all those stuff are only interpretations and representations of the signal. the digital data itself is a series of points(samples) and that's all. it can be shown as points, as a wave, as a spectrum, or identified at the end of a sound system as being the sound of a guitar... representations are only for convenience and the scale could show anything from db to volts to bit values depending where in the system we wish to try and interpret the data. you could show several of those things in audacity by changing some visualization settings. but it doesn't change what the digital data really is.
the PCM system (the lossless formats, excluding DSD) records the vertical value(amplitude) of the signal at a given moment. that's one sample, and the file is made of X samples per seconds. that's all. if the music is overall quieter, then the overall signal will record smaller amplitude values, that's the only difference at the digital level for the file itself. no consideration of quality difference or quantity of information. those stuff are decided by the bit depth(precision of the amplitude) and the sample rate (number of points per second that are recorded), not by the loudness for the file.
Quote:
-your intuition that because the flac file was of smaller size for quiet music, it meant that quiet music might have less information or precision, is shown to be wrong as it doesn't happen when you do the same test with .wav files. wav files result in the same file size whatever the loudness so that argument can't be applied to .wav.
The WAVs were the same file size whether quiet or loud, but the FLAC is able to "throw away" (hide?) wasted data to play back the exact same bits. Nonetheless the FLAC manages to shrink down more with a quieter file. How? It means it manages to find less worthless data to throw away/hide for compression.
-we also know that flac and wave result in the same final information as both as lossless formats. you can go from wave to flac and back, and lose nothing.
Yes, but a FLAC can shrink more of a quiet file, while the WAV isn't trying to shrink anything so keeps the max bits (1144) regardless. Then that means the quieter waveform has less data in it, less detail, less sound. So it's better to record loud. The loudness wars was a good thing. The CDs from the 80s don't have as much information in them. When FLACCed they produce smaller files.
When you have a photograph that is 8 megapixel it has more bits than one that has only 2 megapixels. And the 8 megapixel one has more detail. It can't be shrunk down as much with compression without losing that detail.
The details may not be visible. 4 megapixel or so is already approaching the limits of eyesight for most viewing distances and display sizes. Similar, 16 bit/ 44 kiliohertz was chosen because it seemed to capture enough.
so from those 2 points you can conclude that the smaller flac size is only a consequence of the compression algorithm of flac. not a loss of data or precision or whatever.
no i can't. You can't magically have smaller file sizes without actually losing data (not counting wasted / blank space cut out).
Think about images. You have a 4000x8000 blank canvas in an image editor (Photoshop). In the corner you paste in a photo that takes up only like 1/15th of that blank canvas, leaving he rest white. That white space is easily discarded by compressing. Is it like the extra headroom in a quiet audio file? When using FLAC or mp3? well, yes, but my point is that if your photo doesn't just take up 1/15th of the 4,000x8000 (in native size, without stretching or shrinking, of course, then that photo's "real data" when pasted onto that canvas is, say, 640x480. It's low resolution, by today's standards; it's VGA. You can blow it up, but that's increasing how large the pixels display on your screen (or how large they save, thus increasing file too), but it's essentially the same amount of "data". You didn't increase the detail, although perhaps your ability to take it all in depending on your eyesight. But if you import a photo that is 3000x7000 pixels, and that it's native resolution, you know that's going to have a lot more detail in it compared to the 640x480 photo. It's higher resolution. It resolved (captured) more of reality. So it still has some white space (1000 pixels either side multiplied together), but you can't compress it as much. With a quieter audio file you can compress more. Therefore a louder recording, because it can't be FLACed to as small of file size, it has more detail in it, more of the reality it was trying to capture.
The WAV proof doesn't work as explained above. Because a WAV saves the same amount regardless. That's why I brought up FLACs to begin with, so I could avoid discussion about the destructive compression (lossy) that MP3s do. But MP3s also are smaller when the waveform source was smaller.
I just tried both with Ogg Vorbis and FLAC from Audacity after using the amplify feature to maximize the sound (without clipping), exporting the file, then another where I start over with the same raw data (from the CD) and shrink it down a good bit putting in "-18" into the amplify feature... so the waveform becomes pretty small. But the FLAC and OGG have smaller file sizes for the same length and in Foobar it shows that the -18 version is 589 kbps. That's what it shows in the properties of the files in Windows: http://imgur.com/a/GuVCR
How can it store the same amount of data -- why wouldn't the louder one be the same size (or very close)? The discrepency suggests it is like shrinking an image down in a photo editor to a smaller canvas size -- I mean, not just temporarily zooming in the editor (although that would be talking the amount of data that's in RAM), but what you save to, when putting it on hard disk, and the smaller image is now 380x879 when before it was 1024x768 (or whatever). You know that the 380x879 will not only take up less space on the hard disk, but will have less picture data too. It had to remove some of it to get it that small, unless we're talking vector graphics. I think.
I mean I'm not talking about a GIF or PNG that was originally 380x879 then blown up to 1024x768.
I don't see how the same doesn't apply to audio data. It shouldn't take a lot more space to represent larger waves. For instance, if you have a digital drawing that is (in its original creation) done at 640x480 and you double its size to 1280x960 the program ought to be able to store that data as 640^2x480^2, as a shortcut, so it saves space. It should be taught to square the pixel upon opening. It would in a sense keep its original size if working on multiples rather than a new ratio.
I don't see why it wouldn't work that way with audio data. The program (Foobar, iTunes, whatever) is taught to playback a file at multiples of 2, 3, 4... expanding the audio "size" (or how loud it sounds). That way data could be saved, because for some reason it requires less data to save the same audio when it is smaller waveform.
now in relation to the global resolution of the signal that you'll get out of your headphone, all components and all devices generate some matter of noise. if you record music too quietly, you end up having the music closer to all those noises and the signal to noise ratio will be poor. so from that particular perspective, indeed a signal recorded closer to the maximum amplitude(louder) will be further away from the quiet noise floor, that result in a better SNR and in a way that means a superior resolution. so your initial idea about quiet music and resolution isn't wrong, it's only the relation between smaller file size in flac and signal quality that was wrong.
Is that why the file size is larger then? The noise makes for less data?
I hope this discussion isn't making everyone facepalm too much. I just seem to hear more music, more detail, when it's a fairly loud recording (large waveforms that are nearly clipping), versus moderate to conservative recordings, say from the 80s, even when it was very electronic sorts of music. And classical recordings seem to have gotten better, but they've all also been recording louder, a "casualty" of the so-called "loudness war." But maybe the loudness war is a good thing. They're recording more audible data that we call music. So we should increase the bit depth too, not just to record the canons in the 1812 Overture, and increase the sample rate because it's not just about bat and elephant hearing but you can capture more detail within a given audible range.
But that is a separate matter.
I'll post this even though I figure later today or tomorrow I'll regret it, thinking how stupid I am, not really understand this. I hate that. I always want to discuss things but then I get a bit shy because I'll probably get shot down.