Comparison of PCM and MQA

Audio Gear Talk MQA
530

This topic, and MQA threads in general, has consumed disproportionate amounts of Moderator attention. We think that the debate will benefit from more considered responses, rather than the rapid tit for tat that seems to result in frequent flagging of posts. Accordingly we have slowed posts by users in this thread to once a day.

Just a reminder, if you want to discuss this decision, or moderation generally, open a new thread in the Community Forum section to do so. Posts about those things will be off-topic in this thread.

7 Likes
531

Hi robbi. Thank you very much for the references. I am interested in signal processing (especially dithering, noise shaping and wavelet transforms), and would like to know more, in context to MQA and beyond. So far I have taken effort to learn a few from youtube and personally working out the math. Would it be okay to communicate with you over PM?

532

I did a little bit of analysis with a graph posted online (it was sampled using a laptop adc at 96khz sample rate so not the most ideal way, but useful to get a basic idea).

The result does show up nicely when comparing spectrograms of high res music vs the mqa encoding (music taken from 2l no I think). This is what I’ve understood so far.

First the high sample rate data is apodized. It’s a technique used to ensure less artefacts in the band of interest. Temporal properties, less pre ring/post etc. As it looks somewhat similar to a min phase filter if you look at the bode plot, somewhat trapezoidal in shape (it’s quite a bit different from min phase filter but you can think of its lack of pre ringing to be somewhat similar to how a min phase filter obtains it). To say in simpler words but the reality is a little more complicated, and I’m still trying to work through the rigours of the math behind apodization. (Yes I do understand that it has spectral leakage)

The high frequency noise profile of the adc is characterized beforehand and a transform/mapper is made. With this information the first folding is done. And then the second folding is done using another transform that embeds both the audio content in 20-40khz and the noise profile into the pcm stream. This folding/unfolding algorithm is done using wavelet transforms based on some complex mathematics that I’m still in the process of learning. This is encoded into the pcm stream in the last few bits as some pseudorandom code that the decoder understands, and the algorithm for this encoding and decoding is a completely lossless process (The encoder is designed assuming an apodized signal I think, if so thats a prior condition) if what I understood is right.

This video visually expresses what I’m trying to say.

Regarding the results. Aside from the apodization effects in the ultra high frequencies (30khz plus) everything else looks intact on the spectrogram. And this 30khz+ change loss is not a worry since that loss only accounts for the pre/post ring information. The temporal properties of even the 30khz plus signals are most likely preserved, and sub 30khz are surely intact. If the same was done with a higher sample rate adc with more unfolds I believe the 30khz would become 60 or 75khz.

I don’t quite know about the implementation specifics of what exact transform or weights are used, and I’m still learning about those in general context. There’s too much math rigor on this topic - splines, triangular sampling kernels (wonder what its time domain response would be like, and whether it is uniform), wavelet transforms and what were the basis functions used, and dither (not sure how this is accounted for, haven’t seen this mentioned much in mqa). All these are far more than what a beginner signal processing person like me would be able to understand in its entirety, but I’m making steps to understand it. Even dither, for which there’s quite a bit of literature online is quite mathematically rigorous. But the deeper I explore, the more intuitive and exciting it gets, and I highly recommend anyone else interested to explore further as well (need not be mqa, can be in general, and that’s what I am doing). Hope I get guidance to learn further on the above topics.

4 Likes
533

“MQA Sonic degradation and huge losses”

Very interesting article of someone who had close contact with B.S. and wrote a book about sound quality with a whole chapter dedicated to mqa titled “MQA : A Solution to What?”.

Link to the original page below.

Quotes :

MQA is so dangerous to the recording industry that I dedicated an entire chapter of [Music and Audio: A User Guide to Better Sound to MQA. Titled “MQA: A Solution To What?,” I included discussions of the technical failings of the process and the fact that “folding ultrasonics” doesn’t matter if the original masters don’t have any ultrasonics to begin with. There were sections provided by designers and audio engineers confirming my analysis of MQA. The book also included a lengthy interview with the format’s inventor Robert Stuart, who I have considered a friend and supporter ever since I started recording and releasing high-resolution recordings. Bob was very impressed with my work and even gifted me a couple of very expensive Meridian components years ago. But his efforts to create an MQA-only streaming music world is misplaced and dangerous. And it seems largely motivated by money not fidelity enhancement .

