«

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?

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 here:
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
<(E-Mail Removed)> 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.

In article <(E-Mail Removed)>, Alan Browne
<(E-Mail Removed)> 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.

In article <(E-Mail Removed)>, Alan Browne
<(E-Mail Removed)> 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.

On Wed, 12 Aug 2009 18:14:26 -0400, Alan Browne
<(E-Mail Removed)> wrote:

>nospam wrote:
>> In article <(E-Mail Removed)>, Alan Browne
>> <(E-Mail Removed)> 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.

It depends whether you are a man or a woman.



Eric Stevens

Alan Browne <(E-Mail Removed)> wrote in
news:(E-Mail Removed) :

> 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.

That might depend on what they mean by +6dB and -2dB.

Those figures are pretty meaningless without a frame of reference.

> 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?).

The typical Bayer CFA has 2 greens for every red and blue. That means
lower noise in the green channel, and higher resolution in the green
channel, both things which are most appreciated in the green channel, even
if a converter doesn't try to abstract luminance. Beyond that, the green-
filtered pixels are more sensitive to white light, typically 1/2 stop
better than the blue channel, and 1 stop better than the red channel.
Therefore, with white light, a white or gray subject will have photon
collection roughly 0.25 : 1 : 0.35 ( R : G : B ).

Alan Browne wrote:
> John Sheehy wrote:
>> Alan Browne <(E-Mail Removed)> wrote in
>> news:(E-Mail Removed) :
>>> 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.
>>
>> That might depend on what they mean by +6dB and -2dB.
>>
>> Those figures are pretty meaningless without a frame of reference.
>
> Based on the news release I reasonably assume it's unfiltered (no
> colour) figures for the amount of signal and amount of noise v. a
> comparable conventional (front lit) sensor under the same conditions
> (temp, lighting, test target).
>
>>> 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?).
>>
>> The typical Bayer CFA has 2 greens for every red and blue. That means
>> lower noise in the green channel, and higher resolution in the green
>> channel, both things which are most appreciated in the green channel,
>> even if a converter doesn't try to abstract luminance. Beyond that,
>> the green-
>> filtered pixels are more sensitive to white light, typically 1/2 stop
>> better than the blue channel, and 1 stop better than the red channel.
>> Therefore, with white light, a white or gray subject will have photon
>> collection roughly 0.25 : 1 : 0.35 ( R : G : B ).
>
> All understood, but I was referring to the claim by "Me" about a 1:6:1
> (RGB) sensor site ratio (which you mysteriously snipped out).
>
>
Please don't attribute the "claim" to me - as if it was some speculation
by me based on rumour alone.
It's pretty clear from Sony's data that they use "Clearvid" RGB pattern
array on "Exmor R" (back illuminated) sensors, at least for video. For
still cameras, I haven't seen any information, but with the merging of
the two (still and video) technologies, then it wouldn't surprise me if
compact cameras with video capability come with "clearvid" type sensor.

In article <(E-Mail Removed)>, Alan Browne
<(E-Mail Removed)> 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?
>
> What photo are you referring to?

any 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.
>
> Do you mean colour dynamic resolution or spatial resolution between colours?

i mean luminance versus chrominance resolution of the human eye. the
eye can resolve chrominance 1/10 that of luminance, and bayer samples
it at half, so a bayer sensor actually has 5x as much chroma as the eye
can resolve. the sony sensor cuts to closer what human vision can see.

nospam wrote:
> In article <(E-Mail Removed)>, Alan Browne
> <(E-Mail Removed)> 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?
>> What photo are you referring to?
>
> any 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.
>> Do you mean colour dynamic resolution or spatial resolution between colours?
>
> i mean luminance versus chrominance resolution of the human eye. the
> eye can resolve chrominance 1/10 that of luminance, and bayer samples
> it at half, so a bayer sensor actually has 5x as much chroma as the eye
> can resolve. the sony sensor cuts to closer what human vision can see.
>
But then there's the problem that we don't perceive "blown" / posterised
colour channels with our eyes, but it is an issue with image capture.
Many less (and correspondingly small) R&B sensor sites don't bode well.