Is high ISO like autofixing an underexposure?

Sigma 24-70mm f/2.8 EX DG Macro for Nikon

01/25/2007 03:00:28 PM · #1

I've often wondered if ISO in a digital camera is nothing more than an exposure convenience for automatically adjusting for an underexposed image.

Is there any "real" difference between a capture at a high ISO setting at a particular shutter speed and one taken underexposed at a low ISO and adjusting it later in software?

cryan

Nikon D700

01/25/2007 03:02:46 PM · #2

From my experience the higher ISO will have more noise, but I guess is depends on how underexposed the shot is to begin with. As with slide film, there is no substitute for a correctly exposed photo.

cpanaioti

Canon EOS-40D

Canon EF 70-300mm f/4.5-5.6 DO IS USM

01/25/2007 03:03:25 PM · #3

That's a good question. I wonder if the high ISO shot would have less noise in the shadows since it would be exposed properly.

Something to experiment with maybe.

InDots

Pentax K10D

Tokina AF 28-70mm f/2.8 ATX Pro-SV for Pentax

01/25/2007 03:11:18 PM · #4

If you're under exposing and bet on software to bring the exposure back, you're never got the full detail shadows back (clipping on shadows). The same with over exposing (highlight clipping). In RAW there is some headroom, but also there you should watch if the full RGB scale is coming back (some clipping on only one color could happen, while the other ones are fine: leads to nasty artifacts). So ISO change is important because you're using the right sensitivity for the scene. Yes you would get more noise, but my experiance with RAW is that you better over expose a bit and tune it down than the other way around. The dslr sensor "creates" noise when a photocell is not litt well.

Fisher Price Kid-Tough Digital

01/25/2007 03:16:36 PM · #5

Originally posted by cryan:

From my experience the higher ISO will have more noise, but I guess is depends on how underexposed the shot is to begin with. As with slide film, there is no substitute for a correctly exposed photo.

Is using a high ISO setting on your digital camera actually 'correctly' exposed? Or, is it more like setting a film camera's ISO to 400 for a film actually rated 200 which results in a more 'noisy' negative because it is an 'incorrect' exposure?

LalliSig

01/25/2007 03:19:19 PM · #6

The correctly exposed shot taken at higher iso will blow an underexposed shot out of the water EVERYTIME unless there is something seriously wrong with your camera to begin with. Always increase iso over underexposing, trust me.

01/25/2007 03:22:50 PM · #7

Originally posted by Larus:

I trust you, but the experimentalist in me might just have to go out and prove it for myself. LOL!!

Ivo

01/25/2007 03:25:25 PM · #8

Great subject Steve as I'm sorta clueless on the benefits of messing with my ISO on my D200. It seems that the least amount of noise is at ISO 100 but whenever I go to a low-light setting and boost my ISO to 400 or 800, the noise becomes an issue and I have to NeatImage the heck out of my shots resulting in a somewhat plasticy end-product. I'm at the point where I think the ISO is simply a "flawed" feature on digital cameras only put there to give the impression that it responds similarly to film. Clearly, in my experience, this is not so.

Maybe I'm not using it right or I'm to adjust other settings in my camera in conjuction with ISO? Either way, it seems rather pointless and frustrating when I view my image and it looks like it was pepperred with hot pixels. Any advice or thoughts on this?

Simms

Canon EF 24-70mm f/2.8L USM

01/25/2007 03:28:52 PM · #9

Originally posted by stdavidson:

When I initially read this I though "Oh god, not another pointless question", but after thinking about it, its a real good point.. Would be interested to see the results of your experiments.. Dont listen to Larus, he is still a newbie to this photography lark.. ;)

Brad

Olympus OM-D E-M1

Olympus 40-150mm f/2.8 PRO M.Zuiko

01/25/2007 03:31:11 PM · #10

Originally posted by Larus:

I couldn't agree more Larus.
I usually shoot at 200 & 400, and have no reservations about bumping up to 800.

kirbic

Canon EOS R5

Canon RF 24-70mm f/2.8 L IS

01/25/2007 03:32:10 PM · #11

As Larus posted, there really is a (big) difference, at least a most ISO values. It all has to do with where gain is applied in the system. When you underexpose, you apply "gain" at the end, in post-processing. that's the worst place to do it. What you get is a very posterized image, and *lots* of noise.
On some Canon cameras, the highest ISO setting (3200) is actually a gain-in-software implementation, and this is why it's segregated from the main ISO choices. In this case, you *can* obtain very similar results by just shooting RAW at ISO 1600 and pushing +1 stop in conversion.

cryan

Nikon D700

Sigma 24-70mm f/2.8 EX DG Macro for Nikon

01/25/2007 03:32:47 PM · #12

Yes, setting your camera to a higher ISO does give you a correctly exposed shot, its not pushed or pulled in anyway. The ISO in cameras of today is just like shooting a roll of 100 and then having to change to 1600 for the next series. Obviously we have an advantage nowadays due to being able to change the ISO on the fly. With film when you are pushing or pulling you aren't over or underexposing so much as fooling your camera into thinking it has a different film. So thinking this way, if you had 400 film and shot it and set your camera to shot it at say, 800 you are pushing one stop. The grain you see in an enlarged silver halide. Each film has its own formula and silver size, giving it more exposure results in an enlarged halide. When you push or pull film you also should get it processed to compensate for the push or pull. I hope this makes some sense. Its a technique I've used many times while in the Navy.

cryan

Nikon D700

Sigma 24-70mm f/2.8 EX DG Macro for Nikon

01/25/2007 03:36:05 PM · #13

Ivo, you said you've had to do much post work when you shoot at anything higher than 100. I have shot at 1600 and been able to blow up to 8x10's and not seen as much noise as expected. Take a look at my volleyball shot is my portfolio, they were shot at 1600, the prints of those were absolutely amazing.

Gordon

RIM BlackBerry Curve

Canon EF 85mm f/1.8 USM

01/25/2007 03:36:49 PM · #14

Originally posted by stdavidson:

No it isn't. Higher ISO changes the effective gain of the photosites on the sensor for the linear capture, varying the SNR of the capture (a lower SNR gives more noise in the sensor, because the gain is higher - cf with turning the volume way up on a speaker) Raising the exposure after the fact in software changes the result after linear/gamma conversion.

There's a lot less information in the lower stops of the capture after you've converted the linear sensor data. So capturing it correctly in the first place has the potential to capture much more dynamic range in the shadow details.

That's why the low ISO version, adjusted upwards will very quickly fall apart in the shadow detail, in a way that doesn't affect a higher ISO shot at the effective higher exposure.

Gordon

RIM BlackBerry Curve

Canon EF 85mm f/1.8 USM

01/25/2007 04:07:24 PM · #15

Originally posted by kirbic:

The more interesting converse is what happens with the ISO50 option.
Mostly it seems to lose some dynamic range, for the potential advantage of longer shutter speeds.

kyebosh

Canon EOS-20D

Canon EF 80-200mm f/2.8L

01/25/2007 04:08:09 PM · #16

I learned from my 1D, that getting the exposure right in camera worked WAY better than pushing an exposure that was underexposed. And also that a properly exposed 3200 iso shot didn't look all that bad.

Canon EF 17-40mm f/4.0L USM

01/25/2007 04:08:41 PM · #17

What all you folks are telling me makes sense but I still will conduct my own tests (or/and research results on the Net) to see just how much difference there really is. I just need to find a composition with contrast and good detail in the shadows.

I've always been fascinated by the fact that CCDs and film both generate noise in greater magnitudes in improperly exposed images, but the physics is entirely different.

Gordon

RIM BlackBerry Curve

Canon EF 85mm f/1.8 USM

01/25/2007 04:15:37 PM · #18

Originally posted by stdavidson:

Steve,

Here's the thing I was inarticulately trying to explain.

Raw capture, linear gamma & exposure

Should clear things up.

Message edited by author 2007-01-25 16:15:59.

asimchoudhri

Canon EOS-5D Mark III

01/25/2007 04:18:07 PM · #19

Originally posted by stdavidson:

[...]I've always been fascinated by the fact that CCDs and film both generate noise in greater magnitudes in improperly exposed images, but the physics is entirely different.

I agree... I think about this almost on a daily basis (really...). In particular, the different meaning of the term "grain" in looking at noisy digital vs analog images is something that bothers me... technically with digital it is quantum mottle, and not actual film grains. But, people will make fun of you if you say "That night-shot looks great, but the details in the background are lost due to the quantum mottle."

fotomann_forever

01/25/2007 04:20:30 PM · #20

Let me try to explain another way:

This will work for anyone that has a digital cable box on their television.

Go turn your volume on your cable box way down and then turn the volume all he way up on the television set. Now turn the volume up on the cable box until you are at a comfortable listening level. Hear all that noise?

Essentially, that is what you are getting if you shoot low ISO under-exposed and then bump in post process. Your post processing (the television's amplifier) is going to amplify any noise coming from the sensor (cable box).

Now if you you turn the television down and the cable box up (not so much as to overdrive the TV's amp) the noise disappears. The same happens when you drive the gain on a sensor (ISO) higher and PP at lower levels.

Message edited by author 2007-01-25 16:22:14.

jaysonmc

Nikon D200

Nikon MF 105mm f/2.5 AIS

01/25/2007 04:22:53 PM · #21

My biggest issue isn't so much noise/grain that appears at higher ISO, but rather the loss of color fidelity. Especially at very high ISOs.

Simms