Exposing to minimize noise (when using high ISO)

[unpopular]

There is the same amount of noise at ISO 100 as there is at any higher ISO, the difference is that there is more signal relative to noise. By increasing exposure, you are increasing signal relative to the noise which is constant.

I suppose that in very long exposures you may risk thermal noise accumulation.

 

[analog.universe]

No. There is the same amount of noise at ISO 100 as there is at any higher ISO, the difference is that there is more signal relative to noise. By increasing exposure, you are increasing signal relative to the noise which is constant.

I suppose that in very long exposures you may risk thermal noise accumulation.

Newer bodies even have a built in "dark frame" compensation for long exposure thermal noise. An exposure is taken with identical settings to the one you've asked for, except the shutter is left closed, giving only an image of the noise in the current thermal and electromagnetic environment which is then subtracted from the original. Much more effective than algorithmic noise reduction in post.

 

[unpopular]

My a350 has this feature.

It really only works for stuck pixels afaik. Thermal noise is stochastic and can't be predicted. But either way ETTR doesn't make a lot of sense in low light.

 

[MLeeK]

Does exposing to the right somehow eliminate noise associated with high ISO?If you are at ISO2000 f/2.8 1/100 and have an even meter, upping the ISO to 3200 to expose to the right is only going to introduce even more noise into your shot. Which I do not agree with.

When you under expose you introduce more noise by boosting in post. When you overexpose you hide noise when you reduce exposure in post.

 

[unpopular]

^^ yes, except that ISO adjustments are done at the sensor's analog amplifier, not in post.

I suppose there is some noise introduced by the amplifier as well, but this should be pretty nominal.

 

[matthewo]

Yeah Active d lighting is not very useful and very counter productive if your shooting raw and editing in anything but nikon software. With your ADL on auto inside its going to basically be on high. Problem is it under exposes your raw file and brings back shadow areas. Your pictures probably look decent on the camera lcd, but in lightroom look underexposed, thats cause lightroom cannot read ADL and isnt highlighting your shadows.

I always shoot raw, and turn every control feature on the camera off that doesnt effect the raw file, and cannot be read by lightroom. Other option is shoot raw+jpg and leave the options and just use jpgs that use ADL. But i dont like my camera processing pictures if im going to. I want a clean unaltered slate.

So again ADL does affect a RAW only by under exposing it around 1/2 stop in some instances

That is the Expose To The Right philosophy and by what I have read it makes sense and seems to work quite well. Read This Article.

I've been researching this pretty heavy over the last couple of weeks trying to figure out how to improve my indoor sports shots. That article is about the most succinct and helpful thing I've found so far. Thanks for the pointer!

To follow up and flesh this out: I'm shooting with a Nikon D7000 and a Sigma 70-200 f/2.8 (non OS version). Up until now, I've been shooting in shutter priority mode with ISO fixed at 3200. I've also been leaving Active D Lighting (ADL) set to Auto. AF-C with spot focus. Matrix metering. No exposure compensation.

My shots are always under exposed by at least a full stop, and raising the exposure in LR3 yields horrible results. Very excessive noise and very flat looking images. From my reading, it looks like ADL is partly to blame.

This weekend, I'm planning to make the following changes: Shoot in program auto mode with shutter speed and f/stop both forced. ADL off. Auto ISO on and capped at 12,800. Exposure compensation +0.7. Center weighted metering. AF-C and 21 point dynamic area focusing.

Any comments/criticisms of my thinking here?


WhiskeyTango

 

[Dominantly]

Does exposing to the right somehow eliminate noise associated with high ISO?If you are at ISO2000 f/2.8 1/100 and have an even meter, upping the ISO to 3200 to expose to the right is only going to introduce even more noise into your shot. Which I do not agree with.

When you under expose you introduce more noise by boosting in post. When you overexpose you hide noise when you reduce exposure in post.

Yes if you favor under exposing then trying to fix it in post, you are again being counter productive and will gain noise. Increasing signal gain to gain more light is my least favorite way of doing so. I can accept an ETTR at base or close to ISO, but not at very high ISO

 

[EchoingWhisper]

Expose to the right with high ISO works on Canon sensors which have crazy high read noise but with Sony sensors, just expose to the right with shutter speed and aperture, cause' increasing ISO will do nothing good to your picture other than clipping highlights.

 

[MLeeK]

Does exposing to the right somehow eliminate noise associated with high ISO?If you are at ISO2000 f/2.8 1/100 and have an even meter, upping the ISO to 3200 to expose to the right is only going to introduce even more noise into your shot. Which I do not agree with.

When you under expose you introduce more noise by boosting in post. When you overexpose you hide noise when you reduce exposure in post.

Yes if you favor under exposing then trying to fix it in post, you are again being counter productive and will gain noise. Increasing signal gain to gain more light is my least favorite way of doing so. I can accept an ETTR at base or close to ISO, but not at very high ISO

When you get to ISO 3200, 6400 and are JUST getting proper exposure-you are better off to go up one. It's not a matter of RAISE YOUR ISO!!! It's more of when you get to the point you know you have ISO noise in your images go one step more in order to hide it or reduce it in post processing and keep it from happening in abundance in your mid/top end shadows.

 

[EchoingWhisper]

When you under expose you introduce more noise by boosting in post. When you overexpose you hide noise when you reduce exposure in post.

