dynamic range of digital, larger or smaller?

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

On Sun, 26 Jun 2005 21:55:26 +0100, Eolake Stobblehouse
<eolake@maccreator.spamremove.net> wrote:

>If you read:
>http://www.luminous-landscape.com/columns/1Ds-4x5.shtml
>... you'll see that Mr. Briot says that he gets a *bigger* dynamic
>range from digital (Canon 1DsII) than he does from film (4x5).
>This confuses me, for everything I heard until now indicated that
>digital has a *smaller* dynamic range.
>What is your experience?
>(I have not used film since almost the last time I got laid, so I have
>nothing to compare with.)

That's a REAL GOOD question and can generate mega threads...

OK... it depends on the film! And on the digital cam!

I'd say any digital can beat Kodachrome... and a web cam could probably beat
Polaroid...

But some modern films (is it called Velvia?) can beat digital...

Now if you go to B/W film, that's a different story, since some films I've used
such as Ilford have a useful range of 11 F stops or more... that's a lot....

As for my experience - I can get lots of details out of shadows in digital, but
I've only done color positives in the darkroom, and they come out as is, you
might say, you can't go play with them.

You'd have to hear from someone who has done color printing, I haven't.

And my darkroom is closed!

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

"Eolake Stobblehouse" <eolake@maccreator.spamremove.net> wrote in message
news:260620052155262925%eolake@maccreator.spamremove.net...
> If you read:
> http://www.luminous-landscape.com/ [...] -4x5.shtml
> ... you'll see that Mr. Briot says that he gets a *bigger* dynamic
> range from digital (Canon 1DsII) than he does from film (4x5).
> This confuses me, for everything I heard until now indicated that
> digital has a *smaller* dynamic range.
> What is your experience?
> (I have not used film since almost the last time I got laid, so I have
> nothing to compare with.)
>
> --
> Yours, Eolake
>
> eolake@stobblehouse.com
> http://stobblehouse.com

I think it depends on a lot of factors, with lighting maybe being the most
critical. I've found digital to be close to slide film. I used to print
color, but you can do a lot more with Photoshop and it looks a lot better.

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

On Sun, 26 Jun 2005 17:09:57 -0600, "Roger N. Clark (change username to
rnclark)" <username@qwest.net> wrote:

>Bob wrote:
>
>> On Sun, 26 Jun 2005 21:55:26 +0100, Eolake Stobblehouse
>> <eolake@maccreator.spamremove.net> wrote:
>>
>>
>>>If you read:
>>>http://www.luminous-landscape.com/columns/1Ds-4x5.shtml
>>>... you'll see that Mr. Briot says that he gets a *bigger* dynamic
>>>range from digital (Canon 1DsII) than he does from film (4x5).
>>>This confuses me, for everything I heard until now indicated that
>>>digital has a *smaller* dynamic range.
>>>What is your experience?
>>>(I have not used film since almost the last time I got laid, so I have
>>>nothing to compare with.)
>>
>>
>> That's a REAL GOOD question and can generate mega threads...
>>
>> OK... it depends on the film! And on the digital cam!
>>
>> I'd say any digital can beat Kodachrome... and a web cam could probably beat
>> Polaroid...
>>
>> But some modern films (is it called Velvia?) can beat digital...
>
>No! Digital, has a very high dynamic range, see:
>
> Dynamic Range and Transfer Functions of Digital Images
> and Comparison to Film
> http://www.clarkvision.com/imagedetail/dynamicrange2
>
>Roger

Digital cameras have a maximum range of 12 bits since the file is 12 bits...
scanning film with a 12 bit scanner is a waste of time... if the film was 14
bits how would the scanner see it? That site compares 12 bit digital to a 12
bit scan of film...

Unfortunately I can't find the site I went to originally that tested the range
of various films.

