Dynamic Range War

[unfocused]

Okay, I've noticed a lot of discussion (to put it politely) in other threads about the "dynamic range" of the new 5D III sensor. I'm hoping someone can enlighten me a bit and explain why or if I should care.

I'm not clear what exactly people mean by dynamic range. It seems like at least two definitions are possible.

1) Are you referring to the ability of the sensor to record detail in a scene that has a wide range of light. For example, are we talking about the ability to capture detail in a brightly lit canyon, where the light ranges from near total sunlight to near black. So that, a sensor that has a dynamic range of say "9" would be able to record detail for up to four stops from the midpoint in either direction?

2) Or, are you referring to the ability of the sensor to record discernible differences in light. For example, a range of "9" would mean that on a scale from black to white, there would be nine clear steps visible?

It's been many years since I read the Zone System (and frankly, I found the books excruciatingly boring), but as I recall Adams' basic premise was that film was capable of recording far greater dynamic range than could be reproduced by photographic paper (much less commercial printing). By manipulating exposure and development of the film, he sought to compress the dynamic range recorded by the film, so that it could be aligned with what the final print could reproduce. The general concept, as I recall, was to expose to retain some detail in the shadows and then develop to retain detail in the highlights.

My understanding is that photographic prints even today have less possible range than sensors and computer monitors less than prints. (Although the back lighting of monitors gives the appearance of greater saturation and richness in colors)

So, if I am wrong about this, can someone explain it in understandable terms. And, if I am right, then why should I care at all about dynamic range so long as the final medium is always going to be more limited than the medium used to capture the image in the first place?

[dtaylor]

1) Are you referring to the ability of the sensor to record detail in a scene that has a wide range of light. For example, are we talking about the ability to capture detail in a brightly lit canyon, where the light ranges from near total sunlight to near black. So that, a sensor that has a dynamic range of say "9" would be able to record detail for up to four stops from the midpoint in either direction?

This is what people mean when they say dynamic range. Though the mid point, i.e. the point where a gray tone is rendered middle gray, is not necessarily in the middle of the range. Digital sensors typically have more shadow range than highlight range, and print film typically has the opposite.

My understanding is that photographic prints even today have less possible range than sensors and computer monitors less than prints. (Although the back lighting of monitors gives the appearance of greater saturation and richness in colors)

Monitors have more DR than prints. Some may even exceed sensors.

And, if I am right, then why should I care at all about dynamic range so long as the final medium is always going to be more limited than the medium used to capture the image in the first place?

Because you can compress the captured range into a range that will fit on paper, and your viewer can see the shadow and highlight detail you saw at the scene. More DR also covers more exposure errors.

On that note...I don't know why anybody is talking about 5D3 DR yet. To my knowledge there are no published transmission step wedge tests of the 5D3. DPR is the site which usually does this first. You can safely ignore any and all claims based on noise measurements (i.e. DxO and personal estimates made from available RAW files). Trying to compute DR from the noise floor does NOT result in an accurate measurement of sensor DR in the real world.

[awinphoto]

Hey unfocused... your definitions are both correct to a discernible degree... Yes, the downfall in photography, DR range was/is printing... film had a few more stops than photographic paper back in the day... I think film, especially negative film, had closer to 9 stops give or take (slide film had anywhere from 11-13 if i'm not mistaken but it's been quite a few years).  Photographic paper at the time had like 5-6 stops of DR, so the zone method was created to leverage the developing process, exposure process, and printing to get the most maximum range out of papers limitations.  It was an entire class in itself... a lot of math and testing and experimenting. 

Now with digital, we are once again limited by CMYK and printing... The RGB color gamut is vastly wider than CMYK and commercial printing has not been able to catch up and there is really nothing you can do about it without adding spot colors and such... Modern epson and HP printers have even added Red inks and Orange inks and i think even blue, not to mention light cyan, light magenta, etc... all trying to get the widest gamut possible.  Commercial printers have yet to catch on to do something similar without really adding to the cost of production.

