Sign in with
Sign up | Sign in
Your question

What ISO does the human eye have ?

Last response: in Digital Camera
Share
Anonymous
January 5, 2005 8:21:37 AM

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

Just wondering what the equivalent sensitivity of the human eye is. I'd
guess it's above ISO 400.
--

Alfred Molon
------------------------------
http://groups.yahoo.com/group/Olympus_405080/
Olympus 5060 resource - http://myolympus.org/5060/
Olympus 8080 resource - http://myolympus.org/8080/

More about : iso human eye

Anonymous
January 5, 2005 8:21:38 AM

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

The eye is extremely sensitive (a couple of photons) but is rather poor in
integrating the signal (just around 100 millisecond or less). ISO or ASA
definitions are hence not appropriate for comparison.

Have a look at http://medfmt.8k.com/mf/eye.html for a detailed discussion.

Gregor


"Alfred Molon" <alfred_molonREMOVE@yahoo.com> wrote in message
news:MPG.1c4589bb7ff0451a98a915@news.supernews.com...
> Just wondering what the equivalent sensitivity of the human eye is. I'd
> guess it's above ISO 400.
> --
>
> Alfred Molon
> ------------------------------
> http://groups.yahoo.com/group/Olympus_405080/
> Olympus 5060 resource - http://myolympus.org/5060/
> Olympus 8080 resource - http://myolympus.org/8080/
Anonymous
January 5, 2005 8:21:38 AM

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

The human eye response is very sensitive, but not very fast. In low light it
is very slow. The focal length of the human eye is about 14 to 18 mm. In
the terms of a 35 mm camera it would be about equivalent to a 15 to 17 mm
lens. This is by coincidence!

The field of vision is approximately 160 to about 170 degrees on the
average. The equivalent F stop is about F 2,5 up to about F 11 on a
comparison terms. Each individual is a bit different.

The retina sensitivity also changes as required. This can change to as much
as about 7 to 10 stops. The total range of stops equivalent can be about 18
to 20 stops. Because the retina sensitivity is changing, this means that the
F stop equivalent is changing.

As for the low light condition ASA speed equivalent, this is hard to say.
The longer we are in a dark area, the higher the sensitivity goes up, to a
certain point. From the figures I have read, I would guess it to be
averaging the equivalent to about 1200 ASA or so. This is a sort of guess,
from my personal perception when comparing from what I have seen camera
equipment do. In very bright light it may go down to about 25 ASA
equivalent.

The speed of the human eye response in bright light can be up to about 5 ms
or about 1/200 of a second. This can also vary from one person to the next.

The surface of the retina is spherical. Thus the eye is a spherical lens.
This helps it be in better compensation for the curvature of its lens. The
brain does all the corrections for things to look as perfect as possible.
Because of this, we don't have the visual distortion problems that camera
lenses have.

Are you planning to build one??? Let me know...

--

Jerry G.
=====

"Alfred Molon" <alfred_molonREMOVE@yahoo.com> wrote in message
news:MPG.1c4589bb7ff0451a98a915@news.supernews.com...
Just wondering what the equivalent sensitivity of the human eye is. I'd
guess it's above ISO 400.
--

Alfred Molon
------------------------------
http://groups.yahoo.com/group/Olympus_405080/
Olympus 5060 resource - http://myolympus.org/5060/
Olympus 8080 resource - http://myolympus.org/8080/
Related resources
Anonymous
January 5, 2005 8:21:39 AM

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

GTO wrote:
> The eye is extremely sensitive (a couple of photons) but is rather poor in
> integrating the signal (just around 100 millisecond or less). ISO or ASA
> definitions are hence not appropriate for comparison.

Don't confuse refresh with integration. At low light levels,
the human eye integrates up to about 15 seconds (Blackwell,
J. Opt. Society America, v 36, p624-643, 1946). The ISO
changes with light level by increasing rhodopsin in the retina.
This process takes a half hour our so to complete, and that
assumes you haven't been exposed to bright sunlight during the
day. Assuming you wear sunglasses and dark adapt well,
You can see pretty faint stars away from a city. Based on that
a reasonable estimate of the dark adapted eye can be done.
In a test exposure I did with a Canon 10D and 5-inch aperture
lens, the DSLR can record magnitude 14 stars in 12 seconds
at ISO 400. You can see magnitude 14 stars in a few seconds.
(Clark, R.N., Visual Astronomy of the Deep Sky, Cambridge U.
Press and Sky Publishing, 355 pages, Cambridge, 1990.)

So I would estimate the dark adapted eye to be about ISO 800.

Note that at ISO 800 on a 10D, the gain is 2.7 electrons/pixel
(reference:
http://clarkvision.com/imagedetail/digital.signal.to.no... )
which would be similar to the eye being able to see a couple of
photons for a detection.

During the day, the eye is much less sensitive, over 600 times
less (Middleton, Vision Through the Atmosphere, U. Toronto Press,
Toronto, 1958), which would put the ISO equivalent at about 1.

Roger
Photos, digital info at: http://clarkvision.com


> Have a look at http://medfmt.8k.com/mf/eye.html for a detailed discussion.
>
> Gregor
>
>
> "Alfred Molon" <alfred_molonREMOVE@yahoo.com> wrote in message
> news:MPG.1c4589bb7ff0451a98a915@news.supernews.com...
>
>>Just wondering what the equivalent sensitivity of the human eye is. I'd
>>guess it's above ISO 400.
Anonymous
January 5, 2005 3:11:01 PM

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

Jerry G. wrote:

> The human eye response is very sensitive, but not very fast. In low light it
> is very slow. The focal length of the human eye is about 14 to 18 mm. In
> the terms of a 35 mm camera it would be about equivalent to a 15 to 17 mm
> lens. This is by coincidence!

Here do you get these numbers for focal length? I did a google
search and found many "answers" ranging from 17mm to 50mm
(50 is totally absurd). If you look at this "standard" model
of the eye, it seems that the "standard" focal length is about 20mm
http://hyperphysics.phy-astr.gsu.edu/hbase/vision/eyesc...
The cornea, a meniscus lens, usually has its principle plane in front
of the lens. This implies a longer focal length than the measured
physical size. The lens inside the eye will move the plane
inside the eye, but not by much.

> The field of vision is approximately 160 to about 170 degrees on the
> average. The equivalent F stop is about F 2,5 up to about F 11 on a
> comparison terms. Each individual is a bit different.

The f/stop maximum in the astronomical community is spec'd at
f/3.5 for a dark adapted human eye. With a maximum aperture of 7mm,
this implies about a 25mm focal length. Astronomical telescope
minimum magnification is commonly cited as an f/3.5 light cone,
meaning if you look through a faster system, the eye's f/3.5
optics can't gather all the light.

The information on the web is VERY confusing, and I haven't seen
what I would consider a good reference. If anyone knows of one,
please let me know.

> The retina sensitivity also changes as required. This can change to as much
> as about 7 to 10 stops. The total range of stops equivalent can be about 18
> to 20 stops. Because the retina sensitivity is changing, this means that the
> F stop equivalent is changing.

Retinal sensitivity has nothing to do with f/stop. Like I posted
elsewhere in this thread, the dark adaptation curve shows about
a factor of 60, or 6 stops for the retina. In bright sun, it probably
goes down more. But 18 to 20 stops is not supported
by any data I've seen. Middleton (1958) put the sensitivity
from dark adapted to sun at a factor of 600, or 9.2 stops

> As for the low light condition ASA speed equivalent, this is hard to say.
> The longer we are in a dark area, the higher the sensitivity goes up, to a
> certain point. From the figures I have read, I would guess it to be
> averaging the equivalent to about 1200 ASA or so. This is a sort of guess,
> from my personal perception when comparing from what I have seen camera
> equipment do. In very bright light it may go down to about 25 ASA
> equivalent.

