ISO 1600 vs correcting exposure in photoshop

[dmckinny]

My initial interpretation was that I should set my camera to ISO 1600 and lock that in, which I don't think is the desired interpretation.

Correct. ISO 1600 always has less noise for a fixed exposure, but that doesn't mean we should always use it. High ISO has a price: clipped highlights. That price is always far too high to pay, except when we are in low light, and we no longer need the normal amount of highlight headroom.

OK now, the lights are starting to turn on for me. Daniel, I think I was not comprehending what you meant by 'highlight headroom'. Would the following statement be correct?


Highlight headroom is roughly analogous to available dynamic range. In the case of a low light shot, we slow the shutter as much as is acceptable, and open the aperture as much as is acceptable. Then when there is still not enough light for our needs, we increase the ISO. The trade off here is that by increasing the ISO we are sacrificing dynamic range (losing highlight headroom), in order to decrease read noise and (in the case of digital) amplify the signal on the sensor. This works in low light because there is less dynamic range in the scene, so we can use the higher ISO.

[Daniel Browning]

Would the following statement be correct?


Highlight headroom is roughly analogous to available dynamic range. In the case of a low light shot, we slow the shutter as much as is acceptable, and open the aperture as much as is acceptable. Then when there is still not enough light for our needs, we increase the ISO. The trade off here is that by increasing the ISO we are sacrificing dynamic range (losing highlight headroom), in order to decrease read noise and (in the case of digital) amplify the signal on the sensor.

All correct.

This works in low light because there is less dynamic range in the scene, so we can use the higher ISO.

I would put it differently. Low light scenes tend to have very high dynamic range, and we'd love to shoot them at ISO 100 if we could. It's just that reducing the noise is *so* important that we will sacrifice the headroom to get it. That is, we'll go from 7 stops of highlight headroom down to 3 stops, but the decrease in noise will be well worth it.

[canoli]

wow, interesting discussion.


What I got from this thread is basically this: make sure you ETTR and all will be well, regardless of how you get there - wider aperture, longer exposure, higher ISO. When creative intent dictates you use a small aperture, increase your exposure. If you can't handhold because the required exposure is too long, work off a tripod. No tripod? Increase ISO. The goal is ETTR no matter what. Yes?


The benefits of ETTR, the way I understand it, is you'll be getting the most noise-free images and you won't give up anything that can't be "fixed" in post. Example, a dark scene, ETTR, ends up looking way too bright on your LCD. But if you want to reproduce the scene accurately in your print, you'll have all the latitude you need to recreate that dark scene. If on the other hand you tried to shoot it "correctly" you'd have very little HL detail and a world of noise in the shadows.


Is that all true?


I hope so, because I've been shooting that way for a year now, ever since I learned about the linear capture of digital sensors...

[Daniel Browning]

Thanks for the response, Canoli. I'll try to clarify and be as helpful as I can.

What I got from this thread is basically this:

Your post is close enough. You are using the optimal technique, and that's really all that matters in the real world.


The only nits I would pick are in the vocabulary and the reason "why" you use that technique. You can stop reading now if you're not really interested in the "trivia" of knowing why your technique is better, or what the correct vocabulary is.

make sure you ETTR and all will be well, regardless of how you get there - wider aperture, longer exposure, higher ISO

First of all, ISO is not a part of "Exposure", and therefore it is not a part of ETTR. Exposure does not include "ISO" because it is a word that only relates to the total amount of light the falls on the sensor per area. I realize that many photographers think that "Exposure" means "brightness" and that increasing ISO increases "exposure", but that's not correct. The correct definition is given, for example, in the book "Photography", Eighth Edition, by London & Upton. It is:
<p style="padding-left: 30px;"] Exposure = Intensity (aperture) x Time (shutter speed)


As for "regardless of how you get there" -- I wouldn't put it that way. Because it's very important that exposure be increased before ISO. If you increase ISO when you could have increased exposure, you will get far more noise. In other words, "ETTR then ITTR". Don't "ITTR then ETTR". I'm sure that this is precisely what you meant, but I thought I would point it out for the reader just to make sure there's no confusion.

If on the other hand you tried to shoot it "correctly" you'd have very little HL detail and a world of noise in the shadows.

Actually, shooting it "correctly" (not ETTR) gives you *more* highlight detail. ETTR (and ITTR) gives you less noise in the shadows at the cost of highlights (nothing for free). So you would only do it if you had nothing important in the highlights or the shadows were just so important that it doesn't matter what's in the highlights.

I hope so, because I've been shooting that way for a year now, ever since I learned about the linear capture of digital sensors...

The value of ETTR has nothing to do with the linear capture of sensors. That was a myth made popular by a Luminous Landscape article, where they said that it results in more levels being used (2048 levels in the first stop, then 1024, etc.). That's totally wrong and has no benefit whatsoever because of photon shot noise.


The real reason why ETTR has so much value is because it results in *more light*. More light means less photon shot noise and more distance between any given tone and the read noise of the sensor. I hope that helps.


Kind regards,

[Jon Ruyle]

What I got from this thread is basically this: make sure you ETTR and all will be well, regardless of how you get there - wider aperture, longer exposure, higher ISO.

No. It is true that with a *given exposure* you want the highest iso possible without clipping highlights (high iso means lower read noise). But you always get less noise with a longer exposure and lower iso than with fast exposure and high iso, because the more light you let in the less photon noise you will have. So Daniel is *not* saying we should shoot iso 1600 1/1600 sec instead of iso 100 1/100 sec (at least I don't *think* he is- correct me if I'm wrong, Daniel). He is saying that with a *given* shutter speed and a *given* aperture, we want the highest iso possible without clipping highlights.


