Dynamic Range - Digital sensors

[koolman]

Question: Can someone please explain why digital sensors have a much narrower dynamic range then film ? This is apparent to anyone shooting outdoors in bright light - the shadows are black and the skies blown out.

Will digic 5 improve this? Is it possible to improve it?

[PeterJ]

This article is rather interesting and seems to imply current larger pixel sensors may be able to achieve a few stops extra in the future with better analog conversion:

http://www.clarkvision.com/articles/digital.sensor.performance.summary/

Presumably the sensors themselves will also become a little lower in noise over time. I can't think of any technical reason a camera couldn't be created with lower resolution and lower frame rates but much higher DR using current technology, it probably just wouldn't be a good seller. Give it a while when fps and pixels are high enough for anyone and it might be the next battleground .

[JackSw1ss]

Question: Can someone please explain why digital sensors have a much narrower dynamic range then film ? This is apparent to anyone shooting outdoors in bright light - the shadows are black and the skies blown out.

Will digic 5 improve this? Is it possible to improve it?

i'm not a technician and not an expert of tech stuff, but I know that the range of film is about MAX 12 (EDIT: my friend which is a super duper nerd in film photography told me 12 is really the max...BW that is. whereas the color is no more than 10. Most films are around 6 to 8 like kodakchrome and such, whereas Ilford is around 7, more or less, these are not spot on numbers) while digital is around 9-11....for example, the 5DII if not mistaken is about 9 or 10 or even 12 I think, While the videocameras such as alexa, red etc etc (thus the great performance also of the new canon c300) are very similar to the DR of film, 13-15 (but not much more...but maybe i'm wrong, pick this up with tweezers)

this is all because of materials. The film (with silver...ohh geez, I don't know these terms in english LOL, anyway, the silver grains and layers, the stuff that is sensible to light in the film (which is made by several layers) is more, but not for long, sensible to light than digital)
In the sensor tech i don't knwo alot, but all is caused by the materials, pixels in particular, their sizes in compared to each others and the signal to noise ratio.

If you want I can talk about the human eyes, which I'm more prepared, but as for cameras it's everything related to the materials and electronic signals. The link above I think explains quite good the matter of this issue.

EDIT: this below is basically why the film has more DR (taken from article above) which is basically and much better explained what I was trying to tell you ahahah:

"On the internet, people argue about smaller pixels, and say, after all film has grin (single pixels) with only 1 bit of dynamic range per grain (on or off). But film grain has a 3-dimensional structure and it is grain clumps which provide tonality, not individual grains. The 3-dimensional grain distribution also gives film its characteristic curve so that once a grain has absorbed a photon it is no longer sensitive, and within a grain clump it is the probability of another grain absorbing a photon that gives a logarithmic response of the grain clump, extending its dynamic range. These are properties unlike the 2-dimensional grid In a digital camera electronic sensor. "

[torger]

"Usable dynamic range" is a bit matter of taste and subject. The measured dynamic range as you see on DxOmark is the engineering range, meaning that if a camera has 12 stops range then at 12 stops down from saturation the noise and signal is equally large. However, the useful range in a picture may be only 7 or so on that camera, the bottom stops will just be interpreted as "noise". If you have a 14 bit DA conversion than the range is 0 - 16383 and topmost stop will occupy half of that, the next half of whats left etc, so if you claim 12 stops of DR the bottommost stop has only 4 digital levels (0 - 3) to represent the signal. Not much information there. Still it does not help to extend to a 16 bit DA in general, since the analog noise is typically larger than the digital quantization noise.

The film advantage as far as I can understand it, is that it will compress highlights rather than sharply clip them (I recently saw a Sony patent on such a feature on a digital sensor, so perhaps this comes to digital too) meaning that you can compress highlights directly when shooting, allowing for more shadow space.

How large dynamic range film actually has I don't really know, someone has a number? It depends on film and stuff I guess.

[Meh]

Where does this notion that film has more DR than digital sensors come from?  Ken Rockwell who only shoots jpeg?

It might have been true 10 years ago for some mass market sensors but even then it wasn't significantly more.   Digital sensors produced anytime in the past 5 years have more DR than film.   The response curves are different and the fall off into highlights and shadows is different and to some it's less "pleasing" with digital but digital technically has more DR, which is why you hear talk of "useable DR".   

However, if you're looking at jpegs it's a different story because they are only 8-bit files so a transfer function has to be applied to fit 4096 luminance levels (14-bit) into only 256 luminance levels and highlights and shadows are typically where there is more compression applied.   Keep in mind that LCD displays and prints have much less DR than the original RAW files or film.

I'm not an expert in this and not the best at explaining it so if anyone has any corrections or additional details feel free to correct me.

[gferdinandsen]

Question: Can someone please explain why digital sensors have a much narrower dynamic range then film ? This is apparent to anyone shooting outdoors in bright light - the shadows are black and the skies blown out.

Will digic 5 improve this? Is it possible to improve it?

E6 only has DR of 3.5 stops or so, so I consider digital a Godsend when it comes to DR.  Not to mention how much more DR you can pull out of ACR.

