At this year’s HighEnd in Munich, I was asked from various sides what my opinion of MQA is. I had already published my assessment of MQA 2 years ago, but I am presenting it here again in light of current developments:
The question about the qualitative influence of MQA encoding is controversial. To do justice to that question, we have to cover a wider ground:
MQA is short for “Master Quality Authenticated” and is an encoding algorithm presented by the company MQA Ltd. in 2014. MQA Ltd. promises 3 things for its algorithm:
- Higher playback quality at lower data rates than was previously possible. In fact, MQA Ltd. promises HighRes quality at CD standard data rates. In particular, MQA Ltd. had, at least initially, described its codec as “lossless”.
- Improvements in sound quality by “de-blurring” the music i.a. by correcting errors inherent in the analogue/digital converters used in the recording process, especially the so-called pre-ringing (White Glove Treatment)
- Authenticating the music files and ensuring that the consumer only receives music authenticated by the artists and sound engineers.
The HiFi press was enthusiastic at first. E.g., Robert Harley, editor of The Absolute Sound has referred to MQA as “The most significant audio technology of my lifetime”. Only later on did it turn out that MQA Ltd.’s marketing claims were mostly unfounded. In particular:
- Promise of higher resolution at lower data rates
There are 3 types of audio codecs:
- Uncompressed and lossless codes (e.g. WAV)
- Compressed and lossless codes (e.g. FLAC)
- Compressed and lossy codes (e.g. MP3)
MQA is a hierarchical, compressed and lossy codec (type c.) incl. a kind of copy protection. MQA encodes music files through adaptive differential pulse code modulation (ADPCM) and bit rate reduction.
A MQA data stream is decoded hierarchically: in MQA licensed devices, depending on the type of license, in partial resolution (“1. Unfold” with software decoder) or full resolution (“2. & 3. Unfold” with hardware decoder). In non-MQA-licensed devices, the data stream will be decoded up to effectively 13bits only. Even in MQA licensed devices at 3. Unfold only up to effectively 17bits. The remaining bits are reserved for various “foldings” (see patent description here). The patent description is not particularly transparent (there is a “touchup” function – however that is supposed to work) and the people behind MQA have done little to clarify the exact workings of their codec. However, MQA effectively causes data loss and reduces dynamic range. Additionally, various analyses of HiFi bloggers suggest, that the MQA coding/decoding adds substantial distortion and noise to the resulting analogue signal (Archimago, XiVero, Golden Sound).
In Hi-Fidelity we normally prefer lossless codecs. Lossy codecs do have a role to play, particularly in applications where there is bandwidth limitation, like in mobile uses. In your home stereo there is normally no such limitation, so that there is no good reason to use a lossy compression.
As became clear over time, MQA Ltd.’s claim to reduction of bandwidth requirement is not tenable:
- Analyses of Archimago and Golden Sound suggest, that in CD standard, MQA file size is slightly larger than normal PCM in FLAC format despite reduced bit depth; in highres files data rates correspond to PCM data rates in FLAC format for the reduced bit depth
- the claim to “losslessness” is untrue and in any way and in contradiction to MQA’s own AES paper presentation of Oct. 2014.
- Promise of improved sound quality
How the “de-blurring” of music through ex-post filter correction is supposed to work is a mystery. MQA Ltd. avoids any explanation of its functional details. There are just claims around this including “neuroscience” as the rationale with the often-quoted value of 5µs threshold of human temporal auditory resolution. Modern recording sessions involve a multitude of A/D converters that never get logged. MQA cannot even know which A/D converters were involved in any particular recording session and accordingly cannot make any corrections to actual or presumed errors in these devices.
- Promise of authenticity of the music files
Various reports of artists on Tidal (e.g., Neil Young, Golden Sound, FredericV.), suggest, that their MQA files for which the MQA authentication was granted, do not represent, the files submitted by them. In fact FredericV showed that he could remove 30% of the data in his music files and they would still be shown as “authenticated” by MQA.
The reasons why MQA is nonetheless intensely discussed in HiFi circles has to do with the marketing promises by MQA Ltd. and the different interests of the parties involved:
- Massive advertising by MQA Ltd. with – at least partially – misleading claims (losslessness, touchup funktion, White Glove Treatment etc.)
- For the recording industry, MQA is yet another argument to resell their old archives. I am sometimes smiling at myself for having bought the same album up to 7 times in search of the best sounding version: 1. Original LP, 2. Half-speed mastered LP, 3. Digitally remastered LP, 4. CD, 5. Remastered CD, 6. SACD or DVD-A, 7. Highres download. With each new version, the record industry promised better sound quality. Sometimes that turned out to be true, sometimes not. This time they come up with MQA.
- MQA offers the recording industry a license test and thereby a method to contain the free exchange of music files. It is a belated solution for that long pursued search for a workable copy right protection.
- MQA offers MQA Ltd. license fee income, a) in terms of hardware licenses, and b) in terms of music licenses even for albums, for which MQA is no rights holder.
- For hardware manufacturers, MQA is an argument to sell their customers new hardware
It looks like a system in which everyone profits – only the customer does not.
Some albums do indeed sound better as MQA versions, after they have been newly remastered for MQA from highres source files. The improvement in sound quality is then, presumably, due to the new mastering rather than the encoding technology. Archimago has done a blind comparison of identically mastered highres MQA and PCM files with a pool of 83 audiophiles and musicians in 2017. Similarly, McGill University has conducted a double-blind ABX test, submitted to the AES in 2018. The result in both cases was that the test persons could not detect any difference in sound quality between the highres MQA and PCM files. However, most MQA albums available are not newly remastered but are simply MQA-coded CD-masters and sound like the old CDs, just with reduced dynamic range.
Summing up, MQA does not seem to offer any advantage to the music connoisseur in the context of his or her home stereo. MQA may, under certain circumstances, offer advantages for music listeners under circumstances of reduced data bandwidth, e.g. over the mobile phone or in the car entertainment system. At its core, the MQA technology is a very complex process to show that one can generate music files by bit rate reduction from 24bit to 17bit that are not distinguishable by the human hearing. This could have been achieved a lot easier by bit rate reduction in PCM and FLAC with similar data file sizes – entirely without license fees and costs.
In the context of file based music playback, MQA is actually only relevant for subscribers of the TIDAL streaming service. Tidal calls its highres files “TIDAL Masters” and so far they are all MQA files. TIDAL has announced in April 2023 that it will not add any further MQA albums to its streaming service, but will rather provide new highres files in FLAC format. In April 2023 MQA Ltd. has gone into administration (the UK equivalent to filing for chapter 11).
© Alexej C. Ogorek
References:
- MQA website: Homepage | Experience the best sound with MQA | MQA
- MQA patent description: WIPO – Search International and National Patent Collections
- Golden Sound review: I published music on Tidal to test MQA – MQA Review – YouTube
- Archimago overview: MQA: A Review of controversies, concerns, and cautions – Reviews – Audiophile Style