Using a low iso and low shutter speed is probably more important in most cases, because photon noise usually dominates read noise. This is why I *wish* auto iso worked in the reasonable way (ie, user chooses shutter speed and aperture and camera exposes correctly). If I take a picture in which the shutter speed is four times what it has to be, I've lost 75% of the light I could have had. This is a big deal when you consider how much we pay for fast lenses and sensitive ccd's.

[Jon Ruyle]

Oops-- once again Daniel beat me to it []

[canoli]

Thanks Daniel - I think (!).

If on the other hand you tried to shoot it "correctly" you'd have very little HL detail and a world of noise in the shadows.

Actually, shooting it "correctly" (not ETTR) gives you *more* highlight detail.

Here you're referring to the amount of light hitting the sensor but not the end result, right? I should've said "you'd have very little HL detail..." when you bring it into post and attempt to recover the HLs. You end up with very little HL detail because in post you'd have to boost the shadows so much (to see the HLs) that
the result would be hopelessly noisy. That's what I meant by shooting it "correctly" and getting it wrong. (crossing fingers) Have I got it?

ETTR (and ITTR) gives you less noise in
the shadows at the cost of highlights (nothing for free).

But only if you blow out the HLs, no? The "cost of highlights" is only payable to HLs you care about, naturally. Higher ISO by default doesn't cost you HL detail does it?


I'm guessing "read noise" and "photon shot noise" - if it's not the same as Luminance and Color noise - is a mathematical discussion that's over my head. Maybe the whole discussion is. But thank you for indulging me and for the boost of confidence, telling me that I'm "using the optimal technique," even if I can't talk about it or understand it very well.


So I can safely ETTR all the time (aperture/shutter speed, then ISO) and I'll be using my camera to its fullest potential, capturing the widest possible dynamic range. As long as that's true I'm good to go!


Thanks again Daniel.

[Daniel Browning]

You end up with very little HL detail because in post you'd have to boost the shadows so much (to see the HLs) that
the result would be hopelessly noisy. That's what I meant by shooting it "correctly" and getting it wrong. (crossing fingers) Have I got it?

Close enough for me. [] Personally, I find that the highlights are never noisy enough to be a problem (even in a severely underexposed shot). It's noise in the midtones and shadows that bothers me.

But only if you blow out the HLs, no? The "cost of highlights" is only payable to HLs you care about, naturally.

I agree.

Higher ISO by default doesn't cost you HL detail does it?

It does if you think of it this way:


f/2.8 ISO 400: 3 stops of highlight headroom


f/4 ISO 400: 4 stops of highlight headroom, but now 1 stop of "footroom" has been lost to noisy shadows.


f/4 ISO 800: 3 stops of highlight headroom, back to the same 3 stops of headroom as the first shot.


f/4 ISO 1600: 2 stops of highlight headroom, but now those shadows are far less noisy and are more usable.

I'm guessing "read noise" and "photon shot noise" - if it's not the same as Luminance and Color noise - is a mathematical discussion that's over my head.

It's not the same as luminance and color noise, but it's not over your head, either. Think of it as "read noise = shadow noise" and "photon shot noise = midtone and highlight noise".

So I can safely ETTR all the time (aperture/shutter speed, then ISO) and I'll be using my camera to its fullest potential, capturing the widest possible dynamic range. As long as that's true I'm good to go!

You're doing great!

[Johnny Rasmussen]

Hi, Andy (Johnny)[] !


I&acute;m glad you liked the article. You are right, Google translated it very accurate.


My nickname at highschool was Johnny Cash btw, deep voice and a guitar. :-)





Johnny

[canoli]

Ah...thank you again Daniel! You're a heckuva teacher, and as long as I'm not imposing, I'd love to continue the discussion. (If you'd rather not get into it, I understand. I know sometimes teaching can be tedious...)


The part that baffles me is this:

Higher ISO by default doesn't cost you HL detail does it?

It does if you think of it this way:


f/2.8 ISO 400: 3 stops of highlight headroom


f/4 ISO 400: 4 stops of highlight headroom, but now 1 stop of "footroom" has been lost to noisy shadows.


f/4 ISO 800: 3 stops of highlight headroom, back to the same 3 stops of headroom as the first shot.


f/4 ISO 1600: 2 stops of highlight headroom, but now those shadows are far less noisy and are more usable.

Where do these statements come from? Are they arbitrary figures, just so you can illustrate the differences? Or is "f/2.8 ISO 400: 3 stops..." (and the others) simply true, all by itself?


Aren't headroom values dependent upon the dynamic range, which obviously vary from scene to scene? Or are these statements "constants" in some way?


Sorry if I'm complicating this. I always think of Headroom as a safety margin of sorts, a tonal range that is available in the development process, existing only because the photographer didn't clip the HLs. How much headroom is a function of the exposure (and the scene's avail range of course) and then the process in PP (the creative intent).


But even if I'm correct in thinking of headroom this way, I don't see how "f/4 ISO 400: 4 stops..." (or any of them) is automatically true.


I mean, I see the trend - higher ISOs squeezing the HL headroom - but how or why does it happen?


Again, sorry if I'm complicating this. The answer may very well be staring me in the face but I don't see it yet.


If you feel like continuing, I would really appreciate reading more of your thoughts on this subject.


Thank you very much!