How to determine the dynamic range of a camera - my experiment

[TCampbell]

I don't think there's any reason to use multiple colors. I had wondered about the 18% gray ... thinking that the T1i may be calibrated to 12% gray.

But in any case... to those thinking you'd need a card with multiple gray tones, I don't think you would. Suppose you're shooting a scene that requires a lot of dynamic range ... some objects are well lit and near light and some are dimly lit and near black. It's not as if the dark objects are dark because someone painted them a darker shade. By simply reducing the light in a known and controlled manner that can be gathered from a source of a known consistent brightness you are effectively testing the dynamic range.

The results you're getting are pretty consistent with what I've seen.

Well done!

BTW, I happen to have a Sekonic L-758 light meter. The meter can "profile" the dynamic range of a camera, but to do it they offer a few options. The option they WANT you to use is a special gray card that costs about $140 (if I remember). I noticed BorrowLenses.com will "rent" you the card for about $10 (more reasonable.) The card has, if I can remember (it's on the shelf behind me somewhere) about 24 shades of gray. They want you to photograph the card bracketing for normal, -3 and +3 stops. You then import the images into some analysis software. It does require that you take care to guarantee the lighting is even across the whole surface of the card (the card is fairly large) and between all three exposures (which is trickier than you might think.) If you do it right, they product a camera profile for YOUR camera which is then uploaded into the light meter. When you then sample the subject taking multiple meter readings, the meter tells you what exposure to shoot and plots little arrows along a line showing the limits of what YOUR camera can handle and also arrows showing the meter readings that you sampled. Basically... the meter helps you insure that you'll get the full dynamic range of the image into the shot based on the capabilities of YOUR camera and/or tells you if the image exceeds the dynamic range your camera can handle.

 

[pgriz]

You lost me at math.

What math? I didn't run a regression curve-fit on that.

 

[pgriz]

@ Tim Campbell: I thought that the calibration was lower, since the 18% gray card was coming in at 122 value (when the meter was zeroed) instead of the 127 that I was expecting. So if the camera is actually calibrated to the 12% gray, that would make sense. As for the light-meter, I've got the L-358, and it's working well for what I need. Having done the curve, I feel pretty confident that I can put the highlight detail where I want it. Another thing I discovered doing these tests, is that the Magic Lantern spot-meter thingie (which is supposed to determine where in the range of 0-255 the spot will land) is consistently underestimating the exposure. On the other hand, running the spotmeter on the captured image gives me the same values I get when uploaded to the computer (RAW file) and measured directly. So the testing I've done tells me that if in the captured image it shows a highlight detail has a value of (say) 240, this will match the image when uploaded to the computer. I'm one of those people that needs to connect the dots between when I start with, and what I end up with. This has been a good learning exercise. Thanks for your comments!

 

[480sparky]

.........But in any case... to those thinking you'd need a card with multiple gray tones, I don't think you would. Suppose you're shooting a scene that requires a lot of dynamic range ... some objects are well lit and near light and some are dimly lit and near black. It's not as if the dark objects are dark because someone painted them a darker shade. By simply reducing the light in a known and controlled manner that can be gathered from a source of a known consistent brightness you are effectively testing the dynamic range............

FWIW, I tried this exact same idea on all three of my cameras. I came up with pretty much the same results for all three.... a D60, a D7000 and a D600. And it didn't matter what ISO I used on any of 'em. They all came back as 11-12 stops. In short, using this method my D60 at ISO 1600 had the dynamic range as my D600 at ISO 100.

 

[Mully]

Digital grey card scale is 12% and not 18% (left over from the paper days).

 

[pgriz]

@ sparky: I've heard from some of my shooting buddies, that they are getting better dynamic range with their Nikons compared to the Canons. I'm hoping to replicate the test with actual cameras to see if that's the case.

 

[480sparky]

Digital grey card scale is 12% and not 18% (left over from the paper days).

I used a Robin Myers digital gray card.

 

[Mully]

So do I ......I like it being plastic unlike the old Kodak cardboard ones

 

[480sparky]

So do I ......I like it being plastic unlike the old Kodak cardboard ones

And it's solid color throughout. So if it gets scratched or stained, a little sanding takes care of the problem!

 

[TCampbell]

Digital grey card scale is 12% and not 18% (left over from the paper days).

Yet _most_ gray cards are still 18%. You have to check carefully if you want a 12% card. You can find them, but the majority seem to still mostly be 18% (and I've often wondered why, after all these years, the 12% gray isn't the common one.)

I suspect it's possibly because people use a gray card more commonly for white balance than they do for reflected light metering.

 

[Helen B]

In no particular order:

The Sekonic calibration system for the 758 doesn't work on the raw files - it works on files that have had a tone curve applied, so it isn't great for determining absolute dynamic range - ie what can be done with the raw file. It works best for JPEGs. The Sekonic system/software doesn't tell you the dynamic range of your camera without a lot of fiddling - which is easier to do without the software. (I have the genuine Sekonic card if anyone want more details about this system.)

My comment about using two known reflectivities is that it allows you a check when using shutter speed to vary exposures. If you trust your shutter speeds you don't need it - you can use one reflectivity.

It doesn't have to be 18%, 12% or 50% for a dynamic range test. It doesn't matter as long as it is consistent and evenly lit. (Of course for white balance it is better if it is brighter than 18%.)

One difference between doing a series of different exposures of the same card in the same lighting and one exposure of a multiple density step wedge is the effect of lens flare. With the latter technique flare from the brighter patches will affect the exposure of the darker ones. With a succession of field-filling exposures of one brightness value, the effect of flare will not be as significant. That is one of the reasons for the design of the DSC Xyla. A Stouffer step wedge, on the other hand, was not designed for camera tests.

 

[hirejn]

It's easier just to use a Sekonic meter (L-478 or L-758) and the DTS software, but if you're a math or science junkie and have a lot of time, you could go old school. Of course the Sekonic does cost upward of $400, but if you don't already have a meter, it's worth it.

 

[amolitor]

ISO for a digital camera is a somewhat vague notion, and varies a bit manufacturer to manufacturer. It's surprisingly difficult to get a handle on, in the sense that Ctein had to really struggle with it. I would be very surprised to learn that a metered-correct grey card shot at precisely 127.

Any anyways, isn't it going to be more like 2047 on a 12 bit sensor, and 8191 on a 14 bit sensor?

This talk of 127 and 255 in the context of raw files has me a bit confused.

 

[pgriz]

I use DPP for preliminary processing of my images. The histogram in both the camera and DPP show it ranging from o to 255. When I mouse over the image, the cursor information shows the X,Y coordinates of the pointer position, and the values of the three channels, ranging from 0 to 255. So, those are the values that I am reporting.

 

[amolitor]

So I am guessing that what you're seeing is something like 'the value the software intends to place at the pixel when you convert it to JPEG' or something?

As long as you're not making any adjustments (or are making identical ones to all the images, maybe?) it seems like it ought to be ok.

Thanks for the clarification!