Film versus DSLR dynamic range

Started Jul 24, 2007 | Discussions
Pierre Sottas Contributing Member • Posts: 787
Re: All I was trying to say...

Jay Turberville wrote:

the reality
is that with sensors used in just about every DSLR today, the bits
do, in fact, map directly.

this is where I disagree...

But because Canon has implemented linear
mapping with a 12-bit ADC, we can only get a bit more than 11 stops
of total DR out of the sensor. 12 bits is a limiting factor when
tones are mapped linearly.

No. Because you do not know what happens during the first pre-amplification step at the pixel level BEFORE AD conversion. Mapping with a 12-bit ADC is linear but the processing before AD conversion is not necessarily linear. I gave you an example (the Kodak patent) that explains how to achieve such non-linearity at the price of a slight loss in resolution (a form of binning). There are strong pieces of evidence that Canon CMOS works the same way (search for posts from Julia Borg about binning at high ISO for example).

My point was and is that Luttman's direct mapping of of bit depth to
DR was perfectly appropriate considering that's how sensors are
currently implemented.

it is at least not appropriate for CMOS sensors.

Pierre

Jay Turberville Forum Pro • Posts: 12,917
Re: All I was trying to say...

But because Canon has implemented linear
mapping with a 12-bit ADC, we can only get a bit more than 11 stops
of total DR out of the sensor. 12 bits is a limiting factor when
tones are mapped linearly.

No. Because you do not know what happens during the first
pre-amplification step at the pixel level BEFORE AD conversion.
Mapping with a 12-bit ADC is linear but the processing before AD
conversion is not necessarily linear.

We know that either the output of the 1D-MII is linear at ISO 50, or that Roger Clark goofed up his analysis. That means that the pre-amplification pretty much has to be linear also.

http://www.clarkvision.com/imagedetail/evaluation-1d2/index.html

I gave you an example (the
Kodak patent) that explains how to achieve such non-linearity at the
price of a slight loss in resolution (a form of binning).

Yes, but no evidence that anybody is actually implementing anything like this (excepting possibly Fujifilm).

There are
strong pieces of evidence that Canon CMOS works the same way (search
for posts from Julia Borg about binning at high ISO for example).

OK, I found this thread started by Julia Borg.

http://forums.dpreview.com/forums/read.asp?forum=1021&message=21928279

And she maked the point about non-linear amplification with Canon sensors at high ISO here:

http://forums.dpreview.com/forums/read.asp?forum=1021&message=21965973

So what we have is one probable (I'm inclined to trust Julia's assertions even though she didn't give direct evidence) example of Canon doing providing non-linear processing at high ISOs.

My point was and is that Luttmann's direct mapping of of bit depth to
DR was perfectly appropriate considering that's how sensors are
currently implemented.

it is at least not appropriate for CMOS sensors.

No. It was completely appropriate. Luttmann was discussing maximum DR possible from a digital camera. What a camera does at high ISOs isn't pertinent since that is not where you get the maximum sensor DR.

Though it is interesting to consider the notion that Canon is not using linear amplification at higher ISOs.

It was also an interesting topic. Only a month or so ago I was making the same basic point in an Olympus DSLR thread where someone was showing the advantages of raw formats and highlight recovery. I pointed out that he had just shot at ISO 100 instead of ISO 200 and had used the same exposure settings, he would have had a much easier time dealing with highlights for a JPEG capture.

Julia's thread was worth the diversion.

-- hide signature --
Julia Borg Veteran Member • Posts: 7,280
Re: Bit depth and dynamic range...

It is certainly
possible to compress the tonal values of the scene that has dynamic
range more than 11 stops into this range and still have some
measurable detail in both shadows and highlights.

lens flare limits DR to 10 stops with best zooms stopped down to "sweet spot". with primes the limit is about 11 stops.