[JackSw1ss]

Where does this notion that film has more DR than digital sensors come from?  Ken Rockwell who only shoots jpeg?

Checkout the Zacuto clips on youtube of them comparing the video output of the 1D4/5D2/7D with that of Kodak film. The kodak film has 14.2 & 14.5 stops of DR whereas the 5D has 11.2 but usable is probably closer to 10. Episode 1 of the Great Shootout of 2011 covers DR.

Now I don't know if all film has the same DR as the Kodak film used in that "shootout" but it is a place to start.

http://www.youtube.com/watch?v=k-hqzw5MEa8&feature=channel_video_title

that kodak is motion picture film...not photography film
but still, look at the DR of it.

[JackSw1ss]

E6 only has DR of 3.5 stops or so, so I consider digital a Godsend when it comes to DR.  Not to mention how much more DR you can pull out of ACR.

that is because E6 is the method to process a positive (EDIT:oooopsy daaaisy...a negative, not a positive...the positive is the slide) and obtain a slide (positive)...so slides as well as prints are always lower as DR.
Prints DR is usually in between of 5-7 DR range whereas slides are 3-5 maaax (velvia)

[Meh]

Where does this notion that film has more DR than digital sensors come from?  Ken Rockwell who only shoots jpeg?

Checkout the Zacuto clips on youtube of them comparing the video output of the 1D4/5D2/7D with that of Kodak film. The kodak film has 14.2 & 14.5 stops of DR whereas the 5D has 11.2 but usable is probably closer to 10. Episode 1 of the Great Shootout of 2011 covers DR.

Now I don't know if all film has the same DR as the Kodak film used in that "shootout" but it is a place to start.

http://www.youtube.com/watch?v=k-hqzw5MEa8&feature=channel_video_title

that kodak is motion picture film...not photography film
but still, look at the DR of it.

Right.  My understanding (albeit limited and prone to being wrong!) is that motion picture film is designed differently and optimized for things like DR rather than resolution which is far less important when there's motion... for example in digital video 1080p (2 megapixel) is high-res and 4k (~9 megapixel) is super-high-res.  The transfer functions used in the movie-land are also different.  Consider also that the DSLR video is an h.264 output.

[NormanBates]

here's a link to the zacuto DR tests:
http://nofilmschool.com/2011/06/zacuto-unleashes-great-camera-shootout-2011/

yes, that film is used for movies, not stills, and it is one of the best you could get, but still: see how much better does on the highlights (not so much on the shadows)

also, the DSLRs were tested in video mode, before technicolor cinestyle but set up to maximize their DR, so for stills I guess the DR would be at least a stop higher

[awinphoto]

From what I remember... BW negative film had natively around 7-8 stops of range, you were able to push/pull the film in processing to get some more range, but maybe a stop or two more... Color Slide ISO 50 film you were looking closer to 11 stops... Paper only had closer to 6-7 depending on what paper you were using, RC or standard.  Zone photography allowed you to make the most out of your film/processing/paper to get the absolute most out of your DR.  Digital, other than maybe HDR, which may or may not be perceptually loved, is the closest photographers have to Zone Photography... 

Whether or not digital has more or less DR is dependent on skills, exposure and processing as well as if this is just going to be viewed on screen or printed, if printed, what printer, inks, paper combinations you are using.  In Film we exposed for the shadows, printed for the highlights... digital, we got to do the reverse, expose for the highlights, print for the shadows (just like you would if shooting slide film, you can boost up shadows, you cannot regain lost highlights).  Most people forget this technique and become somewhat lazy.  I think cameras are starting to get to the detail quality that we expected from film, but you will need to really learn exposure to get the most DR Possible.  It is not just camera dependent, it's so much more. 

[awinphoto]

Question: Can someone please explain why digital sensors have a much narrower dynamic range then film ? This is apparent to anyone shooting outdoors in bright light - the shadows are black and the skies blown out.

Will digic 5 improve this? Is it possible to improve it?

The DR that an expert user can achieve using aggressive eposure or exposure bracketing together with optimal development of raw files is quite different from what the average user gets with AE and jpegs.

For the expert/nerd, the main limiting factor is the camera sensor. Digital sensors are basically linear with an obvious saturation point and a noise-floor. Film is (I believe) very different, with smooth roll-offs in high-lights and in shadows. Using one definition one might claim that film has a lot of DR. Using another definition, one might claim that film does not give an _accurate_ recording of the scene brightness over such a large DR. In the end, it usually our eyes that are the judge, and if you prefer the look of film I can only suggest to keep using that.

I do believe that many comparisions of film vs digital is unfair. I.e. digital is compared to film on films terms. For the expert user, this might not matter as she will choose based on her own knowledge. For novices, this could lead to false conclusions.

-h

While I dont consider myself a nerd, Film had sort of a soft S curve in the DR and gradual decline in the shadows/highlights, but just like digital, It was a tad more forgiving in the highlights in negative film (film was essentially transparent (negative) once processed and the exposure burned the image onto it... If you over expose or underexpose maybe 1 stop, you could always compensate by over/under developing...

