A Question About Compressing Files

Thank you. The above is where you contradict what you said not moments earlier when you tell me "That's not right" in response to my saying that a CD is lossy

The CD *is* lossy, just as I said, just as you say above.

 

yep, the only truly lossless way to experience sound is to have the sound waves produced by the actual instruments hit your ears. Shoot, even then, your ears themselves result in signal loss, unless you happen to have perfect hearing and clean the wax out of your ear canals before every listening session . . . .

if by 'original signal' you mean the 'original recording' rather than 'original performance', then sure.
But the entire equipment chain used to record the signal will introduce changes and discard some data from the waveforms actually produced in the studio by the instruments being played.
The original recording != the original performance, in other words. No matter what tech is used to store/process/manipulate the waveforms from there on out, it's always going to be a bit different than it was being in the room while the piece was being played . . .

 

The specific part of the process that you claimed to cause loss in fidelity, the sampling, is not lossy.

 

It absolutely is. It's right there in the word "sampling". A "sampling" is a "sample" of the continuous waveform. It's not the entire thing. It's a sample (subset) of the information available.

 

a character set with that many characters would be highly impractical. It woudl take too many bits to represent that characters themselves, and it wouldn't be any smaller.

 

i hope this thread doesn't get into sampling theory...

 

If that is your definition of "lossy", then SACDs, DVD-Audios, LPs and even original master tapes are "lossy" as well. Every sound recording is "lossy".

 

Yes, which was exactly my point, so I'm glad you brought it up because I meant to!

The CD Standard was merely an arbitrary standard put in place for CDs. Clearly, SACDs are of a higher standard than CDs (which tells you right there that the CD Standard is not "lossless").

If the SACD standard had been the standard originally put in place for digital audio, I guess us audiophiles would be sticklers for maintaining that standard, instead of sticklers for maintaining the (inferior) CD Standard as the "lossless" benchmark for sound.

I'm trying to help myself out of a box that says the CD Standard is "IT", because it isn't. Is it SACD? Is it 320kbs MP3? Is it something else on the continuum that is the end all be all?

 

No, it’s a representation of the waveform that is more compact than the original, and from which the original can be reconstructed. It’s similar to the way that a graph of a function can be represented by an equation, and reproduced exactly from it. The parts of the signal you imagine have been discarded have not been - they can be recreated exactly just from the samples.

 

You're thinking of a lossless codec such as FLAC?

 

Respectfully, it is you who is wrong. You are confusing different connotations of the term "lossy."

 

It isn’t a subset of the information. A point on a curve is not a “subset” of a curve.

 

Actually, what I was doing is pointing out that there *are* different connotations of the term "lossy".

 

Sure, but you specifically mentioned that digital sampling is inherently lossy, but that is an incorrect statement, provided you sample at least twice as frequently as the signal’s highest frequency component. These theories and devices do not emerge from imprecise terminology and enthusiastic handwaving.

 

A curve *is* a set of points. One point is a subset of a *set* (a number) of points.

 

Thanks for your reply. I understand what you are saying. But digital sampling still is not lossy. As has been noted by others above, digital sampling can reconstruct the original waveform, not just a subset of the original waveform. Digital sampling produces quantization error, but that manifests as noise. And if it's below the noise floor of the original analogue recording, then nothing is lost. Even when it it not below the noise floor of the analogue original - such as with DSD encoding, which is only 1-bit - noise shaping can (within limits) nevertheless enable the original waveform to be reconstructed.

Again, sorry, but this is misleading. You're treating an analogue waveform as if it's already a set of digitized/quantized points, and then on that basis claiming that digital sampling is lossy because it captures only some of those points.

 

You are right, of course. What I should have explained is that the samples are used to reconstruct the points in-between, so the fact that there are fewer sample points than unsampled points is immaterial.

 

Turns out I wasn’t 100% correct either.

 

Look, everything is ultimately lossy at some point, even analog. The grain of film; the pigment particles of a print; the ground-up bits of iron oxide on a magnetic tape; the molecules of vinyl in an LP.

 

molecules are lossless

 

Even my cheap-ol' smartphone can easily handle FLAC encoded at 8.

 

Perhaps Quantum computers and qubits hold the answer.