Marc-Hendrik Bremer wrote:
> 
> While you are right that higher accuracy values decrease the level of
> detail, it has other influences, too.
> I just tried it with an old scene and attached the resulting image. I
> increased the accuracy value by far (0.1 instead of 0.000001) and left
> anything else the way it was (after converting to 3.5 language of course).
> Pov reports, that I should decrease max_gradient to lower rentertime  (5.141
> instead of 6). In my opinion that black part is one of those isosurface
> artefacts.
> 

From just the picture i can't conclude much, but Povray's suggestion that
max_gradient is too high does not necessarily need to be correct.  

Christoph

-- 
Christoph Hormann <chr### [at] gmxde>
IsoWood include, radiosity tutorial, TransSkin and other 
things on: http://www.schunter.etc.tu-bs.de/~chris/

Post a reply to this message

Wasn't it nospam who wrote:

>       Been playing with isosurfaces in 3.5 and I am wondering
>if the attached image is a bug or not.  It happens using
>f_noise3d and with f_noise_generator using generator 0 or 1.
>Using f_noise_generator with a generator of 2 does not produce
>the problem.
>       .pov source posted on povray.scenes.text-files

I guess it's way too late to suggest that max_gradient ought to be made
mandatory for isosurfaces.

Or warn "Warning: isosurface has no max_gradient set. The isosurface may
contain holes! Set a correct max_gradient value to get a proper
rendering of the isosurface."

Or let the default value of max_gradient be changed to a value that's
likely to look OK for simple surfaces. I'd suggest a default value of
10.

As things currently stand we're likely to be seeing an *awful* lot of
occurrences of this question in the future.



-- 
Mike Williams
Gentleman of Leisure

Post a reply to this message

Post a reply to this message

Marc-Hendrik Bremer wrote:
> 
> That's undoubted (now Pov found a max_gradient of 6.294), but my point was,
> that setting the accuracy value to high, can lead to artefacts. I'll post a
> simplified code for this in p.t.s-f (under "Low accuracy values in
> Isosurfaces"). Output changes a bit (because of a simplified texture), but
> the behaviour should be reproducible. To exclude that this is a max_gradient
> problem I just rendered it with max_gradient 60. Those black parts remained.
> Perhaps it's another problem, but I can't imagine which.
> 

I can see what you mean.  But these artefacts are different from those
resulting of insufficient max_gradient.  I don't think they are something
new users have to struggle with very often.  

I don't know much about the isosurface algorithm, but maybe that's even
something that could be fixed.

BTW, i recognize some of my textures in that scene...

Christoph

-- 
Christoph Hormann <chr### [at] gmxde>
IsoWood include, radiosity tutorial, TransSkin and other 
things on: http://www.schunter.etc.tu-bs.de/~chris/

Post a reply to this message

Post a reply to this message

Marc-Hendrik Bremer wrote:
> 
> max_gradient problems are far more likely, I agree. There poped up just
> another one in p.b.p-t. Perhaps Warp should also mention that it's a good
> idea to make the container object as small as possible, since you can lower
> max_gradient with a smaler container wuite a bit and things will speed up.
> 

I don't think a smaller container can reduce required max_gradient (unless
the high gradient only occurs outside the smaller container object).  

Making container objects as small as possible should be quite obvious. 
It's also mentioned in detail in the documentation, so i don't think it's
required to put it in the FAQ unless Warp wants to make a general
'isosurface speed' section.

> 
> You think, that's a bug? I would say it's a user error.
> 

I would not say it's a bug, but it could be that the isosurface code can
be modified to avoid it.

Christoph

-- 
Christoph Hormann <chr### [at] gmxde>
IsoWood include, radiosity tutorial, TransSkin and other 
things on: http://www.schunter.etc.tu-bs.de/~chris/

Post a reply to this message

Post a reply to this message

Marc-Hendrik Bremer wrote:
> 
> But a bigger container can require a higher max_gradient (see p.b.b-t
> "isosurface contained_by or evaluation problem"). I'm not sure if this is an
> expectable behaviour, but it seems somewhat logical to me. I finally found
> the description of the isosurface search method from R. Suzuki in p.general:
> RE: Isosurface and function pattern in v3.5
> (Message-ID<3b984214@news.povray.org> if that helps anything). If the points
> d1 and d2 he mentions are further apart, isn't a greater max_gradient
> needed?

No. The gradient of a function is a local property:  <df/dx, df/dy,
df/dz>  The max_gradient value is used to determine if there's a possible
intersection between two points (of arbitrary distance) under the
condition that the gradient between the two evaluated points is maximum.

> But that's all just guessing of course. But it's clear that if the container
> is to big (for what reason ever), the surface will sometimes (but
> reproducible - no "random-number-thingie" as with those pigment-patterns)
> disappear unless you raise max_gradient.

Because of the things mentioned above changing the container size should
have no influence in whether a surface is visible or not (onless you use
'evaluate' of course) but the scene you mentioned really seems to prove
the opposite.

Christoph

-- 
Christoph Hormann <chr### [at] gmxde>
IsoWood include, radiosity tutorial, TransSkin and other 
things on: http://www.schunter.etc.tu-bs.de/~chris/

Post a reply to this message

Post a reply to this message

Marc-Hendrik Bremer wrote:
> 
> Please have a look at Mike Williams reply to "isosurface contained_by or
> evaluation problem" which describes much better than I can, why the size of
> the container has an influence.
> 

Yes, saw it.  Thanks.

In fact i mentioned this possibility in

news://news.povray.org/3BBEE00B.2E13A766%40gmx.de

but then forgot about it myself. :-)

Christoph

-- 
Christoph Hormann <chr### [at] gmxde>
IsoWood include, radiosity tutorial, TransSkin and other 
things on: http://www.schunter.etc.tu-bs.de/~chris/

Post a reply to this message