Yes if you favor under exposing then trying to fix it in post, you are again being counter productive and will gain noise. Increasing signal gain to gain more light is my least favorite way of doing so. I can accept an ETTR at base or close to ISO, but not at very high ISO

When you get to ISO 3200, 6400 and are JUST getting proper exposure-you are better off to go up one. It's not a matter of RAISE YOUR ISO!!! It's more of when you get to the point you know you have ISO noise in your images go one step more in order to hide it or reduce it in post processing and keep it from happening in abundance in your mid/top end shadows.

There is almost no point of increasing ISO to ETTR. ISO is not a part of exposure. Increasing ISO only throws away the highlights while increasing noise. Unless your camera is still using those old sensors with high read noise like Canon's, increasing ISO and amplifying data in post would have almost no difference.

 

[EchoingWhisper]

Yeah Active d lighting is not very useful and very counter productive if your shooting raw and editing in anything but nikon software. With your ADL on auto inside its going to basically be on high. Problem is it under exposes your raw file and brings back shadow areas. Your pictures probably look decent on the camera lcd, but in lightroom look underexposed, thats cause lightroom cannot read ADL and isnt highlighting your shadows.

I always shoot raw, and turn every control feature on the camera off that doesnt effect the raw file, and cannot be read by lightroom. Other option is shoot raw+jpg and leave the options and just use jpgs that use ADL. But i dont like my camera processing pictures if im going to. I want a clean unaltered slate.

So again ADL does affect a RAW only by under exposing it around 1/2 stop in some instances

That is the Expose To The Right philosophy and by what I have read it makes sense and seems to work quite well. Read This Article.

I've been researching this pretty heavy over the last couple of weeks trying to figure out how to improve my indoor sports shots. That article is about the most succinct and helpful thing I've found so far. Thanks for the pointer!

To follow up and flesh this out: I'm shooting with a Nikon D7000 and a Sigma 70-200 f/2.8 (non OS version). Up until now, I've been shooting in shutter priority mode with ISO fixed at 3200. I've also been leaving Active D Lighting (ADL) set to Auto. AF-C with spot focus. Matrix metering. No exposure compensation.

My shots are always under exposed by at least a full stop, and raising the exposure in LR3 yields horrible results. Very excessive noise and very flat looking images. From my reading, it looks like ADL is partly to blame.

This weekend, I'm planning to make the following changes: Shoot in program auto mode with shutter speed and f/stop both forced. ADL off. Auto ISO on and capped at 12,800. Exposure compensation +0.7. Center weighted metering. AF-C and 21 point dynamic area focusing.

Any comments/criticisms of my thinking here?


WhiskeyTango

I always set ADL on extra high. Why? It often helps preventing highlights from being blown out even if I'm shooting RAW. Try to check your meter with your lens cap on while changing ADL from extra high to none and none to extra high, you'll see your metering change.

 

[unpopular]

If you take an exposure at ISO 800 using the same shutter speed and aperture as you had at ISO 200, and nothing is blown then you would have less noise, though brighter image, than if you took the photo at ISO 800 and compensated exposure by two stops and maintained the same equivalent exposure value.

However, you will always have more noise at a higher ISO than at a lower ISO.

I believe, however, the issue is using ETTR while at a high ISO, and not using ISO to implement ETTR. Any net improvement in noise performance by increasing exposure would be cancelled in noise introduced by gain.

 

[analog.universe]

I believe, however, the issue is using ETTR while at a high ISO, and not using ISO to implement ETTR. Any net improvement in noise performance by increasing exposure would be cancelled in noise introduced by gain.

As far as straight analog sensor noise, yes, this is true. The point is that if you boosted the ISO, you have better digital data and are more equipped to handle that noise. You have more shadow information, and have more latitude before you run into quantization errors, thus making your algorithmic noise reduction more effective relative to the amount of detail you give up to it.

 

[Big Mike]

Does exposing to the right somehow eliminate noise associated with high ISO?If you are at ISO2000 f/2.8 1/100 and have an even meter, upping the ISO to 3200 to expose to the right is only going to introduce even more noise into your shot. Which I do not agree with.

When you under expose you introduce more noise by boosting in post. When you overexpose you hide noise when you reduce exposure in post.

Yes if you favor under exposing then trying to fix it in post, you are again being counter productive and will gain noise. Increasing signal gain to gain more light is my least favorite way of doing so. I can accept an ETTR at base or close to ISO, but not at very high ISO

I get what you're saying...and I don't disagree. But the message behind ETTR, that often gets missed, is that underexposing and fixing is post, is the worst thing you can do. So rather than under expose, raise your ISO (thus moving the exposure to the right) and you will end up with less noise than if you had tried to raise the brightness in post.

But yes, if you can reach proper exposure at an ISO that you're comfortable with...then feel free not to raise it.
But some people have found that even lower noise levels can be accomplished (in the final image) by raising the ISO (thus exposing more to the right) and then normalizing the image is post. I'm sure there is a limit, for each camera/sensor/processor where raising the ISO is a worse option...and that's for everyone to experiment with and find what works for them.

 

[unpopular]

I believe, however, the issue is using ETTR while at a high ISO, and not using ISO to implement ETTR. Any net improvement in noise performance by increasing exposure would be cancelled in noise introduced by gain.

As far as straight analog sensor noise, yes, this is true. The point is that if you boosted the ISO, you have better digital data and are more equipped to handle that noise. You have more shadow information, and have more latitude before you run into quantization errors, thus making your algorithmic noise reduction more effective relative to the amount of detail you give up to it.

I don't believe you!

But. I just ran a test, and I'll get back to you.