MQA: The Day After
Here’s what Robert Stuart said about MQA at the end of the YouTube video introducing the technology:
“MQA is a revolutionary new British technology that allows us to capture the nuance and all of the detail of the original music performance but to put it in a file that is small enough to download or to stream today.”
You can check out the video yourself, although it’s pretty much a feel good piece lasting only a few minutes.
The innovation made possible by this new technology is the ability to deliver the benefits of high-resolution audio (96 or 192 kHz/24-bit PCM audio) in a container that is roughly the same size and has the same bandwidth of a CD. It’s easy to think of MQA as a data compression technology based on psychoacoustics but it’s much more than that. The researchers at Meridian have developed an encoding/decoding methodology augmented with metadata about the analog signal that can be used to deliver “lossless” audio at fidelity levels far beyond the specs of the sample rate and word lengths used. They’ve built a really smart end-to-end signal path.
The high-end audio world has moved aggressively towards high and higher sample rate thinking that they will bring improved sound quality. Recording at 96, or 192, or even 384 can result in audibly better recordings but an alternative benefit is the processing that is required to handle artifacts that come with digital audio. Here’s a quote from Robert’s AES paper called, “Hierarchical Archiving and Distribution”:
“It is now widely accepted that one key benefit of higher sample rates isn’t conveying spectral information beyond human hearing, but the opportunity to tackle the dispersive properties of brick-wall filtering. Wider-transition anti-alias and reconstruction filters directly shorten (proportionately) the impulse response and there is also more opportunity to apodize to remove extended pre- and post-rings.”
Another critically important factor (and one that I’ve failed to acknowledge in previous posts) is the resolution of human hearing in the time domain. There is new information that accuracy in the time domain is multiple times more important than the frequency domain. I’m not sure where the cutoff is for this level of accuracy but I’ve read and heard that 5 to 10 microseconds is the range. This requires a sampling rate of 192 kHz…according to this new information 96 kHz/24-bits is not sufficient.
The goal then is obvious. The audio world needs a way of getting the frequency, amplitude, and all important time domains provided by a sample rate of 192 kHz into a package that current bandwidth and storage capacities can handle. Meridian’s MQA does this by creating metadata about the time resolution (and other aspects of the analog signal) into a very low level signal (below the level of the noise) in a PCM file with a much lower sample rate. In a fashion similar to HDCD, the metadata is used to decode the MQA when the appropriate decoding equipment is sensed…otherwise the standard resolution PCM file is reproduced. How clever.
However, as important as this new technology is…it doesn’t make any changes to the original signal. It’s doesn’t reveal any additional nuances or fidelity in the original recording. The claims that it results in an improvement in fidelity over the source are misplaced. Here’s how one audiophile editor said it:
“I heard a wide range of music, from full-scale orchestral to voices to a very quiet piece by the Modern Jazz Quartet from the 1950s. I can still vividly recall the delicacy, ease, and resolution of the cymbals in the MJQ piece. I was also struck by the precision of their placement and how they appeared to float in the air against a completely silent background. The treble was totally unlike any other digital I’d heard, completely free from the metallic hardness and artifacts we assume are part-and-parcel of digital audio. Instrumental timbres were so naturally rendered to be almost eerie in their realism. Voices had a stunning palpability and immediacy that were all the more realistic for their compact image size and the sense that they were surrounded by a natural acoustic.”
What he heard was a standard definition analog sourced recording in MQA. The original source recording was limited by the technology of the 50s. To imagine that the recording suddenly had a “silent background” against which instruments floated or possessed treble that matches a new high-resolution PCM digital recording is hyperbole to be polite. There isn’t any “metallic hardness or artifacts” in a well done high-resolution PCM recording.
MQA is an important stepping-stone in the evolution of music delivery NOT a revolution is sound quality.
I’ll talk about how it’s done tomorrow.
+++++++++++++++++++
I’m still looking to raise the $3700 needed to fund a booth at the 2015 International CES. I’ve received some very generous contributions but still need to raise additional funds. Please consider contributing any amount. I write these posts everyday in the hopes that readers will benefit from my network, knowledge and experience. I hope you consider them worth a few dollars. You can get additional information at my post of December 2, 2014. Thanks.
