Underexposure DOES NOT PRODUCE NOISE

[William Michael]

3 hours ago, digitaldog said:

What is a simple sentence or maybe paragraph to provide to those photographers that seem to think, higher ISO produces the noise? Clearly, this is an exposure issue. If underexposure doesn't produce noise, what is the message that can be stated that is both easy for them to understand and makes the scientist happy? Is that even possible? 

One answer for your consideration - 

When teaching High School Students, and also at Master Classes, especially those which have been discussing and centred upon Available Light Photography, I have avoided the ISO and its relationship to S/N ratio and taken a ' final image outcome approach'.

I use phrases similar to:

"When using High ISO it is very important to nail the 'correct' exposure. This is the exposure which is most suitable for the best quality Final Image. The reason is, when we are using High ISO and underexposing, contrasted to using Low ISO and underexposing: if all else is 'equal' an High ISO and underexposing will increase the appearance of noise in the final image.

"One practical method we can use at the beginning of the Image making process is to Bracket the exposure."     

WW

[rconey]

I feel like I could get flamed really badly here, but.....

This discussion feels as if it is being held in a laboratory.

When I am out hiking, and decide to take an image, I am not thinking about noise. I am thinking about all the technique I learned over the years. Expose to the right, without overexposure (well, mostly, Dog). Bracket. Should use a tripod more but with image stabilization I do that less. Use base iso as much as possible. 

When I get home, I pull up what I have and do the OCD thing and sort through it all. I do blow up magnification to look at noise levels. Throw away most of the images (mostly junk) as Galen Rowell suggested. 

All I can pull out of this discussion is that if I underexpose, signal is reduced and noise is constant. If I amplify  by dialing up exposure I see more relative noise in the underexposed image. Gee, wish I had bracketed better.

Sorry to interrupt the technical discussion.

[digitaldog]

7 minutes ago, rconey said:

I feel like I could get flamed really badly here, but.....

No reason to do so. The reason for the discussion is mainly to define the reason for the noise and dismiss the idea "High ISO causes noise". 

Bracketing: If you can do it, great! Often I can't if I am shooting portraits, sports, action, etc. That's a good deal of what I shoot (note, I shot the 1984 Olympics on Transparency film: no Histograms, nothing that could produce any more than a few frames per second etc: IOW, an old fart). But sure, if you can bracket, go for it. 

Noise isn't a major factor again, thanks to new processing like Denoise. But if I can avoid it in the first place with optimal exposure, I will (and can). 

The goal isn't to make it (overall) technical. The goal is a sentence or two, for those who keep thinking and saying that they love NR because they shoot high ISO. That isn't the cause of the noise; how do we educate them without getting too technical, which I think would ideally bypass, if possible S/N ratios, photons, amplification etc. Might not be possible. 

[digitaldog]

With the generous help (privately) of the fellow (who should go unnamed) that started this by suggesting my text was iffy**, we have come to a better understanding (for me) to come up with this simple statement:

The relative amount of noise changes with exposure. Underexposure produces more noise. If you underexpose, you'll produce more noise.

The statement below was, from his scientific perspective, sloppy. And I now have to agree. But I'm still open to comments unless we can put this to bed.  

**My iffy/sloppy statement was: Underexposure causes noise. His correction was Underexposure DOES NOT PRODUCE NOISE.

[conrad_hoffman]

I think this whole thing can be clearer with some diagrams and analogies. Sensor is a bucket with some mud in the bottom. Fill the whole bucket with water and the concentration of mud is low. That's full exposure and low amplifier gain (low ISO.) Decide to fill the bucket only a couple inches, because that's all the water you have available (low light), and the concentration of mud gets a lot more noticeable. You need high amplifier gain (high ISO) to match that smaller signal (plus the noise) to the full value of the output file.  Or something like that.

[conrad_hoffman]

Something like the attached? It can't be to scale because our noise levels are so low. I don't know anything about current CMOS sensors and the data sheets aren't usually published, but if you look at a scientific CCD, the full well depth is something like 100000 electrons. The noise level is around 3 electrons RMS, so a pretty good ratio. The best sensors are close to what's possible, so I don't expect any cameras will be introduced with wildly better noise performance. Nor do I know anything about amplifier noise, other than it's very low.

[NAwlins Contrarian]

On 5/10/2023 at 7:39 AM, digitaldog said:

No argument.

The photographer reduces (under exposes). The results are more or less noise? That's kind of the OP.

