ISO : in-camera vs raw software

[vetrofragile79]

Hello,i'm new here and this is my first post.

I take a picture 1/20 f16 iso 400 and it is 2 stops underexposed , now i have to choices:

- first way: increase the iso value (signal amplification) directly on camera choosing 1600 iso and shot

- second way: load raw file on camera-raw or lightroom and push up the exposure .

who is the winner in term of noise (luminance and chroma)? On-camera processor or software from pc?

Ok, the best way to make right exposure is increase "real light" ,extend the exposure time or aperture, so the sensor will be hit by more photons...but sometimes i can't do that.

ps: i know there is a plus of noise due to sensor heating in long exposure..

thank you very much

[xpatUSA]

Hello,i'm new here and this is my first post.

Is there difference between the increase in iso directly to your camera and rising through RAW processing program such as adobe camera raw? If there is why?

I would like a clear answer that takes into account the difference between signal amplifier inside the SLR vs software.

thank you very much

Hello and welcome to CiC! I'm sorry to tell that your question is very difficult to understand and so it is not likely that you will receive any "clear answers".

Personally, I do not understand what is meant by "the increase in iso directly to your camera" and neither did I understand "rising through RAW processing program". If you can explain these two phrases in a different way, someone here should be able to help.

I am not trying to be rude. My native language is English english but I do understand the difficulties of writing technical stuff in a foreign language; in a previous life, I did some technical writing in Mexican spanish.

[dje]

I'm not sure whether I understand the question or not either but this example may help clarify things.

Let’s say for a particular scene, you get “proper” exposure with settings of

Shutter speed 1/20 sec, Aperture f/3.5, ISO 100

However you decide the shutter speed of 1/20 sec is too slow. So you increase the ISO setting from 100 to 200 and this now means to get the same exposure you have a shutter speed of 1/40 sec. In this case, the increase in ISO setting has meant that the gain of the sensor amplifier has increased by a factor of 2. It also means that the read noise of the sensor has been amplified by a factor of 2 and as a result the noise is more visible. At this point, the signal and noise is still analogue but both then get converted to digital by the A/D converter.

If you took the shot with 1/40 sec, f/3.5 and ISO 100, it would be under-exposed by a “factor of 2”. You could then raise the exposure in ACR but this would also raise the noise level as you would still be applying an amplification factor of 2 digitally in software.
So overall, there is probably not much difference, but I think most people like to get the exposure right in camera, particularly if you want make use of some of the automatic features of digital cameras such as automatic exposure metering or modes such as aperture priority or shutter speed priority.

Dave

[vetrofragile79]

im sorry for my bad english,im tring to improve it following a summer class

ok, i try to explain what i mean : I take a picture 1/20 f16 iso 400 and it is 2 stops underexposed , now i have to choices:

- first way: increase the iso value (signal amplification) directly on camera choosing 1600 iso and shot

- second way: load raw file on camera-raw or lightroom and push up the exposure .

who is the winner in term of noise (luminance crominace)? On-camera processor or software from pc?

Ok, the best way to make right exposure is increase "real light" ,extend the exposure time or aperture, so the sensor will be hit by more photons...but sometimes i can't do that.

ps: i know there is a plus of noise due to sensor heating in long exposure..

[Polar01]

Vetrofragile79: a real name would help as the handle you picked to way too long to type. Another way as you state f/16 could be switched to f/8 which is two stops, if you are using a DSLR than that is just a simple enough change that can be done almost with out thinking so your "but sometimes I can't do that" is not a true statement. However it maybe with a point & shoot which I do not know about as I do not own one. One other thing about sensor heating in long exposures how long is long, I have taken 3 exposures back to back to back each of which was 20 minutes in length, I did get some however I was shooting in the middle of the night. So if you believe that say 5 minutes or more is too long, than think again not it is not. I was shooting with a Nikon D600 at the time.

Cheers:

Allan

[vetrofragile79]

hello alan thanks for the reply, what really interests me is the difference between reflex and software such as adobe camera raw to increase the exposure using iso.

Gabriele

[revi]

From my experience, you're better off raising the ISO in the camera, than trying to increase the lightness when processing the raw file.
The difference is that the in-camera ISO increase works through amplification before digitisation. Digitisation implies rounding to fixed values,
so the difference between adjacent pixels can be 0, 1, 2, ... after digitisation. Before this step, differences of 0.3, 0.6, 1.1, 2.4... are possible.

Now, if we increase sensitivity by 2 stops, that's an increase of 4x...
Doing that on the raw file, means our adjacent pixels will differ by 0, 4, 8, ... i.e. steps of 4 units
Doing it in-camera, we get differences of 1.2, 2.4, 4.4, 9.6 rounded to 1, 2, 4, 10 in our final image. In other words, we still have all possible values,
and especially the minimum difference is still 1 unit.

So, in-camera ISO increase should get you a less aggressive noise pattern, which will probably be easier to correct as well (as it fits the 'normal'
definition of noise better).

There's a bit more to it, as the camera works in 12 or 14 bits/channel, and post is either 8 or 16 bits/channel.

