High Iso article

(you can make a 150mm f2 for 4/3rds, but you can't make a 50mm f0.6 for a 1/2.5" digicam. Costs increase with lower f-stops, even if light gathering ability stays the same. And f0.6 is somewhat outside the realms of possibility, never mind cost effectiveness.)

Joe0Bloggs said:

pidera said:

What you got is a 'how much better should the lenses be' index.

Read 'comparing different camera systems' :
http://forums.dpreview.com/forums/read.asp?forum=1018&message=16738985

and 'the high iso myth' :
http://forums.dpreview.com/forums/read.asp?forum=1018&message=23033618

Bigger sensors mainly help you out with DOF, not so much with noise.

Rgds, Pieter

Click to expand...

--
R.I.P. my K.M. 5-D 16-4-2007
Ignore everything I post here from now on. All postings 100% pure
hot air completely detached from reality
(well, except for those posts that aren't about taking pictures in
the first place)

Click to expand...

--
R.I.P. my K.M. 5-D 16-4-2007

Ignore everything I post here from now on. All postings 100% pure hot air completely detached from reality

(well, except for those posts that aren't about taking pictures in the first place)

Erik37 said:

Virvatulet said:

I must admit that by the first reading this “test”
didn’t make any sense for me.

Click to expand...

What he shows is how it would look, if the FZ50 sensor was made large
enough for a DSLR, not by making the pixels larger, but by making more
of them.

Click to expand...

It's impossible to 'make the FZ50 sensor large enough for a DSLR' and retain the same per-pixel noise performance, so this 'test' is rather misleading.

Joe0Bloggs said:

Erik37 said:

Virvatulet said:

I must admit that by the first reading this “test”
didn’t make any sense for me.

Click to expand...

What he shows is how it would look, if the FZ50 sensor was made large
enough for a DSLR, not by making the pixels larger, but by making more
of them.

Click to expand...

It's impossible to 'make the FZ50 sensor large enough for a DSLR'
and retain the same per-pixel noise performance, so this 'test' is
rather misleading.

Click to expand...

Ok, how about

"What he shows is how it would look, if the FZ50 sensor's pixels
were made as big as the 10D's, keeping the sensor the same 1/1.8"
size."

Just my two oere
Erik from Sweden, main tool: F Z 5

Erik37 said:

Ok, how about

"What he shows is how it would look, if the FZ50 sensor's pixels
were made as big as the 10D's, keeping the sensor the same 1/1.8"
size."

Click to expand...

It's impossible to interpret his images to mean such a thing... if you made the pixel pitch of the two sensors identical, then using the same focal length you would lay the same number of pixels over the target area. The result would be quite different from what you see here. And I thought his whole point was in trying to show that the FZ50 can equal the 10D in performance using its smaller pixel pitch? (but this is misleading as I pointed out above.)

hum. Does that make sense?

-he talks about having "binned" the pixels. Not sure if the panny is even able to bin pixels. I assume he averaged out noise by downsizing?

-he uses the photoshop zoom function (a simple next neighbor interpolation I guess) to compare the 10D shot at 272% (sic) with a 100% crop from the panny. This definitely doesnt make any sense. Compare a scanning back shot at 300% zoom with the panny at 100% and the panny will look better.

-given the different sensor sizes, it also doesnt make sense to use the same focal length. Its a truism that, given a sufficiently large sensor size disparity, the frame-filling 8MP shot will look better than a crop from a bigger sensor which is otherwise better.

I doubt the panny shot is really binned but even if it is, the comparison is very flawed and from my own testing binning is not useful.

O.
--
http://www.flickr.com/photos/ollivr/
http://www.flickrleech.net/user/ollivr

would be stacking exposures to average out noise.
I was glad to be pointed to a paper in this thread:

http://forums.dpreview.com/forums/read.asp?forum=1027&message=22990225

If I remember correctly the executive summary is you need 4 frames for one stop, 8 for two, 16 for three stops worth of noise reduction and so on...

O.
--
http://www.flickr.com/photos/ollivr/
http://www.flickrleech.net/user/ollivr

Joe0Bloggs said:

Erik37 said:

"What he shows is how it would look, if the FZ50 sensor's pixels
were made as big as the 10D's, keeping the sensor the same 1/1.8"
size."

Click to expand...

It's impossible to interpret his images to mean such a thing... if
you made the pixel pitch of the two sensors identical, then using
the same focal length you would lay the same number of pixels over
the target area.

Click to expand...

Make everthing identical and it will look identical, obviously.

Joe0Bloggs said:

The result would be quite different from what you see here.

Click to expand...

Naturally since the pixel pitch is very different here.

Joe0Bloggs said:

And I thought his whole point was in trying to show that
the FZ50 can equal the 10D in performance using its smaller pixel
pitch?

Click to expand...

No it wasn't. The FZ50 has a much smaller sensor, this will always make
its performance worse.

The point was, looking at a sensor at a given size, what is the effect
of making either many small pixels or fewer big ones.

If we want to know the effect of the megapixel race, this is a more
useful question than comparing 100% crops of sensors of different
size, particularly JPG output from them.

Just my two oere
Erik from Sweden

I think you may have misunderstood the purpose of this test.

Tiffles said:

-he talks about having "binned" the pixels.

Click to expand...

He means downsampled in software.

light_reaction said:

-he uses the photoshop zoom function (a simple next neighbor
interpolation I guess)

Click to expand...

Looks like bicubic to me. Next neighbour would show strong jaggies.

light_reaction said:

to compare the 10D shot at 272% (sic) with a
100% crop from the panny. This definitely doesnt make any sense.

Click to expand...

It does, not for deciding which camera can achieve higher image quality,
but for seeing the effect of "pixel stuffing." A least it's the best test
I've seen so far.

light_reaction said:

-given the different sensor sizes, it also doesnt make sense to use
the same focal length.

Click to expand...

It's the only thing that makes sense given the question he tried to
answer.

Just my two oere
Erik from Sweden

Erik37 said:

Tiffles said:

-he talks about having "binned" the pixels.

Click to expand...

He means downsampled in software.

Click to expand...

Which is binning, isn't it? If not, I need to change some comments in some source code, and I've a feeling that some of the authors of the google results I just got for "binning in software" need to be told, too...

--
Peter

Erik37 said:

It does, not for deciding which camera can achieve higher image
quality,
but for seeing the effect of "pixel stuffing." A least it's the
best test
I've seen so far.

Click to expand...

It's impossible to scale the FZ50 sensor up to the size of the 10D and retain the same per-pixel noise level, so this 'comparison' is rather misleading.