@@shagral Man, after years of playing with this stuff, this is the FIRST explanation that calls out applying exposure correction before any LUT/transform, and this nugget of information on top of all the other info.
Believe me when I say I've spent hundreds of hours exploring RUclips and other web sources to learn more about how cameras work, which specs really matter and why should I do this over that when capturing images. Your channel is by far the clearest and most useful source of information for me! Your videos are a great mix of deep and precise technical knowledge, luminous (no pun intended) analogies (seeing ISOs as dB and knowing that I'm actually playing with the Gain button is a complete game changer for me, same as the bucket of water analogy in your DR vid) and really well conceived pedagogy, reformulating capital info and summarizing the main takeaways at the end. It's a great gift to the world and I really want to thank you for it. I hope your views increase in the future and I hope that more people recommend checking out your work like Camera Conspiracies did for me. I know I will. Cheers from Germany!
Danke schön, Jean! Thank you for taking the time to write such a nice comment. For me the joy of knowing that I helped at least a few people to gain a better understanding of the craft is already reward enough to keep this channel going.
Great job!!! This must be the most no-nonsense video/cine/imaging tech channel I've seen so far, and I've been through lots of them that were mostly wasting my time and others. But... One thing I've met oh so often - why 'negative amplification'? There is a specific word for this, and it's called attenuation.
It's like with anything involving electrical currents or fields. It's a legitimate term, that noone - in the realm of the popular imaging videos at least - is using. Maybe 'negative amplification' sounds more catchy, bombastic, tabloid-style. Good ole' Carlin was so right on bullshit language.
Great work on the video. Thanks for helping me understand this. Also storytelling > image quality EVERYTIME! So just filming stuff people like is the most important battle to wage. It's why people would rather watch Star Trek TNG now versus whatever newer Trek they are making these days...
Yes, story trumps everything, and it's free! As long as you have imagination. Thanks for watching. As a side note, would your dog with 548K subscribers be interested in doing a little promotion for my channel in exchange for a biscuit? 🤣🤣🤣
How awesome to discover your channel! The level of analysis and professional production in your videos is at a different level. As someone with a weak background in camera tech and filming, this is an incredibly useful resource. As a fellow RUclips creator, I struggle to incorporate animated graphics in my own videos. I usually rely on various Final Cut Pro plugins, but they are not easily customizable and rarely give me what I am looking for. Can you give me some advice on how best someone like me could create animated graphics which doesn’t require a big learning curve and is relatively affordable? You do this amazingly well all the time, but time stamp 7:33 is one example I’m thinking of. Thanks! 🙏
Thanks! I just make simple slides in PowerPoint / Keynote with very simple animations and export them as a video. Then add transitions and zooms and stuff in the editor. It's pretty fast and easy.
First of all, I love your videos. They introduce a depth of information that almost no other video on this type on RUclips does. But, I’d like to interject on a few points, which I understood differently. First, while some dual sensitivity cameras have two sets of amplifiers (Varicam 35/LT, FX9 and DGO sensors like the ALEV), the FX3/6 doesn’t. It’s rather an interpolation that happens while down sampling. It’s actually believed that that sensor has around 42MP that are divided into groups of 4, effectively making it 12MP. The two native iso’s are then derived from two modes of signal interpolation - one averaged (lower base) and one additive (higher base). Second, I don’t believe that in general and across the board “native iso” = 0db. I think it’s simple the point where the signal produces the most dynamic range over and under. I actually do believe that Sony, for instance, used to boost the signal to 6db on the Fs7’s native slog ISO, for instance. I’m not sure I’m right about these, but this is at least what I’ve learned over the years.
Hey there, regarding 1) I'd like to see a proof link. And even if correct, this doesn't really change anything for the end user. For all intents and purposes it works like dual base ISO. The higher base is a bit noisier, sure, but at 12,800 it's crazy to complain. 2) that's very easy to verify with cameras that allow iso/gain mode switching. Just put it into 0db and switch to ISO - that's native ISO. Camera manufacturers might have different recommendations for rating/exposing particular gammas (life FS7 you mention), but that's beside the point.
@ oh non of this really matters. Just wanted to interject with a discussion. I don’t think it makes a difference, just find the details interesting. I don’t have a proof link. I got that info from Alister Chapman, who’s maybe not my favorite personality, but he has a fair amount of knowledge. The Fs7, as far as I know, doesn’t allow you to display gain in cine ei mode, where native iso is 2000. That’s why I said that in certain cameras it’s true, in others it’s not. “Native iso” means whatever the manufacturer wants it to mean.
@@shagral This is true and it’s been proven. The IMX510 is actually a 48 megapixel quad bayer sensor. Different cameras implement “dual gain” differently and not all “dual gain” is equal. While the performance between Sony‘s high end cinema line camera’s dual gain and lower end prosumer camera’s dual gain may not be extremely big (which is great for us end users), the way it is achieved is completely different and the difference between different camera brands can be very big. I’m sure different brands have a 3rd, 4th, and so on… more different ways of implementing dual gain but camera brands seem to be very secretive of how they’re doing it so it can be very hard to tell what’s going on under the hood. Here are the links about the IMX510: Sony a7S III has a 2×2 pixel binning IMX510 BSI sensor | The Landingfield Surprise: The Sony A7SIII actually has a 48 Megapixel Quad Bayer sensor! | sonyalpharumors
I’ve always shot under native. As much as I’ve read and been told that native iso is the best, it’s not the cleanest depending on what you’re shooting. My cleanest shots have ALWAYS been achieved by shooting the lowest ISO possible while adding light where I need it when possible. This is why lighting and faster lenses are important.
