Chesky To Release MQA CDs in May

New York-based Chesky Records has just announced its first two MQA-encoded CDs. On May 19, both Camille Thurman's new Inside the Moment and Rebecca Pidgeon's classic The Raven will hit the streets, as it were, with a list price of $18.98. Music lovers with MQA-enabled DACs will be able to experience the full benefits of MQA-encoding/unfolding, while others will enjoy Red Book quality sound.

"The decision to issue these titles as MQA CDs was decided quickly, just 10 days ago, as we were exploring the possibility of working with MQA in our studio," David Chesky explained by phone. "After speaking to Bob Stuart of MQA, we realized we could easily apply MQA to CD. Alan Silverman, whose ARF Mastering is right next door, did the MQA encoding, and both Bob and Spencer Chrislu of MQA double-checked the results."

Once Chesky heard MQA, he was easily convinced of the benefits of MQA CDs. "You get a hi-rez file in the same package as the CD," he said. "It's like listening to the master file on CD. If all goes well, we'll do all our future issues like this."

Chesky also hinted that MQA downloads are in the offing from HDTracks "as soon as the store is ready."

As for the choice of what to issue first, it seems to have been a no-brainer. Thurman, he says, "sounds amazing, like a young Sarah Vaughan, and she plays sax as well." John Atkinson, who was present at one of the concerts where the album was recorded, enthusiastically agrees. "It was an extraordinary experience to hear this young woman channeling Coltrane," he reports.

As for Pidgeon's classic, most audiophiles have already heard it in multiple formats, which will make comparison with the MQA issue easy.

"We needed to press a new batch of The Raven, so issuing it in MQA makes perfect sense," he said. "Now people can play it alongside the other versions and see what they like. The title is also available for download in 176.4/24, which makes for a great comparison."

Chesky feels that MQA sounds a little more analog, relaxed, and organic sounding than other digital issues. When asked to say more about his personal impressions of the format, he opted for language clearly his own.

"Listen, what is real and what is not real? You sit in front of two boxes in your home, and it's a magic trick. But we can't say live is always better, because some halls are over-reverberant, excessively dry, or just plain bad. Sometimes, the recording sounds better than the live experience. MQA addresses digital harshness, and gets the timing and transients right. It's like walking through the woods and smelling the trees vs putting Febreze in your house."

Chesky Records records most of its new titles with an MSB custom converter, specially-designed Crystal Cable, and either B&K microphones or binaural heads. While the label's next title, a 24/48 download-only issue of Chesky's Spanish Poems, featuring soprano Maureen McKay and the Orchestra of the 21st Century, won't be MQA-encoded in its initial offering, most if not all disc releases to follow will be. McKay, whose international career is on the ascendant, came to Chesky by way of Metropolitan Opera conductor Fabio Luisi, who recommended her highly.

Not all labels are following Chesky's position on MQA. Linn Products in Scotland, for example, which was a pioneer in releasing hi-rez recordings, published a strong policy statement on MQA on their website.

COMMENTS
volvic's picture

I would love in the future to compare an MQA encoded CD with music I am familiar with to see if there is a difference. With some SACD's I remember hearing a marked improvement but with other recordings a negligible one. Same with HDCD, for the most part there was an improvement but sometimes I felt I was fooling myself. Let's see what the future holds with MQA on CD.

Anton's picture

Some of those SACDs were likely just redbook CD with a new sticker pretending to be higher rez.

I am fine with MQA as a download toy for now, but will wait a while before clearing my elcassettes, laserdiscs, HDCDs, DVD-As, and other defunct 'greatest sound ever' media off my shelves.

volvic's picture

Yes, the Karajan 60's Beethoven SACD were duds. Could it be possible they just re-mixed and slapped an SACD logo on a 16 bit recording? Not sure, but the sound was no different from the CD. I too will wait, have spent enough during my lifetime to know that patience is the best medicine for my wallet.

hifitommy's picture

karajan's rite of spring was a DG recording and had been poorly multimiked and sounded just that way but it was a very good performance, PLUS you could pressurize a room with it. i would never buy a dg SACD unless i had listened to and was enthralled with it.

multi miking CAN be successfully carried out and not sound like it such as the mehta planets with the LA Phil. what a masterpiece. i have the xrcd and it doesn't compete with the LP and i have not heard the sacd.

NeilS's picture

"...Music lovers with MQA-enabled DACs will be able to experience the full benefits of MQA-encoding/unfolding, while others will enjoy Red Book quality sound."

Undecoded MQA is not Red Book quality, as I understand the conclusions of the analysis posted by "Miska" on computeraudiophile.com

https://www.computeraudiophile.com/blogs/entry/466-some-analysis-and-com...

anomaly7's picture

Seems to me, if MQA removes some data, and repackages what's left only to be unpackaged later during playback, that you should be able to play the CD's you already have with a device that does this on the fly. Of course, it's much better for the MQA industry if you have to buy every recording all over again, but perhaps some manufacturer will make my idea a reality in the future. You could even apply that technology to vinyl record playback, if you digitized the LP output and reworked it for MQA.
Your thoughts Jason, on the future potential for a player that reprocesses your current CD's or digital files with MQA?

David

Jason Victor Serinus's picture

Among other things, MQA addresses timing anomalies in the A-to-D converter used for the master recording. Unless the identity of that converter is unknown, the MQA process is customized. I do not know how that could be done after the fact, within a disc player.

Comments that people will be forced to buy all their recordings once again are off the mark, and fear-based. If you're happy with what you now have, you need do nothing. That's as true for the 20th repackaging of Ella and Louis as it for the MQA version of a recording made two years ago. Then again, if you wish to explore the latest pathway to better sound, you can obtain or, in some cases, upgrade to an MQA DAC, and start by obtaining new recordings that are MQA-encoded.

anomaly7's picture

Thanks Jason, I'll keep an open mind and open my ears when we do an MQA demo in the future with the San Francisco Audiophile Society coming up (hopefully) soon.

AJ's picture
Quote:

MQA addresses timing anomalies in the A-to-D converter used for the master recording.

Where is Bob Stuarts evidence for this claim? I don't see any test data of all ADs used in mastering for the last 30+ years on the MQA or Meridian website. His AES paper certainly presented no such data, but rather used some concocted Matlab filters.
Nor do I see anywhere in Meridian literature that all their digital players have been "temporal blur" players for the last 20+ years. Actually, the claims in their user manuals are quite the opposite, with terms like triangular dither allowing "transparent" playback.
Odd how Chesky never admitted all these years that his recordings suffered such AD maladies either, that he can now hear post MQA.

