"Sweet Spot" for ISO best sensor performance

[lct]

Advertisement (gone after registration)

[...] Utilizing the base ISO, and using the meter only to ensure that all desired data falls within the thirteen-stop range of the sensor, with proper exposure managed in post-production, is, technically, a superior method of exposure. [...]

 

How would we manage to see anything else than black views on EVFs and LCDs then? Just curious.

[icqcq]

Ah: the same magical BS that allows us to see them now: software. I'm not proposing anything different in the handling of the actual data, but of how we meter and expose. The only reason you can see anything on the chimp and the evf are because software is pushing the data around to where you can see it. That wouldn't change; what would change would be how much data it shows: everything the sensor can register, instead of the artificial seven-stops that was brought over from the film cameras/meters/film/thinking.

Edited March 7, 2016 by icqcq

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[pgk]

I usually guess exposure, take a shot, view the histogram and adjust. A great deal easier than most other methods. Once I've set the 'exposure' as I want it, I leave settings alone until the light conditions change. It works well for me and I look on the LCD as a rough guide rather than a precise reviewing tool.

[lct]

Matter of tastes i guess. I'll stick to my clear EVFs and clean isos for now

[pgk]

Matter of tastes i guess. I'll stick to my clear EVFs and clean isos for now

 

It is.I'll stay with the viewfinder and histogram assessment. That said, I don't find noise too objectionable, depending on whether it looks random (like grain) or too 'digital'. M9's good and noise looks 'filmic'. Canon's; well not so good IMO but there you go. Each to their own.

[icqcq]

ISOlationists.

[lct]

Advertisement (gone after registration)

When clean results at 12800 isos will generalize, photogs will feel less and less concerned by isos i suspect but i may be wrong.

[tomlianza]

Use of the ISO setting in a digital camera basically establishes the the camera exposure threshold for a nominal grey point .  It is important to note that exposure is equal to the illuminance X integration time.  There are effectively two shutters in the modern digital camera:  The physical shutter and the CCD/CMOS "shutter".  I use the word "shutter" in quotes because it can be accomplished in many ways depending upon the physics of the chip itself.  The F-stop of the lens is responsible for the illuminance part of the exposure equation.  The shutter speed/integration time controls the time of exposure.  In most physical situations in photography, the exposure time is the most important part of the equation due to image blur, either due to subject matter or camera handling.  If you look at most AutoISO systems, they generally require a maximum allowable shutter time (1/shutter speed).  In the case of the M240 it uses a 1/focal length of the detected lens so for my 75mm Summilux the lowest shutter speed would be 1/75 of a second.  The electric gain following the sensor output is adjusted based upon the anticipated illuminance and shutter speed.  The ISO concept has absolutely nothing to do with 13 stop or 7 stop thinking... it is aimed at establishing a working point to position the camera A/D convertor to get optimal quantization.  Now let's assume that we did not let the ISO shift by fixing the gain.  It is true that you can post process with gain to lighten an image, but you really have a problem with the total number of real levels digitally captured.  Adding analog gain after the chip has, by definition, almost no impact on signal to noise ratio (because both the signal and noise are amplified equally) , but it has a tremendous impact on resolvable grey levels.  In reality, there is really no ISO sweat spot.  In a modern digital camera it really sets the maximum exposure time based upon physical constraints.  Depending upon the physics of the sensor chip, there may be a "sweet"  integration time.  Much of the work I do with CCD sensors has shown that there is a point on most devices where the dark noise increases exponentially at longer integration times.  Below this point, the dark increase is rather linear.  ISO speed is not the product of old thinking, it is the product of having a system that works.  Given the limits of sensors and A/D converters, we would probably have a very similar system if we started from a blank sheet of paper.  

[CheshireCat]

In reality, there is really no ISO sweat spot.

[...]

ISO speed is not the product of old thinking, it is the product of having a system that works.  Given the limits of sensors and A/D converters, we would probably have a very similar system if we started from a blank sheet of paper.  

 

Thanks, very interesting post.

 

However, Jim's experiments seem to find some "sweetness" (or bitterness ):

http://blog.kasson.com/?p=3138

 

What do you think ?

Edited March 12, 2016 by CheshireCat

[pgk]

let's assume that we did not let the ISO shift by fixing the gain.  It is true that you can post process with gain to lighten an image, but you really have a problem with the total number of real levels digitally captured.  Adding analog gain after the chip has, by definition, almost no impact on signal to noise ratio (because both the signal and noise are amplified equally) , but it has a tremendous impact on resolvable grey levels.