Useful as always. A couple of items to think about I felt were missing: 1. Of course in a low light situation you tend to use a Low ISO. You learn it first time you use a ARRI camera BUT if you use a "normal" non-raw camera adding dB BEFORE the AD conversion is way better than doing it in post. Of course it's better to use fast lenses and slow the shutter speed, but if you can't it's still better to add electrical noise rather than digital artifacts. Don't you agree. 2. Second thing: ETTR. As you are not recording on a Linear Color Space but it's a Log CS it's not a smart move to push the signal too much in a part of the compression curve where you have less available nuances. Because it would be above the noise floor, yes, but you would reduce the amount of data in the juicy area of the curve (midtones, skin, etc) to save useless (often useless at least) details in the shadows. Again... Am I missing something? Isn't it better to have them evenly distributed. 3. Dual Native: as you perfectly know the two circuitry are not very comparable. There is often a 2 stops gap in DR between them. Killer feature with a grain of salt.
Anthony, this video was intended for beginners, which you are clearly not :) Of course everything is more complicated. But it's impossible to provide all the caveats and "but's" while still keeping things simple and short. 1. What makes you think you are adding Dbs before ADC? At any rate I would personally never touch ISO unless I'm shooting WYSYWYG (which is never, lol). And with FX3/FX6 high base at 12,800 I literally never deviate from base ISO. The bit about lower ISO was meant more of an illustration to the thinking in terms of gain and how counterintuitive ISO can be. I don't actually think it's a necessary technique with modern cameras. 2. ETTR shouldn't be taken literally, of course, like in photography. A controlled overexposure of just ~1.5 stops is enough to get all of the benefits without the drawbacks you mentioned. And even then it's a situational technique that requires judgement. Exposure is a trade-off, there is no on-size-fits-all formula, you have to decide what to sacrifice and what to prioritise. Besides ETTR has other drawbacks (moving the exposure decision into post production and potentially out of your control can be a pain). 3. Totally depends on the camera. With Sony FX9 and Venice the two ISOs have almost identical DR. When the difference is higher (FX3) Sony tends to not call it "dual base ISO", and even then the difference is about 0,5 stop, not 2. I still believe it's a killer feature, I personally wouldn't buy a camera without it.
@@shagral considering how deep you go into details and technicalities (which is good) it's sometimes difficult to understand who your videos are aimed to. The DR video was way above the (unnecessary?) knowledge you see on a high end set. Regarding 1. Even being a professional you know it is not common to know how the whole circuitry works. Considering you know more, I am actually asking. I thought the ISO process would involve increasing the receptivity (analogic/phisical) of the photocells with an electrical boost (like a mic) and then it would be converted into a digital signal via the ADC. Plus there is a further process of compression of the digital signal into a compressed video file I did not take in consideration in the previous comment. This is another passage where you irreversibly lose data. Again, I don't know, just sharing the train of thoughts but I think in the end it's still better to work with the ISO if all the other resources are unavailable, rather than underexpose and recover in post production. Happy to be corrected if you have more specific tests/data. 1B. Regarding the low ISO with ARRI you actually go lower with ISO as the DR from 100 to 800 does not change and it's only changing where it's leaving more room in comparision with mid grey. So the concept was still interesting to analyze even for other cameras. Moreover we are professional but we jump back and forth from super professional cameras to run and gun situations where the SXT would be a suicide and a C300mk3 is already too much. So having a clearer view about the tech behind cameras at every level is super useful. 2. the ETTR thing was actually more about the question in the end as you have probably analyzed the actual math behind it better than me. I am still trying to wrap my head around the concept: if I have 8 stops of DR in the scene (with a standard camera of course) and 12 stops of DR in the ProRes file, is every stop in the log scale absolutely equivalent to the other? It probably is but still that was the doubt... Using the stops from 5 to 12 or from 3 to 10 is absolutely the same... 3. it's .5 stop on a Xyla chart (i thought more, but did not check). You are the first one to say in your videos that it can be fooled in these heavy NR situations. Maybe 2 is an exaggeration but .5 seems a bit optimistic. And yes grain of salt does not mean it's not very useful to have it, just means remember that there is no such thing as a free lunch. Thanks for the good work
1. I'm pretty sure ISO is just digital gain. Yes, there's a lot of stuff happening before the ADC but I'm pretty sure none of it is directly accessible to the end user, at least on what you call "normal" cameras. I could be wrong. But in practical terms, yes, if shooting Rec.709 I would use ISO as a last resort because camera image processing would yield better results that doing it yourself in post. But that only applies tor Rec.709 (WYSYWYG) shooting. With log if you underexpose, you are in trouble. 1B. I'm not sure what you mean here, but with ARRI ISO is just an exposure index. It basically shifts the available DR between highlights and shadow. Camera still records at native. 2. Yes, I see what you mean. And indeed the higher stops in log are more lossy. More and more linear data is thrown away in the brighter stops. So moving your skintones into stops 10-12 is indeed idiotic, but that's taking ETTR to an absurd extreme. The sweetspot is always the middle, couple of stops above and below middle gray. If you intend for key side of the face to be at 0ev, then shooting it at +1.5ev really has no downsides (unless your highlights are already clipping). 3. This is a fuzzy territory because what is acceptable noise-wise will be different for different people and situations. Practically speaking, I can see the image from 12,800 is noisier than 800 but the exposure latitude is the same when I push and pull it in post. It just means using some NR. I would rather have that option than be totally F'd in low light situations.
@@shagral Just to share the culture. I have just checked a couple of scientific sources and the gain (as I supposed, because it was too logical to be otherwise) is an electrical gain applied to the photodiod. Like in a microphone, it's electrical/analogic noise applied before the ADC.
