Understanding Dual Gain ISO on the Nikon Z6 III and other cameras

Explanation of what Dual Gain ISO is, why it's important, how to find the dual gain ISO level for your camera (using the Nikon Z6 III as an example), strategies for taking advantage of it, how it affects N-Log, and what the future holds for possible improvements. Although this video uses the Nikon Z6 III as an example, the concept applies to all cameras with dual gain ISO sensors, including other cameras from Nikon, Canon, Sony, etc...
To help make this technical subject easier to understand I use analogies and simplifications to the underlying mechanisms involved in dual gain ISO. If you'd like to dive deeper and gain (ahem) an even greater understanding of how sensors implement dual gain ISO I suggest reading the Aptina Dynamic Response Whitepaper on the subject, which Bill Claff hosts on his PhotonsToPhotos site:
www.photonstophotos.net/Aptin...
0:00 What is Dual Gain ISO?
2:15 How to find Dual Gain ISO level
5:17 ISO doesn't cause noise
6:10 Dual ISO shooting strategy
7:19 N-Log and Dual Gain ISO
7:52 Future: Dual ISO readout

[@testcams]

One point I forgot to include in the exposure strategy section is how raw shooters don't ever really need to use the intermediate ISOs between ISO 100 and 800, or the ISOs after 800, since those ISOs provide only modest benefit over ISO 100 or 800 in terms of noise, and actually are a detriment to dynamic range since they amplify the image data at the time of shooting and limit your options to selectively boost the exposure in post while still retaining your highlights. In short, shoot at either ISO 100 or 800, even if that results in a photo that looks underexposed in the camera, since you can normalize the exposure in post and retain the maximum dynamic range. You'll want to measure this for your own shooting because while the improvement in noise shooting at the indicated nominal ISO like ISO 6400 vs "underexposing" at ISO 800 is modest, that improvement goes up as you go further into underexposed territory. For jpeg and video shooters, you'll generally want to continue using the intermediate ISOs because those formats have limited bit depths, which limits how much you can boost the exposure in post before running into posterization and clipping issues. Also, if you'd like to find the Dual Gain ISO of any camera without doing the work yourself you can use Bill Claff's site. Click on a camera on his PDR measurement page. The Dual Gain ISO will be the ISO where there's a noticeable jump in the measured PDR on the graph. Here's the link: www.photonstophotos.net/Charts/PDR.htm

[@xchump]

Please bear with my noob questions.....so are you saying that since I shoot Raw, I would basically never benefit from going over ISO 800? Even if the pictures are underexposed, I should be able to fix that in post, and it will yield better IQ results than going for a higher ISO setting at the time of shooting?

[@boblozano]

Will experiment with this. Will be interesting to see where /if focusing / composition degradation (from the apparently lower light that the photographer uses to make decisions) costs more than the improved headroom gained (pun intended) by avoiding the gap isos (base 1 > gap iso < base 2). That leads to another change that would probably require firmware, in which the actual raw is encoded only with one of the two base isos (or the dual-gain), but the image displayed for the photographer responds precisely to the `apparent` iso selected, either manually or auto.

[@zakiradi450]

This is not true. Plain stupid to be honest.

[@boblozano]

@@zakiradi450 ad hominem is not an argument. Evidence would be useful.

[@guate6]

What are remarkably interesting video. Does this mean that if this option comes to full frame cameras camera would it have to be a hardware change or can it be made by firmware?

[@branimirteodorovic2297]

Finally somebody on KZbin who actually understands how technology works

[@kuau714]

This video is a MUST see for every photographer

[@sharkcookie.YouTube]

The best explanation about this topic I have seen! That's so much.

[@wanderlust0120]

0% intro 0% outro 100% ISO Subscribed ❤

[@shadow479]

This is mind boggling, it breaks the logic of the traditional linear direct correlation of "higher iso = more noise". Great job, sir. Thank you for this test and revelation!

[@user-zr4km5ns8t]

One of the most significant factors I look at in determining whether I need to subscribe to a photography channel on KZbin is based on evaluating the content with an eye to the need to watch something multiple times in order to be sure I have a sold grasp of it. In other words, I'm asking two questions; 1) Is this really necessary information?, & 2) Do I need to go back over this & keep referencing it like a study guide in order to make sure I can retain it? The answer in this case is clearly yes to both questions. Kudos for this work.

[@solardungeon]

Wow!! this is the most informative video on Z6iii noise

[@guyyowell8547]

This is the best explanation of noise and gain as it relates to exposure I’ve ever heard. This is very helpful. It’s interesting that the gain for the Z8 (and probably the Z9) occurs at ISO 500. This is likely due to the higher pixel count, and the fact the pixels are more tightly packed together.

[@RejeRush]

Would love to see a video breakdown of the z6iii vs the z8 in dynamic range, iso performance and how each respective dual gain iso affects an image.

[@rbob4931]

Fascinating to say the least. Well done!!!

[@AmadeuszAndrzejewski]

You have a great talent for explaining technical things in a simple and understandable maner

[@boblozano]

Really superb engineering analysis and testing. Keep it up.

[@johnhernlund539]

Excellent video, great explanations. I will try this same test with my other cameras, to see how they turn out.

[@sharkcookie.YouTube]

I'm patiently waiting for you to post something about shooting video, especially when. It comes to shooting Nraw. I have avoided N-log because it's base ISO is 800 which requires me to use a ridiculously short shutter speed when I'm shooting outside.

[@chantaljacques7664]

Thank you for sharing this on FM. Great information!!!

[@sagardigitallab]

Superb ! 'The best video ' todate-- explaining ISO and Noise relation

[@raudelravelo1169]

Awesome video. Keep it up!

[@jakilahmoulien9070]

this video is gold

[@Hrishi1970]

Wow. Interesting video. Thanks!

[@JohnVijayRaj]

Awesome explanation thank you

[@grahamland7703]

Explains such a lot, had gravitated to shooting at iso 800 as my preferred setting for light challenging situations, very resilient to being pushed. Results at higher iso's tremendously mixed, if exposure slightly out.... But sadly iso 800 has it's limits when you have to shoot movement. Many thanks for lighting my darkness!

[@kjltube]

But if you set your evf to NOT show the actual exposure brightness (instead a brightness that allows you to compose), and shoot manual iso 800, and the appropriate manual shutter speed you need to freeze movement then you should be happy. Unless perhaps if your scene is super dark?

[@Ben_Stewart]

Variable ADU is a problem with newer CMOS sensors. In Astrophotography sometimes it is better to just expose with a higher gain setting rather than little to no gain.

[@dominiclester3232]

A great video, thank you! Will you be selling or sharing your software sometime soon, please?

[@robertLuxama]

That sounds really interesting this software you'v created!! Will it be in sale? Is it for video?

[@Sebas-sg1kj]

What an amazing video! Question: if i need fast shutter speed to freeze movement, should i go for iso 800 instead of using values from 200 to 650 for example?

[@testcams]

If you can shoot at ISO 800 without having to increase your shutter speed further to prevent overexposure then you definitely want to use ISO 800 whenever possible.

[@mihan2d]

I have a question, is the statement true that Z6 mk. 3 would have worse low light performance than a mk. 2 due to its partially stacked sensor due to the sensor design being more complex and therefore the amount of light on each pixel being more limited than a similar "plain" 24 MP BSI 35 mm sensor?

[@RejeRush]

He did tests between the high ISO and dynamic range capabilities of both the z6iii and z6ii in previous videos. Great videos!

[@maldeojocomunicaciones8427]

So, in the case of video, the native Iso of 6400 only applies when using N-log and reverts to 800 when using any other profile?

[@testcams]

Yes, the indicated vs sensor ISO discrepancy doesn't apply to any of the SDR-based profiles. I forgot to mention HLG in my video, which uses a sensor ISO that's 2EV below the indicated vs 3EV below indicated for N-Log, which means HLG ISO 400 is sensor ISO 100 and HLG ISO 3200 is sensor ISO 800.

[@barrymackichan7074]

One other point. A lot of photographers shoot “manual mode” with auto iso. The amount of light hitting the sensor is independent of the iso, so (assuming you don’t cross the line between the dual iso domains) auto bracketing will have no affect on the raw files other than changing iso number, just a number, that’s embedded in it. Similarly, taking exposures for HDR, would give multiple raw files that are essentially identical. For these methods to work, you need to set the ISO to be constant and change the exposures in a way that affects the amount of light going to the sensor - either aperture or shutter speed priority or manual. Am I correct?

[@testcams]

Agreed, there's not much benefit in AE bracketing with Auto ISO, unless the bracket pushed one of the Auto ISO exposures into a Dual Gain ISO, and the software used to create the merged exposure accounted for using the lower EVs of that bracket. But even then if you're bracketing you might as well bracket the physical exposure (shutter speed/aperture) rather than just the ISO.

[@nigerian-nightmare]

Can this be applied to Canon cameras or Panasonic, or is each brand different?

[@Bayonet1809]

Panasonic yes, Canon no. Canon uses their own different sensors.

[@RejeRush]

Question: I shoot weddings and faced with low light conditions that would normally have me going higher than ISO 800. Are you saying I should just keep my ISO at 800 even if the photo "look" underexposed?

[@Bayonet1809]

Yes, the only issue is that unfortunately mirrorless focus systems rely on a bright live-view image for their calculations, so if you are struggling for focus you should increase ISO to make the image brighter.

[@RejeRush]

@@Bayonet1809Thanks. Very interesting

[@aldolega]

Arri and some Canon cinema cameras use dual-gain systems like you describe at the end of the video. The newest Arri sensor reaches 17 stops of DR and it isn't even full-frame!

[@Bayonet1809]

Yes, they use a simultaneous read-out, which is currently not available in stills cameras.

[@Wildridefilms]

@@Bayonet1809 it is available on the G9ii, GH6 and GH7. Though the GH6 only activates it at ISO 800. The other 2 do it from ISO 100 and match the DR of 14 bit readout APSC and sit between 12 bit APSC and 12 bit FF.

[@Bayonet1809]

@@Wildridefilms I am unfamiliar with the internal workings of those Panasonic cameras, but there appear to be no real benefit to whatever they apparently do. Photons to Photos shows the maximum dynamic range of the G9II to be effectively the same as the older G9 (PDR ~10), while the APS-C Sony A6700 manages a PDR of almost 11. That evidence suggests that there is still a significant gap to APS-C, contrary to your claims.

[@Wildridefilms]

@@Bayonet1809 Take a look at cine D or Gerald undone. In video, they sit right in between 12 bit APSC and 12 bit Full frame. I don't know how Photons to photos do their tests, but if it says that the G9 and G9ii are even close in max DR, they're very wrong. There's a huge difference between them. The G9ii is at least 1 stop better, closer to 1.5 stops better than the G9.

[@Bayonet1809]

@@Wildridefilms Oh, you are referring to video. Well that explains it. I was talking about the feature not being available for stills.

[@rsmolkin]

Do you know if Z6 ii has a dual iso as well?

[@testcams]

It does, as does the Z6. All three generations' dual ISO is at 800.

[@tinger20]

Am I understanding the chart correctly, that the 150mp Phase1 has 7 steps?

[@Bayonet1809]

No, it only shows one additional gain level at ISO 400. The other steps in the graph just indicate that Phase skimped on their ADC firmware. Which makes sense since most of their users are not shooting above ISO 100 anyway.

[@b34k97]

Alright, I feel like you've buried the lede here with that very short segment on Dual ISO readout simulation at the end. I'm really curious now, as I think your test shot output looks really good! Much more natural than some HDR shot's I've done in LR previously. I have a few follow up questions here: How does this differ from an HDR Merge in Photoshop or Lightroom? Is your simulation software open source and/or available on GitHub somewhere? Does the software act on (and produce) a RAW (nef) file, or some other file types?

[@testcams]

I have an open-source project for all my raw-based tools at github.com/horshack-dpreview/OctaveRawTools. I haven't added this file yet. It's kind of hacked together at the moment and would need more work to make it usable across all cameras. Keep in mind the dual ISO readout simulation requires shooting two exposures and merging them together into a single raw, whereas a future, sensor-based implementation would accomplish this in a single exposure. Since you have to shoot two exposures you're better off using exposure bracketing, as it will provide noise improvement across a wider tonal range than the EV-grafting of my dual ISO readout simulation.

[@b34k97]

@@testcams Oh cool, thanks for the repo link! Yeah, it makes sense that the real power here would be getting this output without needing to take multiple exposures. I really did like how naturally your program seemed to stitch the two exposures together though... better than I've been able to do with exposure bracketing (though I'm still a beginner at this type of editing, so probably not the gold standard to go by).

[@barrymackichan7074]

It seems to me that the constraints are the resolution of the luminance measure at the shadow end (how close can two intensities be and still produce pixels with different values?) and the ‘overflow’ at the high end that leads to blown out highlights. The second problem seems to me to totally solvable by increasing the number of bits available in a pixel because the constraint here isn’t in the sensor hardware, but simply having enough bits to store the result of a computation. It should cause no problem for post-processing software (although the algorithms will need tweaking). I would think that a logarithmic scale would also fix this, but I don’t know much about them - yet. I have another question - in this two stage process with an analog step on the sensor (?) followed by digital multiplication essentially by the ISO, is one inherently better, provided enough bits are available to avoid overflow?

[@vivalasvegas702]

By July 17th, you’re going to have to explain Triple Gain ISO,🤞🤞🤞

[@boblozano]

www.youtube.com/@testcams any idea whether the z8/z9 or z6iii sensors have the ability to change the gain strategy via firmware, or is it hard-encoded in silicon?

[@testcams]

Good question. I'm guessing the dual gain switch is hard-coded in silicon to the ISO value, which firmware sets in the hardware interface/register, so it's under firmware control but not control that's independent of the ISO.

[@cameraprepper7938]

If you want the highest possible dynamic range in a photo image, then use the base ISO !!!

[@danielvilliers612]

Why use adobe software to do your test. It is known that at least at the start, Adobe induced much more noise to the Z9 images than Capture one. Why at least not use Nikon own free Nx Studio or Capture One, to get a more rounded example.

[@EricCalabros]

Your "hybrid" solution take two shots and combine them, and that's doubling the exposure, so you're reducing shot noise as well as overal read noise. A sensor can't do that with single exposure so the end result is not as dramatic improvement as your experiment.

[@testcams]

The exposure was not doubled in my Dual ISO readout simulation . Both exposures were taken at f/8 1/40, which you can see starting @ 9:19 if you look at ACR's EXIF info display in the upper-right area of the window just under the histogram. Only the ISO was changed, so there is only a reduction in read noise, with no reduction in photon shot noise.

[@EricCalabros]

How combining two 1/40s images doesn't reduce shot noise visibility to equivalent 1/20s shot?

[@BobN54]

Sorry to say, your explanation of dual gain is very confused and muddled. You've tried to make sense of the term 'conversion gain' and been confounded that this name is actually a misnomer - it's a conversion factor - not a gain at all. This is compounded by another misunderstanding, that 'gain' is applied to exposure. This is just incorrect. The exposure is immutable once the shutter is closed. It signifies an amount of light, and nothing that is done electronically or computationally results in the gain of light.

[@testcams]

Thanks for the feedback Bob. I followed Aptina's lead on using "conversion gain", which is how they termed the technique in the whitepaper discussing their invention of the technology. I agree with your general strict use of the word exposure, although its colloquial usage won out in my choice to use it in parts for the intended audience. I drew a pretty clear arc between that colloquial usage and the stricter form of physical light exposure (aperture, shutter) at important parts of the video, so I'm on board with shifting the usage to the stricter form, although I'm sure we probably differ on how to approach and introduce that shift to others.

[@BobN54]

@@testcams The term 'conversion gain' is what the industry uses, but as I said the name is misleading. If you understand it as a 'gain' you'll go wrong. It's a conversion factor and in particular dictates the equivalence between photon and electronic noise. So, if a high conversion gain is used an ADC producing a given amount of electronic noise appears to be less noisy than it would with a low CG. But a high CG limits the amount of charge that can be read from the pixel. Dual CG allows low CG to give a lot of charge for low ISOs and a high CG for low read noise at high ISOs. Charge is directly proportional to exposure, so inversely proportional to ISO. On exposure, I know it's very often misused, but it's meaning is essential to a clear explanation of this topic, so if you misuse it in the explanation you'll just end up with confusion. I think people giving explainers like this need to be careful to get their concepts right, especially core ones such as exposure. As to how to introduce the correct (rather than 'strict') use of exposure, why not just explain what it means? One of the reasons people don't know what it means is that people rarely give explanations. Old text books did.

[@camilo8cheryl]

WoW! Nose bleed👍

[@SirTubeALotMore]

#Nikon, why do you confuse us so much 😢 We shouldn't be forced to watch YT channels to understand your products!

[@chiftele08]

It's not nikon thing, all brands have sensors like this. Nikon has mostly ISO invariant sensors, the dual gain sensors are more in the Canon brand. That's the reason, you heard less about this topic than for example Canon users.

[@b34k97]

What? Electronics are complicated and to get the most out of it, you need to learn how it works at a fundamental level. Simplified user settings, generalized for all users, can only cover so many cases.