Originally posted by m0e
It’s my understanding that all digital recording has some loss in sound quality. That is just the nature of it; after all, they do call it a “sample rate”. Just as analog recording is limited in accuracy by the mechanical limitations of the recording equipment, digital recording does not capture every sound wave sent to the encoder.
True regarding the structure of digital recording. But there have been nauseatingly
long debates with scientists and codec developers from all over the world about the definition of "lossless". I was in a debate recently as a matter of fact on this very topic with several such people. I'll post the link as soon as I can access it.
When I say "lossless is bit-perfect from the source audio", the source audio I'm referring to is the MP3 file (within the topic of this thread, anyway). A FLAC (or WAV) will have perfectly identical sound quality and bit preservation as the source it comes from. You're instead talking about the source of the originally recorded sound, many levels before the MP3 was encoded.
Air vibration (or electric impulse) --> ADC (unless the source is digital) --> recording device --> WAV format (for PC) --> encoding into target compressed format. There is a natural loss with ADC (with sampling, as you point out), but no other until you encode to a lossy format. The ADC loss though, is FAR
from audible for anything CD-level (44.1kHz) or higher, so it's audibly
Originally posted by m0e
Depending on the encoder used it processes the sound waves and encodes the information it “thinks” is important. Then the decoder takes that info and “guesses” as to what sound it needs to fill in the blanks on playback. Basically what I am saying is all digital music is compressed. So going from sounds recorded using one audio encoder and then decoding and recoding to another audio format should result in a loss of sound quality.
Now I could be wrong, it’s not like I have any formal training in this, just an interest in how all things work.
The first sentence in this quote only applies to psychoacoustic
encoding (lossy formats like MP3, Ogg Vorbis, MPC, AAC, and others). FLAC is not a psychoacoustic format, and does not drop any part
of the audio stream, whether you can hear it or not. It does encode (by defining frames, using predictive encoding methods, stereo mode encoding, etc.), but it loses not a single bit of the original source audio. (Again, "source audio" = CD, or WAV, or wherever you encoded the FLAC from...not the musicians playing live or in a studio, which is long before the encoding chain.)
Not all digital music is compressed. A digital recording originally captured at 44,100 samples per second, at a bit-depth of 16 bits, and with 2-channels = 1411kbps, and is indeed uncompressed (after ADC/sampling, anyway). You can compress losslessly to about 60% of the uncompressed file size (+ or -), but lower than that requires dumping bits with either simple quantization encoding (ref: WavPack hybrid, OptimFrog), or more commonly with psychoacoustic encoding. Audio recorded with one audio encoder, decoded, and then re-encoded to another format will NOT result in an inherent loss of sound quality IF the target format is lossless. The final encoding will perfectly
equal, bit-for-bit, the first file in this chain. Any perceived sound quality loss using a lossless target is absolutely placebo
, and in fact this is how placebo is baselined...how I can "test" to see if you're only thinking you hear a difference where none exists. (Believe me, I've gone through hundreds of hours of such baselining and then ABX testing and research on this.)
Originally posted by squall14716
Lossless formats compress the audio, not encode it. It's like putting a txt file in a .zip archive, the file size of the txt file decreases, but nothing is lost. Am I right here, or do I not know what I'm talking about?
Lossless formats do encode the audio in the course of compression (ref: my discussion above on this part). FLAC, Monkey's Audio, LA, WavPack Lossless, and all the others (AFAIK) use an encoding algorithm to compress the target file, but are still fully lossless. Test: Extract a track from CDA to Raw PCM WAV. Encode to FLAC (any setting). Decode back to Raw PCM WAV. Checksum the original WAV and the second WAV. The checksum will prove the files are fully indentical. You're right that lossless encoders also work a lot like *.zip or *.rar, but they do additional encoding to maximize compression of audio. Test: Take a PCM WAV file and Zip a copy. Then FLAC a copy. The FLAC will be much
smaller, because of its encoding algorithm. So it's more than just "compression".
--- I want to reference some discussions I've been involved in with the people who actually design and develop all these encoding formats (FLAC, MP3, Vorbis, MPC, ...) and utilities, but the site's server is down right now. I'll post them when it's back up. ---