ISO - Waste of time when shooting RAW?

[sk66]

Surely an iso invariant camera will give the same RAW file at +0 and iso 102400 as it will at -10 and iso100? Or does "iso invariant" mean "iso invariant only up to 4 stops".

The raw file will be different, the results will be the same (after editing to the same exposure) as long as no highlights (DR) are lost in the brighter file. I.e. the sensor receives enough light that the signal above the noise is 1-127. That can be written as 1-127 (underexposed) or 2-254 (1 stop ISO) and it will make no difference. What you don't want to do is write it as 4-508 (2 stops) as there is nothing above 255, it doesn't exist (it would be 4-255).

Will the D810 show a clear display on live view of an image taken at -5 or even -7 or -10? Or will it show a black image? The Sony camera can show a correctly exposed image even though the recorded image will be almost black.

It can show the correct exposure based on the settings (i.e. brighter than actual). But the image playback (jpeg review) will have the exposure used. Even the Sony will playback the image with the settings used and this is the issue. If the image was recorded significantly underexposed you can't verify optimal settings were used on camera after the fact, the image must be processed externally.

Call me stupid, but I don't understand what these red and blue dots are,

Basically they are red/blue pixels that received enough light to register, but surrounding pixels did not... or they are pixels where the ADC saw *random* signal above the noise floor. Green pixels are used primarily for exposure which is why there are more of them. R/B pixels, along with green are used to calculate color using a 4 pixel grid (Bayer). In low light some pixels may receive no readable light, some may receive a little, and some may receive a lot, due to a high level of photon shot noise (light scatter/randomness). This is what causes dark areas to have a higher level of color noise.
The shot noise will exist in all images from all cameras. But the invariant sensor/camera adds little/no errors to the output due to having a very low noise floor (read noise) and very low write noise (ADC). I.e. it doesn't take much recorded signal to overpower the system noise.

 

[sk66]

I wonder if the photo taken in this thread would be a good example of where an iso invariant camera could come in handy?

https://www.talkphotography.co.uk/threads/local-lake.673585/#post-8100826

Yes.
Here is an example from some pics I took yesterday using the D810.
The image was taken 2 stops underexposed with the ISO nearly at minimum. It was taken in manual mode using highlight weighted matrix metering with auto ISO and -2EC. Alternatively, I could have fixed the ISO at anything between 64 and 400 and let the SS/Ap float instead (allowing 2 stops recovery up to ISO1600 equivalent).

Edit: Technically I could have set a full manual exposure at ISO 64 and let the image go darker/lighter with changes in lighting... the relevant factor is the exposure actually received (SS/Ap). That's just not the way I work, and doing that would be problematic with my other cameras.



Here it is with the Lightroom edits. There is a lot of "recovery" to the edits, with no significant image degradation.



This is the final image on flickr where it can be viewed full size.


Hooded Merganser by Steven Kersting, on Flickr

If I had used the D5 I would use the same method to save the highlights, but I could not recover the shadows/blacks as well. Or I would allow some highlight clipping in order to get a more usable image overall (usually). Preferably it would have been taken in softer/flatter light with either camera where such recovery isn't necessary.

 

[adrianday]

The raw file will be different, the results will be the same (after editing to the same exposure) as long as no highlights (DR) are lost in the brighter file. I.e. the sensor receives enough light that the signal above the noise is 1-127. That can be written as 1-127 (underexposed) or 2-254 (1 stop ISO) and it will make no difference. What you don't want to do is write it as 4-508 (2 stops) as there is nothing above 255, it doesn't exist (it would be 4-255).

I'm not entirely sure I understand what you're eluding to. A 14-bit RAW file provides 16,384 (2^14) levels per channel (RGB). The human eye is capable of resolving around 16.8 million colours. That equates to (2^8)^3. That's 8-bit - 256 levels per channel. The big issue with all of this talk of underexposing by 5EV is that you're effectively only capturing 9-bits worth of data - 512 (2^(14-5)) levels per channel. Whilst this is still beyond the perception of the human eye, there is a risk that extraction of detail (i.e. post production amplification) will degradation of the quality of the image. Loss of shadow detail and banding (due to their being 1/32 of the bandwidth (2^5) per channel prior to amplification) isn't worth the risk IMHO.

The key thing is to expose the image correctly at capture. The D810 features highlight-weighted metering (I'm sure other models will too). It's better than blindly underexposing by 5EV and thinking you can recover in post.

 

[gramps]

So the D850 is ISO-invariant but what is the optimum ISO range?

