Over the last six months or so, many people have asked for my opinion on ISO Invariance, and I didn't really have one, because I hadn't done any tests for myself, so I corrected that, and today will share my results. Let's start with a bit of information about ISO...
Thank you for visiting!
Martin Bailey has been releasing weekly podcasts and blog posts since 2005! Almost all of the 760+ posts here contain a full text article with photographs and illustations, and take at least one day, sometimes three to four days to produce.
You are welcome to listen to the Podcast with the audio player and follow along with the images discussed below.
If you value what we do, please consider a Patreon contribution of $3 or more to unlock the full text of more than 760 posts and gain access to the exclusive MBP Community. There are also higher tiers with various benefits, some including one-to-one Mentorship.
Please visit our Patreon site for full details, and take your photography to a whole new level! | Become a Patron! |
![]() | Existing Patrons please login to access posts and benefits. Thanks for being awesome! |
Image Gallery
ISO 100 to 3200
ISO 100, f/8 for 1 second
ISO Invariance Tests
ISO 3200 100% Crop
ISO 1600 +1 Exposure 100% Crop
ISO 800 +2 Exposure 100% Crop
ISO 400 +3 Exposure 100% Crop
ISO 200 +4 Exposure 100% Crop
ISO 100 +5 Exposure 100% Crop
ISO 100 for 1 sec and ISO 1600 for 1/30 sec 100% Crop
Invariance Test White Cat on White Background
ISO 100 + 1 1/3 Exposure Compensation in Camera
ISO 100 - Zero Comp - Pushed + 1 1/3 in Post
ISO 800 - Zero Comp - Pushed + 1 1/3 in Post
ISO 800 + 1 1/3 Exposure Compensation in Camera
ISO 1600 - Zero Comp - Pushed + 1 1/3 in Post
ISO 1600 + 1 1/3 Exposure Compensation in Camera
ISO 3200 - Zero Comp - Pushed + 1 1/3 in Post
ISO 3200 + 1 1/3 Exposure Compensation in Camera
ISO 6400 - Zero Comp - Pushed + 1 1/3 in Post
ISO 6400 + 1 1/3 Exposure Compensation in Camera
ISO 12800 - Zero Comp - Pushed + 1 1/3 in Post
ISO 12800 + 1 1/3 Exposure Compensation in Camera
Average Scene Tests with X-Rite ColorChecker SG
ISO 100 - Zero Compensation
ISO 100 Pushed 1 Stop in Post
ISO 100 Pushed 2 Stops in Post
ISO 100 Pushed 3 Stops in Post
ISO 400 - Zero Compensation
ISO 400 Pushed 1 Stop in Post
ISO 400 Pushed 2 Stops in Post
ISO 400 Pushed 3 Stops in Post
ISO 1600 - Zero Compensation
ISO 1600 Pushed 1 Stop in Post
ISO 1600 Pushed 2 Stops in Post
ISO 1600 Pushed 3 Stops in Post
ISO 6400 - Zero Compensation
ISO 6400 Pushed 1 Stop in Post
ISO 6400 Pushed 2 Stops in Post
ISO 6400 Pushed 3 Stops in Post
If no images are displayed here, please refresh your browser.
cURL error 28: Operation timed out after 5000 milliseconds with 0 bytes received
To view this content, you must be a member of Martin's Patreon at $3 or more
14 Comments
Rick Nash
Posted at 01:52h, 26 AprilThanks for taking the time to explain ISO invariance. My conclusion based upon your discoveries, when using the Canon 5DSr or 5DIII, in order to keep it simple, regardless of ISO one might as well employ an ETTR strategy and then “fix” in post, Trying to maximize exposure by taking into account ISO invariance would be too complex for myself. I suspect the degree of invariance is influenced by both the amount of light and the colour of the light. In other words, each photograph would be a unique situation.
Are you planning on testing the 80D or 1DXII to see what advantages that sensor may have regarding ISO invariance and DR?
Martin Bailey
Posted at 10:21h, 26 AprilHi Rick,
Thanks for the comment.
