AudioQuest adds MQA Support to Dragonflies via firmware

[Fokus]

21 hours ago, mansr said:

No matter what MQA tries to claim, "rendering" to higher sample rates than 96 kHz is upsampling. Because of the leaky filters used, the upsampled signals contains strong images of the 0-48 kHz frequencies

 

This page http://www.pinkfishmedia.net/forum/showthread.php?t=198428&page=42 contains spectra of what appear to be '192kHz' MQA files, played back through an MQA DAC and then recaptured by a 192k ADC. Nearly all of them have clear signs of leaky upsampling above 48kHz.

 

[Fokus]

Sorry guys, but no.

 

MQA's deblurring is first and foremost the use of short, shallow, leaky anti-alias and anti-imaging filters during music production (downsampling from an ADC running at max rate to 2x rate for storage and distribution) and replay (upsampling to a particular DAC's max rate). That some or all of these filters are minimum phase is not that relevant, since the filters are so short.

 

The rationale is to marry a 2x channel with a 4x or 8x or 16x or whatever x impulse response width. This has been spelled out in MQA literature from day one.

 

 

On top of this, and for low rate original recordings, they have a patent for an all pass filter with a large group delay centered on the ADC's transition frequency. It is not clear if this filter is actually being used.

 

 

[Fokus]

To literally move the preringing of the original filter to after the impulse. A bit bonkers.

[Fokus]

14 hours ago, crenca said:

 

 

which is what a min phase filter does no?

 

Sort of.

 

In our context of audio replay a Meridian-style minimum phase filter's treble rolloff attenuates the recording-side's anti-alias filter's ringing (pre and post),

and imposes its own (post-only) ringing at a lower frequency. (Sidenote: an Ayre, or whomever, -style MP filter does not attenuate enough and does

nearly bugger all for the original ringing, but that's another story.)

 

The all-pass filter in MQA's patent does not attenuate anything, it only delays the signal at the original AA filter's transition frequency, literally moving that one's ringing around. Mind, I have no idea that this is actually being used on original low-rate recordings when preparing them for MQA.

[Fokus]

You need something doing the core decoding (= unfolding from 1x to 2x) first. AFAIK only Tidal, set up properly for doing this, applies.