody? This is like asking what is the sound of one hand clapping (ABX'ed, please) or how many angels fit on a pin's head.

I mean, anybody who was made some research knows that NO format, maybe ever, will be Lossless in relation to the originally produced sound (the artist performance, that is).

Either deal with it or don't. Just don't try to play smart here.

(I am really making an effort here...counting: 1,2,3,4,5...)

You quoted me, but I don't really think you read what you quoted: ... completely digital, from the microphone to the CD ...

I wasn't talking about the air. I was talking about the equipment. (Though if you'd bothered to read the next line, I did mentioned the air later.) If you want to be overly anal, then yes, I suppose it undergoes an analog->digital conversion going into the microphone. But professional microphones will sample at rates far higher than the human ear needs. So that's completely moot.

And if you're going to be so pedantic, you should note that electric is not the same as digital. Electric guitars do produce an electric signal, simply not a digital signal.

And ABXing from a live performance? Ugh. The point of a recording studio is to produce an output which exceeds live performance. That doesn't mean that the equipment can't record data well enough to fully capture a live performance. It just means that such an effort is useless.

I'd bet that well over 99% of the data loss associated with recording comes from the mastering process (read: intentional loss), rather than from equipment limitations.

P.S. Data loss that the human ear cannot detect is completely irrelevant as far as I'm concerned. I don't give a damn if they lose frequencies only dogs can hear, nor do I care if they can't capture (on CD) noise above the threshhold of pain. Such effort is a waste.

First, I should tell you that my argument was not directed at you, but at the original poster. Your second point seemed like the best place for me to squeeze in.

As for your quote, I read it, do you understand it? Microphones have an analog component in them, a transducer*. Hence the reason I make mention that sound is not electric. If it were, then we wouldn't have to worry about the loss due to the transducer, now would we? In either case, there is an analog to digital conversion that must take place, hence the loss there. Also, if microphones sample far higher than the human ear needs, then there must be resampling that occurs since 44.1kHz is barely above what it needs to be.

* I'm not aware of any purely digital transducers in microphones. If there are, then I'm potentially wrong about it being analog. However, turning audio energy into electrical energy is ALWAYS lossy.

As for your 3rd point, which I felt had no bearing to my post, you mention that CD contains more data than air can carry. I felt that this was an error at best and a blatently misleading statement at wost. Air can carry 23kHz tones just fine, whereas CDs can't. Granted it's above the threashold of human hearing, but you didn't mention that until your response.

Finally, although I agree with you on your statement that most audio loss will occur in the mastering process, I'm not certain that it's not in part due to equipment limitations. Otherwise, there would be absolutely no reason for 24-bit 96kHz recordings.

Now before you flame me, be sure at this point to reread my first 2 sentences

there are people who say that it's not necessary at all.

The moment audio is *sampled*, it becomes lossy. Even if a sound it sampled at 1,000,000,000,000,000,000,000 kHz, it is still not infinite, the way our ears hear live audio.

Anyway, I think this thread is more lossy than CD audio will ever be.

I have a headache...I'm gonna go find some aspirin now...

Some things...

Good microphones are just analog, there's no conversion to digital inside. The conversion is performed at a ADC outside the microphone.

Being picky, one could say that the fist lossy stage happens when the mic captures the sound. Here you go from a 3d soundfield to a mono-signal from a single microphone. But this happens more because of the usual micing technique. On a binaural recording this wouldn't happen, since you record exactly what goes to the ears.

On the other side, a microphone can't capture sound with infinite precision, so one could say that microphones are lossy, so the first loss always happens at the mic. However, I believe good mics can capture sound with greater accuracy than the ear has, so this should be no problem in practice. So, in a binaural recording with good mics, one could say that there would no audible loss at all at the mic recording stage, since you record exactly what goes to the ears, with a greater accuracy than the ear has.

In practice, many mics are not transparent, in the sense that they alter in an audible way the sound, but this is intended, because it gives a type of sound that the person making the recording wants. And binaural recordings don't give either the kind of sound the people making recordings want, so to talk about true "transparency" and "lossless" recordings doesn't have much sense even if there were perfect recording mediums.

Now, an AD/DA conversion is no more lossy than recording the signal on tape, for example. In fact it is quite less lossy. This is because an AD/DA conversion does nothing to the signal but adding some minuscule background noise and a cutoff of frequencies over near half the sampling rate. But nothing else, both in theory and using a good ADC/DAC. Dither and the reconstruction filters take care of all the "intuitive" problems of digital recording and playback.