I’m pretty much with you. I think there will be times when I might employ some shooting techniques that make use of what I’ve learned, but in general, I’m going to continue to use ETTR. I just feel more comfortable shooting ETTR and still believe it produces the highest quality images.
I won’t be testing the 80D or 1DXII, as I don’t own these cameras, and I am not supported by Canon etc. so can’t get a loaner either.
Cheers,
Martin.
SM
Posted at 06:23h, 26 AprilThanks for the interesting article Martin! One (slightly unrelated) question – I have often read that larger megapixel images will inevitably look a little soft when zoomed in to 100%. However if you viewed a 50MP image compared to, say, a 24MP image at a set size (e.g. full screen but not 100%) on a good display (e.g. a MacBook Retina), would the 50MP image look ‘sharper’ due to the fact that more pixels are being viewed in the same size space?
I hope that question makes sense!
Martin Bailey
Posted at 10:35h, 26 AprilIf the camera has quality pixels and the ability to resolve light through the lens to a small enough point, then there is no softness in the image, even at this size. My 50MP images from the 5Ds R are as sharp at 100% as the images from my 22MP 5D Mark III.
If the images were shot with a lower quality lens though, for example, resulting in soft images, then to a degree, yes, if you compared it to say a 24MP image at 50% magnification instead of 100%, then they’d probably look very similar. It would really depend on how soft the original is though.
The Retina screens do also make images look sharper. The pixels are so dense that it’s more difficult to see if the image is critically sharp sometimes. There have been times when I’ve thought an image was sharp on my MacBook Pro Retina screen, but then when I double check on my iMac, with less dense pixels, it can appear a little softer. If I nailed focus and there is no camera shake though, a really sharp image will look sharp as tacks on both. Some people check images at 200% on the Retina screen to get around this, but I personally don’t do that. In general you can tell if the image is sharp or not, and I always have the final check on the iMac as well, so it’s not really necessary in my workflow.
I hope that answers your question.
Charlie McDonald
Posted at 07:50h, 26 AprilInteresting stuff. I haven’t thought about exposure techniques for a while now. Even though I’m sporting an older camera it’s nice to hear.
Martin Bailey
Posted at 10:36h, 26 AprilI’m pleased you found this interesting Charlie. Thanks for the comment!
Shane Baker
Posted at 10:30h, 27 AprilThanks Martin! “Invariance” is a new term to me, but in the past I’ve noticed I can push the shadows in my D800 files to blazes. Now I know why.
Martin Bailey
Posted at 10:33h, 27 AprilHi Shane,
Yes, it’s a relatively new term to me too. This is one of the reasons I set some time aside to do these tests to see how my 5Ds R faired. It’s good to at least understand how these things work to help us make better decisions in the field.
Thanks for taking the time to comment!
Cheers,
Martin.
Mark Casebeeer
Posted at 20:28h, 30 AprilThank Martin, another great article. This really hit home for me. I use a 7D ll and at times been confused with post processing results. I always try to ETTR, I just feel I get better results. What is a little confusing is why in post processing can sometimes I lighten a darker photo and it looks good? This has led me to think the same (why am I ETTR)? Your explaining ISO settings does influence this really helps me understand why sometimes it works and other times it’s a disaster. Pushing at lower ISO setting can work but trying to push a underexposed photo in the 3200 – 6400 range just does not get it.
Martin Bailey
Posted at 10:44h, 01 MayThat’s great Mark. I’m pleased this was useful for you.
For sure, the main benefits are when brightening dark subjects shot dark at ISO 100. In other situations there aren’t any real benefits for Canon shooters at the moment.
I was out on Friday shooting some dark night scenes with bright lights, and I could have shot darker, but I continued to shoot ETTR. I just know it better, and feel more comfortable with this method. As dynamic range increases I will probably start to rely more heavily on the benefits of ISO invariance.
Dave T
Posted at 17:33h, 13 MayThanks for a very interesting post Martin. I had never heard of ISO variance but your explanations were very clear and I think I have grasped it now. In wildlife photography I have often heard some shooters talk about underexposing slightly to give them faster shutter speeds and avoid the risk of blowing out exposure if using ETTR. It makes sense in some regards, but I have found that depending on the amount of dark and shadow detail in the image, it would often lead to the introduction of noise when lifting that detail in post production. So, like you I have stuck with ETTR.
