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!
true, it's classical clickbait video. but on the other hand, it's sad, that you "have to" make bs content, just to be popular. and ever worse, it's not about what's right, it's about making clicks and making selfies these days.
the other week he throttled his USB card reader to 15 from 60 and called it like that happens fluidly in the wild. Now he's saying ISO is just another exposure bracketing method to correct in post before heading straight for the drill shot. Ill put it this way I've heard way more whacky conspiracies than this one amigo.
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.
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.
Michael Menzel that being said, I was able to push it 3 stops and then add another exposure instance and push that 1 stop so it should be the same result as 4 stops. -P
@@FStoppers Haha, it was the same for me. But after I invested some (more) hours I saw the value of it. The function are more "raw" than in the other programs, You have more power and for the stuff I do (astrophotography etc.) this is really helpful. But it makes it really hard to learn.
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.
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.
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?
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
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.
Not all of us, just an overly vocal minority. It's like if you worked at a prison, eventually you start to think everyone is a criminal. Most the troll types seem to have nothing better to do with their time. The non-trolls are out taking photos and not really caring what anyone has to say about it. Just enjoy what you do and what you have to do it with.
Well, there would be way more differences because that also dependa on the camera's Dynamic range( Higher d.r. equals to less clipping on highlights( in the case of being overexposed) )therefore the files wouldnt be the same at all, except if all cameras have simillar dynamic range.
No, it doesn't work with overexposed images. If the whites are blown out, there is no way to get back the detail by reducing the exposure. Think of the brightness of the sensor readout as being a number ranging from 0 to 1. If an area is too bright, it's clipped at 1. On the other hand, if something is too dark, it's never truly 0; rather, it becomes a very small number like 0.01. Because there is no clipping at the bottom, the information is not lost. Suppose you have 2 pixels at brightness 0.7 and 0.8. If you double the ISO, which amounts to doubling the numbers, they both become 1, i.e., indistinguishable from each other.
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.
Or, during the conversion from analog to digital, some information is necessarily lost (compression). Is one photosite 0.33333334 or 0.33333335? Digital must decide, whereas analog is boosted with every single photon read and is closer to 0.33333333217452. If boosting happens before digitization, the full uncompressed signal is boosted. If the data is boosted after compression, the signal that was thrown away during digitization cannot be boosted. Information is lost by boosting after digitization.
That's exactly what's happening - and Nikon are very good at it. They just set the firmware to do the work - all can and do this to differing degrees. Hence "iso" is not necessarily fake but is manufactured and can be different - nota physical entity like shutter speed and aperture.
The take-away for me is don't be afraid to underexpose aggressively to preserve highlights. I usually underexpose 1/3 of a stop, but I may change that to 1 stop. Also, anybody who is adamant in their opition should test out their own gear and software and draw your own conclusions that work for you. I like the series, it taught me something and made me think.
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" :)
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.
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
So I did the test and in the end, as much as I'm surprised to admit, there's definitely a difference between ISO400 pushed to ISO6400 and ISO6400. I would say that the difference is absolutely minimal, basically the only thing I see different is similar to the black spots in the video, except that for me, it's white spots. Honestly, I had to zoom in to 100% to see *any* difference at all, but at that level of zoom, you can barely tell what you're looking at. I'd post some center section crops, but don't know how to link them on youtube so they don't get marked as spam.
jhford - How would shooting at a lower shutter speed underexpose more than at a higher one? Maybe I see this backwards to you, but in my mind high = faster lower = slower. A slower shutter allows more light, thus increases exposure. Maybe I am being pedantic, just curious.
Kevin Diaz not sure what you mean? I metered for iso 6400 then turned the camera into manual mode with those settings, so that the underexposure at iso 400 was solely due to the iso setting. I used the same shutter, aperture and focus for all shots
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.
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.
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
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. 😛
Isso mostra claramente quão poderosa é a Nikon D850. Mas estes testes, se feitos com qualquer câmera Canon, o resultado final será muito diferente com ruídos estrondosos.
In the past, when I started shooting RAW, when the picture came out underexposed I thought it did not matter because I finally fixed it in Post. but then I realized, even now, that the best thing will always be to expose correctly, not only for the time saving, but because in the end the quality is usually better. I don’t know if with ISO invariance is the same, but, a part of being a good photographer is to know how to expose well. If I’m in a portrait session, normally I will show to the client some shots, and I don’t want to show underexposed images and have to tell them “oh, don’t worry, I will fix it in post”
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".