‘Another critically important factor (and one that I’ve failed to acknowledge in previous posts) is the resolution of human hearing in the time domain. There is new information that accuracy in the time domain is multiple times more important than the frequency domain. I’m not sure where the cutoff is for this level of accuracy but I’ve read and heard that 5 to 10 microseconds is the range. This requires a sampling rate of 192 kHz…according to this new information 96 kHz/24-bits is not sufficient.’
I fail to comprehend this in the context of the Nyquist-Shannon theorem. I will need more theory to convince me. All this sounds (i.e. the youtube stuff and the announcement) like science interspersed with marketing mumbo jumbo. I have no problems accepting that this compression method provides excellent quality with reduced bandwidth requirements / file size. I just feel the whole thing is presented in marketing speak. Is it lossless or isn’t it? Is ot psychoacoustics or not? Frequency vs Time domain??? We will need more time to tell and distinguish fact from fiction. There we go again…
Forgot to mention that last time I checked I was studying digital speech processing thus it was looooong ago, frequency domain and time domain were just two different ways for expressing the same thing. Well, maybe things have changed since then but somehow I doubt it.
You’re right they are related…very closely related, which is why I’ve always focused on the frequency side.
I’m not saying they are just related. I’m saying they are exactly the same thing expressed differently. You just look at the same thing from a different point of view. You can convert from the one to the other using mathematical transformations (one usually chooses to deal in the freq domain since the mathematics are easier). Now, whether engineering problems (i.e. introduced errors = noise) arise during implementations, which partly invalidate this tautology lies beyond my knowledge domain. This might well be the case, which might explain why frequency vs time may need to be tackled differently in order to represent the one rather than the other with better accuracy. But all this is engineering not theory as far as I understand it. It may well be the case that I’m mising some fundamental thing here so I’m all ears (sic).
If a signal has a rise or fall time of 5 microseconds, that represents a bandwidth of 68 kHz for that content. It makes sense that a sampling rate of 96 kHz is too low.
The frequency and time domains are related by using Fourier transforms or Laplace transforms. I’ve spent my engineering career working with both on a regular basis.
Thanks Dick…I’m not ready to throw in the towel on 96 kHz/24-bit PCM…if the rise or fall time is 10 microseconds that would be 38 kHz and 96 kHz handles that just fine. Right?
You mean 34 kHz. The relationship is linear.
Of course! I should use a calculator! Thanks.
Of course. Talking about time domain as if it was a different thing than frequency domain makes no sense to me. The 5ms = 68kHz signal is, after all, a 68kHz signal which just happens to last for just one cycle! Yes, one the same xxkHz signals we have already spent years debating(?) whether they are audible or not…
Nik…the higher sample rate benefits more than just the higher frequencies…less trouble with filtering, less pre-ringing (and post-ring) on transients.
Interesting, Laplace transforms are a simple way of solving some ODE’s, nothing special otherwise. Surprised they’re being mentioned in the same breath as Fourier, maybe you should clarify.
Dave, I can’t pretend to be a math expert…I learned about Fourier and Bessel functions in grad school when I studied computer music but not in great depth. I’ll have to investigate.
Apologies Mark, the question was aimed at Dick James. BTW, what were you using Bessel functions for; I used them for partially solving PDE’s.
There is undoubtedly marketing speak on the website and in the videos…however, the paper and the patent documents are chock full of compelling information.
Apparently even people like Robert Hartley of Absolute Sound, with all of his critical listening experience and hundreds of thousands of dollars of listening equipment are not immune to the placebo effect — witness his description of the earth shaking revolutionary sound quality of the new MQA technology. To be honest, I don’t understand how a sophisticated listener like this can imagine that a technology which does not improve on the source material but only the packaging of that material somehow sounds transformationally better than the original CD or LP. What is going on here, and what does this say about all the other reviews in his magazine?
Robert is a very good writer, knows a lot about audio engineering, and is the head of a successful magazine, however he’s somewhat stuck in the trap of writing about audio. It’s the same flowery language and overblown descriptions…and about a recording done 50 years ago.