@@shagral As far as I understand point 1., it’s more complicated than that. Different cameras do it differently and some do it before the ADC while others do it after the ADC and others as a complicated combination of both. It’s often referred to as ISO variance and invariance. I believe it is actually wise to raise the ISO to a certain point (when you need the extra boost) on ISO invariant cameras when shooting to a compressed file format because if you were to not raise the gain in camera, and rather raise it in post, the compression artifacts would get magnified and ruin the image as well as other traits that do not pair well with compression such as internal noise reduction. On the other hand, if you are shooting true raw on an ISO invariant camera, it is never wise to raise the ISO in camera above it’s native ISO (native ISO’s if it’s dual gain) as by doing so, you’re literally decreasing your DR for no other benefit in return. Raising the ISO in post will give you the exact same result as in camera but with more highlight information since you won’t be clipping as much information for no reason. I would recommend the same even with compressed raw (speaking from experience), depending on how compressed it is, there’s visually no difference but with the benefit of more DR. ISO variant cameras are a whole nother story since even if you’re shooting raw with them, it may be wise to raise the ISO in camera (when needed) as that would amplify the signal before the ADC giving a much cleaner result then if you were to raise the signal in post especially if shooting to a compressed format. So in many cameras, you’ll actually get better results by raising the ISO in Camera rather than sticking to the native ISO and then boosting it in post. Here are a few videos that explain these concepts the best I’ve seen: ruclips.net/video/gisj6Og3xCI/видео.htmlsi=BS2iEpthQRSW6QLa ruclips.net/video/iiMfAmWbWSg/видео.htmlsi=tOVTcQgBvryltUnL
I used to call myself the "Shooting LOG GAMMA" police. Tryin to explain why so many RUclipsrs misunderstand ISO, amplification, data spaces, etc. The desire of most filmmakers to believe some GAMMA setting and LUT is going to get them on Netflix is overpowering. The descriptions you wrote for your chapters are like a short bible on how cameras really work. The only thing that would keep me going is some people who would comment, "I always wondered why my skin tones seemed blotchy and off" (because they were shooing an aggressive LOG in controlled lighting). Anyway, glad to see others trying to enlighten filmmakers on this fascinating stuff!
It's really hard to learn that stuff organically. You have to really seek it out and wade through heaps of bad info. But yeah, I'm glad it's helped a couple people out. Thanks for watching!
@shagral I saw quite a few videos exposing age old myths poping up recently. The only one that can compare to yours in details and accuracy is by professional colorist who included his grading workflow and practical tips. But your research, production value and pacing are superior. And the tips for LUTs is the extra icing many creators ignore.
Such an informative and useful video, thank you for putting this information out there for all of us who are less informed. Matter of fact, that goes for pretty much every video of yours i've seen so far. Also as some have touched on already, great delivery as well. The way you explain these things finally makes sense to me after wading through so much information that is either just wrong, doen't make sense, or is poorly presented and hard to wrap my head around. Comedic touch is also much appreciated. As I really only shoot photography and have very little experience or knowledge of video, I have a question about your advice to not shoot in LOG if you are shooting a dark scene with no ability to add light and therefor must turn up the gain. How would this pertain to shooting still photography? At that point would it be better for me to shoot jpeg rather than RAW? maybe i'm drawing a connection between RAW and LOG that just doesn't exist, hopefully you can enlighten me on best procedures in this situation... also one other thing to perhaps mention is i use a Fuji camera, so its in fact not a Bayer sensor - a topic which you mentioned at a different point in your video - but I have no idea whether this makes a difference either. If you have time to respond, I want to say thanks for taking the time to explain some of this to a novice like myself. If not, then I understand. Either way you have gained a subscriber and looking forward to your future content
Hey thanks for watching! Not sure I understood your question, but there is no LOG gamma in still photography, as it is not needed (apart from HLG). As for ISO, when shooting RAW stills ISO is only saved as metadata and photos will require noise reduction in post if underexposed. Shooting jpg (or better heif) is definitely an option when you need fast turnaround. But RAW as a backup is always a good idea. In terms of what ISO still looks OK in jpg/heif - totally depends on your camera. Do some controlled exposure tests at home and you will find the breaking point and the exposure sweet spot. With most decent cameras you can go 3-4 stops under native (if native is ISO100, then -3 is ISO 800 and -4 is ISO 1600) and still capture a decent picture. Dual native ISO works in photo mode too. Fuji's pattern is not that big of a deal, it's still a CMOS sensor with a colour filter array, just a different pattern. And as I understand in case of X-Trans it was a mostly a cost-cutting measure - so they could avoid putting an OLPF over their relatively high megapixel sensor while still reducing aliasing. I don't own Fuji myself but I've seen reports of weird artefacts (pink teeth and sclera) caused by that pattern in some edge cases.
Great video and article. Very hard to find good information on this topic. I have been trying to understand why the fuji x-t3 base ISO is 160 (0 gain) for photos and standard video but in f-log base ISO is 640. I'm a little less confused but still not sure how base ISO (0 gain) can have 2 different values excluding a dual base ISO sensor. I understand that the dynamic range setting at DR400 requires a minimum ISO640 because the camera actually exposes 2 stops under using base ISO160 and then brings up the shadows when processing jpegs. I was thinking something similar might be happening here. It's got me wondering how shooting video with ETERNA video profile at DR400 ISO640 compares to f-log ISO640. Subscribed and keep up the excellent information.
Hey, thanks for watching and subbing! The brief answer to your question is at 6:12. Log gammas (f-log video) and linear gamma (raw photos) are exposed differently, hence the difference in base ISO. No gain is being applied, it's just a number that you would have to put into a light meter to get correct exposure for your chosen gamma.
