I would add that there is almost certainly no data loss between DSD and DXD. Any sound difference might be due to how the D to A conversion is done when the file is played back. For that reason, many DACs convert PCM to DSD as part of the conversion process, including the one I use (everything goes to DSD256).
I don’t know how it’s defined there, but the discussion is interesting.
MP3 is considered lossy because there’s an intentional loss during conversion for reducing file size.
What if someone records in MP3, is it then lossless?
16/44 PCM is not considered lossy because the limitations (losses) compared to todays higher rates due to the lower bit depth and sampling rate were not known when it was released as best format at the time?
16/44 PCM would be called lossy when a higher sample rate is down converted to it?
DSD or hires PCM or tape is considered lossless when we record in that format. But all is lossy compared to our (then better) format in 10 years?
So how is a lossless format defined? It doesn’t seem to be its absolute lossless quality (as no format is lossless).
On this matter, Rushton is correct
I have compared, and made available for free file download, the DSD256 results from mixing without PCM processing versus mixing with PCM processing. See the article here. What one hears between these two samples is the impact of the DXD conversion when then output again to DSD. I can’t point you to a DXD-to-DSD file example from Eudora.
But, I can point you to DSD256-to-DXD comparisons from Hunnia Records and from HDTT. The HDTT are transfers from 15ips tapes, but you can hear both the Pure DSD256 and the DXD. On these HDTT files I’m listing, if you purchase the DSD256 file, you can then also download the DXD without any additional cost.
I have an article in process (that may or may not get completed) to compare downloadable samples at 24-bit DXD, 32-bit DXD, and Pure DSD256 of the same music selection. I’d put this on the back shelf, but perhaps there’s enough interest to move ahead with this project.
3 Likes
So we can only use the word lossless related to a conversion process, not to a format…that makes sense!
Then recording in MP3 would be lossless.
If we’d call recording a process, then the above, but also DSD recording would be lossy.
No. To the contrary, the definition he propounds is objective and precise.
6 Likes
Loss of data due to digital compression or format changes
SQ loss may or may not be discernible by hardware or listener
Please do
Yes, but compression then can also just be part of a format change or conversion, as native 16/44 from an absolute point of view is also kind of compressed related to a higher resolution.
We often read this or that is a lossy or lossless format. After what we clarified here, this makes no sense.
There can be lossless compression, eg PCM file conversion to FLAC; in contrast…MP3 creation is lossy, data lost
A visual analog….jpeg compression is lossy and can be readily observed; tif compression is lossless
2 Likes
If you follow @Rushton theory, an mp3 recording would be lossless as long as it remained in the original format.
My understanding is that 16-bit was chosen because the 96dB dynamic range was plenty to meet the limitations of human hearing, and at 44k sampling they could squeeze 75 minutes of music on a 5" disc, plus the 25% overhead for headers and parity data.
I the 1970s Denon were mastering PCM with 13-bit and 14-bit, and it was hailed as better than analogue.
In the 1980s they were doing 16-bit mastering with great success, and 24-bit just makes life a bit easier. 32-bit DXD is just luxury.
A little searching suggests that the general view is that “lossy” is a compressed format that is not what the artist intended as their product. So Spotify is currently lossy as it’s Ogg Vorbis. On the other end of things, anything 24-bit is considered “High Definition”. So 16/44 is in the middle and still considered “lossless” or “hi-fi quality” to the great unwashed.
@Rushton is talking about conversions in the mixing and mastering process. In reality, about 0.0000000001% of music listeners care about that. What matters is the format of the product released to the public. Often this is a 16/44 CD, perhaps a 24/96 download and probably vinyl as well. All would be considered lossless. If Spotify or a radio station distributes them in compressed format, they will be lossy.
There is of course massive snobbery with DSD, and snobs crave confirmation of their place at the top of the social/audio/wealth tree. Of course the people at the other end of things couldn’t care less what other people think about their habits. So the many folk who buy stuff off SoundCloud and the like couldn’t care less about formats and don’t care what other people think. They just like music. Include both my kids in that group.
Yes, that all makes sense.
The interesting thing is, Flac conversion, if being a conversion from the same uncompressed resolution is a lossless conversion and the Flac file a lossless format (only if created from the same resolution non compressed format). Anyway when we play back a compressed Flac file, the playback lost information compared to a higher resolution file or even compared to the same resolution uncompressed format (but in this case just because of the noise it generates within the streamer due to the internal uncompression process). It’s so hard to use the terms correctly, that it’s no wonder that everyone’s talking about different things.
On playback FLAC files are decompressed to the data present before compression
Not so with MP3 files, where compression causes permanent data loss
our audio equipment may insert their issues and colorations
1 Like
Yes. My best insight from today is, that only conversion processes can be named lossy or lossless and formats only if it’s known from where they originated. If that’s known the same format can either be lossy or lossless.
When we get aware of that, we see that most are using terms wrong or just accidently right.
Exactly.
Distributing a file as an MP3 is lossless if the original recording is an MP3 file.
2 Likes
See for example, the hifi voice of the proletariat:
“Now, in numerical terms, ‘lossless audio’ is typically accepted as being 16-bit/44.1kHz – which is the bitrate (16) and sample rate (44.1kHz) of CD music files. That’s right, ‘lossless’ quality as it is considered today is essentially what you have listened to for much of your life (and maybe downgraded from, as you adopted streaming!). That’s why you’ll often see ‘CD quality’ and ‘lossless’ presented as one and the same, as is the case on Amazon Music HD’s website (pictured below).”
CD quality audio (aka, “lossless”):
Same with Qobuz - mp3 = lossy, CD quality = lossless
And from the FLAC official site - it’s 16-bit:
A useful dumbing down fiction when explaining lossy MP3 or FLAC to the uninitiated.
44/16 can be lossy or lossless depending on the format of the original recoding.
As an example, it is incorrect to label 44/16 lossless when the source is 96/24, down sampled and stripped of bits.
6 Likes
You are incorrect.
The MP3 codec is an encoding algorithm to compress audio. The original MPEG-1 codec was designed specifically to carry out lossy compression of 16/44 CD audio data.
It is an international standard, which states:
The system layer supports the following basic functions: the synchronization of multiple compressed streams on playback, the interleaving of multiple compressed streams into a single stream, the initializiation of buffering for playback start up, continuous buffer management, and time identification.
If you record to an mpeg file, for example on a laptop, it is using a lossy compression process.
Only if the file is first recorded at /44/16 and coverted to MP3.
If recorded to MP3 native the MP3 is not lossless; there is no file with more data.
I am sensing you understand this, but enjoying arguing nomenclature.
4 Likes