«

Robert Feinman wrote:

>>Film, however, has only a few percent quantum efficiency, about
>>3%. This in the same 67 square microns, film would convert
>>only ~172,00 *0.03 = 5160 photons. If film could record every
>>photon into density with no noise, it would have a dynamic range
>>of only 5160 or 12.3 stops. But film is not very good. While
>>a single photon may convert a single grain, one really needs a
>>grain clump activated, so many grains.
>
>
> It is widely reported in the technical film literature that 3 photons
> are required to make a crystal developable. How many photons must hit
> the grain for three to interact is another question...
>
> From a practical point of view, I find that using modern color negative
> film, shooting in bright daylight, I can capture the full sun in one
> part of my panoramas and get proper exposure for the rest of the scene.
> What I mean is, that the sky is captured correctly right up to the edge
> of the solar disk. This seems to indicate a pretty high tolerance for
> overexposure. I would guess somewhere in the range of 12 stops.

This depends on what you mean by the edge of the solar disk.
Just off the sun is the solar corona, which is very bright out
several solar radii. This is many more than 12 stops, probably
more than 20 stops. So, you must mean something different.
I do have measurements somewhere to get a real number...

> I see Roger Clark's studies that digital is even better, but this
> doesn't seem to agree with most people's field experience. For example,
> this month's Petersons magazine has two articles about the limitations
> of digital dynamic range. In one the author assumes 7 stops and uses
> multiple images to extract correctly exposed regions and combine.
> In the other the discussion is about the usual tricks to minimize scene
> contrast. So it seems that working pros find digital contrast range
> similar to film transparencies.

The problem with digital is that you can't overexpose, because once you
hit the max signal (4095 on a 12-bit scale), you have zero additional
information. So people cite this as a problem and use it to show digital
lacks what film can do. But they rarely look at the low end, which
is where digital shines, especially when compared to film.

> I don't have any explanation as to why the field results and the
> theoretical (or controlled experimental) don't seem to agree.
> Roger??

The field results agree with lab testing when you use the same
metrics. With digital, you expose to not clip the highlights,
but maximizing the signal. It is a different strategy than with
slide or print film. Those who master it will be able to
extract far more from their images than any film.

But just how much depends on the camera's sensor. The pixel size
determines the full well capacity, or maximum possible signal.
Sensors with small pixels tend to be point and shoot cameras,
with poorer (cheaper) electronics with more noise. So reduced
maximum signal combined with high read-out noise results in
reduced dynamic range. This page,
Digital Cameras: Does Pixel Size Matter?
Factors in Choosing a Digital Camera
http://www.clarkvision.com/imagedeta...el.size.matter
describes the issues.

See also: The Signal-to-Noise of Digital Camera images
and Comparison to Film:
http://www.clarkvision.com/imagedeta...ignal.to.noise

Roger
Photos at: http://www.clarkvision.com

"Scott W" <> wrote in message
news: oups.com...
>> > Here is a photo from my 20D, not a bad exposure, this was taken at 1/60
>> > of a second.
>> > http://www.sewcon.com/photos/IMG_8207.jpg
>>
>> Looks on the verge of over exposure to me.
>>
>> >
>> > Here is I have taken the photo at 1/8000 of a second, 7 stop under
>> > exposed. The ISO and F number were left the same as the 1/60 shot.
>> > http://www.sewcon.com/photos/IMG_8206.jpg
>> > Level were adjusted to pull out as much as I could.
>> >
>>
>> I wouldn't call what you posted "viewable" or even useable.

I'm also unclear what the -7 stop shot is to represent, but let's hold up
here for a moment. The first shot shows a brightness range of about 9 stops:
soft hazy sun and shadows beneath the foreground foliage; I would guess
right about 4 stops difference in incidence readings between the two. No
real blacks that I can see, and good details in both shadows and sunlight
grass. From casual inspection alone, and I emphasize casual, the 20D has a
dynamic range of better than 9 stops; 11 to 12 stops looks easily plausible.
IOW, that's a real nice piece of kit. The days for 35mm slide film are
clearly numbered.