>>
>> Now if you go to B/W film, that's a different story, since some films I've used
>> such as Ilford have a useful range of 11 F stops or more... that's a lot....
>>
>> As for my experience - I can get lots of details out of shadows in digital, but
>> I've only done color positives in the darkroom, and they come out as is, you
>> might say, you can't go play with them.
>>
>> You'd have to hear from someone who has done color printing, I haven't.
>>
>> And my darkroom is closed!
>>

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

Bob wrote:

> On Sun, 26 Jun 2005 17:09:57 -0600, "Roger N. Clark (change username to
> rnclark)" <username@qwest.net> wrote:
>
>>Bob wrote:
>>>But some modern films (is it called Velvia?) can beat digital...
>>
>>No! Digital, has a very high dynamic range, see:
>>
>> Dynamic Range and Transfer Functions of Digital Images
>> and Comparison to Film
>> http://www.clarkvision.com/imagedetail/dynamicrange2
>>
>>Roger
>
>
> Digital cameras have a maximum range of 12 bits since the file is 12 bits...
> scanning film with a 12 bit scanner is a waste of time... if the film was 14
> bits how would the scanner see it? That site compares 12 bit digital to a 12
> bit scan of film...
>
> Unfortunately I can't find the site I went to originally that tested the range
> of various films.

Let's look at the scientific plausibility of 14 bit dynamic range from film.
By the definition and calibration of ISO we can use the full well capacity
of digital sensors to see how many photons are being collected in a
given exposure. E.g. see:
http://clarkvision.com/imagedetail [...] l.to.noise
and references therein.

For an 8.2 micron pixel spacing on a 1D Mark II camera, = 67 square
microns, the electronic sensor collects a maximum of ~52,300 electrons,
corresponding to the brightest parts in a scene (e.g. 100% reflectance
when the exposure is set for an 18% gray card). The CMOS sensors
have quantum efficiencies of 25-40%. Let's assume 30%, so the photons
incident in the green passband is about 3.3 times higher, or about
172,000 photons per 67 square microns. The 1D Mark II sensor has a
read noise of about 7 electrons, so the dynamic range is
about 52,300/7 ~ 7400, or over 12.8 stops, and is thus limited by
the 12-bit analog to digital converter electronics.

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. This is why characteristic
curves for film have a "toe." Film needs many photons, far more than
electronic sensors to record that first low level signal. What that
is needs some study, but it is clearly more than 10 converted photons,
based on comparisons to electronic sensors and their measured read noise.
So, assuming an optimistic 10, then the dynamic range of film would be
about 5160/10 ~ 516, or 9 stops.
This is evident in the response curves at:
http://clarkvision.com/imagedetail/dynamicrange2
in Figure 8b. Note the high noise in the film at the low end.
That is not scanner noise, but noise in the film. Now look at
Figure 10: Kodak's characteristic curve for Kodak Gold 200. The curve
goes much lower because they averaged over a large area in order
to average the noise. But the curve at the low end is of no
practical use for recording image detail.

Scott's test is an excellent illustration of that limit:
expose film and digital at meter and -7 stops and see which can
pull more info out of the noise.

On my dynamic range page (above), the images in Figure 5 clearly show
the same effect, and that is true when examining the film with
a loupe, or a microscope, as well as a with film scanner.

Summary: 14-bits is way overkill for film, and many argue 12-bits
is overkill. I argue 8-bit scans are not enough, so I scan
all my film at 12-bits/channel because that is the next highest
setting on film scanners above 8-bit.

Roger

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

Stacey wrote:
> Scott W wrote:
>
> >
> > It would appear that yes a good digital camera has more range then
> > film, even negative film.
>
> I disagree with that statement. Some are better than slide film but not as
> wide as color negative film.
>
> >
> > 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 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

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

In article <k5cub1ptnvm6uo4uunn631c98qv1pjp6b2@4ax.com>,
Bob <FlintsTone@nospam.com> wrote:

>But some modern films (is it called Velvia?) can beat digital...

Velvia has tiny dynamic range. Underexpose it by a bit, and your picture is
often unsalvagable.

