The matt surface is the surface that the finder image is formed on and so
should be closest to the mirror.
That doesn't sound logical to me. The reference plane will be where the
screen beds against the body, which will be the upper surface. It
doesn't matter that the image goes through the screen - it's
transparent/colourless. As long as the gg is aligned with the reference
plane, focus (having been so before) should be correct. Only if it is
not bedded (and focus would vary across the frame, probably) or
installed the wrong way round will focus be off.
The matte/prism side should be near the pentamirror/pentaprism,
not the mirror. It makes sense because that is a defined plane in the
camera.... the screen can be anything, so long as it's flat against that
surface.
The only thing that bothers me with that argument is that the
index of refraction of the plastic is different from air. That must have
been taken into account in the geometry of the design. So, you'd likely
not be able to change the thickness *too* much. I'd bet it works out that
it'd have to be significantly thicker to screw it up too much.
Just as a note. I'm pretty sure that when I first installed mine
I didn't quite have it in there right. When I actually took a few test
shots at an oblique angle, the screen-focus was slightly different than
the picture taken. I took the screen out, rotated 180 degrees (did NOT
flip it over. It wasn't *that far out), and more carefully put it back
in. Here's the result when focusing on the line between the 9 and 10
towards the middle of the chart.
http://www.ee.vt.edu/~mythtv/imgp3684.jpg
... and here focusing on a piece of tape on a cinder block wall at
about a 20 degree angle to the wall.
http://www.ee.vt.edu/~mythtv/imgp3685.jpg
At f/1.4 seems dead-nuts on to me.
-Cory
--
*************************************************************************
* Cory Papenfuss *
* Electrical Engineering candidate Ph.D. graduate student *
* Virginia Polytechnic Institute and State University *
*************************************************************************
