Sony Exmor R ("back illuminated") sensor in production

http://www.dpreview.com/news/0908/09080601sonycmos.asp

It is in production on the compacts now, so maybe in a year or so for
DSLR's.

http://www.dpreview.com/news/0908/09...ydsctx1wx1.asp

RichA writes:
It is in production on the compacts now, so maybe in a year or so for
DSLR's.

Then some know-nothing will pipe-up that it's "best suited to small
sensors." Fairchild has a 16M medium format sensor that is back lit.

Those two facts aren't incompatible...

-miles

--
`There are more things in heaven and earth, Horatio,
Than are dreamt of in your philosophy.'

RichA wrote:

Alan Browne wrote:
http://www.dpreview.com/news/0908/09080601sonycmos.asp

It is in production on the compacts now, so maybe in a year or so for
DSLR's.

http://www.dpreview.com/news/0908/09...ydsctx1wx1.asp

Then some know-nothing will pipe-up that it's "best suited to small
sensors." Fairchild has a 16M medium format sensor that is back lit.
I've never seen an answer to a question I had about Sony's Exmor R.
Their site claim a one-stop advantage. DPR article qualifies this by
stating there's nearly a one stop gain based on pixel size of 1.75 um =
~120 megapixels on an APS-C sized sensor. The diagrams on the Sony site
clearly show Exmor R with a Bayer array which has only half the number
of R & B photosites. Per 8 photosites, 6 are green, one is red, one is
blue. Do claims for Exmor R rely on this RGB pixel matrix?
Another maker (Kodak?) a few years ago claimed huge advantages by using
some white photosites interspersed in the bayer RGBG array. What
happened with that idea?

Me wrote:
RichA wrote:

Alan Browne wrote:
http://www.dpreview.com/news/0908/09080601sonycmos.asp

It is in production on the compacts now, so maybe in a year or so for
DSLR's.

http://www.dpreview.com/news/0908/09...ydsctx1wx1.asp

Then some know-nothing will pipe-up that it's "best suited to small
sensors." Fairchild has a 16M medium format sensor that is back lit.
I've never seen an answer to a question I had about Sony's Exmor R.
Their site claim a one-stop advantage. DPR article qualifies this by
stating there's nearly a one stop gain based on pixel size of 1.75 um =
~120 megapixels on an APS-C sized sensor. The diagrams on the Sony site
clearly show Exmor R with a Bayer array which has only half the number
of R & B photosites. Per 8 photosites, 6 are green, one is red, one is
blue. Do claims for Exmor R rely on this RGB pixel matrix?
Another maker (Kodak?) a few years ago claimed huge advantages by using
some white photosites interspersed in the bayer RGBG array. What
happened with that idea?

Who knows. Unless you have all the info it's hard to do the sums.

I sort of agree with RichA's "know-it-all"'s remark above (wrt to the
"best suited to small sensors" claim) but I can't see why, once mastered
in fab, that it can't be scaled up to APS-C, FF or MF unless it has a
very high defect rate. But then we never believed that FF would become
as common as it has in the recent 2 years.

Sony's basic claim for the backlit sensor for these first cameras is:

+6dB signal
-2dB noise

for 8 dB total improvement.

Which is about 1.3 stops in that sense. However, such could also mean
that "ISO 250" has the same noise figure as a "natural" ISO 100 sensor
which is probably the right way to state it. (IOW: the natural ISO is
even further away from a very desirable natural ISO 25 or 50 sensor).

Then, because luminance is weakest in the green channel, that is the one
that is doubled in most sensors (with blue/red having the same weight).
I've not seen sensors as highly green weighted as you describe above
(got a link?).

