ISO invariance explained better – and why you don’t need to read this article

Below you can see the shot this came from
No, not the amazing Mystic Meg, but the amazing 16 meg ISO invariant sensor in a Fujifilm X-T1 camera. This was cropped from an underexposed image. You can see the original further down this article.

There has been a lot of talk about ISO invariance of late, all of it sounding mighty complicated. Here’s a short simple explanation of what ISO invariance is, how it might effect the design of cameras in future and is this imperative information? (Standby to be disappointed on that last item)

What the ISO dial REALLY does

On higher-end digital cameras there is an ISO control. This is a volume control. It controls how loudly the camera records your image. That’s all. Until recently, digital cameras have broadly been doing this by controlling its sensor’s ISO variance. This works by changing a hardware setting on the camera’s sensor to lower the threshold at which light photons will trigger an electric current in each pixel, or more accurately, photosite. The more the photosite gets triggered, the greater its apparent volume or brightness, which is then digitally recorded into the RAW file.

On a sensor that is ISO invariant, the ISO control doesn’t change the sensor’s hardware settings. Instead, it instructs the camera software to record the raw file as if all the photosites were set to a louder multiple. The higher the ISO, the higher the multiple. The actual sensor itself doesn’t change to a louder amplitude.

What this means is that on a truly ISO invariant camera, the ISO dial simply sends an instruction along with the raw file that says ‘play back loudly’ or ‘play back quietly’. The difference in quality between, for example ISO 200 and ISO 3200, assuming that in both cases shutter and f-stop have been optimally set,  is exactly nothing. Shocker!

If you take one thing from this post it’s this – the ISO dial is just a volume control

Most cameras now use some formula combining ISO variance and ISO invariance techniques. That is to say, some will use hardware multipliers, some will use software multipliers and most will use a combination of both.

The photos here show the results of under-exposure on a predominantly ISO invariant camera, the Fujifilm X-T1, and subsequent exposure correction of the images in Lightroom.

The original out-of-camera raw file
No this isn’t an accident… well, ok, it is. This is the unprocessed image (see image below) that came from the camera. What do you think, a little under-exposed?
The uncropped image
Here’s an uncropped version of the same image after correcting for exposure. Very little other work has been done to this image aside from running it through my custom Fujifilm preset in Lightroom.
A crop of the crop
Here’s a hard crop. OK, the detail is a bit noisy.

ISO control – the end is nigh

ISO invariance is highly desirable, since it offers much greater tolerance to under-exposure. The brightness with which you capture an image is already becoming less relevant, as the photos here prove. ISO invariance means that only the relative brightness of each image artefact is required information.

For this reason ISO invariance is likely to prevail. As the hardware design of the sensors improves, ISO control becomes increasingly a software issue. A time may arrive in the not distant future when all the exposure control you need in-camera can be achieved with the exposure compensation dial. That, plus the post-processing software to get the exact look you are after.

HDR is also likely to fall out of favour since it will be creating multiple copies of exactly the same image data from the processor. The only difference between them will be the instructions to the raw converter on how to proceed.

It’s worth noting that the photo above used the maximum exposure compensation currently available in Lightroom – equivalent to about 3 stops. Clearly Adobe will need to have a design review sooner rather than later.

Update:  If you’re a Fujifilm user you will be interested to note that Rico Pfirstinger writes the following in his book The Fujifilm X-T1 111 X-Pert tips

The sensor in your X-T1 is calibrated to its native ISO 200, and this remains the same no matter what ISO you set in camera.

Up to ISO settings of 1600, the camera uses a combination of analogue [hardware] and digital [software] signal amplification, and the results of this is burned into the RAW file. Higher ISO settings beyond 1600 are realized by digitally pushing the ISO 1600 RAW file during RAW processing.

 

What have we learned from this?

If you take one thing from this post it’s this – the ISO dial is just a volume control. The higher you turn it up, the brighter your images are recorded. By now, you are probably wondering if you need to even know about ISO invariance. The answer, in a word, is No. If you really want to let out the inner geek, and you have an ISO invariant sensor in your camera, you could choose to set the ISO dial to, say, 200 and then sort out the exposure in the post-processing workflow.

Or you could just carry on taking for granted what you already knew, which is that as cameras get better, you can increasingly get away with badly under-exposing your images.

So there you go. There really was no reason to read this article.

Here's another image that had the exposure set in post-processing
For good measure, here is another example of ISO invariance put to work – or as others might put it, me screwing up the exposure setting. Again, very little work has been done on this image. Camera: Fujifilm X-T1 (16mb X-trans sensor)
And here is the original
And again, this is what it looked like prior to the exposure being corrected in post-processing.
A crop from the above photo
Here’s a tight crop to show that all the essential detail has been preserved in the image

 

Leave a comment