Tidal Masters MQA - Why the Frequency Spectrum Gap?

I'm learning about the new MQA audio format and am playing around with it in the Tidal Masters service. I'm looking for an explanation of why there's a big frequency spectrum gap in a lot of tracks.

► SETUP

• Tidal Masters MQA Service
• Audio inferface set to 96kHz 24-bit
• MQA passthrough disabled (software decoding in Tidal)
• Recording in Adobe Audition to 96kHz 24-bit track

It doesn't make any difference to my life :-) but it's just something I noticed and became curious about (that's what I'm into). Is this part of the mastering process by certain record companies? Is this a limitation of the software decoding? I just don't know.

I'm sure there are many people online who know much more about this that I do and might be able to help me with this one. I'm sure it's a simple explanation or an error in my usage or understanding.

Do you use Tidal? Do you like the idea of Hi-Res / HD / MQA audio being offered on Tidal? Is there really any difference??! I'm interested to hear your thoughts in the comments.

========================

► FIND ME ONLINE:

► GEAR USED IN THIS VIDEO:

Audio Interface: RME Fireface UFX
Editing: Adobe Premiere Pro
Thumbnails: Adobe Photoshop
DAW: Adobe Audition

[Grassy_Gnoll]

We can stream 4k video, NP, but hi-fi music? No, no, no… that's too much bandwidth! smh

[MansRullgard]

Note that the frequencies above the gap are a mirror image of those below it with a slight downward slant. This is because the weak upsampling filter used by MQA lets these images through rather than suppress them as it ought to. Nothing above that gap actually existed in the recording. It is all fake. The 2L recording originally has a high sample rate, and MQA preserves it reasonably well up to 44.1 kHz. If you did a playback and (analogue) capture through a certified MQA DAC at 176.4 kHz (or higher), you'd see similar mirroring in the 44.1-88.2 kHz range.

[markfischer]

Here's the answer. MQA enploys a system its developer calls audio oragami. Oragami comes from the Japanese art of paper folding. The out band signal is segregated during th recording process, reduced in level, and then frequency downshifted. The process is repeated a second time to arrive at the final encrypted signal. The process is reversed on playback to arrive at the original signal. If there is an error anywhere in the process you will get gaps or overlaps in the frequency axis or the amplitude axis. This is how the notch arose from a signal processing failure where the up shift was too great on playback or the downshift was insufficient on recording encryption.

I have 2 comments. As humans cannot hear sounds above 20 khz, any system that records and plays back those frequencies is of no usable benefit and can only potentially produce distortion in the audio pass band or burn up tweeters. The proof that it is inaudible comes from the Fletcher Munson curves shown in full. 20 khz is where the threshold of hearing curve, the softest sound that can be heard intersects the threshold of pain curve.

The system appears to violate the Shannon Nyquist theorem but frankly I'm not interested to investigate how it appears to work. My hunch is that digital interleaving comb filtering is used to time multiplex the signal to get it into the audio passband.

I'm not an audiophile, I'm an electrical engineer and this has absolutely nothing to do with my day job.

[gastonpossel]

That is not a gap. The original file has information up to 24 kHz (probably a 48kHz master), and the spectrum has been mirrored due to bad upsampling.

[georgemaras]

I deed some experimentation because this gaps look familiar, they happen during upsampling and are filter depended it seems that they upsampling everything (thats not available originaly) in high resolution and then encoding it to MQA hopping that sounds better. Not all recordings are high definition but there is a little or no benefit at all doing this. Of course this can create the illusion that the recording is high resolution (visually) but without any audible benefits. Also there is a fact that are perfect sounding 44/16 tracks far more (audiophile) than many 192/24! Dont forget that the (golden ears) almost always use as reference recordings of the 60's, 70's and even 50's to judge their systems! The (quality) thing is very complicated and a good format can quarantee only that if something is good stays good without degradation.

[rarelycomments]

Hi, unfortunately the answers are quite technical but (almost) everything you need to know about the technical aspects of MQA are available in AES Convention Papers 9178 (free) and 9174 (requires membership/payment). Available on the AES website

[Jungmin_Seo]

I have no idea why people really care about ultrasonic area that people can't hear.
People should spend and focus on 20~20khz that we can actually hear!

[mechanickiwi]

The stuff after the dip(right) is added by the mqa process.

[aaronwong]

Quite a few of the modern high resolution releases are done as 24/44.1, however Tidal's implementation of MQA stuff all comes through at 24/88.2 or 24/96. My guess is that you're seeing the result of the original file's filtering.

[paulccng]

Are all tidal music shows the same gap? Or only a few particular songs?

[bobbecker]

Looks like a LPF was applied to the track. Not all tracks have this, however if you dig harder you might see that these tracks are from the same studio and engineers

[sudarshanmaharaj]

could this be a type of watermark similar to watermarks used by universal music. Universal use the watermark between 1 & 3.5khz, maybe tidal or MQA is using theirs at 20khz as this is not audible to most people.

this article goes into more detail:

[xiaobaozha]

Audirvana on MacBook Pro, to Mytek Brooklyn (Full hardware decode) to either various headphones or Naim pre/power to Totem Arro. MQA sound quality variable as per any mastering on any format. With good/great recordings my system has never sounded better, particularly with acoustic music, especially piano, saxophone and acoustic guitar.

The origami folding is not the only process involved with MQA, DSP is also part of the process chain to create the ‘deblurring’ effect as well as specifically using minimum phase filters.

MQA is a disruptive technology as it does not follow the conventional bit rate/sample rate hype train (amongst other reasons), hence the divisive /polarised noise in comment threads...

[philosophiaentis]

There is a reason why there is no MQA encoder and decoder available for the customers and it is becouse it is a fraud. Any one would be able to see the data loss and that is what MQA wants to stay unknown.

[juliaset]

The screen cap is a little dark so I might be mistaken, but it looks like the 2L and the Pink Floyd were not Master.  The Master tracks have an M with a square around it.  I did see it with the Beyonce, but not the 2L or the Pink Floyd. That may explain the lack of the 22K gap.  I’m also a little confused as to why some tracks light up my Meridian Explorer MQA blue LED, but don’t light up the 96K or 192K LEDs.  So some of these may just be CD’s with the MQA flag set.  I have only been listening to the Master tracks using my Explorer for a few days, but it does seem the improvement varies from very good to MIA.

[Roamany-Jones]

Please read Paul Millers findings of upsampled CD files passed off as Hi-Res files in Hi-Fi News and RR. This magazine tells you every month about what you have found out by reviewing the latest download releases. The conclusion he drew from The Pink Floyd album is it is a true Hi-res file, but the synthesisers used on the album are limited to 20khz.

[Vince]

Hi. May I know if offline versions of mqa files (when played in offline mode) experience any downgrade in sound quality as opposed to playing mqa in online streaming?

[aaronwong]

Yeah, I'm pretty sure that's what you're seeing. The stuff that's upsampled from 24/44.1 to 24/88.2 is showing the filter.

[crazyprayingmantis]

Great all we need is another fucking format!!!

[BTStudioAudio]

Hi TDCatTech,

Maybe it's to do with the way MQA folds the higher frequency's (two octaves above the 24kHz) to the lower bits.The range 48 to 96 is folded back and also the 24 to 48 range.
I'm guessing that this folding is going wrong (decoding) in the examples you are giving.

The peak in your example just under the 48kHz is equal strange....