Every major streaming platform rejected it before anyone ran a blind test.
A 1.5 megabit-per-second file outperformed the original master at nearly five times the bitrate. Robert Harley, a 37-year veteran and editor-in-chief of The Absolute Sound with a degree in recording engineering, says the technology responsible is “going to be controversial, but MQA.”
Meanwhile, the market already rendered its verdict. MQA entered bankruptcy in April 2023, and Tidal dropped every MQA track by July 2024.
The Demo That Convinced Harley
Peter McGrath, Wilson Audio’s director of sales and someone Harley calls “one of the great recording engineers,” brought his own 88.2 kHz/24-bit master recordings to Harley’s home for a comparison on Wilson speakers.
He played the raw file at what Harley describes as “like 7 megabits per second.” Then the MQA encode at 1.5.
Harley says the MQA version sounded more alive and more convincingly like a real acoustic space.
His numbers deserve scrutiny, though. Uncompressed stereo at 88.2 kHz/24-bit actually calculates to roughly 4.2 Mbps, not Harley’s “like 7 megabits.”
This may be because he’s including format overhead or rounding generously. But, even by the conservative math, the MQA file ran at a third of the original’s bitrate.
The comparison also wasn’t blind. Everyone knew which version was which. The group was small, the speakers came from McGrath’s own employer, and no one has independently corroborated this specific listening session.
Still, none of that changes the question his demo raises. How can a file with fewer bits outperform the original? Harley has an answer, and it doesn’t come from audio engineering.
The Time-Domain Theory Behind MQA
Human ears can resolve timing differences down to 8-10 microseconds. Standard digital audio filters blur transients across several hundred microseconds. That gap, Harley argues, is where MQA lives.
The intelligence didn’t originate in audio, though. Harley says MQA’s creators adapted it from medical imaging and astronomy, borrowing deconvolution techniques that have sharpened telescope images and medical scans for decades.
For instance, in medical imaging, resolving fine spatial detail means correcting distortions introduced by the imaging process itself. MQA applies the same principle to audio, targeting the temporal smearing that standard digital filters impose on every recording regardless of its resolution.
If the deconvolution works as claimed, the corrected signal recovers transient detail that the original digital encoding threw away.
MQA claims to reduce that smearing to roughly 10 microseconds, which brings digital audio within the range of what human hearing can actually detect.
Other listeners heard the same thing, in different rooms and different years.
Same Trick, Same Reaction, No Blind Test
Harley says he kept hearing that same improvement in later demonstrations. At CES 2014, through Meridian DSP7200 loudspeakers, the impression was so striking he could recall it “nearly a year later” while forgetting every other demo at the show.
Two years after that, McGrath’s opera recording of Tosca in MQA against the 88.2 kHz/24-bit original left him writing that his “jaw dropped, literally.”
Then someone outside Harley’s circle heard it too. At the 2017 Los Angeles Audio Show, reviewer Brent Butterworth attended a public McGrath and Bob Stuart demo on a $153,500 system.
But no one ever ran that test. Every demo was sighted, every system cost six figures, and every recording was McGrath’s own. Without controlled evidence, the listening impressions stayed exactly that.
How Measurements Shaped MQA’s Failure
The rest of the industry wanted measurements, not memories. GoldenSound’s independent analysis found that MQA files on Tidal introduced noise and artifacts, contradicting the technology’s core promise of lossless delivery.
Meanwhile, Apple, Amazon, Deezer, and Qobuz all chose open FLAC for high-res streaming.
So by the time Lenbrook acquired MQA’s assets for $125,000, the technology was worth less than the Wilson speakers used to showcase it.
Harley’s sustained advocacy has drawn scrutiny, though no commercial relationship between The Absolute Sound and MQA or Meridian has been confirmed.
He doesn’t dispute the collapse. He disputes the reason.
“People had a hard time getting their head around the idea that a copy could be better than the original,” he says.
“How can you improve on the original master? And it’s got a bit rate a fraction of the original master.”
He argues that many technically skilled engineers rely too heavily on measurements and not enough on listening.
Even MQA’s co-creator Bob Stuart framed the issue in similar terms, arguing that you cannot reliably predict how something will sound just by looking at an FFT plot.
In his telling, MQA died for exactly that sin.
MQA… McGrath Quality Audio, I assume? The whole thing sounds like an ego-driven con, aside from the technical aspects of the short transients. Improvements are one thing, but vast improvements require vast scrutiny.
Completely wrong about McGrath being what the M stands for.
Good grief.
I had Tidal subscription around 2018 and ended up dropping it around 2021 because i was barely using it and needed to save some money, i used to listen to music using tidal and a set of senheiser headphones, also another pair of astro a40 using the optical output of a Lenovo laptop, it sounded better than any other service like Spotify, last year I got my subscription again, and I even tougth something was wrong on my headphones or system, I started looking online for solutions and then learned that tidal dropped mqa, now I cannot longer tell the difference between tidal and even YouTube.
Mr. Harley is absolutely, unequivocally right. I have a number of MQA recordings, and have the equipment to make use of MQA. As part of my experiment, I also bought the non-MQA versions, on both CD and SACD . The MQA recordings sound not just better, but much better. It’s sad but not enough. People were able to get over their prejudice enough to trust their ears. RIP MQA.
