Invariance

Bill Pierce

Bill Pierce

Well-known
Local time
9:56 PM
Joined
Sep 26, 2007
Messages
1,407
I want to recommend a quick read on invariance. For people who rely on jpgs, it’s not much use. But for raw shooters taking advantage of it can actually insure more acceptable exposures and provide more room for interpretation in the digital darkroom. Using it as a tool is not for everyone; so, as always, your thoughts are more than welcome.

https://photographylife.com/iso-invariance-explained
 
Well Bill, I had to laugh at the first line in the piece:

"ISO invariance is one of the most talked-about topics in photography today . . ."

Having never heard the term "invariance" before, and talking about topics in photography fairly regularly.

I think I know what he's referring to in the article, just never heard the term "invariance" used to describe it. Will read the article and educate myself.

Thanks for the link.

Best,
-Tim
 
New one on me too. I do now have two cameras that can shoot raw, so I'll try to take it in. :)
 
'ISO-less' is an interesting topic and I understand that some of the newer Fuji X-series cameras are approaching that. A 12 page article with ~80 pages of comments ; )

Thanks for the link...
 
My first thoughts when I started reading was this article was about "push processing" digital raw files as referenced in articles like these:

Best way to shoot high ISO with a Leica M9

Just crank up the ISO, not so fast

Been using this technique with my D4, figuring the highest "Real ISO" is about ISO 800, and shooting low light at ISO 800 and increasing exposure of the RAW files in Lightroom.

But after reading the Invariance article, I'm a bit confused. It seems to talk about the same thing, but when you look at the referenced charts, I get really confused.

Here's the chart for my D4:

http://photonstophotos.net/Charts/RN_e.htm#Nikon%20D4_14

It seems to indicate that the Upstream Read Noise (the read noise before the analog amplifier) goes down when you crank up the ISO on the camera. That makes no sense. Trying to find the highest "Real ISO" is more confusing after reading the article and looking at the charts.

Best,
-Tim
 
I understand (at least I think I do) the theoretical underpinnings of the article, but I have to ask whether it makes any real world difference in print quality whether you shoot at a higher ISO or lighten the RAW file in LR, at least to the extent you are careful not to clip highlights. I can't recall an occasion where I have exposed at a higher ISO than 3200, and then only rarely.
 
Timmyjoe said:
But after reading the Invariance article, I'm a bit confused. It seems to talk about the same thing, but when you look at the referenced charts, I get really confused.
-Tim
Click to expand...

"These raw values are not appropriate for comparing camera models because they are not adjusted for gain or area." That's a quote from the site. I think what it means is "Don't worry; trust your results with your camera."
 
I guess what it means, is that it's safe to crank up the ISO to a certain point. But if you go beyond the critical point, the image suffers. But didn't we already sort of know that?
 
Rob-F said:
I guess what it means, is that it's safe to crank up the ISO to a certain point. But if you go beyond the critical point, the image suffers. But didn't we already sort of know that?
Click to expand...

The point is to what degree would the image suffer past that point. With an ISO invariant sensor there would be none as long as you stay within the range.

Good thing is you can confidently shoot for highlight with lower ISO, knowing you can push the shadow for several stops later with zero loss.
 
My digital cameras (Sony A7II and Ricoh GR) are both ISO invariant, and I regularly use that capability. Any time I expect that there will potentially be highlights in a scene that might blow out relative to the average metering - bright sunny days, city lights at night - I always just dial in -2EV to -3EV to cover the highlight, then lift everything in processing. Im not pixel-peeping but I've never felt there was any obvious issues in the results.

Just as an example, this was shot on the Ricoh GR with -2EV
45679200555_f07f51cc53_o.jpg
 
Err, haven't we been taking advantage of that ever since digital photography arrived? This article is hopelessly bloated and goes on and on and on unnecessarily. All it's saying is that sensors have a lot of latitude, and this is embedded in both camera settings (e.g. wide usable ISO range) and raw files (e.g. wide editable dynamic range). And attempting to give this a name ("invariance") that no one uses just adds pointless confusion...

Surely everyone using a digital camera knows that you have this latitude!

The photo below is from 12 years ago with a 6 MP camera that was ancient even then - to emphasise that what this article tells us is old news. It's an extreme case - very low ISO and very underexposed as I was shooting straight into the sun and didn't want to blow the highlights (including the very bright raindrops). See also my postscript under these photos.

I'd imagine every photographer would do something similar - even with film before digital was even an idea, with the aim of dodging and burning later in the darkroom.