Btw, the "temporal blur" "fix" used by MQA, comes at a price - more aliasing distortion. Possibly to audible levels! This article by Dr Lesurf explains it nicely http://www.audiomisc.co.uk/MQA/origami/ThereAndBack.html

No problem with folks liking their music spiced up with some aliasing distortion, but having listened to numerous MQA tracks myself, I would hesitate to call it uniformly "better". YMMV.

crenca's picture

I used to respect Chesky...jumping on the MQA desperation-wagon is not a good sign.

CheskyRecords's picture

Chesky Records prides itself on recording music in the highest possible resolution, providing the music to you in as close proximity to that as possible, and always pushing the boundaries of ways that technology can enhance the listener experience. The launch of our new MQA CD is done to this end. Please be aware that the CD functions as a hybrid. Those who are capable of accessing the full capabilities of the 176/24 are able to do so. Those simply listening for red book audio quality will be able to the same on the same disc. If PCM downloads are what you prefer, those will still be available. This new format isn't replacing any other for us. We're simply providing you another option to experience our high resolution offerings. We're happy that you're listening, regardless which format is your preferred method. Thank you for your continued support of Chesky Records.

Chesky Records Team

crenca's picture

Calling the DRM/IP that limits an end user to the < 16/44 bits "hybrid" is one way to put it, and obviously a bit of a controversy from a consumer point of view. Though if I were you (selling a product) I would do the same thing ;)

Also, note well that on a digital, software, engineering, and mathematical level the < 16/44 bits of an MQA software file is NOT "red book audio quality". The *claim* by MQA and its supporters is that it "sounds like" redbook quality - although there is the concurrent and contradictory claim that it is in fact an improvement on "redbook quality" from a subjective point of view. However you choose to judge it from a SQ point of view, you can not claim that it is PCM/redbook because those are defined from a mathematical point of view - and MQA purposely changes said math by design (for the "enfolding" process, etc.). Also, as MQA admits (reluctantly) that the bit's above the audioband (starting about 40khz if memory serves) are encoded in a lossy manner, it is a bit of a misnomer to say that MQA offers 176/24 capabilities, if by that you mean the same thing as when you compare it to PCM 176/24 - unless you qualify it with some sort of "sounds like" language.

I am heartened and encouraged to hear that you will continue to offer the equivalent PCM resolution. MQA is many things, but in the end I (and many others) judge it to be a failed attempt to innovate that appears to only benefit MQA and possibly labels - the consumer is left holding the $DRM$ bag on this one. I for one support innovation, but then I also don't think consumers should be compelled (by limited supply/choices) to eat its (often) spoiled fruits...

NeilS's picture

If Chesky has decided that for future releases it'll be MQA, undecoded MQA or the highway, I'll be taking the highway.

harryas's picture

why you condemn MQA without seemingly having heard it. You could start by hearing "Master" (MQA) files from TIDAL that are available in around 500 albums there for streaming. If you have a Mac you can try to hear TIDAL and MQA files for a free test period through Audirvana 3 which unfolds MQA-files. I never could imagine hearing such a lively, detailed, clear sound with a hearable stage wide and deep even through my Mac (or, much better, connected through the headphone socket to my system)! How much better would it sound through a MQA-capable DAC! And you can try this without buying expensive appliances. Then tell me again what you think about MQA!

low2midhifi's picture

I read this article, as other recent ones, with all of its world-changing promise. I then read the apocalyptic rebuttal (or prebuttal, if you will) on Linn's site. The dramatic stances of proponents and detractors alike will not find realization of the glorious future, or wretched future, that either side predicts.

This is yet another flavor-of-the-month development in an aggravatingly fragmented track of "innovation" in the recording industry. I'll just stick to what's tried and true, or cheap and reliable, until the dust settles on a feasible long-term alternative.

Until then, the advent of yet another standard, in yet another format, brings to mind a song:

"Johnny come lately, the new kid in town......everybody loves you, so don't let them down...Where you've been lately? .....There's a new kid in town.....everybody loves him, don't they ....And he's holding her, and you're still around......"

I wonder how much better this sounds in MQA or the next flavor-of-the-month?

Jason Victor Serinus's picture

Unless you have heard a well-conducted with/without MQA demo on a revealing system, or taken the time to listen to it on Tidal, and failed to hear a significant difference, this comment, like many others, is based not in fact, but in a priori assumptions and bias. There is no "world changing" promise in my news story, unless you consider "better sound" to be world-changing. And, as much as I love opera, I fail to see how the story is "dramatic." It simply states that David Chesky and his team are convinced that MQA offers better and more realistic sound, and will make that option available to their customers.

Nor would I dismiss a technology now embraced by two major labels, with the third expected to come on board, as a mere "flavor-of-the-month." But such differences in perspective are what makes the world go round. Long may it spin.

readargos's picture

From no less than Andreas Koch appears below. I'll remain optimistic, but it does seem all the reports of controlled comparison could be rigged in MQA'so favor.

http://positive-feedback.com/audio-discourse/questions-answers-mqa-inter...

John Atkinson's picture
readargos wrote:
From no less than Andreas Koch appears below.

Thank you for the link. But I think it fair to point out that Andreas Koch is heavily invested in DSD, another proprietary format that is incompatible with MQA. This disagreement between Koch and Stuart goes back to the SACD vs DVD-A days at the turn of the century.

readargos wrote:
I'll remain optimistic, but it does seem all the reports of controlled comparison could be rigged in MQA's favor.

I can assure you that none of the comparisons that have been written about in Stereophile were "rigged."

John Atkinson
Editor, Stereophile

Solarophile's picture

Considering that we have abundant software like foobar plugins, JRiver playback, dBPowerAmp conversion, support with DoP for DACs. Much of it free.

SACD is a copy protected format. But the DSD bitstream itself is not a proprietary "format". Not in the way MQA is.

Rivalry or not, Koch is technically correct and this has been shown by others.

readargos's picture

Impropriety on the part of any magazine or e-zine; rather that, as Koch's article suggested (especially given the limitted availability of MQA-encoded files in the early reporting), that the "before" MQA encoding samples had been carefully curated to ensure the "after" samples sounded better.

