This article originally appeared in the July/August 2018 issue of WiFi HiFi Magazine.
Has any subject created more controversy among audiophiles than MQA? Developed by Meridian Audio and announced in late 2014, MQA purports to deliver master-quality audio in files that aren't much larger than CD-resolution. Not just CD-quality, not just high-resolution, but audio that matches, as closely as possible, what the microphones picked up at the recording venue.
Since the announcement, MQA's merits have been hotly debated on Internet audio forums. Many knowledgeable commentators have written detailed articles on MQA, analyzing the output of MQA-equipped DACs, and making a strong case that the system is flawed.
One of the best of these analyses was posted in early March on Computer Audiophile by a blogger who calls himself Archimago. In his essay, Archimago links to several other analyses. These articles are well worth reading by anyone interested in high-resolution digital audio. This is the Internet at its best. For a pro-MQA view, it's worth reading a series of columns in Stereophile by Jim Austin.
We've also witnessed the worst of the Internet in this debate, with several posters indulging in character assassination and insult. Some commentators have dismissed their opponents as shills, and bragged about driving them off forum debates. In one memorable exchange, two posters suggested meeting up at an audio show to settle their differences like men!
We see the same behaviour across the Internet, and it's poisoning public discourse throughout the west.
Canadians pride themselves on being the world's most polite people, so if there is anyone who can have a civilized debate about MQA, it's us. In late May, contributor David Susilo discussed his views on MQA in an article on WiFiHiFi.ca. That article is reprinted here, followed by a response by Editor-at-Large Gordon Brockhouse, whose views on MQA have generally been positive. We conclude with David's response to Gordon's comments.
David's Take on MQA
About five years ago, Meridian released MQA (Master Quality Authenticated), promising that when you listen to audio in that format, it would be the absolute best, true-to-life version of the recording. But has it lived up to the hype? I have my reservations.
Way back then when I was given the chance to listen to MQA (A/B comparison), the difference was quite big, with one caveat. The MQA file I listened to was compared to the same song compressed in MP3 at, if I remember correctly, 160kbps. Every now and then, I've come across new MQA demos. But none have been a direct comparison with a non-MQA file of the same bit depth and sample rate, never been level-matched, or never been double-blind tested. From my personal A/B testing (level-matched to 0.25dB accuracy), they sound different. Not better, not worse, but different.
Further, in my opinion, the entire concept of MQA doesn't really mean it is the best of the best. Let's say I produce a song, and I don't want to sell the best quality now, but I want an MQA logo. I can do that simply by going through the MQA process, which includes compressing the music file into MQA and, of course, paying a fee. Then, when the song is played through an MQA-enabled unit, the listener will see the MQA light on and expect, or at least think, that they're listening to the best quality of the recording. As Bob Stuart from MQA told the music industry, "MQA is saving your crown jewels" (read: original untouched master), while at the same time telling the consumer that MQA is better than the original master.
How can it be? How can a file, just because it has MQA certification, be better than the original? That can only happen if something has been changed from the original.
Consider another scenario: I mix my own song and someone else masters the mix, which is common for song releases in different countries. Obviously they will sound different. What if one person has the budget to make the song MQA-certified and the other doesn't? Then even though I'm the creator of the song, the other person can pay to have their version be MQA.
This has been the case for many album releases in different countries. For example, Def Leppard's Hysteria U.S. vinyl release is the MQA version, while the German vinyl is considered the better/preferred one of the recording, or the "hi-res" option.
Another example is Warner's release of the Charice CD, which carries the same name both in the U.S. and the Philippines. Both were released at the same time, and mixed by the same studio, and the same person. One version is geared towards the Southeast Asian market, and the other for the North American market. Which one should Warner choose to be the master for MQA? There can't be two MQA files (conceptually) for a single album.
According to a list on Meridian Unplugged, more than 11,000 MQA-encoded albums are available on TIDAL's Masters tier. And MQA says more than a million tracks have been encoded in MQA.
Out of the approximately 800 albums released in MQA, logically they are questionable in nature when it comes to the "master quality" part. I don't think ZZ Top, Black Sabbath, Bjork, Bread, B-52, Chicago, Faith No More, INXS, and Iron Maiden were ever involved in high-quality recording, let alone the MQA process.
From the technological side, I find MQA to be flawed. More than 15 versions of Miles Davis' Kind Of Blue have been released over the years, with each release better than the one before, and the original master released in stereo and mono (just like the early years of The Beatles' albums), both mixed differently. Even when you look at the master itself, which one is the correct master?
Also, every subsequent release has been marketed as the best version of the recording. So this year, they claim the version to be the best and pay for the MQA certification, and three years from now, they release a sonically better version of it but don't want to pay for the MQA certificate anymore. Then the non-MQA version will sound better than the MQA one, both technically and subjectively. Is there a method to pull the MQA "process" out of the older version and move it over to the newer version? No! Why? Because MQA compression also implements MQA filtering. And that's where I also have an issue.
To simplify things, you have a pristine master, approved by the songwriter, performers all the way to the engineers and the studio itself, and you run it through a filter. How can that be master quality anymore? The untouched version will be the master, the MQA file will be the filtered version.
Then, there is the MQA system flaw. MQA is not, in my opinion, lossless. MQA claims it uses "neuroscience" to achieve the efficient compression technique, and says the processing is lossless. But a quick look at the methodology through the 2013 U.S. Patent Office filing clearly show that the files will be truncated before being streamed and unpacked (as opposed to being compressed and unpacked). Just like MP3, once you truncate the original file, that part of the file is gone forever, regardless the amount of restructuring on the other side. And also just like MP3, MQA compresses music by truncating certain components of sound through perceptual encoding.
Interestingly, at Munich High End 2018, which took place earlier this month, MQA and all of its partners seemed to have been keeping a low profile. Even after many companies displayed the MQA logo at their booths, most presentations were still offered using vinyl or CD, regardless of the music style.
So, is MQA a bad format? No, that isn't what I'm saying. It's much better than MP3, and it sounds very close to lossless. In fact, it's the best lossy compressed format out there to date. If you prefer the sound of MQA-encoded files, by all means, go buy them. But don't think that it is, in every case, the most authentic and measurably best version of the recording you buy.
To paraphrase someone from MQA: when the lossless version of a song in the MQA format is released and you already have the same song in FLAC with the same bit depth and sampling rate (but without the MQA logo), at best, you will get identical sound quality.
To compare high-res FLAC and MQA versions of the same music for this article, Gordon used his desktop system, which includes an iFi micro iDSD Black Label DAC/headphone amp (shown here) driving HiFiMAN Edition X V2 planar magnetic headphones.
Gordon's Take on MQA
Regular readers of WiFi HiFi will know that I've been a proponent of MQA since its announcement in 2014; and in particular, since TIDAL began streaming MQA-encoded high-res audio in early 2017. I have followed the forum debates about MQA with great interest, and some dismay.
I want to comment on some of David's points, and address some inaccuracies. He says "about 800 albums have been released in MQA." In fact, more than 11,000 MQA-encoded albums are available on TIDAL's Masters tier, according to a list on Meridian Unplugged. And, as MQA CEO Mike Jbara notes in the accompanying interview, over a million tracks have been encoded in MQA.
In his conclusion, David advises, "If you prefer the sound of MQA-encoded files, by all means, go buy them." In fact, few of us are unlikely ever to buy MQA-encoded music. While there are a handful of MQA-encoded downloads and CDs available, MQA is almost wholly about streaming.
Noting that multiple remastered versions of popluar back-catalog albums have been offered over the years, David wonders how listeners can be confident that an MQA version is in fact definitive. This is a perennial problem, dating back to the beginning of digital. Some reissues were made following important advances (e.g. adding dither to mask low-level noise and distortion, remastering at higher resolution to allow for gentler A-to-D filtering), and were genuine improvements on earlier versions. There have also been reissues where dynamic compression was added so they'd sound louder, which just sucked the life out of the music.
MQA says it considers provenance a key issue, and wants labels to use their highest-quality masters for MQA versions. While there have been exceptions (I've heard of a few MQA releases with 16/44.1 content hiding in 24/96 containers), in my experience, labels seem to be using their best digital assets for their MQA releases.
David worries that content creators might get locked into MQA. If a label has released an MQA version of an album, and in the future wants to release a new non-MQA version, there's no way to "pull the MQA ‘process' out of the older version and move it over to the newer version," David writes. For music recorded without MQA (the entire back catalog), this is a non-issue. It's like saying an album is irretrievably compromised when a label releases an AAC-encoded version on iTunes. ‘Tain't so of course: the label still has the original assets. In David's scenario, if a label wants to release a new improved non-MQA version of an album, it just pulls out a pre-MQA mix and uses that for the remaster.
Most MQA albums are encoded using automated tools from the best digital master available. But a few, such as the 50th Anniversary Deluxe Edition of Waiting for the Sun by The Doors, receive "white glove treatment." Bob Botnick, The Doors' longtime engineer and mixer, went back to the original analog masters to create this new version, which will be available September 4 on an LP/MQA-encoded CD set, as well as on digital download and streaming services.
For new releases, labels can use their current workflow, and apply MQA encoding at the end of the process, in which case they can retain pre-MQA versions. Studio hardware with built-in MQA processing will allow for new recording projects that are MQA-encoded from the get-go; using that option, MQA will presumably be locked-in. Here, it's worth noting that the big three labels are all MQA shareholders.
Like many observers, David laments MQA's failure to conduct A/B comparisons between MQA and non-MQA versions of the same music during public demonstrations. I too wished for A/B demos in the year following the announcement of MQA. But at events like CES, all MQA and its hardware licensees were doing was playing MQA-encoded music.
This wasn't a case of MQA being evasive, Jbara said during a presentation in Toronto in late June. "Whether you're listening to an MP3, Red Book audio or even high-res, it actually takes a minute for your brain to adjust to the quality of the sound," he explained. "When you're doing A/B testing, you should never listen to a track for less than a minute and a half. Toggling back and forth is really a flawed experiment, because that's not how the brain reacts."
That may sound like a dodge, but it matches my experience. You need to be in a receptive state to make judgements about sound quality. That's difficult if you're listening through unfamiliar equipment in an unfamiliar setting. (I think listening to music is a bit like sex. If you're thinking too much, you're probably not having a very good time!)
In any event, it's now dead-easy to compare MQA-encoded high-res streams from TIDAL with downloads of the same resolution. And when Qobuz launches its streaming service in North America later this year, it will be easy to compare MQA-encoded streams from TIDAL with FLAC streams of the same resolution from Qobuz. As I outlined in last issue's feature on streaming audio, I've had a demo Qobuz account since CES, so I've had ample opportunity to do this.
For this article, I did some more comparisons, mostly through headphones. I compiled playlists of the same high-res tracks from Qobuz (FLAC), TIDAL (MQA), and in my own library (ALAC and FLAC downloads from HDTracks, Pro Studio Masters and other sources), and then did some comparative listening through my desktop headphone setup: Audirvana Plus 3.2.8 software running on a Mac mini, connected via USB to an iFi Micro iDSD Black Label DAC/headphone amplifier powering HiFiMan Edition X V2 magnetic planar headphones. Following the latest firmware update, the iFi Micro iDSD has MQA rendering capability. Audirvana supports playback of TIDAL and Qobuz streams, as well as local content, so I was able to use the same interface and audio engine for everything.
While David S. has his doubts about MQA, he prefers the MQA version to the 1983 CD release.
Were the files used for these comparisons from the same master? It's an interesting question in theory, and one frequently raised in forum discussions about MQA. But my interests are more practical: where can I find the best-sounding version of the music I want to hear?
