sensitometric curves of digital sensors

Archived from groups: rec.photo.digital (More info?)

> The human eye is said to be able to process a lightness range of about
> 1:2000 or 11 f/stops, while analog negative film and digital sensors
> do only master 8 to 9 f/stops. I know where the limitations come from
> in the analog world, but can somebody tell me what limits the
> technicians from building a ccd-chip that has as much a linear
> sensitometric curve as the human eye ?


Cost would have to be the problem here.

Archived from groups: rec.photo.digital (More info?)

"Roger N. Clark (change username to rnclark)" <username@qwest.net> wrote:
> Mr.Adams wrote:
>> Hi,
>>
>> The human eye is said to be able to process a lightness range of about
>> 1:2000 or 11 f/stops, while analog negative film and digital sensors
>> do only master 8 to 9 f/stops. I know where the limitations come from
>> in the analog world, but can somebody tell me what limits the
>> technicians from building a ccd-chip that has as much a linear
>> sensitometric curve as the human eye ?
>> And : Can someone point me to a website where I find sensitometric
>> curves of current digital cameras ?

That's hard, because there is some processing between exposure and the
data file that you see. I do know, however, that the CCD detectors
themselves are almost linear, and as far as I'm aware CMOS sensors are
too.

> The total range of the human eye is much more than 2000. For
> example, at night one can see magnitude 7+ stars, as well as bright
> planets, e.g. magnitude ~ -3 so about 10 magnitudes, or 10,000
> linear. The daytime light the eye's sensitivity decreases by about
> 600, so we have a total range of on the order of 600*10,000 or ~~6
> million.

> The dynamic range of a CCD sensor is limited to by photon statistics
> by how large the well is that can hold the electrons (one electron
> per photon), and read noise. Sensors like the Canon 10D hold ~44000
> electrons, the 1D Mark II 52,000, and then small P&S sensors much
> less. With read noise on the order of 10 to 15 electrons, divide
> the full well capacity by the read noise for maximum dynamic range.
> So the Canon 1D Mark II has a theoretical dynamic range of ~6500, or
> 12.7 stops, being limited by the 12-bit A to D converter with its
> 8.2 micron pixels. The Canon P&S S60 sensor has 2.8 micron pixels, a
> full well of 22000, and a theoretical dynamic range of 10.7 stops.
> P&S cameras with smaller sensors have degraded performance relative
> to the above.

These are two different things. In case one, you're describing the
dyamic range of the eye in two different situations, night and day.
Apart form anything else, the paerture is greatly different in these
two cases.

In case two, you're describing the dynamic range in a single exposure.
But of couse, with a camera you can increase the aperture at night,
just as the eye does.

Andrew.

Archived from groups: rec.photo.digital (More info?)

bmoag <aetoo@hotmail.com> wrote:
> I think the future of digital imaging is not necessarily increasing
> the light sensitivity range of sensors but in creating processing
> algorithms to turn down the sensitivity of the sensor where it is
> being struck by light that is too bright compared to main exposure
> value of the image: selective electronic dodging.

The well is full; how do you propose to stop it overflowing? It'll
take more than an algorithm to do that.

Andrew.

Archived from groups: rec.photo.digital (More info?)

Roger N. Clark (change username to rnclark) wrote:
> Mr.Adams wrote:
>
>> Hi,
>>
>> The human eye is said to be able to process a lightness range of about
>> 1:2000 or 11 f/stops, while analog negative film and digital sensors
>> do only master 8 to 9 f/stops. I know where the limitations come from
>> in the analog world, but can somebody tell me what limits the
>> technicians from building a ccd-chip that has as much a linear
>> sensitometric curve as the human eye ?
>> And : Can someone point me to a website where I find sensitometric
>> curves of current digital cameras ?
>>
>> Thanks for your input !
>> Mr.Adams
>
>
> The total range of the human eye is much more than
> 2000. For example, at night one can see magnitude
> 7+ stars, as well as bright planets, e.g. magnitude
> ~ -3 so about 10 magnitudes, or 10,000 linear.
> The daytime light the eye's sensitivity decreases
> by about 600, so we have a total range of on the
> order of 600*10,000 or ~~6 million.
>
> The dynamic range of a CCD sensor is limited to by photon
> statistics by how large the well is that can hold the
> electrons (one electron per photon), and read noise.
> Sensors like the Canon 10D hold ~44000 electrons,
> the 1D Mark II 52,000, and then small P&S sensors
> much less. With read noise on the order of 10 to 15
> electrons, divide the full well capacity by the read
> noise for maximum dynamic range. So the Canon 1D Mark II
> has a theoretical dynamic range of ~6500, or 12.7 stops,
> being limited by the 12-bit A to D converter with its
> 8.2 micron pixels. The Canon P&S S60 sensor has
> 2.8 micron pixels, a full well of 22000, and a theoretical
> dynamic range of 10.7 stops. P&S cameras with smaller
> sensors have degraded performance relative to the above.
>
> See:
>
> The Signal-to-Noise of Digital Camera images
> and Comparison to Film:
> http://clarkvision.com/imagedetail [...] l.to.noise
> for more details.
>
> Dynamic Range and Transfer Functions of Digital Images
> and Comparison to Film:
> http://clarkvision.com/imagedetail/dynamicrange2
>
> Roger Clark
> Photography at: http://clarkvision.com
>
Is there some source that I can go to to find the capacity of
pixel wells for various camera sensors?
Is this something that is calculated fron the area of a pixel well?
Does the capacity of a well vary from sensor to sensor even at the
same area/pixel?
Thanks for any leads.....Bob Williams

