How to determine DR at high ISO

Submitted by allfive on Wed, 03/23/2016 - 13:29
Forums: 
General discussion

New to FRV.
Does the user manual indicate how to determine DR of various high ISO (1600, 3200, 6400) RAW images?

Iliah

Wed, 03/23/2016 - 13:50

Dear Sir:

If you could qualify what DR is, we can help qualifying it.

allfive

Wed, 03/23/2016 - 14:34

Dynamic Range :)
Actually wish to use FRV to find optimum balance between shutter/aperture and ISO in very low light conditions for my Nikon 600. Large screening room, high overhead lighting, hand held, one or two people speaking where skin highlights can be blown out but lots of shadow surrounding the person. Wanted to input a reasonable dynamic range number and shadow boost in FRV preferences for a variety of ISO's.
You may have a better solution.
Many thanks.

allfive

Wed, 03/23/2016 - 14:47

Also trying to determine how much I can underexpose and "push" in RAW conversion if exposure and high ISO is not sufficient.
I get usable results already but have to work a lot to reduce excessive blue in shadows and keep highlights on skin under control. Trying to find the right exposure to give me the best chance with extreme situation. :)

Iliah

Wed, 03/23/2016 - 14:49

Dear Sir:

Dynamic range (or DR) currently does not have any useful photographic definition. For example, in practice it depends on what are the light sources, and wahat is the print/display size.

Nikon D600 is fairly iso-less camara, meaning engineering dynamic range is decreased close to 1 stop for each increase of ISO by 1 stop.

First thing to know is the level of incident light. You can either use a suitable hand-held meter, or spot-meter with your camera drom a grey card. The result will be the basis for determining the ISO settings range.

If preserving the highlights is important, you will be best off spot-metering from those important highlights and setting the camera to +3 EV.

Colour tints in shadows sometimes can be diminished by using a deep hood on the lens, getting rid of "protection" filters, clearning the camera chamber to reduce flare, and even by trying a different lens.

allfive

Wed, 03/23/2016 - 15:34

Thank you for clarifying my misunderstandings. So I was conflating photographic dynamic range with an EV number spread? Or engineering dynamic range vs scene dynamic range?
 
I think I was confused in the workflow tutorial video where it stated, "we know that the higer the ISO, the narrower the dynamic range",  then went on to clarify how to adjust the underexposure indication in FRV, i.e., Preferences / "underexposure detection below sensor saturation", and set that to an EV number less than the default 8. So initially I was wondering by how much I needed to vary that for different high ISOs, 1600 up to 6400.

You wrote:
"Nikon D600 is fairly iso-less camara, meaning engineering dynamic range is decreased close to 1 stop for each increase of ISO by 1 stop."

Thank you. So at what ISO should I begin decreasing the underexposure preference in FRV from the default 8 for each ISO stop increase? ISO 800? Or should I work from the 5.5 number given in the tutorial for ISO 3200 and adjust up or down accordingly?
 
Thank you kindly for your very helpful advice on how I should best approach my specific shooting situation. :)

Iliah

Wed, 03/23/2016 - 16:04

Dear Sir:

ISO setting on most of the cameras, D600 being among those, is acting similar to microphone amplifier; that is, the signal is captured before the ISO gain is applied. In such situations most of the noise and non-linearities are already baked into the signal, and the gain does not help much when it comes to decreasing overall noise. 

Suppose the camera at base ISO setting allows for 10 stops of photographically usable dynamic range. Increasing ISO by 1 stop, we will be clipping 1 stop in highlights (provided exposure settings, that is aperture and shutter speed are the same; and the light did not change between the shots); but we will be getting a tiny bit more in shadows, typically around 1/6 EV for D600. If we decrease the exposure (again, only shutter speed and aperture are parts of the exposure here, ISO is not) to prevent highlight clipping - and that means that we are going down at least 1 EV in exposure - the total amount of light recorded is decreased, and we will be plugging shodows by close to 1 stop. Net result - whatever we do, increasing the ISO setting by one stop we shrink the total recorded dynamic range (either clipping the highlights, or plugging the shadows, or both) by close to 1 stop.

