MQA vs HiRez: an apples-to-apples comparison

[austinpop]

Hi Mani,

 

i listened to to these two, and have to agree D sounds better to my ears. The beginning of the track, the fret work is much more focused, the bass is tighter, brush strokes have more detail.

 

so in summary, C = D, blurred.

 

i guess that makes C MQA.

[church_mouse]

Mani, I tend to think of everyone on this site except me as American - I know that is wrong.  However, to come across someone just 15 mins drive away is strangely reassuring.

 

Happy listening.

[audiobomber]

I also preferred D, it sounded more natural, less processed, especially on sibilants. I assume C is MQA.

[manisandher]

16 minutes ago, church_mouse said:

Mani, I tend to think of everyone on this site except me as American - I know that is wrong.  However, to come across someone just 15 mins drive away is strangely reassuring.

 

Happy listening.

 

But Warwickshire's the centre of the world... isn't it?

 

Mani.

[manisandher]

28 minutes ago, austinpop said:

Hi Mani,

 

i listened to to these two, and have to agree D sounds better to my ears. The beginning of the track, the fret work is much more focused, the bass is tighter, brush strokes have more detail.

 

so in summary, C = D, blurred.

 

i guess that makes C MQA.

 

Hi Rajiv, it's interesting that so far everyone who's responded has preferred the sound of D. I can understand why, but over here there's something about C that I really like. Maybe it's the fact that I'm upsampling to 705.6 and then sending to a filterless 24 bit NOS DAC:

 

 

To my ears, the closed-form-M filter is the best sounding filter in HQPlayer. But the 'pleasant MQA effect' is even more pronounced when I use XXHighEnd, with its totally non-ringing filter.

 

Mani.

[manisandher]

12 minutes ago, audiobomber said:

I also preferred D, it sounded more natural, less processed, especially on sibilants. I assume C is MQA.

 

Thanks.

 

Mani.

[austinpop]

2 minutes ago, manisandher said:

 

Hi Rajiv, it's interesting that so far everyone who's responded has preferred the sound of D. I can understand why, but over here there's something about C that I really like. Maybe it's the fact that I'm upsampling to 705.6 and then sending to a filterless 24 bit NOS DAC:

 

 

To my ears, the closed-form-M filter is the best sounding filter in HQPlayer. But the 'pleasant MQA effect' is even more pronounced when I use XXHighEnd, with its totally non-ringing filter.

 

Mani.

 

Could you re-enable the link to the original hi-res file - at least for a short while? I've only listened to C and D captures.

[tmtomh]

I know a version of this was posted above, but FWIW, this C minus D - in other words, I time-aligned C and D, inverted D, and then mixed it with C. This is in Audacity (and when I instead invert C and then mix then, I get the identical result, just FYI).

 

 

[tmtomh]

So I just listened to C and D on my main setup, and I was surprised to hear a difference - honestly I didn't think I would.

 

The fretwork at the beginning, is exactly as @austinpop noted: a bit more depth and textured on D than C. I thought it was just auditory memory, as I listened to C first and sometimes I find that the 2nd version I listen to sounds more detailed because I'm more familiar and attuned to the music at that point. But then I went back and forth repeatedly and had the exact same listening impression.

 

Could I replicate this perceived difference in an ABX test? I won't say that for sure - but honestly, in this case based on what I've heard, I would tend to lean towards "Yes," with a good deal more confidence than I would have expected before listening to these samples.

[manisandher]

22 minutes ago, austinpop said:

 

Could you re-enable the link to the original hi-res file - at least for a short while? I've only listened to C and D captures.

 

Here you go (I'll take it down in a few hours' time):

 

https://drive.google.com/open?id=1MudMIZqZRZpMY_-QEo6b7Tk_9i23gNp1

 

Mani.

[manisandher]

11 minutes ago, tmtomh said:

So I just listened to C and D on my main setup, and I was surprised to hear a difference - honestly I didn't think I would.

