We wasted even more time Testing ISO Invariance 

Can I recover images shot through a lens cap with my Leica?

Rubbish. You didn't try MS Paint.

Everyone just don't get it. Tony managed to get EVERY camera blogger/Vloger to say his name and channel at least 5 times... You can't buy better advertising....Great Job Tony!

FStoppers - the number one drill photography channel on youtube.

But really this is a drill advertisement

2016: Fujifilm ISO is fake 2019: ISO is fake 2019 a couple days later: ISO is still fake

I walk around shooting black images because that's what pros do :)

Hail to everybody who remember ASA

Your DeWalt is too noisy. I'm gonna get a Milwaukee.

Conclusion: the D850 is awesome.

"We did all of this work but none of this really matters." Welcome to every hobby I've ever had.

Try the other way around, with over exposing. Would be interesting to see if the results are similar 👌🏼

All these images are basically identical. All tiny little details, are only profile diffences (except from the black dots) could be made to look identical, and in no way (in a blind test) one could find which is better than other. Basically Tony is correct -at least for modern sensor cameras- 10 years ago, with older canons these differences would be absolutely devastating not tiny hair noise signatures.

But what happens if you overexpose with high ISO by 4 stops and then drop the exposure back down in post?! We need more of your time wasted!

As a professional audio chick, I am more concerned about the poor speaker placement on his near-field monitors.

As soon as I read that title from your last vid, I knew you would have a nerd uprising haha

Why don't yall get a camera engineer to explain what's what?

People with cameras are hate and trolls these days. :(

Oh, but you didn't test every camera there is and every program there is. What good is that? I'm also sure you used the wrong subject and the wrong color background. 😛

I'm starting to think that the only difference between image quality in camera brand (AKA: Canon VS Nikon) is just 100% Software, and how good the maths are good in the code that apply enhencement to pics we take in-camera

If the complaint is that "if you boost an image 5 stops it isn't 100% exactly the same, only 99.3%" the basic thesis stands.

You can set any boost in Darktable, the slider only goes up to 3, but when you left click on it you can enter any value.

You're so right. All this doesn't really matter, just enjoy taking photographs! Unfortunately, the huge advances in sensor technology and software has created a "class warfare" amongst the nerds who do not take photographs but simply focus all their energy and hate on any camera brand that they do not own or like. Truly sad.

i did the same test with my A7 III and if i take a ISO 100 and push it to 1600 - its not as good as a 1600 shot. BUT: if i take an ISO 800 Shot and push it to 12800, its indistinguishable. Thats because of the 2nd gain circuit of the A7 III which kicks in at ISO 640. So i took an ISO 50 Shot (which is a processed ISO 100) pushed it 3 stops to get ISO 400 and it looks identical too. So for that Matter the A7 III at least is "dual invariant".

This becomes a best raw post process comparison video too. Thx.

Switching between pictures without putting a complete different picture between them is useless for youtube, because there are using A and B Frames only A Frames are true... When you switch direct to a similar picture there a information from both pictures in the second one specially in the beginning

So Tony was correct. It really shows how poorly Adobe handles raw files. This has also been my experience with the raw files from a 5DSR compared to the Canon software's processing. It's good to know it isn't just a problem with Canon raw processing by Adobe.

Great work. I agree with your conclusions.

Sorry to be so serious, but for the final comparison, you need to do the PS tweaking on the Correctly exposed file as well as the boosted one, in order to compare them. I bet the correctly exposed one can be made better. But great series! I'm thinking two stops under might be a sweet spot, depending how your camera handled highlights. I think I'll try it a bit.

"None of this really matters at all. Get back to takin' pictures" Nuff said

I like how you guys always drill down to the essence of a topic, yet keep it from being boring.

I'd have pushed darktable and luminar an extra stop or fraction thereof to make them appear equally bright... before comparing their shadow noise and equally bright final result as measured by histogram.... Ideally I'd have matched the histogram peak the the same brightness in all raw programs and possibly matched their contrast settings to make the histograms equally spread before comparing shadow noise, as some may default boost shadows more than others.

