MQA vs HiRez: an apples-to-apples comparison

[mansr]

1 minute ago, manisandher said:

With no rendering going on, only decoding?

The deblurring process, if it exists, can rightfully be called a filter. Anything that receives an input signal and produces a related output signal is a filter.

[manisandher]

3 minutes ago, mansr said:

The deblurring process, if it exists, can rightfully be called a filter. Anything that receives an input signal and produces a related output signal is a filter.

 

OK. But to be clear, we're not talking about an anti-imaging filter here, which was what @tmtomh was alluding to.

 

Mani.

[tmtomh]

16 minutes ago, manisandher said:

 

OK. But to be clear, we're not talking about an anti-imaging filter here, which was what @tmtomh was alluding to.

 

Mani.

I didn't intend to refer to anything as specific as an anti-imaging filter. Please don't let anything I might unintentionally have implied throw this discussion off course. Thanks!

[tmtomh]

34 minutes ago, manisandher said:

 

Yes, but there's no digital filtering going on here.

 

Mani.

 

I'm not sure if that's the case - but at any rate, my intent was not to assert that there was. I was just grasping for analogies for what that difference spectrogram is showing.

[manisandher]

1 hour ago, tmtomh said:

As for the strange, almost comb-like difference right at the heart of the midrange, I have no earthly idea.

 

This is so intriguing.

 

 

Could it really be pure coincidence that these comb-like differences sit exactly where our hearing is most sensitive?

 

As for the extremes, maybe MQA is applying some sort of very subtle Munson Fletcher equalisation?

 

Also, let's not forget that everyone who's posted their listening impressions to date has said that they hear a clear difference between the hirez and MQA captures, even with the FFT differences sitting below -80dB or so.

 

Mani.

[Shadorne]

4 hours ago, Miska said:

 

The volume control is quite typical relay ladder. I have similar in my preamp and also the T+A DAC8 DSD for example has such if the variable output is enabled.

 

More interesting is the THX's amplifier module they use in HPA4 which is also used in AHB2 and is responsible for the nice performance.

 

Overall measured performance is quite nice as it is for AHB2 too. I still want to hear how it sounds too, not gonna burn 3000+ on plain specs... But the feature set also matches, because I need balanced in/out plus multiple inputs, because I have stack of DACs... I need to check if it is possible to name the inputs, because that makes life so much easier.

 

I've had also iFi Pro iCAN on my short list.

 

 

The PSU they have is an ordinary SMPS module, probably from some other vendor. But of course they clean it up. Quite a bit like for example Mytek Brooklyn DAC+ (and maybe Benchmark DAC3 too?).

 

 

The benchmark PSU has remarkably low noise. As with most of Benchmark products it is not simply off the shelf from a vendor but tweaked to operate at a suitable frequency up to 1 MHz

 

https://benchmarkmedia.com/blogs/application_notes/inside-the-dac2-part-3-power-supplies

 

 

https://benchmarkmedia.com/blogs/application_notes/152143111-audio-myth-switching-power-supplies-are-noisy

 

 

[Miska]

1 hour ago, Shadorne said:

The benchmark PSU has remarkably low noise. As with most of Benchmark products it is not simply off the shelf from a vendor but tweaked to operate at a suitable frequency up to 1 MHz

 

https://benchmarkmedia.com/blogs/application_notes/inside-the-dac2-part-3-power-supplies

 

The shielded cage part in the picture is very likely (obviously) a standard part from some regular vendor, the yellow part on the main board is not. You can see (if you are familiar with electronics design) there's quite a bit of filtering on the main board the clean up the SMPS output.

 

If you look carefully enough on this picture, you can probably figure out who made the PSU:

 

 

Edit: well, it didn't take me long to find the match, the PSU seems to be one of the TDK-Lambda models:

 

[Shadorne]

52 minutes ago, Miska said:

 

The shielded cage part in the picture is very likely (obviously) a standard part from some regular vendor, the yellow part on the main board is not. You can see (if you are familiar with electronics design) there's quite a bit of filtering on the main board the clean up the SMPS output.

 

If you look carefully enough on this picture, you can probably figure out who made the PSU:

 

 

Edit: well, it didn't take me long to find the match, the PSU seems to be one of the TDK-Lambda models:

 

 

Agree 100%. They tweaked the PSU for audio application or they selected a PSU that operates at the desired frequencies for minimum audio noise. Perhaps it is a custom tweaked version made by TDK-Lambda for Benchmark. I wouldn’t expect them to re-invent a wheel if one was widely available at low cost and with desirable quality. I dont find any of this surprising or shocking. I bet they don’t design and manufacture their own FPGA either. The ESS chip is completely standard too.