And, in my opinion, 24/96 recordings are not neccessary at all. 24/44.1 is enough for recording and editing, but 16/44.1 is enough for playback a properly mastered signal. This is because the noise added and the frequency cutoff at a proper DA stage of a 16/44.1 sampled signal, is beyond what the ear can notice.

Edit: some few additions about the lossy nature of mics.

The neural signal from your ear to your brain is digital. It's quantised (1 or 0), but not sampled (the pulse can happen at any point in time). The transduction process from pressure wave in air to neural signal is very lossy. The neural signal itself is very very noisy, but our brain processes it well.

Microphones and loudspeakers are significantly less "ideal" than any other component in the audio reproduction chain. That's not to say that your ear can't hear a lesser imperfection due to bad sampling through the greater imperfection due to a poor loudspeaker. It's just to say that the design (and placement!) of microphones makes a greater contribution to the sound that the design of the A>D and D>A convertors.

Are CDs good enough? We've been here before - see the FAQ:
http://www.hydrogenaudio.org/forums/index....7516#entry74075

Cheers,
David.

That's a good point. My argument should really be that after the processing our brains do, although we don't really interpret every bit of the sound, it is still impossible for *any* digital recorded format to perfectly match every bit of what we would hear live.

In other words, to sound "perfect" to a person's ears, a recording would have to only lose those bits which that persons brain would lose/discard during mental processing anyway. Two main problems with that...(1) Different people will hear differently, since they have different brains doing the processing, and (2) Even the same person would hear the sound a little differently at each listening session as the brain doesn't process anything the exact same way twice in a row.

Even if these issues are completely non-perceptible 99.99999% of the time, I contend that the word "lossless" is a big, serious word that only means one thing. It's the same concept as being pregnant...you are or you're not, no in-betweens. Once a single bit of data is lost, then you have a lossy recording, whether or not any brain in the world could have ever interpreted that lost bit. So, since technically *any* digital recording of any kind that was ever made is lossy compared to live, the next consideration is transparency, which is subjective to each person who listens to a piece of audio (and on a side note, and as everyone here knows, also happens to be the very center of interest in psychoacoustic audio compression).

Keep in mind that my argument is that a digital recording cannot technically be lossless (as I said before, because of sampling). But calling, for example, FLAC lossless means soemthing else, because you're not comparing it with the original audio source (i.e., live performance), but rather with the original uncompressed (already recorded) audio file it came from.

Is CD audio lossless? It can't be...ever...not technically. Without some miraculous new recording technique, in my opinion, *no* digital recording format could ever be lossless. Even if you set up a million of the world's best quality microphones plugged into the world's best quality digital recording equipment around one guy playing an acoustic guitar. One ever-so-insignificant bit of sound is lost between sample #954 and sample #955. It's now a lossy recording. Maybe not as lossy as a 64kbps MP3, but the term "lossless" could technically never be correctly used to describe it, nonetheless.

A better question, I think, would be "Is CD audio transparent?". Or even better: "Is CD audio good enough?" My answer? I've got not a single complaint...


Edit: Re-reading my post, I realize how extremist my point-of-view sounds. Oh well...I'm a typical Scorpio...black-and-white, no middle ground, ever...it's perfect or it's flawed...period.

If my post sounded rude, I should apologize. I thought you were arguing with me, and I can be overly argumentative at times.


I suppose this is technically correct. So I'll just yield the point. There is always some analog -> digital conversion. But again, if the rate is high enough to capture all a human can hear, it's not lossy, at least not from a practical standpoint. And yes, there is resampling going on. There always is. I'm pretty sure that most mastering is done at above 44.1, so it has to be downsampled before it hits the CD.


Well, I did say "Data loss that the human ear cannot detect is completely irrelevant as far as I'm concerned." The only thing CDs are really missing is the ability to carry tones at volumes loud enough to cause physical pain. And honestly, I'm not so concerned about that.


I firmly believe that most data is lost as a result of compression. That's an intentional process.

Actually, there are microphones which perform the digital conversino internally, but I withdraw my previous statement that they are in common use. I was apparently wrong.

But the loss (perceived or imperceptible) isn't between sample points - your suggestion that digital audio is lossy because of this is wrong, if I'm understanding you correctly (which maybe I'm not!).

