Balancing Exposure and Processing

This thread is about the difference between exposure and ISO, and the use of in camera processing and post processing to achieve a usable image.

I kind of like the old days when Kodachrome was King. Life was simpler. We had fewer choices.

Load your film and then you only had exposure to worry with, and exposure was a combination of shutter and f/stop.

Exposure was how much and how long the film was exposed. Once exposed, every other adjustment was made in the lab when processing the film or printing prints. Good lab work could do wonders.

Digitally speaking exposure is still exposure. It is the combination of two things that control how much light and how long the light reaches the sensor. This has not changed.

What does ISO do?


In the five examples above, the exposure does not change. So, in each of the five shots the sensor has received the same amount of light for the same amount of time. Ponder this for a moment.

The ISO setting then must tell the camera how to process the sensor information and produce an image file. There's a bit of an unknown world in between the sensor and the image file that is in the hands of the camera's processor.

Those who like experimenting in the unknown world are welcome to post to this thread and those whose are interested are welcome to comment.

I'm exploring that unknown world between the analog signal of the sensor and the digital conversion that is our RAW.

Remember, this is only when I can't increase my exposure by either lowering shutter or opening up the lens.

ISO decides how much the analog signal is amplified. There isn't much technical information out there so it's guesswork and then experimenting to check if the results match the guess.

What I think goes on is that at the native ISO the analog signal is converted at something like a 1:1 ratio. Each stop of ISO changes that conversion and multiplies it. ISO 200 would be a 1:2, 400 a 1:4.

The problem with this is that it too multiplies all noise which is clearly visible when we shoot high ISO. This is purely the result of this change. Another negative side effect is it lowering our quality.

What I'm experimenting with is skipping the amplification by only using the native ISO and manually adjust my exposure afterwards in post. The reason for doing so is that my RAW file contains the purest signal my cam could grab while using a specific A and S during those light conditions. It is exactly the same signal the higher ISO used but since I am not changing my conversion ratio, I am also not affecting my quality. My dynamic range will be larger and the colors better at the price of a little more work in post.

I don't know if it works for all cams. If your cam's amplifier results into a better signal than the adjustment in the RAW editor, using ISO is the better choice. But if you can do better in post than what the cam does, don't use ISO, preserve all quality and adjust to your liking in post.

That's all there is to my current approach to shooting RAW.

It's far from an 'unknown world' simply because the entire process has been created by mankind. It's not like someone tossed a prototype digital sensor behind a lens and a file magically appeared on a memory card. Somewhere, some software engineer had to create the code to convert the analog signal to a digital file. And we are now working with distant descendants of that process.

The unknown can become known. One only needs to Google it.

480sparky said:

It's far from an 'unknown world' simply because the entire process has been created by mankind.

Click to expand...

Agreed... ADC's (Analog to Digital Converters) have been around a long time.

All in all, it's a simple process of classifying varying voltage levels into levels of brightness and assigning them a value on a binary scale (0's and 1's). DSLR's would need to have at least an 8-bit ADC to convert the incoming analog signal into 256 shades of black and white but they probably toss in a few extra bits for error correction and other such pedestrian tasks. I don't really know but I'd hazard a guess 12-bit ADC's are probably standard, or at least common, but that's just a WAG on my part.

.....

480sparky said:

The unknown can become known. One only needs to Google it.

Click to expand...

I like that... It would clearly be unimaginable to be back in the days before internet.

The actual difference details between two digital ISO is probably better known than the difference between two film ISO.

It's not unknown in the sense that we don't know what goes on. It's unknown how it is done and what ratios are used. How it affects colors is also something I don't have information on yet.

There's a ton of information out there but about this I find little.

Noise, Dynamic Range and Bit Depth in Digital SLRs

Caveat: I didn't read the whole thing and, assuming I could manage to stay awake through long enough to do so, probably wouldn't understand 95% of it. Still, it looked relevant and I know some people loooove to geek out on this sort of thing, so yeah; go nuts. I'm out the door to go shoot a little.

....

Horoscope Fish said:

Noise, Dynamic Range and Bit Depth in Digital SLRs

Caveat: I didn't read the whole thing and, assuming I could manage to stay awake through long enough to do so, probably wouldn't understand 95% of it. Still, it looked relevant and I know some people loooove to geek out on this sort of thing, so yeah; go nuts. I'm out the door to go shoot a little.

....

Click to expand...

I've read parts of it in regards to noise and it has some interesting information. I also gathered some info at other places and then put some pieces of the puzzle together and started experimenting.

I can't yet fully explain why it works (for the D750) but it does. My night shots I already used the exact same approach and now since two days I use it in birding too. It's not that I can suddenly shoot noise-free since I have the exact same noise I would have when I use high ISO. But by not having the signal amplified and adjusting it in LR, I appear to have less noise in my end result, a higher DR and better colors.

You can check my bird shots of yesterday and today and compare them with those I previously posted. The only thing that changed is that the older are shot at higher ISO while the latest are shot at 100 and adjusted in post.

This article explains how digital cameras work in a fairly straightforward way that is understandable.

How Digital Cameras Work

It also explains what changing ISO does in some detail.

wornish said:

This article explains how digital cameras work in a fairly straightforward way that is understandable.

How Digital Cameras Work

It also explains what changing ISO does in some detail.

Click to expand...

Thanks, there's a part I see exactly what occurs in practice but I lack the knowledge to exactly understand why it happens.

Increasing ISO with analog gain in digital cameras has a similar effect as shifting the histogram higher. Both become brighter, and both also boost the low level noise. Both CAN reduce dynamic range, but only if there are already high values that ultimately get shifted off scale and lost.