Moreover, as Jbara confirms in the accompanying interview, the high-res FLAC downloads being offered by HDTracks, Pro Studio Masters and Qobuz are almost all based on the same digital assets as the MQA versions being streamed on TIDAL, allowing for true apples-versus-apples comparisons. Here are some of my impressions:
"Stairway to Heaven" from Led Zeppelin IV (Remastered): Compared to Qobuz's 24/96 FLAC stream, the 24/96 MQA stream on TIDAL sounded rounder and more embodied; but the Qobuz version sounded sharper and more incisive. The TIDAL version had a greater sense of space; the Qobuz version sounded a little drier. While the difference wasn't huge, I preferred the TIDAL version.
"Don't Know Why" from Come Away With Me by Norah Jones: The 24/192 FLAC stream from Qobuz and 24/192 MQA stream from TIDAL sounded practically indistinguishable. The TIDAL stream seemed a tiny bit more open and organic. I had no real preference.
Shostakovich Symphony No. 5, Boston Symphony Orchestra, Andris Nelsons, cond.: Qobuz's 24/96 FLAC stream and TIDAL's 24/96 MQA stream of this wonderful live performance from Symphony Hall in Boston sounded almost identical. The Qobuz stream seemed ever-so-slightly clearer but drier, and the TIDAL stream ever-so-slightly warmer and more open; but really, this was a situation where any comparison is pointless.
"Starlings" from Break Stuff by the Vijay Iyer Trio: Comparing Qobuz's 24/96 FLAC stream and TIDAL's 24/96 MQA stream of this 2017 jazz album, the decay of Iyer's piano seemed slightly more extended and natural on TIDAL, but instrumental attacks (Iyer's piano, Stephan Crump's double bass and Marcus Gilmore's drums) were a bit crisper on Qobuz. Overall, the TIDAL version sounded a little more organic and spacious.
These impressions match my earlier experiences, outlined in my feature on music streaming in our April issue. As I noted there, whatever sonic advantages TIDAL has, they're not enough to offset Qobuz's vastly greater (and to me, more interesting) catalog of high-res music, and its far superior interface.
I suspect my assessment isn't a whole lot different from David's. Based on A/B level-matched listening, David concludes that MQA files "sound different" from FLAC files of the same resolution. "Not better, not worse, but different."
But that raises a thorny question for David. If you run a "pristine master," approved by the artists and engineers involved in the recording, through an MQA encoder, how can the MQA version be said to be "master quality?" As David remarks, "The untouched version will be the master, the MQA file will be the filtered version."
MQA's response would be that the digital master isn't as pristine as you might think. It has been compromised by anti-aliasing filtering applied during A-to-D conversion, which can cause unnatural ringing artifacts that disrupt timing cues. So even a "pristine" digital master will sound different from the output of the recording console.
MQA's largest hardware licensee is Lenbrook, whose Bluesound products perform MQA decoding. Compared to software decoding, with hardware rendering, the MQA file tells the DAC how it wants to be played.
MQA says its goal is to provide an end-to-end lossless experience. During encoding of existing digital masters, MQA applies a secondary filter that seeks to reduce those artifacts. On new MQA recording projects (or conversions from analog masters), the encoder will use filters that minimize filtering artifacts from the get-go. When an MQA-capable DAC is used during playback, the MQA decoder applies a complementary reconstruction filter tailored for the listener's DAC.
Here we come to the crux of the debate. Critics maintain that MQA's filters have their own artifacts, and that whatever differences people hear between a conventional digital file (FLAC or other lossless format) and an MQA version of the same music are the results of distortion (euphonic or otherwise) caused by MQA's filtering. These effects can be duplicated with well-known digital filtering techniques, they say.
Critics also insist that the MQA process isn't lossless. The most vociferous critics view MQA as a scheme by Bob Stuart et al. to hijack the music and audio industries, and extract a steady stream of royalties.
My view is that MQA is what its backers say it is: an encoding technology that addresses some of the drawbacks of conventional digital recording. What's interesting to me is that when I stop trying to identify differences, and just listen for enjoyment, the MQA streams have an added feeling of solidity and centredness.
