AudioQuest adds MQA Support to Dragonflies via firmware

[miguelito]

16 hours ago, ShawnC said:

That added absolutely nothing 

Disagree. @mansr is just saying he's actually looked at the detail and understands what is happening. There's plenty written by him to look into if you so wish.

[miguelito]

17 hours ago, revand said:

 

Dear Gordon,

 

I read months ago that even if the Dragonflys spec sheets state 24bit/96kHz (USB) capability the DAC chip is capable much more.

My question: What is the maximum resolution the Dragonflys are able to achieve when making the unfolding of the file sent by TIDAL Masters in Flac format after the MQA process?

 

Many thanks

Andras 

The DF specs compared:

 

http://www.audioquest.com/wp-content/uploads/2016/04/draygonfly-spec-sheet-darktheme.pdf

 

The ESS 9010 and 9016 (Black and Red respectively). Both support up to 384KHz sample rate. Given what I see with various MQA files, it seems that the unfolded MQA data is upsampled by the ESS DAC to some rate indicated in the file itself. Some of Joni Mitchell's albums render at 96KHz (so no upsampling) whereas some of the 2L albums render at 384KHz, at least this is what Audirvana indicates.

[miguelito]

1 hour ago, ShawnC said:

More processing - are you talking about the folding/unfolding process?  From what I understand MQA takes that Master, eliminates the faults that were inherent with ADC back then (gets rid of some noise) then adds there filter (which people hate) as part of the process.  That in a nut shell is the way I see it.  The deep technical notes is way beyond my comprehension. Now DRM is entirely different animal.  I still don't see the uproar over all this. Understanding it is one thing but losing your sanity is another (not saying your losing it) but some sure have been rude to those professionals who've tried to chime in.

Takes the master, eliminates the faults, adds a filter... You are repeating the fairly deceitful pr speak from MQA. Lets break it down:

Removes ADC issues: Just about every current production uses a multitude of ADCs, mixing manipulation, and even no ADC at all (eg electronic synths). So the picture of a mike to an ADC is just about as common as Donald Trump in drag. Then the secret time desmearing sauce. There is no secret. Then the lossy compression then unfolding then the choice of an upsampling algo/dsp. All reasonable choices but nothing that is not available to upsamplers out there. 

[miguelito]

3 hours ago, Gonzbull said:

You're partly right about multiple ADCs on a production but the part about mics being hardly used is nonsense. I work in the industry and mics get used everyday. The majority of sounds still get recorded using mics and preamps. Even electronic synths as you call them are recorded via DIs. 

Theres commonly 3 sets of ADCs used. 1st when recording, 2nd when mixing and 3rd when Mastering. 

For sure mikes are used! I meant to say the simplistic picture of one mike to one pre to one ADC to the digital file is deceitful. In reality, many different mikes, to many different pres, to many ADCs, to a mixing console, to a digital master is more like it. Correcting the signature of a recording change, as MQA likes to picture it, is not a possible situation for most recordings.

[miguelito]

9 minutes ago, rickca said:

Are you saying you just want to be an MQA renderer, but you don't have that option?  Why not?  That's what Berkeley is doing on their Alpha DAC Reference Series 2.

What exactly is the BAD ref 2 going to do? Rendering? A quick google search didn't render anything clear.

 

9 minutes ago, rickca said:

When used with MQA core decoding software such as TIDAL HiFi, Roon, Audirvana, etc. the Alpha DAC Reference Series 2 will provide full MQA decoding while minimizing processing noise within the DAC. This approach results in improved audio quality by splitting the MQA processing overhead between external software and the DAC, resulting in lower noise in the DAC and increased D to A conversion accuracy.

At the moment, only TIDAL's native desktop app and Audirvana (and exclusively TIDAL on Windows) can do the "first unfold" to 2x sample rate. Roon does not support this. This means renderers such as the Dragonfly Black and Red will only be able to "render MQA" when used with these apps.

 

As for lower noise/increased D-A accuracy, we shall not go there...