Any recording of the electrical signal which comes from a microphone, whether that recording is analogue or digital, is lossy. However, you can put bounds on what you lose in a digital system. Sample two million times per second, and all frequencies up to one million cycles per second (i.e. 1MHz) will be stored - since the ear can only hear up to 20kHz, that's overkill - certainly lossless as far as human hearing goes!

Then, the bit depth defines the noise floor. 16 bits doesn't match human perception (though it's more than enough for most recordings) - 24 bits is more than enough, though we can't even design convertors of this accuracy, partly because of basic electronics and physics, but also because the 24th bit is quieter than the sound of silence ;-).

There are endless discussion on the limits of digital in the threads in the FAQ - take a look if you doubt either of the above.


The reason recording is lossy is because it doesn't create the exact same sound pressure variations throughout this listening space. The spatial information just isn't there. Stereo isn't bad, 5.1 isn't much better - neither take you to a "you are there" or "they are here" listening experience, though it's possible to get close to either.

There are techniques like Wave Field Synthesis (warning: link to pdf file) and Ambisonics (or even Ambiophonics if you don't mind having your head clamped in place!) that aim to, and (in some cases) even can re-create the audio signal at your ears exactly in theory - all that remains is the distortion introduces by the microphones and speakers.


But all this idea of sound reproduction perfection is so far away from the reality of most modern recordings that it's almost laughable. And, what's more, given a perfect sound reproduction system, our brains won't be happy: there's nothing to look at! So, the best recordings are just a little bit "better" than life. Not too much better - a great record producter once described it as "like putting a little bit of rouge on the cheeks of a beautiful woman" - you take perfect and make it more perfect! Most modern recordings slap bright red paint on the cheeks and blue paint in each eye!

Cheers,
David.

EDIT: added the quote because the discussion had moved on between hitting reply and hitting post!

I believe you are right. And that CD's are therefore lossy. Please let us know when you have achieved sufficient deity to offer hardware and/or software remedies for this serious situation.

Regards,
Madrigal

To hear how differences between people can affects things, The CIPIC HRTF Database for binaural recording is interesting to play with, to hear how differences between people plays a role concerning perception. You need MATLAB to use it though.

Btw: I think AtaqueEG has described this thread very well already

But to suggest that this shows that different people perceive things differently is misleading. I'm sure people do perceive things differently, but different HRTFs do not prove this - in fact, if anything, they suggest the opposite...


We all have different shaped ears - that's simple biology. That's why we have different HRTFs. BUT our brains decode those HRTFs to give us an impression of spatial location. Our brains decode our own HRTFs, so that the sound coming from our ears sounds normal to us. Everyone else's ears are doing the same for them.

If you listen with someone else's HRTFs, your brain is still trying to decode your HRTFs, so the results just sound plain weird, or wrong. But what you hear in this experiment is not what the other person hears - when they use their ears with their own HRTFs (obviously!), their brain decodes their HRTFs and it sounds normal [i]to them/i].


So, the differences due to different shaped ears are removed by the brain, allowing different people to hear, or at least experience, similar things for a given external sound source.

When you inject the source straight into the ear (e.g. using headphones, so bypassing the pinna, and hence HRTF), that's when it gets weird.

Also, if you accept that people know what a sound at, say, straight ahead should sound like to them; and they also know what a sound at, say, 30 degrees left should sound like, then it's clear that real sources at either location will sound like they are, well, exactly where they are! BUT a phantom source created by two loudspeakers (e.g. stereo) may not sound the same to all people, because what the HRTFs are adding, and what the brain is subtracting, are not the same. so the illusion of a central source hanging between two loudspeakers may not work equally well for different people.

Which brings us back to the lossless idea: if you can re-create the sound field that was in the original recording space (within certain parameters of accuracy); not a trick or illusion of it, but the same air movements, then any differences between different listeners won't matter.


What a thread! It all begs the question: what do things sound like to other people? And what do things look like? How do you know that, when I see green and when you see green, we both see the same thing? The relevance to this is that, if you make a perfect copy of a green piece of paper, the copy will still look green to everyone. If the copy changes the colour slightly, then different people's perception of the change will be different. That's why any lossy process is so subjective.

Cheers,
David.

Your question is a very good one and you are clearly trying to understand the theories of sampling and quantisation in intuitive terms. That is good. Please feel free to ignore those that mock you or your questions. They have clearly forgotten what it was to be a novice in a specific domain

In signal theory terms, any real working implementation (not just a theoretical mathematical formula) from analog to digital (consisting of both sampling and quantisation) and back to analog is inherently lossy. This is more often the case, if analog signal is not frequency (or more generally bandwidth) limited, but holds true even if it the analog signal is bandwidth limited.