Thanks for taking the time to do this. It can't be a surprise that software and in camera processors all handle files a little differently, but some people need it to be shown to them. :)
I think that both you and Tony are basically right. When we change the aperture or shutter speed on a camera we are making physical changes to the amount of light hitting the camera's sensor. We can't easily duplicate that in software. In that respect, a digital camera works in the same way as a film camera. But when we change the ISO value in a DSLR we are not making a physical change to the amount of light hitting the sensor or the sensitivity of the sensor. Because the camera can't change anything physical when we change ISO, all it can do is to use in-camera software to give more or less gain to the image. Put simply, digital cameras simulate a film camera's ISO because they can't do anything else. When asked to change ISO, all that digital cameras can do is to boost or reduce gain. The big question then is whether to do that in the camera or in post. But either way we are doing the same thing - using software to change the image. That doesn't mean that we will get exactly the same results in camera as we would from a program like lightroom. All programs and all cameras are slightly different and will give different results. But there is no logical reason that I can see why an in-camera software processing shouldn't be similar to off-camera software processing. Of course we should try to get exposure right (or nearly right) in camera. We should be aware of the noise problems of using high ISO values. The exposure triangle is still the exposure triangle. All that good stuff still applies. But it is relatively unimportant whether we do software simulation of ISO in the camera or in a separate off-camera program.
Correct me if I'm wrong, but in film cameras, when setting ISO, there was no change to exposure, but only the mathematical relation between aperture and speed to achieve proper exposure. The photographer still needed to set these according to his taste.
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.
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. ;-)
So glad to hear this. I've often found myself worrying about having my blacks too black and often tend to over expose. Now I know to err on the side of underexposure. Thanks guys!
I did my own experiment with a D3300 and the underexposed image had a lot of red in it when I brought the exposure up. Lighting was the same for both images, taken seconds apart, with only the ISO changed between the two.
This is an interesting conversation we're having on this topic. Always better to get it right in camera, but it's nice to know that if you are shooting by hand, not stabilized, you can probably shoot underexposed a few stops faster with your shutter speed instead of raising the ISO and still do a fairly good job of pushing the exposure in post. You can remove noise a lot more easily than motion blur. Of course it all depends on how your camera handles amplifying the signal off the sensor. Again-- another reason why you need to KNOW what your gear can do.
What surprises me in all of this is no mention of using ISO as part of the exposure triangle, for creative control of your images. If you just lock ISO at native, you reduce your aperture and shutter speed options at the time of shooting.
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.
The people who claim a camera is ISO invariant, have not tried to push a raw file that has been pushed previously. For example, tone map an ISO 100 image that is pushed 5 stops to match one at ISO 3200, and watch how little latitude remains in ISO100 image. The reason why you are getting incomplete results is that you are not testing the latitude of the pushed image VS the high ISO one.
Thank you for this. I had already watched the Tony version, and this completes the picture. I agree with you that proper exposure is the best plan, and that sometimes entails compromises. Nobody in his or her right mind would deliberately underexpose a shot just to waste time in post.
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.
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.
But if the camera is making the ISO decision after taking the photo as Tony claims, why can we not recover details from the highlights when they blow out.? I am talking about if you had your ISO set to 1600 by mistake for some reason.
Dave McKeegan did a video on Tony Northrup's video on this that you need to see if you haven't already. He talks about the signal coming from the sensor to the AD converter before it is boosted when using non native ISO. If you boost it afterwards using native ISO you get a different result. I don't know enough about camera technology to know how it all works but his explanation sounds good. I do think there is a difference in the processing between cameras and editing software. Maybe you should do one more test and test between a d850 and an older camera that has an older processor. Then you'll know it has something to do with the processor in the cameras.
I think after this follow up video we can pretty much wrap up all the ISO invariance talk and move on now. The closing comments summed things up well enough.
I think that this is a great test. This opens up for a more creative way of photographing. Not being afraid of underexposing with the ISO may be a creative alternativ in some cases.
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 UA-cam 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.
While I've caught you in a comment section, thanks for shooting The Deathbed Vigil. Of all the documentaries I've seen, none is quite as personal, genuine, and interesting.
@@JodyBruchon Hey, thanks! If I'd had a little forsight, I would have actually learned how to shoot a film before jumping into that. It started out as basically just a home movie for the gang at Commodore, but I was convinced by a few friends that it had a wider audience, despite the technical weakness. But that, too, is a good lesson for me even today: the art and message are more important than the tools!
@@DaveHaynie Absolutely. I think one of the biggest tragedies in computing is not only the demise of the Commodore of the 1980s, but also that everyone knows Apple and IBM and Microsoft while Commodore is slowly fading from the minds of many. There was a magic about that company that I don't think can ever be repeated. They caught lightning in a bottle. I have to tell you, man, it really broke my heart to see that extraordinarily polite man you worked with say the things he said about Ali. There must have been so much stress in the last few years. The accounts of what was started and scrapped midway through read like some...I don't know, bizarro bureaucracy black hole fiction, maybe. You guys must have been silently screaming in your offices watching it unfold.
