MQA Users & Discussion Thread

[eltech]

4 minutes ago, davewantsmoore said:

 

He is correct.   A minimum phase filter does not have linear phase.

 

So, that causes the delay, and since it's in the passband, it's audible, right? Well, that's what I'm hearing

[eltech]

4 hours ago, davewantsmoore said:

There is an inference going around that this distortion at playback is a problem. 

I'm not Inferring, I'm saying outright I hear the time delay. 

 

No, it not a crackle, or "noise", that's not the sort of distortion I'm taking about. I am taking about the audible and obvious phase distortion of the minimum phase filter which is blatantly audibly obvious to ME!

I think it sounds awful!

It's distorting the sound!

It is a HUGE problem!

Edited April 23, 2018 by eltech

[wolster]

On 22/04/2018 at 3:24 PM, Eggcup The Daft said:

Sorry, everyone, I haven't been paying full attention. I know @legend is using a Project S2, but how are the rest of you decoding MQA when testing? I have a Dragonfly Red and am a bit dubious about some of the things it is doing, enough to be concerned about whether I'm hearing MQA in anything like a good light.

 

From memory, @legend first used a Dragonfly Red and experienced better results with the Project S2.

[Guest Sime]

@Eggcup The Daft I’m using a Mytek Brooklyn Plus. 

[davewantsmoore]

 

2 hours ago, eltech said:

I hope I haven't taken your quote out of context

You did...  but I follow what you're saying.

 

1 hour ago, eltech said:

 

I'm not Inferring, I'm saying outright I hear the time delay. 

 

I remain unconvinced that the phase of the MQA filter is what is responsible for what you are hearing.

 

2 hours ago, eltech said:

Since it's in the passband, it's audible, right?

There is a lot more to this, than just simply non-flat phase response is "temporal smearing" and "this is audible"  (ie. how archimago puts it).

 

It would be interesting to compare the response of the filter that Ric used in the upsampled files he sent you ..... vs the MQA filter.

 

 

Based on what we know, we would be (and MQA appears to be) more concerned with the filter ringing.... and you can see in the archimago measurements, there appears to be a 'balance' struck between filter phase linearity, and ringing.

 

 

FWIW, I just dialed in the amount of phase distortion shown in his measurement here: https://2.bp.blogspot.com/-fPWwCIIcWgc/WnXcPxrVCjI/AAAAAAAAPp4/Z8EPfXSrKZssayelXk2_PDnDxfpPxYd-wCLcBGAs/s640/04%2BMQA.png .... and set it so I could toggle it on and off at the push of a button.    I wasn't able to tell which was which (on or off) ..... I didn't test whether I could detect it being switched in or out (because there is a tell-tale pause).

[legend]

12 hours ago, wolster said:

From memory, @legend first used a Dragonfly Red and experienced better results with the Project S2.

Yes I thought the Project S2 (~$500) gave significantly better overall SQ than the Dragonfly Red (~$125) as you might hope/expect from their price differences but the Dragonfly was certainly capable of showing the differences (whether one likes them or not!) between MQA and equivalent non-MQA versions.

 

Some reviewers have compared the Project S2 favourably with the Mytek Brooklyn (~2k$) as  a DAC (but not for the Brooklyn's other features).  I would love to make the MQA comparison on an even better DAC - eg the Berkeley but at over $20k it is outside my financial reach!

[legend]

12 hours ago, eltech said:

I'm not Inferring, I'm saying outright I hear the time delay. 

 

No, it not a crackle, or "noise", that's not the sort of distortion I'm taking about. I am taking about the audible and obvious phase distortion of the minimum phase filter which is blatantly audibly obvious to ME!

I think it sounds awful!

It's distorting the sound!

It is a HUGE problem!

 

It is it possible to describe in more detail what you hear as the distortion - and what sort of music you are listening to?  I am genuinely interested to see if I can also hear it - my Project S2 DAC has the ability to switch between filters, including minimum phase, and I have so far only heard very subtle differences with mainly classical music (orchestral & piano) at different file resolutions (44/16 and 192/24).