This has totally changed the entire way i setup my camera and custom modes. The biggest thing to take from this video - " The iso numbers are mostly just made up". I used to waste time trying to memorise high and low native ISO's for different cameras and different picture profile.... TOTAL WASTE OF TIME!!! You simply use High OR Low gain in any picture profile and any camera.
In the Exposure to ISO chart's the 18% Gray point sort of marches down with with Increased ISO, but what I've not seen said anywhere is does the 18% gray point also reflect an alignment to color gamut? (Meaning color gamut shifts between your dynamic range bright vs Dark). Since skin tone's are the most important color to match, does this mean that ISO should be used as a tool to place skin tones correctly in the dynamic range or is Base ISO always better? I've heard on the color grading forums that ETTR causes problems during color grading.
Colour gamut is a three-dimensional palette, it doesn't shift. But the saturation and lightness of colours of course track with exposure. But in case of shooting 10-bit Log or higher quality, you have complete control of that mapping in post production. ETTR only causes problems if A) clipping of one or more of the RGB channels occurred or B) destructive/incorrect colour pipeline is used (LUTs as opposed to colour management or mathematical transforms). Also, ETTR should not be taken literally, usually after ~1.7 stops of overexposure there is zero benefit to exposing even brighter. And even that is only recommended for really low-key scenes. Hope that helps.
Thanks! When shooting RAW the camera only ever records at native ISO. ISO setting becomes EI (exposure index) setting and gets recorded as metadata in the file. EI is basically a monitoring exposure offset, you tell the camera "show me the picture AS IF I had ISO of ____". You can use this to redistribute dynamic range as you desire between shadows (lower EI) and highlights (higher EI).
As a computer engineer, I'm curious as to the path that lead you to your rigger? Is it the heavy metal drumming? I would not doubt it. I do really wish you would have used the language of normalization.
"you lose highlight retention" this does not matter when one is in the dark and there are no highlights. My mystical learning is that adding gain before compression (which tends to be hard in the shadows) is good and therefore one should use a higher ISO when it is the correct thing to do? As it happens.. I have a C200 and I literally build the menu and assigned buttons so that one cannot change the ISO!
It is true that companies can tune their in-camera NR specifically for their sensor, but you also have to factor in the significant restraints they face; if you do it in-camera you need to do the NR in real time with only the camera's internal processor. If you do NR in post you have unlimited time and likely a processor hundreds of times more powerful than the one in your camera.
I have a question about the ISO on my Z Cam F6, which has a dual native ISO of 400 & 2500. When I shoot a scene that is properly exposed at the base ISO of 400, I notice it has more noise compared to when I ETTR at a higher ISO, like 3200. Why does this happen?
Short answer is: "working as intended". F6 uses the Sony A7III sensor if I'm not mistaken and it is a bit noisy and requires ETTR if you want to get squeaky-clean shadows. However, Z-Cam are good at recording a more pristine noise profile with minimal NR and sharpening in camera, so it should withstand NR in post with minimal degradations in quality. It does require more work in post, but that's "cinema cameras" for you. If you want clean shadows out of the box, use ETTR. High/low ISO are almost equal in terms of noise, biggest factor here is ETTR or not. Edit: oh, and do some tests! set up at home in controlled environment and cycle through all the ISOs, over/underexposing and study what happens when normalised in post. That's the best way to find and see the exposure sweet spots // breaking points of your camera. Learn your tools!
Great video on a very misunderstood concept! My question - in a dark situation, instead of using negative gain, is it better to just leave things as-is at your (highest) base-iso, and change 'exposure' (gain) in post???
The goal is to get your waveform above the noise floor as much as possible without clipping the highlights. Negative gain is just a way to lower the noise floor (at the cost of DR in the highlights). But I wouldn't bother with that if you have a modern-ish camera (say 10 bit and ~12 stops of DR). Everything else depends on your camera and scene!
@@shagral I have this question too, in your situration, negative gain lower the noise floor, but at the same time, the brightness of shooting subject lower too, which means the signel and the data we want, also lower and still within the noise floor, how do we get benefit frow negative gain in low light situration? And thank you so much for the video. I am enlightened!
@johnsek4294 low light doesn't mean no light. If we are talking about cinematic exposure (meaning you have some control over light) then you will have to pour more light on your subject to compensate for lower ISO. But you shadows will be cleaner. If it's a total run and gun and you are shooting a black dog on a moonless night then don’t shoot in LOG!
I'm studying filmmaking and recently bought a sony a6600, but i just can't find what the native ISO is, if you google it it just tells you the recomended range wich is completely useless (it's 100 - 32,000)
Quick google search says it's a dual gain camera with low base at 100 ISO and high base at 400 ISO. Seems about right to me. Source: cuchara.photography/2019/06/iso-invariance-and-the-sony-a6500/
😊
Thank you!
@@shagral Man, after years of playing with this stuff, this is the FIRST explanation that calls out applying exposure correction before any LUT/transform, and this nugget of information on top of all the other info.
Believe me when I say I've spent hundreds of hours exploring RUclips and other web sources to learn more about how cameras work, which specs really matter and why should I do this over that when capturing images. Your channel is by far the clearest and most useful source of information for me! Your videos are a great mix of deep and precise technical knowledge, luminous (no pun intended) analogies (seeing ISOs as dB and knowing that I'm actually playing with the Gain button is a complete game changer for me, same as the bucket of water analogy in your DR vid) and really well conceived pedagogy, reformulating capital info and summarizing the main takeaways at the end. It's a great gift to the world and I really want to thank you for it. I hope your views increase in the future and I hope that more people recommend checking out your work like Camera Conspiracies did for me. I know I will. Cheers from Germany!
