Over-exposure can ruin a perfectly good song. Blur’s “Song 2”, Pulp’s “Disco 2000”, Lorde’s “Royals”, Bob Dylan’s “Mr Tambourine Man” and The Doors’ “Riders On The Storm” have been worn down the nub for this fan. Hearing these songs now, they sound like karaoke ghosts – mere shadows of their former selves. That’s partly the result of DJ-ing and partly the result of since becoming an audio journo. There’s a case to be made for not reviewing audio gear with songs that you love.
So it was with some gratitude that the MQA-encoded material hitting my Dropbox last Friday night, along with the corresponding hi-res originals, featured nothing that I even liked, which in turn meant no treasure could be ruined by the intensive A/B comparisons that would follow.
Steely Dan’s “Babylon Sisters” aside, songs like Muddy Waters’ “My Home Is In The Delta” and Ella Fitzgerald & Louis Armstrong’s “They Can’t Take That Away From Me” never, ever normally get a run at DAR HQ. Ditto Ravel’s “String Quartet in F Major 2 Assez Vif, Tres Rythme” (as performed by the Guaneri String Quartet) or Debussy’s “Prelude” (as performed by Joan Rowland).
You might point to my lack of familiarity with these pieces as a shortcoming in assessing MQA (the format) or you might consider my lack of exposure as fundamental to impartiality on what has proven to be the single most controversial topic in the audiophile world this side of Ethernet cables.
A dearth of MQA content, easily dwarfed by the fanfare surrounding Bob Stuart and Peter Craven’s new music format, has all the hallmarks of early years DSD. If anything, this had me kicking of my own MQA listening sessions on the back foot and a little skeptical.
MQA being better, the same or worse wouldn’t matter a jot if the music itself didn’t float my boat. As someone who listens to music that inhabits the netherworld between audiophile approved and chart hits (think: Low to Tom Waits to Moderat), why would I care how MQA sounded if it only mattered to the same old audiophile releases? Does the world really need another version of A Kind Of Blue? Or Hotel California? Or Jazz At The Pawnshop?
Just as reviewers trust DAC manufacturers not to juice sample units with better capacitors and power supply components, these files were accepted on similar terms; that the MQA versions hadn’t been messed with beyond that which would normally take place during the MQA encoding process.
Here’s MQA’s Bob Stuart: “The MQA files were encoded directly from the original files included in the folders. Furthermore, there is no change in level or frequency response in the audio band (MQA is typically within 10mB).”
My review files were clearly marked as loaners and I was happy to delete them all once the A/B comparisons had run their course.
Well, except perhaps “Babylon Sisters” by Steely Dan – song I’d enjoyed to a thousand times before signing on to write about audio and ten thousand times since. It’s one of the few audio demo staples that I (still) truly love.
Not just the song itself but also its host album, Gaucho – production polish be damned, the subject matter tackled by Donald Fagen’s lyrics is unsettling and, at times, a little bleak. A mid-life crisis record that wouldn’t sound out of place piped through a shopping mall PA system.
In my digital audio library sits Gaucho as original CD rip, remastered CD rip and as (remastered) hi-res download.
Why then would I want to keep hold of Gaucho’s first cut in MQA?
Put simply: because it sounds convincingly better than the normal, non-MQA’d 24bit/96kHz file. Lest you thought MQA was just a way to pack, transmit and then unpack hi-res audio via what Bob Stuart calls “audio origami” or “encapsulation”, it isn’t. It’s that. But also more.
We’ll get to the techier SQ-related stuff in a bit. First, audio origami.
MQA can fold a vanilla 24bit/384kHz, 24bit/192kHz or 24bit/96kHz file down inside its 24bit/48kHz container file. Similarly, 24bit/352.8kHz, 24bit/176.4kHz and 24bit/88.2kHz files can be folded up inside a 24bit/44.1kHz MQA container file. In both cases, song.flac becomes song.mqa.flac where FLAC is the file wrapper – this is crucial for backwards compatibility with DACs that aren’t MQA enabled. More on that shortly.
MQA makes hi-res streaming a reality when previously it mightn’t have been. But if MQA (the company) were only concerned with lowering the bandwidth required for hi-res audio streaming, once unpacked, an MQA file and its hi-res cousin would sound (close to) identical. But do they? No, they do not. And from an audiophile perspective, the audible differences aren’t subtle.
