Columns Retired Columns & Blogs |
MQA: Some Claims Examined
"The fundamental problem of communication is that of reproducing at one point either exactly or approximately a message selected at another point."Claude Shannon
Footnote 1: See J. Robert Stuart and Peter Craven, A Hierarchical Approach to Archiving and Distribution"; the description section of the relevant patent; and a Q&A here.
Since its announcement at the end of 2014, Master Quality Authenticated, the MQA encoding/decoding system, has spawned outspoken criticism. Some of the more thoughtful negative reactions have come from engineers such as Dan Lavry, Bruno Putzeys, and Daniel Weiss. Others have been expressed by manufacturers of digital products: the late Charley Hansen at Ayre Acoustics, for example, along with Jason Stoddard and Mike Moffat at Schiit Audio, John Siau at Benchmark Media Systems, and Jim Collinson at Linn Products. Some have been audio writers: Doug Schneider, at SoundStage!, and Paul Miller and Jim Lesurf, at Hi-Fi News. Most vociferous have been anonymous website posters. As Jim Austin remarks in his examination of MQA's decoding of impulse-response data elsewhere in this issue, "the nastiness online is unprecedented."
Typical of the nastiness are the following, taken from the Audio Asylum Web forum: "Nothing more entertaining than seeing Atkinson and Austin continuing to bury their reputation via MQA"; "MQA is a wholesale sellout"; and "If the very experienced editor of the numero uno in hifi mags (as JA is so fond of telling us) is in fact so easily duped, and even worse refuses to admit it and continues to praise what in [another poster's] view is a con, then I don't think he's worthy of much respect."
Am I being duped? I admit that, when I studied Stuart and Craven's 2014 AES paper introducing the concept (footnote 1), I found it theoretically elegant. (Peter Craven is an engineer whose ideas I have grown to respect over the years.) Subsequent demonstrations were convincing. In December 2014 I wrote, in an online news item, that "MQA will allow what appears to be true high-resolution audio to be delivered over the same Internet pipes over which music lovers currently experience at best CD-quality audio from Tidal or Qobuz."
The reduced data rate is due to the fact that, with files sampled at 2x, 4x, or 8x the baseband rate of 44.1 or 48kHz, the information in the first, second, and third ultrasonic octaves can be encoded and packed below the music's baseband noise floor in a 24-bit container. This "audio origami," as MQA calls it, results in a much smaller file than the hi-rez PCM equivalent, yet when the file is unfolded, the resolution and bandwidth of the original file are claimed to be preserved. I was initially skeptical of how the ultrasonic octaves could be separated from the baseband data without introducing artifacts, but studying the digital-signal-processing literature did seem to indicate that it was possible. And it is, of course, how digital-domain loudspeaker crossovers are implemented. In May 2016, I examined this claimed reduction in file size with some of my own 24-bit/88.2kHz recordings for which MQA's CEO, Bob Stuart, had sent me MQA equivalents. Not only were the MQA files smaller than the original PCM files, they were smaller than the uncompressed 16/44.1 CD version of the recording, so that claim by MQA is incontrovertible.
In that 2016 article I also examined the next claim made for MQA: that the encoded ultrasonic data can be encapsulated and buried in the least significant bits, beneath the original recording's analog noise floor. That claim, too, appeared to be correct, though there was a rise, undoubtedly inaudible, in the undecoded noise floor at close to half the sample rate that may well have be symptomatic of MQA's encapsulation of the ultrasonic data.
The most contentious claim made for MQA is that it can compensate for the time-domain errors of both the original A/D converters used to make a recording and the D/A converter used to play it back. This results in the complete recording/playback chain having an impulse response equivalent to a few feet of air, and temporal resolution of the same form and order as that of the temporal sensitivity of the ear-brain. Jim Austin examines the D/A aspect of this claim in his article in this issue, and other writers and I have conjectured that this time-domain compensation correlates with the always-superb sound quality we have heard from decoded MQA files.
However, optimizing the digital chain's behavior in the time domain involves using a very "short" antialiasing filter at the A/D conversion, and a similarly "short" reconstruction filter when the digital data are decoded. The more you constrain the data in the time domain, the less you can do so in the frequency domain. These filters are therefore "leaky," as you can see in the measurements accompanying the Aurender review in this issue, and will thus allow ultrasonic images to fold down into the baseband. Such filters are not new. Wadia's DigiMaster filter and Pioneer's Legato Linear filter both date back to the end of the 1980s, and the Listen filter in Ayre's D/A processors is also short and leaky, with a slow rolloff above the audioband. On the A/D side, converters from dCS and Ayre, both of which I use to make my own recordings, offer a switchable slow-rolloff antialiasing filter.
Does the filter's inherent aliasing degrade sound quality? That will depend on every recording's spectrum, and will be examined by Jim Austin in a future article, along with the question of whether MQA encodes ultrasonic information lossily or losslessly.
Other criticisms of MQA involve its implications for the recording industry, for manufacturers of audio products, and for consumers. I will examine those in next month's "As We See It."John Atkinson
Footnote 1: See J. Robert Stuart and Peter Craven, A Hierarchical Approach to Archiving and Distribution"; the description section of the relevant patent; and a Q&A here.