Thought I'd bring this up again.  Using MegaPov 5.0a here and I see
something of the same effect I had before when a plane intersects with a
isosurface.  Which isosurfaces and what other objects might also be involved
I couldn't say.  Anyway, point is that at the borders, or lines as it were,
where a plane contacts a isosurface there is always a jagged line regardless
of using antialiasing or not.
Seems a fairly serious problem far as correctness of rendering goes.  I
don't know enough about it to say it's affecting every type of isosurface
and object interactions but it looks to be a certainty in a few cases I've
had.
Simplest way to check is just make a isosurface and put a plane through part
of it with distinct color differences and appropriate viewing angle then use
default AA or no AA and compare.  Disclaimer: like I say, I can't be certain
every situation will yield the same affect.
Just wanted to put this out there for any thoughts on it.

Bob
--
omniVerse http://users.aol.com/persistenceofv/all.htm

Post a reply to this message

Bob Hughes <per### [at] aolcom?subject=pov-news:> wrote:
: Anyway, point is that at the borders, or lines as it were,
: where a plane contacts a isosurface there is always a jagged line regardless
: of using antialiasing or not.

  You are most probably just experiencing accuracy problems.
  Isosurfaces are not calculated in a mathematically exact way, as most other
primitives are.

-- 
main(i,_){for(_?--i,main(i+2,"FhhQHFIJD|FQTITFN]zRFHhhTBFHhhTBFysdB"[i]
):_;i&&_>1;printf("%s",_-70?_&1?"[]":" ":(_=0,"\n")),_/=2);} /*- Warp -*/

Post a reply to this message

Bob Hughes wrote:

>Anyway, point is that at the borders, or lines as it were,
>where a plane contacts a isosurface there is always a jagged line
>regardless of using antialiasing or not.

Did you use the max_trace / all_intersections option in the isosurface?

Ingo

-- 
Photography: http://members.home.nl/ingoogni/
Pov-Ray    : http://members.home.nl/seed7/

Post a reply to this message

"Warp" <war### [at] tagpovrayorg> wrote in message
news:39832ecb@news.povray.org...
| Bob Hughes <per### [at] aolcom?subject=pov-news:> wrote:
| : Anyway, point is that at the borders, or lines as it were,
| : where a plane contacts a isosurface there is always a jagged line
regardless
| : of using antialiasing or not.
|
|   You are most probably just experiencing accuracy problems.
|   Isosurfaces are not calculated in a mathematically exact way, as most
other
| primitives are.

Oh dang it, I forgot that was said before too.  I had lowered the accuracy
quite a bit to only 0.001 from the 0.0001 I had used before.  Although,
seemed to me I had seen those jaggies even at 0.000001.  Just how small a
decimal is accuracy supposed to get to before being virtually infinite?  :-)
Had to check again, and yep, even at 0.00000001 I could see jaggies at the
plane/iso boundary.  Considerably smoother than at 0.001 but still obviously
shows the difficulty there is in tracing that boundary.
Hm, what about the 'contained_by' objects?  That have the same trouble?

Bob

Post a reply to this message

Bob Hughes <per### [at] aolcom?subject=pov-news:> wrote:
: Oh dang it, I forgot that was said before too.  I had lowered the accuracy
: quite a bit to only 0.001 from the 0.0001 I had used before.  Although,
: seemed to me I had seen those jaggies even at 0.000001.  Just how small a
: decimal is accuracy supposed to get to before being virtually infinite?  :-)
: Had to check again, and yep, even at 0.00000001 I could see jaggies at the
: plane/iso boundary.  Considerably smoother than at 0.001 but still obviously
: shows the difficulty there is in tracing that boundary.

  'accuracy' is not the only thing that affects the quality of the calculation.
Also 'max_gradient' and 'method' may have something to do with it.

: Hm, what about the 'contained_by' objects?  That have the same trouble?

  The object specified in the 'contained_by' block is just the same as
the regular primitive, and thus mathematically correct.
  You should always specify the smallest 'contained_by' object possible
(which of course contains the volume you want to see from the isosurface).
It helps the accuracy. (For example if your isosurface is a sphere of
radius 1, don't make a container object of radius 100 but of radius 1).