 

BTW I have spent the last few years trying to root out all forms of distortion from my speakers and system generally, particularly 'euphonic' ones.  Hence the move from valve to heavily biased class A/B to Hypex Ncore amps.  And on the speaker front designing ones where the drivers operate largely in their pistonic as well as omnidirectional range; and cabinets that are very heavily braced with CLD to reduce cabinet resonances plus pyramidal shape to reduce internal standing waves as well as frontal diffraction effects; and DEQX-active that corrects for timing/phase errors plus room resonances as well as connecting the drivers directly to the amps with better dynamics etc.  The best example of this so far is the DEQX-Legend system demonstrated at the recent Melbourne Stereonet show - DEQX HDP5 + DEQX Ncore amps + Isobaric Small Red speakers.  Some people liked it - some did not!

 

[Guest rmpfyf]

6 hours ago, legend said:

It is it possible to describe in more detail what you hear as the distortion - and what sort of music you are listening to?  I am genuinely interested to see if I can also hear it - my Project S2 DAC has the ability to switch between filters, including minimum phase, and I have so far only heard very subtle differences with mainly classical music (orchestral & piano) at different file resolutions (44/16 and 192/24).

 

 

The piece was a live female vocal and piano, originally 44/16. Upsampled then to 192kHz using SoX's min and linear phase VHQ filters. Deliberately chosen as extremes.

[legend]

18 hours ago, rmpfyf said:

 

 

The piece was a live female vocal and piano, originally 44/16. Upsampled then to 192kHz using SoX's min and linear phase VHQ filters. Deliberately chosen as extremes.

 

Thanks.  Is there somewhere I can download it?

[legend]

Below are two quotes from the beginning and end of an article by Jim Austin just published in Stereophile magazine.  They pretty much sum up my views on the 'lossy' and 'DRM' aspects of MQA - but are much better expressed!  I look forward to his thoughts on the filtering etc aspects in the next issue.

 

 

[Guest rmpfyf]

1 hour ago, legend said:

Thanks.  Is there somewhere I can download it?

 

The test track or the filter tool? PM me and we'll sort it out.

[legend]

1 hour ago, rmpfyf said:

 

The test track or the filter tool? PM me and we'll sort it out.

Thanks - just PMed.

[davewantsmoore]

On 23/04/2018 at 6:24 PM, eltech said:

I'm not Inferring, I'm saying outright I hear the time delay. 

 

No, it not a crackle, or "noise", that's not the sort of distortion I'm taking about. I am taking about the audible and obvious phase distortion of the minimum phase filter which is blatantly audibly obvious to ME!

I think it sounds awful!

It's distorting the sound!

It is a HUGE problem!

It's seems to me like you're missing some understanding   (don't get me wrong again - I'm not arguing with what you are hearing.... only your explanation of the cause)

 

The linear phase filter has approximately the same amount of echo energy as the minimum phase filter.    You seem to be talking like the linear phase one is "perfect", and the minimum phase one is "bad".     Yes, the phase response for the MP filter is not flat .....  but the "smearing in time" caused by the resampler is still of the same magnitude for each.

 

The difference between them is where the echo is located in time.    In the minimum phase filter, all the echo is located after the impulse energy.   In the linear phase, half the echo is before the impulse.... and half the echo is after the impulse.

 

You can see a graphical illustration here:  http://sox.sourceforge.net/rate-44k1-96k.png   (you're interested in -vs, and -vsM)

 

 

Based on previous studies, it is expected that people are able to hear the pre-echo, more than the late echo....  although thee things are complicated so it's hard to say.    FWIW.  MQAs whole ideal is that neither of these resampling results are ideal .... and it is resampling (wether by the user, or inside a converser) which is reducing the quality of reproduced audio..... and so are attempting to reduce the amount of echo/smear, and optimise where in time is occurs.

 

 

Anyways.   To reiterate, this isn't about me not believing that you genuinely hear something in the sound of MQA that you don't like ...... just that you have grabbed onto something quite specific, and labelling that as "the cause".     Not that I really care what you think so specifically - but it just leads to potentially these types of generalisations pervading the 7 thousand odd people who've looked at this thread.....  other ones have taken decades to banish from 'audio accepted wisdom' 

[Guest rmpfyf]

11 minutes ago, davewantsmoore said:

It's seems to me like you're missing some understanding   (don't get me wrong again - I'm not arguing with what you are hearing.... only your explanation of the cause)

 

The linear phase filter has approximately the same amount of echo energy as the minimum phase filter.    You seem to be talking like the linear phase one is "perfect", and the minimum phase one is "bad".     Yes, the phase response for the MP filter is not flat .....  but the "smearing in time" caused by the resampler is still of the same magnitude for each.

