[Joe Bloggs]

FWIW I agree it is obvious that downsampling to a lower bit depth / sample rate is a lossy process.

 

[watchnerd]

  what's the advantage of MQA over lossless compression of hi-res source?
 

 
Meaning like MQA vs FLAC?

 

[jagwap]

We'll try again:

Is 16bit/44.1khz FLAC lossless?

You missed a few words:

"Is 16bit/44.1khz FLAC a lossless version of a 16bit/44.1khz PCM WAV file?" Yes.


"Is 16bit/44.1khz FLAC a lossless version of a 24bit/96khz PCM WAV file sample rate convered down?" No.

 

[eio]

   
Meaning like MQA vs FLAC?

yes, with the same source sample format.

 

[jagwap]

yes, with the same source sample format.

You may find opinions vary here, and some will be given as fact despite them being opinion.

 

[watchnerd]

You missed a few words:

"Is 16bit/44.1khz FLAC a lossless version of a 16bit/44.1khz PCM WAV file?" Yes.


"Is 16bit/44.1khz FLAC a lossless version of a 24bit/96khz PCM WAV file sample rate convered down?" No.

 

 
Quote:

FWIW I agree it is obvious that downsampling to a lower bit depth / sample rate is a lossy process.

 
Lossless has a very specific meaning in terms of PCM audio: the ability to perfectly preserve 16bit/44.1khz PCM (Redbook) audio.
 
From this point of view, SRC is not a lossy process as the data that is discarded does not result in fidelity below Redbook.
 
This is why different jargon, such as downsampling or decimation, is used to refer to the conversion process...because it isn't lossy when it comes to outputting Redbook-level audio.

 

[jagwap]

Lossless has a very specific meaning in terms of PCM audio: the ability to perfectly preserve 16bit/44.1khz PCM (Redbook) audio.

From this point of view, SRC is not a lossy process as the data that is discarded does not result in fidelity below Redbook.

This is why different jargon, such as downsampling or decimation, is used to refer to the conversion process...because it isn't lossy when it comes to outputting Redbook-level audio.

Your example gave 96kHz 24 bit to 44.1kHz 16 bit. This is ASRC (Asynchronous Sample Rate Conversion) which is imperfect. The amount of precision affects the dynamic range. The best hardware ones I have used have 140dB noise floor (TDM4192) which is considerably higher than the 24dB theory. Not bad, but not a bit perfect reproduction.

Lossless doesn't just apply to Redbook.

Lossless means an encoding which can be restored back to a bit perfect copy. You cannot decimate 96kHz 24 bit to 44.1kHz 16 bit and back again and get the identical file.

 

[jagwap]

Lossless has a very specific meaning in terms of PCM audio: the ability to perfectly preserve 16bit/44.1khz PCM (Redbook) audio.

From this point of view, SRC is not a lossy process as the data that is discarded does not result in fidelity below Redbook.

Wait a minute...

If this is true {which it isn't}, then as MQA fully decoded achieves better than 96.33dB (16 bit) dynamic range, and 22.1kHz (44.1kHz sample rate) {which it does}, then by this definition MQA is lossless.

 

[watchnerd]

  yes, with the same source sample format.

 
Firstly, MQA attempts to deliver high-resolution audio with a smaller data size than it has in FLAC, without losing any fidelity.  In theory this should be useful for streaming situations where bandwidth may be constrained.
 
In practice, the importance of consuming less bandwidth for home audio playback is pretty dubious in an era where streaming of HD video over broadband is common, which is much 'fatter'.
 
But does MQA succeed in making smaller files? Some question this, saying you can SRC FLAC to 18bits, equivalent resolution to MQA, and actually get smaller files.
 
Secondly, and more controversially, MQA claims to deliver better time domain performance by addressing "time smear", which is a reference to ringing associated with filters used in the DA conversion process.  MQA claims this delivers better quality sound, although even they seem to be cautious about exactly how audible this is.
 
How important is this?
 
1. Pre- and post-ringing occurs in the vicinity of the cut-off filter, which for Redbook audio is 22khz.
 
2. Listening tests where subjects listen to different digital filters with different degrees of pre- and post-ringing have not shown a strong preference or even audibility
 
Which leads to:
 
Is MQA higher quality than a 24bit/96khz standard FLAC file?  No, nor does it claim to be.
 
Does MQA playback pass a DBT ABX test where it is preferred vs standard Redbook with statistical confidence?  I haven't heard of any DBT tests of MQA at all.

 

[eio]

You may find opinions vary here, and some will be given as fact despite them being opinion.

reading the white paper and this thread...interesting 

 

[watchnerd]

Wait a minute...

If this is true {which it isn't}

 
Yes, it is true.  
 
Proper SRC of 24bit/96khz to 16bit/44.1khz preserves Redbook levels of fidelity.  Thus, from a Redbook definition, it is lossless.
 
Everything that is discarded is beyond Redbook specs.
 
Explain how that isn't true?

 

[watchnerd]

  reading the white paper and this thread...interesting 

 
Here is Benchmark's critique of MQA:
 
https://benchmarkmedia.com/blogs/application_notes/163302855-is-mqa-doa

 

[jagwap]

Yes, it is true.  

Proper SRC of 24bit/96khz to 16bit/44.1khz preserves Redbook levels of fidelity.  Thus, from a Redbook definition, it is lossless.

Everything that is discarded is beyond Redbook specs.

Explain how that isn't true?

This isn't the definition of lossless.

 

[jagwap]

Firstly, MQA attempts to deliver high-resolution audio with a smaller data size than it has in FLAC, without losing any fidelity.  In theory this should be useful for streaming situations where bandwidth may be constrained.

In practice, the importance of consuming less bandwidth for home audio playback is pretty dubious in an era where streaming of HD video over broadband is common, which is much 'fatter'.

But does MQA succeed in making smaller files? Some question this, saying you can SRC FLAC to 18bits, equivalent resolution to MQA, and actually get smaller files.

Secondly, and more controversially, MQA claims to deliver better time domain performance by addressing "time smear", which is a reference to ringing associated with filters used in the DA conversion process.  MQA claims this delivers better quality sound, although even they seem to be cautious about exactly how audible this is.

How important is this?

1. Pre- and post-ringing occurs in the vicinity of the cut-off filter, which for Redbook audio is 22khz.

2. Listening tests where subjects listen to different digital filters with different degrees of pre- and post-ringing have not shown a strong preference or even audibility

Which leads to:

Is MQA higher quality than a 24bit/96khz standard FLAC file?  No, nor does it claim to be.

Does MQA playback pass a DBT ABX test where it is preferred vs standard Redbook with statistical confidence?  I haven't heard of any DBT tests of MQA at all.

You missed the time smear correction of the ADC.

 

[watchnerd]

This isn't the definition of lossless.

 
Which part of "from a Redbook definition" did you not understand?