However, color negative film also easily spans the same 12 stops. Without
getting into the science of it, I'll just note anecdotally that disposable
cameras have no exposure adjustment at all, and deliver usable exposures
from sunny-16 conditions down into deep shade. The film isn't the problem;
it's the dimestore prints that obscure the magic. Most consumers see only
the 5 or 6 stops in the overly harsh, contrasty prints. The real limits and
capabilities remain hidden behind that strange orange mask. Noise is a
different matter; I'll defer to Mr. Clark's inestimable knowledge on that
topic.

The real magic we're seeing (in the test shots) is Photoshop. We now have
more brightness range than we can print. It's an embarassment of riches,
really. That softly glowing look doesn't work for all subjects, and worse,
the print will be disappointingly flat compared to what we see on screen.

So it comes down to this: Despite advances in film and sensors, we still
have to light and shoot so it will print. Life is too short and too full of
wonders to spend it masking and tugging and pulling in the wee hours of the
night on those extra few stops. Maybe put them to good use to not have to
obsess over metering or histograms. That's worth something. Dunno. To each
their own.

>> > I would really like to see someone take two film photos with this range
>> > using film and get even close to a viewable photo.
>> >
>>
>> Color negative film can deal with a BUNCH of over exposure and still
>> produce
>> a nice print. I'd say anything from 1 stop under to 3 stops over will be
>> able to pull a pretty decent print from i.e. most people would never see
>> the difference looking at the final prints, not a noisy useless mess like
>> you displayed as an example of the dynamic range of digital. Next go 2-3
>> stops over what a good meter reads with a digital camera and see what you
>> have left in the highlights. Nothing! Digital is slightly better than
>> silde
>> film if you shoot RAW but it's still not even close to shooting color
>> negative film.
>>
> My two photos were shot at 7 stops apart, not the 4 you are talking
> about.
> So lets see it, let see you photos with that kind of range.
> Lets see you take two photo 7 stops apart and see what you can get from
> film.
>
> I rather doubt you will get any kind of image at all.
>
> Scott
>

Boat wrote:
>
> However, color negative film also easily spans the same 12 stops. Without
> getting into the science of it, I'll just note anecdotally that disposable
> cameras have no exposure adjustment at all, and deliver usable exposures
> from sunny-16 conditions down into deep shade. The film isn't the problem;
> it's the dimestore prints that obscure the magic. Most consumers see only
> the 5 or 6 stops in the overly harsh, contrasty prints. The real limits and
> capabilities remain hidden behind that strange orange mask. Noise is a
> different matter; I'll defer to Mr. Clark's inestimable knowledge on that
> topic.

Well in fact I took one of those disposable camera into the deep shade.
We were on a hike and rain was perdiced so I did not want to take my
SLR (I should have)so I tried one of those disposable cameras.

In shade this is what I got.
http://www.sewcon.com/photos/0503.jpg
The negative of course was very light, this is the best scan I could
get from it.

As the rain came in and it got darker this is the best I could do
http://www.sewcon.com/photos/0301.jpg

So why bother to take this kind of camera, well the sun will come out
in time
http://www.sewcon.com/photos/sunsout.jpg

What started this thread was a comment by a fellow that said he was
getting more range from his 1Ds Mark II then he had been from film,
that he was delighted in being able to get detail in the shadow where
he had problems before.

I have noticed the same thing, with scanned film I have a very limited
amount of detail in the shadow, but if I use the 20D there is a ton of
room in the shadow.

Both he and I are using antidotal evidence, both noting that we got
more range with digital but Roger has done a careful study of this and
has put it to numbers. There is a common believe that film gives you
12 stops, but have you tested this? The one person I know how has
shows that this is not the case.