The idea is not so strange. I'm sure most of us have heard equipment that excels on some types of material - perhaps it makes our favorite demo tracks sound wonderful - but does not hold up as well with long exposure and over many genres of music.

By the time there is wide exposure to MQA files, it likely will have already succeeded commercially, at which point the quality issue will be irrelevant to the wider, non-audiophile, marketplace.

crenca's picture

A demo is always just that right? Certainly those giving the demo would not try very hard to optimize the PCM playback would they - the DAC's involved were designed from the beginning to highlight/sell MQA were they not?

In any case, I think it is fair to say that most of what your publication (and most others) have written about MQA as far as a "comparison" have been demo's under the direction/control of MQA representatives. The only exception I know of is your very own comparison of recordings you engineered yourself which were given the MQA treatment, and sent back to you for comparison in your own system/room. If I recall correctly (correct me if I am wrong) you could identify the MQA 4 out of 7 times.

Jason Victor Serinus's picture

Peter Mc Grath and John Atkinson have heard their own hi-rez files before and after. Peter just used his in with/without demos in New York City. Their files were not manipulated. Peter and I spoke, and I'll read what he had to say on a panel at AXPONA.

As for the notion that Meridian (whose DAC I heard them on) would manipulate its DAC so that non-MQA would always sound inferior to MQA, that wouldn't be a very smart marketing strategy, would it? As soon as the fraud was exposed, sales would plummet. Even if it weren't exposed, anyone comparing a Meridian DAC (or DAC from any other company that does MQA) to another while playing non-MQA files would choose the other.

I do believe that conspiracies exist. But really now, this is just getting silly...

crenca's picture

Good to hear that others besides JA have had their own recordings (which they should be intimately familiar with) treated and had a chance to compare.

As far as a demo, it is "silly" to suggest that the outcome is not already predetermined. No company is going to "demo" a product in a situation which shows up the one they are selling you.

Are the Meridian DAC designers trying real hard to optimize PCM over MQA?Of course not - they have explicit instructions not to (and it is "silly" to suggest otherwise). At these demo's, are the presenters going to suddenly break out a competing top shelf DAC that is known to put their product in a light that places even a small question mark on it? No. This is not "conspiracy" this is just normal every day (show) business.

When I can walk into my local dealer, sit down in several listen rooms with a multiplicity of competing products and hear MQA completely independent of their control through multiple competing playback chains, then I will have an idea of how it really sounds. When I can visit the dealer across town and when I can return a week or two later to confirm initial impressions, even better.

prof's picture

...while others will enjoy Red Book quality sound.

is what is known in technical terms as a lie.

An MQA CD encodes the extra data as (what will be read on a non-MQA CD player as) high frequency noise. Rather than the -96DB noise floor that Redbook CD customers are accustomed to expecting from 16/44.1 PCM, they will experience a product with a significantly higher noise floor.

That's not Redbook quality, and should not be marketed as such. Stereophile should not be telling its readers that it is Redbook quality.

(If one can trust their patent application, MQA CDs use the 3 least significant bits to encode the MQA data. That is, on non-MQA CD players, effectively a reduction from 16/44.1 to 13/44.1, raising the noise floor to -78DB. According to this old Stereophile article, that's not only not "Redbook Quality", it's significantly worse than 128bit MP3 quality.)

John Atkinson's picture
prof wrote:
An MQA CD encodes the extra data as (what will be read on a non-MQA CD player as) high frequency noise. Rather than the -96DB noise floor that Redbook CD customers are accustomed to expecting from 16/44.1 PCM, they will experience a product with a significantly higher noise floor.

You can't refer to the effect of an increased noise floor without looking at the spectrum of that noise. When I master a CD, for example, I reduce the bit depth of the 24-bit master to 16 bits using a redithering algorithm that shapes the noise floor. This increases the level of the noise floor close to half the sample rate in order to preserve resolution lower down in frequency. (See figs.4 and 7 at www.stereophile.com/content/meridian-518-digital-audio-processor-measurements.) However, it would be incorrect to claim that the resulting CD file is no longer equivalent to Red Book performance, as this rise in noise is in a region where human hearing is extremely insensitive.

You also have to consider that the original recording's analog noise floor has a spectrum that is somewhere between pink and red in nature. This means that in the midrange and below, the effective resolution will be less than 16 bits. But as the analog noise is still part of the original signal, this is still Red Book performance.

In the case of the first MQA CD, the Japanese Piazzola recording, a quick look at the spectra of the undecoded data and the decoded data appear to be identical below 12kHz. The spectrum of the undecoded MQA data starts to rise above 12kHz, reaching a maximum difference compared with the decoded data of 5dB above 15kHz or so. (This spectrum is very similar to what I show in fig.3 at www.stereophile.com/content/inside-mqa.) But this additional spectral content still lies more than 90dB down from peak level, its presence is very likely benign.

John Atkinson
Editor, Stereophile

crenca's picture

...which is a "sounds like" (i.e. subjective) point of view (backed by your objective data). It largely confirms my own listening experience with MQA (via Tidal) through my own "legacy DACs" (as Bob likes to call them) with music for which I have confidence the same master is being used for the MQA and the PCM (of whatever resolution) is being used. However, the thing is is that MQA fudged on the language here. The problem they have is that PCM is an open, digital standard - it is just software and is defined in a mathematically precise way. The consumer is right to stick to his guns and point to the lie (even if it is a mostly benign, white lie as you argue). On the other hand some believe that it is not all that benign (based on their measurements, interpretations, and listening).

prof's picture

When I master a CD, for example, I reduce the bit depth of the 24-bit master to 16 bits ... However, it would be incorrect to claim that the resulting CD file is no longer equivalent to Red Book performance...

16 bits is Redbook performance. With MQA CDs, we're talking about reducing from 16 bits to 13 (using the three LSBs to encode the MCA data).

13 bits is not Redbook quality.

In the case of the first MQA CD, the Japanese Piazzola recording, a quick look at the spectra of the undecoded data and the decoded data appear to be identical below 12kHz.

I didn't emphasize it in my original post, but let me say so now: the reduction of the bit-depth from 16 to 13 bits is not magically reversed by the MQA decoder. The MQA-decoded signal still has a bit-depth of 13 bits (albeit with some additional high-frequency information).