Most people who blindly quote Shannon fail to understand that what works in infinite time in calculus may not work in real-time in a dsp. So, in practise the situation is a little bit worse than in theory (yes, even for sampling only sans quantisation and back to analog).

Now, do analog to signal digital conversions in general have practical audible implications in bandwidth limited audio ment for human consumption?

Sometimes yes, at least for sampling rates much below 44.1 Ks/s and/or word lengths much below 16 bits. However, CD bandwidth at 44.1 Ks/s and 16 bits is sufficient in terms human hearing to capture a good (if not theoretically perfect) audio signal that has been digitized from an analog source.

Whether this can be audibly improved upon (for human listeners) by increasing either the sampling frequency and/or the quantisation space, well... that's a very heated debate. Quite many seem to believe that things can be improved by upping sample rate / bit depth, although some think it's a waste of bits in terms of engineering and what are the practical limits of audio reproduction in most circumstances. That is, some consider that we can't even get anywhere near the 44.1kHz / 16-bit theoretical maximum performance out of our state of the art playback equipment (in an ordinary listening room), so it's not perhaps wisest to try and fix playback problems by increasing sample rate and bit depth in the signal.

However, most seem to agree that a higher sampling rate and longer word length has practical and audible benefits for audio archival and for later processing of the digitized audio signal in such a manner that the processing artifacts do not become as easily audible as they would come with a 44.1khz/16bit original signal.

As to the "is CD lossy" question, that can be answered in many ways. My personal stance is that no, CD is not lossy. The digitisation of analog signals for CD use purposes is often lossy (for good practical reasons). So, the complete preparation of analog material for digital distribution is lossy: microphone recording, pre-amplification, mixing, sampling/quantisation, production, mastering, etc. These all lose information in signal theory terms, when compared to the original analog waveform of an acoustic space in which the recording was done.

However, the transfer of 44.1 Ks/s / 16-bit digital signal to CD does not (and most often is not) lossy. It's just (practically, in terms of data retrieval) a perfect transfer of bits onto a medium that can be re-read over and over again and still get the exact same bits out (practical limitations apply).

As to your lossy chain question, I'm completely lost

As to the implied "is CD enough" question. I'm not going to touch that with a ten-foot pole

Best regards,
Halcyon

PS Do not feel afraid to ask questions that you feel are important or hard to answer yourself. If you don't get an immediate answer, move on and ask elsewhere. Try to set up your own working hypothesis (just as you did above) and try to improve it with others. You'll learn much better by asking questions, trust me (yes, I have some background in educational sciences, so I feel confident in my recommendation).

I feel that's misleading. A pure 10 Khz sine wav count be perfectly captured onto a CD with no loss. That's a rather manufactured example, but you can move from there into more realistic applications, too. A lot of audio can be perfectly captured in 16/44.1.

But there will either be noise or distortion about 90dB below digital full scale. It's not a perfect representation of a pure 10 kHz sine wave because there's other stuff at other frequencies (all be it very quiet) which should not be there.

btw, Halcyon, that's a nice, patient explanation - but he did have the question answered - it's in the FAQ, to which he was directed.

Cheers,
David.

As always you know what you are talking about
I can't see what I wrote that was wrong, but in the context of this thread I guess it could be misinterpreted.
To make it clear, I agree with what you say, and this wasn't ment as an example of how people perceive things differently. I tried to give an example of how things that works for one person doesn't work for another. In this case a HRTF. The goal is as you say, to make both perceive the same thing, but that may not work if the HRTF generated with person A is used with person B. As person B's hearing is trained with his own body/HRTF. As you also say it's hard/impossible to say if they do in fact perseive the same thing anyway, but differences between people affect the training process. Hence you can't just change things that has been a part of this process.

Not sure if I manage to communicate what I try to say, the part quoted above tells me I don't, but I think I have the understanding of it anyway.

In my opinion, something like 24/50 (or maybe 24/48) if properly implemented, is lossless from the ear point of view, under any circumstances, because the dynamic range and bandwidth achieved match or are over than what the human ear can perceive, even over the most demanding situations and the best golden ears.

24/44 maybe could not match hearing of people that can hear isolated ultra-high tones, people that could be say, 0.01% of people, and very young people. But it would be surely enough in my opinion for real world music, where those isolated tones don't exist, or are too soft to be perceived.

16/44 doesn't match max. possible dynamic range of human ear under any circumstance, that is, from the threshold of audibility to the threshold of pain. But it does indeed match the dynamic range of real-world and practical listening, since our listening rooms have noise that is quite over the threshold of audibility, and because nobody in practice gets even close to the threshold of pain when listening for pleasure.

That is an interesting claim, but I cannot find scientific evidence to support your claim (there could easily be some, I'd like to know if there is). The whole hearing apparatus (including muscular coding in middle-ear and basilar membrane + hair cells in inner ear and neurochemical pathways to thalamus and onto cortex) does adapt to signals being fed to them (both over time and almost instanteneously)

This is not AFAIK the same as "remove the differences due to different shaped ears". Especially if we are talking about high level cognitive auditory perception, rather than low level auditory evoked potentials. Perception remains invariably a subjective experience, as you have stated.

Or have I completely misunderstood you (there's always that possibility)



Err... how do you do that? With ordinary headphones you still get the coupling to head and pinnae and ear canal. Although you lose torso and head occlusion to some degree (depending on headphone construction), you don't "bypass HRTF".

Even for canalphones (say Etymotics) the ear canal acts as a (somewhat closed) resonant chamber and as we know, the length of ear canal is not standard from human to human.

Did you mean something else (again)?

I have other comments to various other parts of your wonderful post (including direction discrimination and perception invariance), but have to leave my own comments truncated due real-life calling.

Good thread! Keep the discussion alive and we all learn.

cheers,
halcyon

This could well be (to a statistically large enough degree), but AFAIK currently it's easier to get near 16-bit resolution out of a 24-bit implementation and better amplitude linearity with 96kS/s than 44.1 KS/s for the usable human hearing range.

With higher bit-depth and and sampling rate we can skip some real-time imperfect bandwidth transforms (such as the current de facto multi-bit delta sigma oversampling with requantisation and dither) that are a must have for most inexpensive 44.1kHz/16-bit implementations that try to achieve the theoretical and audible limits of the format.



I agree, although there are other benefits to a higher word length (mostly signal processing/conversion related as above).

Furthermore, to press your point, nobody has the full 120+ dB range of hearing available at one instant in time. Dynamic range in hearing just as in vision is compressed to a usable range for a given signal detection situation. It is smaller than the whole maximum range and adapts based on the signals being sensed. For example the acoustical reflexes of middle and inner ear ensure that nobody will hear much anything from a 20 dB signal for at least a few seconds after a 100 excitation. So, the dynamic range at any time is less than 120 dB and as such will further prevent us from hearing all the limits of 16 bit audio, even though we can detect signals well below the (static random) noise floor of acoustical spaces.

regards,
Halcyon

Yes, but I'm talking about resolution of media, not converters. Good 24-bit converters achieve near-theoric 16-bit performance when playing 16-bit audio. Also, I wrote 50 KHz sampling frequency to leave some room for real-world imperfections of converters in regards to their hf roll-off.



But this higher word length is only needed at the editing stage and inside the converter, not in the media.

I't interesting what you noted about instantaneous dynamic range of human ear, this reduces further the practical requirements of digital reproduction systems.

Wow ! It makes HDCD's dynamic gain a smart way to handle dynamics : to code a gain setting along with the PCM data.

"What a worthy and useful thread. "

"I don't even know where to start "

"Without getting all cranky, I would suggest you go out and do some reading before starting any more "profound" threads like this one."

"Obviously (well, at least to me) we are not concerned with that on this page. We are concerned with "psychoacoustic audio compression". Please for godsakes read up on some audio basics. "

"Is there any actual point to this, Mr. ? Or are you trying to out-l33t somebody? This is like asking what is the sound of one hand clapping (ABX'ed, please) or how many angels fit on a pin's head."

"I mean, anybody who was made some research knows that NO format, maybe ever, will be Lossless in relation to the originally produced sound (the artist performance, that is)."

"Either deal with it or don't. Just don't try to play smart here."

"(I am really making an effort here...counting: 1,2,3,4,5...) "

"Anyway, I think this thread is more lossy than CD audio will ever be."

"I have a headache...I'm gonna go find some aspirin now..."

*****************************************************************

Well, what a delightful bunch you all are.

