I’d first like to give a shout-out to Sound On Sound.
Much of what is written in this post references their phenomenal article on MQA Time-domain Accuracy and Digital Audio Quality.
What is MQA? Do you need it? Is it overrated? What separates MQA from other forms of playback?
All of these answers and more, comin’ up.
Greetings bass head and Welcome aboard!
Stuart Charles here, HomeStudioBasics.com helping YOU make sound decisions leading to a beautiful audio experience that will make you fall in love with music (NOT gear) all over again, so…
What is MQA?
MQA (Master Quality Authenticated) was invented by Bob Stuart, co-founder of the British Hi-Fi manufacturer Meridian Audio.
It represents an efficient way for people to listen to large, high-quality audio files that would otherwise be too bulky to play through a typical streaming app (like Spotify) that requires a lot of bandwidth.
- Related: Tidal vs. Spotify [Definitive Guide]
The process takes master quality, hi-resolution files (authenticated by the artist in the studio), and basically packages them into a smaller size without reducing the quality.
Lossy, compressed files are typically very small and convenient, but come at the expense of an honest representation of the track in question.
Those files aren’t actually authentic as they come from third-party apps such as Apple, Amazon, etc.
The files are given by record labels but are subsequently compressed down to a manageable size for streaming.
MQA circumvents this issue by providing the original file in a small package without sacrificing quality.
Another claim to fame of MQA is its ability to reduce what is known as time (or temporal) smearing.
To understand time smearing, we first must understand frequency response, bit depth, and sample rate.
As outlined in my What is DSD? article, my Bit Depth vs. Sample Rate article, and many other articles and videos on this site and my YouTube channel, humans cannot hear over 20kHz (and cannot perceive high sample rates inherent in say, DSD or ultra-high PCM) no matter what.
This is one of the most basic concepts in audio that gets glossed over time and again by people who have no idea what they’re talking about.
What is interesting to note is the fact that we still may be able to perceive, on some level, the intricacies of a higher sample rates’ effect on our perception of sound; or what is known as temporal precision, which is measured through impulse response.
In simpler terms, frequency response is very important, but the time-domain performance is also just as important; albeit less understood.
In 1929, Von Békésy discovered that humans can actually hear time differences of up to 10µs (microseconds).
For clarity’s sake, a microsecond is equal to one-millionth of a second.
While this does suggest that higher sample rates can be effective in providing a somewhat more resolving signal (i.e. they can improve temporal precision to an extent, thus somewhat reducing the ringing tails after the main impulse), the law of diminishing returns kicks in after about 96kHz.
In other words, anything beyond that is STILL essentially useless from a “more is better/higher is superior” standpoint; something audiophiles still love to argue with me about even despite overwhelming evidence that it doesn’t matter.
There you have it, Diminishing Returns makes yet ANOTHER return in audio.
I’ve talked ad nausea about this concept in headphones and DACS, but now it’s proven yet again that it exists in the source file itself!
- Recommended: Beginners Guide: What is a USB DAC?
Who would have thought?
I’ll say it again; All of the relevant audio information ranges from 0Hz to about 22-24kHz (the yellow triangle below). Anything from 24-48 can be relevant in that there is still a small amount of musical information present above the noise floor (as illustrated in the diagram below), but anything over 48kHz is just negligible noise and doesn’t matter.
Now that you’ve read all that, think for a second about the hundreds of thousands of DACS on the market today, all vying for a chance at your wallet.
Now think about how many of those DACS advertise these ridiculously High PCM rates that don’t matter at all, the DSD crap that … doesn’t matter at all, the Quad, Octa, Hexa, Dexa, Plexi-glass infused Amp/DAC and blah blah blah blah blah.
None of it really matters.
Now, what CAN actually reduce the unwanted pre and post-ringing tails in an impulse response?
Enter Apodization
A concept published in 2004 by Peter Craven, this cascade series of filters allows for almost complete eradication of said tails and temporal smearing; a concept also utilized by MQA.
With a typical 24/192 system, you’ll get about 100µs (100 microseconds) of time smear from an impulse response.
MQA aims to cut that number to around 10µs, right in line with what we discussed above in regards to Von Békésy’s findings; that is, that humans can perceive time differences of around 10µs. Not bad!
The only way this works, however, is in a fully functioning system; that is, the complete end-to-end sampling & encoding process (MQA), all the way to the decoding process utilized by an MQA-certified DAC.
Even if you don’t have an MQA DAC, you can still theoretically hear masters inside Tidal – but only the first unfold of the decoding process.
To hear the most accurate representation of the track, you’ll want the full unpack and that can only be achieved with the appropriate MQA-certified DAC.
But how exactly does MQA provide an ultra-high-resolution track in a tiny package?
