Can anyone explain ISO invariance to me?

double post
 
HoppyUK said:
A bit more on the practical applications of ISO-invariance.

In these debates, it's quite common for some to argue that with an ISO-invariant sensor, exposure settings don't matter - just shoot at base ISO with whatever shutter speed and aperture you want for creative purposes and adjust brightness in post-processing - based on the fact that photon capture will be the same whether whether ISO is adjusted or not. That may be true in very rare instances where base ISO also happens to be right for normal exposure setting purposes, and there's certainly more leeway, but given that best exposure always depends on maximum possible photon capture regardless, the only way of optimising that is to adjust exposure settings in the normal way so you know where you are. With that done, an ISO-invariant sensor may allow you to make further beneficial adjustments.

ISO-invariance changes the popular method of optimising digital exposure known at ETTR - Expose To The Right (of the histogram). This basically means over-exposing to put more detail (photons) into the shadows, then darkening the image back down again in post-processing but with shadow detail retained. It works well but there is a downside because over-exposure will blow highlights so a careful judgement must be made to make sure only unimportant highlights are effected - basically, it's a compromise.

With ISO-invariance, that technique can be changed to ETTL (Expose To The Left) which ensures that all highlights are retained and shadows can be lifted in post with little or no penalty. In some situations with very high dynamic range where you might otherwise need multi-exposure HDR technique, it can make a big difference, eg most back-lit situations where you want detail in the very bright background but also good detail on the shaded foreground. Unlike HDR, ISO-invariance also works with moving subjects as everything is captured in a single shot and since it often means you can use a faster shutter speed, that's another bonus.

A practical downside of ETTL and deliberate under-exposure is that the LCD image goes dark, sometimes very dark to the point of useless, and the histogram will also reflect that.

Exploiting the full potential of ISO-invariance only works when shooting Raw. It's not much use with JPEGs where tonal values are pretty much locked with very little scope for adjustment in post.
Click to expand...
Welcome to the talking to yourself club,now 3 members :D
 
Oi! I'm listening and learning too!

That's my one thing learned for today now, I'm good :D
 
Last edited:
Terrywoodenpic said:
.
That was certainly true for incident light meters. which peg the tone extremely accurately.
However digital cameras, and raw files in particular, can often give a better starting point and final result when exposed either to the left or right of this point to avoid clipping.
In camera Jpegs, though usually near enough, give less opportunity to take advantage of the entire tonal range captured by the sensor, especially in contrasty conditions.
Most photographers are well aware of this, and take such things into account when setting the exposure.

Most digital cameras are for all intents and purposes invariant over three or four stops. That is certainly true in practice for most Fuji X cameras.
Click to expand...

There is only one definition of technically 'correct' exposure - as above - and all cameras are calibrated based on that assumption.

ISO-invariance basically means there is no image quality difference between raising ISO in-camera, and brightening the image in post. That's not the same as being able to pull some extra shadow detail out of any digital image with more or less acceptable results, even if it still looks okay. Sometimes you can, but sometimes you can't do much at all without excessive noise quickly becoming apparent. Different cameras vary a lot.
 
This has gone off track a bit, but it's interesting...
Technically, "correct exposure" records something that is 12(18)% grey luminance at 12(18)% grey (middle of histogram). But that's not really the way DSLRs and reflectance meters work. To start with, most digital cameras have the exposure offset 1/2-1 stop (the fairly arbitrary REI, recommended exposure index). This is probably to protect the hard highlight clipping characteristic of a silicon based sensor... it also contributes to most cameras having ~1 stop recoverable in the raw file.
Additionally, a reflectance meter places whatever you point it at at 12(18)%... it's relying on you to determine what is mid tone and meter appropriately.

These days it would probably be better to think of "mid tone" as being the middle of the DR of the scene and meter for that (which is what matrix/evaluative/average metering does to some extent). In this manner you would always record the maximum DR possible regardless of the camera in use. If the scene has a DR too great for the sensor/ISO it results in some clipping at both ends, which would appear most natural. You can then decide how best to approach the issue if it's a problem. You can shift the exposure either direction in order to retain more at one end while sacrificing more at the other end. You could choose to shift both directions and combine in post. Or you could choose to reduce the ISO and adjust the Ap/SS in order to receive/record more light, which will increase/maximize the amount of DR recordable.
It doesn't really change anything as far as techniques go... it's just a slight shift in defining "mid tones" and in understanding what is happening.

It is probably important to understand that the greater DR capabilities of a more invariant camera exist primarily in the shadows; due to the reduced noise floor... it's always "0-255" readable, it's just that what is 0 readable is reduced. You may decide that the camera is able to see farther into the darks than you generally need, and not as far into the whites as you would generally like. In which case you can choose to just offset the exposure by default (setting your own REI). I.e. you could just set a -1EC in by default. In fact, this is how I typically utilize the ISO invariance of my D810... I set it in manual mode with auto ISO enabled and with something like a -2 EC setting.
Funnily enough (to me anyway) in reality this is pulling the exposure/underexposing; but it is commonly referred to as Exposing To The Right (ETTR), which implies pushing the exposure/overexposing (which ETTR also requires in some instances)... that's probably why some find the ETTR/ETTL concepts confusing.
 
Last edited:
Whatever happened to forum etiquette? Not a peep out of the OP since his first post...
 
HoppyUK said:
Whatever happened to forum etiquette? Not a peep out of the OP since his first post...
Click to expand...
Oh gosh, sorry. Thanks for the reminder.

I do appreciate that people have tried to help, even if I don't understand some of the replies and I don't yet feel able to distinguish those which are genuinely helpful from those which just muddy the waters. That's the trouble when you don't understand something: it is statistically likely that a certain proportion of people who claim to know about any given subject actually don't understand it properly either, but I can't tell which ones they are! (And of course there's the Dunning-Kruger issue.) When I get some time I'll lock myself away in a dark room with a cold towel wrapped around my head and try to make sense of it all.
 
StewartR said:
Oh gosh, sorry. Thanks for the reminder.

I do appreciate that people have tried to help, even if I don't understand some of the replies and I don't yet feel able to distinguish those which are genuinely helpful from those which just muddy the waters. That's the trouble when you don't understand something: it is statistically likely that a certain proportion of people who claim to know about any given subject actually don't understand it properly either, but I can't tell which ones they are! (And of course there's the Dunning-Kruger issue.) When I get some time I'll lock myself away in a dark room with a cold towel wrapped around my head and try to make sense of it all.
Click to expand...

I'm sure you'll get a good enough understanding of it pretty quickly.

There's a little bit on this in DPR reviews there days, for example... scroll down to ISO invariance in this review...

https://www.dpreview.com/reviews/sony-a7-iii-review/6
 
StewartR said:
Oh gosh, sorry. Thanks for the reminder.

I do appreciate that people have tried to help, even if I don't understand some of the replies and I don't yet feel able to distinguish those which are genuinely helpful from those which just muddy the waters. That's the trouble when you don't understand something: it is statistically likely that a certain proportion of people who claim to know about any given subject actually don't understand it properly either, but I can't tell which ones they are! (And of course there's the Dunning-Kruger issue.) When I get some time I'll lock myself away in a dark room with a cold towel wrapped around my head and try to make sense of it all.
Click to expand...

:thumbs:

Dunning-Kruger - is that a fancy lager? :D
 
You must log in or register to reply here.
Back
Top