My recommendation: You should do a test with the sony a7s ii and compare iso 100 to iso 6400 in a LOW LIGHT SCENARIO. Take a look at www.cloudynights.com/topic/546003-read-noise-measurements-for-the-sony-a7sii you can see that the readout noise is quite different as a function of iso. Also, you need to make sure that the readout noise is actually the dominant noise which only happens in low light! . If you have plenty of light, you would be shot noise dominated and the shot noise distribution would be the same. My main gripe is that you guys are looking at data without understanding the physics and you are picking experiments which highlight only a certain behaviour of the noise. This whole 'noise invariance' claim to fame really falls apart when you are looking at low light photography.

Yup ur wrong. My 2nd point 1st. ISO is a film sensitivity rating, the sensitivity of the sensor does not change it is being digitally replicated in the camera software. As such I'm sure the camera is using every tool available in the internal camera processing (color correction, noise reduction, etc) to improve the image quality whilst digitally increasing ISO. So why are you only using one tool setting in post processing. Now to my 1st point the fact that using 4 stops produces vastly different results proves Tonys point that ISO is not a standard and is arbitrarily being set by manufacturers and apperently not only by the camera makers.

As an alternative for all the truly sad nerds who thinks photography is about noise, ISO invariant, etc etc, here's an anti-dote. It's really what photography is: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-bbUmjkhjjZ0.html Enjoy

Let me check for you. Sharingan! ISO doesn't matter.

It really depends on nothing. We are in a era where post processing is a must. So my advice to be able to grow your instagram account and be seen as a good photographer by instagram users, go out, take ugly photos and ugly portraits that have no substance. Then post processing fix the exposure in a mediocre way, fade the photo, add some grain, make the skin and all the colors yellow or green (or the opposite, make then blue), post the before and after photos on instagram. Add a cringy description. You're welcome

RED cinema cameras have advertised for a while that ISO is arbitrary and set post raw capture. I think the differences that are coming out will bring to light the quality of sensors, noise patterns, & software algorithms that separate the lower from higher quality equipment.

It remind me of a saying of Chirac the old french president : "you are really trying to sodomize a bug that didn't ask a thing", instead of asking a sensor scientist that could give an in depth response, and invite him reguraly so we all learn for real, Tony and other youtubers (not Fstoppers) act like they know better, if TN was honest he should title a his video, here my gift as a 1 Million Subc Photography YT channel, another fake drama to shake the community for views(before it was sensor size and the gathering of light and before that it bokeh and it's relation to sensor size ....), the good thing is that it could promote smaller channels, as long as they have something intelligent to say about it, apart from that, all this is a non sense ....

Who cares, if you don't know the camera you use, take a hike.

Nothing better than a correct Exposure.!!! Forget post-processing.!!

I own a D850 and tried this with ISO400 and ISO6400, those two ISO pairs are using the same analogue gain. I can confirm that at these two ISOs there's still a difference, though it's extremely slight and only perceptible when zoomed in 100%. For me it shows as white spots instead of black spots as in the video, but it's definitely there. My guess is that this is down to the difference in the algorithm that Nikon uses to calculate a whole-frame gain calculation vs. how Adobe does it. This is with the margin of error for that I suspect, so I'd still call it ISO-Invariant personally... I know I will underexpose and push if I can't get a high enough shutter speed. Original Comment: Wasn’t the main issue that you aren’t comparing the correct ISOs values because of analog gain past ISO 640? Try doing it with 640vs6400 and ISO64 vs 500

Is it possible that when you set your camera to a higher ISO that maybe the firmware in the camera adds a little noise reduction ?

The part that I think most people should be asking.... is that if those black spots and artifacts are not noticeable by the naked eye and only noticed when you zoom in to the point where you start seeing the pixels.........does it matter that they are there? Nobody is looking at pictures in that way. The consumer sure as hell isn't. So Tony does bring up a very relevant point concerning this ISO topic. The comparisons look virtually identical when zoomed out to default. We deal with something similar in music production when comparing analog vs digital. Analog is analog, so when doing a pitch bend on a synth, it's going to be smooth unbelievably. But in the digital world, that pitch bend is like little small steps. Looks like stairs. And once you start producing music at higher sample rates, those steps (despite being there) are virtually indistinguishable compared to the analog synth. The only people who care about it are producers, analog gear producers, digital producers. But the consumers aren't noticing any of this crap. The same with audiophile people. The stuff they talk about regular consumers don't even give a crap about. So are we trying to impress other photographers, other producers, other people's eyes and ears in our field, or the consumer?