Original
31984809827_1d02154d71_z_d.jpg


Adjusted
524622019_eff30253f7_o_d.jpg


I saw this shot developing with the weird light and colour of the rainstorm as the woman walked up the steps, and ran and grabbed my camera. In my haste I forgot that I'd set my camera to JPG only!

So, this is what you can do with just a JPG file!

Yes, raw would have increased the technical quality. Yes, the shadows have been lifted to an extreme so technical quality has again suffered (e.g. there's unwanted noise). But it's good enough - the exposure and colours match the reality of what I saw, and I've printed it at A3 size.
 
I have not read it properly but I presume this article refers to the characteristic which has resulted in some experienced users recommending that those who own, say, a Leica M8 should shoot these at low ISO settings and deliberately under expose the image, then correcting the image in post. (Shooting in RAW of course).

I have found this to be especially useful since beginning to use Lightroom as it has particularly good capabilities for adjusting different tonal ranges within an image separately. Thus I am readily able to pull back highlights even more without making an already under exposed image even darker then adjust those shadow areas in the opposite direction.
 
Well I never shoot above base ISO on my Leica M9. I always rely on adjustments to the RAW files during processing and this works very well indeed. I've done it for years. Not so with the Sony A7II - I adjust ISO in camera because experience has taught me that its files are far less tolerant of being heavily adjusted afterwards and noise is significant if I do so. I'd suggest that you don't really need to read it up as much as try it. Intriguing that someone has give this a name though.
 
peterm1 said:
...I presume this article refers to the characteristic which has resulted in some experienced users recommending that those who own, say, a Leica M8 should shoot these at low ISO settings and deliberately under expose the image, then correcting the image in post. (Shooting in RAW of course).
Click to expand...

Peter,

This is how I first learned about this concept. I always thought my M9 was a good sunny day camera, but suffered in low light. I read the two article linked to above, and suddenly, shooting my M9 at ISO 640 and EV -1, then correcting in post, gave the camera a whole new life.

Best,
-Tim
 
I utilize negative exposure compensation quite often vs. raising ISO values to maximize highlight retention and raise shadow values in post processing, both in low light and during regular daylight with heavy shadows as I do not like to see beautifully exposed landscapes with large swaths of overexposed, blown clouds in the sky. 1-2 stops isn't unusual and works out very well. Depending on the raw converter you use, DXO's camera/lens profiles tend to extend one's ability to do so and produce excellent results.

Mirrorless cameras with their electronic finders and histograms makes it easy to work out the exposure maximums and minimums needed wile viewing the subject before pressing the shutter release, allowing one to stretch the files in post-processing



 
ISO Invariance

ISO Invariance

What's invariant is the camera's read noise level

There are two main time-dependent noise sources.

One is the read noise which describes noise from electronic components (sensor, analog DC gain and the analog-to-digital controller).

The other is photon (a.k.a. shot) noise. Photon noise is inherent to the conversion of light energy to photoelectrons. Photon noise is beyond human control. It is identical for every camera (when using the same exposure). However the relative amount of photon noise depends on exposure (shutter time and aperture). The relative level of photon noise decreases as exposure increases. For ISO invariant cameras this

When a camera is ISO invariant the noise is dominated by photon noise. Except at extremely high camera ISO settings, the read noise levels are constant. This is the best you can do.

When sensor under exposure is inevitable, the only way to maximize raw-file, signal-to-noise ratio is to understand your camera's read noise vs ISO behavior.

Above base ISO settings older designs relied on DC signal gain to minimize read noise levels. For some cameras the read noise levels would decrease as the ISO parameter increased.

Here's one example.

Using ISO 400 decreases read noise levels compared to ISO 160!

For the sake of completeness, some of the newest sensors use dual-conversion gain technology. Now there are two separate ISO invariant regions.
 
willie_901 said:
For the sake of completeness, some of the newest sensors use dual-conversion gain technology. Now there are two separate ISO invariant regions.
Click to expand...

William - Do you know any sources that give the two ISO's that are the two base levels for the two regions for specific cameras? I think I read the M10 was 200 & 800. It would be nice to know for those of us meshed in protective underexposure.

(Next on the topic list - the exact opposite, maximum exposure, exposure to the right, for capturing the greatest brightness range.)
 
Is there an 'Invariance for Dummies' book? I never let my Histogram go to highlight blow. So, I'm always underexposing (in RAW a little). RAW Therapee is a great editing program to do the retrieval that the author discusses. But I've still not sure after reading this article if I doing it right.
 
I guess I’d rather use high iso and see my exposure in my evf than be 5 stops underexposed and see nothing. With ovfs, you don’t have this issue.
 
You must log in or register to reply here.
Back
Top Bottom