Looks lovely when it works, but it's a right sod to expose. Not the best
film to pick if you're looking for dynamic range, nor is any other E6 film.
You want C41 or B&W films for a dynamic range comparison.

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

according to Fujifilm Research and Development No. 50-2005, page 5;
Fig. 4 -dynamic range of CCD is about a half of that of a film.
I do not know japanese, but they compare DSC at ISO 100 and a film of
ISO 1600.


On Mon, 27 Jun 2005 15:23:52 +0300, "Dimitrios Tzortzakakis"
<dimtzort@otenet.gr> wrote:

*I have done color printing, and I can tell you, it is a pain in the ass.You
*have to collect enough negatives for a printing batch, but usually your
*chemicals are kaputt before that,in B/W I could never print an acceptable
*print, I stained the paper with my fingers, my chemicals were kaputt every
*month, now I gave away my darkroom, and I have Kodak CX 7300 which is the
*best camera I've ever had,my SLR became my nightmare, of course to shoot a
*36 exp.film I needed months,I had slide films in the refrigerator for years
*unexposed, my mother was complaining on the room the papers were taking in
*the freezer (because of the hot greek climate even B/W papers were kaputt
*outside refrigerator in a few months)...Is that enough?

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

Chris Brown wrote:
> In article <6jqub1t3llgg6mklfg7tscg8nho4ks4cl3@4ax.com>,
> Bob <FlintsTone@nospam.com> 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. The dynmic is range is not the ratio of the
highest to the lowest level, if the lowest is zero you would be
dividing by zero. The dynamic range is the ratio of the highest level
to the lowest level that is just out of the noise.

Scott

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

> 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.

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.

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

--
Robert D Feinman
Landscapes, Cityscapes and Panoramic Photographs
http://robertdfeinman.com
mail: robert.feinman@gmail.com

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

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/imagede [...] ize.matter
describes the issues.

See also: The Signal-to-Noise of Digital Camera images
and Comparison to Film:
http://www.clarkvision.com/imagede [...] l.to.noise

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

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

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

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

In article <1119878855.579998.185230@g43g2000cwa.googlegroups.com>,
Scott W <biphoto@hotmail.com> wrote:
>
>
>Chris Brown wrote:
>> In article <6jqub1t3llgg6mklfg7tscg8nho4ks4cl3@4ax.com>,
>> Bob <FlintsTone@nospam.com> 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.

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

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

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

In message <nm46p2-kv2.ln1@narcissus.dyndns.org>,
Chris Brown <cpbrown@ntlworld.no_uce_please.com> wrote:

>In article <1119878855.579998.185230@g43g2000cwa.googlegroups.com>,
>Scott W <biphoto@hotmail.com> 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 <JPS@no.komm>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><

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

In message <3i9nobFk9ecsU1@individual.net>,
Stacey <fotocord@yahoo.com> wrote:

>> 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.

That's a subjective evaluation.

The objective fact, however, is that the most obvious problems with the
image are that the RAW data has visible biases from the readout in both
horizontal and vertical form, and this is potentially removeable, either
by a simple algorithm that averages each horizontal and vertical line of
green pixels, and subtracts the difference between it and the average of
its two neighbors (which will lose some edge detail, and have a few
minor artifacts due to real fluctuations in green due to high-contrast,
high-frequency detail), or an algorithm that looks for the patterns of
line-based offsets, and adjusts them.