@@JodyBruchon Yeah, it was super frustrating. And at that point in time, I couldn't see too far past the tech issues. So I launched six new projects that last year, in the vain hope that I could technology our way out of a problem that wasn't fundamentally a technology problem. Sure, as they spent less money, it started to become a tech problem, but it wasn't to start with. I do wonder, in our new post-digital-boom camera market, if we have enough room for everyone in such a shrinking market. When does the market start to be a tech problem, and who's it going to affect first? Sony's put themselves in a strong position, but if you don't have the right in-house tech, and the right alternative, more-profitable businesses, this could get ugly for some.
I actually learned a lot from this whole ISO thing. It's not something I had ever really thought about before, and watching these videos I think has given me a more solid understanding of how ISO works and how it relates to other things like shutter speed, aperture, sensor size, etc. I even ran my own test (with similar results to what you found) which got me more experience using fully manual settings... something which I don't normally do. Is this stuff super important? Probably not... but I do think having a better understanding of how cameras and image processing works makes me a more competent photographer. So I don't think this is a waste of time - even if the results ended up just confirming your expectations. Next I would be curious to see whether it's better to be underexposed by 4 stops or overexposed by 4 stops. E.g. in a high contrast photo is it better to get your subject properly exposed and try to recover the highlights in post? Or is it better to "protect the highlights" by under exposing a little and bring up the exposure later in post. We've seen that modern cameras give us a ton of flexibility when correcting under exposed images... is the same true of over exposed images? If not, then the takeaway might be that it's better to err a little bit on the side of underexposure.
thanks for your work. I have a comment about those black pixels on your shots. You can see them in every editor you use. They are dead pixels of the sensor. I'm actually surprised how many your camera has. You can use remap sensor feature of your camera to remove them.
What if you shot a bright object (light bulb/reflections) that would max out the sensor to the point of clipping and did this with ISO100 being 'properly exposed' and ISO3200 being totally blown out? Would meeting in the middle end up looking the same? I think you need to make another video.
It looks like Tony was right. Dont worry about the ISO setting too much, since you can recover in post if you messed up. He is also right that it shouldnt be called ISO, because it isnt standardized. Even in the Sony A6000 to A6500 there is a noticeable difference between the claimed ISO and the measured ISO(per DXOmark)
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.
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
I saw Tony’s video on digital ISO‘s being non-standard. He specifically called out the Fujifilm XT3 saying that it’s ISO 100 was really ISO 40. Since I have a Fujifilm X-T3, I tested it against the handheld light meter I used to use with my film cameras in the studio. This light meter was a Seikonic L308. My results showed that the Fujifilm XT3 ISO measurements were spot-on. I also tested my X-Pro3 and it was dead-on too. As far as overexposing or under exposing a digital file, it’s no big secret that modern digital sensors have way more exposure latitude than traditional film. So I don’t see why recovering information from an under exposed raw file is such a big deal.
That's definitely close enough for me. I was shooting an amateur Muay Thai event in a dark bar one time and if I had known I could do this to capture the action better, I definitely would've traded some noise for more crisp action shots.
I've salvaged many photos like this over the years. From weddings to landscapes to moon shots. Mainly in the Nikon lineup as far back as the D7000 to the newer D7500
Nikon and other manufacturers program their own noise processor in camera. So sensor to sensor you going to get different noise characteristics based on the in camera post processing. This is why Sony and Nikon can use the same sensors but when you have a team of developers that can program in processing software to reduce noise you'll get different results with the same sensors but in different cameras. Nikon used to mush the D300 when you shot at higher ISO whether you shot with NR on or Off. Off was a lot more detailed but still mushy. Ppl would just shoot at 1600 the boost in post. Most newer models as we continue to progress are going in ISO less. Fuji is by far the best hands down. Nikon and Sony trail and Canon even today is still at the back of the pack in terms of digital noise, they all have variables and when you push pass 6 to 7 stops (which is a pointless practice) you'll see banding, major color noise and blown pixels.
Results look good :) Note if you want to preserve the highlights more you should be boosting the ISO not just reducing it and underexposing the rest the exposure. Increasing ISO is shifting the gray point to give more space for highlights (hence why cameras often default higher ISO than base for flash to give you more room to help prevent blown highlights in your RAW images).