Alan Browne wrote:
Me wrote:
RichA wrote:

Alan Browne wrote:
http://www.dpreview.com/news/0908/09080601sonycmos.asp

It is in production on the compacts now, so maybe in a year or so for
DSLR's.

http://www.dpreview.com/news/0908/09...ydsctx1wx1.asp

Then some know-nothing will pipe-up that it's "best suited to small
sensors." Fairchild has a 16M medium format sensor that is back lit.
I've never seen an answer to a question I had about Sony's Exmor R.
Their site claim a one-stop advantage. DPR article qualifies this by
stating there's nearly a one stop gain based on pixel size of 1.75 um
= ~120 megapixels on an APS-C sized sensor. The diagrams on the Sony
site clearly show Exmor R with a Bayer array which has only half the
number of R & B photosites. Per 8 photosites, 6 are green, one is
red, one is blue. Do claims for Exmor R rely on this RGB pixel matrix?
Another maker (Kodak?) a few years ago claimed huge advantages by
using some white photosites interspersed in the bayer RGBG array.
What happened with that idea?

Who knows. Unless you have all the info it's hard to do the sums.

I sort of agree with RichA's "know-it-all"'s remark above (wrt to the
"best suited to small sensors" claim) but I can't see why, once mastered
in fab, that it can't be scaled up to APS-C, FF or MF unless it has a
very high defect rate. But then we never believed that FF would become
as common as it has in the recent 2 years.

It's all conjecture until a product actually arrives. It would be nice
if some of the optimistic conjecturing was true, but I doubt it until I
see it with my own eyes.


Sony's basic claim for the backlit sensor for these first cameras is:

+6dB signal
-2dB noise

for 8 dB total improvement.

Which is about 1.3 stops in that sense. However, such could also mean
that "ISO 250" has the same noise figure as a "natural" ISO 100 sensor
which is probably the right way to state it. (IOW: the natural ISO is
even further away from a very desirable natural ISO 25 or 50 sensor).
Not quite sure what they're comparing it with though. It seems perhaps
they're comparing an older CMOS design with back illuminated Exmor R
with column A/D converters he
http://www.sony.jp/products/overseas...r_r/index.html
Newer conventional CMOS with column AD converters in dslrs (D300/90/3x,
a700 and a900) showed some impressive gains (compare D300 with D2x).

Then, because luminance is weakest in the green channel, that is the one
that is doubled in most sensors (with blue/red having the same weight).
I've not seen sensors as highly green weighted as you describe above
(got a link?).
Is luminance weakest in the green channel?
Here's Sony's explanation for "clearvid".
"The method used to generate green is especially important because of
the major contribution made by green to resolution"
For video, then sensors are oversampling and colour resolution at (ie)
1080i downsampled from a 10 or 12mp clearvid sensor shouldn't be a
problem. But this wouldn't bode well for over-saturated red channel or
colour resolving at full resolution.
Of course no reason why back-illumination requires clearvid RGB array -
but all the data I've seen from sony shows Exmor R with clearvid.
http://www.sony.net/SonyInfo/technol...02.html#page04

On Tue, 11 Aug 2009 22:07:37 -0400, Alan Browne
wrote:


Then, because luminance is weakest in the green channel, that is the one
that is doubled in most sensors (with blue/red having the same weight).
I've not seen sensors as highly green weighted as you describe above
(got a link?).

Where on earth did you read and start to believe this nonsense? It is
because the human eye is most perceptive to the green part of the spectrum,
having evolved and living under a green star (the sun), that luminance is
MOSTLY derived from the green channel. This is why 2 green sensors are more
appropriate to mimic human perception. Go educate yourself.

Me wrote:
Alan Browne wrote:
Me wrote:
RichA wrote:

Alan Browne wrote:
http://www.dpreview.com/news/0908/09080601sonycmos.asp

It is in production on the compacts now, so maybe in a year or so for
DSLR's.

http://www.dpreview.com/news/0908/09...ydsctx1wx1.asp

Then some know-nothing will pipe-up that it's "best suited to small
sensors." Fairchild has a 16M medium format sensor that is back lit.
I've never seen an answer to a question I had about Sony's Exmor R.
Their site claim a one-stop advantage. DPR article qualifies this by
stating there's nearly a one stop gain based on pixel size of 1.75 um
= ~120 megapixels on an APS-C sized sensor. The diagrams on the Sony
site clearly show Exmor R with a Bayer array which has only half the
number of R & B photosites. Per 8 photosites, 6 are green, one is
red, one is blue. Do claims for Exmor R rely on this RGB pixel matrix?
Another maker (Kodak?) a few years ago claimed huge advantages by
using some white photosites interspersed in the bayer RGBG array.
What happened with that idea?