The same amount of noise, while capturing less signal (due to less exposure), means lower signal-to-noise ratio, which means more visible appearance of noise.

My sense is that what we perceive is not so much noise as signal-to-noise ratio. The thing we sense is somewhat indirect. This is a far from perfect analogy, but it's somewhat like: we don't actually feel hot or cold. What we feel is heat transfer, which is proportional to relative hot or cold but also affected by other factors. That's why the air inside a 450 degree F oven is more tolerable to our bare skin than 150 degree F water in a pot. In contact with skin, water transfers heat much more efficiently than air, so there's more / faster transfer of heat to our skin from 150 degree F water than from 450 degree F air. Similarly, our eyes (and ears) don't perceive noise so much as they perceive signal-to-noise ratio.

Or at least, that's my impression / understanding.

[paddler4]

I don't think that's a good analogy.

The reason we often see noise in high-ISO images is that the amplification entailed in raising ISO makes the noise visually more apparent. Since there isn't much signal, there isn't much to obscure it as it's amplified.

Often, we don't notice some sorts of background noise because other, louder noises during the daytime obscure them. Now suppose that at night, you amplified the total noise in your environment to the same decibel level you normally have during the daytime. That's analogous to raising ISO. The background noise would be very apparent.

[httpwww.photo.netbarry]

It seems I encounter noise  in a photo when I try to lift an under exposed area, often seen as shadow up in exposure in post, thus amplifying (if that's the right word), the lack of signal to noise ratio in those dark areas.  If you leave darks and blacks as they appear in your preview, you generally won't't see noise, just black if yo leave them as such.. Dos that make general sense?? I tend to expose for the highlights in digital photos as I don't like blown out skis etc., and often like the shapes of shadows.

[digitaldog]

Just now, httpwww.photo.netbarry said:

 I tend to expose for the highlights in digital photos as I don't like blown out skis etc., and often like the shapes of shadows.

But for what kind of data and how? 

Articles on exposing for raw:
http://www.onezone.photos
http://schewephoto.com/ETTR/
https://luminous-landscape.com/the-optimum-digital-exposure/
http://digitaldog.net/files/ExposeForRaw.pdf
https://www.fastrawviewer.com/blog/mystic-exposure-triangle
https://www.fastrawviewer.com/blog/red_flowers_photography_to-see-the-real-picture
https://www.rawdigger.com/howtouse/exposure-for-raw-or-for-jpegs
https://www.rawdigger.com/howtouse/beware-histogram
https://www.rawdigger.com/howtouse/calibrate-exposure-meter-to-improve-dynamic-range

[httpwww.photo.netbarry]

Not sure about data, but how I do, and forgive me if I'm misunderstanding your question, is I shoot in spot meter mode through my EVF that shows a wyswyg in the view finder and just move the spot to where the highlights have enough info in them that I can work with it in post and still get some shadow detail; but if the choice because of the dynamic range in the scene puts one or the other out go gamut, ie all black or all white I will let the blacks be black, and of course there's judgments made about what is important in the scene.  I tend to shoot street type photography's in often varied and changing lighting and in So. Cal often harsh light very contrasty light. The few times I've done portraits in controlled lighting, it's a different game for me.

[digitaldog]

Point your spot meter at a white dog on snow: massive under-exposure.

Point it at a black cat on coal: maybe over-exposure for the JPEG and maybe a very optimal exposure for the raw.

[httpwww.photo.netbarry]

I generally don't spot right directly on the white or black, I point the spot to where it brings in the exposure I want in those areas. Yes shooting in raw really helps.Here's a couple examples. Dark skin and clothes with  white shirt and white shirt with dark background. On the first I would have been moving the spot around the gray of the floor. 

 

Edited May 27, 2023 by httpwww.photo.netbarry

[digitaldog]

Without the raws and viewing a raw Histogram, impossible to comment on their exposure.

[httpwww.photo.netbarry]

Got it. It would take days to find them. I'm just looking as to how they look That's basically how I approach it. Digital camera, analog mind set 😉

[paddler4]

14 hours ago, httpwww.photo.netbarry said:

It seems I encounter noise  in a photo when I try to lift an under exposed area, often seen as shadow up in exposure in post, thus amplifying (if that's the right word), the lack of signal to noise ratio in those dark areas.  If you leave darks and blacks as they appear in your preview, you generally won't't see noise, just black if yo leave them as such.. Dos that make general sense?? I tend to expose for the highlights in digital photos as I don't like blown out skis etc., and often like the shapes of shadows.