When fired from Roon server running on an Intel NUC, over the LAN to a Sonicorbiter SE, via a Curious USB cable and iFi iPurifier 2, into the Mytek Brooklyn DAC (whose front panel MQA logo lights up blue) and then out to a pair of Red Dragon S500 monoblocks driving Spatial Hologram M4 Turbo S open baffle loudspeakers, I found the MQA take on Steely Dan to be more musically convincing than its hi-res counterpart.
This isn’t just a spasm.
One thing I like about MQA is that it puts provenance ahead of sample rate. The A in MQA stands for “authenticated”. Like the “Babylon Sisters”, the Mytek Brooklyn unpacks the Fitzgerald / Armstrong MQA cut to 24bit/96kHz and also turns the front panel MQA logo blue.
Logo colour is not an indicator of sample rate or even audio quality. The green light seen during Muddy Waters’ track indicates MQA content – the first sign of an untampered file – but blue (per the Steely Dan track) takes information provision a step further in letting us know that the MQA file has been approved at the studio level, signed off by either the artist, producer or copyright holder
At the other end of the chain, MQA meets the end user and his/her DAC, inside which the MQA code (software) sits on an XMOS chip. Its job is not only to unfold (decode) the hi-res according to the sample rate limits of the DAC chip but also to optimise the DAC chip’s sound quality by applying a pre-emptive corrective filter to the digital signal prior to conversion to analogue (rendering). This the main reason why the MQA code must sit within I2S reach of the DAC chip – i.e inside the DAC box – and why its effectiveness would be diluted/nullified if placed in an upstream streamer.
Bel Canto’s John Stronczer explains it like this:
“Why embed MQA into the DAC? There are a couple of reasons, using the spline filters you cannot have an apodising filter before or after or else you will compromise the phase and frequency response. You also need to take care with the analog domain filters and ideally any filters after the MQA decoder need to be measured and the MQA filters optimized to take them into account. This will lead to the BEST performance, however I do suspect that a software only implementation could indeed work well but not necessarily at its best, and in the worst case it could actually lead to excessive phase and frequency response errors in some DAC and system applications, so it DOES actually make some sense to control the process in an approved platform and optimize each companies’ or even each products’ filters for best performance, further differentiating any MQA product from another. In other words a specific licensed spline filter will be unique to each design and company.”
As mentioned by Stronczer, the pre-emptive, corrective filter is customised by MQA (the company) according to the fingerprint of the DAC chip, its own filters and (sometimes) the D/A converter’s output stage as measured by MQA. Note that Stronczer also suggests that a software decoder is possible but also probably less effective – a conversation for another day.
One implication for DAC manufacturers signing up for MQA certification is that Bob Stuart’s team will need to recalibrate the filter each time a change is made to the DAC design – e.g. a move from Wolfson to ESS chips means fresh MQA code will be required. FPGA DAC developers whose changes come thick and fast via software updates might see this scenario as less than ideal.
Back to Ella and Louis. The MQA version delivers fuller, more tonally satisfying bass notes and a better sense of the space surrounding Fitzgerald’s voice. Her singing sounds altogether more real, more present and fleshier with MQA than not; there’s more audible evidence of the woman behind it.
Remember: MQA playback is a two part process in a suitably-equipped DAC: 1) hi-res unpacking (decoding) and 2) pre-emptive filter application and decoding (rendering). I’m willing to bet the latter is why I hear better sound quality from an MQA file than the hi-res original. MQA seems to do a better job of making the three dimensional illusion that is high end audio playback more believable.
Nowhere was this more evident than on the aforementioned Debussy piece where audiophile pal Barry, who, more familiar with the piece, had joined me for an afternoon’s listening. He picked up immediately on the more notable decay of the piano’s upper registers and that the MQA file spilled with more recording space information than the hi-res equivalent.
With Muddy Waters’ “My Home Is In The Delta”- which Barry tells me is a well-worn audiophile number – the drum, panned hard left, displayed greater textural detail in MQA form than it did via 24bit/192kHz. The slide guitar sounded more real. Moving in the other direction, from MQA to hi-res, the drum skin lacked texture and reverb, the slide guitar displayed more metallic sheen and the overall presentation sounded comparatively lifeless.