-- hide signature --

Julia

Julia Borg Veteran Member • Posts: 7,280
Re: Film versus DSLR dynamic range

Alfredo Li Pira wrote:

A number of posts in recent threads state that film has more dynamic
range than DSLR sensors. Actually, these measurements

( http://www.clarkvision.com/imagedetail/film.vs.digital.summary1.html ) clearly show the contrary, with slide film having about 5 stops dynamic range, print film about 7 and a DSLR getting to 10 stops. I would be interested in understanding on what are the statements about the low dynamic range of DSLRs based?

"slide film having about 5 stops dynamic range, print film about 7" - sounds more like a blanket statement. at what nominal ASA? exposed as what ASA? processed in what manner?

moreover, grain/noise character/structure is important to perception of dynamic range. equally important is color reproduction in shadows. measured DR may be much wider then perceived DR.

-- hide signature --

Julia

Jay Turberville Forum Pro • Posts: 12,917
Re: Bit depth and dynamic range...

It is certainly
possible to compress the tonal values of the scene that has dynamic
range more than 11 stops into this range and still have some
measurable detail in both shadows and highlights.

lens flare limits DR to 10 stops with best zooms stopped down to
"sweet spot". with primes the limit is about 11 stops.

So Julia, do you thing the non-linearity at the bottom tones in this graph is likely to be the result of lens flare?

-- hide signature --
Julia Borg Veteran Member • Posts: 7,280
Re: Bit depth and dynamic range...

do you thing the non-linearity at the bottom tones in this
graph is likely to be the result of lens flare?

IMHO to a large extent, yes. I checked with a pinhole and got a moderate improvement. on a sensor without microlenses the improvement is more pronounced. comparing output gamuts of different cameras in shadow region is very interesting. the wider is the bottom of the gamut figure the less is flare and the higher is perception of DR.

-- hide signature --

Julia

FrithjofA Contributing Member • Posts: 983
Re: No facts....someboby comfused images with original data

Dave Luttmann wrote:

A 12 bit system CANNOT discern more than 11 stops in a single
capture. Impossible.....end of story.

sorry to say that but you argument only holds for a linear mapping of a dynamical range to some number 8/10/12bit. But your argument does not hold for any monontical but non-linear mapping of real dynamical range to recorde once.

In fact, Nikon disproves you buy imploying compressed NEF images, which are not lossless. They compress the 12bits to 663 values and decompress it back to 12bits. You do not loose the dynamical range, you only loose some steps in between.

Bottom line is: what does the number represent in real life. Can the sensor detect 8 or 12 stops with one exposure. Thats relevant not the image.

Frithjof

Steve Bingham
Steve Bingham Forum Pro • Posts: 26,334
Welcome back, Julia

And, I agree.
Yeah, I make blanket statements too - and you nail me. It's tough out there. :^)

Notice the absence of qualifiers in THIS statement? Slide film differ all over the place - as does color print film - as does negative film.
--
Steve Bingham
http://www.dustylens.com
http://www.ghost-town-photography.com

 Steve Bingham's gear list:Steve Bingham's gear list
Nikon D810 Nikon D7200 Nikon AF Nikkor 85mm f/1.8D Nikon AF-S DX Nikkor 18-140mm F3.5-5.6G ED VR Nikon AF-S DX Nikkor 16-80mm F2.8-4E ED VR +20 more
Julia Borg Veteran Member • Posts: 7,280
Re: No, I am not wrong

folks tend to speak of film latitude as of exposure error that normal printing process can tolerate. latitude is usually defined for particular way of exposing and particular development process (usually it is exposing to nominal sensitivity and normal development). because of H&D characteristic curve dynamic range and latitude for film are not the same - normal printing process will not cope with fragile highlights or extrreme densities.

