First, I'd like to say that I really like this watch scene.  Below I've
written a bunch of tips for this scene, but anyone interested in the
photon mapping patch should read them, since they clarify a few things
where my documentation is lacking.

Bob Hughes wrote:
> 
> I used the photon mapping on my pocket watch picture. Having some
> trouble adjusting for the right effect but then what else is new. Think
> much of my difficulties are with the refractive and reflective surfaces
> of the objects themselves. Then again it could be the 'autostop' angle
> setting I haven't gotten exactly right either. I tried
> 'photons_pass_through' on the dome in a test run and got error messages
> saying: "can't use pass through on refractive and high density, turning
> refraction off", or something to that affect.

The photons_pass_through option was designed for situations where you want
to shoot photons at an object, but there is another object between the
light and the target object.  Normally, photons are discarded if they don't
hit their target object directly.  In this example, if you try to shoot
photons at the watch without specifying photons_pass_through for the
dome, none of the photons that are shot at the watch will get there.  They
will all hit the dome first and be discarded.

Just shoot photons at the watch (by specifying a density), and just specify
photons_pass_through for the dome:
  photons { photons_pass_through ignore_photons } // ignore is used to
                                                  // speed things up

> The two main lights have photons{reflection on refraction off}, the

This isn't really necessary.  I'll have more documentation on this in
the future, but for now imagine this:
Let's say you have a scene with three lights and three lenses (I made
this scene once, and I'll use it for a demo scene).  Now, the three
lights are set up like this:

light_source { loc1, color Red   photons{refraction on}}
light_source { loc2, color Green }
light_source { loc3, color Blue  photons{refraction off}}

The three lenses are set up like this:

object { Lens1 photons { density 0.02*phd refraction on} }
object { Lens2 photons { density 0.02*phd } }
object { Lens3 photons { density 0.02*phd refraction off} }

Now, from the first lens, you'll see caustics from the red and green lights.
From the second lens, you'll only see caustics from the red light.
And you won't see any caustics from the third lens.

So if you turn on refraction in your object, you don't have to for the
lights.  I hope this is kind of clear.

> watch metal has density 0.2 and reflection on only also. The dome glass
> is using refraction only with density 0.2 and the watch crystal is
> refraction only with density 0.15. The mirror is using reflection on and
> density 0.2.

The mirror's reflection looks good.
Don't use a density with the dome (it's not worth it... and if you really
want to, then don't shoot anything at the watch... the photons that go
through the dome will hit the watch).

> Here's the global photon_map settings:
> 
>     gather 23,123
>     radius .12,3,.23
>     jitter .3
>     autostop 6

Jitter could go up to 0.4 (that will be the default in the next version).
Autostop angle is in degrees.  If it doesn't stop shooting photons too
soon, then you've found a good value.

For the gather radius, I'd suggest:
  radius 0.12, 2, 0.12

If the gather size is too small, go to something like:
  radius 0.2, 2, 0.2

> Main thing that doesn't seem right is the wrinkles caustics of the watch
> crystal on the left hand side. There was a worse case covering more of
> it before when the gather was set much lower but I don't understand the
> wrinkling effect much. Think it may be the normal on the dome. It seems
> to be defying my constant adjustments thus far anyway. The normal is
> supposed to be large scale and yet small perturbation. Still working on
> that.

It looks like you've got a lot of variation in the normal for the watch
metal's surface.  That could lead to what you see.  A higher photon
density (and smaller gather radius) might help this (as could decreasing
the amount that the normal is perturbed).

> Oh, and wouldn't be complete without some stats:
> 
>  Win95 OSR2, 233MMX Intel Pentium, 256K L2 cache, 64M sdram, 256M swap
> file (permanent).
>  300x300 pixel res.
>  AA 0.3 AAdepth 2 Jitter 0.0
>  43 seconds to parse.
>  1h 28m 26s total render time.
>  Just less than 6 megabytes peak memory used.

Looks kind of missproportonate pre/regular processing (too few photons used).
I'd suggest you to to multiply all of your densities and gather radii by a
variable and then adjust that variable to get more or less photons.

> All in all a very usable new "feature", thankyou so much Nathan.

You're welcome.  I hope these tips make it work better for you for this and
future scenes.

Now that I look at the scene more closely, I can see some splotchiness on
the far wall (viewed in the mirror).  You may wish to increase the gather
radius (relative to the photon-shooting density) to avoid this.

-Nathan

Post a reply to this message