Overexposure and noise in the highlights?

[rexbobcat]

Just a question for those more technically minded folks.

As everyone knows, exposing to the right and bringing down the exposure minimizes shadow noise. However, if you overexpose at longer shutter speeds (where heat causes son noise) and at higher ISOs, does bringing the exposure down bring out noise in the highlights?

Not that it matters for any reason that would make a difference, but I'm just curious.

 

[Ysarex83 77]

My first reaction that has me scratchin' my head over this is the combination of overexposure and higher ISO. Raising the ISO is underexposing the sensor. So it wouldn't make sense to increase exposure if you've already decreased the exposure (raise ISO).

You referenced ETTR -- ETTR is nonsensical at ISO values above base. It's kinda like saying: I need to be one step to the left. So I'm going to move two steps to the left after I move one step to the right.

So I think you need to pull that higher ISO factor out of the equation and just look at the effect of long exposure/heat.

Joe

 

[rexbobcat]

I overexpose all the time with my X100S at higher ISOs because it reduces the overall apparent amount of noise in the image.

But fine, disregarding ISO entirely, will overexposing and correcting images bring out noise (even if negligible) into the highlights in the same way that bringing up shadows introduces noise?

 

[KmH]

However, if you overexpose at longer shutter speeds (where heat causes son noise) and at higher ISOs, does bringing the exposure down bring out noise in the highlights?

No.
Because thermal noise is not shot noise.

You would use dark-frame subtraction to reduce long exposure thermal noise.
In the dark portions of a digital image the main source of shot noise is dark current.
Dark current is a relatively small electrical current that flows in the photodiode portion of a pixel even if no photons are falling on the photodiode.
The photodiode develops a voltage proportional to the number of photons that hit it during and exposure. Transistors are also part of the pixel and the transistors amplify the voltage developed by the photodiode proportional to the camera ISO setting. The higher the ISO setting the lower the SNR (signal to noise ratio) and the more shot noise a digital photo will have.

This is a technical subject that is really worth understanding well if you want to push the limits of your camera's dynamic range and then try to undo the resulting problems post process.

I highly recommend the book Real World Camera Raw with Adobe Photoshop CS5 which explains the above in terms of exposure on pages 6-16. The guys that wrote the book, Bruce Fraser and Jeff Schewe, are the guys that wrote the software that is the Sharpening/Noise Reduction panel in ACR (Photoshop Camera Raw/LR Develop module)

A related Adobe white paper. Exposure and Linear Capture

 

[Ysarex83 77]

I overexpose all the time with my X100S at higher ISOs because it reduces the overall apparent amount of noise in the image.

Then just lower the ISO and you'll come out ahead with less analog signal amplification and so increased dynamic range.

But fine, disregarding ISO entirely, will overexposing and correcting images bring out noise (even if negligible) into the highlights in the same way that bringing up shadows introduces noise?

Noise from heat build up in the sensor during very long exposures is a real phenomena. We're generally talking exposures in the 30 second plus range and so the camera is tripod mounted. In that situation there's no advantage to raising the ISO. Interesting thought however which never occurred to me. Raising the ISO would shorten the exposure and so reduce the time for heat build up. However raising the ISO underexposes the sensor and we know that's going to cause noise. So is there a possibility that the noise from underexposure (raising the ISO) could be less than the noise from a heated sensor exposed longer at base ISO? Can't say I know for sure but my instinct says no way -- get the ISO to base and make the necessary exposure.

Joe

 

[KmH]

Raising the ISO can't change the voltage each pixel photodiode develops according to how much light hits the photodiode.
Raising the ISO can only change the bias on the transistors that amplify the tiny voltage a microscopic pixel photodiode can develop.
When little light is available, necessitating increasing the ISO, the transistor bias has to be changed so the tiny signal is amplified more.

So how is it that raising the ISO 'under exposes the sensor'?

Active pixel sensor - Wikipedia, the free encyclopedia

 

[Ysarex83 77]

Raising the ISO can't change the voltage each pixel photodiode develops according to how much light hits the photodiode.
Raising the ISO can only change the bias on the transistors that amplify the tiny voltage a microscopic pixel photodiode can develop.
When little light is available, necessitating increasing the ISO, the transistor bias has to be changed so the tiny signal is amplified more.

So how is it that raising the ISO 'under exposes the sensor'?

Active pixel sensor - Wikipedia, the free encyclopedia

Because, raising the ISO can't change the voltage each pixel photodiode develops according to how much light hits the photodiode.

Joe