Yes, this makes sense. When a severely underexposed area is left dark, you can’t see the noise. When you brighten that area, there isn’t much signal to amplify, so you end up brightening the noise enough to be apparent 

[frans_waterlander]

On 5/11/2023 at 7:45 PM, digitaldog said:

With the generous help (privately) of the fellow (who should go unnamed) that started this by suggesting my text was iffy**, we have come to a better understanding (for me) to come up with this simple statement:

The relative amount of noise changes with exposure. Underexposure produces more noise. If you underexpose, you'll produce more noise.

The statement below was, from his scientific perspective, sloppy. And I now have to agree. But I'm still open to comments unless we can put this to bed.  

**My iffy/sloppy statement was: Underexposure causes noise. His correction was Underexposure DOES NOT PRODUCE NOISE.

How about something like: "Full exposure results in the best SNR (signal-to-noise ratio); as exposure is lowered, the SNR gets worse."?
SNR is what really counts and "full exposure" sounds, in my opinion, better and is less provokative than e.g. ETTR, although it comes down to the same thing. This statement also includes both the absolute best possible case and what happens when you lower exposure. "Underexposure" in this context means "less-than-full exposure".

Any constructive comments?

[digitaldog]

14 minutes ago, frans_waterlander said:

How about something like: "Full exposure results in the best SNR (signal-to-noise ratio); as exposure is lowered, the SNR gets worse."?
SNR is what really counts and "full exposure" sounds, in my opinion, better and is less provokative than e.g. ETTR, although it comes down to the same thing. This statement also includes both the absolute best possible case and what happens when you lower exposure. "Underexposure" in this context means "less-than-full exposure".

Any constructive comments?

ETTR is not a good term, but at the time, Michael wrote his story and explained it not too bad. He was using compensation of the lie of a Histogram for JPEG and suggesting one expose to the right of that for raw. Not that any well-trained photographer ever needed a Histogram to optimally expose film, even tricky transparencies. 

Full exposure is OK but a bit vague, but optimal exposure is better, IMHO. Optimal (correct, ideal) exposure is nothing new, and photography 101. Something some of us produced for decades long before Histograms and raw data existed. 
"Optimal exposure results in the best SNR (signal-to-noise ratio); as exposure is lowered, the SNR gets worse"?

EDIT: Optimal exposure is also subjective and something the photographer should control. Clip highlights? Sure, if you want to. Block shadows? Sure, if you want to. 

Edited May 28, 2023 by digitaldog

[frans_waterlander]

Optimal exposure is fine and I understand why and how this term is used. However, (isn't there always an however?) full exposure IMO better reflects what you need to do to maximize SNR.
Optimal exposure may be taken by some people to mean "Do not fully expose as the resulting brightness might by more than what you want your final image to look like." The term full wouldn't leave that mental wiggle room/ambiguity.

Edited May 28, 2023 by frans_waterlander

[digitaldog]

Brightness isn't exposure and is an attributes of the (raw) processing. As Michael explained and called “normalization” in his original article.

[frans_waterlander]

3 minutes ago, digitaldog said:

Brightness isn't exposure and is an attributes of the (raw) processing. As Michael explained and called “normalization” in his original article.

Totally agree, but I used that as an example of how people might perceive the idea of optimal exposure. The best possible SNR doesn't care about people's preferences. It's an absolute with the only limitation of what highlights are allowed to be blown.

Edited May 28, 2023 by frans_waterlander

[digitaldog]

3 minutes ago, frans_waterlander said:

Totally agree, but I used that as an example of how people might perceive the idea of optimal exposure. 

If their perception is wrong, we teach them. As has been done in photography schools or elsewhere for like 200 years. 🤔

And if they say: "Do not fully expose as the resulting brightness might by more than what you want your final image to look like." let them know again exposure isn't brightness (and why) and that Brightness is a perceptual phenomenon one controls through processing (of the raw).

[digitaldog]

If anyone isn't a subscriber and can't get to Michael's original article he wrote: 

Now, load the "properly exposed" frame into a RAW converter, do your usual corrections and then send it into Photoshop. Next, load the "exposed-to-the-right" frame and do the same. But, make sure that you use the RAW converter's gamma, brightness and contrast controls to normalize it first. Now load this frame into Photoshop.

17 minutes ago, digitaldog said:

Brightness isn't exposure and is an attributes of the (raw) processing. As Michael explained and called “normalization” in his original article.