You’re wrong just because I can say whatever I want. More Elia PTW BTS, please!

Has jealous ken been by yet to try and take credit? lol

This is not scientific at all. You only took two samples. You should take 1000 pictures with each setting and then compare the averages.

It would be interesting to see a comparison of the post from the actual lightroom (2019) and a 5/10 years old lightroom. I think the exposure boosting will give us big differences between those two. The thing here is not the small variations after the exposure boost in post, but the really good job that software companies does with each new release.

Have you tried overexposing by 4 stops?

Canon does do it !! I tested... I got same results as you. Actually my canon at 100 with a totally black pic came out very clean in post when I boosted the exposure.

your video makes them all the same when watched on a 320x200 EGA monitor...

Hi there, thanks for the learning. Do you have a sense for how Raw Therapee would perform in comparison?

why would you want to make more work in post? I want to get it right when I take the photo so I don't have to do as much work in post.

Hahaha wow! It wasn't a complete war of time, you came to a conclusion. Gear work!

What! You want us to go out take pictures !?! But I'm photographer I discuss nerd stuff for a living, I don't want to take pictures! 😜

I thought you did a pretty amazing job to be honest....hats off ! Very interesting video 👍

Good advice at the end there! Amen to that!

interesting conversation... never had these issues back in the good old film days ;o)

It makes sense. The boosting the image for high ISO is done before the conversion to RAW, so even with a RAW file are not as good off , as just turning up the ISO. Fine debunking of this "myth" :)

Thanks! It's realy interesting. Nothing as a bit of testing ... I think I'll do the same with my Canon. Good job.

Tony Northrop is a genius! LOL all this glorious watch time loops created by both the supporters and haters making reaction vids 😂... Meanwhile I'm just over here living life and learning lot's from all.

How much did DeWALT pay you?

You actually got Luminar 3 to work without crashing

That photo would never make it to Shutterstock...so what's the point? To much noise = denied.

Thanks for continuing to say "I.S.O." It's the pronunciation that has become accepted in public consciousness by the majority of consumers and retailers.🤣🤣🤣🤣🤣

You're wrong...Haha kidding...It's amazing how these files look similar or almost the same...Although expected and the most important part is when you go home to have sharp pictures and not blurred ones...I don't mind high iso noise as i have the picture and i can build on this...I think is enough with these type of videos,bring the old FStoppers back with nice content please

Thanks. I did a real simple test shooting the same image correctly exposed, while changing the iso (D810) from 100 to 400 to 800. The camera makes a cleaner image at 100.

You're obviously just getting paid by that drill company... All review RU-vidrs are sell-outs these days... SMH !!! (i'm just joking). I agree though, that none of this really matters.

The Nikon D850 has a bump in sensor gain at ISO 400. It is “iso invariant” between iso 64 and iso 400 then invariant between 400 and up but not if you do a test that crosses iso 400.... so just repeat all that testing by taking photos at say iso 800 vs iso 2400 vs boosted in post and then we can talk :-)

Most of these tests miss a critical element: the difference between image DR and potential DR. Most FF cameras are going to bring in 14-bits worth of level per pixel, thanks to the ADC -- it's not related to the sensor's actual DR, but the window into that DR that you're sampling, assuming you're not ISO invarient. Boost it by four stops and you now only have a 10-bit image. And for the purposes of RU-vid and JPEGs, unless you actually shot that full 14-bits worth of DR in your original raw and then compressed it enough to see the effect of that in a JPEG, MPEG, or 8-bit monitor, you won't see a meaningful difference between 14-bits and 10-bits.

ISO Invariance is not new - others have been posting this over 3 years ago. They explained it better than Tony but guess Tony should still get the credit for drawing mass attention due to his popular channel.