The software is coming, and these types of corrections are *not*
noise-removal, per se, like the noise-removal that removes detail.
Detail is more visible with these lines properly blackpointed, where
only random noise remains.
--

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS@no.komm>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><

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

David J. Littleboy wrote:
> <JPS@no.komm> wrote:
> > Chris Brown <cpbrown@ntlworld.no_uce_please.com> wrote:
> >>Scott W <biphoto@hotmail.com> 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?
>
> I think that a lot of people have been using the term "dynamic range" when
> they should have been using the term latitude<g>.
>
> Informally (_very_ informally), the dynamic range of the sensor _ought_ to
> be "mappable" onto any arbitrary range of intensities you like, but I doubt
> it works that way.
>
> Since the sensor response is linear, a sensor with a dynamic range (max
> signal divided by noise floor) of N stops will only be able to respond to N
> stops in the scene. Unless you can adjust the _slope_ of the linear response
> of the sensor.
>
> But I've never heard of a slope of other than 1 (i.e. doubling the signal
> intensity (brightness) doubles the response) in a digital camera. Am I wrong
> in this last bit here???
>
> If I'm right, then Scott is right, and if the slope can be fudged, then
> Chris is. I think.
>
I believe that pretty much all scanners and camera are linear into the
A/D. The only one that I know of that might not be is the Fuji CCD,
which they seem to imply goes non-linear at high signal levels.

Whereas you could make a camera with a non-linear section between the
CCD/CMOS sensor and the A/D it is easier to do the non-linear stuff in
hardware/software.

12 bits is a lot of range and it is unlikely to be the real limit of
dynamic range for any of the sensors out there today. If the sensors
get quieter then expect to see more bits on the converters.

Scott

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

On Mon, 27 Jun 2005 08:28:53 -0500, Don Stauffer <stauffer@usfamily.net> wrote:

>Bob wrote:
>>
>> Digital cameras have a maximum range of 12 bits since the file is 12 bits...
>> scanning film with a 12 bit scanner is a waste of time... if the film was 14
>> bits how would the scanner see it? That site compares 12 bit digital to a 12
>> bit scan of film...
>>
>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.

Then you have a true 2-bit sensor!

>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.

Of course I know that!

What I am suggesting is you cannot measure something with a ratio of 14 bits
with a 12 bit measuring system...

People are confusing number systems with film range, which is conveniently
measured in 'stops' or actually a ratio system of two's.

You cannot just divide out a sensor range by any arbitrary number and get a
meaningfully answer!

A 12 bit file has a capability of 12 stops of measurement. You could measure an
8 bit camera, and it would read as a portion of the available range. You cannot
measure a 14 bit camera or film scanner since it would overflow by 2 bits.

It probably would be better to measure film in DB like all of the other
engineering systems, it would make more sense... a 12 bit camera has a range of
36 DB. an 8 bit camera has a range of 24 DB. a 16bit camera is 48DB. You can't
measure 48db on a 36db system!

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

On Mon, 27 Jun 2005 15:23:52 +0300, "Dimitrios Tzortzakakis"
<dimtzort@otenet.gr> wrote:

>I have done color printing, and I can tell you, it is a pain in the ass.You
>have to collect enough negatives for a printing batch, but usually your
>chemicals are kaputt before that,in B/W I could never print an acceptable
>print, I stained the paper with my fingers, my chemicals were kaputt every
>month, now I gave away my darkroom, and I have Kodak CX 7300 which is the
>best camera I've ever had,my SLR became my nightmare, of course to shoot a
>36 exp.film I needed months,I had slide films in the refrigerator for years
>unexposed, my mother was complaining on the room the papers were taking in
>the freezer (because of the hot greek climate even B/W papers were kaputt
>outside refrigerator in a few months)...Is that enough?

Ya thats funny! I STILL have film in the fridge, probably 15 years old!

Not to mention the B/W paper I got from my Dad who bought it in the 60s....

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

Bob wrote:

> Therefore, when we talk about a dynamic range of 12 bits we are talking about 12
> stops...
>
> I know what you mean about dividing a number by 255 or 4096 or whatever, you
> could apply that to the signal level coming off of the sensor to get true
> dynamic range, HOWEVER we aren't talking about that, we are talking about 12
> STOPS of light or a true range of 2 ^ 12 as a ratio of no light to most light we
> can capture.
>
> That's not the same as dividing a sensor reading or an existing picture by an
> arbitrary binary number.
>
> A 12 bit photo can have an 8 bit photo contained in it, and can be shown on our
> monitor. HOWEVER said photo will also have 4 more bits of range, either in the
> dark or in the light, that can be extracted. An 8 bit photo will not have any
> extra 'range'.