People give digital too much credit... Film, you had to expose properly (same as digital), process it (raw conversion to photoshop/lightroom/aperture), and in most situations, print with contrast filter, dodge/burn, manipluate, etc (post-processing in photoshop, lightroom/aperture)... Just to take a image taken from digital and scanning a slide film and measuring ranges, I'm sure they would be quite comparable... but also then you have to take in to account processing and post processing... All that will affect visible DR and needs to be taken into account when comparing oldschool zone prints.  There's a lot of stuff to take into account

[archangelrichard]

Sorry, NO!

NO!

A million times NO!

Your eyes only do 13 f-stops

film has been measured at 10 - 11 f stops (film is analog, it can have measures like 88.73; digital can not)

digital is usually estimated at 4 f-stops - here's why: It has to do with the sensor. Basic sensors are strictly on - off (sensors are derived from EEPROM - Electronically Erasable Programmable Read Only Memory - which someone discovered by accident could not just be erased but selectively erased forming images!)

Your 14-bit conversions has to have COLOR information as well as brightness, even if the colors are faked (as all digital colors short of a foveon are) so lets look at it this way: 14 bits is 16,384 discreet color / brightness / etc. possible. that is the TOTAL possible (16 bit would give you 65,536 possible, etc.)

Again, film does not have this issue; it can register a near infinite number of possible colors (your eyes often can not detect them) In fact, film's issue is LOSING Dynamic Range - it loses through the f-stop of the lens, the age of the film; etc. so it loses a total of 2 - 3 stops from what our eyes can see; Digital starts from the ground up with the basic sensor seeing only off and on

Can we make better sensors? The first problem is how long it takes to deal with all this information and get ready to do it again - 14 bits is a handful right now; just imagine opening up to 16 bits as the next step, just how much data is that

Then there is the accuracy of the 14 bits (then expanding to 16) - can we even make sensors that can see 1/2 again as little light as we do now (just to get 1 stop difference) with any accuracy (you must see ALL the colors at that level of darkness - and any computer hardware "geek" will tell you that the problem is that your dark blue, green, red, etc. all are not distinguishable from black on computers, the same issue holds for cameras; there is a color there but it is too dark to see it.

There is a "law of diminishing returns" going on here, as people no longer view images on reflective screen (slides, movies) or paper (prints) we are in an age where the limiting factor is screen accuracy and tolerance for detail - what is your screen;s real resolution and it's color response at that resolution and why should camera makers make images that have detail differences you can't see? Your eyes arfe estimated to make out only 10 million colors, so why try to process many more than that?)

On both sides of this page, at the top, are a gradient in Red. In theory, your display is likely 32 bit (24 bits of color plus other information - cheaper ones use 18 bits of color information dithered to fake the rest) but ask yourself - How many colors do you see?

Camera manufacturers are going to the use of "tricks" like HDR - faking the Dynamic Range that the sensors can not see and the result is often cartoonish for the obvious reason that you are attempting to use the higher bright colors from one exposure, the middle from another and the darker from a third and combine them in an unnatural manner to "stretch" them to fit (faking adding more bits of data by eliminating some thereby losing the transitions - think of those gradients being stripes instead so you stretch the "range" by eliminating all the even bit colors)

Can we have better sensors? Got $2 Million for a camera? (not to mention the size increasing, weight increasing, FPS dropping, etc.)

Wiki has a nice article here http://en.wikipedia.org/wiki/Color_depth on color depth that may explain the technicals better; but the point is this: there is a natural reason why digital has, and always will have, lower Dynamic Range than film has - it is in the nature of the beast. Digital is far more a daytime media than film - that is also in the nature of the beast.

Rather than bemoaning this fact, people should learn to deal with what they have, to constructively use digital or film for the effects they want

[Meh]

@archangelrichard... What?

Digital sensors have only 4-stops of DR:  Wrong, why do you think that exactly?

Each 14-bit conversion has color and brightness:  Wrong, each photosite is only measured for luminance levels, color is interpolated after the fact by looking at the luminance of adjacent RGB pixels.

Sensors are strictly on or off:  Wrong, each photosite builds up a voltage as photons are absorbed and converted into electrons.  The voltage is measured and encoded by the Analog-to-Digital converter into a Data Number (DN) that represents the luminance level of the photosite.

Accuracy of the 14 bits:  What do you mean?

[Mt Spokane Photography]

I recently scanned many hundreds of old negatives from my parents estate taken mostly from the early1940's thru the 1950's.  There were a few overexposed and under exposed images that were often recovered nicely, it was amazing how much latitude for over or under exposure there was.  They were not taken with a expensive camera, just a ordinary Kodak camera of the times, but they were great.

When I got to the 35mm images of the 1960's and later, they were usually color, and far inferior to the older medium format images, noticibly inferior to todays DSLR images.  In the 1960's, I have lots of images I took with better quality cameras, a Yashica TLR, Argus C3, Canon FT,  mostly using Kodachrome II or a few with print film, and while these are fairly good, they still do not match those old 1940's B&W negatives from 120 film.

Still, overall, a ordinary point and shoot camera would blow away the IQ of those old images if I were to enlarge them to even 8 X 10.  We definitely have been spoiled by the capabilities of the new technology for ordinary home photos.