In my experience ISO 800, as you suggested, would be a fair estimate for the point, where one needs to start decreasing the underexposure limit. For my purposes, I prefer to start decteasing it from ISO 640, as I print large.

allfive

Sun, 03/27/2016 - 08:23

Thanks, kindly, Iliah. You explanations are quite helpful. (Don't know if the protocol on list is to close a thread with thanks, but I will do so.)

l_d_allan's picture

l_d_allan

Tue, 05/17/2016 - 13:46

I have what may be a related question:
When FRV is showing an image, is there a way for it to display the actual DR (or equivalent term)  of the scene. I realize you can visualize this from the histogram, but perhaps there could be something like the following (with made up numbers):
Lowest RAW RGB value (123, 321, 222)
Highest RAW RGB value (12345, 2345, 10001)
DR per channel (10.0, 6.5, 8.4)
Or a simpler, even more contrived example:
Lowest RAW RGB value (100, 200, 300)
Highest RAW RGB value (400, 200, 2400)
DR per channel (2.0, 0.0, 3.0)

lexa

Wed, 05/18/2016 - 03:52

FRV histogram X-axis is in EV (stops), so it is easy to quick estimate image DR (not lowest/highest values, but 'significant pixel count').

For deep inspection with detailed statistics we suggest to use our RawDigger software. It provides detailed histograms, overall statistics (min/max) and can examine not only entire image, but also arbitrary rectangular selection.

--

Alex Tutubalin/FastRawViewer team

l_d_allan's picture

l_d_allan

Wed, 05/18/2016 - 05:20

HowTo? Edit an OP or reply to have spaces between paragraphs in lines?
 
I only use the LibRaw forums infrequently, so making posts is always a re-learning experience. For several of my posts (including the one to which I hope I'm reply'ing to), I should have used Preview and realized that one CR/LF between paragraphs results in paragraphs having no blank line between them. 
 
I tend to assume (ASS-U-ME) that Edit will be available, so I "draft and save", rather than use Preview. Obviously, operator error on my part.
 
It also doesn't seem possible to delete a post. That would allow me to "realize the error of my ways", and use ^C Copy, delete, and start over but putting in what seem to be extra line breaks. And a reminder to use "Preview" ... again, this is PEBKAC on my part.

l_d_allan's picture

l_d_allan

Wed, 05/18/2016 - 05:10

Thanks for the clarification.
 
I think I'd use the word "less difficult" or "easier" than easy ... and it is still clearly a "quick estimate' even with RD. With the default linear vertical scale for histograms, the left and right sides generally are pretty close to the horizonal line. The non-linear choices do help with that ... a LOT.
 
BTW: Is the option to use a non-linear scale for the vertical axis of the histogram availabe in FRV? Did I miss that? If not, could that be a "Feature Request"?
 
tl;dr? sorry
 
With MagicLantern's auto-ETTR ... which I am guessing you are aware of if not familar with ... you can use "fine tuning" of how much, if any, blown pixels to ignore. For a low contrast scene, you'd use zero %. At the other extreme with the sun in the middle of the frame on a clear day, you might use 10% or even higher. A typical value might be from 0.5% to 1%, which often allows much more exposure than 0% ... perhaps several EV, which seems counter-intuitive.
 
My point is that RD and/or FRV could perhaps allow conceptually similar "fine tuning" of what the "significant pixel count" default would be. 0.5% to 1% might again be default values, unless the 'tog KNEW there were no blown highlights and 0% was appropriate. Also, if the capture was obviously less than ETTR because there were 0 blown pixels, then the OvExp fine tuning could be 0%. That probably applies a LOT with JPEG based blinkies, histogram, Preview, Review, Zebras, etc..
 
I don't think in terms of refining UnExp as OvExp seems Much More Important. That could be more complicated to get "just so", if that was important.

lexa

Wed, 05/18/2016 - 07:14

FRV allows to tune 'overexposed count' for Auto-Exposure feature (in Preferences - Exposure).

--

Alex Tutubalin/FastRawViewer team

Add new comment