I put it at roughly 800, close to your 1200. Assuming ISO 800 for
dark adapted, then 800/600 (the 600 factor from above) = ISO 1.3.

Another question is what is the minimum aperture of the iris? I've
seen it around a mm, but haven't done any measurements. At 1mm, and
a 20mm focal length, you get f/20. Is that the minimum?

Roger
>
> The speed of the human eye response in bright light can be up to about 5 ms
> or about 1/200 of a second. This can also vary from one person to the next.
>
> The surface of the retina is spherical. Thus the eye is a spherical lens.
> This helps it be in better compensation for the curvature of its lens. The
> brain does all the corrections for things to look as perfect as possible.
> Because of this, we don't have the visual distortion problems that camera
> lenses have.
>
> Are you planning to build one??? Let me know...
>
Anonymous
January 5, 2005 10:45:16 PM

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

On Wed, 05 Jan 2005 12:11:01 -0700, "Roger N. Clark (change username
to rnclark)" <username@qwest.net> wrote:

>Jerry G. wrote:
>> The retina sensitivity also changes as required. This can change to as much
>> as about 7 to 10 stops. The total range of stops equivalent can be about 18
>> to 20 stops. Because the retina sensitivity is changing, this means that the
>> F stop equivalent is changing.
>
>Retinal sensitivity has nothing to do with f/stop. Like I posted
>elsewhere in this thread, the dark adaptation curve shows about
>a factor of 60, or 6 stops for the retina. In bright sun, it probably
>goes down more. But 18 to 20 stops is not supported
>by any data I've seen.

The 18 to 20 stops is a bit of a red-herring. To get this figure he
allows the iris to change size over time. If I look into the dark
cave, my iris enlarges, but as I stare at the well lit cave opening,
it contracts. It's similar to letting the camera change it's aperture
to take a sequence of photos that overall, from the first frame to the
last, would also have an 18 to 20 stop range. This is fairly
meaningless.

> Middleton (1958) put the sensitivity
>from dark adapted to sun at a factor of 600, or 9.2 stops

Haven't read that stuff, do you have a url?

--
Owamanga!
Anonymous
January 6, 2005 8:24:43 AM

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

In article <41DC3BC5.40908@qwest.net>, Roger N. Clark (change username
to rnclark) says...
> Jerry G. wrote:
>
> > The human eye response is very sensitive, but not very fast. In low light it
> > is very slow. The focal length of the human eye is about 14 to 18 mm. In
> > the terms of a 35 mm camera it would be about equivalent to a 15 to 17 mm
> > lens. This is by coincidence!
>
> Here do you get these numbers for focal length? I did a google
> search and found many "answers" ranging from 17mm to 50mm
> (50 is totally absurd).

50mm in 35mm film equivalence, which sounds plausible. The field of view
might be huge but what you usually focus on is more restricted (in terms
of field of view).
--

Alfred Molon
------------------------------
http://groups.yahoo.com/group/Olympus_405080/
Olympus 5060 resource - http://myolympus.org/5060/
Olympus 8080 resource - http://myolympus.org/8080/
Anonymous
January 6, 2005 10:13:16 PM

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

Roger N. Clark (change username to rnclark) wrote:

> Jerry G. wrote:
>
>> The human eye response is very sensitive, but not very fast. In low
>> light it
>> is very slow. The focal length of the human eye is about 14 to 18 mm. In
>> the terms of a 35 mm camera it would be about equivalent to a 15 to 17 mm
>> lens. This is by coincidence!
>
>
> Here do you get these numbers for focal length? I did a google
> search and found many "answers" ranging from 17mm to 50mm
> (50 is totally absurd). If you look at this "standard" model
> of the eye, it seems that the "standard" focal length is about 20mm
> http://hyperphysics.phy-astr.gsu.edu/hbase/vision/eyesc...
> The cornea, a meniscus lens, usually has its principle plane in front
> of the lens. This implies a longer focal length than the measured
> physical size. The lens inside the eye will move the plane
> inside the eye, but not by much.

I believe I found the answer to focal length of the eye.
Reference: Light, Color and Vision, Hunt et al., Chapman and Hall, Ltd,
London, 1968, page 49 for "standard European adult":

Object focal length of the eye = 16.7 mm
Image focal length of the eye = 22.3 mm

The object focal length is for rays coming OUT OF THE EYE.
But for an image on the retina, the image focal length is what
one wants. E.g. see:
http://galileo.phys.virginia.edu/classes/531.cas8m.fall...

So this explains the commonly cited ~17mm focal length,
but the correct value is ~22 mm focal length

This then makes more sense for the f/ratio: with an aperture
of 7 mm, the f/ratio = 22.3/7 = 3.2.

Of course these values vary, with cited values from 22 to 24 mm,
same with the aperture. The maximum aperture also decreases
with age.

Roger
>
>> The field of vision is approximately 160 to about 170 degrees on the
>> average. The equivalent F stop is about F 2,5 up to about F 11 on a
>> comparison terms. Each individual is a bit different.
>
>
> The f/stop maximum in the astronomical community is spec'd at
> f/3.5 for a dark adapted human eye. With a maximum aperture of 7mm,
> this implies about a 25mm focal length. Astronomical telescope
> minimum magnification is commonly cited as an f/3.5 light cone,
> meaning if you look through a faster system, the eye's f/3.5
> optics can't gather all the light.
>
> The information on the web is VERY confusing, and I haven't seen
> what I would consider a good reference. If anyone knows of one,
> please let me know.
>
>> The retina sensitivity also changes as required. This can change to as
>> much
>> as about 7 to 10 stops. The total range of stops equivalent can be
>> about 18
>> to 20 stops. Because the retina sensitivity is changing, this means
>> that the
>> F stop equivalent is changing.
>
>
> Retinal sensitivity has nothing to do with f/stop. Like I posted
> elsewhere in this thread, the dark adaptation curve shows about
> a factor of 60, or 6 stops for the retina. In bright sun, it probably
> goes down more. But 18 to 20 stops is not supported
> by any data I've seen. Middleton (1958) put the sensitivity
> from dark adapted to sun at a factor of 600, or 9.2 stops
>
>> As for the low light condition ASA speed equivalent, this is hard to say.
>> The longer we are in a dark area, the higher the sensitivity goes up,
>> to a
>> certain point. From the figures I have read, I would guess it to be
>> averaging the equivalent to about 1200 ASA or so. This is a sort of
>> guess,
>> from my personal perception when comparing from what I have seen camera
>> equipment do. In very bright light it may go down to about 25 ASA
>> equivalent.
>
>
> I put it at roughly 800, close to your 1200. Assuming ISO 800 for
> dark adapted, then 800/600 (the 600 factor from above) = ISO 1.3.
>
> Another question is what is the minimum aperture of the iris? I've
> seen it around a mm, but haven't done any measurements. At 1mm, and
> a 20mm focal length, you get f/20. Is that the minimum?
>
> Roger
>
>>
>> The speed of the human eye response in bright light can be up to
>> about 5 ms
>> or about 1/200 of a second. This can also vary from one person to the
>> next.
>>
>> The surface of the retina is spherical. Thus the eye is a spherical lens.
>> This helps it be in better compensation for the curvature of its lens.
>> The
>> brain does all the corrections for things to look as perfect as possible.
>> Because of this, we don't have the visual distortion problems that camera
>> lenses have.
>>
>> Are you planning to build one??? Let me know...
>>
>
Anonymous
January 7, 2005 5:39:11 PM

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

"Roger N. Clark (change username to rnclark)" <username@qwest.net>
wrote in message news:41DDF03C.8020900@qwest.net...
SNIP
> I believe I found the answer to focal length of the eye.
> Reference: Light, Color and Vision, Hunt et al., Chapman and Hall,
> Ltd,
> London, 1968, page 49 for "standard European adult":
>
> Object focal length of the eye = 16.7 mm
> Image focal length of the eye = 22.3 mm
>
> The object focal length is for rays coming OUT OF THE EYE.
> But for an image on the retina, the image focal length is what
> one wants. E.g. see:
> http://galileo.phys.virginia.edu/classes/531.cas8m.fall...