 

The difference between them is where the echo is located in time.    In the minimum phase filter, all the echo is located after the impulse energy.   In the linear phase, half the echo is before the impulse.... and half the echo is after the impulse.

 

You can see a graphical illustration here:  http://sox.sourceforge.net/rate-44k1-96k.png   (you're interested in -vs, and -vsM)

 

 

Based on previous studies, it is expected that people are able to hear the pre-echo, more than the late echo....  although thee things are complicated so it's hard to say.    FWIW.  MQAs whole ideal is that neither of these resampling results are ideal .... and it is resampling (wether by the user, or inside a converser) which is reducing the quality of reproduced audio..... and so are attempting to reduce the amount of echo/smear, and optimise where in time is occurs.

 

 

Anyways.   To reiterate, this isn't about me not believing that you genuinely hear something in the sound of MQA that you don't like ...... just that you have grabbed onto something quite specific, and labelling that as "the cause".     Not that I really care what you think so specifically - but it just leads to potentially these types of generalisations pervading the 7 thousand odd people who've looked at this thread.....  other ones have taken decades to banish from 'audio accepted wisdom' 

 

The redistribution of spectral energy in time around for an impulse response in the filters chosen is, of course, equal in magnitude though the Q of either is very different. More peak amplitude on the linear, and the ring on the min phase is more than 2x the pre/post on the linear. 

 

So a few things to listen for, then.

[davewantsmoore]

14 minutes ago, rmpfyf said:

More peak amplitude on the linear, and the ring on the min phase is more than 2x the pre/post on the linear. 

 

Perhaps I've confused the semantics, but.

 

The "time smear" for the linear is 0.02s (using the example chart I posed) .... and the time smear for the minimum is 0.02s.

 

The differences is just where the ringing is arranged in time.    Let's anyone thinks you mean that the min-phase is overall ringing "2x".
 

[Guest rmpfyf]

54 minutes ago, davewantsmoore said:

 

Perhaps I've confused the semantics, but.

 

The "time smear" for the linear is 0.02s (using the example chart I posed) .... and the time smear for the minimum is 0.02s.

 

The differences is just where the ringing is arranged in time.    Let's anyone thinks you mean that the min-phase is overall ringing "2x".
 

 

The filters I'd sent @eltech didn't have identical impulse response peaks; it's therefore impossible that their time domain response is identical.

[davewantsmoore]

9 minutes ago, rmpfyf said:

 

The filters I'd sent @eltech didn't have identical impulse response peaks; it's therefore impossible that their time domain response is identical.

Sure... I wasn't inferring that.

 

What I was saying is that the magnitude of the duration is the same for each.    The minimum phase puts all the energy after the response ..... and linear phase put the response in the middle (so there is both a pre and post echo).

 

 

Just trying to dispel the notion that the MP filer contains "all this time smear" ... and the LP filter contains "none".    Otherwise people reading along might get the basics of how this all works quite distorted, and start leaping to incorrect conclusions.

 

Indeed... studies would invite us to believe the the distribution of energy in the LP filter could be audibly worse....   but of course, the world is not ever black and white.

[Guest rmpfyf]

5 minutes ago, davewantsmoore said:

What I was saying is that the magnitude of the duration is the same for each.

 

Not exactly. The example provided has a longer net duration. Giveaway being the difference in peak amplitude. Small differences at high frequencies though that's the game.

 

8 minutes ago, davewantsmoore said:

Just trying to dispel the notion that the MP filer contains "all this time smear" ... and the LP filter contains "none".   

 

That's a valid point. Though to get it smaller, up the sample freq and we're there... hence MQA and hires. 

 

I take @eltech's point of 'that's not the tool the mastering engineer completed their work with, and accordingly it's not the final work they signed off on - so it can only be different'. Similarly, there's no real quality advantage in MQA unless it's an end-to-end process, and even then if it can be proven that ultra-short filters with bugger-all (or variable) pre-ringing make a difference. You need the high sample freq as the post-ring otherwise is very audible. Redbook uses sinc for a reason.