The issue with MQA is that most of it was a scam, only a few actual MQA files were produced properly. Most of the time it was just a little bit of code that would turn on that little blue light with MQA underneath it.
I have equipment that can decode MQA files connected to my reference system and I can hear no difference at all compared to other high quality formats. The only formats that seem to change in sound are digital Vs analogue, or using low quality digital like MP3.
24bit studio recordings for digital or vinyl for analogue I have found are the best formats.
Most of the time
The claim that humans can discern time differences of 10 microseconds seems absurd. The unit of measurement should be milliseconds, not microseconds. I’m a retired electronics tech who now works on guitar amplifiers. The “echo” or “delay” features on modern amps allow users to adjust the timing between the original tone and its echo. There are no microsecond values; only milliseconds. No human can hear audio details that occur during a 10 microsecond pause. I will stand corrected if anyone shows me data taken during a bonified experiment.
“Stop looking at objective tests and listen to my perceived differences which were probably made up by the placebo effect.”
Lost me at the lack of ABX testing.
I met Mr. Harley once upon a time at and audio show we participated in. The man could definitely hear and was fascinated and open to unorthodox equipment and speakers. Like Mr. Harley, we too listened to mqa via Tidal. We tested it against vinyl, remastered cds (even xrcd, up sampled files /flac, sacd, and it consistently was markedly superior to everything else. The instruments just much more life like. The voices had a wholeness to them. It was like information that had been lost had somehow been found. Just magical sounding. While the files were smaller, they obviously had more correct information, and sounded like they were higher resolution than the supposedly higher resolution files. Having had some experience upsampling files myself, I can tell you that 24/192 is not necessarily better than 24/96. And often the higher sampling can introduce noise, a thin brittle sound and hardness. Higher sampling rate does not necessarily equal higher fidelity. Mqa was doing something very very right. And it’s a pity that it was doomed by the segment of audiophiles that listen to their imperfect measurements over their own ears. There is a dac out there that employs some of this technology from mqa that intrigues me. I hope that it may be a silver lining out of a great tragedy.
It’s not that 192 is worse than 96 — it’s that 192 has so much data that, even if you can’t put the issue into words, the fact remains that you are listening to a better sample of the audio. The issues which people tend to have with higher sample rates stems from how the microphones were able to capture sound, which the human ear cannot typically process, but which it is now being forced to process. There are tones & waves, sitting just outside of the hearing range for the extreme majority, which can be unpleasant to experience, depending on the rest of a mastering, or simply the rest of the natural sounds around you. I think that higher sample rates allow normies to experience that potential unpleasantness.
My point of view is informed by my strange situation, as I am deaf in my right ear, but, my left ear has repeatedly tested at a “super listener” or “super hearing” level. It is possible that I am autistic (I’m only getting tested in the near future, though the diagnosis has been suspected for decades), and autistic people, as well as people with an intellect nearly as advanced as mine, have been studied & shown to have a capacity to hear differently — or process more of what is heard — compared to the average listener (average in hearing only, as there are many brilliant people who only have average hearing). In other words: the things you dislike about high sample rates, are your opportunities to experience autism.
Sounds good but that high end market sailed 40 years ago
“They need to see measurements to justify what they’re doing.”
Your ears are not consistent, exactly what should engineers use to be above board on capability? The quality of the initial source will follow ANY format. That’s where the work needs to be done, at the mastering stage. Transfer to record, 16, 24 or 32 (not needed for playback, just mastering) bit or MQA it doesn’t matter, the initial source does. Playback quality is terrible, and the final format isn’t the problem. It is that mastering engineers DO USE THEIR EARS and we hear the awful results of that. We either can’t measure the same way or we can’t hear the same way and, which do you think is the most variable?
Always thought MQA sounded great and was very unhappy when Tidal dropped it. Although Tidal claim to have converted all their files to FLAC there are still are quite a few MQA Albums/Tracks on the platform. They still sound great. After all is it about the sound or the numbers?
Trabajé 25 años en Sony Chile como Engineer Training Meister. Con el advenimiento del formato Mini Disc organizamos una prueba a ciegas con 50 ingenieros de sonido en Santiago de Chile. Fuente A: Formato CD y Fuente B; Formato Mini Disc con con compresión ATRAC III Plus (Adaptive Transform Acoustic Coding), que comprime aprox 4.5 veces la Data original. Bingo: 49 de 50 escuchas prefirieron la fuente B (Mini Disc). De acuerdo a la Intranet de Sony Corp Japan, esto ocurre porque este formato comprimido inteligente se adapta mejor a la compresión que hace el propio cerebro humano con la música, causando menos cansancio auditivo. Por ejemplo, cuando el cerebro recibe una frecuencia fuerte acompañada con una frecuencia débil muy cercana, el cerebro elimina la frecuencia débil, justamente lo que hace ATRAC III Plus, entre otros parámetros biológicos.
If you don’t trust your theory enough to run double-blind tests then you don’t trust your theory, and you’re just selling snake-oil.