-- hide signature --

Julia

arachnophilia Veteran Member • Posts: 3,362
Re: Film versus DSLR dynamic range

"slide film having about 5 stops dynamic range, print film about 7" -
sounds more like a blanket statement. at what nominal ASA? exposed as
what ASA? processed in what manner?

moreover, grain/noise character/structure is important to perception
of dynamic range. equally important is color reproduction in shadows.
measured DR may be much wider then perceived DR.

technicalities are awesome.

there are so very many different ways to expose/develop/print/view film that even the SAME FILM can have dynamic range many stops different. nevermind the differences between slide/print and color/b+w.

 arachnophilia's gear list:arachnophilia's gear list
Nikon D700 Nikon D300S Nikon AF-S Nikkor 70-200mm f/2.8G ED VR Nikon AF-S Nikkor 24-70mm f/2.8G ED Nikon AF-S Nikkor 50mm f/1.8G +3 more
mcd3 Regular Member • Posts: 225
Re: No, I am not wrong

So, a mechanized approach to printing a negative (that already has a dynamic range wider than a "standard" paper) is the standard? A wide dynamic range film will allow larger errors in exposure, particularly overexposure. You may see compression due to the films shoulder, but this can usually be remedied via normal techniques available to a printer. This also will depend upon the contrast range of the subject, and the development applied.

These functions are not different. The layperson just has been told that latitude is margin for exposure error, but in the end, any exposure error is going to create a change that ideally would be corrected in development to protect highlight rendering.

So, perhaps saying that hey are the same is confusing, but they are inseparable in that one is symbiotic with the other, latitude is dependent upon the length of the straight line portion of the films h&d curve.

Julia Borg Veteran Member • Posts: 7,280
Re: No, I am not wrong

mcd3 wrote:

So, a mechanized approach to printing a negative (that already has a
dynamic range wider than a "standard" paper) is the standard?

yes. however I would not call it mechanized, same as I would not call mechanized shooting jpgs.

A wide
dynamic range film will allow larger errors in exposure, particularly
overexposure.

not funny. you at least need to specify what type of film you are talking about - negative or reversal. and even then - too generic of a statement. traditionally, dynamic range has something to do with the minimums and maximums with acceptable S/N ratio, and very indirectly - with errors (like errors in exposure).

So, perhaps saying that hey are the same is confusing, but they are
inseparable in that one is symbiotic with the other, latitude is
dependent upon the length of the straight line portion of the films
h&d curve.

latitude and dynamic range characterize different things, they are different technological parameters, and I wonder why you are trying to mix them together.

-- hide signature --

Julia

Nick Wong Contributing Member • Posts: 560
Thx Lictor

Lictor,

Your response is so knowledgeable and educating.

I've bookmarked this thread already. Thank you for your contribution.

-- hide signature --

Nick Wong

~-~-~-~-~-~-~-~-~-~-~-~-~-~
d@)), d&)ir, !).%, !@-@$, !%-%%, &)-@))
fm@, fm#a, @), #%, %), %), %, ! )

http://photos.nickwong.net

 Nick Wong's gear list:Nick Wong's gear list
Nikon D800 Leica M-Monochrom Nikon Df Fujifilm X-T1 IR
Julia Borg Veteran Member • Posts: 7,280
Re: Film versus DSLR dynamic range

Anyway, you don't want to use a
media with huge DR, because you will get a very flat image as a
result with no contrast whatsoever.

presuming one does not know postprocessing and how to compress DR there into something useful.

The advantage film still has over digital is elsewhere. Linear range
is similar, but film wins outside of that range. It's due to the
way digital and analogic react to saturation.

it is more a matter of raw conversion vs. film development. in both cases latent images are pretty linear.

When you saturate a digital device, it clips. In 8 bits, anything
above 255 is 255. That's the dreaded white hole in your blown
highlights.

when you saturate film it clips too - right to the base or to max density.

btw 255 is not used any more in processing of digital captures.

Worse, not all channels clip at the same time - causing
color shift when the highest channel clips while the other keep
responding.

same as with film if color temperature is un-balanced.