Simply put, they use what they call the origami method of unpacking (this is referred to as encapsulation; not conventional lossy downsampling).
It’s also important to note that while traditional sample rate conversion employs anti-aliasing filters to negate the unwanted artifacts in the recording, MQA does not.
This strategy allows for a completely clean signal without the unwanted timing blur that goes along with the more conventional method of conversion.
It also does not largely matter anyways since, again, any aliased elements will always be infinitely quieter than anything in the audible range – i.e. you still cannot hear them.
But does MQA sound better?
This is perhaps the most important question and one that I’ve personally wrestled with for quite a while.
I know others have as well. From a theoretical standpoint, it would seem to make a world of difference.
Within the scope of a practical listening experience?
I lean towards yes, but it’s really hard to say as you’ll find out later.
I do think Sound on Sound’s analysis is excellent and echo almost the same exact sentiments, but I’ll put it in my own words.
The music just takes on a somewhat more live flavor. The instruments sound crisper, livelier, and more vibrant.
They have more personality. The Soundstage is better and everything sounds more liquid-smooth and natural.
The Timbre of instruments sounds better and more correct.
Attack, sustain, and decay are just that much more realistic and immediate. What is Timbre?
Now is it a monumental difference? I would say no. Could it be a placebo? Absolutely.
It’s more subtle than you would think, but to me, it was definitely “there” for whatever that’s worth.
Even if you knew nothing about what I’ve discussed today, you’d likely find the music to sound crisper and lusher.
At first, I chalked that up to the music just being a bit louder, but upon further research into the actual science behind MQA, my tune has certainly changed.
Again, Golden Sound’s analysis throws a bit of a monkey wrench into this which we’ll also outline in a bit.
Whatever differences I perceived at first make complete sense after further investigation into these concepts.
With a great set of headphones like the K712, you’re getting as close to a ‘You’re there’ feeling as can be realistically expected.
But remember, you still have headphones on your head.
In my exhaustive research over the years, I think a lot of people overexaggerated the effects of a headphone with a good Soundstage, particularly in the case of AKGs.
It’s not that the headphone doesn’t provide an out-of-your-head experience; it certainly does.
But the wild and unsubstantiated “It’s like you’re there!” claims can be safely discarded if you happen to be reading this.
It’s really not like that.
A headphone like the K702/K712 can give you a more “in-person” type of feeling, but those moments are fleeting.
- Recommended: AKG K612 vs. K712 vs. K702 vs. K701 vs. Q701
They come and go.
It’s not as if they just linger for the entire duration of a listening experience, causing you to feel like you’re in the front row of a Phish concert high as a kite.
A good speaker setup to me is still superior to a great set of cans with regard to the illusion of being there, but even that isn’t perfect.
I had a chance to listen to $60,000 Mark Levinson speakers at an Audio Advice Music Matters show a couple of times in 2018 and 2019.
While they sounded utterly breathtaking in every way, I still had to close my eyes and concentrate super hard to feel even one iota of being in front of a band.
So you can imagine how out-of-control audiophiles get when they try and describe that^ within the context of tiny headphone drivers that are inches away from your ears.
- Related: What is a Headphone Driver?
Golden Sound’s Rebuttal
Around 2021, Golden Sound came out with a video basically claiming MQA was very “iffy”, and it certainly made the rounds.
Fortunately for me (and you) I took notes and still have them on paper.
After going over them, we’ll make an ultimate, final determination on whether or not MQA/Tidal is a legitimate form of lossless music.
Here are some of Golden Sound’s main points condensed.
Please keep in mind these are jotted notes from 2021 or so and may be a bit erratic:
MQA is proprietary
and nearly impossible to test.
- We cannot encode music in MQA ourselves or use test files to find out what is actually going on.
- Unfolded MQA may not be the same as native hi-res.
- We can’t get a true digital unfold and check it against the original.
Thus, MQA makes it very difficult to test claims and figure out what’s going on.
We need to find where real original masters and MQA versions are available.
Good ways to test:
- 1kHz Sine Wave – Used for testing distortion and other performance metrics of audio hardware.
- Impulse Response Test – This represents all frequencies used to test filter/reconstruction performance and digital processing.
Golden Sound then made a test track, putting all of the following inside:
- Impulse Response
- White Noise
- Square Wave
- 32-tone test signal
- RMAA test sequence
He also published an 88.2 kHz version with test tones of 35k and 40k Sine waves, as well as full 20k – 44k sweeps.
All of this is to find out if anything gets removed in the compression process.
He then determined that a 44.1kHz and Native FLAC file were identical in Deezer but not in Tidal.
Issues in Tidal:
- High-frequency noise was present.