Danke schön, Jean! Thank you for taking the time to write such a nice comment. For me the joy of knowing that I helped at least a few people to gain a better understanding of the craft is already reward enough to keep this channel going.
this channel is a hidden gem! Keep up the great work!
Thanks a ton!
It really it! I keep coming back to this one explanation of it and the Gerald Undone gag was perfect!!! haha
Great job!!!
This must be the most no-nonsense video/cine/imaging tech channel I've seen so far, and I've been through lots of them that were mostly wasting my time and others.
But... One thing I've met oh so often - why 'negative amplification'? There is a specific word for this, and it's called attenuation.
Makes sense! Like with microphones.
It's like with anything involving electrical currents or fields. It's a legitimate term, that noone - in the realm of the popular imaging videos at least - is using. Maybe 'negative amplification' sounds more catchy, bombastic, tabloid-style. Good ole' Carlin was so right on bullshit language.
Very professional POV on the topic. Thank you so much for sharing with us.
Fantastic! Gold advice right here 10:15 This exactly why I was having an issue - I was adjusting exposure AFTER I applied the LUT
Glad it helped!
I was shocked when I saw your subscriber amount.. Great video!
Thank you! We are all at the mercy of the algorithm 🥲
Great work on the video. Thanks for helping me understand this. Also storytelling > image quality EVERYTIME! So just filming stuff people like is the most important battle to wage. It's why people would rather watch Star Trek TNG now versus whatever newer Trek they are making these days...
Yes, story trumps everything, and it's free! As long as you have imagination. Thanks for watching.
As a side note, would your dog with 548K subscribers be interested in doing a little promotion for my channel in exchange for a biscuit? 🤣🤣🤣
This is literally the best video on the subject I have seen!
"oh wait, that's the other guy's thing..." 🤣🤣 killed me
How awesome to discover your channel! The level of analysis and professional production in your videos is at a different level. As someone with a weak background in camera tech and filming, this is an incredibly useful resource.
As a fellow RUclips creator, I struggle to incorporate animated graphics in my own videos. I usually rely on various Final Cut Pro plugins, but they are not easily customizable and rarely give me what I am looking for. Can you give me some advice on how best someone like me could create animated graphics which doesn’t require a big learning curve and is relatively affordable? You do this amazingly well all the time, but time stamp 7:33 is one example I’m thinking of.
Thanks! 🙏
Thanks! I just make simple slides in PowerPoint / Keynote with very simple animations and export them as a video. Then add transitions and zooms and stuff in the editor. It's pretty fast and easy.
First of all, I love your videos.
They introduce a depth of information that almost no other video on this type on RUclips does.
But, I’d like to interject on a few points, which I understood differently.
First, while some dual sensitivity cameras have two sets of amplifiers (Varicam 35/LT, FX9 and DGO sensors like the ALEV), the FX3/6 doesn’t. It’s rather an interpolation that happens while down sampling. It’s actually believed that that sensor has around 42MP that are divided into groups of 4, effectively making it 12MP. The two native iso’s are then derived from two modes of signal interpolation - one averaged (lower base) and one additive (higher base).
Second, I don’t believe that in general and across the board “native iso” = 0db. I think it’s simple the point where the signal produces the most dynamic range over and under. I actually do believe that Sony, for instance, used to boost the signal to 6db on the Fs7’s native slog ISO, for instance.
I’m not sure I’m right about these, but this is at least what I’ve learned over the years.
Hey there, regarding 1) I'd like to see a proof link. And even if correct, this doesn't really change anything for the end user. For all intents and purposes it works like dual base ISO. The higher base is a bit noisier, sure, but at 12,800 it's crazy to complain. 2) that's very easy to verify with cameras that allow iso/gain mode switching. Just put it into 0db and switch to ISO - that's native ISO. Camera manufacturers might have different recommendations for rating/exposing particular gammas (life FS7 you mention), but that's beside the point.
@ oh non of this really matters. Just wanted to interject with a discussion. I don’t think it makes a difference, just find the details interesting.
I don’t have a proof link. I got that info from Alister Chapman, who’s maybe not my favorite personality, but he has a fair amount of knowledge.
The Fs7, as far as I know, doesn’t allow you to display gain in cine ei mode, where native iso is 2000. That’s why I said that in certain cameras it’s true, in others it’s not. “Native iso” means whatever the manufacturer wants it to mean.
@@shagral This is true and it’s been proven. The IMX510 is actually a 48 megapixel quad bayer sensor. Different cameras implement “dual gain” differently and not all “dual gain” is equal. While the performance between Sony‘s high end cinema line camera’s dual gain and lower end prosumer camera’s dual gain may not be extremely big (which is great for us end users), the way it is achieved is completely different and the difference between different camera brands can be very big. I’m sure different brands have a 3rd, 4th, and so on… more different ways of implementing dual gain but camera brands seem to be very secretive of how they’re doing it so it can be very hard to tell what’s going on under the hood. Here are the links about the IMX510:
Sony a7S III has a 2×2 pixel binning IMX510 BSI sensor | The Landingfield
Surprise: The Sony A7SIII actually has a 48 Megapixel Quad Bayer sensor! | sonyalpharumors
I’ve always shot under native. As much as I’ve read and been told that native iso is the best, it’s not the cleanest depending on what you’re shooting. My cleanest shots have ALWAYS been achieved by shooting the lowest ISO possible while adding light where I need it when possible. This is why lighting and faster lenses are important.
native ISO as the point of 0 db! Finally, it makes sense to me, thank you! (Now subbed)
Thanks and welcome aboard 🙌🏻
Subscribed! Looking forward to your future content.
Awesome, thank you!
Useful as always. A couple of items to think about I felt were missing:
1. Of course in a low light situation you tend to use a Low ISO. You learn it first time you use a ARRI camera BUT if you use a "normal" non-raw camera adding dB BEFORE the AD conversion is way better than doing it in post. Of course it's better to use fast lenses and slow the shutter speed, but if you can't it's still better to add electrical noise rather than digital artifacts. Don't you agree.
2. Second thing: ETTR. As you are not recording on a Linear Color Space but it's a Log CS it's not a smart move to push the signal too much in a part of the compression curve where you have less available nuances. Because it would be above the noise floor, yes, but you would reduce the amount of data in the juicy area of the curve (midtones, skin, etc) to save useless (often useless at least) details in the shadows. Again... Am I missing something? Isn't it better to have them evenly distributed.
3. Dual Native: as you perfectly know the two circuitry are not very comparable. There is often a 2 stops gap in DR between them. Killer feature with a grain of salt.
Anthony, this video was intended for beginners, which you are clearly not :) Of course everything is more complicated. But it's impossible to provide all the caveats and "but's" while still keeping things simple and short.
1. What makes you think you are adding Dbs before ADC? At any rate I would personally never touch ISO unless I'm shooting WYSYWYG (which is never, lol). And with FX3/FX6 high base at 12,800 I literally never deviate from base ISO. The bit about lower ISO was meant more of an illustration to the thinking in terms of gain and how counterintuitive ISO can be. I don't actually think it's a necessary technique with modern cameras.
2. ETTR shouldn't be taken literally, of course, like in photography. A controlled overexposure of just ~1.5 stops is enough to get all of the benefits without the drawbacks you mentioned. And even then it's a situational technique that requires judgement. Exposure is a trade-off, there is no on-size-fits-all formula, you have to decide what to sacrifice and what to prioritise. Besides ETTR has other drawbacks (moving the exposure decision into post production and potentially out of your control can be a pain).
3. Totally depends on the camera. With Sony FX9 and Venice the two ISOs have almost identical DR. When the difference is higher (FX3) Sony tends to not call it "dual base ISO", and even then the difference is about 0,5 stop, not 2. I still believe it's a killer feature, I personally wouldn't buy a camera without it.
@@shagral considering how deep you go into details and technicalities (which is good) it's sometimes difficult to understand who your videos are aimed to. The DR video was way above the (unnecessary?) knowledge you see on a high end set.
Regarding 1. Even being a professional you know it is not common to know how the whole circuitry works. Considering you know more, I am actually asking. I thought the ISO process would involve increasing the receptivity (analogic/phisical) of the photocells with an electrical boost (like a mic) and then it would be converted into a digital signal via the ADC. Plus there is a further process of compression of the digital signal into a compressed video file I did not take in consideration in the previous comment. This is another passage where you irreversibly lose data. Again, I don't know, just sharing the train of thoughts but I think in the end it's still better to work with the ISO if all the other resources are unavailable, rather than underexpose and recover in post production. Happy to be corrected if you have more specific tests/data.
1B. Regarding the low ISO with ARRI you actually go lower with ISO as the DR from 100 to 800 does not change and it's only changing where it's leaving more room in comparision with mid grey. So the concept was still interesting to analyze even for other cameras. Moreover we are professional but we jump back and forth from super professional cameras to run and gun situations where the SXT would be a suicide and a C300mk3 is already too much. So having a clearer view about the tech behind cameras at every level is super useful.
2. the ETTR thing was actually more about the question in the end as you have probably analyzed the actual math behind it better than me. I am still trying to wrap my head around the concept: if I have 8 stops of DR in the scene (with a standard camera of course) and 12 stops of DR in the ProRes file, is every stop in the log scale absolutely equivalent to the other? It probably is but still that was the doubt... Using the stops from 5 to 12 or from 3 to 10 is absolutely the same...
3. it's .5 stop on a Xyla chart (i thought more, but did not check). You are the first one to say in your videos that it can be fooled in these heavy NR situations. Maybe 2 is an exaggeration but .5 seems a bit optimistic. And yes grain of salt does not mean it's not very useful to have it, just means remember that there is no such thing as a free lunch.
Thanks for the good work
1. I'm pretty sure ISO is just digital gain. Yes, there's a lot of stuff happening before the ADC but I'm pretty sure none of it is directly accessible to the end user, at least on what you call "normal" cameras. I could be wrong. But in practical terms, yes, if shooting Rec.709 I would use ISO as a last resort because camera image processing would yield better results that doing it yourself in post. But that only applies tor Rec.709 (WYSYWYG) shooting. With log if you underexpose, you are in trouble.
1B. I'm not sure what you mean here, but with ARRI ISO is just an exposure index. It basically shifts the available DR between highlights and shadow. Camera still records at native.
2. Yes, I see what you mean. And indeed the higher stops in log are more lossy. More and more linear data is thrown away in the brighter stops. So moving your skintones into stops 10-12 is indeed idiotic, but that's taking ETTR to an absurd extreme. The sweetspot is always the middle, couple of stops above and below middle gray. If you intend for key side of the face to be at 0ev, then shooting it at +1.5ev really has no downsides (unless your highlights are already clipping).
3. This is a fuzzy territory because what is acceptable noise-wise will be different for different people and situations. Practically speaking, I can see the image from 12,800 is noisier than 800 but the exposure latitude is the same when I push and pull it in post. It just means using some NR. I would rather have that option than be totally F'd in low light situations.
@@shagral Just to share the culture. I have just checked a couple of scientific sources and the gain (as I supposed, because it was too logical to be otherwise) is an electrical gain applied to the photodiod. Like in a microphone, it's electrical/analogic noise applied before the ADC.
@@shagral As far as I understand point 1., it’s more complicated than that. Different cameras do it differently and some do it before the ADC while others do it after the ADC and others as a complicated combination of both. It’s often referred to as ISO variance and invariance. I believe it is actually wise to raise the ISO to a certain point (when you need the extra boost) on ISO invariant cameras when shooting to a compressed file format because if you were to not raise the gain in camera, and rather raise it in post, the compression artifacts would get magnified and ruin the image as well as other traits that do not pair well with compression such as internal noise reduction. On the other hand, if you are shooting true raw on an ISO invariant camera, it is never wise to raise the ISO in camera above it’s native ISO (native ISO’s if it’s dual gain) as by doing so, you’re literally decreasing your DR for no other benefit in return. Raising the ISO in post will give you the exact same result as in camera but with more highlight information since you won’t be clipping as much information for no reason. I would recommend the same even with compressed raw (speaking from experience), depending on how compressed it is, there’s visually no difference but with the benefit of more DR. ISO variant cameras are a whole nother story since even if you’re shooting raw with them, it may be wise to raise the ISO in camera (when needed) as that would amplify the signal before the ADC giving a much cleaner result then if you were to raise the signal in post especially if shooting to a compressed format. So in many cameras, you’ll actually get better results by raising the ISO in Camera rather than sticking to the native ISO and then boosting it in post. Here are a few videos that explain these concepts the best I’ve seen:
ruclips.net/video/gisj6Og3xCI/видео.htmlsi=BS2iEpthQRSW6QLa
ruclips.net/video/iiMfAmWbWSg/видео.htmlsi=tOVTcQgBvryltUnL
I used to call myself the "Shooting LOG GAMMA" police. Tryin to explain why so many RUclipsrs misunderstand ISO, amplification, data spaces, etc. The desire of most filmmakers to believe some GAMMA setting and LUT is going to get them on Netflix is overpowering. The descriptions you wrote for your chapters are like a short bible on how cameras really work. The only thing that would keep me going is some people who would comment, "I always wondered why my skin tones seemed blotchy and off" (because they were shooing an aggressive LOG in controlled lighting). Anyway, glad to see others trying to enlighten filmmakers on this fascinating stuff!
It's really hard to learn that stuff organically. You have to really seek it out and wade through heaps of bad info. But yeah, I'm glad it's helped a couple people out. Thanks for watching!
Thanks for the video. Your explanation of ISO helps unravel a few mysteries.
Thanks for watching! Glad it helped you
@shagral I saw quite a few videos exposing age old myths poping up recently. The only one that can compare to yours in details and accuracy is by professional colorist who included his grading workflow and practical tips. But your research, production value and pacing are superior. And the tips for LUTs is the extra icing many creators ignore.
Such an informative and useful video, thank you for putting this information out there for all of us who are less informed. Matter of fact, that goes for pretty much every video of yours i've seen so far. Also as some have touched on already, great delivery as well. The way you explain these things finally makes sense to me after wading through so much information that is either just wrong, doen't make sense, or is poorly presented and hard to wrap my head around. Comedic touch is also much appreciated.
As I really only shoot photography and have very little experience or knowledge of video, I have a question about your advice to not shoot in LOG if you are shooting a dark scene with no ability to add light and therefor must turn up the gain. How would this pertain to shooting still photography? At that point would it be better for me to shoot jpeg rather than RAW? maybe i'm drawing a connection between RAW and LOG that just doesn't exist, hopefully you can enlighten me on best procedures in this situation... also one other thing to perhaps mention is i use a Fuji camera, so its in fact not a Bayer sensor - a topic which you mentioned at a different point in your video - but I have no idea whether this makes a difference either.
If you have time to respond, I want to say thanks for taking the time to explain some of this to a novice like myself. If not, then I understand. Either way you have gained a subscriber and looking forward to your future content
Hey thanks for watching! Not sure I understood your question, but there is no LOG gamma in still photography, as it is not needed (apart from HLG). As for ISO, when shooting RAW stills ISO is only saved as metadata and photos will require noise reduction in post if underexposed. Shooting jpg (or better heif) is definitely an option when you need fast turnaround. But RAW as a backup is always a good idea. In terms of what ISO still looks OK in jpg/heif - totally depends on your camera. Do some controlled exposure tests at home and you will find the breaking point and the exposure sweet spot. With most decent cameras you can go 3-4 stops under native (if native is ISO100, then -3 is ISO 800 and -4 is ISO 1600) and still capture a decent picture. Dual native ISO works in photo mode too.
Fuji's pattern is not that big of a deal, it's still a CMOS sensor with a colour filter array, just a different pattern. And as I understand in case of X-Trans it was a mostly a cost-cutting measure - so they could avoid putting an OLPF over their relatively high megapixel sensor while still reducing aliasing. I don't own Fuji myself but I've seen reports of weird artefacts (pink teeth and sclera) caused by that pattern in some edge cases.
Good job on the video!
Great video and article. Very hard to find good information on this topic. I have been trying to understand why the fuji x-t3 base ISO is 160 (0 gain) for photos and standard video but in f-log base ISO is 640. I'm a little less confused but still not sure how base ISO (0 gain) can have 2 different values excluding a dual base ISO sensor. I understand that the dynamic range setting at DR400 requires a minimum ISO640 because the camera actually exposes 2 stops under using base ISO160 and then brings up the shadows when processing jpegs. I was thinking something similar might be happening here. It's got me wondering how shooting video with ETERNA video profile at DR400 ISO640 compares to f-log ISO640. Subscribed and keep up the excellent information.
Hey, thanks for watching and subbing! The brief answer to your question is at 6:12. Log gammas (f-log video) and linear gamma (raw photos) are exposed differently, hence the difference in base ISO. No gain is being applied, it's just a number that you would have to put into a light meter to get correct exposure for your chosen gamma.
This was very helpful! Thanks!
Thanks for watching!
This has totally changed the entire way i setup my camera and custom modes. The biggest thing to take from this video - " The iso numbers are mostly just made up". I used to waste time trying to memorise high and low native ISO's for different cameras and different picture profile.... TOTAL WASTE OF TIME!!! You simply use High OR Low gain in any picture profile and any camera.
Yes, just like a roll of film has only one ISO. Thanks for your continued support Jeff!
In the Exposure to ISO chart's the 18% Gray point sort of marches down with with Increased ISO, but what I've not seen said anywhere is does the 18% gray point also reflect an alignment to color gamut? (Meaning color gamut shifts between your dynamic range bright vs Dark). Since skin tone's are the most important color to match, does this mean that ISO should be used as a tool to place skin tones correctly in the dynamic range or is Base ISO always better? I've heard on the color grading forums that ETTR causes problems during color grading.
Colour gamut is a three-dimensional palette, it doesn't shift. But the saturation and lightness of colours of course track with exposure. But in case of shooting 10-bit Log or higher quality, you have complete control of that mapping in post production.
ETTR only causes problems if A) clipping of one or more of the RGB channels occurred or B) destructive/incorrect colour pipeline is used (LUTs as opposed to colour management or mathematical transforms). Also, ETTR should not be taken literally, usually after ~1.7 stops of overexposure there is zero benefit to exposing even brighter. And even that is only recommended for really low-key scenes.
Hope that helps.
Great video! What about RAW? Shoot at native iso? Or...
Thanks! When shooting RAW the camera only ever records at native ISO. ISO setting becomes EI (exposure index) setting and gets recorded as metadata in the file. EI is basically a monitoring exposure offset, you tell the camera "show me the picture AS IF I had ISO of ____". You can use this to redistribute dynamic range as you desire between shadows (lower EI) and highlights (higher EI).
VERY interesting, thank you.😇
Thanks for watching!
Now it makes sense to me
As a computer engineer, I'm curious as to the path that lead you to your rigger? Is it the heavy metal drumming? I would not doubt it. I do really wish you would have used the language of normalization.
"...and we are working with beyer pattern sensors here..." How DARE you ignore us Fuji shooter! There are DOZENS of us!
"you lose highlight retention" this does not matter when one is in the dark and there are no highlights. My mystical learning is that adding gain before compression (which tends to be hard in the shadows) is good and therefore one should use a higher ISO when it is the correct thing to do? As it happens.. I have a C200 and I literally build the menu and assigned buttons so that one cannot change the ISO!
It is true that companies can tune their in-camera NR specifically for their sensor, but you also have to factor in the significant restraints they face; if you do it in-camera you need to do the NR in real time with only the camera's internal processor. If you do NR in post you have unlimited time and likely a processor hundreds of times more powerful than the one in your camera.
Excellent presentation. Using Keynote?
Good old-fashioned PowerPoint 😂
I have a question about the ISO on my Z Cam F6, which has a dual native ISO of 400 & 2500. When I shoot a scene that is properly exposed at the base ISO of 400, I notice it has more noise compared to when I ETTR at a higher ISO, like 3200. Why does this happen?
Short answer is: "working as intended". F6 uses the Sony A7III sensor if I'm not mistaken and it is a bit noisy and requires ETTR if you want to get squeaky-clean shadows. However, Z-Cam are good at recording a more pristine noise profile with minimal NR and sharpening in camera, so it should withstand NR in post with minimal degradations in quality. It does require more work in post, but that's "cinema cameras" for you. If you want clean shadows out of the box, use ETTR. High/low ISO are almost equal in terms of noise, biggest factor here is ETTR or not.
Edit: oh, and do some tests! set up at home in controlled environment and cycle through all the ISOs, over/underexposing and study what happens when normalised in post. That's the best way to find and see the exposure sweet spots // breaking points of your camera. Learn your tools!
@ Thank you for the info. Definitely will do some exposure test.
Great video on a very misunderstood concept! My question - in a dark situation, instead of using negative gain, is it better to just leave things as-is at your (highest) base-iso, and change 'exposure' (gain) in post???
The goal is to get your waveform above the noise floor as much as possible without clipping the highlights. Negative gain is just a way to lower the noise floor (at the cost of DR in the highlights). But I wouldn't bother with that if you have a modern-ish camera (say 10 bit and ~12 stops of DR). Everything else depends on your camera and scene!
@@shagral I have this question too, in your situration, negative gain lower the noise floor, but at the same time, the brightness of shooting subject lower too, which means the signel and the data we want, also lower and still within the noise floor, how do we get benefit frow negative gain in low light situration?
And thank you so much for the video. I am enlightened!
@johnsek4294 low light doesn't mean no light. If we are talking about cinematic exposure (meaning you have some control over light) then you will have to pour more light on your subject to compensate for lower ISO. But you shadows will be cleaner.
If it's a total run and gun and you are shooting a black dog on a moonless night then don’t shoot in LOG!
I'm studying filmmaking and recently bought a sony a6600, but i just can't find what the native ISO is, if you google it it just tells you the recomended range wich is completely useless (it's 100 - 32,000)
Quick google search says it's a dual gain camera with low base at 100 ISO and high base at 400 ISO. Seems about right to me. Source: cuchara.photography/2019/06/iso-invariance-and-the-sony-a6500/
I recommend watching this entire video at least 2 times.
I fully support this recommendation.