Balancing Exposure and Processing

[J-see]

I read somewhere not every ISO stop works in the same manner and that some do a pretty decent job while others are horrible at it. There's a difference in how they implement the gain but it's hard to impossible to find any technical explanation on ISO behavior for Nikon cams.

 

[J-see]

I finally have some data affirming what I'm doing.

Shadow Improvement of Photographic Dynamic Range versus ISO

If you want to check your gain, that place has all the information.

 

[J-see]

I've been doing quite some testing the last days and found a reasonable approach that works for me.

These are some shots I took each time lowering shutter one stop (of light) and increasing ISO one and another shot keeping ISO at 100. Then I normalized the exposure for the 100 and adjusted it to the higher ISO. The colors are not identical but that can be fixed.

Control:



ISO 100 - 200



ISO 100 - 400



ISO 100 - 800



ISO 100 - 1600



ISO 100 - 3200



I can push ISO down about three stops without much difference for the end result. That way I preserve my colors and DR better with a minimal price to pay. Only ISO 6400 I need to figure out the best method.
But for the rest I only need 100-400 and I can shoot up to light levels requiring ISO 3200. If I shoot with auto-ISO set to max 400 and -3EV, I should have the best quality for those situations. If light is lower, I need to do something else. At 800 ISO all goes down too rapidly.

 

[J-see]

A test with night shots.

Manual, fixed A&S but each next shot I increased ISO according ETTR (100-200-320-400-800 ISO). It was too dark to blow out highlights.

I processed the 100 to my liking and then applied the same settings on the rest only lowering their exposure to the 100 and setting black and white at their clipping points.

I then cropped to 100% and scaled that to 200% in PS. All are saved using the same EXIF but not necessarily in the correct order.

 

[J-see]

Here's some figures showing my DR loss in the highlights for DR gain in the shadows when using ISO while shooting 14bit.



As you see, besides that small bump there's zero improvement up to ISO300 and after that, the actual gain in the shadows in minimal. When you compare that with the steady loss in my highlights, you can start to wonder when it is worth using ISO.

There happens more than just loss and gain for DR and it is different for every cam but it isn't bad shooting with this knowledge in mind.

It affirms the little difference you see in the shadow details between the shots above.

 

[J-see]

Here's the gain/loss for my D3300.



If you start thinking about these numbers, one can start to wonder if ISO is not above all a setting for shooting JPEG. Not unlike WB, contrast... etc. The one is like, back in the day, shooting film while the other is shooting instant film, like Polaroid. The first needs to go through the lab or dark room to be properly exposed while the other gets exposed in cam and rolls out as a finished product.

 

[J-see]

Ok, I redid the ISO processing of the indoor test. 1 control shot at ISO 100 and then 1 at ISO 3200 correctly exposed (according cam) and one with the same A&S but ISO 100. 5 stops exposure increase to the ISO 100 and all have their B&W point set to the clipping point. I used the cam monochrome to only show the shot at luminosity values.





Noise close-up:





Midtone region:





Light region:

 

[J-see]

I'm almost there with post-exposure. The monochrome shots looked close to identical so the luminosity values were correct. The only problem left are the colors that somehow get affected by ISO-less shooting.

I adjusted the camera profile and corrected the RGB channel and shadow tint and in that seems to be the solution to this problem. I still need to fine-tune it but they slowly start to look alike.





Once I have that sorted out, I just need to save different exposure presets and apply those on import.

 

[Blade Canyon]

I shot "ISO-less" last night at a wine bar. It was a very low-light situation. The RAW files interpreted by Adobe were much different than the JPEGs created by the camera. I think I left the camera on ISO 400, subscribing to the theory that it's okay to use gain to push it a little to the right. (I also brightened her eyes a little in post.)



The EXIF says I was at -1 exposure compensation, but the camera was in manual mode so that had nothing to do with it.

 

[J-see]

I shot "ISO-less" last night at a wine bar. It was a very low-light situation. The RAW files interpreted by Adobe were much different than the JPEGs created by the camera. I think I left the camera on ISO 400, subscribing to the theory that it's okay to use gain to push it a little to the right. (I also brightened her eyes a little in post.)