Now regarding dynamic range, it is the level of stops and subtitles that you see throughout the entire range of the print...  DR is best seen as an S curve... most cameras can get most of the middle ranges but struggle getting the information in the extreme highlights and shadows in which the DR range gets elongated and such.  The 5d2, for example had a DR, depending on the testing source, of around 11-12 stops of DR.  I would guesstimate most consumer digital cameras out there on the market today should capture around 9-10 stops of DR easily if not more.  Now the question remains, if nikon/canon/sony/phase one, etc develop a sensor that could capture 13,14,15 stops of DR, most of it, as mentioned above would be in the subtleties in the highlights and shadows, but whether any would show up in print is another thing.  Of course with the increasing development of digital frames, projectors, HD monitors, etc... you can make a good presentation, but for professional photographers delivering paper prints to clients, until not only cameras but printers/ink/output/cmyk development continues as well, it really is a futile argument.

Edit... epson a few years ago came out with the R2800 which was supposed to give more DR by having multiple black/gray inks to give more definitions in subtle tones, as well as give cleaner B&W prints without any color tinting that was prevalent in standard B&W inkjet printing... While successful, It also struggled in the color part because where they added in blacks and gray inks, they took away in color inks to compensate... now they have the R2880 and the R3000 which is trying to blend the technologies... It's still a work in progress however these printer companies update their printers almost less frequently than the 1d series does and commercial printing really hasn't changed a whole lot in the last half decade. 

[Larry]

if I am right, then why should I care at all about dynamic range so long as the final medium is always going to be more limited than the medium used to capture the image in the first place?

Despite whatever limitations of the final medium, if there are discernible differences in the print that result from DR differences in the sensors, most of us would care.

The old, out of tune piano with some sticking keys will never deliver the whole tune, but 10 fingers will still do a better job than 5 ;-)

[epsiloneri]

Dynamic range is the ratio between the brightest signal you can reliably detect, to the faintest signal you can reliably detect. This is close to your definition 1. Your definition 2 is closer to the numerical dynamic range, but is not necessarily related to the photographic dynamic range (because with DR = brightest/darkest the number of steps in between don't matter).

BTW, the human eye has a DR of about 30 stops (though not simultaneously). It's fortunate reproductions don't have that kind of DR, as it would easily be quite uncomfortable to watch. Just imagine actually being blinded by a photograph of the Sun...

[Mt Spokane Photography]

This image shows why we would like more dynamic range in a camera.  I took this on a bright day, and the camera exposed for the bright sky.  I then switched to manual exposure and exposed for the person, but kept the image as a example.

A higher dynamic range might let you get good detail from both the bright and the shadow areas, without enough dynamic range, you must chose the area you want to be able to show the detail.

Black and white film typically had very good dynamic range, and you routinely got good prints from a bit of under or over exposure.  Digital is not so forgiving, its because there is less dynamic range.

Most of the discussion involves how much range you can theoritically extract from a raw image, and is not necessarily related to the quality of the final image.

[awinphoto]

Black and white film typically had very good dynamic range, and you routinely got good prints from a bit of under or over exposure.  Digital is not so forgiving, its because there is less dynamic range.

Only if you knew you screwed your exposure in the beginning and developed your film to push/pull times to compensate... Otherwise you were left with a muddy mess... Also DR may or may not also help your image... much like epsiloneri said your eye has a huge DR, but what your eye does is compensate on the fly... so in your scene, you see the sky, your focus of your subject goes oof and the sky slightly darkens to reveal the detail in the clouds, sky, etc... then you focus on your subject, it dilates, lets more light in, your subject appears normal, and then you lower your vision, look in shadow your eye see the shadow detail fine... It does all this by not focusing/processing all the information at the same time and focusing on everything at the same time... Kinda like walking indoors and outdoors or when outside, looking in the distance trying to discern something in the distance, squinting forces your eye to focus exactly on said subject and your eye recacluates the light so you can see it.  Our eyes is the most advanced camera/lens system ever created and we mostly dont even think of it...  So in theory, if you nailed exposure on the guy, the shadows would have been brought up nicely and you would get white skys... increased DR may give you nicer shadow and maybe discern a cloud shadow or two but probably couldn't make this a perfect photo unless we get the same DR as our eyes. 

[dtaylor]

For the record...

* Slide film has 5-8 stops depending on emulsion.

* Early generation DSLRs were around 8 stops.

* Current DSLRs are in the 10-12 stop range. (Note: the newest FF bodies from Canon and Nikon haven't been tested yet.)

* Print film has 9-14 stops depending on emulsion.

