ISO Is Totally Fake

ISO is known as one of the exposure factors. The higher the ISO the more brightness. Camera manufacturers are even boasting about how high their ISO is going when they want to sell you a camera.

In this new video from Tony Northrup, he is explaining why ISO in modern cameras is fake. First things first, ISO is an international standard. ISO is an acronym for International Standard Organization. Back in the days of analogue cameras, they needed a standard for how sensitive film is to light. It would not make much sense to have two companies making two types of film, with different light sensitivity, calling both of them ISO 100.

Today many digital cameras are basically ISO-less. This is also called ISO invariance. It does not matter to the final photo whether you photograph it on ISO 100 or ISO 6400 given the shutter and aperture stays the same. Confused? Obviously, the ISO 6400 photo is brighter out of camera, but if you increase, the ISO 100 photo with the same amount of stops in post processing you will end up with the same amount of noise in the final photo as the ISO 6400.

As Northrup points out, that gives a potential for photos at ISO 3. Yes ISO 3. It is basically the same process as stacking images in post and averaging out the noise in the photo. Check an earlier article of mine to see how I apply this to my drone photos. What is really important for a clean photo is the "signal-to-noise ratio."

Check out the video above. If you are tech savvy this might not be new to you, but it is still very important to how you think about your settings when capturing your photo.

Log in or register to post comments

41 Comments

revo nevo's picture

add 5 stops to 5Dmk3 raw file and you will see your worst nightmare so no not every sensor is ISOless

Ivan Lantsov's picture

in film it ASA not iso

Adriano Brigante's picture

It's been ISO since 1974...

Ivan Lantsov's picture

I older than that youngster

Adriano Brigante's picture

What are you talking about? The International Organization for Standardization published the related standard in February of 1974 (called ISO6:1974, "Determination of ISO speed of monochrome (black-and-white), continuous-tone photographic negative materials for still photography").

Ivan Lantsov's picture

I learn ASA, you learn iso, what it matters? in 1962 kodachrome is asa 25. Is good color not like today, today color is junk

Adriano Brigante's picture

And in 1942, Agfacolor was 15/10° DIN. What's your point?

Ivan Lantsov's picture

not fight you go away

imagecolorado's picture

I'm guessing Tony doesn't understand electronics very well.

Why is that your guess? He perfectly described what is actually happening with ISO and digital sensors these days.

imagecolorado's picture

Photography Iso tables standards are actually multiple standards for measuring film characteristics per given amounts of light. Digital cameras are electronic devices, so there is no direct correlation except mathematically. There are multiple implementations of how ISO sensor sensitivity is calculated, the manufacturers may choose which standard they use. The choice is based on how their electronics are designed and the mathematics are correlated to provide a consistance of exposure within the industry. Variations from model to model and brand to brand exist but the mathematical relationship has remained constant (logarithmic at present)

Iso invariance is a made up term to explain the vague generalities of sensor and circuit designs which the typical person would find beyond their technical ability in any other way but for generalization perceptions of the results of changing settings.

Nothing he says addresses the degradation of signal quality when the gains of amplifiers are adjusted to compensate for different amounts of light collected. His claim that a camera is iso invariant and any iso can be used and then adjusted in post processing is false. Signal degradation will occur using that false technique. Some cameras will do better than others. Improper operation can't be masked with buzz phrases.

I see no reason to change the terminology either. ISO as it relates to digital photography doesn't really alter anything mathematically from the photographers standpoint and it does maintain continuity of exposure theory. What happens inside the camera is the manufacturer's attempt to keep that continuity.

Keith Davis's picture

I always thought some noise was the result of signal conversion from analog to digital out of the sensor. ISO represents amplification of signal rather than the sensitivity of the sensor. Otherwise high ISO does not introduce noise but the amplification lets us see the noise that is already there. The less noise at higher ISO would be an indicator of the quality of the sensor and software. So an acceptable image at a higher ISO would be a good measure of performance if ISO were truly standard... I think.
So bottom line to me is ISO is an important measurement if taken with a grain of salt.
Butch

Add to the mix the difference between F numbers and T and the external meter can be way off. The only problem with his comparison might be that he use different lenses with different T values :-)

revo nevo's picture

Nope his comparison is right. Fuji is darker than others (and I use Fuji)

William Faucher's picture

Possibly because the T-stop of Fuji lenses is higher than that of other lens brands. This would cause darker exposures.

revo nevo's picture

NO! You can adopt canon lens and it would still be darker. THis is just the way Fuji metering works

William Faucher's picture

Do you have any empirical evidence of this? I am curious to see some test results.

William Faucher's picture

Interesting read, but this just furthers my point. This test doesn't factor in the variables of T-Stops. It's a small but critical thing to factor out all variables. You would need to shoot with the exact same lens at the exact same shutter speed, ISO, aperture. Only variable being the body, only then can you accurately determine if it is the sensor that makes darker shots or not. Without this method, comparing is truly worthless, and produces flawed results.

I'm not disagreeing with you, it is entirely possible that Fuji's metering does make things darker. I don't shoot fuji myself. I am merely pointing out that unless a test factors this in, results are skewed and ultimately worthless/misleading.

revo nevo's picture

Than do your own test. That would mean that fuji can't make as bright glass as others including zeiss and others that make glass for fuji cameras

Fuji even has two T stop rated lenses for sale (fuji mk lenses) and they are T 2.9

And with auto settings Tstop does not matter since camera does not care and does not know about it. Camera makes the scene bright as it tinks it should be and it makes it a bit darker than others

So 0 Ev is darker on Fuji than on Nikon. T stop does not have anything to do with it since it would compensate it with slower shutter speed in A Mode

As I stated on his video, back with film, Kodak VPS was rated at 160, but we all set it as iso 100. DXO regularly reports lenes rated at 1/4-1/3 of a stop slower than stated by the manufacture. Use the information as a guide and adjust as needed.

