Dynamic range of current Canon sensors

[kanellopoulos]

Slide film, I have read more than twice, has a latitude of about five

f-stops. How do current Canon sensors compare to that? Has anybody

experimented?

 

BTW, I plan to buy one or two ND grad filters from Singh-Ray for use

with my 20D (and hoping for a FF future). Apart from the 3 f-stop hard

one, which else would you add?

[conraderb]

Nick - I think that many people would say that 5 stops is more like a bad print film, and that slide film has quite a bit more 7-8. But hey - I'm not an engineer.

 

I do know that many wedding photographers are happy with the dynamic range that they get from their 20Ds, D70's, etc, etc. What kind of work do you shoot? Landscapes, I presume?

[dave_trayers]

Maybe closer to 8 stops?

<p>

<a href="http://www.dpreview.com/reviews/canoneos5d/page23.asp">http://www.dpreview.com/reviews/canoneos5d/page23.asp</a>

[peterlyons]

Hey Nick, just a thought... if you're using Photoshop CS2 and shooting static subjects, the HDR (High Dynamic Range) feature could work out as a alternative to those fine but expensive Singh-Ray ND filters.

[kanellopoulos]

Yes, my main interest is landscapes.

One of the well known photographers who wrote about 5 f-stops is John Shaw. Perhaps his text is now outdated. I don't know.

[mcgarity]

I don't know what sensor you guys are using but my 300D certainly doesn't have anywhere near 7 to 8 stops of latitude. With Jpegs it has about 4 1/2 to 5 stops. You can get more by shooting in RAW. But to do that you have to decode the RAW file twice and combine the images in post processing. And if you are going to do that you get better results if you bracket your exposures and combine two shots instead of trying to use just one.

[mcgarity]

And incidently combining bracketed shots in post processing is a better way to go than using graduated neutral density filters. Its more precise and is actually a heck of a lot easier than fiddling around with the filters.

[kanellopoulos]

Yes Larry, but you need a tripod. And where I go, it is not always easy to carry one. In fact, it would be impossible to keep pace with my friends.

[mendonphoto]

...a VERY sturdy tripod and absolutely zero subject movement in the frame. ND-Grads don't work for everything, but when they do, it sure beats compositing in PS.

 

Slide film does indeed have about 5 stops. DSLR sensors have more and it's been tested and verified time and time again. Jpeg images are limited because of the file format, not the sensor. There are several RAW image processors that will let you get more than 5 stops out of a good RAW file. Heck, the default Canon processor will do it - just save your files as 16-bit tiffs and process in PS.

[mark_chappell]

<I>Slide film, I have read more than twice, has a latitude of about five f-stops. How do

current Canon sensors compare to that? Has anybody experimented?</i><P>

 

I suppose it comes down to how you define 'latitude' but no slide film I used in 30+ years

had 5 useful f-stops of exposure latitude. More like 3. My DSLRs (Canon 1DII, 10D) have

perhaps 5 stops and you can squeeze out a bit more in photoshop or the like. DSLRs are

definitely better than slide film in this respect. I've not used print film to any degree, but I

think Conrad Erb's got it backwards: print film has more useful dynamic range than slide

film -- or so say many people with more experience with it than me.<P>

 

I find my 3-stop 'hard' grad ND to be very useful for landscapes under lots of

circumstances. Less commonly, my 2 stop soft-edge gets used. Both are Singh-Ray.

[jack_rabin]

For analyses using Canon gear, go to NormanKoren.com or Clarkvision.com.

Both have done a nice job answering this question.

Depending on Canon model, you are in neighborhood of 8.5 stops. The

Canon 1-series bodies, with larger sensors and photosites, may even

approach 10 stops.

 

But, you can't really "capture" that f/stop range in hard contrasty outdoor

lighting - real lit scenes - because the highlights would overexpose before the

shadows unblock. Useful range is likely between 6-7.5 stops on the APS-C

sensors and maybe 7.5-8 on 1-D bodies.

 

And, if you did capture it, you could only bring it out if you shoot RAW 12-bit

and process into 16-bit TIF (certainly not JPG). And finally, you could not

really print that many stops, nor see them on most modest monitors.

 

I have satisfied myself that this range is real though by doing tripod shots of

complex scenes in soft outdoor light. The recovery of useful shadow detailis is

an amazing thing. Ceetainly killed slide film for me.

 

Up until a couple years ago web forums were full of concerned folks claiming

digital only had 5 to 5.5 effective stops exposure range. This was because

there was (and remains) trouble holding highlight detail in hard contrasty light

with digital, just like slide film. With digital it's due to linear gain. Metering

strategies and circuitry advances have improved on this. But that is different

than overall contrast range, dyanamic range, exposure latitude, whatever one

terms it.

 

http://postit.rutgers.edu/uploads/Peach%2Ejpg

Ignore ugly grass in background...

All I know is with slides there was NO way I could capture black of the box,

maintaining highly saturated red highlight details of the peaches, and not

having unnatural Velvia look (I must be only person who doesn't like Velvia's

fake saturated color).

