Against the Grain I do not prefer MQA


#1

Now I have a proper setup to test these things once and for all.

For the SGCD, DSD trumps all notion of MQA -PCM. This newfounded obsession with MQA is just hype to my ears.

The magic that PCM->DSD conversion provides is several magnitudes greater than fiddling about with PCM.

For experimentation I have tested Tidal MQA of several tracks vs. my 24 bit downloads… both upsampled the same way to DSD to my SGCD.

MQA sounds muddy, strings and piano blend together and sound like they are being constrained.

The local 24 bit comes alive more like the instruments are here in the room. Violin has so much more emotion and expression. The MQA took away the violin’s voice and made it sound generic. Non-MQA you get a sense of the player’s personality. Several tracks sampled from Max Richter - Three Worlds.

I will stand up and say that PS Audio is ahead of the curve. DSD sounds like music, MQA sounds like a band-aid. I can’t wait for a DS Jr!


#2

Not sold on MQA, but your test is flawed. You are testing the MQA with out it being decoded. Then converting it to DSD. Whatever you are using to convert files to DSD, is not designed to deal with the data folded in there. And as far as I know there is no way to unfold MQA completely then send it to whatever you use to convert PCM to DSD.

We have pretty much gone through this many times, there is seldom any way to know if the MQA version came from the same master. So that is another variable.

All you have proved to yourself, is that taking an unfolded MQA file and running it through a format converter doesn’t work. Most of us could have told you that.


#3

@jeffstarr

Audirvana unfolded to its max rate available for the track of 24/48. The MQA blue dot is on in the software… etc etc.

I have also compared Files of 24/96. Audirvana unfolds to that as well.

Both native and MQA going to DSD sound way better than straight MQA PCM.

I have 24/96 of Natalie Merchant… which is also on MQA which will unfold in A+. Will compare tonight.

While we don’t have proof that things come from the same master… why worry about a format that doesn’t sound as good… same master or not? I will compare more and lets see if there is a trend to the sound signature.


#4

Best I could tell the Audirvana is doing the same thing that Tidal’s player does, the first unfold. So the sample rate converter is still dealing with data it is not designed to convert. They say on their website you need a MQA DAC to again double the sample rate. Which is misleading because of the other things that MQA supposedly does.

“why worry about a format that doesn’t sound as good… same master or not?” Where do you get the idea that I am worried? What I am doing is pointing out that your conclusions are based on a faulty comparison. If you are using Tidal, and the MQA are sourced from better masters, that is a benefit. The DAC that I am using does not do DSD, the PC I use is not a Mac, so I can’t use Audirvana.

I have read your other posts where you have found converting files to DSD sound better. I am happy to hear that you found something that sounds good to you.

What does worry me is that MQA becomes the only version available. What also concerns me is that Tidal will raise the price for access to better versions that were sourced when they used MQA. Depending on price, I will probably switch to Quboz when it becomes available, to get access to true hi-rez files.

Do you understand that even if MQA does not really deblur or correct timing errors having access to a better master is a benefit. I can’t afford to buy hi-rez files of all the music I listen to thanks to Tidal.

And to make it very clear, I am not pro MQA, but seeing as we are stuck with it, why wouldn’t I let Tidal send my DAC 24/88 or 24/96 of a master that has not been downsampled to 16/44. There is one other benefit you don’t have a brickwall filter at 22khz.


#5

To add to your point… not all files unfold to 192.

The max richter files only unfold in software, that’s it. Same for the 24/96 files. The only ones doing more are 24/192. This is misleading and you have to dive down in the details to really realize this.

In fact Audirvana let’s you track what is happening. Most 24/48 limited files are actually upsampled and you can’t turn off the upsampling! I don’t agree with this and this is misleading! There is a huge lack of granularity.


#6

@jeffstarr

i apologize if I’ve mentioned to you before or if you already are aware but Audirvana is coming to Windows very soon.


#7

See here https://www.audiostream.com/content/mqa-decoding-explained

I am comparing files that complete all unfolding in software and don’t receive any extra benefit in hardware.

Comparing Natalie Merchant, unfolded to 96 - this is the max rate.

MQA to DSD, clearly beneficial to straight MQA, the instruments sound more like real instruments. Vocals are less boomy, guitar sounds much better, nice string harmonics. Acoustic bass sounds a little bloated. Vocals are jumping in front of guitars. Vocals are forward in my opinion. While the strings are better than straight MQA, still something off.

PCM to DSD, relaxed, more harmonic detail in the strings. Vocals are more natural. Speakers completely disappear. Vocals are no longer forward… better seamlessness of instruments. This is the only version that I can hear realistic string vibrations of the acoustic bass. The others are all bloated. This is the only one where the instruments don’t blend together and sound live in natural space.


#8

I’d like to add that I am not trying to trash MQA but rather just open the possibility that MQA by itself is not the be all end to sound quality. I am not implying anyone has stated this. Rather there is so much hype with MQA and so many variables present its easy to generalize MQA is a guaranteed win for SQ.

However, if you take the Tidal app with MQA and compare to Audirvana with straight PCM (no upsampling), you may be surprised at your results. Implementation is HUGE.

Keep in mind that all of the ADC fixings take place in software for all songs and the other goodies are simply tuning the upsampling protocols. However, these upsampling protocols are all based on min phase filters. I do not prefer these filters. This is most likely a big reason why I feel I can do better with the Izotope upsampling in Audirvana where I control the slope and only use pure linear phase.

I have compared linear phase upsampling in A+ to the same with DSD conversion and the DSD on the SGCD wins hands down.

All of this may vary by DAC. If you have a non ESS chip I have no idea if you will have my preferences. Heck, some people swear by min phase (i don’t know what they are smoking ) and that is fine. Also there is a chance if you have a different DAC with a different power layout, maybe something else works better.

So far however, I can upsample and DSD convert PCM 16 bit TIdal streams that sound significantly better than MQA unfolded to 24/96.

All DSD conversion is not created equal… Roon is ok but doesn’t have the flexibility or options to really get there. Its a fine option however and will be an improvement over PCM.


#9

Hold the phone… Just tried out something that works alot better.

Using the FabFilter Pro Plugin in A+ I can now use a custom filter to filter out aliasing.

Setting the upsample filter to 3db/octave produces extremely low ringing and insane instrument separation. Nyquist has to be adjusted to 1.3 in order to prevent rolloff.

Now I turned down the internal AA filter to minimum level since it is a bell filter and only cuts at the Nyquist and not above.

Next using the FabFilter Pro I set a 96/db High Cut filter set for -3db point at 20khz (I am willing to roll the highs a little here) with Medium latency Linear Phase filtering. I tried the Natural Phase filtering and while it sounds really fast and has better transients… somehow it doesn’t sound like music to me, sounds analytical and precise, but no emotion.

I am in contact with A+ to see if we can fully disable the internal AA filter; its not that effective and we can do better with plugins.