|
![](/i/fill.gif) |
AlexLens wrote:
>> Making a lens out of a lathe works very poorly. The reason is that
>> none of the available spline types will give you a G2-continuous [*]
>> connection on the control points. However G2-continuity is essential
>> to optical surface design.
>>
>> Since most lenses are modeled with polynomials anyway, I would
>> advise you to use an isosurface with the corresponding polynomial.
>>
>> Jerome
>>
>> [*]
>> http://en.wikipedia.org/wiki/Geometric_continuity#Geometric_continuity
>> --
>> mailto:jeb### [at] free fr
>> http://jeberger.free.fr
>> Jabber: jeb### [at] jabber fr
> -----------------------------------------------------------------------
--------
>
> Thanks for the reply!
>
> I had no clue the cubic splines were not G2 continuous as they usually
are. The
> documentation page for the cubic spline makes it seem like it is necess
arily G1
> continuous, however. I'm guessing that POV-Ray only needs to know the
first
> derivative to know how a photon would interact with a surface, so this
method
> still should work.
>
Note that you example used quadratic splines, not cubic... However,
even cubic spline is not G2 continuous (it is C2, but the 2nd
derivative is 0 on the control points, which breaks the G2
property). It is indeed G1 and pov will have no problem knowing how
a photon interacts with the surface. The reason you need G2
continuity isn't because of POV, but because of physics: the focal
point is different on either side of a non G2 point and the
difference is then amplified as the light travels away from the
surface, which quickly makes for very big differences.
> As for the Isosurface, I don't have an analytic function to input to PO
V-Ray, so
> maybe that wouldn't work?
>
No, if you can't get an analytic function it won't work. Where do
your spline control points come from?
> What do you all recommend as the fastest method to make this lens just
given the
> numerical data I have?
>
I did some research a couple of years ago on how to connect Bezier
curves in a G2 way. I can try to dig it up if you are interested.
However, IIRC the constraints it imposes on the curve mean that you
probably won't be able to use it.
Jerome
--
mailto:jeb### [at] free fr
http://jeberger.free.fr
Jabber: jeb### [at] jabber fr
Post a reply to this message
Attachments:
Download 'us-ascii' (1 KB)
|
![](/i/fill.gif) |
|
![](/i/fill.gif) |
Am 14.06.2010 21:30, schrieb AlexLens:
> Right now I am trying to intersect it with a plane to bound the flat side of the
> lens. Is that how you would close this?
No, you'll really need to close the lathe object itself, by placing both
the first and last point on the axis.
> Also, are some of my settings unnecessary, it seems to be taking longer than I
> would expect it to..maybe this method overall is not the best way to implement
> this lens from data points?
I'd suggest linear spline; with that many data points, it shouldn't make
any noticeable difference in precision, and you'll have a much easier
job closing it with a plane.
Post a reply to this message
|
![](/i/fill.gif) |