 

If you look at my example in post 50, could you explain how increasing the ISO would help because I believe that the limiting factor is the exposure of the highlight area of sky - blow this any further and the colour and tonality in the sky burns out and a very important area within the photo is lost, negating much of the point of the photo. There IS noise in the lightened shadows, but having shot many images this way and having tried similar images using a higher ISO setting in the past, I find that using base ISO gives me the results I prefer most. 

[ECohen]

So what your saying is: 

 

Shoot RAW. Set my camera at 200 ISO and leave it there.

If I can get my exposure "traditionally" shutter speed and f/stop....great.

 

If not as long as the "proper exposure" fits with in the 7 stop range of the sensor, forget the LCD and chimping, just shoot and fix it in Camera RAW. 

"Proper exposure" with this style of shooting, means enough  to stop the action or enough depth of field?

 

So from now on, concerts/ dark cafes/ auto racing/anything inside/walks in a dark woods/my dog running in the backyard/ small f stop shots..... Shoot ISO 200, expose for the highlights  and fix the shadows "exposure" in RAW........Really?  

And now "exposure".... the word loosely means brightness ?

 

I made some tests and this does work but before I adopt this way of working for everything, I need to weigh the computer time ...Its not a click or two, all dark frames don't have the same fixes It's not a bulk fix per outing and some of you suggest ?

 

 I find it hard to believe that this "style " of shooting is how you all use your camera?  

That said, I do this anyway on some things but for the most part I want a properly exposed shot out of the camera for a lot of reasons, and I do understand the difference between film and digital. 

 

So are you all saying that this is a manufacturing problem, because we are in a transitional period in camera design? According to you all, ISO should be removed as an "exposure" choice? Then why does Leica still include it?

 

I did like that someone on this thread suggested Leica change its metering and it would have to change its processing for an good looking LCD test shot  too.

 

So this is really how you work? ......in our digital world.

Edited March 12, 2016 by ECohen

[icqcq]

ECohen:

 

No, I'm saying the sensor (according to data found online: FWIW) has 13 stops of latitude, but the meter in the camera lies to you and tells you you only have seven stops to play in, and, based on the ISO you dial in, it lies about which particular seven stops of the 13 you have access to at the moment. Logically, it would be better if the camera/meter let us work throughout the sensor's 13 stops of latitude, without the artificially imposed limits of 'filmic thinking': film had a more limited latitude range than the sensor does, but you had the flexibility of shooting films of differing sensitivities.

 

No, I doubt anyone here is really shooting this way: we're learning about what's going on digitally, and talking hypothetically about where tradition has rode roughshod over new technology, and what would be better for everyone if what would work better didn't involve having to relearn everything, so we don't relearn things, we just take old vocabulary and old thinking and overlay the new stuff and say it's the same, when it's not.

 

At least, all of this thinking preceded Tomlianza's input....

 

Tomlianza:

 

So... I'm stuck at this: you're saying that ISO is our way of 'warming up' the sensor, saying, in effect, "Watch for light in this range"? Which is to say, although the sensor has a thirteen-stop range, in reality it struggles to manage light throughout that simultaneously, so ISO is a way of forcing the photographer to limit his/her expectations of latitude to what the sensor can reasonably manage?

Edited March 12, 2016 by icqcq

[pico]

Boosting so-called ISO amplifies the processing across all sensor cells. There is no ISO above the base. It is all a competition between noise and signal.

.

[lct]

Could someone explain to me how EVFs could stay clear and clean if we underexpose by 2 or 3 stops as suggested above? Just trying to comprehend what remains a total mystery for me sorry...

[icqcq]

The native (and only) ISO of the sensor is 200. You shoot at 3200: you've pushed the sensor four stops: that information is buried, in the same way that shooting at 200 and underexposing by four stops would be buried. But because you've dialed in ISO3200, the software says, "Ah! I will now push the information up four stops, such that my master might chimp it!" Exposing 'properly' at ISO3200, and underexposing four stops at ISO200, are exactly the same thing in terms of what information the sensor captures (except, not quite, not exactly, because we've seen that running the ISO up can force the software to compromise the data, but that's after-the-fact, and one of the things that keeps this thread running); the only difference is that the software compensates for that by moving information around (first for the meter; then, after exposure, for the Chimp/histogram, and, finally, in the data/file) so that you can see the information in graphic form (i.e., "Oo: look at the pretty flower!").