Thanks for the info.

Bart
Anonymous
January 8, 2005 8:04:50 PM

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

I was told many years ago that the human eye is the equivalent of
about a 45mm lens on a 35mm camera. That is why a "normal" lens is a
50mm lens... not sure why they made it 50mm and not 45mm, I guess it
makes th math easier to multiply focal lengths. In the old days you
could choose from a 50mm, a 100mm, a 150, 200, etc. length lens

robert Strom

On Thu, 06 Jan 2005 19:13:16 -0700, "Roger N. Clark (change username
to rnclark)" <username@qwest.net> wrote:
Anonymous
January 9, 2005 3:18:57 AM

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

"Robert Strom" <notavailable@yahoo.com> wrote in message
news:bk40u0l6pulhr92osp82tt9jksgbep6e48@4ax.com...
>I was told many years ago that the human eye is the equivalent of
> about a 45mm lens on a 35mm camera.

That is because an average print at normal viewing (say 10 inch or
about 25 cm) distance has a Field-of-View resembling that of a lens of
about 45mm focal length for 35mm film.

Bart
Anonymous
January 9, 2005 12:09:49 PM

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

Ah, but so does film :-) There are a number of edge effects,
especially if you play games during development, like developing with
no agitation. There are a few adjacent pixel effects in CCDs, but not
nearly as much as film.
Anonymous
January 10, 2005 4:58:39 AM

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

Kibo informs me that Alfred Molon <alfred_molonREMOVE@yahoo.com> stated
that:

>Just wondering what the equivalent sensitivity of the human eye is. I'd
>guess it's above ISO 400.

The question isn't meaningful, because the human eye has
locally-compensating sensitivity. This is why it is that when see
'negative' ghost images when you look away from something very bright in
a much darker scene.

--
W
. | ,. w , "Some people are alive only because
\|/ \|/ it is illegal to kill them." Perna condita delenda est
---^----^---------------------------------------------------------------
Anonymous
January 16, 2005 8:36:50 PM

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

Alfred Molon <alfred_molonREMOVE@yahoo.com> wrote:

>Just wondering what the equivalent sensitivity of the human eye is. I'd
>guess it's above ISO 400.

In colour mode, or B&W?
--
Ken Tough
Anonymous
January 17, 2005 12:56:38 AM

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

Eye can adjust its sensitivity depending on light conditions, by millions of
times. Otherwise you couldn't look straight into Sun at noon and then see faint
stars at night. Also if you compare eye to the camera you'll notice that camera
can collect light over long period of time where eye can not and "longer
exposure" of eye to faint source of light doesn't increase perceived
sensitivity of retina. So if you compare apples to apples, you would need to
keep camera's exposure somewhere shorter than 1/10 of sec. and see at what ISO
the camera picks up as much detail as your eye can. I would think it would be
in thousands ISO for B&W and at least 1600 for color vision. With digital
camera it would be easy to estimate. Are you the game?
Anonymous
January 17, 2005 4:07:12 AM

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

On Sun, 16 Jan 2005 17:36:50 +0200, Ken Tough <ken@objectech.co.uk>
wrote:

>Alfred Molon <alfred_molonREMOVE@yahoo.com> wrote:
>
>>Just wondering what the equivalent sensitivity of the human eye is. I'd
>>guess it's above ISO 400.

No kidding! I'd guess thousands of times more! You're comparing
apples and oranges. The human eye is infinitely better than any
camera/film ever made. It focuses faster, has huge latitude,
excellent sensitivity, etc etc.
Anonymous
January 17, 2005 5:18:40 AM

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

On Mon, 17 Jan 2005 01:07:12 GMT, secheese <sec@nbnet.nb.ca> wrote:

>On Sun, 16 Jan 2005 17:36:50 +0200, Ken Tough <ken@objectech.co.uk>
>wrote:
>
>>Alfred Molon <alfred_molonREMOVE@yahoo.com> wrote:
>>
>>>Just wondering what the equivalent sensitivity of the human eye is. I'd
>>>guess it's above ISO 400.
>
>No kidding! I'd guess thousands of times more! You're comparing
>apples and oranges. The human eye is infinitely better than any
>camera/film ever made. It focuses faster, has huge latitude,
>excellent sensitivity, etc etc.

When it is young <g>


--
Scott in Florida
Anonymous
January 17, 2005 10:22:04 AM

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

In article <3SxeInASoo6BFwoX@objectech.co.uk>, Ken Tough says...
> Alfred Molon <alfred_molonREMOVE@yahoo.com> wrote:
>
> >Just wondering what the equivalent sensitivity of the human eye is. I'd
> >guess it's above ISO 400.
>
> In colour mode, or B&W?

Colour - but this is actually an old thread with lots of replies.
--

Alfred Molon
------------------------------
Olympus 4040, 5050, 5060, 7070, 8080, E300 forum at
http://groups.yahoo.com/group/MyOlympus/
Olympus 8080 resource - http://myolympus.org/8080/
Anonymous
January 18, 2005 4:08:31 PM

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

This assumes that the eye has a DQE of roughly 30%.

Joe
------------ And now a word from our sponsor ------------------
Do your users want the best web-email gateway? Don't let your
customers drift off to free webmail services install your own
web gateway!
-- See http://netwinsite.com/sponsor/sponsor_webmail.htm ----
Anonymous
January 20, 2005 10:30:07 AM

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

Joseph Miller wrote:

> This assumes that the eye has a DQE of roughly 30%.

Where do you get this number? I am pretty sure my data
and references can not be off by a factor of 10.

If you look back this thread, I wrote:

Don't confuse refresh with integration. At low light levels,
the human eye integrates up to about 15 seconds (Blackwell,
J. Opt. Society America, v 36, p624-643, 1946). The ISO
changes with light level by increasing rhodopsin in the retina.
This process takes a half hour our so to complete, and that
assumes you haven't been exposed to bright sunlight during the
day. Assuming you wear sunglasses and dark adapt well,
You can see pretty faint stars away from a city. Based on that
a reasonable estimate of the dark adapted eye can be done.
In a test exposure I did with a Canon 10D and 5-inch aperture
lens, the DSLR can record magnitude 14 stars in 12 seconds
at ISO 400. You can see magnitude 14 stars in a few seconds.
(Clark, R.N., Visual Astronomy of the Deep Sky, Cambridge U.
Press and Sky Publishing, 355 pages, Cambridge, 1990.)

So I would estimate the dark adapted eye to be about ISO 800.

Note that at ISO 800 on a 10D, the gain is 2.7 electrons/pixel
(reference:
http://clarkvision.com/imagedetail/digital.signal.to.no... )
which would be similar to the eye being able to see a couple of
photons for a detection.