 

[sk66]

I'm not entirely sure I understand what you're eluding to.

A 14 bit raw file does not necessarily have 16,384 levels. At ISO 1600 the D810 is recording less than 8bit/color (20bit) and tonality (7.8bit). This is not due to the ISO, it is due to not collecting enough light. It doesn't matter how many bits you encode it with, it will not increase. Just like writing the file as a 16bit tiff/dng doesn't increase the data. Even at base ISO it is only just above 8bit color (25.7) and is at just under 9bit tonality. Additionally, a scene may have less than 14bit. And just because it is a 14bit encoder does not mean it has that fine of accuracy.
Edit: a 14bit pipeline is NOT writing the last EV with the same accuracy/steps as the rest combined (1/2 the data), it is writing everything at 14bit... The difference is that the last stop has twice as much signal/data generated.

I do agree that darker scenes should be exposed brighter (ETTR) in order to capture more data. But not at the risk of blowing highlights (preferably) and not by using ISO (*maybe* camera dependent).

The key thing is to expose the image correctly at capture. The D810 features highlight-weighted metering (I'm sure other models will too). It's better than blindly underexposing by 5EV and thinking you can recover in post.

The key with an ISO invariant sensor/system is the actual exposure (SS/AP), not the ISO.

 

[sk66]

So the D850 is ISO-invariant but what is the optimum ISO range?

For me w/ the D800/810 it has been ISO 64-1600 *depending on the actual light captured.* I would not expect the higher resolution of the D850 to increase that, if anything I would expect a decrease if you want to use the file at the larger potential sizes. Smaller pixels have a smaller FWC and less DR (a normalized image will have the same/better characteristics due to oversampling/combining).

 

[Bollygum]

P


Perfect understanding



OK Stupid

I'll try again. Lets look at our example of the image (lets call it a photograph) on the table covered by our Opaque piece of special glass. This is a composite image consisting of the photograph on the bottom (which includes our (1) Photon Noise & (2) Front end Read noise). (3) The Back End Read noise, which includes the ADC and Digital amplifier is represented by the Opaque piece of glass lying on top of our photograph and covers it completely.

We know that the ADC ads errors (imperfections/noise) to our final image. The 'red & blue' dots represent these imperfections or what is refereed to as 'Back End Read Noise'. As we move a potentiometer on this opaque piece of glass it turns clear but still contains the imperfections (red & blue dots) . The potentiometer is just like the exposure slider (ISO slider) in lightroom - it is the digital amplifier part of the Back End Read Noise . As the opaque glass goes clear we can start to see the photograph under it but the imperfections in the glass also become more visible as they are 'amplified.'.(Increased)

This Back End read Noise is applied to the whole of our composite Image as it lies on top of it and we have to 'look through' it. With a Non ISO Invariant sensor this degrades the resultant image quality.

If we can make the opaque glass (Back End Read Noise) perfect with no imperfections (red & blue dots) we have a true ISO invariant sensor that doesn't care what digital amplification is applied to it as it has no imperfections anyway.

Does this help?

Not a lot. It seems to be a very convoluted way of explaining something. Let me ask a question. Is the iso amplification analogue or digital?

 

[sk66]

Is the iso amplification analogue or digital?

It can be either or both... depends on the camera and what ISO you are talking about.

 

[Bollygum]

It can be either or both... depends on the camera and what ISO you are talking about.

Fortunately I succumed and read an article on iso invariance. https://www.google.com.au/amp/s/photographylife.com/iso-invariance-explained/amp It explains that for most ranges of iso the (iso invariant) camera will use digital amplification. So no loss and potential gain in DR. But there are still some points at which the camera uses analogue amplification. ISO 1600 on a Nikon D800 is an example. The 800 will not be iso invariant across iso 1600, though any increase in noise will probably be quite small. The Sony A7R2 has a point of analogue amplification at iso 640 I believe.
I am not sure how much difference this will.make to my photography as I have never been keen on using high iso. I tend to use iso 100 and iso 800 which, purely by chance, seem to be a very good choice. I may have to reassess with the A7R3

 

[Fraser Euan White]

Not a lot. It seems to be a very convoluted way of explaining something. Let me ask a question. Is the iso amplification analogue or digital?