At some stage in our lives, most of us realise that other people don't always act in *exactly* the way that we would like. They may say things that annoy, upset or frustrate us, but sooner or later we come to realise that 99% of the time, this was out of simple ignorance or failure to understand rather than a deliberate attempt to cause offence. We learn this quality called "tolerance," which allows society to function in an imperfect world full of fallible, imperfect individuals who do not have absolute knowledge of everything under the sun.

This process is often called "growing up." A pity so many people on this board never seem to manage it.

Thanks and goodbye.

My sentiments exactly...

This thread is not the easiest one to follow, but after reading more of the FAQs and other links posted previously in this thread, it still seems to me that the point in the recording process at which the audio has its last "lossless moment" is just before the sound hits the microphone, and just before the first electron enters a recording device from any electric instrument's signal processor. This is being discussed to death, I know, but it still seems to me that an analog recording would have the only chance to theoretically be lossless, but likely could never be in a real-world situation. I would still contend that digital, by it's nature, cannot be lossless. If digital could record sound without sampling the sound, then it could have at least a chance of being lossless.



I am without a doubt a newbie here, and I defer to your knowledge and references on this matter in all ways except on one issue. I would think that no matter what bounds you put on what's lost in a digital system, the fact remains that something is lost. Even if no human ear could ever hear it. My only point is that once that happens, the term "lossless" would never again apply. If any part of a sound is lost, but still "lossless as far as human hearing goes", it's still *not* lossless. My only point that differs here is that "lossless" is a sacred word with an absolute definition in the realm of recorded audio. One that effectively prevents it from ever correctly being used to describe recorded audio.

The concept of a recording being lossless as far as human hearing goes would imply that LAME --alt-preset insane is lossless to my ears, since I've never been able to ABX any sample I've tested with it. So if I truly cannot hear any difference between it and the original PCM WAV file, then is that MP3 format lossless to me? I argue "no", but it is, obviously, transparent. Those are the concepts that I think are getting overlapped too much in this thread. Lossless, by nature, guarantees transparency (and is part of what makes it a sacred term). Transparency can come from a recording format that is quite lossy, however, for the vast majority of people, such as from CD audio. Transparency is much easier to achieve than losslessness (is that a word?).

In the example you give of bounds in a digital system, if all frequencies up to 1MHz per channel are captured, I completely agree that it is overkill for satisfying human hearing abilities. But if there is one, tiny, menial, meaningless, but still existant frequency flaoting in the air in front of a microphone at 1,000,142 Hz, and if that frequecy was not captured on a CD, then the CD is lossy. Even if the thing lost can never, ever be heard by a human. It's still perfectly transparent, but it lost the title of "lossless" once any type of bounds were introduced.

I have no argument against the fact that "CDs are good enough". I think CDs are outstanding media, and I've never had a complaint about sound quality from one (except for crappy mastering, but that's a different issue entirely). But they still shouldn't be described with that giant, absolute, holy, sacred word: Lossless. That's all I'm saying. WAV to FLAC is lossless. Live to CD is not. (I'm not presuming to "teach" anyone here, as everyone in this thread likely has more technical knowledge than I do in these matters, I'm only using those as examples to communicate my point.)

And I honestly spent a good bit of time reading through the links you posted, and through other related information to try to educate myself as much as possible on the matter of CD audio, recording techniques, digital vs. analog, soundfield reproduction and everything else I could think to look up. But my argument is more philosophical than technical however, as I stand guard with vigilance in front of the word Lossless.

A couple thoughts:

You can look at a CD-medium for distributing recorded music in two ways:

1) To reproduce, as faithfully as possible, the music that was recorded

or

2) To reproduce, as faithfully as possible, the music that the artist/group wants you to hear

Subtle differences I know, but in my opinion, good CD's are not just a recording of a band playing (exception: live shows). A band should use the studio and recording process to create something that is different (not worse or better, different) from what is heard at a live show....so in this sense the "goal" of a CD (distribution of music) does not have to be entirely about replication of what was heard in the studio that day.

And another point, think of a song you hear, and think of how it was recorded....it was tracked, dubbed, cleaned up, compressed, etc, etc...can the two even be compared (in terms of loss and losslessness)?

If you're joking then ignore this. If you're not...