During the day, the eye is much less sensitive, over 600 times
less (Middleton, Vision Through the Atmosphere, U. Toronto Press,
Toronto, 1958), which would put the ISO equivalent at about 1.

Roger
Photos, digital info at: http://clarkvision.com
Anonymous
January 20, 2005 2:08:32 PM

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

Roger N. Clark (change username to rnclark) wrote:
> Joseph Miller wrote:
>
>> This assumes that the eye has a DQE of roughly 30%.
>
>
> Where do you get this number? I am pretty sure my data
> and references can not be off by a factor of 10.
>

This number comes from working with vision scientists and surveying the
literature. I am talking abot the maximum DQE of the humand dark
adapted eye. Whether this is useful DQE is another thing. We routinely
measure CCD DQEs in our CCD development lab (we make custom CCDs for
astronomical use), and the question of the eye's DQE is something we
often discuss. There are several vision scientists assocaited with our
organization. The question is what fraction of the photons striking the
eye produce a physiological response, a recognition that a photon has
been absorbed. The answer I have most often seen is roughly 20-35% at
the wavelength of peak sensitivity. Incidentally, the CCDs we are now
using have DQEs above 90% at their peaks. Just about the only light
lost is reflection, so good anti-reflection coatings are crucial.
Anonymous
January 20, 2005 11:21:54 PM

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

Joseph Miller wrote:

> Roger N. Clark (change username to rnclark) wrote:
>
>> Joseph Miller wrote:
>>
>>> This assumes that the eye has a DQE of roughly 30%.
>>
>>
>>
>> Where do you get this number? I am pretty sure my data
>> and references can not be off by a factor of 10.
>>
>
> This number comes from working with vision scientists and surveying the
> literature. I am talking abot the maximum DQE of the humand dark
> adapted eye. Whether this is useful DQE is another thing. We routinely
> measure CCD DQEs in our CCD development lab (we make custom CCDs for
> astronomical use), and the question of the eye's DQE is something we
> often discuss. There are several vision scientists assocaited with our
> organization. The question is what fraction of the photons striking the
> eye produce a physiological response, a recognition that a photon has
> been absorbed. The answer I have most often seen is roughly 20-35% at
> the wavelength of peak sensitivity. Incidentally, the CCDs we are now
> using have DQEs above 90% at their peaks. Just about the only light
> lost is reflection, so good anti-reflection coatings are crucial.


I have never seen in the literature the DQE of the human eye.

You seem to indicate your answer is an assumption.
But there is direct data, which I indicated in my previous post.

1) the faintest star visible in a telescope, using the Blackwell
(1946) criterion, is magnitude 14.2 (Clark, 1990; 1994), . This is
for an equivalent eye integration time of 6 seconds and 50%
probability of detection by a trained observer (Blackwell, 1946).
A Canon 10D camera with a 5-inch lens and 6-second exposure
time reaches about magnitude in 14.2 in 6 seconds at ISO 800 with
similar probability of detection. If the eye was 10,000, then
the magnitude difference between the camera and 2.7 magnitudes
(a factor of 12.5) and the eye could detect magnitude 16.9
stars with a 5-inch telescope.

This directly compares with real data, and has no assumptions about
the eye, and ISO 800 fits best. Because comparisons are different
between digital and the perception in the eye plus brain, I
would put an error bar on this number of about +/- 50%,
thus if a more accurate answer was in the iso 400 to 1600 range,
it would not surprise me. The data do not indicate iso 10,000,
however.

References:

Blackwell, J. Opt. Society America, v 36, p624-643, 1946.

Clark, R.N., Visual Astronomy of the Deep Sky, Cambridge
University Press and Sky Publishing, (book of 355 pages), 1990.

Clark, R.N., How Faint Can You See?, Sky and Telescope,
106-108, April, 1994.

A reference on digital camera faintest star:
http://clarkvision.com/astro/moon_saturn_jupiter.d60.v1
Saturn's satellite Tethys, magnitude 10.3 was detected in 1 second
at ISO 200, 5-inch lens, D60 camera. At iso 800, 6 seconds
that is 24 times fainter, or 3.4 magnitudes, = 13.6 magnitude
limit. The Canon 10D is almost 2x less noise than the D60, or
about 0.6 magnitude fainter, thus the 10D has a 14.2 magnitude
limit with the 5-inch lens. These values should be accurate to
a few tenths of a magnitude.

Roger
http://clarkvision.com
Anonymous
January 21, 2005 2:30:09 AM

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

In article <41f002ed@darkstar>, Joseph Miller says...

> the wavelength of peak sensitivity. Incidentally, the CCDs we are now
> using have DQEs above 90% at their peaks. Just about the only light
> lost is reflection, so good anti-reflection coatings are crucial.

You are using back-illuminated CCDs ? Only this type has such a high
quantum efficiency.
--

Alfred Molon
------------------------------
Olympus 4040, 5050, 5060, 7070, 8080, E300 forum at
http://groups.yahoo.com/group/MyOlympus/
Olympus 8080 resource - http://myolympus.org/8080/
Anonymous
January 21, 2005 4:17:48 AM

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

"Roger N. Clark (change username to rnclark)" <username@qwest.net> wrote:
>
> So I would estimate the dark adapted eye to be about ISO 800.

Sounds about right. I find that shooting at 1/30, f/1.4, ISO 1600 captures
more than I see at night in the city.

David J. Littleboy
Tokyo, Japan
Anonymous
January 21, 2005 8:32:25 AM

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

"Roger N. Clark (change username to rnclark)" <username@qwest.net> wrote in
news:41EFC06F.1030403@qwest.net:

> Where do you get this number? I am pretty sure my data
> and references can not be off by a factor of 10.

Well, they've been pretty darn off in every other thread... maybe not by a
factor of ten, but by a factor of 3 or so? absolutely.

--
http://www.neopets.com/refer.phtml?username=moosespet
Anonymous
January 21, 2005 11:04:41 AM

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

Jon Pike wrote:
> "Roger N. Clark (change username to rnclark)" <username@qwest.net> wrote in
> news:41EFC06F.1030403@qwest.net:
>
>
>>Where do you get this number? I am pretty sure my data
>>and references can not be off by a factor of 10.
>
>
> Well, they've been pretty darn off in every other thread... maybe not by a
> factor of ten, but by a factor of 3 or so? absolutely.
>

Bullshit. John is a troll who has been attacking me in
multiple threads. He makes statements like the above
to incite controversy. He has been discredited in multiple
threads with his wild and unsubstantiated claims.
A google search will show his true colors.
Anonymous
January 21, 2005 4:04:38 PM

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

I afraid your experiments on the faintest star you can see do not
measure the DQE of the eye. A star image on the retina is not usually
confined to one rod (or cone), but spread out over many. Thus the
photon flux per retinal detector element is quite a bit smaller than the
total photon flux from the star. Whether you recognize a star is there
at the limit of vision is a complicated issue. If the brain or retina
did integrate, that would be irrelevant to DQE. Also, as an aside, the
eye is constantly jiggling around at the pixel level, so a star image is
far from fixed even at the sub-second time scale. If I get a few
moments, I will provide you some references to laboratory measurements
of the DQE of the (living) human eye. I read a recent one that claimed
above 80%, but I didn't believe that result; I can't remember where I
saw it off the top of my head.