Scott

"Scott W" <> wrote in message
news: oups.com...
> In shade this is what I got.
> http://www.sewcon.com/photos/0503.jpg
> The negative of course was very light, this is the best scan I could
> get from it.
>
> As the rain came in and it got darker this is the best I could do
> http://www.sewcon.com/photos/0301.jpg
>
> So why bother to take this kind of camera, well the sun will come out
> in time
> http://www.sewcon.com/photos/sunsout.jpg

The third one illustrates lens flare, not the shoulder of the response
curve. The "lens" on the disposable I disassembled was a single hardened
droplet of mostly clear plastic. Coated optics will do better, but can still
be problematic. As for the other two, I wasn't there when you took them, and
I wasn't there to watch you scan them. My experience was/is different from
yours.

> Both he and I are using antidotal evidence, both noting that we got
> more range with digital but Roger has done a careful study of this and
> has put it to numbers. There is a common believe that film gives you
> 12 stops, but have you tested this? The one person I know how has
> shows that this is not the case.

I have my own collection of unprintable shots combining cumulus tops in full
sun and main subject detail in shade. That's an EV range of 17 down to 4 on
my spotmeter. 12 stops is routinely possible, even if not ideal.

Anyway, I was agreeing with your observations about the 20D. And also
reminding that it still has to get onto paper at some point.

In article < .com>,
Scott W <> wrote:
>
>
>Chris Brown wrote:
>> In article <>,
>> Bob <> wrote:
>> >
>> >Digital cameras have a maximum range of 12 bits since the file is 12 bits...
>>
>> That's an indication of precision, not range.
>>
>> The dynamic range is the differencee between the highest intensity that
>> records as black, and the lowest intensity that records as white in a single
>> shot. That black will typically be recorded as zero, but you get the same
>> dynamic range regardless of whether you record the white as 255 (8 bits) or
>> 4095 (12 bits), other things being equal. In the latter case, you just have
>> more shades between them.
>
>Well no you are not right here. If we assume a linear A/D converter,
>and they are, then if you want 12 stop of range you will need at least
>a 12 bit converter.

No you don't. You can get a 12 stop dymanic range with any bit depth you
like. Try the following thought experiment - imagine you have an image that
has those 12 stops of dynamic range, and it's been digitised with 12 bit
precision. The image (we'll assume it's monochrome for simplicty - it doesn't
affect the point) represents black as 0 and white as 4095.

But you can't display that image on your graphics card - you need to drop it
down to 8 bits of precision, so you run it through a convertor - black is
now 0 and white is now 255.

That image still represents exactly the same dynamic range - what was white
before is still white now, and what was black before is still black now, and
they still correspond to the same captured light levels. All we have changed
is the precision with which we can represent the intermediate greys.

You are confusing range and precision.

Chris Brown wrote:
> In article < .com>,
> Scott W <> wrote:
> >
> >
> >Chris Brown wrote:
> >> In article <>,
> >> Bob <> wrote:
> >> >
> >> >Digital cameras have a maximum range of 12 bits since the file is 12 bits...
> >>
> >> That's an indication of precision, not range.
> >>
> >> The dynamic range is the differencee between the highest intensity that
> >> records as black, and the lowest intensity that records as white in a single
> >> shot. That black will typically be recorded as zero, but you get the same
> >> dynamic range regardless of whether you record the white as 255 (8 bits) or
> >> 4095 (12 bits), other things being equal. In the latter case, you just have
> >> more shades between them.
> >
> >Well no you are not right here. If we assume a linear A/D converter,
> >and they are, then if you want 12 stop of range you will need at least
> >a 12 bit converter.
>
> No you don't. You can get a 12 stop dymanic range with any bit depth you
> like. Try the following thought experiment - imagine you have an image that
> has those 12 stops of dynamic range, and it's been digitised with 12 bit
> precision. The image (we'll assume it's monochrome for simplicty - it doesn't
> affect the point) represents black as 0 and white as 4095.
>
> But you can't display that image on your graphics card - you need to drop it
> down to 8 bits of precision, so you run it through a convertor - black is
> now 0 and white is now 255.
>
> That image still represents exactly the same dynamic range - what was white
> before is still white now, and what was black before is still black now, and
> they still correspond to the same captured light levels. All we have changed
> is the precision with which we can represent the intermediate greys.
>
> You are confusing range and precision.