Thanks for taking the time ! good job guys !!

So it is 2022, and I still think that Tony was 'mostly' correct about ISO. a) it is completely pointless these days what ISO 100 means (and camera reviews should actually investigate what ISO 100 means on a new body instead of just boosting about the ISO range when that means nothing). b) Dual gain is not well understood and that it is sometimes better to shoot at higher ISO to get less noise (and also not understood by Tony in his first video). c) It really does make shockingly little difference and being very underexposed due to choosing a too low ISO probably doesn't mean you have to retake the picture.

The Nikon D7200 in Adobe Raw +5 EV is very similar to in camera ISO 3200. See for yourself www.dpreview.com/reviews/nikon-d7200/13 DPreview concluded in 2015 " D7200 is probably the closest to being a truly ISO Invariant camera that we've seen. This means that you could reduce ISO (and the hardware amplification that goes with it) by 5EV in order to retain 5EV of highlight detail, then adjust the brightness afterwards with no additional noise cost."

For those who still want to argue here are the ISO specifications. Please qoute the revelemeant ISO standard. www.iso.org/obp/ui/#iso:std:iso:18383:ed-1:v1:en www.iso.org/committee/48420/x/catalogue/p/0/u/1/w/0/d/0

Man you guys REALLY don't know your stuff and this comes across as amateurish. I cringed when you said the reason to use Capture NX was to get Nikon images looking the same as in camera processing. Seriously, have you guys ever actually tried to do that? Easy test - shoot RAW + JPG then compare in camera JPG to Capture NX JPG produced from the RAW file. Make sure you use the same picture control and settings. There's a good argument for under-exposing rather than ETTR (Expose to the right) which was popular a few years ago. You can't recover a blown highlight. Correct exposure will always win though.

As I already wrote under another youTuber - basicaly you got it right - every camera just has ONE (1) base iso. The only difference you see after pushing 3, 4 or even 5 stops is due to ON-SENSOR denoise, by the way - Canon have been the very first ones to use that, which of course is more powerful if you choose to use iso1600 compared to iso100. After seeing the difference between those images, 1600 vs 100, that is just caused by the lower use of one-chip denoise (even if shooting "RAW") as using iso100, being lower, and iso1600, being higher. ;-)

I'd like you to compare 12-bit RAW vs 14-bit RAW. This is the thing should influence much on the image quality in case of 4-5 stops underexposure. I did this on my own several years ago with my Nikon D7000 and I was impressed how RAW file can be handy in case of underexposure (and even overexposure) cases.

Can we please stop talking about the Northrup's channel? Their videos are super click-bait, they totally have an attitude of our opinion is better than yours, and Tony likes to pretend he comes up with calculations for "true exposure" or "true DOF" and all that stuff, dude isn't a mathematician or a physicist so why would I care. Channels like theirs and Jared Polin regressing the dialog around modern photography to stupid "We found out which 70-200 is the best!!@?!@!##$@!$!##!"

Isn't the D850 ISO invariant ONLY IF you are in the same analog gain range? I'm seriously asking because i don't really know that camera... If so you can't compare ISO100 pictures and ISO1600 since there are at different analog gain steps. He has done this in both videos, not sure if intentionally to push the results towards his point... I think anyway Fstoppers are really taking Tony's words too literally and seriously: Even in a ISO invariance i would expect some small differences (came on, nothing is perfect, it's just a concept... ) The fact that you can push the exposure up 5 or 4 stops from ISO400 and get almost the same results than with much higher ISOs is already impressive and good enough to define it as "Invariant" isn't it?

What you do not understand is ISO noise can be different from base of 100 to say 400 (Sony Mark 2 and 3 cameras) but no difference between 800 to 51200 but above is bad. The point of ISO Invariance is say you are shooting a city at night to reduce highlights being blown out you can use say the 800 get a dark image but increase exposure in post to bring out better shadow detail with same noise as if shot at 6400 with highlights blown out. Now for the goodie, if you are used to capturing sunsets/rises in HDR (3 @ +/- 2ev) you no longer need the tripod or do an HDR. You can increase SS getting a dark image but in post just increase exposure to brighten and bring out shadows. I can shoot at ISO 50 (yes 50) adjust zebras away (to keep highlights in check) and adjust 0 to -3ev dail and good SS (for lens mm) and brighten in post and play with sliders to get as good if not better single image while just walking about. Also think of indoor shots without a flash same difference dark image looks better without a flash in post. Just saying you have not played with real world capture and all PP'ing programs will do a great job even with pixel peeping.