I have often wondered why the ISO of consumer digital cameras is so low.
They can have lots of gate structures on the CCD surface, but they
use lenslet arrays to attempt to get around them. The ISO of CCD
cameras we use are roughly in the 30,000-40,000 range

Roger N. Clark (change username to rnclark) wrote:

> Joseph Miller wrote:
>
>> Roger N. Clark (change username to rnclark) wrote:
>>
>>> Joseph Miller wrote:
>>>
>>>> This assumes that the eye has a DQE of roughly 30%.
>>>
>>>
>>>
>>>
>>> Where do you get this number? I am pretty sure my data
>>> and references can not be off by a factor of 10.
>>>
>>
>> This number comes from working with vision scientists and surveying
>> the literature. I am talking abot the maximum DQE of the humand dark
>> adapted eye. Whether this is useful DQE is another thing. We
>> routinely measure CCD DQEs in our CCD development lab (we make custom
>> CCDs for astronomical use), and the question of the eye's DQE is
>> something we often discuss. There are several vision scientists
>> assocaited with our organization. The question is what fraction of
>> the photons striking the eye produce a physiological response, a
>> recognition that a photon has been absorbed. The answer I have most
>> often seen is roughly 20-35% at the wavelength of peak sensitivity.
>> Incidentally, the CCDs we are now using have DQEs above 90% at their
>> peaks. Just about the only light lost is reflection, so good
>> anti-reflection coatings are crucial.
>
>
>
> I have never seen in the literature the DQE of the human eye.
>
> You seem to indicate your answer is an assumption.
> But there is direct data, which I indicated in my previous post.
>
> 1) the faintest star visible in a telescope, using the Blackwell
> (1946) criterion, is magnitude 14.2 (Clark, 1990; 1994), . This is
> for an equivalent eye integration time of 6 seconds and 50%
> probability of detection by a trained observer (Blackwell, 1946).
> A Canon 10D camera with a 5-inch lens and 6-second exposure
> time reaches about magnitude in 14.2 in 6 seconds at ISO 800 with
> similar probability of detection. If the eye was 10,000, then
> the magnitude difference between the camera and 2.7 magnitudes
> (a factor of 12.5) and the eye could detect magnitude 16.9
> stars with a 5-inch telescope.
>
> This directly compares with real data, and has no assumptions about
> the eye, and ISO 800 fits best. Because comparisons are different
> between digital and the perception in the eye plus brain, I
> would put an error bar on this number of about +/- 50%,
> thus if a more accurate answer was in the iso 400 to 1600 range,
> it would not surprise me. The data do not indicate iso 10,000,
> however.
>
> References:
>
> Blackwell, J. Opt. Society America, v 36, p624-643, 1946.
>
> Clark, R.N., Visual Astronomy of the Deep Sky, Cambridge
> University Press and Sky Publishing, (book of 355 pages), 1990.
>
> Clark, R.N., How Faint Can You See?, Sky and Telescope,
> 106-108, April, 1994.
>
> A reference on digital camera faintest star:
> http://clarkvision.com/astro/moon_saturn_jupiter.d60.v1
> Saturn's satellite Tethys, magnitude 10.3 was detected in 1 second
> at ISO 200, 5-inch lens, D60 camera. At iso 800, 6 seconds
> that is 24 times fainter, or 3.4 magnitudes, = 13.6 magnitude
> limit. The Canon 10D is almost 2x less noise than the D60, or
> about 0.6 magnitude fainter, thus the 10D has a 14.2 magnitude
> limit with the 5-inch lens. These values should be accurate to
> a few tenths of a magnitude.
>
> Roger
> http://clarkvision.com
>
Anonymous
January 22, 2005 1:09:09 AM

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

Joseph Miller wrote:

> I afraid your experiments on the faintest star you can see do not
> measure the DQE of the eye.

I never said it did. Nor is knowing DQE relevant to the question of
what is the ISO.

> A star image on the retina is not usually
> confined to one rod (or cone), but spread out over many. Thus the
> photon flux per retinal detector element is quite a bit smaller than the
> total photon flux from the star. Whether you recognize a star is there
> at the limit of vision is a complicated issue. If the brain or retina
> did integrate, that would be irrelevant to DQE. Also, as an aside, the
> eye is constantly jiggling around at the pixel level, so a star image is
> far from fixed even at the sub-second time scale. If I get a few
> moments, I will provide you some references to laboratory measurements
> of the DQE of the (living) human eye. I read a recent one that claimed
> above 80%, but I didn't believe that result; I can't remember where I
> saw it off the top of my head.
>
> I have often wondered why the ISO of consumer digital cameras is so low.
> They can have lots of gate structures on the CCD surface, but they
> use lenslet arrays to attempt to get around them. The ISO of CCD
> cameras we use are roughly in the 30,000-40,000 range

The "high" iso claimed for CCDs is bogus. It is all in what different
people define as an image. If you use the same definition,
you get similar results. In recent comparisons by astrophotographers
on the yahoo groups digital_astro list, people have directly
compared what can be obtained with CCDs doing astrophotography,
and with DSLRs. The difference was generally a only
factor of 2 to 4. But the CCDs were used with no filters,
and the DSLRs of course have the color filters on the sensors.
When compensated for the filter bandpass and transmission,
the sensors are essentially equal, as one would expect.
Thus the definitions of what is an image is different
in order to come up with such claims.

If you are using similar CCD different definitions, this might be
why you get the higher ISO for the human eye than when
one compares digital cameras.

Roger

>
> Roger N. Clark (change username to rnclark) wrote:
>
>> Joseph Miller wrote:
>>
>>> Roger N. Clark (change username to rnclark) wrote:
>>>
>>>> Joseph Miller wrote:
>>>>
>>>>> This assumes that the eye has a DQE of roughly 30%.
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> Where do you get this number? I am pretty sure my data
>>>> and references can not be off by a factor of 10.
>>>>
>>>
>>> This number comes from working with vision scientists and surveying
>>> the literature. I am talking abot the maximum DQE of the humand dark
>>> adapted eye. Whether this is useful DQE is another thing. We
>>> routinely measure CCD DQEs in our CCD development lab (we make custom
>>> CCDs for astronomical use), and the question of the eye's DQE is
>>> something we often discuss. There are several vision scientists
>>> assocaited with our organization. The question is what fraction of
>>> the photons striking the eye produce a physiological response, a
>>> recognition that a photon has been absorbed. The answer I have most
>>> often seen is roughly 20-35% at the wavelength of peak sensitivity.
>>> Incidentally, the CCDs we are now using have DQEs above 90% at their
>>> peaks. Just about the only light lost is reflection, so good
>>> anti-reflection coatings are crucial.
>>
>>
>>
>>
>> I have never seen in the literature the DQE of the human eye.
>>
>> You seem to indicate your answer is an assumption.
>> But there is direct data, which I indicated in my previous post.
>>
>> 1) the faintest star visible in a telescope, using the Blackwell
>> (1946) criterion, is magnitude 14.2 (Clark, 1990; 1994), . This is
>> for an equivalent eye integration time of 6 seconds and 50%
>> probability of detection by a trained observer (Blackwell, 1946).
>> A Canon 10D camera with a 5-inch lens and 6-second exposure
>> time reaches about magnitude in 14.2 in 6 seconds at ISO 800 with
>> similar probability of detection. If the eye was 10,000, then
>> the magnitude difference between the camera and 2.7 magnitudes
>> (a factor of 12.5) and the eye could detect magnitude 16.9
>> stars with a 5-inch telescope.
>>
>> This directly compares with real data, and has no assumptions about
>> the eye, and ISO 800 fits best. Because comparisons are different
>> between digital and the perception in the eye plus brain, I
>> would put an error bar on this number of about +/- 50%,
>> thus if a more accurate answer was in the iso 400 to 1600 range,
>> it would not surprise me. The data do not indicate iso 10,000,
>> however.
>>
>> References:
>>
>> Blackwell, J. Opt. Society America, v 36, p624-643, 1946.
>>
>> Clark, R.N., Visual Astronomy of the Deep Sky, Cambridge
>> University Press and Sky Publishing, (book of 355 pages), 1990.
>>
>> Clark, R.N., How Faint Can You See?, Sky and Telescope,
>> 106-108, April, 1994.
>>
>> A reference on digital camera faintest star:
>> http://clarkvision.com/astro/moon_saturn_jupiter.d60.v1
>> Saturn's satellite Tethys, magnitude 10.3 was detected in 1 second
>> at ISO 200, 5-inch lens, D60 camera. At iso 800, 6 seconds
>> that is 24 times fainter, or 3.4 magnitudes, = 13.6 magnitude
>> limit. The Canon 10D is almost 2x less noise than the D60, or
>> about 0.6 magnitude fainter, thus the 10D has a 14.2 magnitude
>> limit with the 5-inch lens. These values should be accurate to
>> a few tenths of a magnitude.
>>
>> Roger
>> http://clarkvision.com
>>
Anonymous
January 22, 2005 3:27:01 AM