To be honest, I don't see a major problem with this thread. It started as a likely argumentative (and possibly worn-out) topic, and a little bickering broke out, but so what. Maybe I'm subconsciously blanking out all the nastiness, but I think the majority of this thread is a valid, civilized discussion. Sure, my first post was one of sardonic humor, bound in abstract cynicism. It wasn't the first time I've jumped into a thread with a semi-humorous (but respectful) protest, only to return later to join the discussion. And it certainly won't be the last time I do it. B)

And though I may differ on certain points with people here who I clearly acknowledge as smarter than I am about recorded audio, that's how people learn: communicate thoughts, listen to responses, learn facts, and change beliefs when appropriate. And then we share what we learn with others, and the cycle continues...

Anyway, a little bickering is just going to happen in an otherwise serious discussion. It's nothing to get in a fit over, as long as it doesn't become a flame war (which this thread is far from becoming IMO). If it were, a moderator would be in here handing out warnings, I'm sure. I think your advice of "tolerance" is valid, but it's already being used here.

If something in a thread breaks forum rules, report it. If you just don't like a thread, then don't read it. There! Problem solved!

"Lossy Audio Formats
The term "lossy audio formats" is used on audio formats [mp3, wma, ogg, etc.] that compress audio files by removing/filtering audio data. Once an audio file is converted to a lossy audio format, the original quality is lost forever (Reurich, 2002 d).

These are formats that compress audio files to an extremely small size and still sound good to most people. The bitrate is determined by how much data the file format is using for each second of audio-stream, and will also determine how good the quality will be. The higher bitrate, the better sound. If you want a near-CD quality, a bitrate of 128 Kbit/s should be suitable, but is also depending on the compression method used."

The above definition can be found here.

The key word here is compress. Now, with that in mind, is your (anyone's) argument still valid? And, in answer to the original question, theoretically, as the arguments suggest, yes, but in following the definition of lossy audio, no.

Later.

That's a good reference. I wouldn't presume doubting someone's definition of "lossy audio formats" without a valid argument, and I may sway in calling CD audio "lossy".

But I still can't call it "lossless" since it can't possibly capture all of the source audio data, the same way LPAC (for example) captures every bit of the data in a PCM WAV file (it's source audio data). The crux lies in the fact that a CD's source audio data is floating in the air (so to speak), and has severe inherent difficulties to capture all of it "perfectly"...it's technically not an infinite amount of data, but it's more than has ever been captured in a digital format. This same task is much easier for a lossless codec such as FLAC, LPAC or Monkey's since their source data is simply another file...reasonably finite and quantifiable.

From the same site, here's the definition of "lossless"...

Lossless audio algorithms return audio data that is identical to the original audio quality (Audio Box, Inc, 2000). This makes the file size generally larger than with lossy audio compression, and less useable for the Internet. With lossless audio formats you will see a reasonable file size decrease (usually around 30%) but not quality loss at all (Reurich, 2002 b ).

The core of my argument is that CD audio cannot return audio data that is identical to the original audio quality. I consider "identical" to be a word just as serious as "lossless", and would be very strict in it's definition. "Identical" means exactly identical, not just "sounds exactly the same to a human".

But thank you for linking a formal definition of these terms.

Hmmm...another consideration...

If we use only the Reurich definitions for a moment..."lossy" refers to only compressed formats, and lossless does as well. But uncompressed audio has to also be definable as either lossless or lossy, right? Or are these terms taken, and different terms are needed, perhaps?

[brief sarcasm=on]

If we need a new term to describe the integrity of CD audio, I vote for "Schmooo". CD audio is schmooo. All in favor? All opposed? See how easily we could resolve this debate? Just as long as no one talks about schmooo. (Same rules as for Fight Club.) I don't want to hear anyone asking "Is schmooo lossless?"

[/brief sarcasm]

Edit: Fixed formatting mistakes

I have a similar question I am just dying to ask: is 256kbps MP3 CD quality?

(for those of you who have never been on the - allas deceased - r3mix forums: please don't answer this question, makes people go )

Although I do wonder wether the current 16-bit CD format doesn't limit a bit the dynamic range of the music, compared to the good old LP?

This is a very subjective impression, let me explain: my dad has an old original and well-kept LP lying around from CCR, Willy And The Poor Boys. I happened to pick it up on CD a couple of days ago and compared the two on the same amp and speakers at my dad's (I don't own a turntable). It's probably just me or the remastering quality but the LP sounds "warmer, fuller" then the CD, as if the CCR boys can really play all out, pedal to the metal.

It's just an impression I guess. And BTW: yes I can ABX it because the LP has little cracks and pops from time to time .