OMG I'm splitting a gut reading the comments below. ASA...ISO...OMG...WT_!!!

Why do photographers waste so much time over inane issues like this. Just learn your craft - HONE your craft - have actual skills instead of relying on gear, software, and shortcuts....how can you expect to be respected if you don't bring real skills to the table? Learn how to expose a photo correctly...and everything else goes out the window....easy. The ONLY photographers who should care about all this ISO talk is astrophotographers. And that is because they need to know- is it better to shoot my image at 10 sec @6400iso or 20 sec @3200iso....which is cleaner in post? No one else should care.

Here's what happens when you underexpose a RAW image by 5 stops: the top 5 bits of the sensor's dynamic range record nothing but black. So if your RAW files are captured in 14-bit color depth, you will effectively get 9-bit RAW images at 5 stops underexposure. When you boost the exposure by 5 stops in your photo editor, you're effectively shifting all the image data 5 bits higher, filling in the vacant top 5 bits of the DR and blacking out the bottom 5 bits. The image is now brighter, but still limited to 9 bits of dynamic range, so the boosted near-black shaddow noise looks coarser than the noise in an unboosted, properly exposed image.

I have Sigma SD Quattro. Usually have to shoot with iso100-400, because the noise is really bad in iso800 and foveon sensor is real reason for it. Transforming x3f files to dng with wrapper, in lightroom edit have to add more 2-3 stops of light to picture for getting ”right results”.😅 Maybe the hardest camera i have ever used, but not impossible to shoot. Only if you know what are you doing.

You should have used ACDSee, On1 Raw or Topaz Studio. Anyway ISO is a setting specific to a camera model. It is primarily used to set jpeg output standard for colour and brightness. Not for RAW though raw manipulation may be part of the in camera balancing act. The whole Idea of ISO as described by Northrop is flawed as it is intended to disrupt. Rather like his flawed total light concept.

Only tested this on my a7III from ISO 100 to ISO 25600 but saw no difference when properly exposed in Lightroom from 1600 to 25600. I did see a difference in everything from 100 to 800, although it is supposed to be ISO invariant from and starting at 800. I shot at night and natural light, so there is room for error at that 800 ISO, but if it is the same in an uncontrolled environment it will be the same in a studio. The difference here is really in dynamic range, and that will be affected if you shoot at higher iso in camera but won’t be (won’t be or at least won’t be as much) if you push it in post.

How about this for an experiment? Shoot an image at iso 100 and another at 1600. Match the exposures and then see how each image reacts to some progressively hard editing.

While a clarification from your previous video is appreciated, however I am disappointed once again. If you stated your objectives and assumptions as described here AT THE BEGINNING of your previous test, that would have been fair. Instead you sounded clueless there and now, after doing some background research, instead of owning your sloppy video, you are defining who is calling you out a nerd. "I do not think anything I said was incorrect", that is right, simply you just had no idea of what you were talking about. Not cool buddy

Dp review test all there cameras for ISO invariance, if you dig through the actualy quite complex technical background Bill Claff’s Photonstophotos website, it’s all out there, Tony was right but didn’t emphasise the caveats and exceptions, but frankly it want the point of the video.

The mythbuster series conclude the amount of the noise and grains you may get by underexposing the photo... OK, fine... but one of the reasons we crank up the ISO is to be able to hand hold the camera and shoot at the F-stop of the interest.... interesting though... I tried with my D810 and I got very similar results...

