MQA bails on Rocky Mountain Audio Fest*

@Tim Müller

I have agreed with some of your points recently, but not sure if you are trying sarcasm or humor, but not sure it is working.
Or perhaps you have moved into another realm of though??

I do not want to see my opinions or thoughts, look as if we are debating an us or them thing, think there is room for all views, and I truly welcome those that find 24 bit advantageous and truly like to see their ideas about how it benefits.

Me saying I do not see a lot of benefit, does not mean I disrespect those that do, and in fact I would love to agree with them, and possibly may in the future.

I see those with both views making great points, and feel you may be going a tad off the rails.

 

Well

Indeed, the funny is strong in this little thread.

It sidelined so often and so widely off topic that it's a miracle it didn't capsize. BTW, I have absolutely no idea why we are arguing here about DSD but I'll try to help steer @Tim Müller back on track: IMHO the synergy between MQA and DSD is so right, so obvious, so ...natural, that it's evident or even axiomatic that it is going to be the final nail in the coffin of PCM. Yes, MQA turned me into a newt, but now I feel better and can see the truth.

 

I could do, but I am afraid it would be too off-topic in this thread.

My point is:
It is common wisdom amongst audiophiles that more bits and higher sampling rates provide better details and better transient response.
Furthermore, the potential benefits of 24 bits and 96kHz or 192kHz are not so easily recognized in daily practical listening. So, my conclusion is, that 24 bits and 192kHz is not much of an improvement, and we still need more bits and higher sampling rates, to make the benefits more recognizable. Whether 32 bits or 128bits, whether 768kHz or 3072kHz are sufficient, is debatable. But the general route for better audio quality is clear:
Higher bit resolution and higher sampling rate.

MQA on the other hand, is going the opposite route: Lossy compression, lower sampling rates and less bits.
I feel, this is opposite to common audiophile wisdom.

 

As far as MQA goes, yes when companies like MQA come along, promise the earth but wrap it up in the way they have.

Then the record companies come along with the hardware manufacturers and try to sell us gear for an unnecessary vehicle to deliver another copy of Rumours that we probably don't need.

And the consumer gets stiffed again. No thanks.

 

I love how one fine poster got this off topic from MQA to HiRes and nobody blinks an eye but in another thread on box sets somebody mentioned postage and the Gort had a hissy fit. Interesting.

 

Sampling rate and bit depth are two different things entirely, and should not be used interchangeably...

http://thehub.musiciansfriend.com/tech-tips/sample-rate-and-bit-depth-an-introduction-to-sampling

If you have some practical experience with MQA vs hi rez or red book formats, I'd love to hear about it, but the generic old low rez/medium rez/hi rez speculative argument has already been done to death around here...

 

Not really what I was talking about, but ok...

 

Well, kind of. When the evangelism starts to appear extolling the virtues of formats that are pretty questionable, then yes, I'll add my comments too.

 

That is the same stairsteps argument like in
MQA bails on Rocky Mountain Audio Fest*

You are moving in circles.

 

If this even higher bits and sampling rates are providing an increase in SQ, they must provide less increase than going from Redbook to 24/96.

 

Yes, but you forgot to take the lossy compression into your account.

 

Common knowledge? Among whom?

No, you clearly do not understand sampling theory. The anti-alias filter constrains the input signal and its rate of change. There is no fine detail that cannot be captured, because the anti-alias filter ensures that the bandwidth of the incoming signal is limited to the Nyquist limit.

You do realize that a PCM waveform as it traverses a serial connection like S/PDIF, AES-EBU, etc. looks exactly the same?

I'm glad you've cleared that up!

And yet the best DSD converters have measurable SNRs exceeding 125 dB. Why is that?

Are aware that the output of a DSD ADC is the input signal with the addition of some (high-frequency) shaped noise, and that the shaped noise is essentially removed by the analog low-pass filter in the output of a DSD DAC, leaving just the wanted input signal? This has been proven mathematically and in practice!

Where are you pulling these figures from?

Repeating the same nonsense over and over isn't going to make it true. You can provide no justification for the above, overlooking the technical challenge in making a better than 32-bit converter.

Since you clearly do not understand the topic, please see:

https://repository.upenn.edu/cgi/viewcontent.cgi?article=1144&context=ese_papers

http://www.numerix-dsp.com/appsnotes/APR8-sigma-delta.pdf

 

Common audiophile wisdom would also appreciate the potential benefits of reducing temporal blur in digital sampling. MQA does that. Which is what I find interesting about MQA. Does the sonic benefit of temporal deblurring overcome the downsides of being lossy and losing some bits? That's the big question. Only way to find out is to listen to MQA with an open mind and open ears.

 

No, it doesn't do anything like that.
It's only the yes-men and the sock puppets try to make you fool into believing that.
Archimago's test revieled that nobody could anything like that.

To quote from my post:
It is common wisdom amongst audiophiles that more bits and higher sampling rates provide better details and better transient response.
MQA are doing the opposite: less bits and lower sampling rate. And lossy compression on top. Wonderfull!

 

Yes, MQA actually does everything like that. It is correcting for temporal blur. That's it's main thing. That's what it does.

In order to hear the benefits of temporal deblurring you need to play the files back on an MQA DAC. The software only decoding doesn't do that. You need a DAC with full MQA decoding support to hear the potential benefits of temporal deblurring.