No.
See figure 7 at:
http://www.clarkvision.com/imagedetail/dynamicrange2
Figure 7 shows the 8-bit data from a jpeg file plotted on top of
12-bit data from a raw file of the same image. The dynamic range of
the two are the same.
>
> You can't do that by simply re-sampling any photo to any bit number.
>
> And scanning with a 12 bit scanner limits your readings to LESS then 12 bits due
> to error.

When you are scanning film, you are digitizing the film optical density,
which is NOT the same as dynamic range of the original
scene; it is much compressed.
>
> You could try scanning the film twice, once with a max light value calibration
> and then again with a lower calibration, and mix them into a larger file...

If you have a good scanner, this is not necessary.
>
> One other thing about film scanning, the range you ultimately get will depend on
> the brightness of the lamp compared with the density of the film, you have to
> have enough light so that the min value of light is above zero, or you won't be
> reading the proper range.

It also depends on the noise floor of the scanner. You need a less bright
light if you have a low noise sensor.
>
> Not to mention the non-linearity of film...

Exactly. 12-bits on scanned film is a much greater scene intensity
range than recorded in density range. For example, see figure 10 at:
http://www.clarkvision.com/imagedetail/dynamicrange2
The Kodak characteristic curve for Kodak Gold 200 goes over a scanner
DN range of 1000 to 65,535 for a scene range of 10 to 65,535 (arbitrary
DN units). This says that the film density range is only a factor
of 65, for a scene range of 6,500! The scanner does not need as much
dynamic range as the film recording the original image.

Roger

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

Stacey wrote:

> Scott W wrote:
>
>
>>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.
>
>
> So? This is a test of your scanner, not the film, which doesn't seem to be
> up to the task.. And again that noisy mess you posted as 7 stops under
> isn't a "useable image" by any stretch of anyones imagination, other than
> maybe your own?

I didn't read that Scott said it was a usable image.
The point is that some information can be extracted from the
digital image, and film in the same conditions would show much less,
and I agree. He challenged anyone to do as well with film.
See Figure 5 at:
http://www.clarkvision.com/imagedetail/dynamicrange2
which shows similar results of digital compared to film,
and a quantitative analysis is shown on the page too.

Roger

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

Roger N. Clark (change username to rnclark) wrote:
> Stacey wrote:
>
> > Scott W wrote:
> >
> >
> >>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.
> >
> >
> > So? This is a test of your scanner, not the film, which doesn't seem to be
> > up to the task.. And again that noisy mess you posted as 7 stops under
> > isn't a "useable image" by any stretch of anyones imagination, other than
> > maybe your own?
>
> I didn't read that Scott said it was a usable image.
> The point is that some information can be extracted from the
> digital image, and film in the same conditions would show much less,
> and I agree. He challenged anyone to do as well with film.
> See Figure 5 at:
> http://www.clarkvision.com/imagedetail/dynamicrange2
> which shows similar results of digital compared to film,
> and a quantitative analysis is shown on the page too.
>
Exactly so, there are those who say digital cameras have about the same
range as slide film. If there were the case then underexposing by 5
stops and there would be about nothing left, under exposing by 7 and
there would be nothing left. The point of my underexposed photo was
not that it was good but rather that there was still range left after
going down 7 stops, something that slide film could never do. How much
range is left is hard to tell but I believe it is more then you would
find if you were to do this test with print film.

Roger's test is more compelling, since it shows the output from
film, both print and side, compared to the 1Ds Mark II, which has about
the same range as the 20D.