Well first off you much understand that when we convert to 8 bit it is
normally in a non-linear way, this mean the a step in light from level
0 to level 1 is no 1/255 of level 255, it is much smaller, how much
depends on the color space you are working in.


The next point is that when we deal with dynamic range we are dealing
with the smallest signal that can be seen above the noise to the
highest level, if the system is electrically very quite then the lowest
signal that can be detected will be level one on the A/D converter,
since this are linear that would be 1/4095 of the highest level.

If there is electrical noise, and there always is, then we add in what
is call quantizing noise, this turns out to be 1/sqrt(12) of one bit
level , when dealing with the RMS value of the noise.

Noise limits the dynamic range of a system and when we use an A/D
converter just the fact that we are converting to a digital level adds
noise. The goal is to have the quantizing noise not add much noise,
this is not hard to adchive. If we have a system with electrical noise
with an RMS level of say 1 A/D level the effect of the A/D would rise
this noise to 1.04 of a level. But keep in mind that an RMS noise of 1
A/D level will be bouncing a lot between +2 levels and -2 levels and so
there will be a fair bit of noise on the signal.

The number of bits in the A/D converter sets an upper limit to the
dynamic range of a system but it does not set the range, which can be
much lower depending on the noise levels in the system.

Scott

Boat wrote:

> The third one illustrates lens flare, not the shoulder of the response
> curve. The "lens" on the disposable I disassembled was a single hardened
> droplet of mostly clear plastic. Coated optics will do better, but can still
> be problematic. As for the other two, I wasn't there when you took them, and
> I wasn't there to watch you scan them. My experience was/is different from
> yours.

The third one in fact is not lens flare but fog, not film fog but real
water droplets in the air kind of fog.

> > Both he and I are using antidotal evidence, both noting that we got
> > more range with digital but Roger has done a careful study of this and
> > has put it to numbers. There is a common believe that film gives you
> > 12 stops, but have you tested this? The one person I know how has
> > shows that this is not the case.
>
> I have my own collection of unprintable shots combining cumulus tops in full
> sun and main subject detail in shade. That's an EV range of 17 down to 4 on
> my spotmeter. 12 stops is routinely possible, even if not ideal.
>
> Anyway, I was agreeing with your observations about the 20D. And also
> reminding that it still has to get onto paper at some point.

Less and less is ever getting to paper, more and more is simply being
viewed on the computer screen, which also has a limited range.

But even with paper there are reasons why you would like a bit more
range, with more range you can do more dodging and burning, for me
using software. I have a large number of prints where without the use
of this I would have no choice but to blow out the sky if I want the
foreground to look at all good.

Negatives do have enough range to do a fair bit of dodging and burning,
but it is hard to get to it, I have to run my negative through the
scanner twice, at different exposures to get the full range out of the
film and even then I don't feel like I have the full range that I get
with the 20D.

Let me sum up buy saying this, I believe that negative film has a bit
less range then my 20D. I believe that for practical cases I get a lot
less range with film then I do with my 20D (due to limitations of my
scanning hardware). For the most part there is enough range in my
scanned photos to get a good print from them, even a little selective
adjustment of levels.

Scott

In article <. com>,
Scott W <> wrote:
>Chris Brown wrote:
>> In article < .com>,
>> Scott W <> wrote:
>> >
>> >
>> >Chris Brown wrote:
>> >> In article <>,
>> >> Bob <> wrote:
>> >> >
>> >> >Digital cameras have a maximum range of 12 bits since the file is
>12 bits...
>> >>
>> >> That's an indication of precision, not range.