Archimago's test did not require or need a MQA DAC to do the test. The test files were doctored up pseudo-MQA files. Files that were doctored up to be sort of like MQA but able to play that sort of MQA type of sound on DACs that don't have MQA decoding. Archimago's test was not a test of actual real MQA. To consider it a test of actual MQA is just wrong. I refuse to allow that pseudo-MQA test to be used to bash MQA. Because it was not a true test of MQA. Those of us who understand what the test was do not consider it a test of true MQA. If you want to bash MQA then bash it for the right reasons. Archimago's test is not one of those right reasons. I listened to the test files for Archimago's test. The original 192 and 96 source files from 2L sounded better than the doctored up listening test files that Archimago used for the test.

 

I searched Google for the term temporal blur just now. The first three pages had references mainly to video. One to a book on psychology. Anything audio related was purely MQA. Never mentioned as an issue in any other audio field or historical article in those three pages. Which struck me as odd given the supposed nature of the problem for us audiophiles.

Temporal blur: as I read on another forum just now, if you can't dazzle them with brilliance, baffle them with bullsh*t.

 

"Temporal deblurring" is a marketing term by Meridian. I can guarantee that the engineers at Meridian didn't start out their designs and ideas about MQA by googling "temporal deblurring" and going "OK, how do we do that?". Nope. Real engineering doesn't involve ideas and solutions that you can find in google. I guarantee that the engineers at Meridian already knew more about temporal timing effects of different filters and the effects of those temporal timing issues on the overall audio than they could ever find in a general google search. Google Scholar might be a better source for some info about research on things like temporal resolution in audio sampling. But even that is going to be limited. If you're an engineer at Meridian and have to learn about temporal resolution by doing a Google Scholar search then you've already lost your job. Real engineering books and talking to peers and doing your own research is how engineers learn and discover how to make the next generation of things. They don't learn from google.

So the fact you found nothing about temporal blur in a general google search means nothing. Real engineering solutions can't be found in a general google search. Imagine if the engineers designing the CD relied on Google to let them get ideas and figure out how to design the CD.

 

On a related note, have any real engineers endorsed MQA? I mean independent engineers not employed by MQA. So far, it only seems to be endorsed by audio journalists. Could someone link to digital audio engineering specialists' thoughts on MQA?

 

Hello Ham Sandwich,
Clearly from your statement, you have not understood that indeed the test IS a true reflection of MQA Core decoding as exactly what you would hear comparing software MQA decoding from TIDAL versus an equivalent 2L at 88/96kHz. The only thing that was a "doctored" was the MQA-like upsampling filter. The upsampling I used is very close to what Mytek uses in the Brooklyn when playing back 88/96kHz MQA decodes.

As I explained with the Audioquest Dragonfly Black and Mytek Brooklyn results, the DAC firmware isn't doing much more than upsampling with varying amounts of dither, noise shaping, and applying one of 16 minimum phase "leaky" upsampling filters. Sorry, no magic in any of this and if you experiment with these filter settings, you will quickly realize that differences are miniscule.

Remember, as Bob Stuart said, MQA typically achieves "15.85 bits" of resolution. The price to pay with the high frequency reconstruction is bit-depth resolution. Again, no magic here; there's no free lunch. Suppose we start with an original 24/96 hi-res file and we want to decrease the data rate to an equivalent 24/48 file (say for streaming), one has 2 options:

1. We simply downsample to 48kHz while maintaining 24-bit resolution and give up the ultrasonic frequencies above 24kHz = STANDARD downsampling.
2. We sacrifice 24-bit depth to "typically 15.85 bits", and encode the ultrasonic frequencies from 24-48kHz in a lossy fashion = voilà MQA encoding & decoding. Throw in some stuff about "deblurring" while you do this of course.

Which of the 2 do you choose? Do you think there's going to be a massive difference in sound quality?

Personally, I think solution 1 is just fine. 24/48 sounds great and in many cases would be easier to compress than MQA for streaming. Plus, since time domain performance is linked with bit-depth, one could argue that maintaining true 24-bit resolution provides better time-domain performance in the filtered reconstruction. Finally, I don't like MQA's poorly antialiasing filters. You can easily see the artifacts from these "leaky" filters:
Archimago's Musings: MEASUREMENTS: AudioQuest Dragonfly Black 1.5 - PART 2 (On "MQA Rendering")

See the section at the bottom when I reconstructed "Love Drought" from Beyonce to demonstrate the nature of the filtering being used after I saw the YouTube video. It's pretty obvious that MQA is adding ultrasonic "hash" that should not be there.

 

I've just been mansplained!

 

Unless the filter is tuned to my DAC that I'm listening to it is not a good representation of MQA. I'm using a Schiit Gungnir multibit DAC. The sample files didn't sound good to me. I preferred the original 2L files. Maybe if I was listening using a Mytek DAC it would be different? I don't like the sound of the Mytek DACs so it's not a DAC I would ever own, and it's not a DAC I would want to do a listening test with. If the sample files were tuned for the Mytek DAC filter then that may explain why I didn't like the sound of the sample files.

I still don't understand how you can do a MQA listening test that doesn't have the sample files specifically tuned for the DAC that people are using to listen to those files. Without that it is just not MQA. May be listening to MQA-like filters, but it's not an MQA listening experience.

 

If you can't find diddly independent objective material on temporal blur outwith MQA marketing fluff, how can this be "common audiophile wisdom"? That suggests knowledge you've already implied doesn't exist outwith MQA until recently.

If you take into account the seeming issues with MQA, objective opinion - at best - suggests the benefits are questionable, all things MQA considered. You can't have it both ways.

 

But what You write is only straw man arguments. Surely You must see this.

 

Yep, that isn´t that a common thing in audio?