I have recently read quite a bit about the new 1DXII and seen some presentations by a couple of Canon sponsored photographers, who showed the results of shooting in low light and it appears that the new model is somewhat different to previous models as it deals much better with noise in the shadow and dark areas of an image. Lifting that detail in post production has shown it to be far more tolerant with cleaner results. That camera is out of my reach financially, but I am hoping that the technology will cascade into other models down the line, like the 5DIII replacement.
So maybe, there may be case for me to re consider shooting to the right in the future as the newer models role out. It’s something I hadn’t thought much about, but thanks top your article I will do in the future.
Martin Bailey
Posted at 18:37h, 14 MayIt’s certainly an interesting development Dave.
I think based on my recent tests, that I could for example shoot my white on white winter scenes a little darker, to avoid the occasional over-exposed image, and no that I could increase the exposure in post to bring the whites back. I could do the same with darker scenes too.
It will feel strange, because I’ve been exposing to the right so long it’s more natural for me, but I know the option is there now.
With the larger pixels of the 1D X Mark II I’m sure that’s going to be the case. As I mentioned on G+, I won’t be buying a 1DX2 but as dynamic range increases in all cameras it’s going to be interesting to see how this affects the way we shoot.
Iratxo
Posted at 00:53h, 14 AprilVery interesting post Martin. Thanks a lot for your time. I have recently listened to this and your ETTR posts. You always say that exposing to the right improves overall SNR in the images you get. It kept me thinking a lot on my way to work.
The noise in any analogue to digital capture system has always two sources: thermal noise and quantisation noise.
When using ETTR what you do is minimise the quantisation noise introduced by the ADC in the camera. Added to that, the logarithmic way the eye sees makes us more sensitive to this noise as we have a great latitude in our eye.
But, there is other very important noise in the system: the thermal noise. The thermal noise is the base noise the sensor, as an analogue system, has. It is always present. When we expose the sensor with light we have the sum of the desired signal (the light photons in the sensor) plus the thermal noise of every photodiode.
Using higher ISOs is said to be worse for noise, but in reality what it is is worse for the signal. Let’s say we are shooting in Av Mode. Asume that for a given scene we need a shutter speed of 1/100 for ISO 100. This means that we are going to have a value of say S photons at the sensor and a value of N noise. This gives an SNR of S/N. If we increase the ISo to 200 and change the shutter speed to 1/200 to maintain the aparent exposure level, what we get in the sensor is S/2 photons and the same amount of noise N. The SNR is then (S/N)/2.
If the value of N is low enough then the quantisation noise introduced by the analogue to signal converter is going to be the main source of noise, but if it is not, then the main source of noise is going to be the thermal noise.
IMHO this is why shotting with the highest possible quantity of light in the sensor is the best rule of thumb always. This usually means (but not always) using the lowest possible ISO, thus increasing real exposure (aperture and shutter time).
The truth is that the ISO is not changing the real exposure, only the apparent result.
Martin Bailey
Posted at 10:30h, 14 AprilBeautifully written explanation Iratxo!
Of course, what you say is completely accurate, and I agree one hundred percent.
In an ideal world, where we have enough light to create images that are bright enough without increasing the ISO, then creating images as you say, with “real exposure” is ideal. This is what we do when it’s bright enough.
But, there are situations when there is not enough light to capture the subject quickly enough to avoid subject movement. For example, photographing an eagle in flight on an overcast day. At ISO 100 the shutter speed might need to be lower than a 1/8 of a second, and of course even for a panning shot that’s too slow to get a sharp image.
When shooting handheld in inside a Himba hut, the only way to get a shot without you and/or the subject moving during the exposure is to increase the ISO. And, when it’s necessary to increase the ISO, in my experience, it’s still better to increase it so much that you Expose To The RIght. This results in the least amount of noise in the final image.
I hope this makes sense.
Regards,
Martin.