Properly use Expose to the Right and that will give you most shadow recovery but its hard to not verify you aren't clipping the highlights in camera when the previews/zebras/histograms are based off the JPG results instead of raw sensor data. Otherwise you are stuck getting a camera with more dynamic range. Overall just use ISO when it makes sense.
I can now imagine some street photographers shooting 5 stops underexposed so that when people ask them to delete the photo, the photographer could simply claim that the camera is broken and it could only take black images!
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.
Why not use ISO-AUTO basically all of the time? Set Aperture and Shutter speed as required by the situation and then let the camera come up with an ISO that is reasonably close... And adjust exposure in post if needed?
Only "love" to this video! I think that the entire excitement is about Toni's video title "ISO is a total fake." No it is not, regardless if a specific camera has a sensor appearing close to ISO invariant or not. Entirely different aspect altogether.
I wouldn't say that "none of this matters at all" speaking as a product photographer I need my photos to be as clean as possible before I send them off to my graphic artist to get the highest quality photos for when we publish in magazines. I would love a larger comparison across multiple camera brands, maybe add in some mixed lighting and do more complex scenes to really put it all to the test
But would heavily underexposing retains more details in highlight tho? If so then will shooting at 100 ISO always be better in high dynamic range scene?
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.
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!
And damaged his credibility still further. Way to go Tony - social media climber and disseminator of false information.
Tony "I want to thank our sponsor Squarespace" Northup
true, it's classical clickbait video. but on the other hand, it's sad, that you "have to" make bs content, just to be popular. and ever worse, it's not about what's right, it's about making clicks and making selfies these days.
And in the end like always he was 95% right...
@@uncoy This video backs up what Tony was saying, Images look "almost" identical.
Can I recover images shot through a lens cap with my Leica?
Yeah, there's a procedure for that. Remove the cap, go back to where you shot the image, shoot the image again. Fixed! :D
Yeah, those $95 lens caps allow for that!
Wait !! What?
@@abhijitmeti1611 Only half-snark here... you can't actually shoot through a Leica lenscap. Some of them do actually cost $95.
@@DaveHaynieyeah bro I known but what's that special in them
FStoppers - the number one drill photography channel on youtube.
I walk around shooting black images because that's what pros do :)
In reality it's because you left the lens cap on. :P
Awesome!!!! LOL!!!
@@BookmansBlues 😂🤣😂
Stan KMT - and the shoot in P mode too, since that stands for pro!
@@BookmansBlues Yeah but Sony has such badass DR that you can still recover the images even with the lens cap on.
2016: Fujifilm ISO is fake
2019: ISO is fake
2019 a couple days later: ISO is still fake
But really this is a drill advertisement
FStoppers always selling out :)
Imiy I thought it would be a Tide advertisement
LOL!!! I was thinking the same thing, WHAT'S WITH THE FEKKIN DRILL!? 😂🤣😂 You gotta admit, it's one sexy drill tho :P
the other week he throttled his USB card reader to 15 from 60 and called it like that happens fluidly in the wild. Now he's saying ISO is just another exposure bracketing method to correct in post before heading straight for the drill shot.
Ill put it this way I've heard way more whacky conspiracies than this one amigo.
have you learned anything about photography? it's clearly an impact driver!!
Conclusion: the D850 is awesome.
I have a D810. It’s pretty good too.
Pentax k1 kills it easily😏
Hail to everybody who remember ASA
Tore Hansen Salute! 🖖
Just don't forget to pronounce it 'A - Suh".
Hear, Hear!
Seems like a long time ago. Oh boy I am getting old... I even remember ilford ID11 for those ASA films...
Tony has certainly raised a DIN in the photo nerd community. 😉
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.
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.
And without Lightroom to save them ,
I've used Canon, Nikon, Sony, Panasonic, Olympus, Fuji and they are all great.
t Lim so true did Ansel Adams worry about this you don't get famous chasing technical stuff its all about the creator
@@jamesridley184 He spent hours in the darkroom to get it right
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.
I dunno man, that's a lot of noise. I don't think it holds up.
@@cmkimbrell67 The noise was basically the same between ISO 3200 and ISO 100 (boosted in post).
"We did all of this work but none of this really matters." Welcome to every hobby I've ever had.
U nailed it buddy
I feel your pain.:-)
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.
Michael Menzel I have to say, using Darktable for 20 mins made me want to completely remove it from this video. That program....okay I can’t even! -P
Michael Menzel that being said, I was able to push it 3 stops and then add another exposure instance and push that 1 stop so it should be the same result as 4 stops. -P
@@FStoppers I kind of wish you'd looked at RawTherapee.
@@FStoppers I feel the same way about photoshop.