 

The fretwork at the beginning, is exactly as @austinpop noted: a bit more depth and textured on D than C. I thought it was just auditory memory, as I listened to C first and sometimes I find that the 2nd version I listen to sounds more detailed because I'm more familiar and attuned to the music at that point. But then I went back and forth repeatedly and had the exact same listening impression.

 

Could I replicate this perceived difference in an ABX test? I won't say that for sure - but honestly, in this case based on what I've heard, I would tend to lean towards "Yes," with a good deal more confidence than I would have expected before listening to these samples.

 

Thanks.

 

Yeah, it's amazing when you think about it - your and Paul's null tests show that the differences sit at a pretty low level. And yet, people can hear them... quite clearly.

 

I still think there's more to MQA than meets the eye (looking at an FFT plot). I mean, I think there's a reason why those differences exist, in the specific way they do. And I still believe there's something interesting happening in the temporal domain with MQA.

 

Mani.

[austinpop]

9 minutes ago, manisandher said:

 

Here you go (I'll take it down in a few hours' time):

 

https://drive.google.com/open?id=1MudMIZqZRZpMY_-QEo6b7Tk_9i23gNp1

 

Mani.

 

Thank you. I have it. I also found the MQA file on Tidal. I will do some listening tonight when I next have some free time.

 

Note that my current chain is comprised of several "wondrous machines," not least of which is the Ayre QX-5 Twenty DAC, which has always sounded best fed native sample rates.

[tmtomh]

32 minutes ago, manisandher said:

 

Thanks.

 

Yeah, it's amazing when you think about it - your and Paul's null tests show that the differences sit at a pretty low level. And yet, people can hear them... quite clearly.

 

I still think there's more to MQA than meets the eye (looking at an FFT plot). I mean, I think there's a reason why those differences exist, in the specific way they do. And I still believe there's something interesting happening in the temporal domain with MQA.

 

Mani.

 

Yes, my (admittedly completely unscientific) sense is that something is happening temporally too and since these frequency comparisons are comparing the entire track (or in the case of Audacity I think about 2 mins worth), perhaps there are small temporal differences that are being masked in the null-test comparisons.

 

I also agree that the amplitude of the differences is very low. At the same time, I run null tests fairly often - to compare CD masterings/pressings, for example, and to see what effect downsampling and dither has on HD files vs a CD rip of the identical music and mastering. And while the differences often are very low-level like this, I am constantly surprised that, when I play the difference file and turn up the volume, it sounds like music - it's not spurious noise artifacts, static, and so on. 

 

Personally I think a lot of what we struggle with in these discussions is that fact that it's difficult for many of us to truly internalize and grasp at a tangible level just how the digital data we're looking at correlates at the minutest level with what we actually hear when we run that data through a DAC, analogue output stage, amplification, transducers (speakers/headphones), and finally soundwaves hitting our eardrums and getting process by our whole ear/brain hearing mechanism.

 

This is one reason why a firm grasp of digital sampling theory - and as @mansr often says, a firm grasp of the math involved - is so important. Otherwise we have no hope of being able to distinguish between an actual potential correlation on the one hand, and a metaphor on the other.

[PeterSt]

4 minutes ago, tmtomh said:

comparing the entire track (or in the case of Audacity I think about 2 mins worth), perhaps there are small temporal differences that are being masked in the null-test comparisons.

 

That.

I have said numerous times that this can't be expressed in dB numbers. Let alone average it over ... how many samples ?

[look&listen]

48 minutes ago, manisandher said:

I still think there's more to MQA than meets the eye (looking at an FFT plot). ----    And I still believe there's something interesting happening in the temporal domain with MQA.

Yes, agree, & despite hate mongering anti-MQA propaganda campaign.

Thx for another fine thread!

 

OT-  visual migrane just now subsiding, so can read again. Interesting to 'see' mechanisms of perception through old age, drugs, &c.  

[PeterSt]

2 minutes ago, look&listen said:

drugs, &c.  