No, an LP has a dynamic range of about 13 bits.



Actually, lossless is a word defined in the digital domain. Any word or physical theory is valid under a given field of application, under a given context.
For example, all the arguments used so far here to explain that music recording can't be lossless assume that sound is a continuous real function of time and space that for each position of space and time (x, y, z, t) defines a real number that describes the pressure of the air. But is this a valid assumption ? Yes, under most, if not all practical circumstances. But it is still a simplified model. A more accurate description would be that sound is a flux of phonons subject quantum mechanics laws.

So asking if something is lossy or nor doesn't make sense without specifying lossy under what circumstances.

Lossless for human ear ? That's rather called "perceptually transparent".
Lossless for a digital file ? That means that it allows to reconstruct the original file bit for bit.
Lossless for live recording ? As long as bit depht and sample rate are above the mics, amplis and speaker noise and frequency limits, a digital recording can be.
Lossless for musical performance ? That requires three dimentional continuous analysis (x, y, z, t), but can be achieved in theory as long as the resolution is higher than the air's acoustic properties.
Lossless for physical air pressure ? As long as we record enough quantum informations about phonons so as to have more info that is necessary to define the concept of "pressure", that has only a macroscoping meaning.
Lossless for the phonons's flux ? That might be achieved under forthcoming supesymetric theories. This may or may not cover the time-space concept at Plank's limits, and virtual particles defining vacuum's energy.

Whether this can be continued endlessly is a question that divides specialists (physicists and philosophers). Buddhists philosophers have strong arguments showing that this is endless, and thus that lossless is impossible. That's also the belief of many phisicists, but not all. Stephen Hawkins for exemple thinks that a final physical theory exists and can be found. Under this theory, lossless may or may not be possible. Buddhists philosophers ask "if this theory exists, Why does it exists and not another one ?". Answer this question, and the theory is not complete since it has a cause outside itself, in another theory. Don't answer it, and the theory has no proof or background.

Under that strict definition of "lossless", no real-world recording system can be lossless, be it digital or analog, now or ever, since it's impossible both to capture and store the variations of pressure with infinite detail.

@rjamorim

Yep, i agree to you

Lego, F1Sushi, What do you want to tell with this ?

That you are not interested in this thread ? Then why don't you go and post the exact same thing in EVERY thread that didn't provide some news ?

In one word, your posts were certainly the most useless of the whole thread.

However, I'll take you point of view into account and move this into the off-topic section.
This way, there will be no point anymore in saying that "this thread is unuseful".

Thanks for moving this thread elsewhere...and my apologies - my sarcasm occasionally gets the best of me.

And if sound is a continuous real function of time and space, then that would imply that it would have an infinite number of real numbers to describe all positions, right? That would seem to lend toward the theory that recording could never be lossless. And in your second description, even if there were not an infinite amount of data, it would still seem to be too much to reasonably capture in a lossless way. Something along the lines of one minute of audio becoming a terabyte of recorded data, for example (or some other extreme amount).


I agree that it's important to define the scope of the term "lossless". I have taken the position that it is an absolute term rather than a subjective one. I took the liberty of indexing your examples above. -a. is subjective and hardly (or not) quantifiable, so lossless wouldn't apply. -b. is quantifiable and absolute, so I agree, lossless can easily be achieved. -c. is the tough one...is it an infinite function? If so, then lossless capture would be impossible there as well, regardless of how high resolution would because you're dealing with a continuous wave (or maybe the concept of capturing with continuity is just beyond my ability to understand/believe?) -d. would result in a finite quantity, unless I'm mistaken, so I agree that lossless would be possible by that definition. -e. sounds likely not possible to be lossless at this time, but I agree that extending and proving current scientific theories might make it possible.

To oversimplify things for a minute, this entire debate makes me think of what it would have been like in, say, 1940, when the speed of sound was the golden ring for aviators. So many people said, and even tried to prove theoretically, that traveling at or beyond the speed of sound was a scientific impossibility. Then 1947 came.


Interesting that this can be analyzed equally with religion, philosophy and science.

"How did the universe begin?"

"Will we ever be able to go faster than the speed of light?"

"Will lossless recording of live audio ever be possible?"



I guess my final thought is that I believe lossless recording can be possible, but it will take more science than we've currently mastered. And, for that matter, probably more philosophy and religion as well. Maybe in five years. Or ten. Or a hundred. Or maybe in three months...who knows?


Edit: Fixed formatting mistakes.