View attachment 135770

The EXIF says I was at -1 exposure compensation, but the camera was in manual mode so that had nothing to do with it.

I think the problem with JPEG is the bit-size. I don't know about the D800 but I shoot 14bit with the D750 and that contains loads of information compared to a JPEG.

I need to test it some more but it seems to work. The D800 wins about 1/3th of a stop in the shadows directly after ISO 100 so there is some direct gain. For me the first 300 deliver nothing so by the time I get my first 1/3th, I sacrificed a whole lot for that.

The only problem is the color shift which is a bit annoying at the moment and the only reason I shoot up to 400. After the third stop, colors shifted too much and it's a whole lot of work to correct them. Until I found a formula.

Also after 400 my SNR takes a dive which I try to avoid.

I encountered a couple more online that apply ISO-less shooting but not many. I think we have to map this territory ourselves and check when it is good to use and when it isn't worth it. If cams keep evolving and they can push the technical side some more, ISO will become something of the past. At least for RAW.

I just check the D800 vs D750 and both the cams are fairly similar when it comes to loss and gain isn't that different either. So what works for you should work for me. There might be some difference because of sensor pixels.

 

[J-see]

A lot of the color problems can already be fixed by switching camera profiles. Here's one processed roughly and then using different profiles in the camera calibration.

SOOC: -3EV so this is an ISO 6400 equivalent.


Adobe standard:


Flat:


Neutral:


Cam standard:


I shoot with the aRGB profile but for these shots, adobe standard isn't the right choice during post processing and neutral or flat should be selected to start with.

 

[J-see]

I've been shooting all day at ISO 100 regardless the subject or conditions. I opened up as wide as possible for any given situation and slowed down what I could. Besides that, I just shot.

This one I had to increase exposure about 2 1/2 stop in post but with the snow it's hard to say what ISO I would have needed. The meter goes bonkers with all the white.



I have no shot with similar settings to compare them too. The closest is this ISO 1000 which I shot some time ago in rather good light conditions. It's while I was testing the D750.



Both cropped at 100%. Noise reduction disabled.





This is the "darkest" I shot today; about 3 stops in post.







Tomorrow I'll try to push it beyond three stops and see what that does.

 

[J-see]

Here's a radical example.

I shot the first with auto-ISO up to 6400. Light was very bad and with this shutter speed it picked 5600. All shots are processed with LR's standard 25 sharpening but zero noise reduction.

I cropped a part out of it.



Then I shot it again but now using ISO 100 instead.

SOOC:



Processed identical but I adjusted exposure in post by 6 stops.



WB is identical for both but the colors differ. It's hard to say which is correct since they both are very different at bit depth and tonal range. I need to test this some more until I know what to do and can fine-tune it.

Now both cropped at 100%

 

[Marcel]

Here's a radical example.

I shot the first with auto-ISO up to 6400. Light was very bad and with this shutter speed it picked 5600. All shots are processed with LR's standard 25 sharpening but zero noise reduction.
Processed identical but I adjusted exposure in post by 6 stops.



WB is identical for both but the colors differ. It's hard to say which is correct since they both are very different at bit depth and tonal range. I need to test this some more until I know what to do and can fine-tune it.

Now both cropped at 100%

View attachment 136244

View attachment 136245

As much as the noise seems similar, I think that you are loosing color information when you underexpose that much. This might be why the color doesn't match in the shadows specially.

 

[J-see]

As much as the noise seems similar, I think that you are loosing color information when you underexpose that much. This might be why the color doesn't match in the shadows specially.

No matter how counter-intuitive this sounds but the underexposed shot has more color information and a wider tonal range than the correctly exposed. The sensor received the exact same information but during A/D conversion, it had to limit that information because of the ISO used while the "underexposed" didn't suffer that problem.

The colors are indeed slightly off but that's actually a problem which can easily be fixed once I know how to adjust them. At a luminosity level both shots are fairly identical and colors is just a matter of adjustments.