@@FStoppers Haha, it was the same for me. But after I invested some (more) hours I saw the value of it. The function are more "raw" than in the other programs, You have more power and for the stuff I do (astrophotography etc.) this is really helpful. But it makes it really hard to learn.
Your DeWalt is too noisy. I'm gonna get a Milwaukee.
Rubbish. You didn't try MS Paint.
:-) the best comment today!
😂🤣😂
Its upgraded to Paint 3D
^winner
LoL
I like how you guys always drill down to the essence of a topic, yet keep it from being boring.
oh the pun
😂🤣😂
If you use a large bit do you get a bit more noise than using a small bit?
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.
All this proves is that some cameras have carry more data making their ISO less specific.
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.
This becomes a best raw post process comparison video too. Thx.
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?
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
As a professional audio chick, I am more concerned about the poor speaker placement on his near-field monitors.
what would be the correct place?
@ does the audio change depending on which side the speaker is on the table??
What's your IG? Im also an Audio Engineer. Let's network!
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.
As soon as I read that title from your last vid, I knew you would have a nerd uprising haha
People with cameras are hate and trolls these days. :(
Not all of us, just an overly vocal minority. It's like if you worked at a prison, eventually you start to think everyone is a criminal. Most the troll types seem to have nothing better to do with their time. The non-trolls are out taking photos and not really caring what anyone has to say about it. Just enjoy what you do and what you have to do it with.
Great work. I agree with your conclusions.
Try the other way around, with over exposing. Would be interesting to see if the results are similar 👌🏼
Well, there would be way more differences because that also dependa on the camera's Dynamic range( Higher d.r. equals to less clipping on highlights( in the case of being overexposed) )therefore the files wouldnt be the same at all, except if all cameras have simillar dynamic range.
No, it doesn't work with overexposed images. If the whites are blown out, there is no way to get back the detail by reducing the exposure. Think of the brightness of the sensor readout as being a number ranging from 0 to 1. If an area is too bright, it's clipped at 1. On the other hand, if something is too dark, it's never truly 0; rather, it becomes a very small number like 0.01. Because there is no clipping at the bottom, the information is not lost. Suppose you have 2 pixels at brightness 0.7 and 0.8. If you double the ISO, which amounts to doubling the numbers, they both become 1, i.e., indistinguishable from each other.
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.
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!
I'll save him the trouble...you blow out the highlights.
Once the ISO gain is applied in camera I dont think you can recover information from there so you lose dynamic range
this is what I commented on their last video, you can't claim a camera is ISO invariant if the reverse isn't also true!
I would like to see an 8 or even a 10 stop underexposure and while you are at it... a 8 or even a 10 stop over exposure. you can do eeeet!
Nothing better than a correct Exposure.!!! Forget post-processing.!!
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 ?
Tony tFuntek that def seems to be happening! -P
Most cameras don't do it, that I know only Sony atm unfortunately does it on long exposure
Or, during the conversion from analog to digital, some information is necessarily lost (compression). Is one photosite 0.33333334 or 0.33333335? Digital must decide, whereas analog is boosted with every single photon read and is closer to 0.33333333217452. If boosting happens before digitization, the full uncompressed signal is boosted. If the data is boosted after compression, the signal that was thrown away during digitization cannot be boosted. Information is lost by boosting after digitization.
That's exactly what's happening - and Nikon are very good at it. They just set the firmware to do the work - all can and do this to differing degrees. Hence "iso" is not necessarily fake but is manufactured and can be different - nota physical entity like shutter speed and aperture.
The take-away for me is don't be afraid to underexpose aggressively to preserve highlights. I usually underexpose 1/3 of a stop, but I may change that to 1 stop. Also, anybody who is adamant in their opition should test out their own gear and software and draw your own conclusions that work for you. I like the series, it taught me something and made me think.
Underexposing 1/3 or 1 stop is FAR from aggressive LOL.
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" :)
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.
Thanks for taking the time ! good job guys !!
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
Well he's using a D850 not the A7iii so its ISO 400 for the dual gain and I was hoping he was going to do that comparison of 400 vs 6400.
@@Cryect2Well, I have a D850 and i guess I could just do it myself
So I did the test and in the end, as much as I'm surprised to admit, there's definitely a difference between ISO400 pushed to ISO6400 and ISO6400. I would say that the difference is absolutely minimal, basically the only thing I see different is similar to the black spots in the video, except that for me, it's white spots. Honestly, I had to zoom in to 100% to see *any* difference at all, but at that level of zoom, you can barely tell what you're looking at.
I'd post some center section crops, but don't know how to link them on youtube so they don't get marked as spam.
jhford - How would shooting at a lower shutter speed underexpose more than at a higher one? Maybe I see this backwards to you, but in my mind high = faster lower = slower. A slower shutter allows more light, thus increases exposure. Maybe I am being pedantic, just curious.