* I've seen some B&W emulsions, when properly processed, yield 18 stops.

[awinphoto]

For the record...

* Slide film has 5-8 stops depending on emulsion.

* Early generation DSLRs were around 8 stops.

* Current DSLRs are in the 10-12 stop range. (Note: the newest FF bodies from Canon and Nikon haven't been tested yet.)

* Print film has 9-14 stops depending on emulsion.

* I've seen some B&W emulsions, when properly processed, yield 18 stops.

I think you got your print film and slide film mixed... Slide film typically had much wider DR than color negatives... So much so that many pro's who shot color would shoot slides, and then try reexposing the slides on negative film when needed to print.  But they were two separate beasts... Negatives you exposed for shadows, print for highlights, slides, you had to expose slightly for highlights.  B&W film had maybe upwards to 9 or so when we did our denistometer tests, but it really was dependent on ISO and brand. 

[qwerty]

As neuroanatomist pointed out in another thread, people (myself included) always want a little bit more.  However, I really was expecting a boost in DR with the new sensor.  I do hope that the people reporting a lack of significant improvement are wrong (and they generally list ways their analysis might be off); however it does have me a little concerned.

If you look at DxoMark scores for dynamic range (see http://www.dxomark.com/index.php/Cameras/Compare-Camera-Sensors/Compare-cameras-side-by-side/(appareil1)/680%7C0/(brand)/Nikon/(appareil2)/485%7C0/(brand2)/Nikon/(appareil3)/483%7C0/(brand3)/Canon ; click on measurements, the dynamic range), you can see that the 5dII dynamic range asymptotes at low ISO while the DR for the two Nikon sensors keeps improving as you reduce ISO.

I have some background in statistics, but know nothing about sensor design.  However, I did expect that Canon would have an amazing low-iso dynamic range with their latest generation of sensors, if only to keep up with Nikon.  Right now, Nikon's full frame sensor from 2008 and crop sensor from 2010 beat Canon's best full frame sensor by a wide margin (almost 2 full stops). 

I am not very knowledgeable about film photography, but my recollection is that, beyond the reported dynamic range, film is more forgiving than digital; with digital, if you blow a highlight, it is blown completely and utterly beyond recovery (your Spinal Tap brand amp won't go to 12, no matter how hard you try).  With film, since it is analogue, it is a more gradual process; it becomes progressively harder to distinguish between different highlight areas, but there is still some minuscule difference.


I really hope that the people who are claiming no such improvement for the 5D III are wrong.  My photographs will certainly be limited by my skill and not my camera's DR, but I would (rationally or not) feel better about purchasing a 5D III if I felt like it was optimal in every way.

[awinphoto]

As neuroanatomist pointed out in another thread, people (myself included) always want a little bit more.  However, I really was expecting a boost in DR with the new sensor.  I do hope that the people reporting a lack of significant improvement are wrong (and they generally list ways their analysis might be off); however it does have me a little concerned.

If you look at DxoMark scores for dynamic range (see http://www.dxomark.com/index.php/Cameras/Compare-Camera-Sensors/Compare-cameras-side-by-side/(appareil1)/680%7C0/(brand)/Nikon/(appareil2)/485%7C0/(brand2)/Nikon/(appareil3)/483%7C0/(brand3)/Canon ; click on measurements, the dynamic range), you can see that the 5dII dynamic range asymptotes at low ISO while the DR for the two Nikon sensors keeps improving as you reduce ISO.

I have some background in statistics, but know nothing about sensor design.  However, I did expect that Canon would have an amazing low-iso dynamic range with their latest generation of sensors, if only to keep up with Nikon.  Right now, Nikon's full frame sensor from 2008 and crop sensor from 2010 beat Canon's best full frame sensor by a wide margin (almost 2 full stops). 

I am not very knowledgeable about film photography, but my recollection is that, beyond the reported dynamic range, film is more forgiving than digital; with digital, if you blow a highlight, it is blown completely and utterly beyond recovery (your Spinal Tap brand amp won't go to 12, no matter how hard you try).  With film, since it is analogue, it is a more gradual process; it becomes progressively harder to distinguish between different highlight areas, but there is still some minuscule difference.


I really hope that the people who are claiming no such improvement for the 5D III are wrong.  My photographs will certainly be limited by my skill and not my camera's DR, but I would (rationally or not) feel better about purchasing a 5D III if I felt like it was optimal in every way.