Dollars and Nonsense

Companies like TIDAL and nugs net are charging a hefty premium for content encoded with MQA. The cost of the MQA encoded format is twice the cost of the MP3 and $2.00 more than a lossless “HD” version encoded using FLAC. Give me the “HD” FLAC version. FLAC is an excellent codec, it’s FREE, and lossless. In fact, MQA “comes in a lossless FLAC file.” Why pay more for something you can get for free?

I hope that MQA fails miserably and is relegated to the dust bin of failed audio technologies. It is not good for the industry. It is not good for fidelity. And it’s not good for music reproduction. The only thing MQA is good for is MQA Limited, Bob Stuart, and its investors.

Source :

6 Likes
534

So they don’t have any advantage from seeing MQA fail then…?

535

There is a lot of history being left out of the article you are quoting by Mark Waldrep. I won’t comment because it will only contribute to the ongoing cycles of negativity that the moderators are attempting to remove. But I’d recommend avoiding this source. It is heavily biased, as should be clear from the wording.

2 Likes
536

My parting comment on this is to be quiet honest unless MQA is forced on the industry I wouldn’t worry about it

Studios do not use it as part of production even after a decade, you are not hearing what the artist heard this is an impossible statement to make to the consumer, it’s not lossy in the words of MP3 but it does not preserve frequencies in the hi res domain that were there originally, the long term goal is to have one consumer controlled file on all platforms which you have to buy into to ‘better’ the sound, now if audiophiles are happy with that then that’s all well and good, I myself would prefer digital standards that the studios use and alter the file how I want to hear it not what a pre set filter system thinks how I should hear it. The hi res genie was let out of the bottle years ago, certain companies unfortunately want to put it back and make you pay for it going forwards

On my last count I believe there’s around 1.5 million tracks showing out of a 70 million alleged in Tidal so putting this into perspective Hi Res is a niche market still in lossless streaming and MQA is carrying approx 2% of Tidal’s total output, plus they only have Tidal propping this up where is the expansion coming from. Amazon / Qobuz / Deezer are FLAC PCM and Spotify is soon to go lossless unless they persuade Apple then they are stuck, my concerns are why buy into technology that doesn’t have a good road map going forwards

Enjoy your music whatever bottle you drink from

3 Likes
537

[Moderated]
Why should we avoid reading anything from a Ph.D with 40 years of experience with music, recording, playback and all the mediums like cd, dvd, bluray, sacd, … Is that just because he tells the truth about mqa [Moderated]?

Some background :

Mark Waldrep, aka Dr. AIX, has been producing and engineering music for over 40 years. He learned electronics as a teenager from his HAM radio father while learning to play the guitar. Mark received the first doctorate in music composition from UCLA in 1986 for a “binaural” electronic music composition. Other advanced degrees include an MS in computer science, an MFA/MA in music, BM in music and a BA in art. As an engineer and producer, Mark has worked on projects for the Rolling Stones, 311, Tool, KISS, Blink 182, Blues Traveler, Britney Spears, the San Francisco Symphony, The Dover Quartet, Willie Nelson, Paul Williams, The Allman Brothers, Bad Company and many more. Dr. Waldrep has been an innovator when it comes to multimedia and music. He created the first enhanced CDs in the 90s, the first DVD-Videos released in the U.S., the first web-connected DVD, the first DVD-Audio title, the first music Blu-ray disc and the first 3D Music Album.

So he’s quite an expert, and sorry but I believe an expert with 40 years of experience more than some mqa marketeers that still don’t want to admit mqa is lossy and call it “more efficiently packing”.

Since mqa marketeers like it visual, well here is a good comparison between mqa and jpeg.
When you save a jpeg in photoshop you can choose the compression rate, the mqa encoder does similar things.

Similarities with an mqa encoder :

image
small file = MQA-CD = high compression = low quality

image
large file = 24b MQA = low compression = high quality

When you save a lossless TIFF as a JPEG, a perfect white pixel with RGB value (255 255 255) might become (250 250 255). Almost white, but not white.
And if you increase the compression rate, the file gets smaller, but the picture gets blurred.

Like JPEG, MQA is lossy. The smaller the JPEG/MQA file, the higher the compression rate, the more it blurs the resulting picture/track.