Kevin Diaz not sure what you mean? I metered for iso 6400 then turned the camera into manual mode with those settings, so that the underexposure at iso 400 was solely due to the iso setting. I used the same shutter, aperture and focus for all shots
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.
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.
Hi there, thanks for the learning. Do you have a sense for how Raw Therapee would perform in comparison?
Why don't yall get a camera engineer to explain what's what?
"None of this really matters at all. Get back to takin' pictures"
Nuff said
I don’t care about what people say about you Lee, keep your OCD alive bring more videos of this nature, I love them!
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
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. 😛
Isso mostra claramente quão poderosa é a Nikon D850. Mas estes testes, se feitos com qualquer câmera Canon, o resultado final será muito diferente com ruídos estrondosos.
Gene Waddle he also should test all the different white balances with every Nikon lens, and all the different color drills. :)
And no full frame vs. APS-C comparison. WEAK! ;-)
@@happythec1am I don't know why he didn't think of that.
AND, no sheep. what the heck ????
In the past, when I started shooting RAW, when the picture came out underexposed I thought it did not matter because I finally fixed it in Post. but then I realized, even now, that the best thing will always be to expose correctly, not only for the time saving, but because in the end the quality is usually better. I don’t know if with ISO invariance is the same, but, a part of being a good photographer is to know how to expose well. If I’m in a portrait session, normally I will show to the client some shots, and I don’t want to show underexposed images and have to tell them “oh, don’t worry, I will fix it in post”
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".
And Roberto has the correct answer. At last someone paying attention.
Thanks for taking the time to do this. It can't be a surprise that software and in camera processors all handle files a little differently, but some people need it to be shown to them. :)
I think that both you and Tony are basically right.
When we change the aperture or shutter speed on a camera we are making physical changes to the amount of light hitting the camera's sensor. We can't easily duplicate that in software. In that respect, a digital camera works in the same way as a film camera.
But when we change the ISO value in a DSLR we are not making a physical change to the amount of light hitting the sensor or the sensitivity of the sensor. Because the camera can't change anything physical when we change ISO, all it can do is to use in-camera software to give more or less gain to the image. Put simply, digital cameras simulate a film camera's ISO because they can't do anything else. When asked to change ISO, all that digital cameras can do is to boost or reduce gain. The big question then is whether to do that in the camera or in post. But either way we are doing the same thing - using software to change the image.
That doesn't mean that we will get exactly the same results in camera as we would from a program like lightroom. All programs and all cameras are slightly different and will give different results. But there is no logical reason that I can see why an in-camera software processing shouldn't be similar to off-camera software processing.
Of course we should try to get exposure right (or nearly right) in camera. We should be aware of the noise problems of using high ISO values. The exposure triangle is still the exposure triangle. All that good stuff still applies. But it is relatively unimportant whether we do software simulation of ISO in the camera or in a separate off-camera program.
Correct me if I'm wrong, but in film cameras, when setting ISO, there was no change to exposure, but only the mathematical relation between aperture and speed to achieve proper exposure. The photographer still needed to set these according to his taste.
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.
You, Tony and others i consider the best youtubers in the photography art and youtube hype... just keep it going.
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.
;-)
So glad to hear this. I've often found myself worrying about having my blacks too black and often tend to over expose.
Now I know to err on the side of underexposure.
Thanks guys!
I did my own experiment with a D3300 and the underexposed image had a lot of red in it when I brought the exposure up. Lighting was the same for both images, taken seconds apart, with only the ISO changed between the two.
This is an interesting conversation we're having on this topic.
Always better to get it right in camera, but it's nice to know that if you are shooting by hand, not stabilized, you can probably shoot underexposed a few stops faster with your shutter speed instead of raising the ISO and still do a fairly good job of pushing the exposure in post. You can remove noise a lot more easily than motion blur.
Of course it all depends on how your camera handles amplifying the signal off the sensor. Again-- another reason why you need to KNOW what your gear can do.
you are 100% right. Shoot the image correctly but if you get it wrong don't toss the image out, you can probably recover it.
What surprises me in all of this is no mention of using ISO as part of the exposure triangle, for creative control of your images. If you just lock ISO at native, you reduce your aperture and shutter speed options at the time of shooting.
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.
The people who claim a camera is ISO invariant, have not tried to push a raw file that has been pushed previously. For example, tone map an ISO 100 image that is pushed 5 stops to match one at ISO 3200, and watch how little latitude remains in ISO100 image. The reason why you are getting incomplete results is that you are not testing the latitude of the pushed image VS the high ISO one.