- Small differences throughout the audible band of musical content.
- Clipping.
- Transients overshooting giving the impression that it’s louder.
That last point is important because it’s something I may have experienced initially (and something I mentioned earlier).
Other notes:
- MQA does not preserve phase integrity. The ringing rebounds before stabilizing.
- Audibility of pre and post-ringing is not sufficiently proven one way or another and may be signal dependent.
- Phase linearity/phase distortion has been demonstrated to be audible (more so with headphones than speakers).
- Different frequencies move through the filter at different speeds throwing off the Timbre – which relies on multiple frequencies at the same time.
- The unfold presents problems, only rectified by the HQ player’s poly sine MQA filter
Tidal MQA folding of 88.2kHz to 44.1 was messy.
There is more high-frequency noise even in parts of the track that should have been silent.
The Main Point
MQA files are not lossless, but if I could interject here, this isn’t exactly news. PS Audio said the same back in 2018.
*If no MQA DAC or unfolding is used, issues will arise compared to lossless FLAC.*
This is also super important because of what we touched on earlier according to Sound On Sound’s analysis.
The amount of noise that MQA adds is proportional to the amount of ultra-sonic content that MQA has to reconstruct.
60dB 1kHz Sine Wave Result:
- Dynamic range is handled poorly.
- High-Frequency noise around 45dB below the signal (Is this MQA adding dithering?)
In a 44.1kHz MQA file, produced from a hi-res master, the noise is 20dB above (even noisier).
DAC unfolded a track to hi-res 352 (kHz) when it was never a high-res track, to begin with. It was simply upsampled.
Tidal also doesn’t serve the lossless file, only the MQA file. Tracks labeled “master” cannot be heard as lossless.
Unfold/White Noise
Ultra Sonic attenuation performance and filter design
- MQA’s filter has poor attenuation.
- There is still persistent High-Frequency Noise
- The Square Wave has excessive overshoot.
- It’s not beneficial to the listener or the integrity of the song.
- The upsampling filter introduces artifacts.
- The complete unfolding process does essentially nothing.
88.2 File with Sine Wave test
Notes:
- Golden Sound added a 35k and 40k Sine, too high for a normal 44.1k file.
- Can MQA retain this info? The Sines in the MQA 44.1 file have been aliased down into the audible band @ 9.1k and 4.1k (NOT below the noise floor). This basically results in more noise.
- After the complete unfold, the 35k and 40k sines were in their correct spots. So MQA’s claim of restoring high-frequency content is partially true, although there is still a lot of noise.
- There are more artifacts in lower frequencies.
- Poor dynamic range in audible and ultrasonic bands when MQA is unfolding ultrasonic content.
- In the full decode, problems were still present.
- The claim that the process retains 100% of the original recording is clearly false.
- It is also not folded back down to inaudible levels as claimed by MQA – another main theme.
Final Thoughts
Some people on Reddit claimed Golden Sound’s findings were a scam in their own right, but I’d have to disagree.
Did you watch the video? Did you take notes and at least try to understand where he was coming from?
I think the main point is not necessarily that MQA is a “scam” per se (though it still may be, who knows), but rather, the methods Tidal uses to achieve what they believe to be “lossless” (i.e. the Origami method mentioned in the open, the unfolding process, etc.) may actually be flawed protocols.
This is evident in the amount of noise/artifacts, poor transients, etc. that were found in Golden Sound’s tests which to me was kind of the overarching theme here.
Whether or not his tests are sound/valid is an argument for the comments perhaps.
Leave one down below and let me know your thoughts!
The Full Unfold
It’s also important to note that many claim you must actually have a DAC capable of the full unfold, which corresponds with the literature I outlined earlier from Sound On Sound.
Golden Sound also corroborates this in the video, so I’m inclined to believe that MQA isn’t quite the monster people were making it out to be after his video was released (Yay overreactions), but there are also elements to the process that seem kind of suspect upon further investigation.
For me?
Canceling my subscription boiled down more to the fact that I simply wasn’t using the app – not nearly enough to justify the cost, anyhow.
Not when I have Spotify and a ridiculous amount of playlists.
Did Golden Sound’s video play a role in helping me make that final decision to cancel? Absolutely.
But I welcome a discussion on this and am open to alternative viewpoints.
Well, that’s about it for today my friend! I hope you’ve enjoyed this article on What is MQA? and came away with some valuable insight.
If you love what I do here and want to support the blog and channel in a more personal way, check me out on Patreon and discover all the value I have to offer you.
Questions? Comments? Requests? Did I miss the mark on something? Please let me know down below or Contact me!!
Is MQA worth it to YOU? I would love to hear from you. Until next time…
All the best and God bless,
-Stu