Fortunately I succumed and read an article on iso invariance. https://www.google.com.au/amp/s/photographylife.com/iso-invariance-explained/amp It explains that for most ranges of iso the (iso invariant) camera will use digital amplification. So no loss and potential gain in DR. But there are still some points at which the camera uses analogue amplification. ISO 1600 on a Nikon D800 is an example. The 800 will not be iso invariant across iso 1600, though any increase in noise will probably be quite small. The Sony A7R2 has a point of analogue amplification at iso 640 I believe.
I am not sure how much difference this will.make to my photography as I have never been keen on using high iso. I tend to use iso 100 and iso 800 which, purely by chance, seem to be a very good choice. I may have to reassess with the A7R3

I did post a link many, many pages ago to this - post No.78; some 171 posts ago!! (But you obviously chose not to read it as you felt you couldn't learn any more about the subject)....................I won't post anymore as I was trying to explain the full theory in simple terms (as you really weren't grasping what was being said) ; if you had read this several days ago when asked it would have saved a lot of time and effort! (It may well be worth going back a couple of pages and reading the quote from Canon that I posted for you that you clearly didn't read )

Cheers,

Fraser

P.S. - since you feel you understand the theory now Bollygum, after reading the concise three-four page article instead of my 'convoluted' couple of paragraphs, you could maybe explain where the theory of shooting at base ISO comes from (I'm not stating the theory is correct but it is what this original thread started off as and not just ISO invariance)?

Interestingly when I was quoting parts of the same article you were replying with:

This is all very silly

This is utter rubbish

???

 

[Bollygum]

I did post a link many, many pages ago to this - post No.78; some 171 posts ago!! (But you obviously chose not to read it as you felt you couldn't learn any more about the subject)....................I won't post anymore as I was trying to explain the full theory in simple terms (as you really weren't grasping what was being said) ; if you had read this several days ago when asked it would have saved a lot of time and effort! (you still haven't got the theory correct so I can take pride in the fact you don't understand a professionally written article either )

Cheers,

Fraser

P.S. - since you feel you understand the theory now Steve , after reading the concise three-four page article instead of my 'convoluted' couple of paragraphs, you could maybe explain where the theory of shooting at base ISO comes from (I'm not stating the theory is correct)?

Interestingly when I was quoting parts of the same article you were replying with:





???

You know what my biggest problem with this thread is? Too many words and not enough information. No, I didn't read all 249 posts. Does anybody?
To me, the simple fact that most ISO expansion is purely a digital multiplication explains everything. All the other stuff is padding - even in the article that I quoted, most of the complexity seemed irrelevant.
You claim that I still don't understand the theory. Can you explain what relevant bit I don't understand? And what an odd thing to take pride in.


As for my comments about things being silly and utter rubbish. Mostly they were wrong, but in what I kept trying to say (ie iso 100 to iso 102,400 or 10 stops) they were right, though not as right as I originally thought. You ignored that because you apparently wanted to talk about pure theory. As a photographer I live in the real world of real cameras and it concerns me how the real cameras work. In my case my A7R3 is iso invariant from iso 100 to iso 400 and from iso 640 to iso 102,400. With a Nikon D8xx the figures would be iso64 to iso800 and iso 1600 to iso102,400. That didn't take 249 posts to explain.

But as a final word. Thank you for trying to explain. I did have the wrong idea to start with and you and Steven have helped to get me to the right point. I'm not sure if it will change my photography as I am high iso averse and I will always try to collect more light rather than use high iso (or underexpose). Oddly enough I think this may have more relevance for video, which is clearly not iso invariant, as it is essentially similar to jpeg. The reason for this is that we are very limited in what we can use as shutter speed in video and we have to predict the future in that we start shooting at one setting and the lighting could change unexpectedly. The use of S Log and subsequent improvements on that will change video significantly.

 

[Bollygum]

P.S. - since you feel you understand the theory now Bollygum, after reading the concise three-four page article instead of my 'convoluted' couple of paragraphs, you could maybe explain where the theory of shooting at base ISO comes from (I'm not stating the theory is correct but it is what this original thread started off as and not just ISO invariance)?

???

If a camera is truly iso invariant all photos are taken at the base iso and the signal digitally multiplied to fit the selcted iso. ie it makes no difference what iso you select and in the worst case you could clip some data in bright areas making it worse to select a high iso..

 

[Fraser Euan White]

If a camera is truly iso invariant all photos are taken at the base iso and the signal digitally multiplied to fit the selcted iso. ie it makes no difference what iso you select and in the worst case you could clip some data in bright areas making it worse to select a high iso..

Hi Steve,

when we refer to an ISO Invariant sensor we are only referring to the last part of the digital capture; It is the 'opaque' piece of glass or the technical term 'Back End Read Noise' - a true ISO Invariant sensor adds no back end read noise. The Iso Invariance has nothing to do with the other two types of noise; Photon Noise and Front End Read Noise present in the complete Image (Photograph).

Understanding the whole may help to explain the 'base ISO' theory; would you or anyone else like me to go on or am I boring the ti*ts off everyone?

Fraser

 

[Bollygum]

Hi Steve,

when we refer to an ISO Invariant sensor we are only referring to the last part of the digital capture; It is the 'opaque' piece of glass or the technical term 'Back End Read Noise' - a true ISO Invariant sensor adds no noise at this stage. The Iso Invariance has nothing to do with the other two types of Noise; Photon Noise and Front End read noise present in the complete Image (Photograph).

Understanding the whole may help to explain the 'base ISO' theory; would you or anyone else like me to go on or am I boring the ti*ts off everyone?

Fraser

I'm happy to read what you have to say, but first, do you agree that a Nikon D8xx camera is (more or less) iso invariant between iso64 and iso 800, and also between iso1600 and the maximum iso? The Sony A7R2&3 have a similar break point, but at iso640 not iso1600. I think this is important for photographers, rather than those interested in camera theory.
Having worked for most of my life in large systems design and performance management (IT), I tend to think in digital and I understand complex systems quite well. Opaque pieces of glass are a bit hard to imagine in any system that I have had to deal with, so it is not an analogy that has much meaning for me. Consequently I am likely to get the intended meaning wrong.
On the other hand I do understand that once a signal is digital then no multiplication can add data. It merely shifts the data by adding zeros to the left.

 

[sk66]

To me, the simple fact that most ISO expansion is purely a digital multiplication explains everything.

This is actually incorrect/backwards... most *usable* ISOs are analog gain. The extended ranges (hi/lo) are typically digital. Some are digital when they should be analog, and some are dual stage (analog + digital). Basically, the computer in your camera, in terms of exposure, is not as smart as a dedicated program/computer... but it *can be* smarter when it knows more about the sensor/system characteristics than a 3rd party program does (proprietary stuff) .
But analog vs digital isn't a significant factor in terms of how things work...

 

[Fraser Euan White]

I'm happy to read what you have to say, but first, do you agree that a Nikon D8xx camera is (more or less) iso invariant between iso64 and iso 800, and also between iso1600 and the maximum iso? The Sony A7R2&3 have a similar break point, but at iso640 not iso1600. I think this is important for photographers, rather than those interested in camera theory.
Having worked for most of my life in large systems design and performance management (IT), I tend to think in digital and I understand complex systems quite well. Opaque pieces of glass are a bit hard to imagine in any system that I have had to deal with, so it is not an analogy that has much meaning for me. Consequently I am likely to get the intended meaning wrong.
On the other hand I do understand that once a signal is digital then no multiplication can add data. It merely shifts the data by adding zeros to the left.

That's it Steve - you have finally beaten me into surrender


A2 by Fraser Euan White on Talk Photography


I feel I am on a hiding to nothing with this discussion! I do not own a Nikon D8xx nor do I ever intend to own one, I don't own a Sony either, don't wish to own one and have never, or am I likely to research either camera!

You want me to explain things simply for 'photographers' yet you are quoting ISO numbers for certain cameras and want me to make comment on them? Then you explain how digital amplification adds data by adding zero's to shift the data?

However, in the next sentence you state you can't understand that laying an opaque piece of glass on a photograph means you can no longer see the photograph then, if you were to move a slider to make the opaque glass clear you could then see it.I stated that this is just like the exposure slider in lightroom, yet you state you can't relate to that?

I am obviously getting it completely wrong trying to explain the base ISO theory; if your 'IT trained brain' can not relate to analogies that I think a GCSE student would relate to then there really isn't any point wasting anymore time.

It may come as a surprise but talking about 'The base ISO theory' may just involve.........erm............theory.

 

[Bollygum]

Ah well, I've got it wrong, but the results as shown on DPReview etc are the same as if I got it right - so my wrong model works rightly (to borrow a Trumpism). And for a practical photographer, that is what really counts. Understanding how the whole system works is the key to my photography, so that is my focus and I sometimes get impatient with a focus on what are, to me, irrelevant details. Still, I have to thank you for making me aware what iso invariance means in the real world. Even if I got it all wrong.

 

[adrianday]

I'll put the kettle on for a nice cup of tea.

 

[Fraser Euan White]

Ah well, I've got it wrong, but the results as shown on DPReview etc are the same as if I got it right - so my wrong model works rightly (to borrow a Trumpism). And for a practical photographer, that is what really counts. Understanding how the whole system works is the key to my photography, so that is my focus and I sometimes get impatient with a focus on what are, to me, irrelevant details. Still, I have to thank you for making me aware what iso invariance means in the real world. Even if I got it all wrong.

Steve,

all I will say in conclusion is that I find your photographs excellent

Fraser

 

[Bollygum]

Steve,

all I will say in conclusion is that I find your photographs excellent

Fraser

Thanks Fraser. I wish I could get enthused about exactly how iso invariance comes about, but I just can't. It is enough for me to understand what happens and what it means for me. I read about it and I just glaze over in the bits that explain how it happens. I just focus on what happens and making a model of it. So I apologise for the frustration you must feel. Life's too short for lengthy arguments over ....... what or how.

 

[gramps]

Thanks Fraser. I wish I could get enthused about exactly how iso invariance comes about, but I just can't.

You are not alone and I believe it is being over-hyped as it's something new/different.
Exposure compensation should be able to do all that is needed in most cases and if you do that with an ISO invariant camera maybe so much the better but with the bonus that you actually still get to see a good representation of what the image looks like on the LCD screen.

 

[HoppyUK]

You know what my biggest problem with this thread is? Too many words and not enough information. No, I didn't read all 249 posts. Does anybody? <snip>

Does anybody read every post? Yes, if I want to learn something. And definitely yes if I want more information or enter into debate.

249 post and half of them are from you, yet now we learn you didn't read everything, didn't follow explanatory links, and your judgement was prejudiced by a firm conviction that you knew better and everybody else - on here, and multiple other respected sources - was being "very silly" and talking "utter rubbish." I wish I hadn't read every post now.

Credit to you for owning up though, that's a rare quality around these parts

 

[Bollygum]

You are not alone and I believe it is being over-hyped as it's something new/different.
Exposure compensation should be able to do all that is needed in most cases and if you do that with an ISO invariant camera maybe so much the better but with the bonus that you actually still get to see a good representation of what the image looks like on the LCD screen.

I use both types of cameras and I already find that I avoid using the older cameras for really dark work (sounds like I'm up to no good ). I also tend to underexposed the new cameras by more than I used to, but I can't see always using iso100 catching on.

 

[Bollygum]

Does anybody read every post? Yes, if I want to learn something. And definitely yes if I want more information or enter into debate.

249 post and half of them are from you, yet now we learn you didn't read everything, didn't follow explanatory links, and your judgement was prejudiced by a firm conviction that you knew better and everybody else - on here, and multiple other respected sources - was being "very silly" and talking "utter rubbish." I wish I hadn't read every post now.

Credit to you for owning up though, that's a rare quality around these parts

I'm impressed. Are you a lawyer?

 

[woof woof]

You are not alone and I believe it is being over-hyped as it's something new/different.
Exposure compensation should be able to do all that is needed in most cases and if you do that with an ISO invariant camera maybe so much the better but with the bonus that you actually still get to see a good representation of what the image looks like on the LCD screen.

ETTR can't do it all, not all the time anyway.

I think people are maybe missing a point. With ISO invariance you can do something as simple as raise the shadows with less penalty that you'd very possibly pay with non invariant kit.

Anyway, it's all been said multiple times now. I would however urge people to read up on this if only for interest.

 

[gramps]

ETTR can't do it all, not all the time anyway.

I think people are maybe missing a point. With ISO invariance you can do something as simple as raise the shadows with less penalty that you'd very possibly pay with non invariant kit.

Anyway, it's all been said multiple times now. I would however urge people to read up on this if only for interest.

Exposure compensation works in two ways + & -
The discussion started out with the scenario of a black image on the rear LCD screen and the image being 'rectified' in processing ... personally I doubt many togs will want to use that extreme end of working 'blind' on capture.

 

[woof woof]

Exposure compensation works in two ways + & -
The discussion started out with the scenario of a black image on the rear LCD screen and the image being 'rectified' in processing ... personally I doubt many togs will want to use that extreme end of working 'blind' on capture.

I think your mind is closed to this and if that's the case then so be it but personally I think it's a subject at least worth reading up on.

 

[gramps]

I think your mind is closed to this and if that's the case then so be it but personally I think it's a subject at least worth reading up on.

Nothing of the sort, a mind closed is not open to reviewing it and considering it ... I have done both, even tried it in the field and come to my own considered conclusion.