Overexpose to reduce noise

Pippan wrote in post #18976672
When there's enough light.

Yes, but in the real world, how often does the presentation of the question of whether or not to use a higher ISO setting cause one to suddenly decide to give more real exposure? Your initial choice of Av and Tv values was chosen for a reason, and if you change it, it is not because you could have just raised the ISO with the same original settings, too. There should be no choice between these options; they are not mutually exclusive.

You get more total light by increasing actual exposure, and you get a little less noise by keeping the exposure but raising the ISO setting.

Pippan wrote in post #18976798
Has true ISO invariance been achieved yet?

It's a matter of degree. The only cameras that are totally unaffected by ISO setting for a manual exposure are those that don't use variations in analog gain, like the Sigmas and some early medium format cameras.

If any camera uses more gain at ISO 200 than it does at ISO 100, then it will probably do a push to ISO 100,000 better from the 200 setting than from the 100 setting, even if you don't notice a difference between pushing to 100,000 from 3200 vs 6400, or using an ISO 200 exposure from 100 vs 200.

The character of the post-gain noise is very important here. Any camera with relatively fine, random post-gain read noise will be more visibly ISO-invariant in the high ISO ranges than a camera with blotches and banding in that post-gain noise. At high ISOs, the relative strength of pre-gain to post-gain read noise is at a very high ratio, but if the post-gain read noise has blotches and banding, and the pre-gain does not or has much less, then that post-gain noise can be much more visible than its measurements would suggest. The Nikon D5 is a great example - it has very low pre-gain read noise, and this dominates the noise character up until about ISO 208K or so, and then once the camera gets above 208K, depending on the color quality of the light, horizontal banding starts to quickly dominate, because the analog gain stops at 102,400, but is really needed to go higher in that camera for the high ISO settings that it offers.

John Sheehy wrote in post #18976882
Yes, but in the real world, how often does the presentation of the question of whether or not to use a higher ISO setting cause one to suddenly decide to give more real exposure? Your initial choice of Av and Tv values was chosen for a reason, and if you change it, it is not because you could have just raised the ISO with the same original settings, too. There should be no choice between these options; they are not mutually exclusive.

You get more total light by increasing actual exposure, and you get a little less noise by keeping the exposure but raising the ISO setting.

In my opinion there are numerous cases in which these choices are not so rigid. I may have originally chose a shutter speed is x for a reason, but that doesn't mean I can't get the same overall image with a longer shutter speed (or shorter). To some extent aperture has a "liveable" range for the effect you are trying to create as well. You can make choices but you are allowed to change your mind either for artistic purposes ... or because you want to add a little overall exposure.

Pippan wrote in post #18976800
Actually I look forward to that day. Then ISO for raws would be irrelevant and optimum brightness realised in processing with no noise, or loss of DR or colour quality.

There have already been cameras with no variable gain, and cameras in which variable gain is pointless. The Canon G9 I bought years ago uses variable gain for all ISO settings, and looks pretty much the same under-exposing its ISO 80 by 4.33 stops as it does shooting at ISO 1600, despite 4.33x as much gain at ISO 1600, but that fact just means that you don't need to expose RAWs to the right at a higher ISO setting, and has no reflection on the amount of noise. "ISO invariance" does not mean low noise, compared to other cameras. It is more about making exposure/ISO decisions, given a certain camera in use.

Camera A could be completely "ISO invariant" and camera B could be very "ISO-variant", and camera B might still give less noise in all cases with the same exposure, regardless of ISO setting.

John Sheehy wrote in post #18976882
Yes, but in the real world, how often does the presentation of the question of whether or not to use a higher ISO setting cause one to suddenly decide to give more real exposure? Your initial choice of Av and Tv values was chosen for a reason, and if you change it, it is not because you could have just raised the ISO with the same original settings, too. There should be no choice between these options; they are not mutually exclusive.

You get more total light by increasing actual exposure, and you get a little less noise by keeping the exposure but raising the ISO setting.

Well John, my initial choice is almost never Av or Tv, it's M without auto-ISO, but I live in a very sunny place* with usually so much light that even with a CPL filter at ISO100 I normally have plenty of leeway with shutter speed and/or aperture. So only if it gets darker and I can't slow SS or open aperture any more, only then do I increase ISO to take the signal into the right hand sector of the histogram.

*I don't often have time to participate in these sort of discussions, though I learn a lot from them. However, where I live is so sunny I've just had an operation to remove multiple skin cancers and I can't do much. 'Stirring the possum' here is keeping me entertained.

jcothron wrote in post #18976894
In my opinion there are numerous cases in which these choices are not so rigid. I may have originally chose a shutter speed is x for a reason, but that doesn't mean I can't get the same overall image with a longer shutter speed (or shorter). To some extent aperture has a "liveable" range for the effect you are trying to create as well. You can make choices but you are allowed to change your mind either for artistic purposes ... or because you want to add a little overall exposure.

