32 Megapixel Canon?

Started Jan 25, 2008 | Discussions thread
bobn2 Forum Pro • Posts: 56,311
No, not wrong

JJ Simpson wrote:

bobn2 wrote:

Cultured Vulture wrote:

bobn2 wrote:

Scale up the G9 pixel density to a FF sensor, you get 239MPixel. The
problem's never been making sensors with lots of pixels, it's getting
the noise performance, but more importantly, providing the processing
and memory power to handle all those pixels. A more interesting
alternative (as hinted in posts above) would be to use tiny pixels,
but oversample, that is a physical 200MPixel sensor, producing
20MPixel output, with digital low pass filtering applied, instead of
a physical AA filter. I suspect what stops it is availability of low
cost digitisation and signal processing with sufficient bandwidth to
make it work at a sensible rate.

I love it, what size card could you use. 1 TB for how many shots?
Diffraction limited at 1.4.
Respond to rudeness with civility, it really annoys them.



You've not understood. The camera would produce 20MPix output, and
would need the same size card as a 20MPix camera. Resolution, and
therefore diffraction limiting, would be just the same as a 20MPix
camera. Effectively what it would be doing was merging 10 tiny sub
pixels into one output pixel. This is what oversampling is about -
take many more samples than you actually need, do the signal
processing digitally. This would allow 'software' low pass and colour
separation filters which can be tailored much more effectively than
pieces of glass.

That maybe only true for jpg but for RAW, I think you'll find that
all 60mp (20x3)
would be recorded to the card (80MB - 14bit) and the raw converter
would do the oversampling and output the final image. Thats why the
Foveon (Fuji) sensor raw files are so big.

Believe me, I do understand this stuff. Your definition of 'RAW' is based entirely on current camera technology, not an oversampled camera. The output of this camera would be 20MPix, derived from 180,000,000 'sub pixels' assuming three times oversampling in each linear dimension (your sums are even wrong). I'm not talking about RGB subpixels, I'm talking about oversampling. It's a completely different principle. The image would be oversampled using tiny pixels, and processed to produce output at a lower resolution - similar to 'pixel binning' but done using a digital signal processor, and therefore more tunable. That in turn may be compressed using JPEG as with any 20MPix raw output.

BTW, Foveon and Fuji sensors are completely different to each other, in principle and construction, and neither is the same as the oversampled sensor described here.

Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow