Exposure adjustment to RAW files--any downside?

[Pieter 12]

The title says it all. I sometimes will stop down resulting in underexposure to get the depth of field I want without having to increase the ISO. In post processing I will then make an exposure adjustment to the RAW file to get back to the proper exposure. Does this add noise or other unwanted artifacts, etc?

 

[rdeloe]

What camera are you using? Under exposure will lead to more noise when you lift the exposure to get back to where you want. However, for GFX (as an example) Jim Kasson has shown that up to a certain ISO (can't remember which one!) underexposing by a certain amount (let's say three stops) gives you the same result in the end as raising ISO to get the corresponding exposure. For example, if you normally shoot at ISO 100, raising ISO to 800 gives you the same result in a RAW file as underexposing by three stops.

It's amazing how new software tools can change accepted wisdom on questions like this. I use Lightroom, and I've noticed that AI noise reduction is so good that I no longer fuss about heavy lifts in the shadows, or shooting at higher ISO when I need a certain shutter speed and still want the image to be readable on screen (so underexposure is not an option).

 

[Pieter 12]

What camera are you using? Under exposure will lead to more noise when you lift the exposure to get back to where you want. However, for GFX (as an example) Jim Kasson has shown that up to a certain ISO (can't remember which one!) underexposing by a certain amount (let's say three stops) gives you the same result in the end as raising ISO to get the corresponding exposure. For example, if you normally shoot at ISO 100, raising ISO to 800 gives you the same result in a RAW file as underexposing by three stops.

It's amazing how new software tools can change accepted wisdom on questions like this. I use Lightroom, and I've noticed that AI noise reduction is so good that I no longer fuss about heavy lifts in the shadows, or shooting at higher ISO when I need a certain shutter speed and still want the image to be readable on screen (so underexposure is not an option).

It is a Leaf AFi II 7 back. It doesn't perform that well over ISO 200, I like to keep the ISO at 100. But it sounds like the noise would be the same as shooting at the higher ISO. Does the noise get reduced if you over expose and adjust in post then?

 

[rdeloe]

I don't know what would happen because that's a very different sensor technology.

Is it up and running? Easily the best way to answer your question is to see for yourself! Make some test shots of the same scene, one at ISO 200 underexposed two stops, and one at ISO 800 with the correct exposure. How does the ISO 200 version compare to the ISO 800 version when you lift the 200 version up two stops in post?

 

[Pieter 12]

I will.

 

[SrMphoto]

Most older sensors benefit from rising ISO (less noise).

 

[Paul2660]

Is the back CCD or CMOS. If CCD I would not expect higher ISO to help. I always felt the best thing for my P45+ was base ISO and maybe ISO 200 anything higher was just too full of noise. IQ160 was slightly better and when pixel binning was used gave good results to ISO 800 albeit at at loss of full resolution, instead of 60 Mp 15.

Paul

 

[tenmangu81]

Most modern sensors are ISO equivalent : you can compensate a low light either by increasing the sensor sensitivity (which is just an electronic amplification) on shooting, or by increasing exposure when post-processing. It works well until 3 or 4 EV.
For CCD sensors, it's not so obvious, though.

 

[SrMphoto]

Here are my comments on the topic and posts in this thread.
- Increasing ISO will decrease noise when keeping the exposure (shutter speed, aperture, and scene light) invariant.
- We are talking about ISO invariant ranges where ISO change does not significantly change the noise. Many modern sensors have an ISO invariant range above the dual conversion gain point, but many older sensors are not ISO invariant. PhotonsToPhotos publishes shadow improvement charts that document the ISO invariant ranges.
- The noise, and thus IQ, is determined by the exposure, not ISO. However, rising ISO decreases the maximum possible exposure. Shooting at base ISO allows for maximum possible exposure and, therefore, the least noise.
- ISO setting does not change sensor sensitivity.
- When ISO is within the ISO invariant range, we can keep ISO lower and the image darker without compromising IQ (OP’s question). However, if the sensor is not ISO invariant, keeping the ISO lower than possible (without increasing the exposure) will produce more noise.

 

[Paul2660]

My point is what is the sensor in the camera the OP is using. It appears to be an older Leaf back, and if it's CCD, ISO invariance doesn't matter. CCD backs (Phase used Koday and then Dalsa for the early IQ backs. They were pretty much a one ISO camera and pushing the ISO was shown to be pretty pointless in plenty of posts on this site and others. I believe the OP has a Dalsa sensor, thus CCD. On this forum years ago, Wayne Fox produced a very detailed post with a IQ180 (also Dalsa) and showed very clearly that pushing the ISO range on a CCD sensor was pretty pointless. His post should still be open on this site from 4 to 5 years ago and it had very detailed images.
The only CCD where this issue was somewhat not the same was the IQ260, where P1 tried to play some tricks to give the back an actual higher ISO range at full resolution. I owned one for several years and found that it was not the case.
Paul

 

[mristuccia]

I'm curious, how does this "invariance" concept of shooting darker photos at base ISO fit with the ETTR one which recommends exposing the image to the rightmost part of the histogram when it is possible?

 

[tenmangu81]

Actually, the most important feature is not noise absolute value, but signal/noise ratio. Because most part of the noise is random, increasing the exposure time by a factor of 4 will increase the S/N ratio by a factor of 2.

 

[rdeloe]

I'm curious, how does this "invariance" concept of shooting darker photos at base ISO fit with the ETTR one which recommends exposing the image to the rightmost part of the histogram when it is possible?

Coming at it from a purely practical perspective, I'm underexposing only because I need to prevent highlights from blowing out. In many cases, it's still ETTR because even 3 stops of underexposure (according to the meter) may be leaving areas where highlights are on the edge of blowing out. Even with 2.67 stops under, there's still lots of detail in the areas that look black in the unmodified RAW. Two stops under probably would have worked because I have a tiny bit of headroom on the highlight side of the histogram. In the camera's histogram (which is of course based on the JPEG) it looked like the highlights were blown.



We all find our own rhythm. With a GFX 100S, I try to stay at ISO 100. If shutter speed is a concern, I use whatever exposure I need even if that means underexposure, or sometimes I'll raise ISO if I find the dark image too annoying. My personal maximum ISO on a GFX 100S is 1250. After that ISO noise is ruining the image for me, although I should revisit that in light of how effective Lightroom's AI noise reduction tool is now.

 

[MartinN]

A ccd back needs lots of light, therefore a sturdy tripod and slow shutter speed or powerful studio strobes …

 

[SrMphoto]

I'm curious, how does this "invariance" concept of shooting darker photos at base ISO fit with the ETTR one which recommends exposing the image to the rightmost part of the histogram when it is possible?

Invariance comes into play when you cannot shoot ETTR at base ISO. Assume you have exposure where you can use ISO 1600 without clipping and are in the ISO-invariant range. That means you can shoot at ISO 800 or 400 with the same exposure (underexposing) without any IQ degradation.
To get the best IQ, you shoot ETTR at base ISO. Sometimes, you cannot saturate the sensor at base ISO, in which case you have an ISO choice to make. In the ISO invariant range, you can "underexpose" to avoid highlight clipping (safer).

 

[SrMphoto]

Coming at it from a purely practical perspective, I'm underexposing only because I need to prevent highlights from blowing out.
In many cases, it's still ETTR because even 3 stops of underexposure (according to the meter) may be leaving areas where highlights are on the edge of blowing out.

Yes, you are doing ETTR if you are just below clipping relevant highlights, even if that means applying negative EC.

Even with 2.67 stops under, there's still lots of detail in the areas that look black in the unmodified RAW. Two stops under probably would have worked because I have a tiny bit of headroom on the highlight side of the histogram. In the camera's histogram (which is of course based on the JPEG) it looked like the highlights were blown.

<skip>

We all find our own rhythm. With a GFX 100S, I try to stay at ISO 100. If shutter speed is a concern, I use whatever exposure I need even if that means underexposure, or sometimes I'll raise ISO if I find the dark image too annoying. My personal maximum ISO on a GFX 100S is 1250. After that ISO noise is ruining the image for me, although I should revisit that in light of how effective Lightroom's AI noise reduction tool is now.

High ISO does not increase noise; low exposure does. Because automatic metering couples ISO with exposure settings, it may look as if high ISO creates noise, but high ISO is just an indication that your exposure is low. Understanding the separation of ISO and exposure allows me to create technically better images.

 

[rdeloe]

High ISO does not increase noise; low exposure does. Because automatic metering couples ISO with exposure settings, it may look as if high ISO creates noise, but high ISO is just an indication that your exposure is low. Understanding the separation of ISO and exposure allows me to create technically better images.

You are correct of course. However, I find when I'm explaining things to people and I don't know how much background they have in the underlying technical details, it doesn't hurt to go with a more straightforward explanation that will mirror their actual experiences.

So yes, there is no such thing as an "exposure triangle" and anyone who says ISO is a variable in exposure is technically incorrect. However, for the practical photographer who isn't interested in the underlying electrical engineering, they can look at their images and see that high ISO produces images that have more noise. Personally, I'm not going to make a fuss as long as people understand (1) that increasing ISO does not magically make the sensor more sensitive to light, and (2) the underlying cause of the noise they are seeing is insufficient exposure.

For people who like to go into the weeds, this site is looking a bit dated now, but I think it still does a good job explaining what's going on without being overwhelming for most people. https://clarkvision.com/articles/iso/

 

[Pieter 12]

So I believe I have some sort of grasp on how the sensor (in my case, a CCD with a native ISO of 50) and increased ISO ratings (once again, in my case a max of 800) relate to IQ. What no one has put for so far is the advantages and or disadvantages of increasing exposure on a RAW file in post. Because, obviously, one is not increasing exposure it is a simulation of sorts, just as ISO beyond the sensor's base seems to be.

 

[SrMphoto]

So I believe I have some sort of grasp on how the sensor (in my case, a CCD with a native ISO of 50) and increased ISO ratings (once again, in my case a max of 800) relate to IQ. What no one has put for so far is the advantages and or disadvantages of increasing exposure on a RAW file in post. Because, obviously, one is not increasing exposure it is a simulation of sorts, just as ISO beyond the sensor's base seems to be.

If I understand Paul's explanation of the sensor behavior, you gain nothing by raising the ISO in the camera. I would keep it at base ISO and push it in the post.
Could you test it for yourself and let us know? Take two images of a ColorChecker or similar at base ISO and ISO 800 but at the same exposure, and lift the darker image in the post. Do you observe a difference?

 

[Pieter 12]

What I can do now is show a detail of the same image, as shot at ISO 100 and the exposure raised 2-3/4 stops in C1. Noise is apparent in both, but the +2-3/4 seems noisier, maybe some of that could be because it is less murky.