 

What we (OK, it's all me: I'm the provocateur in this thread, and I'm entirely content with that; I would like to thank the grand minds that are providing scientific input, and staunchly advocate against the twitterheads whose attention-spans are exhausted by the 146th character of any post) are proposing is that we get rid of the artificial limits imposed by the old "Seven Stop Film Latitude" thinking, and push on into the "My CMOS Has Thirteen Stops Of Latitude: ISO/Seven Stops of Latitude Is Dead!" thinking. What we* are proposing is that the manufacturer gives us a way to use the full latitude of the sensor with our full knowledge and consent, instead of confining us to an artificially and traditionally imposed seven stops (see tomlianza's contribution above for caveats and possible modifications to this thinking, which I've yet to fully integrate), doing away with ISO as a traditionalist holdover that is being imposed on a new technology that exceeds it.

 

While we're at it, I'd like to propose a new nomenclature for apertures, based on percentages, instead of this whimsical "Why Darling! Can't you see it's sequentially based upon the square-root of two, interspersed with doubling starting at one, praise be to Hipparchus!" nonsense.

Edited March 13, 2016 by icqcq

[lct]

I just wanted to comprehend how could the EVF stay clear and clean if we underexpose the way you suggest.  It would become noisy this way wouldn't it? If you believe you have answered my question already just forget it as i'm not clever enough to understand what you say sorry.

[icqcq]

No, lct, I prefer to think that I have not yet explained my thinking clearly enough, and a better explanation might help me understand it better myself.

 

What I'm saying - based on my understanding from those more technically savvy than I - is that the EVF would stay clear and clean because the EVF is always being told, by the software, what to show you. If the software tells it to show you a deep, dark, muddy image, it will. If the software tells it to show you a beautiful, bright image with a full range of information, it will.

 

More importantly, the software could use the same data to tell the story in those two, very different, ways. The software is controlled by, in our case, Leica. But Leica, like most manufacturers, has chosen to emulate what we knew about and from film, into the software it programmed for our M240s. Leica has created a synthetic film experience for us, because that is familiar and we already know how to work within that, comfortably.

 

But it appears that, if we pushed ahead and accepted what digital actually has to offer, we would have a different experience. If Leica programmed the camera software to reflect what the CMOS is capable of, we would have a bit of a learning curve, and that includes unlearning some things that many of us learned with great effort over a great deal of time, and which often, as a result, seems 'natural,' 'innate,' or, even, 'unlearnable.' Today I was teaching a friend the basics of photography, and when I began to describe what we've been talking about in this thread, she - with no experience - couldn't comprehend why there would be an artificially ('traditionally') imposed seven-stop limit on a light-meter or histogram. Absolutely baffled, which made me realize that what we're discussing and talking about here isn't 'natural' or 'innate,' but learned, and, therefore, unlearnable, or - better - we could learn new and better ways of thinking that are better suited to the new medium in which we are working. There are parallels - analogues - brought over, but they are not perfect, and as digital continues to evolve, we were better off evolving with it, than forcing it into existing vocabulary and traditional thinking that are, simply, convenient, rather than better.

 

Of course, and, curiously, we are among the most traditional of digital photographers: we are shooting, of all things, rangefinders. With Manual Focus lenses. And Manual Exposure! What a bunch of throwbacks we all are!

 

And yet... and yet.

Edited March 13, 2016 by icqcq

[lct]

[...] What I'm saying - based on my understanding from those more technically savvy than I - is that the EVF would stay clear and clean because the EVF is always being told, by the software, what to show you. If the software tells it to show you a deep, dark, muddy image, it will. If the software tells it to show you a beautiful, bright image with a full range of information, it will. [...]

 

The firmware could then show a well exposed jpeg whilst the raw file would stay underexposed if i follow you well, right?

[icqcq]

Exactly. The software is interpreting the data provided; that's actually a good idea: a jpg that visually shows range captured within the sensor latitude, while the raw file would simply wait for proper extrapolation in post-processing.

Edited March 13, 2016 by icqcq

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[lct]

Good, now the photographer will need to control such underexposition of course. So he will need a unit of measure for that. It can't be f stops, shutter speeds, asas nor isos (if i understand well) so what unit would you suggest to use instead? EVs? Sorry to ask all these questions...