Thank you for this. I had already watched the Tony version, and this completes the picture. I agree with you that proper exposure is the best plan, and that sometimes entails compromises. Nobody in his or her right mind would deliberately underexpose a shot just to waste time in post.
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.
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 wrong just because I can say whatever I want.
More Elia PTW BTS, please!
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.
But if the camera is making the ISO decision after taking the photo as Tony claims, why can we not recover details from the highlights when they blow out.? I am talking about if you had your ISO set to 1600 by mistake for some reason.
Dave McKeegan did a video on Tony Northrup's video on this that you need to see if you haven't already. He talks about the signal coming from the sensor to the AD converter before it is boosted when using non native ISO. If you boost it afterwards using native ISO you get a different result. I don't know enough about camera technology to know how it all works but his explanation sounds good. I do think there is a difference in the processing between cameras and editing software. Maybe you should do one more test and test between a d850 and an older camera that has an older processor. Then you'll know it has something to do with the processor in the cameras.
Are you sure you turned off the noise reduction in camera?
I think this may be the reason why the high ISOs noise have difference
I think after this follow up video we can pretty much wrap up all the ISO invariance talk and move on now. The closing comments summed things up well enough.
Cool ISO comparison video. As someone who sometimes shoots band photography at ISO4000+ this was really interesting and useful
I think that this is a great test. This opens up for a more creative way of photographing. Not being afraid of underexposing with the ISO may be a creative alternativ in some cases.
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 UA-cam 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.
While I've caught you in a comment section, thanks for shooting The Deathbed Vigil. Of all the documentaries I've seen, none is quite as personal, genuine, and interesting.
@@JodyBruchon Hey, thanks! If I'd had a little forsight, I would have actually learned how to shoot a film before jumping into that. It started out as basically just a home movie for the gang at Commodore, but I was convinced by a few friends that it had a wider audience, despite the technical weakness. But that, too, is a good lesson for me even today: the art and message are more important than the tools!
@@DaveHaynie Absolutely. I think one of the biggest tragedies in computing is not only the demise of the Commodore of the 1980s, but also that everyone knows Apple and IBM and Microsoft while Commodore is slowly fading from the minds of many. There was a magic about that company that I don't think can ever be repeated. They caught lightning in a bottle. I have to tell you, man, it really broke my heart to see that extraordinarily polite man you worked with say the things he said about Ali. There must have been so much stress in the last few years. The accounts of what was started and scrapped midway through read like some...I don't know, bizarro bureaucracy black hole fiction, maybe. You guys must have been silently screaming in your offices watching it unfold.
@@JodyBruchon Yeah, it was super frustrating. And at that point in time, I couldn't see too far past the tech issues. So I launched six new projects that last year, in the vain hope that I could technology our way out of a problem that wasn't fundamentally a technology problem. Sure, as they spent less money, it started to become a tech problem, but it wasn't to start with.
I do wonder, in our new post-digital-boom camera market, if we have enough room for everyone in such a shrinking market. When does the market start to be a tech problem, and who's it going to affect first? Sony's put themselves in a strong position, but if you don't have the right in-house tech, and the right alternative, more-profitable businesses, this could get ugly for some.
I actually learned a lot from this whole ISO thing. It's not something I had ever really thought about before, and watching these videos I think has given me a more solid understanding of how ISO works and how it relates to other things like shutter speed, aperture, sensor size, etc. I even ran my own test (with similar results to what you found) which got me more experience using fully manual settings... something which I don't normally do. Is this stuff super important? Probably not... but I do think having a better understanding of how cameras and image processing works makes me a more competent photographer. So I don't think this is a waste of time - even if the results ended up just confirming your expectations.
Next I would be curious to see whether it's better to be underexposed by 4 stops or overexposed by 4 stops. E.g. in a high contrast photo is it better to get your subject properly exposed and try to recover the highlights in post? Or is it better to "protect the highlights" by under exposing a little and bring up the exposure later in post. We've seen that modern cameras give us a ton of flexibility when correcting under exposed images... is the same true of over exposed images? If not, then the takeaway might be that it's better to err a little bit on the side of underexposure.
thanks for your work.
I have a comment about those black pixels on your shots. You can see them in every editor you use. They are dead pixels of the sensor. I'm actually surprised how many your camera has.
You can use remap sensor feature of your camera to remove them.
What if you shot a bright object (light bulb/reflections) that would max out the sensor to the point of clipping and did this with ISO100 being 'properly exposed' and ISO3200 being totally blown out? Would meeting in the middle end up looking the same? I think you need to make another video.
OK, no surprises, but I’m still glad you did this test. I learned a number of things from it.