 

When it comes to electronics I am not a fan of building everything discretely in house from scratch - almost impossible to do that efficiently, correctly/reliably or cost effectively as a specialized part supplier -  discrete is rarely as good as carefully selecting good parts and tweaking for the specific application desired.

 

It looks like they went to some trouble to ensure their circuit board is well shielded (multi-layered with grounded top and bottom)  - I would expect the circuit board to be their own design. 

 

Anyway nothing remarkably surprising at all.

 

If you see something wrong with their approach what is it? To me it looks like a rather sensible or practical approach. 

[fas42]

We still seem to be discussing whether Audacity messes things up by arbitrarily adding dither - so this is from the manual, https://manual.audacityteam.org/man/dither.html. Does anyone wish to dispute anything stated there, or suggest that the methods used are inadequate?

[Don Hills]

Regarding filters in MQA: There is a filter in the encoder to split the signal into two parts (for example, 0-24 KHz and 24-48 KHz). The "low" part is stored in the top 15 bits of each 24-bit sample. The "high" part is encoded and stored in the lowest 8 bits of each sample. The decoder uses a matching but complementary filter to rejoin the decoded "high" part to the "low" part. In theory, the decoding filter should exactly reverse the effect of the encoding filter.

[PeterSt]

42 minutes ago, Don Hills said:

In theory, the decoding filter should exactly reverse the effect of the encoding filter.

 

Which is why I personally wouldn't call it a filter. Hence, deblurring can't be about that (if it exists).

[Fokus]

1 hour ago, PeterSt said:

 

Which is why I personally wouldn't call it a filter. Hence, deblurring can't be about that (if it exists).

 

Conceptually it is a set of four filters, with rigid design constraints. But as Don said, once executed at the encoding side and then at the decoding side they should become invisible and are indeed not related at all to deblurring or the laze downsampling and upsampling filters.

 

(However, and slightly off topic, it appears as if the reconstruction filter MQA DACs tend to use by default for non-MQA 1x material is one of these four filters. It also is still an open question as to

any negative effects of the one encoding-side low-pass downsampling filter on the quality of an undecoded MQA signal. There is no free lunch in mathematics, one would expect a negative impact here.)

 

 

[PeterSt]

13 minutes ago, Fokus said:

and slightly off topic, it appears as if the reconstruction filter MQA DACs tend to use by default for non-MQA 1x material is one of these four filters.

 

Not so off topic is what I wanted to say myself but didn't and which closely looks to what you're saying there, is :

 

What about MQA 1x material which is thus MQA but which remains 44.1 / 48.

To me it would be mor interesting to see what happens with that, but I wouldn't like to put a burden upon Mani at a. finding a master in 44.1 which is the same as the master used for MQA (resulting in 44.1) and b. making a copy of that for us/you to judge which will be the farthest from easy, I think, because how to avoid the upsampling/filtering any self respecting DAC would do. OK, Mani could use his NOS1 DAC and apply the very same upsampling filtering for either in software, which makes it subject to XXHighEnd again and not Roon.

Well sort of, and more possibilities exist.

 

Anyway, we must understand that the deblurring as such is something different from packing a hires file in a low res container. Thus start out with a low res to begin with, but which *is* MQA and you may end up with the real thing. And mind you, for me the nature of MQA is (audibly) just the same in a 44.1 / 48.

[crenca]

6 hours ago, PeterSt said:

 

Not so off topic is what I wanted to say myself but didn't and which closely looks to what you're saying there, is :

 

What about MQA 1x material which is thus MQA but which remains 44.1 / 48.

To me it would be mor interesting to see what happens with that, but I wouldn't like to put a burden upon Mani at a. finding a master in 44.1 which is the same as the master used for MQA (resulting in 44.1) and b. making a copy of that for us/you to judge which will be the farthest from easy, I think, because how to avoid the upsampling/filtering any self respecting DAC would do. OK, Mani could use his NOS1 DAC and apply the very same upsampling filtering for either in software, which makes it subject to XXHighEnd again and not Roon.

Well sort of, and more possibilities exist.

 

Anyway, we must understand that the deblurring as such is something different from packing a hires file in a low res container. Thus start out with a low res to begin with, but which *is* MQA and you may end up with the real thing. And mind you, for me the nature of MQA is (audibly) just the same in a 44.1 / 48.