 

Question is whether that difference is pleasing or not. 

[eltech]

On 24/04/2018 at 7:05 AM, legend said:

It is it possible to describe in more detail what you hear as the distortion

 

To answer your question requires me to perhaps provide some credentials of hearing acuity.

 

A few years ago I was interested in an amplifier made by a famous Australian amplifier manufacturer in Victoria who specifically designs transistor amplifiers to sound like a valve amp. I went to his house for an audition. He started out by playing me some music on his lowest price amplifier. I said, its nice, but I can hear the second order distortion. It sounds a lot like a valve amp. Its not my cup of tea. He then plugged in his more expensive model and we had a listen, and I said, that's better, but I can still hear that second order distortion. So, he put on his most expensive amplifier and it did have much lower audible second order distortion, so I took that one home for a “home demo”. In a week he rang me to ask what I thought, and I told him that though I thought it was the best of the lot, with an obviously clean signal path, the second order distortion was still too high for my liking and I preferred my existing amplifier because of its lower audible distortion.

He told me that I was pretty much the only person he knew who was able to pick it out, and hear the difference in distortion between amplifiers.

 

I recently obtained a linear tracking turntable. Fortunately it uses a Audio Technica cartridge which allowed me to install my AT150Mlx stylus onto the P-mount cartridge. I had a listen and was immediately pleased to hear a marked noticeable reduction in tracing distortion which had always annoyed me with pivoting tonearms.

What I hear with a pivoting tonearm is a change in the distortion spectrum as the pivoting tonearm traverses the playing area of the record. The best way for me to describe it is to say that the tone of instruments change throughout a side. Its probably most noticeable with high hats, but violins suffer from it too. Second order distortion also blunts the attack of the high hat, and second order distortion seems to spread out the sound and slow the attack. You'll have to forgive my use of language here because English has very few words to accurately describe sound. There is a “creamyness” to second order distortion which is audible to me. To explain, low distortion is like drinking clean fresh water, and second order distortion is like drinking pouring cream.

I do apologise for this flowery language but I'm struggling to convey in a meaningful way what I hear, only that I hear it, and its obvious to me.

 

When I listed to the first track on the 2L website I listened through one speaker only. I was using a professional USB DAC. I listened to the 16/44 version and heard the violin sounding prominent, I heard the bow on the strings, and extended high frequencies, and the female vocalist was clear and prominent. When I then listened to the same track which was encoded with MQA I heard a perceived lessening of amplitude of the high frequencies, it sounded like someone had used EQ to lower the highs. Weird!. I then noticed that the violin seemed also like it had sunk into the mix, and the attack on the violin was no longer present - no bow on string sound. I then listened to the non-MQA high res version and found that it sounded much closer to the 44/16 version than the MQA version. My preference was for either version not encoded with MQA. Things that are audible to me with MQA is a change of tone, a change of attack, and inexplicable perception of a softening of the mid and top end, and all of this together gave me a strong perception of a huge loss in clarity and involvement with the music.

The files @rmpfyf gave me dont have exactly the same effect, with his minimum phase file I found a slight dulling of attack, and an excessively soft presentation to the mid and highs, and there was a creaminess to the female vocal. I suppose I could say it sounded a bit smeared, however his files didn't exactly lose the clarity in the same way I perceived the MQA files to have lost clarity.

I can imagine that some people would subjectively enjoy the minimum phase file sent to me by @rmpfyf but the obvious loss of cleanliness and clarity is too much for me to enjoy. I much preferred the 44.1 file.

I hope this reply gives you some insight into how I hear things.

Edited April 25, 2018 by eltech

[davewantsmoore]

2 hours ago, rmpfyf said:

Not exactly. The example provided has a longer net duration.

Not quite sure exactly what you mean.... in the linked example, the MP extends to 0.02s ... and the LP extends to 0.01s....  but because the LP begins at -0.01s, then the total duration of 'smearing' is 0.02s.

 

2 hours ago, rmpfyf said:

I take @eltech's point of 'that's not the tool the mastering engineer completed their work with, and accordingly it's not the final work they signed off on - so it can only be different'

Yes, of course....  that's the whole point (improvement).

[Guest rmpfyf]

11 minutes ago, davewantsmoore said:

Not quite sure exactly what you mean.... in the linked example, the MP extends to 0.02s ... and the LP extends to 0.01s....  but because the LP begins at -0.01s, then the total duration of 'smearing' is 0.02s.

