selmslie Loc: Fernandina Beach, FL, USA

We [should] already know that visible noise (signal to noise ratio aka S/N or SNR) is the determined by exposure, not ISO.

I have studied this at length and determined that S/N is primarily the result of shot noise. Sensor read noise does not begin to influence the SNR until well below middle gray (around EC-5 or darker).

If we develop an image from raw, there are times when you find that the original exposure may have resulted in a dark JPEG from the camera.

Increasing the ISO in the camera or reducing it and using the Exposure slider to offset it might produce the same image, assuming the camera is ISO invariant, which nearly all modern cameras are (but that's a different topic).

What happens to noise when you move the Exposure slider to the right to brighten the image? With each stop added via the Exposure slider we increase the apparent brightness by one stop. But we also double the effect of shot noise and this lowers the visible SNR.

So what would happen if we took two images using the same exposure (aperture and shutter speed) and equalized the brightness with the Exposure slider?

It has been suggested that the image that used the higher ISO would have less noise. Of course, we already know it would be better because it will be closer to ETTR (exposing to the right) but that may actually be because the shadows are recorded better.

To determine if there is really a measurable difference, I tested three cameras, a Fuji X100t, a Sony A7 II and III and a Nikon Z7. As luck would have it, all three are ISO invariant.

Below are the results of the test.

The samples for each camera used the same exposure (aperture/shutter speed). The SD column is from a 150x100 pixel selection at the center of the image.

The A7 II has not Bayer array so the effective ISO settings are actually one stop higher than what was set. ISO 25600 does not work right but all of the other ISO settings in the analog range are proportional and ISO invariant. The log(SD) changed by almost exactly one stop with each doubling of the ISO. For the other cameras I just tested two ISO settings.

As we can see, there may be a very small difference in the noise level changes but they are within a reasonable range of the actual difference in ISO.

If you have been following this so far, it conclusively shows that raising the ISO does not actually alter the amount of visible noise. It's safe to say that there is no benefit to lowering the ISO and using the Exposure slider to correct the apparent underexposure.

The higher exposure does not suppress any noise. In fact, with an ISO invariant sensor, it has absolutely no effect.

ETTR remains the recommended approach because it collects more clean information in the shadows, if you decide to amplify it.

User ID

Depends. On paper or in real world ?!?

And, acoarst, I did NOT read your essay and charts. I only noted its length and numerous charts. Youre just on the same old pointless fishing expedition as youre always on.

selmslie Loc: Fernandina Beach, FL, USA

It's pointless until someone suggests something that is counter-intuitive, which happens frequently.

The only way to settle such a question is with objective research.

I have presented these findings to demonstrate how such questions can be answered without posting links or opinions from outside sources who may be stating as fact something that they have never really examined carefully.

fredpnm Loc: Corrales, NM

I have found that Topaz DeNoise and even Lightroom's Denoise function do wonders for unwanted image noise...and that's even without all the dialog and charts...

I'm sure other programs will equally deal with unwanted image noise.

In this case and application knowing any of the science is generally of no value

User ID

Your same old fishing expedition as already seen over and over again, complete with the usual charts and such.

selmslie Loc: Fernandina Beach, FL, USA

The best way to deal with noise is to avoid it in the first place. That's easy to do if you do most of your photography in daylight. It's just a matter of getting a proper exposure.

The biggest mistake is to go looking for noise by viewing an image at 100% and then trying to fix something that nobody would have seen or cared about in a normal print or image.

selmslie Loc: Fernandina Beach, FL, USA

If it doesn't interest you, why bother to comment.

billnikon Loc: Pennsylvania/Ohio/Florida/Maui/Oregon/Vermont

Increase ISO, you increase noise. Decrease ISO, you decrease noise. This has always worked for me.

Juy Loc: Delaware

I may be thick headed but what is SD?

selmslie Loc: Fernandina Beach, FL, USA

If you increase the ISO you will probably reduce the exposure. It's the reduction in exposure that increases the noise.

All of the tests in the table were taken at the same exposure. After correcting for the underexposure, all of them ended up with the same amount of noise within +/- 0.03 stops (within the margin of error). The only one that is a little higher is the X100T (0.06 stops).

selmslie Loc: Fernandina Beach, FL, USA

SD is Standard Deviation.

A high SD means the noise will be easier to see.

When you increase the brightness with the Exposure slider you also increase the visible noise. The SD doubles with each stop of Exposure increase.

The same thing happens when you recover information from the shadows.

Juy Loc: Delaware

[quote=selmslie]SD is Standard Deviation.

Thank you

billnikon Loc: Pennsylvania/Ohio/Florida/Maui/Oregon/Vermont

Sorry, don't buy it.

selmslie Loc: Fernandina Beach, FL, USA

If all you capture are birds in daylight, you probably will never see noise in your images.

It might show up in the deep shadows but that's easy to cover up simply by raising the black point.

Otherwise, get yourself a copy of RawDigger and find out for yourself. It will be the best $20 you ever spent.

Rick from NY Loc: Sarasota FL

Why is anyone bothering to engage with this guy? He knows he’s right and we are all physics 101 morons. Replying to him is what gets him off.