So, yes, they should be identical (more precisely, the difference should be less than -78DB) across the entire audio band. But they will be identically-inferior to an (in this case hypothetical) 16 bit Redbook CD of the same material.

Jason Victor Serinus's picture

Inquiring minds want to know.

prof's picture

For what it's worth, I am a Professor of Physics at a large public university in the southern US (I'm sure you can figure out which one from my IP address).

Not sure how that advances the discussion, though.

In the interests of doing the latter, let me say that I have nothing against MQA-encoded "24/48 FLAC" files.

There, the MQA data is encoded in the 8 least-significant bits, reducing the bit-depth from 24 to 16. On a non-MQA-aware device, that is identical (where "identical" means the difference is less than -96DB) to 16/48 PCM, which is certainly CD quality. On an MQA-aware device, you get something roughly equivalent to 16/96 PCM.

Trading bit-depth for an effectively higher sample rate may be worthwhile in that case.

But I find it pretty strange that anyone at Stereophile would be advocating reducing the PCM bit-depth from 16 to 13 bits, for the sake of including a (much smaller amount of) ultrasonic data.

But that is what you're advocating (the MQA CD).

harryas's picture

I thank the pro(f)s for analyzing and discussing technical information on MQA in detail. I think this is very important. As a high-end music lover, unfortunately I don´t understand much of that. For me hearing and enjoying music and being touched by the music and its sound are the basics that I rely on. And when I HEAR really palpable differences that´s my criterion. So again: If you don´t want to invest money for testing, try TIDAL "Master" streams even through a Mac with Audirvana 3.0 (for the test periods it´s totally free!!) and compare your HEARING EXPERIENCE to that of other Hi-Res-files.
Offcourse I have different ears than others, but for the experience of music I would like to know what others HEAR (or not) comparing these hi-rez files. ENJOY!

John Atkinson's picture
prof wrote:
I find it pretty strange that anyone at Stereophile would be advocating reducing the PCM bit-depth from 16 to 13 bits, for the sake of including a (much smaller amount of) ultrasonic data.

Again, you can't mention effective bit depth without looking at the spectra of the noise floor. As I mentioned in my post yesterday, if you look at the noise floor analysis in fig.3 of one of my recordings (made in a very quiet hall) at www.stereophile.com/content/inside-mqa, the random analog noise is pink/red in character. This means that it is highest in the bass and lowers as the frequency increases. To use your analogy, this also means the the bit depth for the desired music signal is lower than 16 bits at low frequencies and for the example I show, is still less than 16 bits at the top of the audioband.

Now, if you can replace the LSBs with higher-rez/rate data encoded as random noise - something that Peter Craven and the late Michael Gerzon described in a paper published in the Journal of the AES in 1995 and arrange that the level of resultant pseudo-random noise never rises above the original random noise floor at all frequencies (other than above 15kHz, as mentioned above), you have not compromised the effective resolution of the music. (Peter Craven is one of the progenitors of MQA.)

prof wrote:
But that is what you're advocating (the MQA CD).

I readily admit both that the MQA encoding of 16/44.1k material for issue on a "Red Book" CD involves psychoacoustic sleight of hand - heavily dependent on the spectrum of the recorded music - and that the margin for audible error is much lower compared with a 24-bit wrapper. In those regards I am as skeptical as you. But the proof of whether this can be done does not depend on what I regard as misleading arguments on bit depth but on listening to the results. That is where we at Stereophile are coming from on this subject.

John Atkinson
Editor, Stereophile

NeilS's picture

Speaking as a non-technical reader, whether an undecoded MQA CD sounds as good, better or worse than a non-MQA CD seems to me a subjective issue, where there isn't a "right" answer.

But that seems to me to be not addressing the issue that was originally raised - the statement that "...Music lovers with MQA-enabled DACs will be able to experience the full benefits of MQA-encoding/unfolding, while others will enjoy Red Book quality sound..."

As I understand it, an undecoded MQA CD is objectively not Redbook CD quality. If that is so, the above statement appears to me to be misleading, and should be amended.

John Atkinson's picture
NeilS wrote:
As I understand it, an undecoded MQA CD is objectively not Redbook CD quality. If that is so, the above statement appears to me to be misleading, and should be amended.

In my response above, I explained how with real music signals, if the MQA-encoding has been correctly performed - and yes, that is a big "if" - the undecoded CD will indeed offer the same "Red Book" sound quality.

John Atkinson
Editor, Stereophile

prof's picture

OK. I was trying to keep things "simple", but -- if you insist -- let's take account of the spectral content of the noise.

Let's start with your Redbook CD. If you just took the 24 bit master and truncated to 16 bits (dropping the 8 LSBs), your recording would have quantization noise (the rounding error associated to representing the amplitude of the original audio waveform by a 16 bit integer). The quantization noise would be white (frequency independent) at a level of -96DB.

But you're a smart guy so, instead, you apply dither (essentially, randomly twiddling the 16th LSB). Correctly-applied, this shifts the frequency spectrum of the quantization noise, tilting it towards higher frequencies. The noise at low frequencies (where our ears are most sensitive) goes down, while the noise at high frequencies (where are ears are less sensitive) goes up.

Great!

Now let's prepare that same 24bit master for an MQA CD. First we truncate to 13 bits.

Let's say we now just pad with zeroes. That would be a "Redbook" CD (nominally 16bit PCM), but with a "true" bit-depth of 13 bits. Again, we have quantization noise. If we don't do anything, it's white noise at -78DB. But, as before, we can apply dither: tilting the spectrum of the noise towards high frequencies. The result is a noise spectrum that looks similar to the one of your first CD, but which is uniformly 18DB higher.

Finally, instead of padding with 3 zeroes, we can encode the MQA data in the 3 LSBs. This add an additional source of noise, when played on a non-MQA CD player. You are absolutely right that we do not know the spectral content of that noise (since we are not privy to the details of how the MQA data is encoded in those 3 bits), but let us make the most optimistic assumption possible: that it is mostly high frequency noise, where we are not so sensitive.

Under this assumption, the MQA-encoded CD sounds little-worse than the "13bit, padded with zeroes" CD that we discussed previously. The latter has a (frequency-dependent) noise floor which is uniformly 18DB higher than the "true" 16bit Redbook CD that we are comparing it with.

On an MQA CD player, the 3 bits of MQA data gets "unfolded" into high frequency (above 20KHz) information, but the 13 bit nature of the rest of the recording remains, as does the noise floor: 18DB higher than a Redbook CD.

prof's picture

I readily admit both that the MQA encoding of 16/44.1k material for issue on a "Red Book" CD involves psychoacoustic sleight of hand - heavily dependent on the spectrum of the recorded music - and that the margin for audible error is much lower compared with a 24-bit wrapper.

The proper comparison is not with 24bit PCM, but with 16bit PCM where all 16 bits are used to encode musical data (rather than the 13 bits used in an MQA CD).

Your argument seems to be that with noise-shaping, 13 bits are "good enough" for Stereophile listeners, and there's no audible loss in quality if the 3 least-significant-bits (LSBs) are repurposed for some other use.

I will defer to you as to whether 13bit PCM is "good enough" for Stereophile listeners. I would just dispute calling that "Redbook quality," which it plainly is not.

John Atkinson's picture
prof wrote:
Your argument seems to be that with noise-shaping, 13 bits are "good enough" for Stereophile listeners...

That's not what I wrote at all. Please do not put words in my mouth.

prof wrote:
and there's no audible loss in quality if the 3 least-significant-bits (LSBs) are repurposed for some other use.

No, that's not what I wrote either. What I said was if the recording had an analog noise floor was sufficiently high and of a typical spectral density, then there would room for a hidden data channel - hidden except in the top octave, as you and I have agreed about - that may well be inaudible yet allow the MQA codec to unfold higher sample-rate data.

prof wrote:
I will defer to you as to whether 13bit PCM is "good enough" for Stereophile listeners.

And once again, that is not what I wrote. I said that the only way of determining if an MQA-encoded CD works as claimed is by listening. This is just as need to be performed to assess the effects of other perceptual- and program-based codecs.

Going forward, if you wish me to take you seriously, please stick to what I have actually written and not your baseless conjectures of what you think I meant to say.

John Atkinson
Editor, Stereophile

prof's picture

No, that's not what I wrote either. What I said was if the recording had an analog noise floor was sufficiently high and of a typical spectral density, then there would room for a hidden data channel - hidden except in the top octave, as you and I have agreed about - that may well be inaudible yet allow the MQA codec to unfold higher sample-rate data.

Sorry. It has been unclear heretofore (to me, at least) what your argument was.

Clearly, "if the recording had an analog noise floor [that] was sufficiently high" then the noise added by the digitization (whether at a bit-depth of 16 bits or 13 bits) is irrelevant (except perhaps, as you say, in the top octave).

I certainly agree with that statement. With such source material, there's plenty of place to hide the MQA data.

Whether that's a good tradeoff (given that the MQA data is encoding high-frequency information "missing" from the 44.1kHz sampling, and we've just injected a bunch of extraneous noise into the top octave) is certainly open for debate.

But I assume that we can also agree that not all musical source material has such a high noise floor that 13bit PCM is adequate to capture it (in the sense that the noise-shaped quantization noise is small compared to the pre-existing analogue noise).

16bit PCM was chosen for the Redbook Standard for a reason.

Going forward, if you wish me to take you seriously, please stick to what I have actually written and not your baseless conjectures of what you think I meant to say.

I will attempt to do so. But, I confess to still being befuddled by your argument.

crenca's picture

Which is essentially, that 13 bits is "good enough" for even the most resolving playback chains and that any SQ differences/gains will be found in recording mics/process/playback chains, etc.

While I am open to this perspective, and would even say I have in part "confirmed" it for myself through Tidal playback, I do not see what MQA offers over and above the downside of having to shave things down to 13 bits. I suspect their "de-blurring" claim is marketing mumbo jumbo for repackaged and known filtering (as Koch says) and the whole package of MQA is really just a "soft" DRM Trojan Horse that really does not benefit the consumer at all (whether he is just a "music lover" or an eccentric "audiophile"). Barring MQA being forces upon us by labels (when/if they deny PCM to us) I don't think MQA has any real chance in an open marketplace...

prof's picture

While I am open to this perspective, and would even say I have in part
"confirmed" it for myself through Tidal playback,...

Not really.

The MQA "24/48 FLAC" that you streamed on Tidal encodes the MQA data in the 8 least-significant-bits of 24-bit PCM (reducing the "true" bit-depth to 16 bits). That means that what you were listening to was equivalent to 16/48 PCM (i.e., Redbook CD quality).

Here, we're talking about MQA CDs, where the "true" bit depth is only 13 bits ... a totally different animal.

NeilS's picture

I may be misunderstanding what you wrote, and don't want to belabor the point to your annoyance but I read your "...the undecoded CD will indeed offer the same "Red Book" sound quality..." as a statement about an undecoded MQA CD having acoustic transparency compared to a non MQA CD.

I believe the contention is instead that an undecoded MQA CD does not satisfy "Red Book" CD quality standards, as pointed out by "Miska", "prof", and others.

Anton's picture

This new format stuff is all fun and games....until Neil Young gets involved.

Anybody listening to his Blu Ray music disc set?

Pono?

If Neil gets on board with MQA, I am outta here.

Archimago's picture

Hey guys, maybe I missed a reference here... But where is there any talk of MQA CD being 13 bits + 3 bits encoding?

We'll know when we actually get the data "in hand" and can run it through some DACs. But until then I would just as much suspect that there's nothing here but embedding some MQA code in the 16th LSB and triggering the DAC to upsample with MQA's type of digital filtering while hyping up the filtering as improving time domain performance.

We shall see. Either way, to suggest that the "improvement" would be highly worthwhile seems ridiculous.

prof's picture

That's what they say in their patent:

  • Use 8 LSBs to encode the MQA data in "24/48 FLAC" (as used, e.g., on Tidal).
  • Use 3 LSBs to encode the MQA data on MQA CDs.

What they actually do is anyone's guess, but 3 bits sounds pretty plausible: use fewer than 3 bits and there's not enough room to encode any useful data; use more and you've degraded the sound quality so much that not even true-believers will call it "Redbook quality."

Archimago's picture

Thanks prof.

Will look into this once any of these MQA CD's are released.

I'm going to guess that they don't *dare* do this to the music! Especially with an audiophile recording. Whatever time domain benefit (that in itself appears to be nonsense) would be meaningless in the face of such bit depth reduction.