 

[miguelito]

33 minutes ago, citsur86 said:

I have been using the Tidal App on my MacBook and I am getting the dark purple MQA color on my DFR.  Are you sure it is only TIDAL on Windows?

I meant to say on mac you can either use the TIDAL native app or Audirvana. Audirvana does not have a Windows version, so on Windows you can only use the TIDAL app.

[miguelito]

1 minute ago, rickca said:

Miguelito, the 2nd paragraph of my earlier post is a direct quote from a Berkeley press release, not my opinion.  See https://www.computeraudiophile.com/forums/topic/30786-berkeley-audio-design-adds-mqa-rendering-to-alpha-dac-reference-series-2/

Ok got it. It is a bit unclear from this whether the BAD will be a renderer or full MQA decoder.

[miguelito]

2 minutes ago, abrxx said:

Sorry, I didn't fully understand your answer. I think we all understand that when Tidal is playing an MQA track, the datastream being sent to the MQA renderer (in this case the Dragonfly) is sent at either 88.2 or 96 kHz, that is, the first unfold has been done.

Correct.

 

2 minutes ago, abrxx said:

You stated that the next fold (if required) is done by the micro controller, which takes us to 192Khz. I'm simply asking if the original track was at 384kHz, does your code do the final unfold from 192 to 384? In all cases I understand that actual data going into the DAC is at a folded rate of 88.2 or 96Hz. 

What the microcontroller seems to be doing it simply interpreting the embedded information in the unfolded PCM stream that includes configuration instructions for the ESS DAC to set filters and upsample to a particular rate. The ESS DAC is capable of upsampling all the way to 384KHz. What upsampling rate is used as well as the filter to be used is determined by the MQA info embedded in the unfolded PCM stream.

[miguelito]

4 minutes ago, citsur86 said:

So if you are using the Dragonfly Red or Black and having Tidal do the "core" unfold, and it sends the Dragonfly 96KHz, then what was the point of the update to the Dragonfly to make it able to do the second unfold if it can only go up to 96KHz?  

The update enabled the microcontroller in the DF to interpret the embedded MQA info in that unfolded stream. That info will tell the ESS DAC in the DF to upsample to a certain rate and to set particular filters. Before the update, this info was missed so whatever upsampling and filters where used where unrelated to the MQA information.

 

You also get a purple light with the update giving you the (misguided?) comfort of an MQA file being rendered.

[miguelito]

1 hour ago, revand said:

Does this mean that the Dragonfly Red/Black at making the second unfold of the music file already unfolded by the desktop app or Audirvana Plus 3 to 24bit/88.2 or 24bit/96 kHz (at "rendering") cannot provide the resolution of the original master?

Strictly speaking, the resolution of a 384KHz master was long lost in the folding... Even if you had a 24bit/96KHz master, once you fold it to 24bit/48KHz you cannot get back the original data. MQA Co. argues it is "perceptually lossless" meaning that the unfolded file (48KHz back to 96KHz) is "psychoacoustically indistinguishable" from the original 24/96. I will leave it to you to judge what that means.

 

The second step in decoding is the application of filters to render the file. This is at best "informed upsampling". Consider the Dragonfly's ESS DAC: It does not know anything about MQA, all it does know is to set specific upsampling filters as instructed by the microcontroller.

 

So when someone says "recreating the original" or some BS like this, I just roll my eyes...

 

Does it sound good? Sure. Does it prevent you from using room corrections? Yes. Does MQA sound better than PCM with room corrections? For those that enjoy the latter the answer is clearly no.

[miguelito]

6 minutes ago, PeterSt said:

 

You must be a bit careful with this ...

As I understand, the DF doesn't go beyond 96KHz anyway, so there is not much to upsample, assumed the original file was 96KHz (or 88KHz).

This, while upsampling is also a possibility (by MQA hardware). But I think this is completely outside the subject.

The 2L recordings go to 88KHz (first unfold) into the DF and get upsampled to 352KHz in the DF's ESS DAC.

 

Quote

To be hopefully more clear : the "upsampling filters" as you mention them are nothing of the kind when we'd talk about reconstructing the native 96KHz (etc.). Let's call that "unfolding process" and you're at what it is.