 

Not quite in the filter as I applied it - I deliberately took it to extremes. It is of course possible to design a filter as you suggest.

 

1 hour ago, eltech said:

The files @rmpfyf gave me dont have exactly the same effect, with his minimum phase file I found a slight dulling of attack, and an excessively soft presentation to the mid and highs, and there was a creaminess to the female vocal.

 

Interestingly I gave @eltech no forewarning of what the filters were - just that one was 'minimum phase' and one was 'linear phase'. These comments are however consistent with how the filters were employed - there was less peak amplitude and a longer ring.

 

If this had to be extended as a test I'd suggest running the sample frequency as hard as a given DAC can sustain to get the time domain responses as sharp as possible, then to play with filter design to see if a more pleasing outcome can be obtained. Some DACs in practice work just as such and employ DSPs to do the heavy lifting on resampling per se. I don't mean 192-220kHz as is common for xover or (more increasingly) room correction DSP work, we're talking 384kHz.

[Guest Eggcup The Daft]

On 22/04/2018 at 8:37 PM, eltech said:

What are you observing?

I've been doing some more listening to test tones (that makes me a sad case) and it looks like the blame may go elsewhere.

What I discovered initially was that playing high frequency tones from Tidal at a high volume (higher than I would listen to through the headphones I have available) caused the pitch of the tone to lower, or a loud artefact, so that there is an audible tone - for a 20kHz test tone, at full volume, the audible tone at full volume is lower than an 8kHz tone in pitch. This is quite a shock.

Closer listening reveals that the tone starts to become audible at the sort of volume level I'm listening at normally with my EL-8s. If I go much above that volume level with music from Tidal, then the sound becomes blurred and "shouty", so I suspect that the effect I'm hearing with the tones is there with music also. The effect is not there if I play back files (normal or hi-res) through other players - I can turn up the sound and it is clean up to dangerous listening levels.

 

If I go through the same routine with the SoundBlaster Recon3Di chipset on my PC, I get static from Tidal at the moment, but it's intermittent and I can report similar if less dramatic effects than with the Dragonfly. The low tone is different. Again, this does not happen with files played back from other players.

 

As an aside, the volume control in the Tidal player is worse in terms of producing the lower tones than if I force maximum volume and let the Windows volume control (which I understand uses the control in the Dragonfly) control volume.

 

The tones I'm using in Tidal are the "HERMES PH1 SOUND-EFFECTS / ULTRA HIGH FREQUENCIES (HEARING TEST HI PITCH)" files. These like all the test tone discs on Tidal are not available in MQA/Master format. You may want to try checking these from Tidal through your DAC to see if you have similar potential faults.

 

I have to point the finger in part at the Tidal player...

[davewantsmoore]

1 hour ago, rmpfyf said:

Not quite in the filter as I applied it - I deliberately took it to extremes.

What did you do differently?

 

[legend]

16 hours ago, eltech said:

 

To answer your question requires me to perhaps provide some credentials of hearing acuity.

 

A few years ago I was interested in an amplifier made by a famous Australian amplifier manufacturer in Victoria who specifically designs transistor amplifiers to sound like a valve amp. I went to his house for an audition. He started out by playing me some music on his lowest price amplifier. I said, its nice, but I can hear the second order distortion. It sounds a lot like a valve amp. Its not my cup of tea. He then plugged in his more expensive model and we had a listen, and I said, that's better, but I can still hear that second order distortion. So, he put on his most expensive amplifier and it did have much lower audible second order distortion, so I took that one home for a “home demo”. In a week he rang me to ask what I thought, and I told him that though I thought it was the best of the lot, with an obviously clean signal path, the second order distortion was still too high for my liking and I preferred my existing amplifier because of its lower audible distortion.

He told me that I was pretty much the only person he knew who was able to pick it out, and hear the difference in distortion between amplifiers.

 

I recently obtained a linear tracking turntable. Fortunately it uses a Audio Technica cartridge which allowed me to install my AT150Mlx stylus onto the P-mount cartridge. I had a listen and was immediately pleased to hear a marked noticeable reduction in tracing distortion which had always annoyed me with pivoting tonearms.