If they did this, it really would be game over for MQA in the eyes of any self-respecting audiophile who values honesty and for whom the term "high fidelity" still has any meaning whatsoever.

John Atkinson's picture
Archimago wrote:
We'll know when we actually get the data "in hand" and can run it through some DACs.

MQA kindly sent me the MQA-encoded Piazzolla CD. There was about half a second of room tone before the music starts on Track 1. I examined this background noise by playing the CD on an Ayre C-5xe player (no MQA decoding) then feeding the CD data to a Meridian Ultra DAC with MQA decoding.

You can see analyses of the noise floors of the two analog signals output by the two DACs at www.stereophile.com/content/piazzolla-mqa-cd-noisefloor.

You can see that while the undecoded noise floor has a slight excess of energy centered on 20kHz, the spectra of the undecoded and decoded floors match very closely over the entire audio band. So if the MQA encoding is reducing the resolution as proposed by "prof," this loss still lies below the level of the original recording's analog noise at all audio frequencies.

One thing puzzles me about this graph, which is the rise in the noise of the decoded spectrum above 30kHz and peaking at -112dBFS at 70kHz. The A/D converter I used, an Ayre QA-9, has a relatively flat noise floor up to 96kHz - see fig.10 at www.stereophile.com/content/ayre-acoustics-qa-9-usb-ad-converter-measurements - so the Meridian DAC does appear to be responsible for this behavior. More later.

John Atkinson
Editor, Stereophile

prof's picture

You can see that while the undecoded noise floor has a slight excess of energy centered on 20kHz, the spectra of the undecoded and decoded floors match very closely over the entire audio band. So if the MQA encoding is reducing the resolution as proposed by "prof," this loss still lies below the level of the original recording's analog noise at all audio frequencies.

That's not what I said (now who's misquoting whom?).

What I said was that both the undecoded and the decoded MQA CD have a "true" bit-depth of 13 bits.

If the MQA data in the 3 LSBs is mostly high frequency, then the noise floor, through the rest of the audio band, will be identical.

It will also be 18DB higher than it would have been, on a "regular" Redbook CD, where all 16 bits are used to encode the audio signal.

(As I said, above, the "right" comparison is with a version of the recording with a true 16-bit resolution. You could downsample the 24/192 FLAC file, mentioned there, to 16/44.1 and then compare the noise-floor of the resulting file with that of the MQA CD -- either the undecoded, or the decoded version, as through most of the audio band it makes no difference.)

Edit: I will admit, though, that -- below 10kHz -- the noise spectrum that you plot here looks more like what I would expect for room noise than quantization noise.

John Atkinson's picture
prof wrote:
John Atkinson wrote:
You can see that while the undecoded noise floor has a slight excess of energy centered on 20kHz, the spectra of the undecoded and decoded floors match very closely over the entire audio band. So if the MQA encoding is reducing the resolution as proposed by "prof," this loss still lies below the level of the original recording's analog noise at all audio frequencies.

That's not what I said (now who's misquoting whom?).

My apologies, I thought you had clearly said that to accommodate the MQA's hidden data channel, the data on the CD had to be truncated to 13 bits. That is indeed a loss of resolution, as the noisefloor will be 18dB higher than with a 16-bit word length.

prof wrote:
What I said was that both the undecoded and the decoded MQA CD have a "true" bit-depth of 13 bits.

I must have missed that. But you are incorrect.

prof wrote:
If the MQA data in the 3 LSBs is mostly high frequency, then the noise floor, through the rest of the audio band, will be identical.

Now I am confused, as that is in part what I was saying.

prof wrote:
It will also be 18dB higher than it would have been, on a "regular" Redbook CD, where all 16 bits are used to encode the audio signal.

Actually, all 16 bits are used in all "Red Book" CDs. My point, that you don't seem to grasp, is that with all real recordings, the analog noisefloor will be higher than the 16-bit noisefloor, that because the spectrum of the noise will be pink or red in nature, the difference between peak level, 0dBFS, and the level of the noise will be reduced.

You can see this in the graph comparing the spectra of the decoded and undecoded analog noisefloors at www.stereophile.com/content/mqa-cd-noisefloor-repeat. (Note that I have expanded the vertical scale of this graph compared with the earlier posting to make it more clear what is happening.) You can see that other than the rise in the noise close to 20kHz, the spectra overlay exactly, meaning that at lower frequencies, the MQA encoding doesn't compromise the dynamic range/resolution of the original signal.

To address the possible criticism that perhaps the analog noisefloor in this graph is actually due to the MQA encoding, I purchased the 24-bit/48kHz download of the Piazzolla album and examined its noisefloor when decoded. That spectrum is shown in www.stereophile.com/content/piazzolla-decoded-mqa-flac, plotted to the same vertical scale. You can see that below 17kHz or so, this spectrum is identical to that of the MQA CD. (A/D levels were again matched to within 0.1dB.) As the original data had a 24-bit depth and there is no question that the hidden data channel with a 24-bit wrapper lies well below the 16th bit, this graph proves that the analog noisefloor of the MQA CD is that of the original recording.

So having got that out of the way, in my next posting, to come in a day or so, I will add some comments on auditioning this MQA-encoded Piazzolla recording in all its various flavors.

John Atkinson
Editor, Stereophile

prof's picture

My apologies, I thought you had clearly said that to accommodate the MQA's hidden data channel, the data on the CD had to be truncated to 13 bits. That is indeed a loss of resolution, as the noisefloor will be 18dB higher than with a 16-bit word length.

If there is a hidden data channel, then it has to be encoded somewhere. In their patent application, they say it's encoded in the 3 LSBs of an MQA CD (and, FWIW, in the 8 LSBs of an MQA "24/48 FLAC" file).

As you seem to agree, that would raise the noise floor by 18DB relative to a "conventional" Redbook CD, which used all 16 bits for PCM data.

My point, that you don't seem to grasp, is that with all real recordings, the analog noisefloor will be higher than the 16-bit noisefloor...

Actually, I do grasp that. The question on the table was: would it be higher than a 13-bit noise floor?

My best interpretation of your measurements is that the MQA CD is not using the 3 LSBs to encode a hidden data channel.

It looks to me like (per Archimago's suggestion) all 16 bits are being used to encode PCM data and there is no hidden data channel.