Archived from groups: rec.photo.digital (More info?)

Bob Williams wrote:

>
>
> Roger N. Clark (change username to rnclark) wrote:
>
>> Mr.Adams wrote:
>>
>>> Hi,
>>>
>>> The human eye is said to be able to process a lightness range of about
>>> 1:2000 or 11 f/stops, while analog negative film and digital sensors
>>> do only master 8 to 9 f/stops. I know where the limitations come from
>>> in the analog world, but can somebody tell me what limits the
>>> technicians from building a ccd-chip that has as much a linear
>>> sensitometric curve as the human eye ?
>>> And : Can someone point me to a website where I find sensitometric
>>> curves of current digital cameras ?
>>>
>>> Thanks for your input !
>>> Mr.Adams
>>
>>
>>
>> The total range of the human eye is much more than
>> 2000. For example, at night one can see magnitude
>> 7+ stars, as well as bright planets, e.g. magnitude
>> ~ -3 so about 10 magnitudes, or 10,000 linear.
>> The daytime light the eye's sensitivity decreases
>> by about 600, so we have a total range of on the
>> order of 600*10,000 or ~~6 million.
>>
>> The dynamic range of a CCD sensor is limited to by photon
>> statistics by how large the well is that can hold the
>> electrons (one electron per photon), and read noise.
>> Sensors like the Canon 10D hold ~44000 electrons,
>> the 1D Mark II 52,000, and then small P&S sensors
>> much less. With read noise on the order of 10 to 15
>> electrons, divide the full well capacity by the read
>> noise for maximum dynamic range. So the Canon 1D Mark II
>> has a theoretical dynamic range of ~6500, or 12.7 stops,
>> being limited by the 12-bit A to D converter with its
>> 8.2 micron pixels. The Canon P&S S60 sensor has
>> 2.8 micron pixels, a full well of 22000, and a theoretical
>> dynamic range of 10.7 stops. P&S cameras with smaller
>> sensors have degraded performance relative to the above.
>>
>> See:
>>
>> The Signal-to-Noise of Digital Camera images
>> and Comparison to Film:
>> http://clarkvision.com/imagedetail [...] l.to.noise
>> for more details.
>>
>> Dynamic Range and Transfer Functions of Digital Images
>> and Comparison to Film:
>> http://clarkvision.com/imagedetail/dynamicrange2
>>
>> Roger Clark
>> Photography at: http://clarkvision.com
>>
> Is there some source that I can go to to find the capacity of
> pixel wells for various camera sensors?
> Is this something that is calculated fron the area of a pixel well?
> Does the capacity of a well vary from sensor to sensor even at the
> same area/pixel?
> Thanks for any leads.....Bob Williams
>
The full well capacity of sensors is hard to come by.
Some is published in the engineering and scientific
literature, which you can't get online unless you pay
a subscription. I've listed the ones I've found on my
web page above, as well as the ones I have derived.
It is not hard to derive if one can get raw data
out of the camera. There are links to methods on my
digital.signal.to.noise web page. If you find other
sources, please let me know. I have found some specs on
sensors, but then one must find what sensor is in what
camera. Apparently, there are far fewer sensor models than
there are cameras models.

One can get an idea of the full well capacity but the pixel
area or pixel spacing. But is is not a direct relationship,
as different designs are more efficient in one way or
another. See Table 3 at:
http://clarkvision.com/imagedetail [...] l.to.noise

I sure would like to know the full well capacity of the
really small pixels like in the canon S70 (2.3 micron
spacing). I am willing to do a few more analyses if people
have a camera that has raw output and can produce a
16-bit linear tif, have a macbeth color chart and want to
spend some time imaging, making a CD and sending it to me.
I'm interested in getting a range of cameras, and not
any model specifically. I you like to see data for really
small pixels, consumer DSLRs, like the 20D, and the top
end cameras like the 1Ds Mark II, or others with really large
pixels.

Roger