MQA Deep Dive - I published music on tidal to test MQA

[DimitryZ]

Not really sure what a CD has to do with anything but it does show how digital technology and terminology can be confusing. Since MQA is about hi-res it should be tested to determine how well it can playback a encoded hi-res file as compared to the same file compressed in FLAC...both objectively and subjectively.

I think it's clear that MQA is competitive with FLAC at moderate hires (say up to 96KHz) and dramatically outperforms it at higher frequencies such as audiophiles demand - DXD.

 

[Raindog123]

There is no question that it is mathematically lossy.

But so is a CD release from a modern DXD master.

The relevant question is "compared to what?

Don’t forget the “in lower Shannon triangle”, whatever this is

 

[DimitryZ]

Don’t forget the “in lower Shannon triangle”, whatever this is

Again, this is a technical forum and we are all encoraged/expected to learn new things.

Give it a try?

 

[Grooved]

Can someone explain this, please ?

part of a 24/96 FLAC file from Qobuz - directly loaded in the software (to confirm with the one below that the digital recording does not change the signal)

Same 24/96 FLAC file from Qobuz play in Roon and digitally recorded from Roon output, then loaded in the software

Same track but 24/48 MQA played and decoded with Core Decoder of Roon to 24/96, digitally recorded from Roon output, then loaded in the software

 

[scott wurcer]

"Hollow post?" It's one of very few with actual numbers - audioSCIENCEreview?

Your comparison of BERT (bit error rate) to waveform null in dB makes no sense at all.

 

[dc655321]

I did an order of magnitude analysis of this upthread and came up with these values for distortion for various codecs:

Repeatedly pointing to your "analysis" does not make it any more correct.

 

[amirm]

Can someone explain this, please ?

What is the track name?

 

[Raindog123]

Again, this is a technical forum and we are all encoraged/expected to learn new things.

Give it a try?

Absolutely! Have a reference/link?

 

[DimitryZ]

Your comparison of BERT (bit error rate) to waveform null in dB makes no sense at all.

Because it's not bit error rate. It's an estmation of amplitude deviation in the analog wave form. Where dB amplitude equation would apply.

Would you offer a competing analysis?

 

[DimitryZ]

Absolutely! Have a reference/link?

Just go over Amir's posts.

 

[Grooved]

What is the track name?

The beginning of the 1st track from the last album you checked and bought from another site - Terry Riley - Sun Rings, that appeared to be a 44.1 one that have been upsample on all "HiRes" version
I will checked tomorrow on other ones that don't have been upsampled

 

[DimitryZ]

Repeatedly pointing to your "analysis" does not make it any more correct.

So far, no one offered a competing assessment.

The important number is for MQA. By the data published by Archimago (as I was told), MQA nulls against "etalon" LPCM extremely well. I looked at the peaks at low frequency, which is the most conservative way. Correctly averaged and weighted, the number will be only better.

So only more unresolvable by our systems - aka lossless.

 

[scott wurcer]

Because it's not bit error rate. It's an estmation of amplitude deviation in the analog wave form. Where dB amplitude equation would apply.

Would you offer a competing analysis?

This is, (BTW a real world computer is far better than this). One bit in 1e9 dropped does not correspond to -180dB amplitude deviation, the two concepts have no relationship to each other. At the simplest level which bit an LSB or an MSB there are orders of magnitude difference?

Mathematically Lossless Codec:

Mathematically perfect algorithm running on a real-world computer and network, will still have an error once in a great while. Let's pick a really tiny value - 1 in a billion or 1E-9. This results in the error against the original file of -180dB. Great performance!

 

[DimitryZ]

This is, (BTW a real world computer is far better than this). One bit in 1e9 dropped does not correspond to -180dB amplitude deviation, the two concepts have no relationship to each other.

So your point is that mathematically lossless replay over typical home network should do better and should null against the original to better than -180dB. Excellent point. And using error as a placeholder for amplitude is incorrect. In the context of "order of magnitude analysis" I was searching for "a really small number"

However, once MQA reaches -120dB+, it is beyond our systems' resolution abilities and is, therefore, physically indistinguishable from mathematically lossless playback.

Disagree?

 

[JohnYang1997]

So your point is that mathematically lossless replay over typical home network should do better and should null against the original to better than -180dB. Excellent point.

However, once MQA reaches -120dB+, it is beyond our systems' resolution abilities and is, therefore, physically indistinguishable from mathematically lossless playback.

Disagree?

That depends on where you go from here.
If you are to say hence MQA is no difference than mathematical lossless then no.
If you are to say the value in real world then yes.
However this is based on assumption that MQA does null better than 120dB. It's not yet proven.

 

[pjug]

So far, no one offered a competing assessment.

The important number is for MQA. By the data published by Archimago (as I was told), MQA nulls against "etalon" LPCM extremely well. I looked at the peaks at low frequency, which is the most conservative way. Correctly averaged and weighted, the number will be only better.

So only more unresolvable by our systems - aka lossless.

You are trying to quantify how well the MQA tracks PCM? I don't think you can get anything meaningful from your eyeball analysis. You could run DeltaWave comparing decoded MQA vs PCM files.

 

[JohnYang1997]

You are trying to quantify how well the MQA tracks PCM? I don't think you can get anything meaningful from your eyeball analysis. You could run DeltaWave comparing decoded MQA vs PCM files.

How do you get decoded MQA without hardware btw?

 

[pjug]

How do you get decoded MQA without hardware btw?

You can use virtual cable to make a recording if you have a player that does the MQA decoding.

 

[mtristand]

Pretty much nonsense. "Zero loss of content" is only important for computer code - so if you are a computer, you should be concerned.

All systems operate in the real world. A mathematically lossless playback originating in your computer will accumulate errors once it reaches your playback device. I did an order of magnitude analysis of this upthread and came up with these values for distortion for various codecs:

Mathematically lossless: -180dB
MQA: -120dB
Great lossy codec: -60dB

Errors in the first two are beyond our equipment's SNAD. They can't physically be reproduced, much less heard. Therefore, they are both lossless to the listener - which has always been MQA's DESIGN INTENT.

"Lossless to the listener" is still misleading. You're still losing data, so it's lossy, not lossless.

One major quality-assessment metric of a lossy compression algorithm is to what extent the result is indistinguishable from the original by humans. Just because something happens to be hard to differentiate or indistinguishable in certain contexts doesn't mean you can now call it lossless, because you're losing data... so it's lossy.

And either way the point is practically moot because MQA is applying a "deblurring" transformation to statistically-assessed music anyway, so you're not technically getting the original regardless, but rather some new version.

But I'll cut right to the chase here - what would your defense of this claim be (I extend the same question to Amir):

Using a unique ‘origami’ folding technique, the information is packaged efficiently to retain all the detail from the studio recording. While MQA retains 100% of the original recording, an MP3 file keeps just 10% of the data.

What do you believe this claim is saying?

Do you believe the layperson would interpret it the same way? How do you honestly believe they would interpret it?

 

[dmac6419]

You can use virtual cable to make a recording if you have a player that does the MQA decoding.

You can?