To be fair that recording was probably subject to a lot less processing than a modern recording and, despite the limitations of the equipment of the time, might have benefited from a “simpler is better approach”.
The MJQ recording is a terrific recording done when analog and simple ruled the day. My point is that it didn’t improve when processed through MQA. You got the same fidelity ast he original analog tape played on a great deck and system…and a high-resolution PCM transfer using great conversion etc would equal the analog tape or MQA version. The only advantage the MQA offers is a smaller footprint.
….and it the great question, whether this ‘small footprint’/the smaller size will engage more people for downloading/paying/listening to higher res’ed files?
If MQA is a completely transparent process of encoding and decoding, then it will make high-resolution files more accessible. The key question still remains whether anyone will care and whether the artists and labels will start making better sounding recordings.
Isn’t there the possibility that their ADC is producing a better result than what is used today? Hasn’t ADC had vast improvements since the introduction of the CD? Is that what Robert Harley is hearing?
I doubt it. I don’t think Robert is actually hearing anything new…he’s writing about the sound quality when MQA is not about sound quality. It’s about making files smaller without any loss. Yes, ADC have gotten better. But they’ve been good enough to handle analog recording fidelity for a long time.
I was always told that the human heard sounds as coincident if they were separated by 18ms or less. And that 27ms was the threshold for easy distinguishment. What has changed?
I can’t honestly say that I’ve done extensive reading on this…but the new numbers I hear are 5-10 ms.
Isn’t FLAC “small enough to download or to stream today” ?! That format is only just becoming established in a mainstream sense: a new format war could have more people choosing neither, rather than buying into it.
Much is made of Hi-Res involving frequencies above the range of human hearing. However, have there been any proper tests as to whether some of these high frequency sounds can sympathetically affect much lower ones, even in the bass?
It depends on the specs of the audio that is placed in the FLAC container. There have been some proper tests that establish that there are very real benefits to higher sample rates…whether we “hear” the ultrasonics or not is another thing.
If they mention “psychoacoustics”, then it means the format is not lossless. I’ll pass in favor of ALAC and FLAC, thank you very much. Then again, I have zero interest in streaming.
This is a lossless encoding/decoding process…and the nature of the scheme was informed by studying psychoacoustics and neuroscience…according to the information I have read.
It’s dang near 2015 and bandwidth speeds – storage space are no longer major issues for the majority of us.
What is a problem is industries profits. Besides the software the hardware people need a new “got to have” product to sell. So now we have a new compression system you’ll just have to buy.
Stewart asks about Stereophiles Robert Hartley’s comments that the new system makes huge audible improvements when we know this is technically not possible. Just more snake oil my friend, Stereophile has to support its advertisers if they expect the advertisers to support them.
It’s sad that a fantastic advance in audio like digital has led to the biggest explosion
in BS and snake oil that this industry has ever seen..
Clearly, the MQA development is important. If Meridian has really figured out a way to stuff 11 pounds in a 10 pound bag within existing formats, then we’re closer to streaming HD-Audio. I think that ultimately where we’re headed. But the whole think hinges on better quality to start with and that’s simply not happening.
Having to buy new hardware/new licences just because Meridian can ‘stuff 11 pounds in a 10 pounds bag’ – no way!
If you’ve already got download from HDtracks or others that you like in PCM, there is not need to adopt MQA.
“I can still vividly recall the delicacy, ease, and resolution of the cymbals in the MJQ piece. I was also struck by the precision of their placement and how they appeared to float in the air against a completely silent background. The treble was totally unlike any other digital I’d heard, completely free from the metallic hardness and artifacts we assume are part-and-parcel of digital audio. Instrumental timbres were so naturally rendered to be almost eerie in their realism. Voices had a stunning palpability and immediacy that were all the more realistic for their compact image size and the sense that they were surrounded by a natural acoustic.”
Sincere congratulations, you’ve just re-invented the yeast-grown sounding Upsampled MP3s which, as long as I listen to them, feature real-world sound sharpness & density, absolutely silent background {while being played through a very advanced Class B power amplifier}, 120 dB dynamic range, precisely focused 3D soundstage, nearly extreme detail & a degree of refinement {gloss} somewhat higher than DVD-A. Well, actually, there should be much to improve upon them.