(Further evidence for this is that the MQA-decoded spectrum rolls off sharply above 20kHz, exactly like the conventional CD, and totally unlike the 24/192 FLAC file.)

As the original data had a 24-bit depth and there is no question that the hidden data channel with a 24-bit wrapper lies well below the 16th bit...

What is this "24-bit wrapper" of which you speak? It's not on a Redbook CD.

Archimago's picture

Thanks John.

Interesting. So it could very be that they are not embedding MQA data to affect the lower 3 bits in a significant way. That's good but still leaves the question of what benefit this provides ultimately to the sound quality...

Of note is that the MQA decoded output seems to have no significant "unfolding" of ultrasonic frequency material. This speaks to my suspicion that it's just embedding MQA material in the 16th bit LSB with no actual "unfolding".

prof's picture

I would have thought that the fact that this is an MQA CD is something that would be communicated out-of-band (say, in a data track that would be read by an MQA CD player, but ignored by a conventional one).

Their patent application says nothing about how an MQA CD is recognized, but an out-of-band indicator clearly wouldn't work with John's setup (conventional CD player + MQA-enabled DAC).

Edit: Indeed, out-of-band signalling wouldn't work more generally with MQA-enabled DACs (whether the source of the PCM content is an "MQA CD" or an "MQA 24/48 FLAC file"). So, even though they don't say so, presumably an MQA-encoded PCM stream is just recognized by looking for the header at the start of each MQA data frame. If there's no actual MQA data, the rest of the frame could just be filled with the LSBs of the original PCM content.

The fact (for instance) that both traces show a steep drop above 20kHz (indicating that the MQA-enabled DAC is applying a low-pass filter to throw away any ultrasonic information, just as the conventional DAC does) is consistent with the possibility that there is no "MQA data" encoded on this MQA CD.

prof's picture

Dunno how to get a copy of this MQA CD; John Atkinson says he has a copy, but he may have gotten it from the record label.

A 24/192 FLAC version is available. So the thing to do would be to downsample the latter to 16/44.1 and then see how the MQA CD compares with it.

I'm gonna wager: "Not well..."

rt66indierock's picture

Thank you for your comments. If “the margin audible error is much lower compared to a 24 bit wrapper” then there is a problem with mainstream music because there is little to no margin for error in many recordings.

You seem to confirm with your comment that “MQA encoding correctly performed” that the conversion process is more time consuming and error prone than I was lead to believe talking with MQA representatives at T.H.E. Show and RMAF in 2016. And if you have to double check the results with Bob Stuart and Spenser Chrislu as David Chesky did you have a slow process of MQA recording suitable only for niche markets.

But this is all academic at this point because only one of my nine reference albums has been converted to MQA so I can’t test it. And I can’t buy MQA versions of the music I listen to as either downloads or CDs in the United States to just casually listen to MQA. Sorry but TIDAL won’t cut it to compare MQA to other formats.

John Atkinson's picture
rt66indierock wrote:
If “the margin audible error is much lower compared to a 24 bit wrapper” then there is a problem with mainstream music because there is little to no margin for error in many recordings.

To the best of my knowledge, the majority of music today exists as 24-bit integer or 32-bit float master files (not including DSD). Since the "Mastered for iTunes" program started 5 years ago, iTunes required 24-bit files for transcoding to AAC. So the margin for error with MQA masters for streaming is not an issue. The margin issue only raises its head with an MQA-encoded CD.

John Atkinson
Editor, Stereophile

AJ's picture
Quote:

MQA addresses timing anomalies in the A-to-D converter used for the master recording.
-JS

Quote:

To the best of my knowledge, the majority of music today exists as 24-bit integer or 32-bit float master files (not including DSD).
-JA

Yes, I agree. That begs the question what "Timing Anomalies" is MQA "fixing" with 24bit and 96K + encoding?
Another "solution" in search of rumored audiophile "problems"? ;-)
Last I checked, for file size reduction (assuming this is even needed given todays increasing bandwidth), 24/96 FLAC can be done lossless.

rt66indierock's picture

John,

That sounds good but the reality is Apple requires headroom in their Mastered for iTunes submissions because of the clipping present in 24 bit tracks. I agree it should not be there but it is present with a lot of the music they receive.

As for the majority of music today existing as 24-bit integer or 32 bit I’ve heard the opposite but can’t comment further.

John Atkinson's picture
rt66indierock wrote:
John Atkinson wrote:
To the best of my knowledge, the majority of music today exists as 24-bit integer or 32-bit float master files (not including DSD).
As for the majority of music today existing as 24-bit integer or 32 bit I’ve heard the opposite but can’t comment further.

I was writing about modern best practices from conversations with mastering engineers and from my own experience. See, for example, the article on the making of Stereophile's new release at www.stereophile.com/content/tight-lines. But yes, I am sure there are also some that only exist in 16-bit formats. On an AES panel in the 1990s that I attended, Bob Ludwig was complaining about being send DAT tapes for him to prepare the final master. But that was a long time ago.

John Atkinson
Editor, Stereophile

rt66indierock's picture

A couple of thoughts can to mind as I’m preparing my April update of MQA is Vaporware. You are focused on the container and I’m focused on the music itself. Yes you can store music in a big container of 24 or 32 bits but if the music only has a dynamic range of 12 or 14 bits I have to question the necessity of the container size being over 16 bits.

In a bit of irony since I can’t comment myself let me quote Bob Stuart. "Principally because of the combination of environmental noise and microphone self-noise (plus tape noise with analogue masters), very few recordings achieve let alone exceed 16-bit dynamic range. Add to this the fact that we can hear signals within noise only to about 10dB below the noise level (see olive curve in Figure 21) and it follows that bits 19 to 24 carry no useful information."

As far as modern best practices go, I’m not hearing results. It’s still the same people messing up recordings now just as they messed them up in the past. Record companies can make good sounding records they just choose not to.

prof's picture

Yes you can store music in a big container of 24 or 32 bits but if the music only has a dynamic range of 12 or 14 bits I have to question the necessity of the container size being over 16 bits.

I think that's the point of MQA's "24/48" FLAC format: use the "superfluous" 8 bits of the 24bit container to encode something useful (could be a treasure map; could be some encoded version of the ultrasonic content that would have been available in a higher sample-rate file).