Another problem with digital is that CCD/CMOS are absolutely linear :

again, latent image is very close to linear on the film too.

twice the amount of light will cause the output of the sensor to
double.

twice the amount of light will cause twice the number of silver halogen grains to restore to silver forming latent image. H&D curve is a curve for developed film, not for exposed film.

On the other hand, analog does not clip, it compresses.

in development.

That's why to be effective, digital needs a more DR than film.

DR is DR.... to use it takes some skills.

Because you will have to sacrifice some of that DR (the part that has
a linear response) to simulate the non-linear portion of film (the
shoulder you had for highlights) and get a smoother transition from
highlights to pure white. Likewise in the shadows with the foot of
the film response.

nice theory, but seems wrong. can you explain the same using definition of DR as S/N ratio? with numerical examples?

-- hide signature --

Julia

Slough Senior Member • Posts: 1,809
Re: Film versus DSLR dynamic range

I was curious about the difference between DR and latitude, and I believe it is like this.

A film has two processes, namely exposure and development. A pre-developed film has a potential namely the information stored in the latent image i.e. the chemical reactions caused by light. Once developed, the negative (or positive) has a DR which can be measured. Development does not bring out all of the image in the latent information. Develop it for less time than usual, and you will favour highlights, but the grains in the shadows will not appear. Develop it for longer than usual, and you will bring out the shadows, but the highlights will be burnt out. In other words, latitude is a measure of how much you can push or pull the film, whereas DR is a measure of the amount of information in the processed image. For a DSLR the latitude is what you recorded in the electronic pixels i.e. in the RAW image. How you convert that to an image determines the dynamic range. Clearly you can have DR less than or equal to the RAW latitude, but not more.

Someone might wish to correct this as and where wrong.

Truman Prevatt
Truman Prevatt Veteran Member • Posts: 9,072
Re: Film versus DSLR dynamic range

That is very applicable to color films. However, with B&W film, you can modify the development to match the latitude and the DR if you want. Between different developers, different dilutions, different time and agitation and such things as water bath development you can get all the DR that the film has to offer if that is your desire.

The basis for the zone system was the old "rule of thumb" in B&W photography - expose for the shadows and develop for the highlights.

With digital you get what you get. Or "expose for the highlights and pray for the shadows."

Truman

Slough wrote:

I was curious about the difference between DR and latitude, and I
believe it is like this.

A film has two processes, namely exposure and development. A
pre-developed film has a potential namely the information stored in
the latent image i.e. the chemical reactions caused by light. Once
developed, the negative (or positive) has a DR which can be measured.
Development does not bring out all of the image in the latent
information. Develop it for less time than usual, and you will favour
highlights, but the grains in the shadows will not appear. Develop it
for longer than usual, and you will bring out the shadows, but the
highlights will be burnt out. In other words, latitude is a measure
of how much you can push or pull the film, whereas DR is a measure of
the amount of information in the processed image. For a DSLR the
latitude is what you recorded in the electronic pixels i.e. in the
RAW image. How you convert that to an image determines the dynamic
range. Clearly you can have DR less than or equal to the RAW
latitude, but not more.

Someone might wish to correct this as and where wrong.

-- hide signature --
 Truman Prevatt's gear list:Truman Prevatt's gear list
Fujifilm X-Pro2 Fujifilm X-H1 Fujifilm XF 18mm F2 R Fujifilm XF 35mm F1.4 R Fujifilm XF 55-200mm F3.5-4.8 R LM OIS +9 more
The Davinator
The Davinator Forum Pro • Posts: 22,234
Re: Film versus DSLR dynamic range

Julia Borg wrote:

Anyway, you don't want to use a
media with huge DR, because you will get a very flat image as a
result with no contrast whatsoever.

presuming one does not know postprocessing and how to compress DR
there into something useful.

The advantage film still has over digital is elsewhere. Linear range
is similar, but film wins outside of that range. It's due to the
way digital and analogic react to saturation.

it is more a matter of raw conversion vs. film development. in both
cases latent images are pretty linear.

