Is MQA good enough to get Analog Guys to enjoy it? Or still cause Digital Fatigue?

I am unsure what oversampling achieves in a DAC that is a DS model, as that would oversample in the DAC anyway. I guess it's origins were to avoid the issues of a filter? Oversampling first in software, then the DAC doing it again, I don't get the logic, or have heard the sonic benefit of that route.

Yuri Korzunov said:

DSD have simpler demodulator (in DAC) than PCM.

Click to expand...

The Mojo article I posted seems to say the opposite, and that DSD increases modulation noise. I have no idea of either is true as concentrate on Redbook or 96K in an R-2R. I will get to hear the Lampi GG later this spring, so am open minded about the subject.

astrostar59 said:

I am unsure what oversampling achieves in a DAC that is a DS model, as that would oversample in the DAC anyway. I guess it's origins were to avoid the issues of a filter? Oversampling first in software, then the DAC doing it again, I don't get the logic, or have heard the sonic benefit of that route.

Click to expand...

In DAC may be not only oversampling but conversion to sigma-delta modulation (a.k.a. DSD).

All these actions need for successful work with analog filter.

I suppose, oversamping by DAC may be applied not for all input sample rates.

Though all depend on implementation. Using DAC as "black box" we can try found input sample rates+bit depths combinations that sound better (defined by implementation of DAC).

astrostar59 said:

The Mojo article I posted seems to say the opposite, and that DSD increases modulation noise. I have no idea of either is true as concentrate on Redbook or 96K in an R-2R. I will get to hear the Lampi GG later this spring, so am open minded about the subject.

Click to expand...

Of course, Mojo know about DACs more that me.


1. As rule, modern PCM DAC contains both DSD modulator and demodulator for easier implementation.



2. Modulation noise is a complex issue. It depend on used sigma-delta modulator first.

The modulator should push energy of the modulation noise to higher frequencies as far as possibly.

Other hand it can decrease stability of work of demodulator (modulator's feedback loop issue).

Easier implementation of these conditions on D128 and higher.


3. Second, 1-bit modulated signal should be deep filtered at higher frequencies.

As example, in my software I applied 3 kninds of filter by cut edge (band): 20...24 kHz, 20 ... 27 kHz, 20 ... 100 kHz (depend on input and output sample rates).

I recommed use 20 ... 24 kHz cutting. But some people more like maximal 20 ... 100 kHz.

I suppose it may depend on implementation of playback tools.

DSD64 is most critical for filtering. 20 ... 27 kHz and 20 ... 100 kHz cause very significant level of noise at such bands.

However, all this noise placed in ultrasound. It may be listened for some apparatus due non-linear distortions (intermodulation products in audible range). Also it consume part of useful dynamic range.

DSD128 have lesser level of noise for 20 ... 27 kHz and 20 ... 100 kHz bands.

DSD1024 fine work in 20 ... 100 kHz band even, as I remember.

Resume:

Modulation noise in demodulated signal depend on DSD sample rate and implementation both DSD modulator and demodulator.



4. Resistor DAC is too complex for implementation. Because for achieving DSD's demodulator precision there need very high precision of used resistors.

Some time ago I calculated it and got some currently unreal precisions. But I claim nothing, because don't remember these figures.

Also there may be resistor's temperature stability issue.

astrostar59 said:

I am unsure what oversampling achieves in a DAC that is a DS model, as that would oversample in the DAC anyway. I guess it's origins were to avoid the issues of a filter? Oversampling first in software, then the DAC doing it again, I don't get the logic, or have heard the sonic benefit of that route.

Click to expand...

DSM in audio conversion is an technique which reduces the quantization noise resulting from an converter having less bits of resolution than have the digital samples fed to it, by relocating that quantization noise from the audio band to the ultrasonic band. One of the necessary prior steps for that is the opening up of ultrasonic bandwidth for the relocation. The opened ultrasonic bandwidth is a function is the (over)sampling rate, just as it is for any digital signal. So, oversampling is utilized open the needed ultrasonic spectrum.

Interesting points Yuri. I wonder where Audio Note, Zanden and the Aqua La Scala sit as they have no filtering in the digital domain. In theory this should sounds really bad, but it actually sound excellent. I have heard the Audio Note DAC 4 with and without the filter and it sounds better without. Quite possibly the 'artefacts' that the filter in Redbook was trying to erase are not as bad an effect as the ringing and intermodulation caused by a typical filter in 44.1? And quite possibly up sampling the data to then apply a filter, the act of up sampling caused timing errors and smearing of the music? I am not an expert but relating to how I have heard it applied this way in the recent past.

Quite possibly current DACs can avoid most of these problems by using multi bit (24 bit instead of 16 bit? I was no so impressed by the CH Precision DAC as the latest demo I had of that design, in that case it was up sampling everything.

Interesting points Yuri. I wonder where Audio Note, Zanden and the Aqua La Scala sit as they have no filtering in the digital domain. In theory this should sounds really bad, but it actually sound excellent. I have heard the Audio Note DAC 4 with and without the filter and it sounds better without. Quite possibly the 'artefacts' that the filter in Redbook was trying to erase are not as bad an effect as the ringing and intermodulation caused by a typical filter in 44.1? And quite possibly up sampling the data to then apply a filter, the act of up sampling caused timing errors and smearing of the music? I am not an expert but relating to how I have heard it applied this way in the recent past.

Quite possibly current DACs can avoid most of these problems by using multi bit (24 bit instead of 16 bit? I was no so impressed by the CH Precision DAC as the latest demo I had of that design, in that case it was up sampling everything.

astrostar59 said:

Interesting points Yuri. I wonder where Audio Note, Zanden and the Aqua La Scala sit as they have no filtering in the digital domain. In theory this should sounds really bad, but it actually sound excellent. I have heard the Audio Note DAC 4 with and without the filter and it sounds better without. Quite possibly the 'artefacts' that the filter in Redbook was trying to erase are not as bad an effect as the ringing and intermodulation caused by a typical filter in 44.1? And quite possibly up sampling the data to then apply a filter, the act of up sampling caused timing errors and smearing of the music? I am not an expert but relating to how I have heard it applied this way in the recent past.

Quite possibly current DACs can avoid most of these problems by using multi bit (24 bit instead of 16 bit? I was no so impressed by the CH Precision DAC as the latest demo I had of that design, in that case it was up sampling everything.

Click to expand...

Sample rate or PCM/DSD as itself guarantee nothing.

At ideal hardware (without non-linear distortions) fully non-filtered digital signal (44 or 768 or DSD or other) will sound identically ideally filtered signal.

We don't hear ultrasound (by major modern theory). Hence any ultrasound aliases for minimal 44 and above we can't hear.

However it will consume half of dynamic range for aliases. But it is not sound quality matter in current context.

If apparatus (from analog part of DAC to speakers) cause non-linear distortions, that generate audible products of intermodulations by ultrasound, we get necesity filter ultrasound or aliases (upper half of sample rate) only.

Real apparatus always cause intermaodulation products in audible range. But the products may have level at edge of audibility or below.

So even for real apparatus filtering may be avoided, I suppose. May be the DAC 4, that you mentioned, have so low intermodulations, that filter cause higher distortions, than intermodulations.

But it is hypothesis only. For exact answer need learn the DAC 4 with precise measurement tools.

Modern multibit DAC have sigma-delta modulation (a.k.a. DSD) phase. So quality of multibit remains implementation matter. More precise released 16 bit may give better quality, than 24 bit with higher level of errors.