Jack

[ky2]

<i>"Depending on Canon model, you are in neighborhood of 8.5 stops. The Canon 1-series bodies, with larger sensors and photosites, may even approach 10 stops."</i><br><br>

I love the internet. Everyone can post something that can later be quoted and glorified upon. HOWEVER, either these guys don't know what dynamic range is, their test is flawed, or they've been exposed to euphoria-infusing substances. My experience with DSLRs (RAW only) is that sensor latitude is about the same a slide film, with highlight response better handled by slides. This may not what digital fans want to hear... but im not a Reichmann or an Askey.

[oceanphysics]

Most of you don't seem to have any idea what latitude is. It is not the same thing as dynamic range. It's basically the amount your exposure can be off and still be acceptable. It's subjective, but slide film has essentially no latitude. Maybe half a stop. Print film has a stop or two, depending on whether it's over or under, how picky you are, the subject involved, etc. Digital has almost no latitude in the overexposure direction for most subjects.

[chris_floden]

EOS 1 series Canon sensors have a sensitivity of 6 stops, or "Zones" if you're an Ansel Adams follower. The 20D is close to 5 stops or Zones.

[andrew robertson]

12 stops. 12 bits = 12 stops. It's that simple (with RAW anyway). Feel free to experiment

for yourself. The bottom 4 bits or so are pretty 'shallow', so the color is very rough and

there isn't tons of detail. 8 'good' stops is easily doable, especially with the proper kind of

RAW conversion. You'll never get even 8 stops onto paper without compressing the heel

and toe, though, so I don't think dynamic range is as critical as many other people.

[mcgarity]

12 stops? Right. I have a bridge I'm selling. You interested?

[mike dixon]

<i>12 stops. 12 bits = 12 stops. It's that simple (with RAW anyway). </i><P>

No. It means there are potentially 4,096 (2 to the 12th power) levels of gradation between the "darkest" and "lightest" info for that channel; it doesn't mean that the "lightest" that the sensor can capture is 4,096 times brighter than the "darkest" it can capture.

[mark_chappell]

What Mike Dixon says.

[andrew robertson]

Mike, I would recommend experimentation as you don't seem to believe me.

 

With RAW conversion, I have been able to 'save' images that were underexposed up to 8

stops, and retain detail if not tonality. Go ahead, try it. Put your camera in M mode, meter

something, and take a picture. Then take another one that is 2 stops under. Do it again.

Then do it a fourth time. If you're using a 10D like I am, you may even be able to go

deeper than 8 stops under medium exposure.

 

Getting an image with really wide latitude often involves making two or more

developments of the same RAW file and then merging them later. This doesn't work at all

with JPGs - here you're limited to about 5-6 stops.

[andrew robertson]

Larry, as you doubt me I don't think I could ever convince you. You should try it yourself at

home, following the procedure I have laid out above. Then YOU can buy ME a bridge.

[boris c hann]

"Slide film, I have read more than twice, has a latitude of about five f-stops. How do

current Canon sensors compare to that?"

 

The absolute dynamic range of a Canon sensor is narrower than for E6 film, but (and it's a

big but) the usable RAW exposure range is considerably better than for E6.

[dave chew]

To the second part of your question: I have all four of the standard "Galen Rowell" series: 3S Hard, 3S Soft, 2S Hard, and 2S Soft. The 3-stop hard and 2-stop soft are the two I use the most. These are the only two I take if travelling light. These are also the two Galen would recommend whenever workshop students didn't want to plunk down $400 for all four.

[mcgarity]

"Getting an image with really wide latitude often involves making two or more developments of the same RAW file and then merging them later"

 

Andrew I made this very point at the beginning of this thread. I do this all the time. But a usable 12 stop spread from one exposure? I have been down this road a time or two and my own personal experience doesn't line up with that. I have found that when something is badly underexposed, even if I recover an image of some kind, its of such poor quality it not useful for anything.

[mark_houlder2]

the number of bits used to store information has no bearing whatsoever on anything except the amount of information you can store: dynamic range or exposure latitude have nothing to do with it. the 'bits' in question are an abstraction necessary because digital computers can only differentiate between 'on' and 'off', and for them to be able to store up to 4096 discrete values, 12 'bits' of 'on' or 'off' values are needed.

 

how does this affect exposure lattitude or dynamic range? it doesn't. you could have 128-bit storage but if the sensor can only capture 2 stops of dynamic range, the dynamic range is 2 stops, not 128. BUT it would be capable of storing (that's storing, not necessarily sensing) 4096 different tones (per channel) inbetween those two extremes. of course, that's not to say that canon sensors *don't* have 12 stops of DR - i really don't know (though I highly doubt it). But just because they store colour in 12-bits per channel really doesn't mean anything except that they store colour in 12 bits.

[thomas_sullivan]

12 stops!? Jeez, I wish. I agree with those that say that a (20D) Canon is very near E6, and that with RAW processing one can achieve extremely nice images up to 1 stop off in exposure. sometimes if I "nail" the "expose to the right" philosophy dead on (ie...meaning it's not "really" underexposed, it's just I want to bring out the shadows a little more) I can tweak maybe another 1/2-3/4 stop out of it. But past that the underexposed areas get too much noise in it for me. Perhaps using CS2's RAW exposure offset AND the lumanance and color noise filters I can get maybe up to 2 stops off and an "acceptable" pic (meaning it's slightly "mushy" from the noise reduction options). But that is absolutely it.

 

Never tried the multi-exposure layer thing yet....really haven't had a tripod with me in a looooooooooong time....heh.