It looks like Tony was right. Dont worry about the ISO setting too much, since you can recover in post if you messed up. He is also right that it shouldnt be called ISO, because it isnt standardized. Even in the Sony A6000 to A6500 there is a noticeable difference between the claimed ISO and the measured ISO(per DXOmark)
This topic is legitimately big enough now to where we need an engineer that works for one of the big camera developers to explain what's happening.
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.
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
Good advice at the end there! Amen to that!
I saw Tony’s video on digital ISO‘s being non-standard. He specifically called out the Fujifilm XT3 saying that it’s ISO 100 was really ISO 40. Since I have a Fujifilm X-T3, I tested it against the handheld light meter I used to use with my film cameras in the studio. This light meter was a Seikonic L308. My results showed that the Fujifilm XT3 ISO measurements were spot-on. I also tested my X-Pro3 and it was dead-on too. As far as overexposing or under exposing a digital file, it’s no big secret that modern digital sensors have way more exposure latitude than traditional film. So I don’t see why recovering information from an under exposed raw file is such a big deal.
That's definitely close enough for me. I was shooting an amateur Muay Thai event in a dark bar one time and if I had known I could do this to capture the action better, I definitely would've traded some noise for more crisp action shots.
Personally I'd rather get the exposure correct when I take the image, but if I'm off, it's nice to know that it's a lot more forgiving in post.
Why did you compare PhotoShop with the other softwares at 5:26? Was it accidentally?
At 5:57 you compared the rest of them to the Nikon Capture.
Love it when non-tech peeps are trying to understand how things work
I've salvaged many photos like this over the years. From weddings to landscapes to moon shots. Mainly in the Nikon lineup as far back as the D7000 to the newer D7500
Nikon and other manufacturers program their own noise processor in camera. So sensor to sensor you going to get different noise characteristics based on the in camera post processing. This is why Sony and Nikon can use the same sensors but when you have a team of developers that can program in processing software to reduce noise you'll get different results with the same sensors but in different cameras.
Nikon used to mush the D300 when you shot at higher ISO whether you shot with NR on or Off. Off was a lot more detailed but still mushy. Ppl would just shoot at 1600 the boost in post.
Most newer models as we continue to progress are going in ISO less. Fuji is by far the best hands down. Nikon and Sony trail and Canon even today is still at the back of the pack in terms of digital noise, they all have variables and when you push pass 6 to 7 stops (which is a pointless practice) you'll see banding, major color noise and blown pixels.
Did you try the "Enhance photo" in Camera RAW? Would be interesting to see.
Results look good :) Note if you want to preserve the highlights more you should be boosting the ISO not just reducing it and underexposing the rest the exposure. Increasing ISO is shifting the gray point to give more space for highlights (hence why cameras often default higher ISO than base for flash to give you more room to help prevent blown highlights in your RAW images).
What if I want to preserve highlights without adding extra noise off the bat so I have the most leeway to play with shadow boosting later?
Properly use Expose to the Right and that will give you most shadow recovery but its hard to not verify you aren't clipping the highlights in camera when the previews/zebras/histograms are based off the JPG results instead of raw sensor data. Otherwise you are stuck getting a camera with more dynamic range. Overall just use ISO when it makes sense.
2 videos in a row missing Tony's point about ISO invariance after a certain threshold. Nice.
I can now imagine some street photographers shooting 5 stops underexposed so that when people ask them to delete the photo, the photographer could simply claim that the camera is broken and it could only take black images!
Thumbs up for telling people to stop pixel peeping and jsut get back to taking pictures.
There is always a analoge boost of the voltages so their will ‘always’ be a slight different between under exposed shots and normal shots
Great video man! I appreciate your attention to detail and telling it the way it is!
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.
Why not use ISO-AUTO basically all of the time? Set Aperture and Shutter speed as required by the situation and then let the camera come up with an ISO that is reasonably close... And adjust exposure in post if needed?
Only "love" to this video! I think that the entire excitement is about Toni's video title "ISO is a total fake." No it is not, regardless if a specific camera has a sensor appearing close to ISO invariant or not. Entirely different aspect altogether.
I wouldn't say that "none of this matters at all" speaking as a product photographer I need my photos to be as clean as possible before I send them off to my graphic artist to get the highest quality photos for when we publish in magazines. I would love a larger comparison across multiple camera brands, maybe add in some mixed lighting and do more complex scenes to really put it all to the test
But would heavily underexposing retains more details in highlight tho? If so then will shooting at 100 ISO always be better in high dynamic range scene?
Those black dots happen because it's detail that isn't captured by the sensor, thus increasing the exposure will not bring anything back.
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.