MQA

Guys, guys, all this talk of stroking is leaving me confused. Should I, or should I not be wearing pants for this conversation?

 

But isn't this a theoretical argument as well?

Either way, the above is true only if also assuming a perfect analogue brickwall filter. Which is never true. For instance, @baldr is positively appalled by brickwall filters.

and

So it seems 44.1 kHz sampling rates raise real engineering conundrums. From where I stand, there are three ways to tackle the engineering tradeoffs:
1. Use NOS and an analogue brickwall filter, which seems to degrade the "perfect reconstruction" of the analogue waveforms
2. Use NOS and go filterless (à la Metrum), if you assume that there is little content of importance above 12 kHz (esp. considering the one of the highest fundamentals for instruments is at about 8 kHz) and that aliasing is overall a minor issue.
3. Use oversampling, which may introduce its own engineering porkies (unless you have access to some burritos).

In all of these cases, a higher sampling speed seems to decrease the engineering constraints. But what's more, my understanding is that in other domains (like imaging or networking) the general rule is to use sampling speeds at 1.5 or 2 times the Nyquist frequency (i.e. applied to audio, something approaching 96 kHz given the common 20 kHz band limit). Overall there seems to be an argument that slightly higher sampling speeds would be beneficial for audio playback applications...

 

Essentially what all modern DS D/A IC's do internally AFAIK. Except multiplying by higher ratios

 

True, though one should be careful with possible confounding factors. Will be playing with these things myself once (hopefully) I get my hands on @Torq's DBT box.

My understanding is that oversampling usually relies on approximations (e.g. Parks-McClellan), which requires dropping all original samples and replacing them with approximations. This is one of Schiit's selling points, and seems like a genuine engineering complication to me. May be wrong, though...

 

I strongly suspect that Audacity's downsampling sucks ass. One day I might make two vinyl rips, one recorded at 44.1 and one at 176.4 and compare both again. I'm sure that letting the ADC do the downsampling would be a better option than downsampling in software - at least Audacity, but I'm not convinced that they will sound the same. The Audacity downsampled one sounded harsh, to the point where I was wondering wtf was going on.

 

No solid example, no. Just this comment here:

 

Downloaded Adobe Audition and did some 10 minute tests. Take this as a "first impression" and not a definite answer. Only have the GO450 and Leckerton on hand right now, since my Gungnir and Rag are at the speaker rig in the middle of the room 4m away from my computer. All Sabre DACs have sounded better at high sample rates to me, so it's likely that it would be harder to tell apart (and more fair) using the Gungnir.
Down-sampling using Adobe Audition (quality set to 100%) is much better than the best quality using Audacity, but still not quite there. When simply downsampling (both with and without the Pre/Post filter), it sounds notably more closed-in and more digital (harsher, grainier) than the source file (which also sounds digital, dynamically flat, closed-in, grainy and generally just sounds like shit compared to the real thing). Using Adobe Audition to upsample the file back to 176.4kHz actually helps with the GO450, but I can still tell the two files apart (using foobar ABX comparator plug-in to avoid placebo).
The differences here are of an entirely different nature than the difference between 16 and 18/20/24bit. Hard to say which one has a bigger effect. The 16bit file (using triangular dither) sounds different from the 24bit source file, even if the source file probably doesn't get anywhere close to 16, let alone 24, effective bits. The 16bit file sounds lifeless compared to the original. Dynamically flat with lacking plankton (reverb doesn't linger as long).

 

So my question would be...

Let's say I have a ripped CD's PCM wave file which is 16bit 44Khz.

Would it be better for me to use some high quality Audio software(Adobe Audition or something) to upsample the PCM wave file to 16bit 176.4Khz file and transfer to my Sony ZX2 DAP for playback? This will no doubt increase the file size by alot.

 

Oh lol, my comments weren't with regard to MQA at all. But I do think it's possible that MQA could sound better than regular 16/48, but also probably worse than 24/96. That doesn't mean that I will go out and buy Hi-Res or MQA unless I know the mastering was much better than the CD version. 100% of my music is 16/44.1.

 

I scrolled past a lot of this rather bickery thread as I am still groggy, but I do know a good deal about sample rates and bit depth in both theory and practical experience.
When it comes down to it, MQA is trying to solve a "problem" that doesn't really exist by bizarre means. I just don't see what's "broken" with PCM that requires DSD or MQA. I'd much rather have lossless 16/44.1 than a lossy and noisy 24/?? file, even if it is noise-shaped.

If you want to know how different sample rate converters perform: http://src.infinitewave.ca
In the last few years most DAWs have really stepped it up. Even Audacity 2.0.3 looks good, SoX does a good job. iZotope is still killer.

MQA will end up in the dustbin of audio history, as a footnote next to every other format that was going to fix everything, and software devs and hardware manufacturers will be bothered to support MQA by the people who manage to drop four figures on MQA music before the format tanks.

The production chain is crucial. Mixing and mastering need to be done carefully. But any change in those processes are exponentially more noticeable than any change in file format.

The MQA people have been putting down big, big cash to make sure every audio writer is writing about their process, doing Q&As, flying out to demonstrations, granting it legitimacy. And a bunch of those writers are probably buying in, genuinely convinced. But there is the classic conflict of interest that new formats give them something to write about, and a new type of gear they must review because now the readers want to know (because they told them about the format in the first place, and how it is "must have"), and now the makers have more money for advertising, etc.

 

It's funny how this T3 MQA article has a picture of a Sony NWA-ZX2 walkman that does not support MQA.
http://www.t3.com/news/t3-explains-what-is-mqa-audio

 

On the contrary, the problem is real- the MQA people don't have as many wheelbarrows full of cash as they'd like, so they'd prefer if every stage of the chain paid to use their format- that's pretty much what it boils down to. Beyond that, it seems to be of debatable value, of course.