There are dozens more sources showing mqa is lossy.
That’s also what I hear. PCM is much more detailed and crisper, more natural, and it IS what the artist intended, as MQA is a post process always starting from a digital file.

Here another article stating mqa is lossy from 2016
and also mentioning aliasing artefacts… (= mqas weak filter)

https://cbenchmarkmedia.com/blogs/application_notes/163302855-is-mqa-doa

It’s perfectly fine if you want to avoid these sources and stop commenting on them so admins don’t have to jump in all the time. I share that feeling. Like it is perfectly fine for others to share and read those sources as they might find confirmation of what they are experiencing with mqa and why a regular pcm still seems better for them. It’s a free world. If we all were thinking the same it would be a sad world, and I noticed mqa is trying really hard to make that happen. I don’t want such a sad world.

1 Like
538

I had to laugh as there is nothing as lossy as a hot Rod Hi Fi system. Cable losses, crossover losses, amp loses speaker design compromises. Should I even mention Vinyl compromises and losses. That’s even before we get into perfect matching and room acoustic…
The upshot is however, how does it sound? Does it please you, does it make you enjoy the music and I am certain, with every enthusiast, it does, in the main until the next nagging doubt of an upgrade comes along.
Well, the sound of MQA that I listen to, pleases me very much indeed, as to make the insignificant to human perception losses complained about here, irrelevant.

Back to K.D. Lang in MQA now…

5 Likes
539

KD Lang’s output on CD are impeccable nothing MQA can offer changes that

1 Like
540

Umm, no. That’s not how .jpgs work.

2 Likes
541

It isn’t how MQA works either.

2 Likes
542

I said :

So I wasn’t complaining about “insignificant losses”.

Yes it is. That’s exactly how it works. JPEG uses DCT Discrete Cosine Transform which is a lossy compression, that’s also used in other lossy formats like mp3 and aac. They approximate the signal but compression artefacts appear when heavy compression is applied.
(From wikipedia “Discrete cosine transform”)

Because it’s lossy JPEG offers no guarantee the original pixels are preserved. Just like lossy sound formats offer no guarantee the original sound as recorded is preserved.

Here an image detail after lossless and after lossy compression
image

Those pixels are NOT white anymore are they?

JPEG is lossy. Just like MP3, OGG, ACC and MQA are. MQA doesn’t use the same compression method, but is also lossy.

  • PCM is like TIFF (large file, excellent quality)
  • MQA is like JPEG Low compression (smaller file, acceptable quality for marketing)
  • MQA-CD is like JPEG High compression (very small file, clearly audible, in short : crap)

Look up on wikipedia : “mqa”. The first sentence reads LOSSY compression.

You loose details. And I care for details.

If I go to the Louvre I expect to see the real Mona Lisa (PCM) and not some pixelated xerox that they tried to improve by applying some eyeliner, blush and lip stick (MQA).

543

I very often use a Chord Hugo 2 direct to high-end headphones. That eliminates most of the normal, hi-fi sound degradation due to multiple components being used. Even so, with other, and even less resolving systems, I can clearly hear the losses to classical music done by the MQA encoding on TIDAL. I can clearly hear the damage to jazz (eg: Miles Davis). I can clearly hear the vocals aren’t as loud on post-MQA music. I can clearly hear that they have run post-MQA music through some kind of process that sounds similar to me to what 3D plug-ins do. I can clearly hear that the bass is boosted on the old music. None of this is pleasing!

Given all the technical claims they made regarding “de-blurring” etc. have been thoroughly debunked, both because they are impossible, and because a number of us have tested MQA DACs and the MQA encoding process and found that they blur and add distortion to the music, the only logical conclusion is that whatever it is that is “pleasing” could probably be done simply with a 3D plug-in, on low settings, to get the same result.

Distorting music and claiming that they’ve invented some new process that turns musical lead to musical gold is Mastering Quality my Ass!*

*As in the animal, and “asinine”, which has the same Latin origin, should the mods misunderstand the reference.

5 Likes
544

Any losses are, of course, insignificant to human perception. All the musical information is protected. I laugh when people state they can clearly hear these losses as that is incredible as said losses don’t exist but the power of the mind can create them for you if you want them bad enough.
I’d love someone to come round and point these losses out to me on my system because I know how music sounds, I am human and have what I consider normal perception.