Time to punch in Ravel on the Roon remote. Again, I heard superior separation from the MQA version but also a greater sense of musical flow from string players. My notes read “Oozing class”. This A/B unmasked the widest delta between the kind of hi-res audio file that we’re used to hauling around and the lighter, leaner MQA version.
In more general terms, and depending on the musical content, MQA draws player outlines with greater clarity and fleshier tonality. More easily ‘seen’ is each player’s soundstage positioning. Decay, from percussive ticks to guitar licks, becomes more obvious too.
But if the MQA versions share a common trait, one that isn’t as apparent in the standard hi-res PCM files, it’s a greater sense of ease. The kind that vinyl lovers talk about when they complain that digital playback sounds too brittle or too rigid. This would seem to fly in the face of Benchmark’s “bottom line” assertion that “MQA is no substitute for listening to the original high-resolution files.”
This commentator’s findings don’t mean the Brooklyn DAC playing MQA files will best the Aqua La Scala playing Redbook content – it doesn’t. Complicating the issue somewhat is that I personally prefer listening to Redbook (or hi-res PCM) decoded by the PS Audio DirectStream Jr than MQA via the Mytek Brooklyn. The PS Audio has a more ‘full cream’ flavour compared to the Mytek’s leaner espresso hit.
Sticking with the Mytek though, a move from hi-res PCM to MQA on the Brooklyn brings SQ improvements similar in size to switching out the source from a budget streamer to a more high end solution. Or like moving from a stock Sonos Connect to one modified by Wyred4Sound.
What of the majority who don’t (yet) have an MQA DAC? Aren’t they left out in the cold? One thing that the MQA story (so far) has shown us is that audiophiles detest being left out of the conversation and will advance an opinion on MQA’s audible merits whether they’ve heard it or not.
This is where our story gets REALLY interesting.
According to Bob Stuart, the third and most important strand of the MQA trifecta takes place before the music even reaches the end user’s DAC, the download store or streaming service.
MQA isn’t a codec, it’s a process. And that process begins at the studio. The MQA algorithm corrects time domain inaccuracies (pre- and post-ringing) caused by the original A/D converter. This is either done at the studio, soon after the recording is made or, for existing recordings, with one of the handful of MQA-equipped computers dotted around the globe.
The implication being that MQA can potentially correct the temporal smearing of almost every recording in existence. Success here largely depends on whether the original A/D converter’s make/model is known. If it is known, precise settings can be applied, otherwise a best guess filter is applied.
Bob Stuart explains: “What is happening here is that the encoder (using system metadata and/or AI) resolves artefacts that are obviously different in each song according to the equipment and processes used. When these distracting distortions are ameliorated then the decoder can reconstruct the analog in a complementary way.”
“Removing ‘pollution’ is not lossless” [my emphasis].
This means each MQA file’s DNA is different to that of the standard hi-res PCM equivalent. In strictly data terms, the MQA encoding process is lossy – it is no longer the studio master as archived by the record label. Bob Stuart’s proposition with MQA is that the originating hi-res PCM file is filtered in order to make it sound better with any D/A converter.
Stuart continues: “…the magic is that when these distracting elements are not present our perceptual system has an easier time grouping and streaming and locating the sound objects in the presentation.”
“With regard to the filtering aspect of this, thirteen years ago we showed apodizing could help at playback, but to get temporal resolution down to the 10us level or below we must work on the whole chain. Doing so allows us to reduce blur by at least an order of magnitude compared to conventional 192kHz.”
This begs the question: how do PCM hi-res and the corresponding MQA compare when played back via a non-MQA DAC?
Without MQA in the DAC, no un-folding takes place – only the first 44.1kHz or 48kHz is parsed – and the pre-emptive filter correction isn’t applied. Only the MQA encoding process is heard. Would hi-res PCM, untouched by MQA, sound better, worse or the same when A/B-d against MQA’s time domain corrected, sample-rate limited equivalent?
With a PS Audio DirectStream Jr DAC sat below the Mytek Brooklyn in my HiFi Racks hifi rack and with files in hand, the answer was only a morning’s listening away.
I’ll cut to the chase: for the same reasons stated above, but to a lesser degree, the MQA files handled by the PS Audio DAC bested the sound quality of the 192kHz or 96kHz hi-res versions in every single case, from Muddy Waters to Steely Dan to Ravel and (especially) to Ella and Louis.