 

&d

[audiobomber]

52 minutes ago, tmtomh said:

This is one reason why a firm grasp of digital sampling theory - and as @mansr often says, a firm grasp of the math involved - is so important. Otherwise we have no hope of being able to distinguish between an actual potential correlation on the one hand, and a metaphor on the other.

None of that is necessary for me to know. I'm just an audiophile and music lover. I only need to listen for what I like/dislike. The sibilance in C was a bit bothersome to me. Any other sonic differences were not annoying enough, nor attractive enough, to matter. 

[mevdinc]

I listen to music everyday and each time for a good few hours.
My initial impression of MQA was that I thought it sounded better.

 

When Tidal started to provide MQA nonMQA versions of the same music then I started to compare the 2 several times.

Yes, they definitely sound different.

 

As a listener I managed to observe 2 things about the MQA versions:
1) They defiantly sound 2-3 DB louder, which results in making the MQA versions more impressive/bigger.
2) You get more highs or should I say the highs seem to have more detail. But after a while it becomes tiresome, fatiguing.

I now always prefer the native versions over the MQA versions.

 

Best.
Mev

 
 

[tmtomh]

1 hour ago, audiobomber said:

None of that is necessary for me to know. I'm just an audiophile and music lover. I only need to listen for what I like/dislike. The sibilance in C was a bit bothersome to me. Any other sonic differences were not annoying enough, nor attractive enough, to matter. 

Well of course - one doesn't have to know anything about math to know what one hears, what one likes and dislikes, and to enjoy music. I didn't mean to imply otherwise - sorry if I did.

 

But my point is that one does have to have some regard for the scientific principles and mathematical concepts in order to be able to participate meaningfully in conversations about how observed data might correlate to what we hear. If you have no interest in why D sounds better to you than C, that's cool - honestly, not being snarky. But if someone were, for example, to claim that C sounds worse because of ultrasonic noise MQA adds, then one would need to have some basic theory of how or why that ultrasonic noise might be the cause of sibilance or reduced depth or texture in some of the sounds one hears. (In that case, I think one would have a tough time making the argument that ultrasonic noise was the cause.)

[fas42]

Deliberately didn't read through the thread before listening - and still haven't done so ... first thoughts, where's the treble?!! First track listened to was a bit dim; OK, the other one will be better - but it was worse! Uh oh, it's an audiophile track - none of that nasty, treble stuff around here, thanks!

 

So, the winner will be, the one that squeezes a bit more life from the events captured ...

[fas42]

Had a second listen, to confirm impressions ... sample C wins by default, because D is dead and buried, as far as sparkle is concerned - so, HiRez is a complete failure for conveying those subtle high frequencies, if these samples were examples of doing that.

 

Now, to look at what's going on in the waveform ...

 

 

[fas42]

So, is something going on in the time domain, in the posted waveforms, that's easy to see ... you betcha!

 

 

Okay, this is left channel only, of C vs D, where C has been aligned with D by removing samples from the beginning of C - the timing is at the top. Both have been upsampled to 352800 to improve aligning accuracy, and to see what's going on - and the key thing is that brickwall equalisation has been applied, you're only seeing the content above 10k in the above, all below 10k is attenuated near 120dB.

 

Note: the amplitudes have not been amplified to make this clearer, the levels at the left are the actual volume, compared to full scale of 1.

[audiobomber]

6 hours ago, tmtomh said:

Well of course - one doesn't have to know anything about math to know what one hears, what one likes and dislikes, and to enjoy music. I didn't mean to imply otherwise - sorry if I did.

 

But my point is that one does have to have some regard for the scientific principles and mathematical concepts in order to be able to participate meaningfully in conversations about how observed data might correlate to what we hear. If you have no interest in why D sounds better to you than C, that's cool - honestly, not being snarky. But if someone were, for example, to claim that C sounds worse because of ultrasonic noise MQA adds, then one would need to have some basic theory of how or why that ultrasonic noise might be the cause of sibilance or reduced depth or texture in some of the sounds one hears. (In that case, I think one would have a tough time making the argument that ultrasonic noise was the cause.)

