Over the previous 16 posts, plus the work that I‘ve done on expose-to-the-right (ETTR) (you can see it here), I’ve demonstrated – to myself, at least – several things:
- ETTR can produce low noise results when employed at or near the base ISO of the camera.
- It’s possible to mess with the camera settings, particularly the in-camera white balance, and obtain credible approximations to the actual raw histogram, allowing ETTR with some precision.
On the five cameras I’ve tested (the Nikon D4 and D800E, the Leica M9, the Sony NEX-7 and RX-1), there are small to zero (to negative, in the case of very high ISO settings) gains in the signal-to-noise ratio (SNR) to be had by cranking up the ISO setting in the camera over its base level. Thus, ETTR is only an effective way to improve the SNR in the final image if you leave the in-camera ISO setting at the base level.
Unfortunately, there are many photographic situations where ETTR at base ISO produces unacceptable f-stops or shutter speeds. Dim light, camera or subject motion, depth-of-field considerations, and other practicalities conspire to keep the photographer from getting enough photons to the sensor. What’s the best strategy in that all-too-common circumstance? There are two poles.
The first is to set an acceptable combination of f-stop and shutter speed, then crank up the ISO to keep the histogram on the right. This has the advantage of allowing the photographer to use the same technique regardless of the lighting level, but it has some disadvantages.
- The sacrifice of headroom and margin for error with no gain in image quality.
- Slower, more fiddly exposure calculations (trial exposures, spotmetering, etc.) when they’re not necessary.
The opposite approach is to leave the ISO setting in the camera at base ISO, and when it’s impractical to expose to the right, just underexpose and fix things in your favorite raw development program. This also has disadvantages:
- The preview image may be too dark to see.
- In the case of cameras with electronic viewfinders, the finder may be useless, or nearly so.
- The photographer may be unaware of how far underexposed the shot is, and not know if acceptable results can be obtained in post=processing.
- The SNR may be slightly short of optimum (this is the least significant problem, and the effects may be so small that it’s unreasonable to consider it a problem at all).
As is true with most things in life, the extreme approaches are not the best ones. I’m sure there are many middle-ground solutions that will work well. In the next post, I will explore a possible centrist solution.