Under Expose for Lower ISO: adjust levels later, BETTER results???

Started 3 months ago | Discussions thread
Dale Cotton
Senior MemberPosts: 1,904
Like?
Re: Under Expose for Lower ISO: adjust levels later, BETTER results???
In reply to elliottnewcomb, 3 months ago

I did a few quick tests shooting raw+JPEG in high contrast lighting keeping aperture and shutter the same but varying the ISO. What I'm seeing is that the higher the ISO the more the bright areas blow but the the more darker areas are cleaner. Conversely, the low ISO shots with underexposure hold the highlights better but the darker areas are dramatically worse.

As I understand it, engineers have a number of variables to work with when dealing with increases in ISO. The main one is to boost the gain on the electrical signal from the sensor proportionately to the increase in ISO. So a change from ISO 400 to 800 is a one stop change in exposure, which doubles the signal strength. If this boost is applied equally to all pixels then those that are already near overexposure will be pushed into overexposure. Sort of like adding ice to a full glass of water -- some of the water will slop out over the top.

The upshot is that a simple boost in gain costs you a stop of dynamic range for every stop of ISO increase you apply.

A high contrast scene contains areas that are dark, bright, and in the middle. While the bright areas may be on the verge of blown highlights (overexposure) or already blown, the darker-than-midtone areas are, by their very nature, underexposed. So using a low ISO setting with underexposure via aperture or shutter speed means that the darker areas will get even less light than they would otherwise. This in turn means that they will be more damaged by noise than otherwise.

The upshot is that a stop of underexposure gives you half the light in all areas, including dark areas, which simplistically means a doubling of the noise in the dark areas.

So simplistically, you're trading off dynamic range for noise levels. Pick your poison. However, this logic does not account for the fact that camera engineers have other tools in their toolkit to help lessen the impact of each increase in ISO. Here is a link to the relevant DxOMark RX100 measurements. You have to choose which graph you want to look at from there, so look at the SNR (signal-to-noise) and Dynamic range graphs. If I understand the numbers correctly -- and that's a big if given my shaky math skills -- it looks as though the loss of DR varies from one ISO to another. And similarly, the gain in noise varies from  ISO to ISO. For example, if you switch from 800 to 1600 ISO the decibel number drops roughly by 3; but if you switch from 1600 to 3200 decibels drop by 4. Conversely, if you switch from 800 to 1600 ISO your DR drops from 10.3 to 9.1; but if you switch from 1600 to 3200 your DR drops from 9.1 to 8.3.

So in theory you could base your ISO vs. underexposure decisions on those particulars (sort of like a blackjack fanatic trying to game the house by memorizing tables of numbers). But the general rule based both on the DxO data and on my quick tests seems to be as simple as -- you're trading DR for noise, take your pick based on which is more valuable in each situation.

I'll be very interested to see to what degree our more technical forum members will shoot holes in this chain of reasoning!

Reply   Reply with quote   Complain
Post (hide subjects)Posted by
Yes!New
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow