ISO Invariance (ISO is Fake Follow-up) 

I SO confused ~

I’m actually really enjoying these technical videos. 👍🏿

You should make more videos like these, talking about the more technical stuff

Tony is trolling all his critics.

I follow Tony like he's the New York Yankees in these RU-vid photography controversies. They throw everything at him and he bats it away.

That was actually very interesting! After 7 years doing photography, I didn't know about this "ISO drop gap" ! Thank you !

Every time he said "ENHANCE!", I expected him to solve a murder mystery.

Never try to out-geek a committed geek. Thanks, Tony. Your turn, @Fstoppers!

Tony, I've read and viewed, including photo samples, a fair amount on ISO invariance. Yet your illustration is quite the most lucid exposition I've found.

Wow! Those signal to noise drops are REALLY interesting!

I was the same, when I was a kid, used to take everything apart to understand how it really works ;-)

Going for a higher iso does give you slightly less dynamic range. So it becomes a noise vs range battle. What do you guys prefer? Less noise, less dynamic range. Or more noise, but better dynamic range?

I love these videos, especially when people talk crap on other channels instead of digging into what was discussed. Fstoppers was probably the best/most fair. Well deserved victory lap! All kidding aside these videos are really what breaks your channel away from all the others. Would love to see more of these kinds of videos with one to one comparisons. And my second fav are the videos where you two are drinking and answering questions. Chelsea is a hoot!

“They did a really...good job of it.” Just say what you’re feeling Tony, I heard that microscopic pause 😂

I stopped following f stoppers when they were arguing whether a photo with the moon in front of clouds was real or fake.

Tony - Thank you so much for this. I have the “ISO Setting/Read Noise” numerical table downloaded for my Sony A7riii to my phone for quick reference (until I get it memorized). Really enjoyed this discussion. This will give me a great deal of flexibility when shooting in different light conditions. For instance, when the sun is going down and I’m at ISO 500 but my shutter speed is not quite enough to stop the action that I want stopped; now I can jump to 1600, have slightly better quality, and move up the shutter speed to get the stop motion image I want (assuming that I have the depth of field that I desire). Options are good, I used to think that ISO had a linear relationship, now this opens up so many more options for clarity & creativity!!! Noise for my Sony A7riii: 160 = 640 200 = 800 250 = 1000 500 > 1600

I don't think testing with an actual image is a good idea. Digital cameras are not as digital as you would think. First in the chain is the sensor. The sensor outputs an analog voltage for each and every pixel. Following the sensor is a variable gain amplifier, this is the device that controls the effective ISO. Then there is an analog to digital converter. The best measurement would be to choose a shutter speed. Cover the lens completely so that no light enters. Record a raw frame for each available ISO setting. Then, for comparison sake, sum the values of all the pixels for each of the frames. What you are looking at is the dark noise from the system. If you graphed this data, you would probably see the curve go very slowly at first, because many of the values coming from the sensor and amplified by the variable gain amplifier are below the sensitivity of the analog to digital converter. At some ISO point, the amplifier amplifies the noise enough that the A/D is able to sense it, then it should rise at a predictable linear rate. This would be the true way to determine the noise floor of the camera. Changing ISO does nothing to increase noise, it simply magnifies the noise created in the sensor, of course it also amplifies the image at the same rate. So all image to noise ratio is created in the sensor at the time of exposure regardless of the ISO setting.

Tony, ignore the haters, I love technical videos. And many more people love them. Keep them up :D

Photons to Photos proved that for the Fuji XT3 best iso to use is : From ISO 160 until 320 The noise increase slightly but after 320 the noise is worst than ISO 800. So it's good to avoid any numbers between iso 320 and 800. From 800 to 1250 iso noise is similar. So your usable iso's will be 160 until 320 or 800 until 1250 and over that the noise always increasing. I did test my camera and that was true,couldn't believe it

the sweet spot on my EOS 2000D is around 1630 (indicated) ISO, I usually shoot at 400 so might be time for me to start to really play with my shutter speed and aperture settings...

In Photoshop and Lightroom (as well as other programs) there are many ways to edit a picture, your argument is just you trying to convince folks you actually know what you're doing. Sad

So for someone (using Sony A7iii) that doesn't want to have to push ISO a huge amount in post (lots of extra time involved), the best way to work is that if I can't shoot at base ISO of 100, I should go straight to 640 rather than bothering with anything between the 200 and 500 range?

Tony, I love how unaffected by critics you seem. You don't seem to take things personally. It's really refreshing.

Good follow up. Would be good to really drive home how this relates to dynamic range.

ISO is not fake . It is a standard. Some cameras are fake yes. But not all. People must learn to read the light in candels and lux. And use a professional light reader like Sekonic.

This was really interesting. Guess I won't be shooting iso 800 anymore, it's either 100 or 1600 haha

Love this kind of technical video! Really learned something. A question here: Why is there a noise drop at certain ISO?

Keep doing technical, as most other otherwise knowledgeable YT channel providers often get the technical details wrong, while you are almost always correct. Thanks.

I've not done any of the tests but have started to employ ETTR to good effect. I shot some pics at 6400 ISO and the images had hardly any noise - not exactly what you are talking about, but my point here is to expose correctly - as far to the right as you can - and what happens is you reduce the S/N ratio and thus noise in the darks. Shooting on a Canon 6D. Comparing these results to shooting at 100 ISO but underexposed and getting bags of noise all over the image. Metre right and push to the right.

Maybe it is time to change our vocabulary. For instance, new light emitting diode (LED) light bulbs now lists Lumens and their Watts equivalent. Lumen is the actual measure of light (that is produced by the bulb) but there needs to be a translation to the Watts-equivalent that we are all used to for incandescent bulbs. So maybe we need to learn new terms like signal-to-noise ratio and signal gain, while also showing the equivalent in photographer-speak, like stops of noise and ISO (respectively). It reflects the actual physics of semiconductor sensors, while also bridging to the 100 year old nomenclature created in the chemistry-based film days. As we transition to using both sets of vocabularies, it will become more comfortable, and also make more sense. Thanks for these and other videos, Tony and Chelsea.

bias reviews. canon fan boy you did not speak of canon ,

Ok. so for my D850, I should either shoot below ISO 150, or above 400 based on photonstophotos. Great info.

In your last video, you did not just "hint at". You made a categorical statement that changing the ISO setting in my camera to increase the (apparent) exposure in my camera is exactly like moving the slider in Lightroom. There was no hedging. It's good that you're correcting things now, but don't pretend that you didn't say what you said.

Yeah more vids like this - it's great knowledge

Tony, you are on to an amazing topic that so many people totatly mis understand. Can we all stop saying "ISO" and just call it what is REALLY is...."gain". Also, the term "exposure" only applies to shutter speed and aperture. "ISO/Gain" happen AFTER exposure is COMPLETED. What are we exposing image sensors to? LIGHT! Gain is NOT light. If your photosite is 50% full, 1 million "ISO" will not change that photosite's 1/2 full status!

This is a super useful video. There are bunch of nerds who want to know about these things! Big thanks Tony!

Good video Tony. That website is very useful. It might be an idea to do a video on dynamic range tradeoff using the same website. I could imagine some people noticing the big drop at say, 800 iso for noise on a dual gain sensor, and deciding to use that as their "base iso" without factoring in dynamic range.

Tony, It was very nice. This is a video for the ones curious about the machine. Its not intende to impact ones photography. So, haters, please refrain and let us, who are interested in the inner workings of the machine enjoy it.

I really like your technical videos...and even if something you said wasn’t 100% correct 100% of the time, it’s still fun to talk about. Any chance you could find out how to determine a safe reciprocal rule for using electronic shutter?

Thanks for the follow-up video Tony! One question tho, does that works the say way when shooting video? Meaning using ISO 640 rather than 500 for lower noise??

I knew about certain ISO settings being cleaner, especially at full stop settings like 100, 400, but I didn’t realize that it was significantly different at lower ISOs for my camera. Also, this and the ISO is fake video really explains the methods you did in the “ISO doesn’t cause noise” video, which confused me and a ton of other people when you would raise the exposure to match the pictures in Lightroom. Watching these videos ahead of time would have made it make so much more sense. Glad to know that my camera is much noisier at 200 than 400, now! Thanks for these videos!

Great info Tony. Understanding read noise at different ISO's is super useful. Thank you!

HAHA Tony said I.S.O. @1:15... GOT 'EM! 😁

I love when you make these videos! I learn something and the comments that people leave you are comedic relief lol

Its a shame your microphone was locked down. I was hoping the video would end with you dropping the mic.

I did a test of the Canon 80D. I shot at 100 ISO and at 3200 ISO and then pushed exposure 5 stops for the very "under exposed" 100 ISO. There was no noise difference.

Actually very interesting, I’ll have more control over my astrophotography results now. Thanks!

If only the other guys were half as knowledgeable as you are and were half as questioning as you are... Thank you for pointing out again and again the truth that people are too butthurt to accept.

You should get Moose back on, have a "real professional" photographer clear this all up. 🤣🤣🤣

If you have such camera, you can shoot on much faster speeds in A mode for example. What I mean - If you compensate -3 stops in A and shoot moving objects you will receive 3 stops faster shutter and much better chance to have a picture without motion blur. This can be a difference, between unusable and usable photo.

This video was really helpful. But without this information, the last video could perhaps be a little misleading.

This is the kind of stuff Thom Hogan is testing in his books (on Nikon cameras), which can be pretty useful if you want to get a better understanding of how your specific camera works, and how to capture optimal data. Tony is doing a great job at explaining all of this. Honestly, I think Tony is at his best when he gets to explain to a general audience something that's very technical and geeky. A lot of times, I feel like I already have a very good understanding of what he talks about, but it's always good to see how he explains it, which can make the topic even more intuitive. Great job Tony!

Nice job Tony! You tackled the issue very well, and I’m now obsessed with geeking out over those iso charts

So if it is too dark to use base ISO. It is better to underexpose, so the histogram should be on the left. You don't risk to loose any information about the highlights and you know you can raise shadows without loosing anything too. Different thing if there is enough light to use base ISO. In that case it could be better to expose to the right, to have lower noise in the shadows.

Sweet. Now look forward to 10 videos with his name in the title and a pic of him in the thumbnail with half assed information.

Well, so much for the "photonstophotos" site. Obviously this guy likes Nikon. He neglected to test the Sony A99 ii camera completely. He failed entirely in testing the Fujifilm finepix S5 -- NOTED for having high Dynamic Range -- probably because he neglected to activate the appropriate Mode in the S5! Result: What Else doesn't this guy know? No HHHR for the OM Systems OM-1 -- which he even fails to name the company properly! But, if all you want to show is Nikon D850 material -- you have your site!

Wait how do you get lightroom to push more than 3 stops worth?

Lol. Totally useless information but I enjoyed every minute of it. Nice job on not showing hate towards your critics. I find it healthy to have a civil discussion on these things. Many folks are unable to do that.

Finally, RU-vid guessed right about what video I wanted to watch next. This is exactly the video that I was hoping to see after watching your old video about ISO being fake. Sometimes it's your old, "evergreen" "nerdy" content that i appreciate the most.

I've just done the suggested test with my Canon 80D, with the Canon 24-70 f/4 lens. An outdoor shot with blue sky, clouds, dark green trees and shadows. Min ISO is 100, max is 6400. Comparing the same aperture (f/16) and shutter speed (1/4000) with the histogram more or less centered at ISO 6400, I shot at ISOs going down by factors of 2 to 100. Then I took the ISO 100 shot and imported into Pshop (after converting to DNG as Pshop no longer reads 80D raw format!). First thing, it made a big difference if I increased the exposure in ACR (the Adobe dng to pshop converter) or just adjusted levels in Pshop, it is much cleaner using ACR even though it can only do 5 exposure "levels" (are those stops?) Using the Pshop Levels tool added more noise, chromatic aberration, and also spread the histogram introducing many big gaps. Comparing the ISO 100 up-exposed with ACR versus ISO 6400, they look very similar, no obvious difference in noise at 100% crop, slightly more noise in the underexposed ISO 100 at 200% crop. There is a big color shift however, the ISO 6400 sky is much less blue when imported directly via ACR from the dng file. I was able to correct that by lowering the temperature in ACR while importing the ISO 6400 dng, but based on the actual sky color per my eye, the ISO 100 underexposed was more accurate color when up-exposed in post at the default temperature setting and white balance. Photonstophotos is consistent with this, they say the 80D is fairly invariant, with noise pretty linear except the stops in between ISO 100 and 200; this seems a feature of multiple cameras so there must be some funny processing thing to get those in-between settings right at the bottom. I use NeatImage ReduceNoise Photoshop filter, it cleaned things up quite nicely though not like shooting at real ISO 100. Overall, it still makes sense to me to set ISO in camera, but underexposed images or areas can clearly be rescued a lot in post. Thanks for this, it's helpful to know the limits of the equipment. Will try will film perhaps, but there the aspects of development and scanning add lots more variables so perhaps not worth the experiment.

Yay! Tony -- as opposed to FStoppers -- knows how to actually *pronounce* ISO (EYE-so, not eye-ess-oh). Not dissing FStoppers. Love the site. Learn much from them. But on this occasion, oh honey, you don't take on the Tony Northrup. He dah bomb. (That, and anyone could understand he was dead-on right. ISO -- pronounced, EYE-so -- is arbitrary. And he is also my hero for saying, It's just for camera nerds. Whether it's useful to know this or not is really debatable. Fun to know, but... how much bearing does it have upon the way we set up photos, control light, frame the shot, make magic instead of snapshots? I don't know 'bout *you* but I'm one dem nerds!)

I couldn't verify the findings with my A7Riii. According to the chart, ISO 500 and ISO 1250 should be equal. Real test shots showed more noise at ISO 1250. Then I tried ISO 500 vs ISO 800. The chart shows significantly less noise at ISO 800. Again, real tests showed slightly more noise at ISO 800. Am I missing something here? btw I basically set the ISO on auto and adjusted the shutter speed to raise or lower the ISO. All pics were equally exposed. This to me matches what really happens when out shooting.

Hey Tony and Chelsea. I've watched this previous video about ISO and also all the other video's on RU-vid that followed up on your initial video about ISO being fake. Apart from that, I have a question I can't figure out for myself. It has to do with sensors. What I don't understand, is why a sensor, let's say micro 3/4th of 24 MP is relatively cheap and can be found in camera's under 400$ while a full-frame sensor of 24 MP costs over 2000$ generally speaking. In the 3/4th, all the pixels are put togheter much tighter and in production, it's more 'difficult' to create than having a sensor with less pixel density. My logic says it should be the other way around. Like with CPU's in computers, the smaller the architecture, the more expensive and the longer the process to create. Maybe you can address this for me, that'd be very helpful.

When i test my Fuji X-T1 with the following: Exposures from 800 iso to 6400 iso with the same shutterspeed, same DR and same WB. All photo's look a bit different in color, sometimes a bit darker but with iso 1000 the colors are fading very bad. At iso 800 colors gets a bit better. (i just went 3 stops btw) When i test photo's from iso 200 to 6400 and adapting the shutterspeed to get the same exposure,... all photo's looks excectly the same in color, sharpness and,.. well no big difference in noise from 200 to 800 and somewhat noice almost the same from 1600 to 6400. With both tests there was no noice reduction added in camera or with export. So what is the point of isoless sensors? i still don't get it :\

D810 does not have a drop-off like D850 and Z6 (between ISO 636 and 800). Good thing, I don't shoot between that ISO range with my Z6. For daylight, I use the lowest native ISO for landscape. For nightscapes, ISO 800 and up.

Well, I disagree. I am more often pleased with the camera getting the exposure right, even with higher ISO, than if I underexpose and try to stretch it in post processing software. Your mileage may differ, but while there is some merit to saving underexposed shots in post, I find it is better to expose right in camera, even with higher ISO. Funny that is how the camera manufacturer says to do it. . . . Just my take. JR

Ok so, let me clarify (& correct me if I'm wrong) ... For the Sony a7 III, if I have to correctly expose a photo by setting the ISO to 503, then I could just shoot at ISO 100 and raise the exposure in post. This will result in about the same amount of noise. Similarly, an image that needs proper exposure higher than ISO 640 can just be shot at ISO 640 and just be adjusted in post to get the same noise level. Is that right? That's why in minute 10:35, the noise is different because ISO 51,200 is only similar down to ISO 640 whereas ISO 100 can only be stretched out to ISO 503. That is also why these other youtube camera reviewers are getting different noise levels due to this "dual gain." The other reason being that their camera may not be truly ISO Invariant. Did I understand that correctly?

Didn't know about the noise graphs. Raises a question about auto ISO. I often shoot with manual exposure time and aperture and with auto ISO. (Actually a Sony a7rm3.) Does anyone know if it is smart enough to let the ISO go up to 630 (or whatever the base ISO is) as long as there is no resulting over-exposure? I have noted that it quite often sets the ISO to 640, but don't know if it avoids 300-500 if it can without over-exposing.

I never leave comments on RU-vid, but this video is so good that I had to tell you! These technical videos are my favorite, might be because I’m an engineer :D

Tony, if I understand this correctly, and my main concern is noise, I should avoid ISOs 200 to 399 and jump to ISO 400 on the Nikon D850. Why is the Canon 5D Mark III so jagged? What does graph for the Canon 5D Mark III says about Canon's ISO?

This was by far a much more informative and well researched video than the one about ISO being Fake. I think the problem here is that once again, Tony has cherry picked data and failed to really dig into the contradictions when they arose. The bit where you show that highlights being clipped will make the photos unusable even after adjusting for exposure in post, contradicts the point previously made about exposing for ISO 51,200. I suspect that if you repeated this test where you exposed for 51,200, but captured the secondary image at ISO 100, that you'd see a similar result. Clipped sensor data is not invariably recoverable. You began to touch on this when you dug into Photons vs Photos, but I think you failed to make that connection, but it is an important one. You've shown us this really compelling data in an elegant fashion, but didn't finish the experiment by rigorously interpreting it. What you've ultimately done is prove that ISO isn't Fake, in fact it very much matters when you run the risk of clipping out both highlights and shadows. But secondarily, perhaps there is an ISO range(s) where ISO doesn't matter as much, or where there are sweet spots on a per device basis. It's not easy to wrap up in a sexy package like "ISO is Fake!", but we also shouldn't be so brazen as to say ISO is Fake, thus can just be fixed in post so don't worry about it. It's fundamentally bad advice for any photographer of any skill level, especially beginners or budding amateurs.

So I didn't understood the whole thing. There is a logarithmic graph. It's displayed linearly, doesn't this mean the Snr grows exponentially? ISO iso-invariance achieved through a "straight" line? Does it mean that at iso 1600 and ISO 800 +1 gain the Snr is identical if the line is straight? While a dual gain graph makes sense, have you guys seen the 6d mki graph? It's a freaking Rollercoaster ride.

Most Canon's are not ISO invariant. To be clear dynamic range drops rapidly with ISO, so if DR is important, use the lowest ISO-invariant ISO settings and push exposure. Also for static scenes, take many shots, say 10-20 and then you can stack them and use median filter and eliminate noise, win-win, high DR and low noise.

Great video. Thank you. What's the deal with iso 640 on the a7iii? I thought you start at 100 and double the number, 100, 200, 400, 800, 1600 etc etc. So should I be using 640 instead of 400 or 800? A little confused

Why not repeat the same test overexposing instead underexposing? Thanks :-)

Damn, ....Lee almost had you but,...., "NORTHRUP SCHOOLED" AGAIN,......

It is not we are mad at the search for answers, it is your method of getting the results. You are using a 3rd party software to do what the camera mfg. does in camera or with their brand software. That is not a valid test, ever. Also, your first video was ISO doesn't matter because ISO is not a standard blah blah blah. Two things you still have not done to try to get truly valid results. Test your ISO the camera reads with a dedicated high end light meter, from ISO 100 to at least 6400 to see if the camera ISO and what should be used is accurate. 2nd, use Nikon software for Nikon cameras, etc. Use their Raw converter and see if it looks the same. They are the one that would be able to make the image the same as the proper shot to begin with. Photoshop, Capture 1, and Lightroom all use different file setting so you can't use that as a test procedure as it is already flawed. Try those 2 things and it will answer the questions. Using the wrong methodology in every video actually proves ZERO and that is what irritates us. I think ISO matters, if for no other reason then to give some sense of what the exposure should be based on prior experience. I know I need ISO 3200 with my F1.8 lens to get a proper exposure in lower light. If my camera is more dark than normal or more light than normal, I know the camera ISO is not accurate and the maker has fudged the numbers. If the light meter shows F1.8, 1/60 sec, and ISO 1600, and my image looks proper with those settings, well then I know my camera is exposing properly. In that case ISO actually does matter as it would match the light meter. That also holds true with camera tests of higher ISO. If I know I need ISO 3200-6400 for most of my shoots, and the camera meters correctly, and the ISO is accurate, then the High ISO test shots will let me see if there is an improvement in higher ISO when a new model is released.

Using the D850 as an example. If Nikon can do something to make noise fall-off the linear graph at ISO 400, is there any reason they cannot add more fall-off points at extra cost; say at 1600 as well. i.e. is it a £ tradeoff at the moment or is it physics? Sorry - don't have the right tekky words to ask properly :)

I have watched the videos on this topic and this is still not clear to me. I understand that the lowest native ISO will lead to the least amount of noise all other things being equal within the same camera. But it seems that any manufacturer could arbitrarily say that their lowest ISO was 64 or 100 or whatever. So if the exposure was exactly the same, and was shot with a Nikon at 64 ISO vs a Sony at 100 ISO - would there really be less noise? Has this ever been tested? Is there something special about the Nikons' sensor or optics that is different that the other manufactures? Please help.

Hahahahaha god I love these Tony follow up videos

Your iso invariance test starts out with an iso of 640. Starting with an iso of 100 would get different results. That camera is iso invariant above 640, which you probably know. Most cameras are not iso invariant. Nobody would intentionally drastically underexpose an image and then correct it later but you did get a lot of reaction. This is crazy, I'm going out to take some shots.

Why did you not try the other cameras from their drop off point? I own a t-t2 and happen to know that the drop related to dual gain lays around iso 800, so is you compared between 800 and 12800 or something the difference would have been minimal.

As far as I understood this video, not even the new Sony cameras are ISO invariant, because there is a good reason the base ISO does not start at ISO 640... So Tony's rant in the first video he did was misleading, but it seems that he did not want to mention this in this second video. Huh?? Tony, that's not how I knew you... Sad. If you stretched your point to far in the first video just acknowledge it, where's the problem. Everyone can get a little bit carried away by a good argument, but you should be able to acknowledge it.

Also there is a question always popping up in my mind, kind on this topic if you can help.. if I set my camera to record photos on a lower mp. 12 instead if 16 for example. Will my full image, no zoom, have less noise visible. What I am asking is, is it worth it. 6d and 6d2 appears to have exactly the same sensor and because 6d2 is more MP it has a bit more noise. Which if what I suggest is true means that if I use a 6d2 with lower MP I will have similar final results. Hope I do make sense, and find an answer :) cheers

What an interesting review and concept. I thought it was all about noise. (S/N ratios). It is about range. All the more reason that polar mounts and trackers are a must for astro-photography. Dont forget or subordinate DSLRs to Mirrorless cameras.

Hmm, not sure the bottom of this subject has been reached yet. How did we end up with this interpretation of ISO? Is this the best we can do with digital sensors to simulate film grain, or is this the byproduct of it being too hard technically to get a sensor to behave in a similar way, is this a technical compromise? I'm sitting here wondering whether an ISO invariant camera is honestly better or worse than the alternative, from wanting a more detailed pro/con comparison; the output may be the same, but urgh, the technical nerd in me wants to know, but the artist says, settle down you big OCD looney. I wonder how a really expensive camera like a $50k hassleblad deals with ISO.

If cameras apply variable amplification before the conversion of analog to digital, why are there these articles on ISO that say ISO is not causing amplification? Seems there's a lot of contradictory info. I just assumed cameras have some ability to adjust amplification of the signals coming from the sensors.

Can anyone please tell what's the 2nd optimum ISO for Nikon D750 for astro and nighttime long exposure photography?? Like ISO640 in case of Sony A7 iii ..

I don't quite know why this didn't start with a woops I'm sorry, as you basically explain here how your last video on exclusively this subject was horribly wrong... And the information was readily available (the normal do review camera reviews for example). Good to see a correction though.

Tony, you do this so often!!! You will make a claim, which frankly is often moot. This will then trigger all kinds of other people to refute....and then you will bore with us a counter point.....so chill and move on!!!!

Good job. You are learning. Astronomers have known this for years. Look up photon to electron exchange. The camera counts electrons, not photons. Yes, photon kick of the electrons that are counted. When you turn the ISO up the number of photons to electron ratio is read differently. So when you have a low photon to electron ratio you have the limit to how high the ISO should be set. One photon to one electron counted.

Are 12 bit cameras and 14 bit cameras directly comparable? My 90D appears to have more noise than my old 1000D within the common ISO range, according to Photons to Photos

I note over on sonyalpharumours they have purported specs for a new Sony sensor having a “PGA”. This is a programmable (variable) gain amplifier and likely explains these steps in noise. There is an equation used in my field called the Friss equation that is used to determine the total noise in a system of amplifiers. One of the axioms that comes from this is the noise of the first stage is most important but only if you have significant gain in this first stage. So assuming the on chip PGA is at least as quiet as the rest of circuitry then increasing the gain of this stage will decrease the overall system noise. I’m an astrophotographer with a Sony a7riii and I always use ISO 640 or sometimes a little bit higher.

So what you are saying. ISO does matter, the second image was more grainy. so iso 5100 it would just be pixulated. If iso is too high you can not recover highlights. finally. sales pitch for presets

If a Sony A7 III at ISO 636 has 1.558 Read Noise, thats saying you should not use ISO 503 having 4.199 Read Noise? Because of the drop there?

5D Mark IV I hear is pretty much ISO 400 and above. Have you tested it? The thing I didn't hear you address though is, when you adjust the ISO in camera it adjust it before you hit the analog digital converter and if you do it this way, it came out of the converter on whatever ISO you had the camera on, some people were saying that caused artifacts. I don't know, haven't tested it.

So here’s a question. Say I’m shooting a small event, and I want to stay in Aperture Priority so I can be fast paced. If I’m in auto iso, what should I set the min maximum at? If there’s a huge difference on the a7iii between just iso-500 vs 640, and I set the minimum to 640, I miss the opportunity to set the minimum to iso-100 for situations where there’s enough light. If I set the maximum iso too low, now I have to worry about the camera forcing the shutter speed too low.

Dummies are Dummies @TonyNorthrup so keep thinking and investigating.

I only understood ISO as an electrical boost. This is interesting. Also looking at the charts made me understand some things. Amazing how this isnt general knowledge. If you wouldnt know youd refrain going up further and perhaps give up on a shot while in truth it might be beneficial to go up.

I think this is a great topic. I’m also a computer professional and a bit of a geek... and I have been into a couple online spars with those who are rigid about ISO and it’s effects. The fear is about grainy photos and high iso... folks thinking ‘just lower the iso and it lowers the grain’. This shows that there’s a lot more to it than that and it depends on each digital camera.