It is serious but if you were joking, I’m happy to laugh 
1 Like
You can actually see what is missing here. Hearing is a different story, but just knowing that was enough for me to ditch the MQA. 
1 Like
Who can hear over 18-20 kHz?
I think only children.
How common, on Tidal, are these “16/44.1 MQA” tracks (actual effective bit-depth = 13 bits) as opposed to “24/44.1 MQA” (actual effective bit-depth = 16 bits)?
I thought that “16/44.1 MQA” was only for MQA CDs and that, for streaming, MQA files were supposed to be “24/44.1” or “24/48” (i.e., have an actual bit depth of 16 bits).
My dog can I won’t do him the disfavor to cut off his favorite frequencies when he sits down with me for a nice listening session… 
4 Likes
I don’t know lads, why don’t you start your own threads on it?
I’m flagging all this crap as off topic
SQ wise MQA is an off topic by design! 
1 Like
You realize this is the same argument that was made for MP3, right?
3 Likes
Thats as maybe… But in this case it’s true… Show me what’s missing, sounds sublime here…
You are worrying about the dip? Which just states it was a CD to start with?
Where do we get the notion that 16/44.1 MQA has an effective bit depth of 13 bits? Or that 24/44.1 has an effective bit depth of 16 bits?
Sure. I keep forgetting, everyone’s not as ancient as I am. Here’s a bit from howstuffworks.com
To make a good compression algorithm for sound, a technique called perceptual noise shaping is used. It’s “perceptual” partly because the MP3 format uses characteristics of the human ear to design the compression algorithm. For example:
- There are certain sounds that the human ear cannot hear.
- There are certain sounds that the human ear hears much better than others.
- If there are two sounds playing simultaneously, we hear the louder one but cannot hear the softer one.
Using facts like these, certain parts of a song can be eliminated without significantly hurting the quality of the song for the listener. Compressing the rest of the song with well-known compression techniques shrinks the song considerably – by a factor of 10 at least. When you’re done creating an MP3 file, what you have is a " near-CD-quality " song.
Look familiar? I suppose Meridian would say, “near studio-master quality”.
And a skeptic’s view of audiophilia’s rejection of MP3:
Using this technique, some of the audio has been removed. Fortunately, you probably won’t mind, since the parts removed were the ones that your ear would probably have screened out anyway. Any serious audiophile, will of course hear the difference, but that’s why MP3 is called “near CD quality” sound. But then, serious audiophiles claim to hear the difference between the earlier analog sound recordings and the new digital ones, noting that the digital managed to lose something.
2 Likes
Then we have my experience, and that’s another thing…
From any reference on MQA. See, e.g, Wikipedia.
On 24-bit MQA tracks, the 8 least significant bits of each sample are devoted to the MQA data, leaving an effective bit-depth of 16 bits.
On 16-bit MQA tracks, the 3 least significant bits of each sample are devoted to the MQA data, leaving an effective bit-depth of 13 bits.
On an MQA-aware DAC, the MQA data is “unfolded” to recover
- some of the ultrasonic content (above the Nyquist frequency)
- a choice of “custom” reconstruction filter (supposedly matched to the filter used in mastering the track).
You never recover the bit-depth that you sacrificed to encode that information.
1 Like
I wonder who wrote that? I would use Wikipedia for general commentary but wouldn’t rely on it for authority, certainly not regarding MQA (or anything too controversial), where there’s a campaign against it pushing their “truths”.
What happens if 13 bits sound better than 16?
Oh, for @#$&'s sake! That 16/44.1 MQA devotes the 3 least significant bits of each sample to the MQA data isn’t an “opinion”. It’s just how MQA works. Where did you think the MQA data was stored?
I’ll leave it to you to work out the implications .
3 Likes
My understanding is that if a 16 bit 44.1 kHz MQA file also has a 44.1 kHz ORFS (common among many Tidal new releases), then nothing is MQA encoded in the noise floor other than the few signaling bits needed to select the specific digital filter for 88.2 kHz upsampling. No bit depth lost, no ultrasonic encoding, no real first unfold, just upsampling.