Scott

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

nailer wrote:
> someone already has stated that such comparisons are futile due to
> large variety of CDD and films of various quality.
> In my view, only very expensive CDD devices are better than film (in
> that respect at least).
> You may have your own opinion. If you know Japanese, you can read the
> article. I don't.
>
This is an issue that many have a great interest in and this whole
thread was started when the OP read an article in The Luminous
Landscape by Alain Briot, were he stated that he was getting much
better range out of his 1Ds Mark II then he had gotten out of film.
Note that this was not a case of stating that every CCD would beat
every film, it was a case of the 1Ds Mark II beating whatever film
Alain used. Now Alain did not put this forward so much as a detailed
study as he was showing what he got when he shot film and when he shot
digital and he was giving the reasons that digital was working better
for me.

People, being people, questioned Alain's statements that he was
getting more range from his digital camera then from his film camera.
Note that Alain never claimed that there was not more range in film
what he claimed was that for his workflow, using his scanner he got far
more range with the Canon then he ever did with his 4 x 5 camera.

So the question raised by the OP was, what are other peoples
experience. This does not seem at all like a futile question to me,
other will be using different film and different digital cameras. Of
all the people who have replied to the OP's post Rogers has done the
most work in this area. People may have a hard time understanding his
curves (this would appear to be the case) but his test photos are hard
to miss, see here.
http://clarkvision.com/imagedetail/dynamicrange2/

These are pretty reveling photos, they clearly show that for the film
scanner combination that Roger used the 1D Mark II had far more range
then film.

If others have photos showing much more range out of film it would be
great to see them.

Why does this all matter? if you are thinking of giving up on film and
switching to digital you would like to really know what you can expect.
We are going through a fascinating transition between film and
digital, for a while it was the case that if were taking photographs
that were not demanding, like for a newspaper, you could live with the
lower quality photos that a digital camera produced. But now things
are changing, and in more and more cases if you want the highest
quality you shoot digital.

Scott

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

"David J. Littleboy" <davidjl@gol.com> wrote in message
news9q22o$ehv$1@nnrp.gol.com...
SNIP
> Since the sensor response is linear, a sensor with a dynamic range
> (max signal divided by noise floor) of N stops will only be able to
> respond to N stops in the scene. Unless you can adjust the _slope_
> of the linear response of the sensor.

The sensor's stored charge is essentially linearly proportional to the
light flux incident on the sensor, with a possible exception when
approaching saturation levels where sensors may exhibit 'some'
non-linearity.

> But I've never heard of a slope of other than 1 (i.e. doubling the
> signal intensity (brightness) doubles the response) in a digital
> camera. Am I wrong in this last bit here???

Before the amplification and ADC step, yes, it's a rather linear
response. Some sensor designs also allow to switch on an anti-blooming
circuitry, which will reduce true dynamic range by draining signal as
it starts approaching saturation levels.
<http://www.molecularexpressions.com/primer/digitalimaging/concepts/ccdsatandblooming.html>

In general, this may be a helpful primer on CCD technology (principles
apply mostly to CMOS devices as well):
<http://www.microscopyu.com/articles/digitalimaging/ccdintro.html>

This is a fun (with JAVA simulation) tutotial on S/N ratio:
<http://www.microscopyu.com/tutorials/java/digitalimaging/signaltonoise/index.html>


Bart

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

In message <VJcwe.17$xG2.320@news.uswest.net>,
Don Stauffer <stauffer@usfamily.net> wrote:

>The noise level is the average signal when the sensor is exposed with
>zero light, as with aperture covered by completely opaque cover. The
>values from all pixels should be RMSed, the result is the noise level.

It sounds like you're saying two different things here. First you say
the noise level is the average, then you say the average square-root
difference from the average is the noise level. The latter I can agree
with, if in fact, that is what you meant.

>In any digital camera there are several sources of noise. A test is
>better than analytically computing noise (from things like well size, etc.).

There certainly are, but the noise level is certainly not a hard limit
on usable latitude. It might serve as a point of reference, but
certainly not a concrete threshold.

--

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS@no.komm>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><