-- 
main(i,_){for(_?--i,main(i+2,"FhhQHFIJD|FQTITFN]zRFHhhTBFHhhTBFysdB"[i]
):_;i&&_>1;printf("%s",_-70?_&1?"[]":" ":(_=0,"\n")),_/=2);} /*- Warp -*/

Post a reply to this message

"ingo" <ing### [at] homenl> wrote in message
news:8F80D55A1seed7@204.213.191.228...
|
| Did you use the max_trace / all_intersections option in the isosurface?

Thanks, just tried it and unfortunately only a slower render.  I think
mostly I'm overly sensitive to the boundaries because the surfaces shift
from isosurface to flat plane suddenly.  About like having a blob of gelatin
in the corner of a room.  The change from smooth curving surface to
straight-lined wall is a large difference.
In fact if I do turn up the AA sufficiently there's a reasonable smoothing,
and I'm also neglecting to mention that jitter can't be used to break up the
stair steps of the jaggies either while doing animation.
I still see these darkened seams along the two surfaces which doesn't get
affected much, if at all, by AA but it's very view angle dependent.  Seen
from glancing angles it looks like separations almost.  This is a reflective
and semitransparent plane just like another time I saw this happening.
Again, I think Warp is totally right about the accuracy keyword.  Partly my
fault for being over zealous while working (playing) at this.

Bob

Post a reply to this message

"Warp" <war### [at] tagpovrayorg> wrote in message
news:39833716@news.povray.org...
|   You should always specify the smallest 'contained_by' object possible
| (which of course contains the volume you want to see from the isosurface).

Oh good gosh!  I had it approx. 1.6 units higher than needed to be!  Making
it about 5 times higher in the y than the tallest mountains.
Dumb mistake.  I think when I was adjusting to get correct mountain tops it
was something else causing faulty peaks so I had the upper limit to the
container way up.
Sure glad you said that.

Bob

Post a reply to this message

so after all of this have you tested isosurface interaction with other
shapes yet?  how about replacing your plane with a huge cube?  i'd be
interested in your results; especially if you ever get the jaggies to go
away.  keep us posted.

Bob Hughes wrote:
> 
> "Warp" <war### [at] tagpovrayorg> wrote in message
> news:39833716@news.povray.org...
> |   You should always specify the smallest 'contained_by' object possible
> | (which of course contains the volume you want to see from the isosurface).
> 
> Oh good gosh!  I had it approx. 1.6 units higher than needed to be!  Making
> it about 5 times higher in the y than the tallest mountains.
> Dumb mistake.  I think when I was adjusting to get correct mountain tops it
> was something else causing faulty peaks so I had the upper limit to the
> container way up.
> Sure glad you said that.
> 
> Bob

Post a reply to this message

"ryan constantine" <rco### [at] yahoocom> wrote in message
news:39835221.CE9E0252@yahoo.com...
| so after all of this have you tested isosurface interaction with other
| shapes yet?  how about replacing your plane with a huge cube?  i'd be
| interested in your results; especially if you ever get the jaggies to go
| away.  keep us posted.

I'm convinced now it's all in the 'accuracy' and nothing else.  When trying
to render as fast as possible for both a decent render and render time it's
not good to use accuracy 0.00001 or less, that's why I was getting horrible
effects at 0.01 and still not good at 0.001.  I think I can get by okay with
0.0001 though, but it is much slower than would otherwise be.  Another
problem is the AA depth, can't use +r2 for instance regardless of the +a
setting.  In other words, I often cheat on the settings for animations
compared with stills because the movement typically hides the imperfections.
Oh, I tried as you suggested, a box is no different that I can tell.
Basically just human error on my part, the computer can do a great
isosurface if you let it.
I still haven't figured the mysterious transparent rotating object when
motion blur is applied which I was encountering long ago.  I happened across
it again a couple months ago and then it vanished again, figuratively
speaking, before I was able to keep the pov script as it had been at the
moment.
I don't want to cry wolf all the time so I hope my observations aren't taken
as such.

Bob

Post a reply to this message

no fault, Bob. I try to break code all the time ;)

Grim

Post a reply to this message