I very basically processed these since I was only concerned about noise differences. That's seems fairly similar no matter what approach.

Once I start fine-tuning the processing, I can do much more with the ISO 100 than the ISO 5600 because the RAW contains more information. It's then just a matter of getting it out.

What's also fantastic about ISO-less or RAW+ shooting is that it is very relaxed shooting. Set the minimum shutter and aperture required and shoot. The rest is moved to post. Evidently this is only for shots where ISO is the only option left to get more light. Else I prefer slowing down or opening up.

 

[Eyelight]

An observation.

Whrereas the sensor receives the same amount of light at the same exposure settings regardless of ISO, in low light, the sensor doesn't receive enough light to create all the colors in the image. This is when the camera ISO setting tells the processor to "build" the rest of the image.

What is missing from the ISO 100 shots is information added by the in-camera processing to an image that wasn't saturated enough on the sensor to show the true colors.

I think some of the examples shown are remarkable, and I'm still interested in where the breaking point is. But would have to add that I'm not seeing a reason to bypass the camera's capability to fill in the missing information in an underexposed image.

 

[STM]

What you have listed above are what are referred to as "equivalent exposures". It would be the same as ASA 100 (I still can't get used to saying ISO) and 1/125 second @ f/4 and 1/250 sec @ f/2.8. Both yield the same exposure, all other things being equal. One of the unique, though somewhat cumbersome at times, features of Hasselblad "C" lenses is that you set aperture and shutter speed (the shutter is in the lens not the camera) on rings on the lens barrel opposite each to set exposure. You have to depress a lever to unlock them. Once set, when you rotate one ring the other ring moves with it and each click will be an equivalent exposure. Later lenses like the CF had the two rings which moved independently of each other, but you could depress a button on the aperture ring which would lock them together and again give you equivalent exposures.

 

[J-see]

An observation.

Whrereas the sensor receives the same amount of light at the same exposure settings regardless of ISO, in low light, the sensor doesn't receive enough light to create all the colors in the image. This is when the camera ISO setting tells the processor to "build" the rest of the image.

What is missing from the ISO 100 shots is information added by the in-camera processing to an image that wasn't saturated enough on the sensor to show the true colors.

I think some of the examples shown are remarkable, and I'm still interested in where the breaking point is. But would have to add that I'm not seeing a reason to bypass the camera's capability to fill in the missing information in an underexposed image.

I think the color shift has less to do with the light levels. It is true that the less information the sensor receives, the less accurate the RAW file will be. But that same shift in colors happens during all light levels. I can shoot the same shot increasing ISO each time and thus overexposing but when normalizing in post, the colors are fairly identical. The same is not true for shooting low and increasing exposure in post.

I have few problems with the color shift up to three stops but the moment I go beyond those, the color shift becomes fairly strong but from what I see, mainly limited to the shadows. There's a 0.5EV difference in the shadows between 100 and 6400 so maybe there it got clipped in one color channel. I have to compare them.

I'm reading up upon it and it seems WB affects the % of colors that are present in each channel. I think the problem also could be situated there. The luminosity levels are identical and the colors are but a profile attached to the RAW. It's when loading this profile and adjusting it in a RAW editor, there's some issue with how the colors scale in relation to exposure increase. This might be true for every exposure difference but I assume since the difference in exposure is seldom as huge as what I am doing, we don't really notice.

 

[Marcel]

As much as I applaud your courage to try proving this theory, I'm not convinced of the value of this. Of course, you can recover a raw file, but getting the sensor to "grab" more light in the shadows specially is a huge plus for me. Now time might prove me wrong, but where I come from (film days), when you have nothing on the negative, you're left with nothing on the print.

 

[J-see]

I discovered the problem. Dynamic range.

The scene has a wider dynamic range than the cam's capacity thus something has to go. For the high ISO this loss is evidently in the highlights while for low ISO, it is in the shadows.

When checking the clipping it becomes obvious which channel paid the price.



It was not unknown to me clipping would occur differently but I didn't make the link until now.