Gordon Cahill's picture

Yep. Same with Velvia. And we used to "discuss" whether it was really 32, 40 or 50... Once upon a time we'd buy a new camera and then test it to see how the metering compared to our old camera. Then we'd make an adjustment and go shoot instead of worrying how it compared to another camera from another brand that we didn't own.

Stated ISO has been loosely applied by cameras manufacturers since they developed the standard. Nothing has changed except that now it's a big deal.

Gordon

Jan Kruize's picture

This man is a star in making a simple thing complicated.

AJ Scalzitti's picture

No, this is incorrect. ISO is a mix of software and electronic gain, understanding your specific camera and when it add's gain is important. For example my older Canons worked at multiples of 160 but everything else was software (your PC does that part better); so you were using electronic gain at 320, 640, etc

Logan Detty's picture

https://youtu.be/bE5ugJEfDdU

Please watch my finding on this while taking RAW compression into account

Spy Black's picture

For me I merely use it as a reference for computing a good exposure, it's not really important that it may not be correct. It would be nice if we did stick to the ISO standard of density measurement at rated ISO however.

I think he kinda misses the point about manufacturers stating such high ISOs, and that is that sensor X is capable of being pushed further and still have better high ISO IQ than sensor Y.

Paul Gosselin's picture

There are a bunch of misconceptions here.

First: No, shooting at low ISO and increasing the gain in post is not equivalent to shooting at a higher ISO. See example with ISO 100 and ISO 25600 (on a Nikon D5500). To understand this, we need to understand that the noise in camera can be split in three types:
1- The shot noise: This is a physical quantum noise due to the randomness of the phenomenon converting light into electric charges. It is unavoidable in our cameras. Its amplitude is dictated by how much light is accumulated. The more light we capture, the lower the noise (relatively to the incoming light).
2- The noise from the circuit: Every electronic component produces additional noise for various reasons. Amplifying the signal from the photodiodes prior to the analog-to-digital conversion (ADC) allows to reduce the relative impact of the noise from the ADC stage.
3- The quantization noise: Because we convert the signal in numbers with a finite resolution, the lowest values cannot be very accurate. Take JPEG with 8 bits per channel: for each channel the value can go from 0 to 255, but the lowest value jumps from 0 to 1, to 2, to 3… there is no 2.34.
Increasing the ISO would not change anything with regard to the shot noise (1). However, depending on the implementation on the camera, it can reduce the importance of the circuit noise (2). Also, it reduces the quantization noise (3).
Now, you must know that cameras start having low circuit noise (2), and with RAW (16 bits) the quantization noise (3) can show only for very very low values in the recorded file. So, sure, shooting at ISO 100 or ISO 800 in modern cameras would not change anything. But in very low light situation, ISO is still something real.

Why ISO is not the same for every camera? Well, because of how it is defined for digital camera. For JPEG, there is an objective norm: “this much light must give an image this grey”. But this makes no sense for RAW files, as they describe the data from the image capture rather than an output image (while the norm apply to an output image, including the software treatment that is not fixed with RAW files.). So for RAW files, this is a matter of the manufacturer's opinion on recommended exposure.

So does ISO mean nothing? And it is also true that the manufacturers can say whatever they want about the max ISO?
Well yes… and no. Offering an ultra-high ISO setting that would only give you noise is a risky move. This kind of unjust advertising can really hurt the reputation of the brand.

Can we implement arbitrary low ISO? Not so easily. If there is to much light, there are to much electronic charges that accumulate during the capture process and the photodiode saturates. So basically, as explained by Northrup, you must actually take several pictures and end up averaging them. As has been explained by Northrup as well as Mads Peter Iversen, you can actually do it yourself if what you want is a long exposure: take many pictures, average it. And true, having this possibility inside the camera would be nice and would not be very hard to get. (On the other hand, doing it off camera allows you to realign the shots and circumvent the need for a stable tripod.)
Whether or not you merge the pictures on or off camera, your biggest limitation will be the buffer size, then the writing speed of your memory card. Depending on what you are photographing, the pause between each picture will probably not be an issue… but it might sometimes.
But is it an arbitrary low ISO? No, not really. Why? Because if you cannot reduce you shutter speed further and there is still to much light to shoot at f/1.4, you still cannot shoot at f/1.4.

This! I registered only to upvote and comment on your summary! There are lots of debates over this, and some people have SOME knowledge about sensors, and signal processing, and stretching the terms to validate their OPINION, which is based on vague conceptions. This needs to be cleared up, because it is very misleading. Anyone can try this at home comparing 5stop shifts, and see that these claims are not true, and the difference varies extremely between cameras and sensors. People need to understand, that between the sensor readout and the raw data storage, there is more than a single wire. Some manufacturers use intricate analog signal processing circuits, some use more digital post processing. The balance between those decides 'how true' are Tony's claims, thus how "iso invariant" is a sensor. Thanks for summing this up! Spread the truth!

Rashad Hurani's picture

a PhD in radio communication engineering here. We used to say radio com is all about noise; these debates are not debatable in comms since early 20th century, now it is reaching you. I heard a LOT of confusing, and flat wrong arguments, about the noise issue in digital cameras that make me laugh especially because those who dive in it are not experts, yet dare to give highly specialised technical judgment.

Bottom line: Paul's is the best and most correct technical argumentation you can find on the subject, as far as I read and heard

More comments