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

In article <41f16fa2$1@darkstar>, Joseph Miller says...

> I have often wondered why the ISO of consumer digital cameras is so low.
> They can have lots of gate structures on the CCD surface, but they
> use lenslet arrays to attempt to get around them.

The spectral efficiency of (most) standard CCDs peaks around 20-30%,
while back-illuminated CCDs have QE up to 100%.

> The ISO of CCD
> cameras we use are roughly in the 30,000-40,000 range

I wonder what CCDs you use and what noise levels you have. Anyway, by
amplifying the signal before A/D conversion you can get any ISO you want
(but not any signal-to-noise ratio you want).
--

Alfred Molon
------------------------------
Olympus 4040, 5050, 5060, 7070, 8080, E300 forum at
http://groups.yahoo.com/group/MyOlympus/
Olympus 8080 resource - http://myolympus.org/8080/
Anonymous
January 22, 2005 9:42:46 AM

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

"Roger N. Clark (change username to rnclark)" <username@qwest.net> wrote
in news:41F11A09.5060809@qwest.net:

> Jon Pike wrote:
>> "Roger N. Clark (change username to rnclark)" <username@qwest.net>
>> wrote in news:41EFC06F.1030403@qwest.net:
>>
>>
>>>Where do you get this number? I am pretty sure my data
>>>and references can not be off by a factor of 10.
>>
>>
>> Well, they've been pretty darn off in every other thread... maybe not
>> by a factor of ten, but by a factor of 3 or so? absolutely.
>>
>
> Bullshit. John is a troll who has been attacking me in
> multiple threads. He makes statements like the above
> to incite controversy. He has been discredited in multiple
> threads with his wild and unsubstantiated claims.
> A google search will show his true colors.

wild and unsubstantiated eh?
that's why you had to change your "test" picture to get rid of jpeg
compression?
that's why now, instead of before, you have suddenly put 'references' to
other sites on your site?
because they're so wild and unsubstantiated?

--
http://www.neopets.com/refer.phtml?username=moosespet
Anonymous
January 22, 2005 1:57:30 PM

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

In article <41F1DFF5.3020701@qwest.net>, Roger N. Clark (change username
to rnclark) says...

> The "high" iso claimed for CCDs is bogus. It is all in what different
> people define as an image. If you use the same definition,
> you get similar results. In recent comparisons by astrophotographers
> on the yahoo groups digital_astro list, people have directly
> compared what can be obtained with CCDs doing astrophotography,
> and with DSLRs. The difference was generally a only
> factor of 2 to 4. But the CCDs were used with no filters,
> and the DSLRs of course have the color filters on the sensors.
> When compensated for the filter bandpass and transmission,
> the sensors are essentially equal, as one would expect.

In astrophotography back-illuminated CCDs are used, which have QE three
to four times higher than standard front-illuminated CCDs used in
digital cameras. In other words these CCDs are three to four times more
sensitive.

Another factor should be that CCDs in astrophotography are cooled down
to very low temperatures (something you can hardly do in a DSLR), which
further reduces the noise - and therefore increases the effective ISO.

But of course it is possible that not all astronomers use this top
equipment.
--

Alfred Molon
------------------------------
Olympus 4040, 5050, 5060, 7070, 8080, E300 forum at
http://groups.yahoo.com/group/MyOlympus/
Olympus 8080 resource - http://myolympus.org/8080/
Anonymous
January 22, 2005 1:57:31 PM

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

Alfred Molon wrote:

> In article <41F1DFF5.3020701@qwest.net>, Roger N. Clark (change username
> to rnclark) says...
>
>
>>The "high" iso claimed for CCDs is bogus. It is all in what different
>>people define as an image. If you use the same definition,
>>you get similar results. In recent comparisons by astrophotographers
>>on the yahoo groups digital_astro list, people have directly
>>compared what can be obtained with CCDs doing astrophotography,
>>and with DSLRs. The difference was generally a only
>>factor of 2 to 4. But the CCDs were used with no filters,
>>and the DSLRs of course have the color filters on the sensors.
>>When compensated for the filter bandpass and transmission,
>>the sensors are essentially equal, as one would expect.
>
>
> In astrophotography back-illuminated CCDs are used, which have QE three
> to four times higher than standard front-illuminated CCDs used in
> digital cameras. In other words these CCDs are three to four times more
> sensitive.

That does not explain the factor of ten to twenty of your
ISO 20,000 to 30,000 over digital camera CCDs.
Here is one of the top CCD being used by astronomers,
the STL-11000M, which uses the KAI-11000M/CM chip
http://panther-observatory.com/Review_STL11000M.htm
The page compares the cooled KAI-11000M/CM to a Canon 10D camera.
The creator of the web page, Johannes Schedler, concluded
"I compared both cameras in respect of S/N ratio and came to a factor
of at least 4 for the luminance part.
The Canon CMOS chip itself is very good and the reduced sensitivity
is caused essentially by the Bayer color filter matrix and the bad IR
filter characteristics (lost Ha)."

The IR filter refers to the IR cutoff which cuts off hydrogen
alpha light from nebulae.

The KAI-11000M/CM Kodak chip is an 11 megapixel 35mm size
chip.
http://www.sbig.com/sbwhtmls/large_format_cameras.htm#s...


Here are some comparison specs:

Sensor pixel size Full well Read Noise Gain
(microns) (electrons) (electrons) (e/adu)
KAI-11000M/CM 9 50,000 13 0.8
Canon 10D* 7.4 44,000 9-15 0.7
Canon 1DMarkII* 8.2 52,000 15 0.8

* at ISO 1600

The reported read noise on the 10D seems variable. I'll look
at my own data (haven't done that yet). But in either case,
you see that the gain in electrons per adu (that is the digital
number you get from the A to D converter are essentially the
same. Read noise is also similar. There is no way, with that
gain and read noise, that the cooled CCD can have anything but
an iso comparable to the digital cameras. So again,
iso 20,000 to 30,000 is incorrect.

I could not find if the KAI-11000M/CM chip is back side
illuminated. But even so, we would be talking 3 to 4x,
not 10 to 20x higher iso.

>
> Another factor should be that CCDs in astrophotography are cooled down
> to very low temperatures (something you can hardly do in a DSLR), which
> further reduces the noise - and therefore increases the effective ISO.

Cooling and lower dark current noise does not increase ISO.
If so, then we would see even wider differences between
DSLRs and P&S cameras. It all comes down to how one defines
ISO. Apparently the camera manufacturers have one specification,
and the CCD people are touting a different one.
I believe the camera manufacturers are specifying a certain
well depth for a given exposure, similar to density on film,
and the high claim iso people are basing it on signal-to-noise
relative to film.

Roger

>
> But of course it is possible that not all astronomers use this top
> equipment.
Anonymous
January 22, 2005 4:23:26 PM

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

"Jon Pike" <Anonomoose@spamlesshotmail.com> wrote in message
news:Xns95E5F13762BE4LessThanPerfectInc@24.71.223.159...
> "Roger N. Clark (change username to rnclark)" <username@qwest.net>
> wrote
> in news:41F11A09.5060809@qwest.net:
SNIP
>> Bullshit. John is a troll who has been attacking me in
>> multiple threads. He makes statements like the above
>> to incite controversy. He has been discredited in multiple
>> threads with his wild and unsubstantiated claims.
>> A google search will show his true colors.
>
> wild and unsubstantiated eh?
> that's why you had to change your "test" picture to get rid of jpeg
> compression?

No idea what you're talking about, and frankly not interested either.

> that's why now, instead of before, you have suddenly put
> 'references' to
> other sites on your site?

Roger has always had references to other webpages on his site.

> because they're so wild and unsubstantiated?

Apparently your claims are...

Bart
Anonymous
January 22, 2005 8:36:10 PM

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

In message <41F276C8.3090707@qwest.net>,
"Roger N. Clark (change username to rnclark)" <username@qwest.net>
wrote:

>Here are some comparison specs:
>
> Sensor pixel size Full well Read Noise Gain
> (microns) (electrons) (electrons) (e/adu)
>KAI-11000M/CM 9 50,000 13 0.8
>Canon 10D* 7.4 44,000 9-15 0.7
>Canon 1DMarkII* 8.2 52,000 15 0.8
>
>* at ISO 1600
>
>The reported read noise on the 10D seems variable.

10D RAW values at ISO 1600 are pretty useless. 1600 on the 10D is ISO
800, with a stop of under-exposure, and a doubling of the RAW numbers 0
through 2047, and 4095 for anything greater than 2047. Then, half of
the even values are offset by 1 in very simple patterns in the image.
The following is an image where the even numbers in an ISO 1600 10D
image are black and the odd numbers are white (you will never see this
in a linear tiff that scales the data and/or subtracts the blackpoint):

http://www.pbase.com/jps_photo/image/38841732/original

The speckles of exception to the pattern are mapped-out pixels,
substituted with interpolation.

On the 10D, ISO "1600" is really 800, distorted by the fabricated odd
numbers. "3200" (or "H") is really 1600, distorted by the fabricated
odd numbers (created to fool histograms?).

On the 20D, 1600 is 1600, and 3200 is 1600 (with no odd-number
nonsense).
--

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS@no.komm>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
Anonymous
January 22, 2005 9:35:41 PM

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

In article <41F276C8.3090707@qwest.net>, Roger N. Clark (change username
to rnclark) says...

> That does not explain the factor of ten to twenty of your
> ISO 20,000 to 30,000 over digital camera CCDs.

Actually I wasn't the one who mentioned an ISO 30000.

> Here is one of the top CCD being used by astronomers,
> the STL-11000M, which uses the KAI-11000M/CM chip
> http://panther-observatory.com/Review_STL11000M.htm

For that CCD the QE peaks at 50% while with back-illuminated CCDs the QE
peaks at 100%. I'd guess the KAI-11000M/CM is used by amateurs while in
the big telescopes of the research centres they use back-illuminated
CCDs (but as I said, I'm just guessing here).

<snip>

> Cooling and lower dark current noise does not increase ISO.
> If so, then we would see even wider differences between
> DSLRs and P&S cameras. It all comes down to how one defines
> ISO. Apparently the camera manufacturers have one specification,
> and the CCD people are touting a different one.
> I believe the camera manufacturers are specifying a certain
> well depth for a given exposure, similar to density on film,
> and the high claim iso people are basing it on signal-to-noise
> relative to film.

If noise is lower (due to cooling) you can amplify more the signal
before A/D conversion, thereby obtaining a higher ISO. I have no idea
however how much noise (total sensor noise) is lower at very low
temperatures and therefore how much gain you would obtain by cooling the
sensor.

To summarise it, you have to multiply the QE gain due to back-
illumination (3x to 4x) with the gain obtained by removing the colour
matrix (possibly also the infrared filter) and with the gain obtained by
operating the CCD at very low temperatures.
--

Alfred Molon
------------------------------
Olympus 4040, 5050, 5060, 7070, 8080, E300 forum at
http://groups.yahoo.com/group/MyOlympus/
Olympus 8080 resource - http://myolympus.org/8080/
Anonymous
January 22, 2005 9:35:42 PM

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

Alfred Molon wrote:

> In article <41F276C8.3090707@qwest.net>, Roger N. Clark (change username
> to rnclark) says...
>
>
>>That does not explain the factor of ten to twenty of your
>>ISO 20,000 to 30,000 over digital camera CCDs.
>
>
> Actually I wasn't the one who mentioned an ISO 30000.

Sorry. It was Joseph Miller, and he claims up to 40,000.

>>Cooling and lower dark current noise does not increase ISO.
>>If so, then we would see even wider differences between
>>DSLRs and P&S cameras. It all comes down to how one defines
>>ISO. Apparently the camera manufacturers have one specification,
>>and the CCD people are touting a different one.
>>I believe the camera manufacturers are specifying a certain
>>well depth for a given exposure, similar to density on film,
>>and the high claim iso people are basing it on signal-to-noise
>>relative to film.
>
>
> If noise is lower (due to cooling) you can amplify more the signal
> before A/D conversion, thereby obtaining a higher ISO. I have no idea
> however how much noise (total sensor noise) is lower at very low
> temperatures and therefore how much gain you would obtain by cooling the
> sensor.

Well, there is one small ugly fact of physics that limits
this gain idea: photon counting statistics. With a gain around
1 electron per adu, you are getting everything there is.
Modern DSLRs and P&S cameras are photon noise limited.
Improving the gain with lower read noise will not really
help much.

For example, the Canon 10D, with a full
well capacity of ~44,000, gives a signal-to-noise on an
18% gray card of 89, and photon noise of 89 photons.
The read noise of ~15 is already much less than the photon
noise, that is why the system is photon noise limited.

In the shadows, say 0.5% of full well, or 220 photons,
the photon noise is 15, about equal to the read noise.
So reducing the read noise to zero, only increases the signal=
to-noise from 11 to 15. This is a help, but in your
typical well exposed image would barely be noticeable,
and then only in the darkest parts of the image.

>
> To summarise it, you have to multiply the QE gain due to back-
> illumination (3x to 4x) with the gain obtained by removing the colour
> matrix (possibly also the infrared filter) and with the gain obtained by
> operating the CCD at very low temperatures.

Again, the gain of operating the CCD at low temperatures
is a negligible effect on ordinary photography with
normally exposed images. The low temperatures serves
to reduce noise from dark current, but even here a
new strategy has been developed to circumvent this problem
to a large degree: for long low light exposures with
a ambient temperature electronic sensor, simply take
multiple shorter exposures and add them together.
Astronomers are taking many hour long exposures with
DSLRs a few minutes at a time, and the resulting images
are gorgeous.

Roger
Anonymous
January 23, 2005 1:04:27 PM

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

In article <41F2C56D.2070001@qwest.net>, Roger N. Clark (change username
to rnclark) says...

> Well, there is one small ugly fact of physics that limits
> this gain idea: photon counting statistics. With a gain around
> 1 electron per adu, you are getting everything there is.
> Modern DSLRs and P&S cameras are photon noise limited.
> Improving the gain with lower read noise will not really
> help much.

I used to think thermal noise and read noise were dominant. But maybe
you are right, who knows.

> Again, the gain of operating the CCD at low temperatures
> is a negligible effect on ordinary photography with
> normally exposed images. The low temperatures serves
> to reduce noise from dark current, but even here a
> new strategy has been developed to circumvent this problem
> to a large degree: for long low light exposures with
> a ambient temperature electronic sensor, simply take
> multiple shorter exposures and add them together.
> Astronomers are taking many hour long exposures with
> DSLRs a few minutes at a time, and the resulting images
> are gorgeous.

But in this case, since light levels are so low but exposure times are
long, thermal noise should dominate, so the cooling should be effective.
--

Alfred Molon
------------------------------
Olympus 4040, 5050, 5060, 7070, 8080, E300 forum at
http://groups.yahoo.com/group/MyOlympus/
Olympus 8080 resource - http://myolympus.org/8080/
Anonymous
January 23, 2005 4:07:39 PM

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

Alfred Molon wrote:

> In article <41F2C56D.2070001@qwest.net>, Roger N. Clark (change username
> to rnclark) says...
>
>
>>Well, there is one small ugly fact of physics that limits
>>this gain idea: photon counting statistics. With a gain around
>>1 electron per adu, you are getting everything there is.
>>Modern DSLRs and P&S cameras are photon noise limited.
>>Improving the gain with lower read noise will not really
>>help much.
>
>
> I used to think thermal noise and read noise were dominant. But maybe
> you are right, who knows.

We do know , and definitively. See:

The Signal-to-Noise of Digital Camera images
http://clarkvision.com/imagedetail/digital.signal.to.no...

Even some P&S cameras are photon noise limited. The Nikon D70,
, Canon 10D, 20D, 300D, and 1D Mark II are all photon
noise limited, by direct testing, and by comparison of
noise data on dpreview.com, probably most cameras being
produced today are photon noise limited.

>
>>Again, the gain of operating the CCD at low temperatures
>>is a negligible effect on ordinary photography with
>>normally exposed images. The low temperatures serves
>>to reduce noise from dark current, but even here a
>>new strategy has been developed to circumvent this problem
>>to a large degree: for long low light exposures with
>>a ambient temperature electronic sensor, simply take
>>multiple shorter exposures and add them together.
>>Astronomers are taking many hour long exposures with
>>DSLRs a few minutes at a time, and the resulting images
>>are gorgeous.
>
>
> But in this case, since light levels are so low but exposure times are
> long, thermal noise should dominate, so the cooling should be effective.

Yes, cooling is effective but as I previously stated, multiple
short exposures have proven no need for cooling. Here are some
examples:

http://astro.nightsky.at/Photo/Gal/M31_WN.html

http://astro.nightsky.at/Photo/Neb/B33_WN.html

http://www.clarkvision.com/galleries/gallery.astrophoto...

Roger
Anonymous
January 24, 2005 2:08:06 AM

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

In article <41F4040B.8030502@qwest.net>, Roger N. Clark (change username
to rnclark) says...

> > But in this case, since light levels are so low but exposure times are
> > long, thermal noise should dominate, so the cooling should be effective..
>
> Yes, cooling is effective but as I previously stated, multiple
> short exposures have proven no need for cooling.

The question is what ISO boost you get by cooling. If you operate a CCD
at -let's say- -50°C how much lower are noise levels (than at +20°C)?
--

Alfred Molon
------------------------------
Olympus 4040, 5050, 5060, 7070, 8080, E300 forum at
http://groups.yahoo.com/group/MyOlympus/
Olympus 8080 resource - http://myolympus.org/8080/
Anonymous
January 24, 2005 2:08:07 AM

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

Alfred Molon wrote:

> In article <41F4040B.8030502@qwest.net>, Roger N. Clark (change username
> to rnclark) says...
>
>
>>>But in this case, since light levels are so low but exposure times are
>>>long, thermal noise should dominate, so the cooling should be effective.
>>
>>Yes, cooling is effective but as I previously stated, multiple
>>short exposures have proven no need for cooling.
>
>
> The question is what ISO boost you get by cooling. If you operate a CCD
> at -let's say- -50°C how much lower are noise levels (than at +20°C)?

For short exposers, essentially no boost. Even if read
noise were reduced a little, for normal exposures, the signal
is so large that read noise is negligible. For long
exposures, cooling reduces noise from dark current. It does
not change gain, and it does not change the number of
electrons per photon, so there is no ISO change at all.

With DSLR's being able to take 10+ minute exposures, they
already have pretty low dark current noise, but I would be
interested in seeing some numbers on the sensors. I have
not measured it myself. I'll add it to my list of a gazillion
other things to do. But from what I'm seeing with the multiple
frames being added, I think it is a small increase. In fact,
the improvement is impressive enough that I'm even seeing
the cooled CCD folks doing short exposures and adding them.
There are other advantages of multiple short exposures, such as
if one gets messed up, like bumping the camera during an
exposure, simply throw it out of the average.

Roger
Anonymous
January 24, 2005 3:45:37 AM

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

In article <41F423D8.5060206@qwest.net>, Roger N. Clark (change username
to rnclark) says...

> For short exposers, essentially no boost. Even if read
> noise were reduced a little, for normal exposures, the signal
> is so large that read noise is negligible. For long
> exposures, cooling reduces noise from dark current. It does
> not change gain, and it does not change the number of
> electrons per photon, so there is no ISO change at all.

In a CCD ISO is defined among other factors by the gain of the
amplifier, meaning that you can easily boost ISO by amplifying more the
signal.

But it only makes sense to do so if noise is low enough, i.e. if noise
is low enough you can amplify more the signal and thus have a higher
ISO.
--

Alfred Molon
------------------------------
Olympus 4040, 5050, 5060, 7070, 8080, E300 forum at
http://groups.yahoo.com/group/MyOlympus/
Olympus 8080 resource - http://myolympus.org/8080/
Anonymous
January 24, 2005 3:45:38 AM

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

Alfred Molon wrote:

> In article <41F423D8.5060206@qwest.net>, Roger N. Clark (change username
> to rnclark) says...
>
>
>>For short exposers, essentially no boost. Even if read
>>noise were reduced a little, for normal exposures, the signal
>>is so large that read noise is negligible. For long
>>exposures, cooling reduces noise from dark current. It does
>>not change gain, and it does not change the number of
>>electrons per photon, so there is no ISO change at all.
>
>
> In a CCD ISO is defined among other factors by the gain of the
> amplifier, meaning that you can easily boost ISO by amplifying more the
> signal.
>
> But it only makes sense to do so if noise is low enough, i.e. if noise
> is low enough you can amplify more the signal and thus have a higher
> ISO.

It also only makes sense if you are not fully digitizing
the electrons. Once you get to 1 electron per digital
number, you can not get any advantage to additional gain.
We are already there. See the gains factors on my page:
http://clarkvision.com/imagedetail/digital.signal.to.no...
see table 1.

Roger
!