Film is MOST CERTAINLY NOT linear. I suggest you check out the response curves of various B&W films....nothing could appear less linear. If you have ever processed your own B&W film, you'd already know this!

When you saturate a digital device, it clips. In 8 bits, anything
above 255 is 255. That's the dreaded white hole in your blown
highlights.

when you saturate film it clips too - right to the base or to max
density.

Very, Very hard to clip certain films. I can routinely overexpose Tri-X and NPS160 by 4 to 6 stops and easily retain highlights. Let's see you drag 6 stops of overexposure in a digital file back into a useable image.

btw 255 is not used any more in processing of digital captures.

Worse, not all channels clip at the same time - causing
color shift when the highest channel clips while the other keep
responding.

same as with film if color temperature is un-balanced.

Another problem with digital is that CCD/CMOS are absolutely linear :

again, latent image is very close to linear on the film too.

No, it's not.

twice the amount of light will cause the output of the sensor to
double.

twice the amount of light will cause twice the number of silver
halogen grains to restore to silver forming latent image. H&D curve
is a curve for developed film, not for exposed film.

On the other hand, analog does not clip, it compresses.

in development.

That's why to be effective, digital needs a more DR than film.

DR is DR.... to use it takes some skills.

Because you will have to sacrifice some of that DR (the part that has
a linear response) to simulate the non-linear portion of film (the
shoulder you had for highlights) and get a smoother transition from
highlights to pure white. Likewise in the shadows with the foot of
the film response.

nice theory, but seems wrong. can you explain the same using
definition of DR as S/N ratio? with numerical examples?

Actually, not wrong. It is just part of the latitude that some film has over digital capture.

-- hide signature --

Julia

 The Davinator's gear list:The Davinator's gear list
Canon EOS D30 Canon EOS 10D Nikon D2X Fujifilm X-Pro1 Fujifilm X-T1 +17 more
NikonManFlorida Forum Member • Posts: 70
Re: Film versus DSLR dynamic range

DR is not simply a function of emulsion vs. CCD/CMOS ... e.g. Fuji's S5 sensor has much greater DR than most (if not all) other digital sensors, albiet with limited resolution.

Its all relative ...

Slough Senior Member • Posts: 1,809
Re: Film versus DSLR dynamic range

Truman Prevatt wrote:
That is very applicable to color films. However, with B&W film, you
can modify the development to match the latitude and the DR if you
want. Between different developers, different dilutions, different
time and agitation and such things as water bath development you can
get all the DR that the film has to offer if that is your desire.

The basis for the zone system was the old "rule of thumb" in B&W
photography - expose for the shadows and develop for the highlights.

With digital you get what you get. Or "expose for the highlights and
pray for the shadows."

Truman

Slough wrote:

I was curious about the difference between DR and latitude, and I
believe it is like this.

A film has two processes, namely exposure and development. A
pre-developed film has a potential namely the information stored in
the latent image i.e. the chemical reactions caused by light. Once
developed, the negative (or positive) has a DR which can be measured.
Development does not bring out all of the image in the latent
information. Develop it for less time than usual, and you will favour
highlights, but the grains in the shadows will not appear. Develop it
for longer than usual, and you will bring out the shadows, but the
highlights will be burnt out. In other words, latitude is a measure
of how much you can push or pull the film, whereas DR is a measure of
the amount of information in the processed image. For a DSLR the
latitude is what you recorded in the electronic pixels i.e. in the
RAW image. How you convert that to an image determines the dynamic
range. Clearly you can have DR less than or equal to the RAW
latitude, but not more.

Someone might wish to correct this as and where wrong.

Very interesting. Thanks.

Toermalijn
Toermalijn Forum Pro • Posts: 15,860
Re: Welcome back, Julia

As does digital camera's.

Keyboard shortcuts:
FForum MMy threads