What does the second and third “unfold” of MQA files actually do?

[gmgraves]

If the first unfold, generally done in software such as the Tidal app, Audnirvana, and Roon outputs files of either 88.2 or 96 KHz sampling rate, what do the second and third unfolds from the Renderer do? I called AudioQuest, and was dumbfounded to find that their technical guy didn’t know! He suggested I contact MQA in England. But before I do that, I thought I’d ask here. Thanks in advance for any help on this issue.

[mansr]

The first "unfold" decodes compressed data hidden in the lower 8 bits of the 24-bit MQA stream. The output has twice the sample rate of the input.

 

The second "unfold" is a trivial upsampling using one of 16 predefined filters as indicated in MQA metadata. We know how to decode this.

 

There is no third "unfold." Anyone claiming otherwise is either a liar or repeating lies.

[gmgraves]

1 hour ago, mansr said:

The first "unfold" decodes compressed data hidden in the lower 8 bits of the 24-bit MQA stream. The output has twice the sample rate of the input.

 

The second "unfold" is a trivial upsampling using one of 16 predefined filters as indicated in MQA metadata. We know how to decode this.

 

There is no third "unfold." Anyone claiming otherwise is either a liar or repeating lies.

Thanks. The notion of a “third unfold”, I got from the MQA website. They talk about it but don’t say anything other than that the renderer does the second and third unfold and gives the listener the audio that is exactly what the engineers and producers heard from the studio playback. 
So what you are saying, if I understand you properly, and I respect your knowledge, is that the Core Decode is really all one needs for MQA playback.

[mansr]

4 minutes ago, gmgraves said:

So what you are saying, if I understand you properly, and I respect your knowledge, is that the Core Decode is really all one needs for MQA playback.

That is correct.

[gmgraves]

3 hours ago, mansr said:

That is correct.

Thanks again. 

[Norton]

8 hours ago, mansr said:

The first "unfold" decodes compressed data hidden in the lower 8 bits of the 24-bit MQA stream

Other than unpacking compressed data does the “first unfold” apply any proprietary DSP as well?  In other words is there a reason why you might anticipate that it would sound different from a straight  24/96 file?  

 

I also note that some MQA files stay as “24/48” after the initial decode, what’s going on (or not going on) there?

 

Thanks.

[mansr]

3 hours ago, Norton said:

Other than unpacking compressed data does the “first unfold” apply any proprietary DSP as well?  In other words is there a reason why you might anticipate that it would sound different from a straight  24/96 file?  

In cases like this, it's useful to look at the encoder and decoder as a single black box. What comes out may well be somewhat different from what went in. How much of the processing resulting in that change happened in the encoder compared to the decoder isn't really relevant, though one can reasonably assume that the encoder is doing more work since it doesn't have to run on a puny XMOS processor.

 

3 hours ago, Norton said:

I also note that some MQA files stay as “24/48” after the initial decode, what’s going on (or not going on) there?

Simply put, you've been double-scammed. A lot of modern pop recordings are mixed and mastered in 24/44.1 or 24/48. The sane and honest thing to do would be to leave well enough alone and deliver these as plain FLAC. That wouldn't involve MQA though, so they've come up with a completely useless scheme whereby the encoder compresses the low 10 or so bits into 8 which are put into the bottom bits of the MQA stream. Above this they add some noise so the file played without a decoder becomes a little lower than CD quality. The decoder removes the noise and unpacks the compressed data, yielding a file that is almost as good as the original. In these cases, the FLAC-compressed MQA comes out a little larger than using FLAC directly without MQA.

 

In short, MQA gives you lower quality at higher bit rate, then makes you pay for the privilege.

[daverich4]

20 hours ago, mansr said:

The second "unfold" is a trivial upsampling using one of 16 predefined filters as indicated in MQA metadata. We know how to decode this.

For some reason I vaguely remember reading a long time ago in this thread that filters 9-16 are actually just duplicates of 1-8? Or, I could have just dreamt it. 

[mansr]

3 minutes ago, daverich4 said:

For some reason I vaguely remember reading a long time ago in this thread that filters 9-16 are actually just duplicates of 1-8? Or, I could have just dreamt it.

The value in the file is 5-bit number, so the possible range is 0-31. In practice, renderers only implement 16 filters, treating 16-31 as duplicates of 0-15. Examining various MQA files, I have only ever come across about half of those 16, and some are much more common than others.

[esldude]

Reminds me of Nixon hosting Khrushchev.  Looking at the apparent rocket launcher tail-lights on the big fins of a 1959 Cadillac, Khrushchev asked, "what do those do?".  The somewhat puzzled Nixon, looked at them, and said, "Those?....They sell Cadillacs."  

 

I'd assess the same level of importance to any of the unfolding in regards to MQA. 

 

[PeterSt]

20 hours ago, Norton said:

I also note that some MQA files stay as “24/48” after the initial decode, what’s going on (or not going on) there?

 

This is what XXHighEnd would show you, when it is so. It is my understanding that the current "standard" (a convention) depicts that there's always an upsampling step in order when the file is under 88.2 sampling rate. I (XXHighEnd) don't comply to that ... (and you (XXHE) would upsample yourself anyway, per your own means and up to the rate your DAC can handle).