While I too hope to see improvement in the new 5d3... if it helps in any way, it also has that cool HDR function... from what I could tell in reviews you could set the ranges as little to as far apart as you want and it will either do it in camera or save the individual files for you to do it later based on your preference, and got several different settings...  If it's as good as advertised, it could be one way to really boost DR in your images without it looking fake.  I'm excited to give it a whirl. 

[unfocused]

Thanks all. This is a very helpful and reasonable discussion.

Now, to flog the horse's corpse just a bit more: looking at Mt. Spokane's example, what I have a hard time wrapping my head around is that it seems to me the problem is not that the sensor fails to record enough dynamic range, but rather the sensor does not know how to compress the range that exists in the scene.

It still strikes me that the problem is that the distance between the various tones from shadow to sky needs to be narrowed. Granted, this particular image is not properly exposed for the skin tones, but if it were, the sky would be blown out. Is that a problem with too little dynamic range? Or is it more correctly a problem with too much dynamic range?

Finally, I would just add that in this example, my first reaction is to see an image that would look about the same if shot with slide film. We've always had a problem with these kind of lighting conditions and the old way of dealing with it was to change the conditions by shifting position or moving the subject.

Again, I appreciate the responses here. Candidly, my non technical take on all this is that the improvements that we are talking about with modern sensor technology are primarily around the margins. Whether it is ISO, noise, dynamic range, etc. etc., it sure seems to this old film photographer that we are light years ahead of where we used to be.

[neuroanatomist]

Now, to flog the horse's corpse just a bit more: looking at Mt. Spokane's example, what I have a hard time wrapping my head around is that it seems to me the problem is not that the sensor fails to record enough dynamic range, but rather the sensor does not know how to compress the range that exists in the scene.

In order to compress it, the sensor must first record it - and that's the problem. 

Consider your eyes - they have what amounts to a situationally flexible dynamic range.  When you are in bright sunlight, but then walk into a nearly pitch-black room, at first you can't see anything - but after a while, your eyes accomodate and you can see...until you walk out into the sunlight again, which monentarily blinds you (although you adjust more rapidly in that direction). 

A sensor does not have that flexibility - it's dynamic range is fixed.  You can adjust the exposure to capture a higher or lower portion of the scene's total dynamic range, but you're limited.  Imagine a scene with 13 stops of dynamic range in it from deepest shadow to brightest highlight.  If your sensor has 11 stops of DR, you can choose to set the exposure so you capture stops 1-11 of the scene and blow two stops of highlights, or capture stops 3-13 of the scene and lose two stops of shadow detail.  IF you had a sensor with a 13-stop DR, you could capture the whole range of the scene in one shot. 

Yes, output media are often more limiting (and JPG images are limited to 8-bits) - but the point is, if you capture the full tonality of the scene in the RAW file, you can then choose which portions to keep, or compress the tonal range of the scene, as desired.  But if your sensor doesn't have a sufficiently large DR to capture lowest lows to highest highs in the first place, those data are gone and cannot be recovered later.

[unfocused]

Thanks. Got it.

http://www.facebook.com/photo.php?fbid=2010338070756&set=a.2010337630745.95914.1612860769&type=3&theater

Not sure this will work, but if it does, you may see an example of why I think we are pretty much just dealing with the margins. Shot of the amphitheater in Arles. I wanted the highlights to blow out and the shadows to go near black. Not sure I would have liked the effect otherwise.

It's quite mind-boggling to stand there and realize that two thousand years ago, people were walking through these same halls on their way to their seats, where they were going to watch gladiators fight to the death.  Anyway, I really do appreciate everyone's comments. This is why I enjoy this forum.

[dtaylor]

I think you got your print film and slide film mixed... Slide film typically had much wider DR than color negatives... So much so that many pro's who shot color would shoot slides, and then try reexposing the slides on negative film when needed to print. 

Slide film never had wider DR than color negative. Check the characteristic curves published by manufacturers. The best slide film I'm aware of, in terms of DR, was Astia at roughly 8 stops. That's where print films began.

The only reason slides were ever shot to negatives and then printed is because you can't directly print a slide with standard darkroom printing processes. The resulting print would be a negative of the slide.