XXHighEnd

Hi Peter!
I have been using 2.11 with great enjoyment. It is stable, robust and very discriminating of quality of original music files. I am having some very elementary question that I don't see where to fit in any topic area but thanks for your direction for me to post here.

My questions are:
To take advantages of the custom Arc Prediction filters, do I need to play a track using upsampling? For example, if I play a Red Book track I need to choose 4x, 8x or 16x upsampling and also choose filters accordingly to the extend that DAC supports?

And if a DAC is only capable of 44.1k, and when I play a Red Book track from 44.1k to 44.1k out, are these filters still working, or improving SQ? I heard something and hope it's not imaginary.

What are the filters in the group [NOS] for? ony for Phasure DAC?

Hi there Zheng,

Let's see whether I can answer your questions adequately;

First off, digital filtering always comes along with upsampling. This is hard to explain, but I could try to convince you that only the upsampling step allows to have the result mostly the same as the original. Think like injected additional samples while the original samples remain. But mind you, this is typical for Arc Prediction and this means of filtering is far from common (at all).
The more upsampling steps, the more samples can be injected in between the originals. Thus, with 16x upsampling, 15 samples are injected between each 2 originals. Let's say that they imply (and hunt for) a better transition from the one original to the next original. Not "one big jump" but in 16 small (calculated) steps now. Or in two steps when upsampling is just 2x.

"Custom" Arc Prediction is not really different, BUT in this case a second stage filter is applied on top of the normal Arc Prediction. This is done the most mildly where the (possible) transients of Arc Prediction are sustained. In this case, however, the original sampling points are changed, as in any normal digital filter doing this, but with the notice that this is "measured" (by me with the analyser) to not noticeably change the transients. Do notice that a normal filter will change the transients notably, me always stating that it will changes violins into flutes (think micro transients, implied by the bow and the resin onto the string - the more resin the more "jumping" (transients) on the string).

The more times upsampling, the better the filtering can be applied (this is a story in itself). But notice that this is generally so, because it also can depend on the behaviour of the DAC at higher sampling rates (it needs to process more data and this in itself can be audible in a negative sense). And indeed it is so that some people play our NOS1 DAC at 352.8 instead of 705.6 (upsampled RedBook); less current is used now, which may mean that more reserve is there for stuff like again transients.

As you can see, the transients are almost a hobby-horse to me. The accurate rendering of them is key (in my view).

When playing out 44.1K (from 44.1K) no filtering can be active. Still it may be so that the software may "touch" other parts which again imply processing, but now in-software only. And mind you, as we know, XXHighEnd is "tweaking" all over the place and this is about processing (all over the place). Also more physical things may be in order, like you outputting 44.1K all right, but maybe outputting in 24 (or 32) bits, while the original is 16 bits. Officially this can't matter, but now physically it may imply half more data (or double) towards the DAC, which now must process it.
And as we also know by now, everything matters ...

What are the filters in the group [NOS] for? ony for Phasure DAC?

It is that you mention them, otherwise I never even saw them. I suppose they sneaked in the production version, from some testing a long time ago. Please don't use those.

I hope it is all clear a little !
Kind regards,
Peter

Wow thanks Peter! Yes it helps me to understand the process a lot more now. It reminds me of certain aspects of digital image processing. If I need to do any adjustments in color, light or anything, it’s the best practice to do it when image is in very high resolution and color depth, even if your end product needs only to be a small image of 800 pixels in limited color depth.