ISO on the R5

BobKnDP wrote:

FingerPainter wrote:

BobKnDP wrote:

I presume that this is a trolling post.

Neither the post nor the video in it are trolling. Tony is perfectly correct.

The noisiness of an image depends on its Signal to Noise Ratio (SNR). The SNR depends primarily on how much light you captured in the photo. How much light you captured in the photo depends on the exposure, the surface area of the sensor, and the quantum efficiency of the camera. The latter two tend to be constants for a given camera. The exposure depends on:

1. the amount of light in the scene reaching the lens (scene luminance)
2. the T-stop of the lens (approximated by f-stop), and
3. the length of time the sensor is exposed to light (shutter speed)

The widely held view that the noisiness of an image depends on the ISO setting is an error caused by mistaking correlation for causation, and perpetuated by the omnipresent but fundamentally flawed conceptual model called the "Exposure Triangle".

I admit that I didn't watch the whole video at first. Comparing a greatly underexposed image with its brightness boosted to match a higher ISO shot really put me off. I presume that the quantization noise in the underexposed image was mostly responsible for the poorer SNR.

That's almost certainly not the case. The much lower exposure was responsible for the poorer SNR.

Nothing in the video is false, but it's playing with semantics. If you set the camera to a high ISO, and take a normal exposure with it, it will be noisier than a normal exposure taken at a lower ISO.

Yes, but why? It is because the shot taken with the high ISO captured much less light.

And speaking of semantics, what's a "normal exposure"?

The lower ISO exposure will be longer (by the ratio of the ISOs).

... assuming the camera is in control of the shutter speed.

It makes little sense to compare identical exposures with different ISOs.

It makes sense for two different reasons. One is to ascertain experimentally whether change in exposure or change in ISO setting is the actual cause of the change in SNR. The experiments show that it is change in exposure that is the actual cause. The other is to deal with practical matters of settings choices when an increase in ISO doesn't cause a change in exposure.

Here's an example. Suppose you are shooting a pair of fast-moving subjects that are not at the same distance from the camera. You need a certain minimum shutter speed to freeze subject motion adequately. You need a certain aperture to get both subjects in the field of acceptable sharpness. But when you use that shutter and aperture at base ISO you get an image that is too dark. Should you use the same shutter and aperture but increase the ISO setting to get correct image lightness?

People who mistakenly believe that raising ISO causes noisier images will tell you "No. Stay at base ISO and lighten the image in development". People who know that raising ISO doesn't increase image nosiness if exposure is left unchanged will tell you to go ahead and raise that ISO, so long as it doesn't cause clipped highlights. If you follow that latter advice you will actually get a slightly less noisy image on most digital cameras.

Another practical reason deals with use of Auto-ISO. People who wrongfully think that raising ISO itself causes noisier images fear using Auto-ISO because they fear that the camera may raise the ISO if Auto-ISO is set, and if it does, the image will be nosier than if the camera had stayed at base ISO and underexposed. In fact, Auto-ISO will only raise ISO when it cannot raise exposure any more, so the increase in ISO has no negative effect on image noisiness.