But there is a big difference between analog and digital.

Analog is the real world, and has no discrete steps. Analog has no clipping end point.

For example, this is where processing to remove the orange cast from color negatives differs. This is a very minor issue when using color filters in the analog light, trivial to do, no bad effects. But trying to shift digital data that much involves huge clipping issues, never very satisfactory. For example, film scanners do this job by increasing the exposure time of the blue and green channels, relative to the red channel. This mimics analog action of a color filter, and is much superior than trying it digitally later.

The main problem I have with ISO (since I know what it does) is that it is applied during A/D conversion and thus irreversible. I can shoot at 100 ISO and increase the exposure in post to match the high ISO shot but I can never lower the exposure of the high ISO and match the 100 shot. There's too much loss in the RAW itself. Mind you, I can match the exposure just fine but not the quality.

Any lowering of data loses dynamic range. Lowering existing data shifts higher values to lower values, leaving an unfilled gap at the top, which is less dynamic range. This is why the cameras do NOT say ISO 50, they call it LO 1, LO 2, LO3. Not the same specs as ISO 50.

I didn't bother testing the Low range on mine but I know for sure that any increase in ISO affects the quality and range of the data. Even when the DR is wider than the cam's capacity, the only difference between high ISO and normal (100) is that the clipping occurs differently between both but the higher the ISO, the more it clips in comparison.

wornish said:

This article explains how digital cameras work in a fairly straightforward way that is understandable.

How Digital Cameras Work

It also explains what changing ISO does in some detail.

Click to expand...

Interesting stuff: "As ISO is increased in a digital camera, less electrons are converted into a single ADU. Increasing ISO maps a smaller amount of dynamic range into the same bit depth and decreases the dynamic range."

Blade Canyon said:

Interesting stuff: "As ISO is increased in a digital camera, less electrons are converted into a single ADU. Increasing ISO maps a smaller amount of dynamic range into the same bit depth and decreases the dynamic range."

Click to expand...

That's the problem when conversion is done by different ratios into a value that remains subject to the same minimum and maximum.

Here's how ISO affects my cam at certain aspects of its quality. ISO does not just add some noise as some might believe. This is true for every cam once you increase ISO. The numbers or % loss of yours is not necessarily identical to mine.


ISO 100
SNR 18%: 44,9dB
Dynamic range: 14,53Ev
Tonal range: 9,59 bits
Color sensitivity: 24,8 bits

ISO 200
SNR 18%: 42dB
Dynamic range: 13,73Ev
Tonal range: 9,09 bits
Color sensitivity: 24 bits

ISO 400
SNR 18%: 39,5dB
Dynamic range: 12,86Ev
Tonal range: 8,7 bits
Color sensitivity: 23,1 bits

ISO 800
SNR 18%: 36,8dB
Dynamic range: 12,1Ev
Tonal range: 8,27 bits
Color sensitivity: 21,7 bits

ISO 1600
SNR 18%: 34dB
Dynamic range: 11,17Ev
Tonal range: 7,81 bits
Color sensitivity: 20,2 bits

ISO 3200
SNR 18%: 31dB
Dynamic range: 10,39Ev
Tonal range: 7,32 bits
Color sensitivity: 18,7 bits

ISO 6400
SNR 18%: 27,9dB
Dynamic range: 9,47Ev
Tonal range: 6,8 bits
Color sensitivity: 17,2 bits

When shooting my ISOless style, it is also good to remember I shoot RAW 14bit lossless. In how much this makes a difference with 12bit I don't know.

I understand very well what you are saying, but please consider this.

The problem with this approach is that you reduce the number of discrete levels available for the image. If you shoot with high ISO, you fill the entire histogram, let's say 256 levels in each channel. In your approach all this information is compressed into, let's say, 25 discrete levels in each channel. Then, you expand these 25 levels into 256 levels in post-processing. By doing this, you will introduce discrete artefacts, who's appearance will depend on the image: banding, pixelation, loss of quality, noise (although different than high-ISO noise).

When the amount of light reaching the sensor is low, it creates noise. Whether you amplify the image onboard (high ISO) or in post-processing (adjusting exposure), the noise will be there (it might be in a different form, though). There is no magical way of removing noise in low light.

My histogram is not an accurate representation of my data. It's very easy to see that. If you open a correctly exposed shot in post and overexpose or underexpose it by some stops, the histogram moves to the left or right but that does not imply you lose data. It's just about data that is shown or not shown at this moment.

What you see in my original shot is the exact same. The data is there but when I load the shot, it displays according the settings of the image I see. Once I start adjusting the black and white point and correct exposure, my histogram normalizes to what we are used to. I don't affect any level while high ISO affects them very much.

My photon shot noise is indeed there no matter what I do. But I am in control over exactly what I increase while ISO does so indiscriminately in the whole shot.

Here's one of my night shots. The SOOC, processed version and one overexposed. I added the histograms.







All three contain the exact same RAW data in the file. It's just differently displayed and that's what the histogram is about. It's relative to the values of how the shot is displayed on your screen. That's why the histogram of the original shot I show here is not necessarily a good or accurate indication of the data it contains.

This far I don't see any disadvantages besides it being awkward to see when importing. I'm party used to it since my SOOC night shots look that dark too.

But I'm only shooting it since a couple of days and time may tell if there are limitations. It's possible when light is even lower I might lose info in the darker parts because there's too little for 100 to convert to a different level but enough when multiplied by ISO. I don't know that yet.

I know during the night it only has advantages this far and I'm shooting 100 since a while for those.

I said it before but just in case; I only do this when I can no longer increase my light intake by either opening up or slowing down. It's when ISO is the only option left, I choose not to.