And this works up to 384KHz (or 352.8 more practically (DXD like from 2L).

88 or 96 to 352 or 384 is upsampling in the ESS DAC. The MQA embedded data that the microcontroller decodes tells the ESS DAC what upsampling settings/filter to use. There are no MQA owned anything in that DAC.

[miguelito]

10 minutes ago, DarwinOSX said:

Lost or slightly diminished?

You are never going to get PCM streaming which means it will always be an outlier used by people who read forums like this one.

Lost in the mathematical sense. Once you allow for "psychoacoustically equivalent" you're doing medical perception not scientifically verifiable facts.

[miguelito]

9 minutes ago, PeterSt said:

 

Maybe I am unsure what you mean by the latter ("owned") ? language problem ...

What I mean is that there is no knowledge of how to unfold MQA in the ESS DAC. My understanding it that the firmware update for the Dragonfly only touches the microcontroller code, not the ESS DAC firmware. Maybe @Wavelength can clarify.

 

Quote

If you only know that MQA can unfold to 384 when the file natively is that.

Uhm ... don't ask me how that would work when still the transport (stream) is 44.1 or 48 but it is a typical thing I take for granted. Anyway it is not upsampled.

This, while upsampling to max 384 is a feature of MQA.

You're taking too many things for granted, methinks...

 

Quote

I suppose you are saying that the DF just doesn't go beyond 96. Fine. But this is not related to MQA's capabilities (and upsampling as such I don't call a capability to begin with, but alas, it can do that too).

No. What I am saying is in fact EVERY renderer will start from a 88 or 96 KHz stream. I have seen nothing in the MQA documentation that negates that the first 2x unfold is ALWAYS done in the same exact way, be it exclusively in hardware or split between a software unfold and a further rendering.

 

So in other words, MQA is ALWAYS decoded in two steps:

1- Unfold to 2x rate

2- Post-process to target rate

 

If I missed docs that indicate I am wrong, please point me. Thank you.

[miguelito]

46 minutes ago, abrxx said:

 

Yes it is. He's just using the same marketing language ("folds") for completely different things. Whether its un-encapsulating the 48/24 to 96/24, or the MQA-secret-sauce upsampling, he calls both unfolding. Additionally, each time the rate doubles, its an unfold. So to go from 48/24 to 192/24 is TWO unfolds. And so on....

But that's my point... The ESS DAC in the DF doesn't know anything about these unfolds. It only knows about upsampling. I acknowledge there's an interesting twist in choosing upsampling filters on a per-file basis, but lets not call it what it is not.

[miguelito]

20 hours ago, DarwinOSX said:

 

Bob Stuart talks about three unfolds.

 

 

No he does not. Three regions, two "folds".

 

Moreover, he says that region C is "noise" that allows the DAC to work at a higher speed. This is pretty much what upsampling does: guess higher frequency information such that you don't need a brickwall filter with all it's phase (ie time blurring) issues. This video is pretty convincing that what is done to "recover" C is pure upsampling.

[miguelito]

15 minutes ago, mansr said:

Could it be that A+ displays the format of the file rather than what is sent to the DAC? If the file sample rate exceeds what the DAC supports, it is presumably downsampled. If there's one thing everybody agrees on, it's that the Dragonfly is limited to 96 kHz input.

That's what it does. The MQA file includes information on how high to upsample. So for example Joni Mitchell's 'Both Sides Now' only goes to 96KHz (ie no upsample in the DAC, possibly some filter applied), whereas some 2L recordings indicate 352KHz (which I take MQA is telling the DAC to upsample the incoming 88KHz stream to 352KHz).

 

The fact that the DF is limited to 96KHz is fine as the first unfold never goes beyond that anyway.

[miguelito]

On 5/26/2017 at 8:39 PM, ShawnC said:

Here's Han's review

 

 

This review is all wrong. He has no idea how this works and it can make people misunderstand how to use the DF to render MQA. 

 

And btw, he is NOT decoding MQA at all! What a moron.