That ‘pop’ sound you just heard was the seal breaking on a fresh can o’ worms.
I don’t really want to qualify this finding but I feel that I must given its (probably) controversial nature. This is what I heard, alone in my apartment. You might hear it differently had you access to these same files played back through the same hardware. The hardware you could buy – but before you venture an opinion on how MQA sounding better on a non-MQA DAC cannot possibly be true, ask yourself: do you have access to these same MQA files? Conversely, we don’t yet know if other MQA files will sound as good as the sample set provided to me by Bob Stuart via Dropbox.
Also, know that I repeated this same experiment with the AURALiC Vega DAC. No MQA software on its XMOS chip but also no FPGA silicon and no Ted Smith code. Instead an ESS Sabre decoder. And guess what? Same result: MQA sounds better than hi-res even when the DAC chip doesn’t see north of 48kHz when faced with MQA source files.
Here we note AudioStream’s Michael Lavorgna concluded similarly in his own MQA review:
“Does MQA encoded music played back on a non-MQA DAC sound better, worse, or the same as playing back the file in its native resolution? While it depended on the recording, un-decoded MQA file did not sound worse than the original native file and in some cases, the un-decoded MQA version sounded better.”
The implication is that you don’t need an MQA-capable DAC to feel the benefit of an MQA file. Which is great news for those not (yet) ready to drop big cash on new hardware.
Listeners wanting to get MQA ready and taste the very limited amount of content already available for download can grab a Meridian Explorer2 for US$299. My findings with MQA listening don’t suddenly make my own counter-arguments about catalogue size disappear. Nor do they address technical concerns (heard elsewhere) about MQA encoding. Those stories and the audio engineering debates that fuel them remain real and will continue to be covered here as time allows.
Downloads though – the future they are not. We know their sales are trending downward. Although their official position is one of delivery method impartiality, MQA probably aren’t betting the farm on HDTracks or Qobuz selling their wares.
Readers who don’t want to buy their music again, I suspect Bob Stuart hears you. He probably doesn’t want you to buy it all over again either. Stuart’s making his money elsewhere – at the studio level, at the DAC hardware level and (probably, eventually) at the cloud streaming level.
The way I see it, MQA’s future proliferation will be made (or broken) by streaming service adoption. And right now that means Tidal. The immediate and face-slappingly-obvious benefit of MQA to Jay-Z’s company is the ability to stream hi-res content in Redbook-sized containers. “Babylon Sisters” 24bit/96kHz file size is halved once encoded to MQA: 131.6Mb to 73.6Mb. The Ravel piece goes from 221.3Mb (as 24bit/192kHz) down to 62.6Mb (as MQA).
Another benefit for Tidal is the need to archive/stream only one file. MQA will only unfold according to the maximum sample rate handling of the MQA-enabled DAC and will play back just fine on a normal DAC.
The Mytek Brooklyn’s MQA decoding can be toggled on/off according to taste. Off, it’ll behave like a non-MQA DAC. That’s useful if you don’t want the hi-res portion of the file that some DAC engineers are claiming is partly lossy. The time-domain correction within the MQA file will remain.
Tidal Hifi subscribers with MQA-capable DACs get the double whammy of a full hi-res stream decoded by an MQA-optimised signal path. Those not in possession of such hardware will, based on my listening experience, see an uptick in sound quality without lifting a finger. Or wallet.
Tidal is the key that will unlock the door to MQA’s proliferation. It has the power to bring MQA to a much wider audience, side-stepping the need to re-buy one’s library, and for – fingers crossed – the same price as the existing HiFi service’s monthly fee. Even if MQA content added to the HiFi service causes the monthly fee to rise by a couple extra bucks, it’s still better value than dropping US$20+ per MQA download.
Moreover, as we’ve seen with exclusive releases from Kanye West and Beyonce, Tidal has the industry clout to make sure MQA-encoded albums aren’t only the same old same old, especially now that Warner Music Group are on board – a licensing deal that Stuart says took three years to finalise.
Imagine a world in which Tidal streams MQA content invisibly to its HiFi subscribers. A world in which Beyoncé albums are treated the same as Diana Krall but with the potential to improve the listening experience beyond the realm of audiophile nerds like you and me.
Further information: MQA