 

Bob James' "The New Cool" is one (in 48) that is available in MQA and 48 PCM on Tidal...just off the top of my head

[manisandher]

On 10/18/2018 at 9:46 AM, PeterSt said:

What about MQA 1x material which is thus MQA but which remains 44.1 / 48.

 

Hey Peter, I'll see what I can come up with.

 

Mani.

[manisandher]

On 10/18/2018 at 4:28 PM, crenca said:

Bob James' "The New Cool" is one (in 48) that is available in MQA and 48 PCM on Tidal...just off the top of my head

 

No, the PCM is 16/44.1 on Tidal. It may be available as a 24/48 PCM elsewhere, but even if I could find it, I'd have to check if the MQA was derived from the same master.

 

Making apples-to-apples comparisons isn't as straightforward as it might appear.

 

Mani.

[crenca]

11 minutes ago, manisandher said:

 

No, the PCM is 16/44.1 on Tidal. It may be available as a 24/48 PCM elsewhere, but even if I could find it, I'd have to check if the MQA was derived from the same master.

 

Making apples-to-apples comparisons isn't as straightforward as it might appear.

 

Mani.

 

Yes, but all PCM on Tidal is 16/44....  The Hi Res version is 24/48 (available usual places) and the MQA version reports as 24/48, so as close your going to get to being sure "MQA was derived from the same master..." without inside info or post analysis I suppose.  When Tidal was batch dumping MQA albums by the 100's a day this was one.  I am confident (as I can be) this is the same master...

[Miska]

Just came across this one, which is interesting in itself:

http://www.2l.no/hires/DXD-DSD/index.html

 

Although big question I have is why he didn't run the second Horus at DSD256 instead of DSD128? Well, now the DSD version is at least 200 MB smaller...

 

This would be so much more interesting with more complex music.

 

[Kyhl]

I avoided reading until I could listen, not wanting to skew my thoughts.

 

D was more enjoyable to listen to.  C was brighter and had a flatter soundstage.  On D the soundstange came across more natural with instruments easier to place in space.  The voice on D was smoother and easier on the ears.  Hope that wasn't MQA.  ?

 

Now I'll go back and read the thread to find out which was which.

[crenca]

What's the score?  Has anyone chosen C (MQA) over D (hi res - 24/88.2 correct?)?

[fas42]

It's interesting to see how MQA works in the flesh ... on the basic laptop setup I use it adds a bit of enlivening punch to audiophile recordings, which tend to be very 'dead' - so MQA sometimes wins. I suspect on recordings that have plenty of dynamics as part of the mix MQA will add nothing, or make it worse.

[Kyhl]

On 10/16/2018 at 5:29 AM, Miska said:

 

You can listen MQA content, either local files or Tidal stream decoded by Roon and upsampled by HQPlayer.

 

If you have HQPlayer Embedded, you can also do the same in various other ways too.

 

I listened C and D through my usual desktop test/monitoring setup; RME ADI-2 Pro AE (with DSD Direct enabled) running at DSD256 and Holo Audio Cyan DSD running at DSD512 to Schiit Jotunheim headphone amp and Sennheiser HD800 headphones. ADI-2 run by HQPlayer Desktop with poly-sinc-ext2 filter and Cyan run by HQPlayer Embedded with poly-sinc-mqa-mp filter.

 

I feel that C has slightly accentuated but messy highs. While D has more air.

 

 

Accentuated but messy highs is a much better description than what I wrote.

You nailed it.

 

My fear was that a slow rolled off filter could hide the messy highs of C making D sound better if it was using an early roll off.

[kumakuma]

23 minutes ago, fas42 said:

It's interesting to see how MQA works in the flesh ... on the basic laptop setup I use it adds a bit of enlivening punch to audiophile recordings, which tend to be very 'dead' - so MQA sometimes wins. I suspect on recordings that have plenty of dynamics as part of the mix MQA will add nothing, or make it worse.

 

I'm confused. You said you didn't have a DAC and only listen to CDs. How are you listening to MQA releases?

[fas42]

5 minutes ago, kumakuma said:

 

I'm confused. You said you didn't have a DAC and only listen to CDs. How are you listening to MQA releases?

 

People have posted files that supposedly are the "unfolded" versions. As in this thread. Of the tiny amount of this material I've heard, just over the laptop sound system, I've gone both ways on which is "best".

[kumakuma]

Just now, fas42 said:

 

People have posted files that supposedly are the "unfolded" versions. As in this thread. Of the tiny amount of this material I've heard, just over the laptop sound system, I've gone both ways on which is "best".

 

You really have no idea at all.

 

And you wonder why people have a hard time taking you seriously.