What I hear with a pivoting tonearm is a change in the distortion spectrum as the pivoting tonearm traverses the playing area of the record. The best way for me to describe it is to say that the tone of instruments change throughout a side. Its probably most noticeable with high hats, but violins suffer from it too. Second order distortion also blunts the attack of the high hat, and second order distortion seems to spread out the sound and slow the attack. You'll have to forgive my use of language here because English has very few words to accurately describe sound. There is a “creamyness” to second order distortion which is audible to me. To explain, low distortion is like drinking clean fresh water, and second order distortion is like drinking pouring cream.

I do apologise for this flowery language but I'm struggling to convey in a meaningful way what I hear, only that I hear it, and its obvious to me.

 

When I listed to the first track on the 2L website I listened through one speaker only. I was using a professional USB DAC. I listened to the 16/44 version and heard the violin sounding prominent, I heard the bow on the strings, and extended high frequencies, and the female vocalist was clear and prominent. When I then listened to the same track which was encoded with MQA I heard a perceived lessening of amplitude of the high frequencies, it sounded like someone had used EQ to lower the highs. Weird!. I then noticed that the violin seemed also like it had sunk into the mix, and the attack on the violin was no longer present - no bow on string sound. I then listened to the non-MQA high res version and found that it sounded much closer to the 44/16 version than the MQA version. My preference was for either version not encoded with MQA. Things that are audible to me with MQA is a change of tone, a change of attack, and inexplicable perception of a softening of the mid and top end, and all of this together gave me a strong perception of a huge loss in clarity and involvement with the music.

The files @rmpfyf gave me dont have exactly the same effect, with his minimum phase file I found a slight dulling of attack, and an excessively soft presentation to the mid and highs, and there was a creaminess to the female vocal. I suppose I could say it sounded a bit smeared, however his files didn't exactly lose the clarity in the same way I perceived the MQA files to have lost clarity.

I can imagine that some people would subjectively enjoy the minimum phase file sent to me by @rmpfyf but the obvious loss of cleanliness and clarity is too much for me to enjoy. I much preferred the 44.1 file.

I hope this reply gives you some insight into how I hear things.

 

Many thanks for your detailed reply – much appreciated.  It seems that we listen for similar things – but may interpret them differently sometimes!

 

I agree with you entirely about the negative effects of 2nd order distortion.  Although it can be ‘euphonic’  and can mask other forms of ‘nastier’ distortion elsewhere in the reproduction chain it does also blunt dynamics etc and make things sound woolier if nice but much the same.  My favourite analogy is it is like adding 3 tea-spoons of sugar to everything.  However I am surprised (though impressed!) that you can hear such a difference between the inner and outer edges of an LP with a pivital tone arm given that even the best (most expensive) cartridges rarely have a THD of less than 1%, much of it 2nd order, plus all the resonances of tone arm and table etc – which is why I rarely listen to my TT system these days.

 

@rmpfyf also kindly sent me copies of the music (piano plus female voice) in the original 44/16 plus 192/24 with linear phase and minimum phase.  I only had time for a quick listen last night but thought the most obvious difference between the original and upsampled ones was the bass notes of the piano near the beginning – to me they sound better defined, even subjectively deeper.   I thought there was also a small difference in the high frequencies of the female voice but this difference seemed slightly greater between the linear and minimum phase.  But I think this is where we differ.  To me the minimum phase sounded slightly less harsh – dare I say it, with less digital glare! Any change in the dynamics of the piano did not jump out at me.

 

I repeat I only had time for a quick listen and hope to do so again tonight in more detail.  The differences I heard were not great (something akin to the changes in brand of polypropylene capacitors I was listening to recently) and may depend on the system through which I was listening – Project S2 DAC + Ncore 500W amps + passive Kantu 10 loudspeakers. I also want to try listening through some other systems, particularly the DEQX active ones.

[Guest rmpfyf]

15 hours ago, davewantsmoore said:

What did you do differently?

 

VHQ filters (see below on pic ripped from Archimago) - smaller peak amplitude on the min phase, different Q. The filter may 'act' as long in time but not with the same amplitude response magnitude.

 

Consistent with comments from @eltech and @legend

 

It is of course possible to tweak these and get a hybrid. There might be a scenario in the middle that is more liked by more than either. Want to try it?