As you pointed out, you can make a few adjustments with the noise removal sliders and get it to look lust like the ISO1600 file, but then you can also make those same adjustments to the ISO1600 file to make it look even better still. The bottom line is...there's little difference between an under-exposed image (taken at a low ISO) that's pushed in post and a correctly exposed image (taken with a higher ISO) but there is still a difference and we all want the best image possible, so correctly expose in camera or, perhaps, allow it to be under-exposed by just one stop to ensure you don't blow out the highlights, then recover the shadows with only a one stop push.

Not sure in summary your conclusion is still use ISO to meet metering desired requirements... I think makes sense but still photos will be post-process even for RAW to JPEG formatting...today digital world has improve a lot...and post becomes a tool to fix undesirable outcomes from the camera... and I think is not a crime to tweak then... obviously we want and specially I want to get out from the camera the perfect picture and leave the post as the most basic corrections as maybe vignetting, barrel distortion...maybe reduce highlights, many of these tweaks can be done in the camera body which are nothing less than post processing on the camera body...

What most photographers don't understand (because camera controls mislead them) is that RAW dynamic range STARTS at the point where the sensor saturates. All sensors deliver maximum output at this point, and as light intensity is decreased, the sensor's output decreases linearly (proportionally). As light intensity decreases into the shadow region, the sensor's output descends into the range where electronic noise becomes noticeable. The point where the sensor's output drops below the average noise level is (by definition) the dynamic range limit of the camera. The reason the camera's ISO setting has an imperceptible effect on its DR limit is because ISO boosts both the sensor output AND its inherent noise level by the same amount. Where the ISO setting does make a difference is in how the image is digitized by the camera. As the sensor output descends toward black, fewer and fewer bits are used to digitize its output level (each stop darker is digitized by half as many bits). At near-black levels of intensity, so few bits are used that color and saturation discrimination become very coarse. That is why you see an increase in chroma noise in the shadows when you boost exposure by several stops to compensate for an underexposed image. The cause is NOT that the sensor is contaminated with near-dark color noise, it's purely a digitization artifact caused by insufficient bit-depth. If you raise the ISO to properly expose the image, those near-black regions will be digitized with more bits, and that will improve color and saturation discrimination. The moral is that you should use higher ISO settings when you want to capture the highest quality of shadow detail. You should use lower ISO settings when you want to make sure that the brightest highlights will not be blown out.

All I want is enough sensitivity to light to get a photograph in situations that I can look at with my eyes. The numbers, the differences in cameras, don't mean anything to me save for giving me settings to try. If Canon interprets that differently in all of it cameras, I don't care if 400 ISO = 400 ISO. I don't care what you call it. I don't care if the "experts" agree. I care that the picture turns out.

Expose for H/L, software adjustments for shadows, the exact opposite to how photographers used film, where many photographers exposed for shadows and developed for H/L.

The only time when the time of applying digital gain will make a difference is that a camera sensor & analog to digital converter is capable of delivering more bits of bandwidth per pixel then the picture file format is able to store. Its like using audio compression the wrong way which leads to clipping and lost details. I think that no "consumer" camera will apply analog gain to emulate "ISO" settings - that would be way too expensive and unreliable... . What really makes me wonder is that the ISO name giving organization isn't taking any action against those camera manufacturer whose ISO values do not correspondent to what those values should be. If they don't maybe a consumers rights organization will file lawsuits of misleading marketing. If this happens ISO will renamed on digital cameras as what it really is: (digital) gain.

Ok, ok .. I like it .. Now, where's the link for the drill on Amazon? Lol .. :-) Seriously though, this was a well done video. Moral of the story, the *free* Nikon software is the best .. no surprise there. After all, who would know the specifics of the raw algorithm map better than the manufacturer? Did you see the word *free*?

I think your testing could be off as the country you moved to exposes at least 12 stops of beer by the pool and a ISO (Intentional Sober Operations) of happier daylight ;) Seriously though, in the old days I would push and pull film from its native ISO if I didn't have the right choice and then process it to the ISO I shot at, not the ISO of the film. This is not that different. It can be done, even Professionally but there are always limits to everything when it comes to light! The beauty is... the ability to do something necessary to get the shot otherwise missed completely.

2 videos in a row missing Tony's point about ISO invariance after a certain threshold. Nice.