Who knows. Unless you have all the info it's hard to do the sums.

I sort of agree with RichA's "know-it-all"'s remark above (wrt to the
"best suited to small sensors" claim) but I can't see why, once
mastered in fab, that it can't be scaled up to APS-C, FF or MF unless
it has a very high defect rate. But then we never believed that FF
would become as common as it has in the recent 2 years.

It's all conjecture until a product actually arrives. It would be nice
if some of the optimistic conjecturing was true, but I doubt it until I
see it with my own eyes.


Sony's basic claim for the backlit sensor for these first cameras is:

+6dB signal
-2dB noise

for 8 dB total improvement.

Which is about 1.3 stops in that sense. However, such could also mean
that "ISO 250" has the same noise figure as a "natural" ISO 100 sensor
which is probably the right way to state it. (IOW: the natural ISO is
even further away from a very desirable natural ISO 25 or 50 sensor).
Not quite sure what they're comparing it with though. It seems perhaps
they're comparing an older CMOS design with back illuminated Exmor R
with column A/D converters he
http://www.sony.jp/products/overseas...r_r/index.html

Newer conventional CMOS with column AD converters in dslrs (D300/90/3x,
a700 and a900) showed some impressive gains (compare D300 with D2x).

Then, because luminance is weakest in the green channel, that is the
one that is doubled in most sensors (with blue/red having the same
weight). I've not seen sensors as highly green weighted as you
describe above (got a link?).
Is luminance weakest in the green channel?
Here's Sony's explanation for "clearvid".
"The method used to generate green is especially important because of
the major contribution made by green to resolution"
For video, then sensors are oversampling and colour resolution at (ie)
1080i downsampled from a 10 or 12mp clearvid sensor shouldn't be a
problem. But this wouldn't bode well for over-saturated red channel or
colour resolving at full resolution.
Of course no reason why back-illumination requires clearvid RGB array -
but all the data I've seen from sony shows Exmor R with clearvid.
http://www.sony.net/SonyInfo/technol...02.html#page04


I always understood it to be weakest as opposed to best resolution and
why there are more greens. Resolution is reason enough as well.
Whatever the underlying reason, it still seems excessive that a sensor
would put 6 green to 1 each red and blue as opposed to the more
conventional 1:2:1 in other sensors.

In article , Alan Browne
wrote:

I always understood it to be weakest as opposed to best resolution and
why there are more greens. Resolution is reason enough as well.
Whatever the underlying reason, it still seems excessive that a sensor
would put 6 green to 1 each red and blue as opposed to the more
conventional 1:2:1 in other sensors.

the eye's colour resolution is actually not all that good.

nospam wrote:
In article , Alan Browne
wrote:

I always understood it to be weakest as opposed to best resolution and
why there are more greens. Resolution is reason enough as well.
Whatever the underlying reason, it still seems excessive that a sensor
would put 6 green to 1 each red and blue as opposed to the more
conventional 1:2:1 in other sensors.

the eye's colour resolution is actually not all that good.

Talking about sensors, however.

In article , Alan Browne
wrote:

I always understood it to be weakest as opposed to best resolution and
why there are more greens. Resolution is reason enough as well.
Whatever the underlying reason, it still seems excessive that a sensor
would put 6 green to 1 each red and blue as opposed to the more
conventional 1:2:1 in other sensors.

the eye's colour resolution is actually not all that good.

Talking about sensors, however.

were you planning on not ever looking at the photo?

if the image is intended for human viewing, then how much the eye can
resolve is very important. why capture what you won't be able to see?
that's why bayer sensors work so well and why foveon is a solution in
search of a problem.

on the other hand, if it's for image analysis then you'd probably want
better colour resolution.