[snip]

>The next point is that when we deal with dynamic range we are dealing
>with the smallest signal that can be seen above the noise to the
>highest level, if the system is electrically very quite then the lowest
>signal that can be detected will be level one on the A/D converter,
>since this are linear that would be 1/4095 of the highest level.
>
>If there is electrical noise, and there always is, then we add in what
>is call quantizing noise, this turns out to be 1/sqrt(12) of one bit
>level , when dealing with the RMS value of the noise.
>
>Noise limits the dynamic range of a system and when we use an A/D
>converter just the fact that we are converting to a digital level adds
>noise.

We seem to have got subtly sidetracked here, from representation to
conversion. Bob's orginal claim (it's there above) was about the number of
bits in the file, not the A/D convertor, and he seems to be confusing bits
and stops.

In message <nm46p2->,
Chris Brown <_uce_please.com> wrote:

>In article < .com>,
>Scott W <> wrote:

>>Well no you are not right here. If we assume a linear A/D converter,
>>and they are, then if you want 12 stop of range you will need at least
>>a 12 bit converter.

>No you don't. You can get a 12 stop dymanic range with any bit depth you
>like. Try the following thought experiment - imagine you have an image that
>has those 12 stops of dynamic range, and it's been digitised with 12 bit
>precision. The image (we'll assume it's monochrome for simplicty - it doesn't
>affect the point) represents black as 0 and white as 4095.

>But you can't display that image on your graphics card - you need to drop it
>down to 8 bits of precision, so you run it through a convertor - black is
>now 0 and white is now 255.

>That image still represents exactly the same dynamic range - what was white
>before is still white now, and what was black before is still black now, and
>they still correspond to the same captured light levels. All we have changed
>is the precision with which we can represent the intermediate greys.

>You are confusing range and precision.

You are confusing capture with display, I think. The zero point is
meaningless in capture, as anything but a frame of reference.

Zero doesn't really exist in most displays, but it certainly exists in
sensor captures. In fact, photon-capture zero is usually well above RAW
number 0. It is 129 out of 4095 on my Canon 20D.

What *is* important, is what the ratio of the brightest light captured
at the clipping point is to the lowest signal *above* zero, and you
absolutely *NEED* at least 12 bits of linear capture to have 12 stop of
dynamic range capture at the pixel level (although it does not guarantee
that much dynamic range at the highest capture frequencies).

If you don't like calling this dynamic range, what would you like to
call it?
--

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><

In message <taTve.6$>,
Don Stauffer <> wrote:

>The digitizer quantization is only one factor in the total dynamic
>range. Suppose I have an electronic sensor that has only a four to one
>dynamic range, i.e., the max signal is only four times the noise level
>of the sensor.

What do you mean by "the noise level in the sensor"?

If you digitized to 4096 linear levels in two different systems, and
each had a blackframe with values of 0 to 1024, but one was distributed
like this:

* *
* *
* *
*********
---------------

and the other like this:


*
*
* *
****** *******
----------------

What would be the noise level for each?

Would they have the same dynamic range?

>Nothing prevents me from putting this noisy signal into a 12 or even 16
>bit digitizer. Yet the result is NOT a 12 bit or 16 bit dynamic range.
>It is a 2 bit dynamic range. The sensor quality IS important when
>talking about the dynamic range of a digicam.

The top curve might come close to what you say, but the bottom curve is
more like digicam reality. On my 20D, a a blackframe results in noise
centered on 129, and significant 8 to 25 raw values in either direction
(visible part of bell-curve on histogram), depending on ISO and
tenperature, as well as exposure time.

"The noise" more closely follows the standard deviation than the width
of the noise.

The bit depth, minus losses due to blackpoint offset, is a limiting
factor to dynamic range at the lowest ISOs, regardless of how dynamic
range is measured or defined. Geting more bit depth does not
necessarily increase dynamic range by an equal number of stops as bits,
but when digitization *is* the limiting factor (as it is at ISO 100 on
most DSLRs), extra bits (if digitization is of sufficient quality) do
bring on more dynamic range.
--

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><