It's now 35 years since the launch of the CD. Digital recording has steadily improved over that span. So when we compare FLAC and MQA streams of well recorded content, we can expect to hear only slight differences, if we hear any at all. But the differences I hear are the ones MQA purports to deliver. It's not night-and-day, but it is worthwhile.
David's Take on Gordon's Take
After I posted my views on MQA on WiFiHiFi.ca, Gordon invited me to participate in a printed discussion of the format. I thought this was a great idea, as we both have a similar understanding of how MQA works, but different opinions on its merits and shortcomings.
First of all, I have to apologize for misstating the number of MQA albums out there. I did not realize that there are 11,000 MQA albums on TIDAL. Not until Gordon mentioned this did I realize I was viewing TIDAL's Master listings incorrectly, hence my erroneous count of 800 albums.
After browsing TIDAL for four hours, I could find only a couple of hundred albums that I'd want to listen to. While I'll experiment and explore on a streaming service like TIDAL, like most audiophiles, I'm a collector at heart; so finding only 300 albums I want in my collection makes it seem like MQA is going more for quantity than quality.
Universal Music Japan has released 100 MQA-encoded Audio CDs; and I find this a more attractive option than streaming and downloads. When I heard about this, I tried to buy all available MQA-encoded audio CDs so I could do a direct A/B comparison with my CD collection. I know this is an unusual approach, but it works for me. But out of those 100 discs, there are only three with music that I have in my library; and that includes a sampler CD. So these first releases are not great titles to begin with.
Making things worse, my confidential source in Japan told me that Sony has no plans to release MQA-encoded CDs. If Sony is not going to be in the game, about 30% of the world's music catalog will not be available in MQA format on physical media. Not a good indicator for a codec that was announced almost four years ago (especially for me, as about 60% of my CD collection is from Sony).
As much as I agree with Gordon about how labels will use their digital masters for MQA releases, it still can't change the reality that what is considered the "best" today may not be considered "best" three years from now. I also agree that the studio will not (or at least, should not) throw away their multi-track master just because they have created an MQA mix. I still have the multi-track tapes and hard drives of all my recordings, should one day I want to remix or remaster them; but that's not the issue here.
The issue is MQA's claim to offer the best quality possible. What if today's release is mastered in MQA, and three years from now there is another remaster of the same song that is considered the best at that time. What is it going to be called? "MQA Again?" "MQA the sequel"? Won't it be confusing owning two MQA files of the same song? It's easy for streaming, as the streaming service can just substitute the new file for the old one. But how are the labels that are selling MQA files (as downloads or physical media) going to handle this problem? I thought MQA promises the best master available? If so, then MQA should have a way to recall and destroy the previous MQA file. If not, then the very idea of MQA is flawed.
I agree with Gordon if the master was taken from two-channel analog, converted to digital, remastered in digital, then completed in MQA, the system may perform as promised. But even that is questionable, as MQA does not know exactly what type of ADC was originally used in the studio. Depending on the analog circuitry, ADC chip and other variables, the characteristics of the anti-aliasing filter will differ, so ringing artifacts can't be fixed with a one-size-fits-all MQA filter.
Remastering can be also done on the original multi-track, whose sources can include both analog and digital tracks. But sometimes even analog tracks have undergone A-to-D conversion for adding digital effects, then turned back to analog to be recorded onto an analog multi-track tape. Sometimes all the tracks were recorded in analog, with different layers of digital and analog effects inserted into the track before creating the two-channel mixdown (which then can be recorded in analog or digital). Sometimes even the two-channel digital master may be turned back to analog to add warmth, (this was quite popular in the late ‘80s and early ‘90s). All these variables make the claim that an MQA filter can fix anti-aliasing artifacts laughable.
At the end of the day, as listeners, we go back to the choices on offer. If you prefer the MQA version, go with it. Just don't think it's the end-all-be-all solution MQA claims it to be. Now let me get back to the MQA version of Mike Oldfield's Tubular Bells, which I love even better than the 1983 CD release.
For more on MQA, check out our one-on-one interview with MQA, Ltd. CEO Mike Jbara.
Photo at top by David Susilo