If (as on a CD) your bit depth is already "only" 16 bits, then you can't play that trick (without a loss in quality).

Unless, of course, you have a "hidden data channel with a 24-bit wrapper [which] lies well below the 16th bit".

I'm not sure what that's supposed to mean, but it sure sounds cool.

michaelavorgna's picture

Tallying up the comments on these MQA CDs, what we are left with is the possibility that the CD-layer may sound the same, better, or worse than the CD while the MQA-layer may sound the same, better, or worse than the high-res download.

Hmm. How can we possibly resolve this perplexing conundrum for ourselves? I see a few options; a) continue to comment and read other comments about the technology, b) listen, or c) ignore the entire thing.

Of course only one of these options will actually answer the relevant question for potential buyers while the other two get you the same result.

Michael Lavorgna
Editor, AudioStream

AJ's picture
Quote:

b) listen

Ah, you're advocating trust your ears only listening, aka blind testing. Like the ones used by major orchestras to select players based solely by b) listen. Bravo!

Now of course, if one has poor and/or no trust in their ears, then d) "Experience" might be the more appropriate option. That is, one determines by looking, knowing, believing, etc, etc + listening.
Under that scenario, I'm sure many audiophiles will opt for the MQA experience!!

michaelavorgna's picture

If one has poor and/or no trust in their ears, I recommend investing in another hobby.

AJ's picture
Quote:

If one has poor and/or no trust in their ears, I recommend investing in another hobby.

Or stay far away from good/trust your ears b) listen (aka blind) tests.
Where one must trust their ears, not eyes, beliefs, manufacturers claims about "blurring", etc, etc, etc

michaelavorgna's picture

...to the brain, last I checked ;-)

AJ's picture
Quote:

...to the brain, last I checked ;-)

Yep, along with those eyes, reading all this purported "blurring" and whatnot.
Hence d) Experience.
As I said earlier Michael, if some audiophile prefers their "listening" experience spiced up with a nice little dose of anharmonic aliasing distortion and HF EQing and imagine this was somehow the "artists intent" 10, 20 or 30 years ago...well so be it. We all have our preferences ;-).

michaelavorgna's picture

Here's a little something from our "About Us" page on AudioStream:

Here's our approach to the hi-fi hobby in a nutshell:

The experience of enjoying listening to music is the ultimate goal of our hi-fi hobby.

prof's picture

...we are left with is the possibility that the CD-layer may sound the same, better, or worse than the CD while the MQA-layer may sound the same, better, or worse...

What is this "MQA layer" of which you speak?

michaelavorgna's picture

.

prof's picture

Ah! I see. Like "24bit" is a metaphor for "hi-rez".

michaelavorgna's picture

...until somebody gets hurt ;-)

low2midhifi's picture

Our friends from north of the border have their own assessment of this format. I find this assessment to be a fair one. I'll leave it to the reader to make up his/her mind, but I encourage the interested to read on:

http://www.soundstagehifi.com/index.php/opinion/1057-mqa-one-year-later-...

I hope that the debate does not lead to outright hostilities in Chicago/Rosemont next week. A few years back a Flac -vs- Apple Lossless ("tastes great"/"less filling") debate almost led to such an outcome in a dealer seminar that I attended.

Maybe the sponsors of next week's show can set up a boxing ring, or arm-wrestling table, if this debate is not settled through people's ears. Last year it was apparent, from the demeanor of some of the assembled grandees, that there were some discussions that got quite heated before the show opened up to the public.

On the subject of our friends north of the border, I'm hoping there won't be a presentation (MQA or otherwise) of a certain musical performer's new album in every other room next week. I also hope that others, usually up on the 12th floor, will give "Stimela" (aka "Coal Train") a much needed rest. The "mish-mash food on that iron plate" needs a break, and from all versions of recorded media.

Let's just agree on the Cubs (or Sox), Deep Dish Pizza, and the fantastic proximity of O'Hare Field to the show's venue next week and all will be fine at Axpona 2017 ;).

michaelavorgna's picture

...of an audio technology without listening to it is a neat trick.

johnnyangel's picture

The cheerleading for MQA has left a bad taste in the mouth of this longtime Stereophile reader. But for the sake of argument, let's accept that it has a valid role in improving the quality of streamed audio.

That said, no one seems to be asking what the point is of issuing a MQA-encoded CD. Is there some claim that this could equal the quality of an SACD or Blu-Ray audio disc?

I have a Parasound CD-1 that plays Redbook CDs quite nicely, and an Oppo player that does quite well for SACDs. Tell me about something that will actually advance the state of the art and I might be interested, but I don't see why this is better than what I have.

John Atkinson's picture
johnnyangel wrote:
The cheerleading for MQA has left a bad taste in the mouth of this longtime Stereophile reader.

I don't believe we have been "cheerleading" for MQA. We have been reporting on the technology employed by this codec, commenting favorably on what we have heard in careful comparisons, and, as in this thread, trying to dispel the misinformation being spread by others, some of whom are not as disinterested as might be thought.

johnnyangel wrote:
That said, no one seems to be asking what the point is of issuing a MQA-encoded CD.

The promise, that has yet to be substantiated through listening, is that without compromising the CD sound quality, an MQA-encoded CD when played through an MQA-capable DAC will offer the benefits of a high sample-rate recording.

John Atkinson
Editor, Stereophile

AJ's picture
Quote:

an MQA-encoded CD when played through an MQA-capable DAC will offer the benefits of a high sample-rate recording.

Could you elaborate on what these benefits might entail for consumers?

johnnyangel's picture

Thanks for the response, John. I think you missed my point, though, which is that we can already purchase high-sample rate recordings on optical discs via SACDs (many available for a lot less than $18.98 these days) or Blu-Ray audio discs. And, a lot more people have SACD or Blu-Ray decks than have MQA-capable DACs. So if I'm Chesky, what's the point of this development?

Re-reading Jason's article, I see Chesky claims that an MQA-decoded CD "sounds a little more analog, relaxed, and organic sounding than other digital issues." Better than SACD or Blu-Ray then?

Anton's picture

I am eager to see the replies to your opinion/questions!

AJ's picture
Quote:

what the point is of issuing a MQA-encoded CD. Is there some claim that this could equal the quality of an SACD or Blu-Ray audio disc?

Flashbacks of HDCD come to mind ;-)
(and all the similar raves about those, in yesteryear)