I believe I am participating meaningfully in this conversation. How MQA works is for audio engineers to figure out. Whether it has financial promise, that is for marketers to determine

 

I don't need to know why MQA sounds better or worse than non-MQA, but I would like to know whether it does. I haven't bothered to download an MQA software upgrade for my iFi DAC, because I don't have any MQA files or source.

 

Honestly, I don't see this technology going anywhere, even if it is an improvement on what we have now. I was just curious about MQA so I listened to the provided files. I don't know at this point whether C or D was MQA. The answer will guide my opinion of MQA going forward.

[pkane2001]

1 hour ago, fas42 said:

So, is something going on in the time domain, in the posted waveforms, that's easy to see ... you betcha!

 

 

Okay, this is left channel only, of C vs D, where C has been aligned with D by removing samples from the beginning of C - the timing is at the top. Both have been upsampled to 352800 to improve aligning accuracy, and to see what's going on - and the key thing is that brickwall equalisation has been applied, you're only seeing the content above 10k in the above, all below 10k is attenuated near 120dB.

 

Note: the amplitudes have not been amplified to make this clearer, the levels at the left are the actual volume, compared to full scale of 1.

 

Here are the same two files (C and D) with a slightly larger scale, taken from about 2:17 mark. No upsampling or filtering applied, but the phase corrected to a fraction of a sample, as well as a minor phase drift corrected out.

 

Red represents D, the original PCM, while blue, C is the MQA encoded one. These are obviously not the same, although the differences are fairly minor, and appear to be in the higher frequencies:

 

There's no additional ringing detectable in the cepstrum plot, at least nothing obvious:

 

 

 

[tmtomh]

35 minutes ago, audiobomber said:

I believe I am participating meaningfully in this conversation. How MQA works is for audio engineers to figure out. Whether it has financial promise, that is for marketers to determine

 

I don't need to know why MQA sounds better or worse than non-MQA, but I would like to know whether it does. I haven't bothered to download an MQA software upgrade for my iFi DAC, because I don't have any MQA files or source.

 

Honestly, I don't see this technology going anywhere, even if it is an improvement on what we have now. I was just curious about MQA so I listened to the provided files. I don't know at this point whether C or D was MQA. The answer will guide my opinion of MQA going forward.

 

Hang on - I didn't say or imply that you were not participating meaningfully in this conversation. I said that in order to participate meaningfully in a conversation about what might be causing audible differences between MQA and non-MQA files, one needs to have some sense of the scientific and mathematical concepts.

 

In fact, I specifically wrote, "If you have no interest in why D sounds better to you than C, that's cool" - and in your comment I've quoted above, you wrote, "I don't need to know why MQA sounds better or worse than non-MQA." That's the same thing.

 

You've every right to your view and your perspective. But I'm not disagreeing with you, nor am I dismissing what you say or trying to insult you - in fact, I've taken pains to explicitly clarify that I am not doing that. So if you're going to respond to what someone says, please take the time to fully and accurately read what they've written.

 

That said, I hope you get the opportunity to listen to multiple "apples to apples" comparisons between MQA and non-MQA files, but it isn't easy to do so, because one has to ensure the mastering is identical, and it's difficult to extract the decoded stream as @manisandher has (or at least, you're not often going to run into comparisons where the person has taken the trouble to do what he's done).

 

Finally, MQA would say that the comparison we've done here isn't valid, because the MQA file did not have the final unfold/render, which includes "customized compensation for the DAC." In other words, from MQA's point of view, it's impossible to do an MQA vs non-MQA comparison online amongst a group of people like this, because for them the sound of an MQA file is not completely shaped until it's playing back in real time, through your particular DAC, using the "custom" final digital filtering that MQA has selected for your model of DAC.