On 14/09/14 19:46, Larry Colen wrote:
Toralf Lund wrote:
On 14/09/14 16:31, Igor PDML-StR wrote:
Zos,
I guess, you were asking about the reference to the effective
resolution of the lenses combined with a sensor.
Here is the link:
http://petapixel.com/2014/09/04/why-i-want-to-switch-to-nikon-but-cant-tony-northrup-throws-gas-on-the-canikon-debate/
(from the thread "OT? In case you are having some insomnia
tonight..." )
http://www.mail-archive.com/pdml@pdml.net/msg703996.html
He starts talking about the interplay of the high resolution lenses
and the sensor shortly after 4:00.
Personally, I'm too lazy to look up that right now, but I'm thinking
that it might make sense to have a sensor resolution of up to 8x the
one
of the lens. I'll leave it as an exercise to find out how I came up
with
that number (told you I was lazy.) But I'm also wondering if one could
make that 6x instead, and also reorganise the sensor (see above.)
Nyquist rate times the bayer pattern.
Double should be good enough to get unaliased luminance data, but you
need to look at the sample rate for each color. What you are really
asking for is that each color be sampled at over the nyquist rate.
I wonder if there are pathological cases where the repeating pattern
is at 45 degrees and you need another 1.4 times the resolution to
avoid aliasing.
What if it is a curvy (meandering) line?
Igor
--
PDML Pentax-Discuss Mail List
PDML@pdml.net
http://pdml.net/mailman/listinfo/pdml_pdml.net
to UNSUBSCRIBE from the PDML, please visit the link directly above and follow
the directions.