I can enjoy music in any format but MQA just sounds so right and tireless.

It’s great listening to Radio Paradise MQA for some of the MQA music not normally available in the UK, so let’s hope this is the first crack in the licence issues that stops this great music being available for all in MQA. The standard CD stream is still there along with AAC and MP3 for those who love those codecs and how they present the music. Blurred and Lossy… Enjoy the music however you listen.

1 Like
545

I’m sorry, but quoting a Wikipedia article you don’t understand doesn’t make it so. Here’s a real life example. The JPG is a Save As from the Tif. We can discuss this further if you’d like but as this is a Roon Forum and not a Photo Forum it won’t be here. If you’d like to jump over to DP Review or The Luminous Landscape we can continue there.

Pure White JPG
Pure White JPG1012×1054 38.1 KB
Pure White Tif
Pure White Tif1014×1052 38.3 KB

546

This is all getting very silly now. Yes compressing a plain white page isn’t going to find much opportunity to alter the white. The point is, artifacts are produced when varying values are compressed and a block of white (or any recurring pattern) will be influenced by the values around it. But who cares, this is how it works, either we accept the artifacts, or find another way. Just like with MQA.

4 Likes
547

… like mqa.

Your statement “the standard cd is still there” is false. The standard CD stream is not always still there on Tidal, on the contrary.

There are basically 2 possible ways they take, depending on the original format (16/44 or 24bit)

  1. 16/44 (redbook) PCM
    before mqa : Tidal offers the 16/44 PCM for streaming
    after mqa : Tidal offers the 16/44 (sorry 15/44) mqa, the original PCM is always removed !!!

(because they don’t want you to hear how much better the lossless version was)

  1. 24/xxx (hi-res) PCM
    before mqa : Tidal offers a downsampled 16/44 pcm, you won’t see the 24/xxx PCM
    after mqa : Tidal offers 24/xxx mqa.
    The 16/44 pcm can safely stay on Tidals servers, because Tidal and MQA are happy when people compare the 24/xxx mqa - that is 2.5x bigger - with the 16/44 pcm. Most people won’t realize they are comparing 16/44 pcm and 24/xxx mqa.

So never one can see the original PCM an mqa was created from. Not on Tidal.

This shows that Tidal and MQA are always hiding the original files.
If MQA was better, they had no reason at all hiding the original PCMs.

For honest comparisons you need a second subscription to a real lossless service like Qobuz, because with Tidal alone you NEVER knew or will know how a hi-res track sounds like.

Honest comparisons are :

  • Lo-res : 16/44 PCM vs 16/44 mqa
  • Hi-res : 24bit PCM vs 24/xxx mqa
    PERIOD

On Tidal you can only compare a Lo-res 16/44 PCM with a 24/xxx MQA.
That is EXACTLY what MQA Ltd did on CES2016… play a lo-res 16/44 CD vs a hi-res 24/192 MQA.
They never let people listen to the original 24/192 master :slight_smile:

I rest my case.

image

5 Likes
549

I have both Tidal and Qobuz. Some Tidal MQA sounds better and some Qobuz High Res sounds better. It’s the master that makes the difference not MQA vs non-MQA.

4 Likes
550

Your comments now contradict themselves. You say that the MQA you listen to “sounds so right”, which means it is different from the original. I have listened to those differences and can hear the losses to the music in jazz and classical recordings after MQA processing. We’re not talking about the losses from the 3 bits of encoding, we’re talking the consequence of running the music through the MQA processing, which involves more than the origami compression, but includes some kind of DSP processing, and alters the sound in an audibly detrimental way. If they weren’t audibly different, the MQA versions would sound the same as the original.

Whenever people post technical information, the argument from yourself and PV is “just listen”. Whenever people DO listen and disagree, your argument is a cut-and-paste of MQA’s technical arguments, which have been disproven.

I’ve posted tracks I’ve listened to, and I don’t see any attempt to listen to them and give feedback about your own experiences. On the other hand, I’ve listened to music yourself, other people on here, and people elsewhere have mentioned that they thought sounded better with MQA and listened to them through multiple equipment (some MQA compatible, some not) to try and understand where people got their impressions from. I think you should do the same.

4 Likes