New demo

[tEd]

To expand on Humus' point, here are some very regular sparse sample patterns (up to 16x):

The red ones are those that form a rotated and skewed grid over the whole screen. The blue ones are those where the "lines" don't exactly continue over pixel borders.

In screenshots those patterns will look very good, maybe with the exception of some thin lines (e.g. antennas, electricity lines) that are exactly aligned with the pattern grid.

In movement, however, those edges aligned with the pattern grid will not move completely smooth, e.g. even with the 16x pattern there are situations where such a moving edge will only jump between 3 intermediate steps.

Here are some irregular (well, some aren't) patterns that should give very good results in any situation:

I notice that there is only 1 sample position per row and column. If that was intentional, would you mind explaining to me why? Does it have to be that way for a sample pattern or is it that way simply because it is the optimal approach?

That's what defines a sparse sample pattern. Not more than 1 sample per row/column. It's better as when samples are aligned.

 

[aranfell]

I notice that there is only 1 sample position per row and column. If that was intentional, would you mind explaining to me why? Does it have to be that way for a sample pattern or is it that way simply because it is the optimal approach?

Consider a nearly horizontal or nearly vertical edge sweeping across the image. With N samples that are all in separate rows and columns, the result is N increases in intensity as the edge sweeps across a pixel. So in this case, a sparse pattern produces nearly the same edge AA quality as an N^2 sample ordered pattern, which still has only N increases in intensity because the edge hits N of the samples at once (or nearly at once).

The question of edge AA quality is more difficult for lines at other angles. Any time an edge crosses two or more samples at the same time in a pixel (or at nearly the same time), the quality of the edge AA is reduced, because the number of evenly spaced changes in intensity is reduced. So it is desirable (though impossible) to have a sparse pattern at all angles. But generally edge aliasing shows up worse for the near-horizontal and near-vertical cases, so those are the ones to pay the most attention to.

Enjoy, Aranfell

 

[KimB]

Well, you missed that not only are all samples at different rows and columns, but the second criterion is that their positions are chosen such that their distance from one another is maximized. This gives a finite set of possible configurations.

 

[Xmas]

Wow Xmas, do you design sampling patterns for a living?

Nah, unfortunately not. I just seem to have too much spare time lately...

I'm actually wondering for quite some time now, how come no company has not employed him yet. That means if he actually can or wants to at the moment.

I'm "busy" writing my diploma thesis (3d graphics related) atm which I hope will be finished in a few months (and for some reason, lots of completely unrelated ideas cross my mind...) . But after that, who knows

Well, you missed that not only are all samples at different rows and columns, but the second criterion is that their positions are chosen such that their distance from one another is maximized. This gives a finite set of possible configurations.

Distances are maximized with an ordered grid, or, if having a sparse grid is priority, with the regular sparse sample patterns I showed in the first image. But those are not the ones that result in best quality overall. You need to hit some fine balance between samples being not too close to each other and adding irregularities to the pattern.

 

[KimB]

Well, right. But from what I remember, if you do the calculations, you get a number of results if you use those two criterion, some of which will be the irregular patterns, others which will be the ordered patterns.

 

[Ailuros]

I'm "busy" writing my diploma thesis (3d graphics related) atm which I hope will be finished in a few months (and for some reason, lots of completely unrelated ideas cross my mind...) . But after that, who knows

Aha! Publish or perish Best of luck for that one and anything to follow.

Question: I have had a look at the demo by itself but didn't have time to combine the application's sparsed SSAA with control panel AA yet. How good are the chances that application and driver control panel AA sample never overlap to reach that theoretical maximum of 66xAA?