I stumbled upon this elsewhere. I have a book I should be reading but . . .
Tidal ‘HiFi’ is NOT lossless - GoldenSound
TLDR: Tidal ‘HiFi’ is not lossless. It just streams the MQA version of tracks with some metadata removed to prevent most DACs from recognising it as MQA and limits streaming to 16 bit.
Not all streams from Tidal HiFi are MQA. I would say that fast majority aren’t MQA.
How can you tell which are and are not?
If as Paul is quoting, Tidal, before streaming, is stripping the MQA information out of the stream, which would cause Roon to see it as standard FLAC; you would have no way of knowing if it was a neutered 44.1 MQA or Redbook CD.
What is interesting is that in the new marketing material Tidal does not reference source content resolution anymore, but, instead puts up streaming rates.
This was reported last week, if you want true FLAC you are better looking at Qobuz.
Thank you, I hadn’t seen that.
Thanks for sharing
A few years ago, Tidal was launched as “lossless” when Spotify was not. And I think they did gain the interest of many listeners focusing on sound quality. But ever since they have entered the partnership with MQA, “lossless” is not longer their priority. They are prioritizing “shiny marketing stickers” (MQA, Dolby Atmos…) for a while now. That does not take away some of the qualities of Tidal and of the sound quality of some of these formats (never heard Dolby Atmos), but lossless is not longer their priority.
I can understand their approach. If shiny stickers better sell than lossless… well go for it. But what bothers me is that, in this process, they have betrayed their honesty and promise. They still market their HiFi offer as lossless, and as documented by “GoldenSound”, this is not the case (because of MQA…)
The Golden Sound article quoted by the OP gives all the gory details.
Gory Details indeed.
I suppose that answers my question about the CD quality of the mid price band (CD only subscription), its a butchered MQA if the file is Master Sourced
My problem is I am stuck as Tidal is my (in South Africa) only CD quality source without having a “lash up a DAC USB cable” to use Apple Music . At least the ALAC files are probably CD Quality
Be interesting to see any MQA fans wading in here … 
Let’s not 
We’ve all had enough of that.
But I am sure they will be tempted , I agree the MQA war is not a nice read
It’s impossible to have a discussion about the detail of streamed digital music, when there is no standard terminology. The suppliers deliberately rename and obfuscate in order to make unsubstantiated claims and avoid easy comparison. I’m not attacking anybody personally; it’s standard marketing technique to create the impression of unique advantage. And then there’s the standardised terminology, that I can’t understand anyway,… like “upsampling”; which I think always means the same thing, but I cannot imagine why it makes any difference to increase the bit rate after you’ve reduced it.
If the original master for an album was 24-bit / 96 kHz, is the 16-bit/44.1 kHz “CD Audio” version “lossless”?
If the original master for an album was 24-bit / 96 kHz, is the 16-bit/44.1 kHz “CD Audio” version “lossless”?
Exactly. The “losslessness” conversation is poorly understood and poorly formulated. A normal PCM file is only lossess, bitwise, for a short period of its history. It’s lossless from the output at production to the input to a user’s system (assuming no loudness normalization by the streaming service). All of the processes on both sides of that are lossy or, at minimum, change the bitstream. Converters are lossy, DSP processes including sample rate conversion change the stream. There are no other accepted criteria of losslessness, e.g. dynamic range, waveform shape, etc.
In that sense, MQA is exactly as lossless as PCM. Both format types are encoded in FLAC (usually) from production output to reception for playback and both are lossless. I believe the sample rate changes in MQA, from fully decodable to non-decodable, are done by the rights holders rather than by the streaming service but don’t know that. In any case, the reduction is what’s intended as a source file.
Interesting question. Is digital lossless at all? Is anything recorded (analog or digital) lossless (compared to live event) ? We need to provide a bit of context, else there is no possible conclusion.
In the 70’s and 80’ and early 90’s no one was talking about “lossless”, I think. “lossless” became a thing when “lossy” also became a thing. And I would think that the explosion of mp3 in the late 90’s made of all us aware that there existed such thing a “lossy”. “lossless” or “lossy” are terms that are used to describe the encoding of sound information, e.g., mp3, flac… typically to reduce the amount of data necessary to store the information. This encoding is independent from the sampling frequency and quantization.
So, within this context, lossless does not apply to your question, in the most accepted form of language. This is not how it is used to describe the loss of information of the the typical psycho-acoustic based compression algorithm, such as mp3, ogg, aac…
To your question tough, if the acoustic form fits within the possibility of the PCM encoding of 44.1kHz/16kHz, i.e., with a bandwidth limited to 20kHz, and a dynamic range limited to 96dB, there is no extra information to store in the 96/24. The original acoustic waveform could be reconstructed with either version. If there is content above 30kHz for instance, there yes the conversion from 96/24 to 44/16 will cause a loss.
According to “GoldenSound”'s measurements, MQA introduces noise beyond that of quantization. So MQA “lossyness” is not like the limitations of PCM, it comes on top.
The ideas in MQA need to be understood. The algorithm is designed around a clear definition of the dynamic range and frequencies that are audible to humans and that exist within music data. That includes a thoroughly measured and evaluated knowledge of the noise floor in recordings. It also includes lengthy listening tests in pro studios to confirm the audibility or inaudibility of any noise in the final output. Some of the measurements being flogged by certain individuals are based both on misunderstanding what accurate measurements would be, and sometimes refusal to concede that a noise source at 90kHz and -100dBFS would always be inaudible and would be utterly masked anyway by normal audio data in the file.
PCM processing doesn’t have an established noise floor definition either; it only states the output bit depth. In many cases with PCM, noise shaping is used, and modulation noise can be seen and measured in the output.
But let’s not get into another lengthy go-round on the design of MQA, though.
In conventional digital terminology the terms “lossy vs. lossless” usually apply to compression algorithms (typing “lossy vs. lossless” into Google backs this up). The meaning is pretty cut and dried, if you can’t reproduce the original data from the compressed version then it’s lossy.
Over time, in different domains, arguments have been raised that if you can’t notice the losses then this might also be termed “lossless” in some way. I first saw this when some marketing leaflets, and people I worked with, started to refer to JPEG 2000s lossy compression as “visually lossless”. My counter to this is that it depends who, or what, is doing the looking. Put more generally calling something “lossless” just because a particular observer didn’t notice the loss feels like a deliberate attempt to move the terminology gateposts.
Altering the bitstream shouldn’t be conflated with the “lossy” debate, many DSP/upsampling processes are reversible and wouldn’t be considered lossy. Much of what you list is applied to the original stream on a “per-performance” basis, e.g. room correction, or upsampling via HQP.
the terms “lossy vs. lossless” usually apply to compression algorithms (typing “lossy vs. lossless” into Google backs this up). The meaning is pretty cut and dried, if you can’t reproduce the original data from the compressed version then it’s lossy.
The problem is this: lossy codecs like MP3 are a compression process applied at the output of the production file. So there is a direct bit comparison available to look at before/after compression. MQA on the other hand affects the filtering and processing from recording through to playback. So there is no direct point-of-comparison to the bitstream before/after compression. It’s conceptually different and has to be compared to the original by looking at dynamic range, waveform, and audibility. Even the MQA 44.1k that started as 44.1k files are often redone from master tapes, and not just encoded.
DSP/upsampling processes are reversible and wouldn’t be considered lossy.
Most DSP processes aren’t strictly reversible, especially close to the noise level. Upsampling involves anti-imaging filters so isn’t lossless. Careful design keeps it as lossless as possible.
If you didn’t already know - would you be able to tell either way?