So called MQA CDs that purport to have >44.1 kHz ORFS, however, may be a different story.
AJ
Jacques,
That’s not right.
“On 24-bit MQA tracks, the 8 least significant bits of each sample are devoted to the MQA data, leaving an effective bit-depth of 16 bits.”
See this block diagram for an accurate view of how 24 bit data is distributed and then recovered in MQA, for a specific example of a 96/24b signal. Furthermore, the 17 actual bits are noise shaped, giving an overall resolution of nearly 20bits. https://cdn.shopify.com/s/files/1/0321/7609/files/MQA-Block-Diagram.png?8802298321645544022
"On an MQA-aware DAC, the MQA data is “unfolded” to recover
- some of the ultrasonic content (above the Nyquist frequency)
- a choice of “custom” reconstruction filter (supposedly matched to the filter used in mastering the track)."
No, it recovers all of the available ultrasonic content above Nyquist. The only restriction to that is due to the (gentle) attenuation of the filter itself . However that’s true of all anti-alias filters. If you tell yourself that you need to preserve the inaudible ultrasonic frequencies exactly, then use minimally a 192 kHz sample rate and choose a filter flat to 96kHz.
Remember the basic properties of audio signals. They roll off in amplitude at 1/f and don’t extend beyond about 70-80 kHz, rarely even going above about 50 kHz. So anyone arguing that you need to preserve well out beyond 90 kHz doesn’t understand the signal.
“You never recover the bit-depth that you sacrificed to encode that information.”
You sacrifice noise but not information. Reading the published papers on MQA, the encoded bit depth depends on the noise floor of the signal itself (which is analyzed before encoding). It’s also known how far below the noise floor humans can detect a signal and the MQA shaped noise allows for that. In other words, you wouldn’t hear anything different if you played back the original 24b signal because the bottom bits are all noise from the recording (mikes, preamps and hall noise). That’s the principle of MQA design. This is one of the original papers from Stuart/Craven but there are others that explain it too.
1 Like
About many things.
I am well-familiar with noise-shaping dither.
The block diagram you linked to confirms exactly what I said (see Fig 7A). 17 bits come out of the LF/HF splitter. But only the 13 most significant bits are losslessly encoded as the 13 most significant bits in the output PCM. The next 3 bits are “lossy”.
Following that are 4+4=8 bits of encoded MQA data.
I’ll be generous and say that the effective “lossless” bit depth (before noise-shaping") is 16 bits. A critic would say “13 bits”.
Why are we leaving this one alone? The whole discussion in the other thread (from which our comments were yanked) was about 16/44.1 MQA on Tidal.
Sorry, but everyone (including Bob Stuart) agrees that the compression block in Fig. 7A is lossy. It is just not possible to fit losslessly-compressed 96kHz data into the 8 LSBs of a 48kHz file.
I heartily agree that the ultrasonic “content” here is
- inaudible
- mostly electronic noise
(Obviously, you understand that if there really were audio content out to 70-80 kHz, you’d need to sample at more than twice that frequency to capture it. 24/48 MQA is only supposed to unfold to 96kHz, which would (lossily!) capture audio signal out to 48 kHz.)
You don’t sacrifice noise.
By truncating to 16 or 13 bits, you raise the level of the quantization noise.
- With 24/44.1 MQA, you lower the SNR from 24 to 16 bits (which, IMHO is perfectly fine).
- With 16/44.1 MQA, you lower the SNR from 16 to 13 bits (which is not fine).
This is not about the SNR of the original recording. I suggest you Google “quantization noise”
Dithering helps (especially noise-shaping dither). Everyone mastering a 16/44.1 audio file applies noise-shaping dither. But applying dither to a 13-bit signal does not get you anywhere close to high fidelity.
Again, I am perfectly fine with the proposition that 24 bits are superfluous and (provided you dither the 16 most significant bits), you can use the 8 LSBs of a 24bit PCM file for steganography, with no audible loss of quality.
I don’t think the same is true of 16bit PCM.
2 Likes
Whatever “available” means…