That also explains why under-exposure is so bad, esp. at high ISO...

Wrt long exposures: Part of the noise specific to long exposures can be removed by what's called 'black frame subtraction':
take an photo with the same exposure as you used for your image, but with the lens cap in place, then subtract that 'black'
photo from your image. Some of the long exposure noise is hot pixels, i.e. pixels that give more and more signal, even when
there's no light hitting them. (This signal is low enough that it's invisible for short exposures) Such a black frame can even
be taken at a later time, and used for other pictures taken with the same ISO and exposure time (F-stop is not important).

Remco

[Inkanyezi]

Your question makes very much sense to me. Raising ISO in the camera or doing it in the software have many similarities, but as pointed out in tje post above, there is an actual "ladder-step" quality of the digital values, and in less exposed areas, those values are coarser than in those that have much exposure.

Hence, if you alter the curve in RAW processing, there is a risk of posterisation in the dark regions, which you avoid by increasing ISO in the camera, which works before digitising. Thus your exposure values have a smoother gradation in the very dark areas when raising ISO in the camera. The noise will be the same in both cases, but the gradation will be different, and sometimes that difference may be visible.

And, as not all intuitively understand, altering "exposure" in RAW processing does not change exposure. There has been confusing terminology in this thread around the matter, so I repeat. You cannot, will not, never, change exposure in post processing, as exposure is the amount of light that hits the sensor. But as you so eloquently put forth in your original question, what you change is in fact ISO, as long as we understand ISO to be the amplification of the signal - as well as the noise.

Posterisation in the low levels, the digital steps instead of a smooth analogue gradient, is the difference between the two methods.

It is also possible to "raise ISO" later in post processing, in the image manipulation software, but then there are even fewer digital levels, which is particularly evident in the brighter levels. The risk of visible posterisation increases when fiddling with the curve in 8 bit image formats.

[dje]

I've been thinking about this a bit more since my original post and came to the conclusion there was more too it than I indicated. I see that in the meantime Remco and Urban have contributed with the point about digitisation which makes a lot of sense to me. I have just done a simple test in which I compared noise between two shots - both with the same shutter speed and aperture but one with ISO 400 and the other with ISO 100. The latter was under-exposed but that has been compenated for in ACR. As you can see, the second shot has significantly worse noise. I can't claim my tests to be absolutely objective but I think they illustrate the point that it is better to get the exposure right in camera. In addition to what the others have said, I also wonder if the de-mosaicing introduces more noise in the "under exposed in camera" shot.

These shots are 100% crops of a dark area.

[dje]

One other point - noise generated after the ISO amp at the input of the A/D converter will not change with ISO setting but it will be increased if digital gain is introduced in software.

Dave

[vetrofragile79]

thanks to everyone really, you responded to my initial question: "it is better to use the iso in-camera" because the analog amplificator allows more accurate values ​​with respect to digitization.

Another question, I hope not too much to ask, but reading the tutorial did not really understand what is the link between (fixed pattern noise, random noise, banding noise) and chroma noise-luminosity noise .. What connects the first three to the last two?

best regards

Gabriele

[jcuknz]

un-cancelled19 June
The point is that one doesn't need 'raw' software to do it just the regular better editing programme with either levels or curves tool.
My opinion is that it is better to do it with the gear designed to be working together than with a second lot of 'general purpose' 'gear'.
But it is worth being aware of what is possible so that one can control other parameters in the making of the exposure to get the best possible end result. Photography, the endless compromise.

Here I didn't have enough light to shoot in the Mining Museum out of Colorado Springs so took the shot at a exposure I knew OIS would give me a sharp result ... around 1/20 f/2.8 and raised it in editing. If one tries to do it with the normal brightness tool the highlights get blown badly. Lev els/curves raises the dark and mid-tone areas only.
A pressure locomotive which is pumped to a high pressure to provide motive power inside a mine.

I do not do this very often, very rarely actually, but I know I can do it for instances like this.

EDIT ... PS to #13 if your editor has the ability to split the image, Photoshop into LAB and PSP into CYMK you apply noise reduction to the L or the K layer respectively.

[revi]

(...)
Another question, I hope not too much to ask, but reading the tutorial did not really understand what is the link between (fixed pattern noise, random noise, banding noise) and chroma noise-luminosity noise .. What connects the first three to the last two?

Afaik, nothing:
The noise is present, and causes random (more or less...) variations in pixel brightness and colour. But there's a number of ways to
express the colour and brightness of a pixel, with RGB being one of them. Another system uses a luminance axis and two chrominance
axes (yellow/blue and red-green).

The interest of using this system to deal with noise is that the eye is more sensitive to luminance detail than to chrominance detail.
So we can be a lot more aggressive when reducing chrominance noise without too much loss of image quality (noise reduction
usually softens the image). And, luminance noise (wich is grayscale noise) is easier accepted than chrominance noise.
(note that this is mainly valid when dealing with an RGB image, I'm not sure there's an advantage in using luminance/chrominance
in a RAW image, where each pixel has only one colour, and usually the red and blue channels suffer most from noise.)

[vetrofragile79]

Thank u !