I didn't say that you can't change your mind; what I said is that choosing the actual exposure is not mutually exclusive with changing the ISO setting bias against whatever actual exposure you have chosen. IOW, "should I change the exposure or should I change the ISO bias?" is not a practical question; it is an artificial academic construct. If one understands what they are doing, they will choose the exposure they want, and then, on top of that, they have the opportunity to also expose to the right at a higher ISO if they feel that there is a benefit with their camera.

By ‘eck, some people aren’t ‘half over complicating a very simple idea.
What does it matter which of the three variables you use to over expose the image? Obviously you would use the one that has the least effect on the type of image that you are taking.


For example: a floodlit building at night. You would choose an aperture that gives you the required depth of field and a low ISO. The shutter speed has the least effect on the image so that is what you would use to overexpose the image.

Another example: a football match played under floodlights. You are going to choose a shutter speed that will freeze the action and an aperture that will isolate the player from the background, that means that the only variable is the ISO. So you would adjust the ISO to overexpose the image.

An example of where you would use Aperture to increase the exposure would be a panning shot, because depth of field is largely irrelevant in that instance.

Pippan wrote in post #18976897
Well John, my initial choice is almost never Av or Tv,

"Choice of Av and Tv values" includes M mode. What are you setting the dials to in M mode? Av and Tv "modes", which imply "Av-priority" and Tv-priority", are not the same thing as Av and Tv values. Every camera operation in any exposure modes involve Av and Tv values.

That's the edge or wall case, where the camera runs out of ISO setting range at the low end. Such conditions could allow ISOs of around 1 (one) if the cameras had it, with a fast lens on a sunny day (1/80 at f/1.4). Operating against that wall has a different set of considerations than floating within a range of elevated ISO settings. Against the wall, you are wasting light to avoid over-exposure. Away from the wall, light is precious, and how much you want to capture comes first, and then choosing the most advantageous ISO bias is an option with possible benefit, depending on what tonal levels you need, and what camera you are using.

Well, that's what I thought we were talking about, mostly. The value of a RAW histogram still applies, though, against the wall or floating at higher ISO settings. It's just that at the wall, it is always actual exposure being adjusted when you expose to base ISO's right or left. Away from base ISO, the decisions uncouple.

john crossley wrote in post #18976901
By ‘eck, some people aren’t ‘half over complicating a very simple idea.
What does it matter which of the three variables you use to over expose the image? Obviously you would use the one that has the least effect on the type of image that you are taking.


For example: a floodlit building at night. You would choose an aperture that gives you the required depth of field and a low ISO. The shutter speed has the least effect on the image so that is what you would use to overexpose the image.

Another example: a football match played under floodlights. You are going to choose a shutter speed that will freeze the action and an aperture that will isolate the player from the background, that means that the only variable is the ISO. So you would adjust the ISO to overexpose the image.

An example of where you would use Aperture to increase the exposure would be a panning shot, because depth of field is largely irrelevant in that instance.

It matters because only 2 of the 3 factors actually deal with light, the other factor is a "volume knob" to amplify/magnify the perception of that light. That is the crux of the discussion for the past 20 years, how one juggles the two that actually impact exposure with the one factor that is an amplifier and all the grief that might come with that.

Many, many times, one cannot just "choose an aperture" because they will likely hit the widest aperture the lens allows. Ditto with shutter, one cannot flexibly change the shutter due to what is being shot.

In sports with poor lighting, you don't choose an aperture for DOF control, because you will likely ALWAYS be wide open to allow room for a fast shutter and lower ISO. However in many situations, the lighting is so poor that even with an f1.8 lens and ISO of 6400, you still have too slow a shutter. Drama, concerts, comedy shows, etc all are related.

The mistake that is made is when folks are under the impression that higher ISOs create much more unmanageable noise, so they underexpose and/or slow down the shutter, which means two things in the sports area anyways:
- They are going to multiply the noise by changing the exposure in post (which means you are not doing analog gain but rather mathematical gain, and analog gain is preferable)
- They get blur and think there is nothing that can be done, or the lens is to blame, or the camera.

In higher ISO situations, ETTR and ITTR are your friends, go to a higher ISO than one is comfortable with in order to expose to the right to give you the latitude in shutter speed (because you will already be aperture-constrained). In post processing, bring the exposure DOWN, then clean up the noise. This will almost ALWAYS give a superior result to the alternative. I personally think it is ALWAYS superior but I am willing to consider there might be a few situations where that could not be the case.

This is why a fully working auto ISO is so nice. You can lock down your other two parameters, and set up EC to the right by 1/3 or 2/3, set up the metering mode to accommodate that EC, and let the ISO float around as needed. It makes HAMSTTR pretty easy now. Add in highlight priority on the newer bodies like the 7D2 and 5D4, and you gain additional benefits, something I learned from John, and now it is part of my sports setup every time.

