Dynamic Range measurements


Hello, I realize tests of cameras and I want to determine the dynamics of the sensor with RawDigger. Here's how I do it:
I shoot a gray and uniform scene. For the first shot I let the camera automatically expose for a VR Exposure = 0. Next i surexpose the target by EV=+7. Next i shoot i dark with the lenscap at 1/1600 s f/16 to have a black and quick exposure

So i have
- 1 shot at EV=0
- 1 shot at EV=+7
- 1 shot darknoise (1/1600 s with lenscap)

I try to mesure the Sony RX10

I open the raw file with rawdigger to mesure the darknoise and with the histogram i read the max of the statistics

Red 60
green 1 44
blue 88
green 2 52

Next, i open the raw EV=+7

Red 15796
green 1 15796
blue 15796
green 2 15796

All the channels are at 15796

to evaluate the dynamic range i make this calcul : Dyn Range = (log (B) - log (A))/log2 where B is the result at EV=+1 and A dark noise

i found this :

Red 8,01977603373876
Green 1 8,46723501070998
blue 7,48784003382305
green 2 8,24683193431926

Do you think my calcul is correct ?

Next i calculate for EV=0 and take the min + the max of the statistics / 2.

For the surexposition i calculate (log (B) - log (A))/log2 where B is the max for EV=+7 ans A the result for EV=0

i found this


So for me, the max surexposition for the RX10 is 2,3EV.
But when i shoot a targer with +2EV i can't recover all the detail in the high level. The max of overexposure is +1,7EV

How can you explain this ?
Thank you

First: some RAW converters

First: some RAW converters apply hidden exposure compensation (positive) to compensate in-camera underexposure. It is well known 'feature' of Adobe ACR/LR: http://photographylife.com/adobes-silent-exposure-compensation

Second: you use (max-min) as noise estimate. This results in heavy over-estimation of noise (esp. if hot pixels are present on your sensor).  Use 'sigma' column (4th field in RawDigger statistics) as variation /noise/ estimation.

Also, if you divide signal range to noise in dark areas, you'll estimate some kind of 'technical' dynamic range. In real photograhy the 'practical' dynamic range is more interesting.  To measure it you may shot something with small and not very contrast details (such as piece of newspaper). Use wide bracketing, e.g from -8EV to +3EV with 1EV step and you'll know much more about your camera after you see the shots.

Thank you for the answer.

Thank you for the answer.
You mean that i can shoot a target with differents exposure and juste look at the overexposure / underexposure value to determine when the camera is out of range. It will be great !

No. If you need precise under

No. If you need precise under-exposure display, you need to evaluate dynamic range of your camera (at given ISO and light) according to your needs/taste, than set Underexposure limit according to it.

The practical dynamic range is depends (with fixed ISO/light) on the detail size and contrast you want to see in shadows.

Here is my DR target: http://blog.lexa.ru/sites/blog.lexa.ru/files/images/DR-image-8bit-thumb-... (I print it using jet printer, then shot with different exposure).

Here is how I do comparison: http://blog.lexa.ru/sites/blog.lexa.ru/files/images/desktop.img_assist_c...

As you can see large (30pt) and contrast( 1.5EV) details are visible, while medium sized (18pt) details are not (on the same image)

So, if you plan to make large size print (so you need small details), the practical DR is smaller.  If you want to make screen-size preview (so your large 16Mpix or so image will be reduces to 1-2Mpix screen size), you need only large details, so DR is larger.

Here the complete text about my Canon 6D dynamic range estimation: http://blog.lexa.ru/2013/04/15/eos_6d_dinamicheskii_diapazon_na_nizkikh_... It is in Russian, try Google translate or so. Feel free to ask if translation is not good enough.

one more question

I try another solution with a Stouffer chart : 41 step for 13 Stop each steph is 1/3 EV
I shoot with the X-T1 and expose for high light
i open the Raf. It is possible too determine the dynamic range with the over / under exposure alerte ?
you can see the image here : https://drive.google.com/file/d/0B707sUilawn1UTVCZGxJUXpjT1U/edit?usp=sh...
the over and under exposure here : https://drive.google.com/file/d/0B707sUilawn1QUVLSEdzQ2J3X0k/edit?usp=sh...
and the RAF file here : https://drive.google.com/file/d/0B707sUilawn1YnhCc0VyajhNYW8/edit?usp=sh...
With your tool, i can read about 26 or 27 stop for 9 EV. Do you think it's a good way ?
Thank you

Dear Sir,

Dear Sir,

There is hardly anything of substantial contrast past 8 stops down from maximum. Part of the problem may be flare, yet another - cosine law, and have you checked how even is the light? Say, if you shoot the step wedge left to right and right to left with the same exposure, are the readings from the same patches are close enough? Stouffer transparency targets are made on black and white film, silver grain and raster add complications again. As a result, it is hard to know how much contrast is left (or, in other words, what is the sensitometric step between the neighboring patches). As you can see, histogram peaks are not at 1/3 EV from each other, and below 6 EV from maximum they really start to degrade.

An easy to shoot target is http://www.amazon.com/Sekonic-Corporation-401-757-Exposure-Profile/dp/B001JQLH8I - it is all matt, and with proper lighting it provides an excellent estimation of the loss of contrast due to flare and glare. I bet your readers will appriciate such a test very much, especially when different lenses and cameras are compared. Still, for a good estimate of photographically useful dynamic range an old test with a newspaper (if I remember correctly Alex already have suggested it, using a low contrast target and progressively underexposing it) is most telling.

Best regards,
Iliah Borg

Add new comment