This setup on the 5D4 has allowed me to shoot JPG for sports (I also have raw available if needed), and then I do a small amount of image cleanup for post. I allow the camera to shoot up through ISO 32000 this way and within an hour, I have 300-400 photos ready to go. Before on other bodies, I had to work from the raw to JPG, and it was about 2-3x the work.

This is a 75% crop from an image, from a JPG from the 5D4 using these parameters, with minor cleanup. No undue noise, skin/peach fuzz/jersey fabric all intact, highlight/shadow differences are well controlled, etc IMO. I shoot these about 2/3 EC, and ISO floats around to accommodate my desires.

TeamSpeed wrote in post #18976928
In higher ISO situations, ETTR and ITTR are your friends, go to a higher ISO than one is comfortable with in order to expose to the right to give you the latitude in shutter speed (because you will already be aperture-constrained), and then bring the exposure DOWN in post, then clean up the noise. This will almost ALWAYS give a superior result to the alternative. I personally think it is ALWAYS superior but I am willing to consider there might be a few situations where that could not be the case.

That can be true so long as the camera is actually increasing gain for higher ISOs. Once the camera passes that point, there is no reason to risk highlights, no matter how much post-gain read noise a camera has. Canon sensors tend to only go a stop past max analog gain, if at all, but something like the Nikon D5 goes 5 stops past max analog gain. If I had a D5, I would never let the ISO setting be as high as the camera goes. At top ISOs, the histogram of read noise doesn't even fit in the RAW value range, so noise is clipped and then filtered, and average signal levels become non-linear, making color more problematic
than it needs to be.

Of course, HTP offsets your RAW highlight headroom in combination with any EC setting you use. If you use +2/3, combined with HTP, you are effectively at -1/3. The difference is, HTP makes images a stop brighter than the same actual exposure combined with the same ISo value would otherwise give. Of course, Canon has decided to cripple HTP by limiting it to a lower maximum ISO value than non-HTP, which doesn't make a single stitch of sense; HTP has the least (if any) downside at the highest ISOs, and there is no reason why once a camera reaches maximum analog gain, that it can't use 1-stop HTP for the next stop higher ISO, 2 stops of extra highlights for the next stop, etc. That would offer smaller files, actually, with the RAW file compressions that Canon uses.

TeamSpeed wrote in post #18976928
It matters because only 2 of the 3 factors actually deal with light, the other factor is a "volume knob" to amplify/magnify the perception of that light. That is the crux of the discussion for the past 20 years, how one juggles the two that actually impact exposure with the one factor that is an amplifier and all the grief that might come with that.

Many, many times, one cannot just "choose an aperture" because they will likely hit the widest aperture the lens allows. Ditto with shutter, one cannot flexibly change the shutter due to what is being shot.

In sports with poor lighting, you don't choose an aperture for DOF control, because you will likely ALWAYS be wide open to allow room for a fast shutter and lower ISO. However in many situations, the lighting is so poor that even with an f1.8 lens and ISO of 6400, you still have too slow a shutter. Drama, concerts, comedy shows, etc all are related.

The mistake that is made is when folks are under the impression that higher ISOs create much more unmanageable noise, so they underexpose and/or slow down the shutter, which means two things in the sports area anyways:
- They are going to multiply the noise by changing the exposure in post (which means you are not doing analog gain but rather mathematical gain, and analog gain is preferable)
- They get blur and think there is nothing that can be done, or the lens is to blame, or the camera.

In higher ISO situations, ETTR and ITTR are your friends, go to a higher ISO than one is comfortable with in order to expose to the right to give you the latitude in shutter speed (because you will already be aperture-constrained). In post processing, bring the exposure DOWN, then clean up the noise. This will almost ALWAYS give a superior result to the alternative. I personally think it is ALWAYS superior but I am willing to consider there might be a few situations where that could not be the case.

This is why a fully working auto ISO is so nice. You can lock down your other two parameters, and set up EC to the right by 1/3 or 2/3, set up the metering mode to accommodate that EC, and let the ISO float around as needed. It makes HAMSTTR pretty easy now. Add in highlight priority on the newer bodies like the 7D2 and 5D4, and you gain additional benefits, something I learned from John, and now it is part of my sports setup every time.

This setup on the 5D4 has allowed me to shoot JPG for sports (I also have raw available if needed), and then I do a small amount of image cleanup for post. I allow the camera to shoot up through ISO 32000 this way and within an hour, I have 300-400 photos ready to go. Before on other bodies, I had to work from the raw to JPG, and it was about 2-3x the work.

This is a 75% crop from an image, from a JPG from the 5D4 using these parameters, with minor cleanup. No undue noise, skin/peach fuzz/jersey fabric all intact, highlight/shadow differences are well controlled, etc IMO. I shoot these about 2/3 EC, and ISO floats around to accommodate my desires.

In all honesty, that is just a very long winded version of what I said in my previous post.

Pippan wrote in post #18976798
Has true ISO invariance been achieved yet?

Some Nikon, Sony and Fuji cameras have been claimed to be. Maybe not to the full extent you required but to my untrained eyes from what I've seen still great.