Speaking of organic noise from digital cameras, I was blown away when I learned that The Holdovers was shot digitally. Apparently, they shot the Alexa at 1280 ISO to add natural texture in camera, before hitting it with heavy film emulation in post (with added grain/halation/gate weave). I was pretty much convinced that it was shot on specially made film stocks from the 70s, but it wasn't!
I did the same thing with a short I shot on the Mini LF. I pushed it to 1600. It’s Black & White, super high contrast. The noise pattern looks great. 3200 on the Mini LF looks like 16mm
Bro, holy shit! I've been taking pictures for a year with always 200 ISO. Just tried this on my GX80 and my images are not only more sharp but has a little more life. Thank you so much!
This is crazy. I’ve refused to watch this video for a few weeks now but I thought for sure this was clickbait. But you’re right, I’ve actually been doing this wrong 😂😂😂 thanks
The bit about 'gain' you moved through quite quickly is actually critical as it explains much more literally what's happening at the sensor level. This is because each camera actually only has one ISO rating, you mention it as the 'base' iso, anything above or below that is gain added to the sensor signal /after/ the photo diode well, so it doesn't actually make the sensor more sensitive to light. It just takes that same sensor signal and add signal gain down the image pipeline. Think of it like gain in audio. This is why the 'rated' iso for a camera system often has an equal distribution of usable stops above below the 0EV line, and 'iso' ratings above and below that native sensitivity don't incrase the total amount of usable stops (which WOULD be an increase in sensitivity) but instead just shift where those stops of range are assigned, giving you more available stops in the highlights or more in the shadows, but not both. This is just like adding 'gain' in audio (or in colour finishing for that matter). You can't add gain to find sound that wasn't sensed in the first place by the recorder, you can only move the whole measurement up (and the noise floor with it), same thing with light. The reality is that our hardware digital sensors can actually record way more light than a single ADC can capture and convert into signal, that's why ARRI run dual 14bit ADCs, to increase the range of signal the recorded file is able to hold at any one time without clipping off the top/bottom values.
while i do agree with most of your points, I do want to point out certain architectures like the Exmor in the Sony Venice line do have two sets of metal wirings for each of its native EI, which effectively shifts the curve at no cost to signal to noise ratio. This leads to my disagreement with your approach of the information, because I don't think it is necessary to go into electrical engineering under a video that is meant for general consumption. there is an old training video for engineer students where the presenter is just saying technical sounding nonsense and is expecting the audience to follow. sure, what you are saying might sound sane to your peers, but that belongs in a white paper, not in an educational youtube video.
Different gain circuits certainly do give the camera a second or even third base ‘rating’, not to mention DGO sensors like the Alexa ALEV mentioned. My comment was more trying to cover the fundamental idea of what iso adjustments are actually doing (adding or subtracting signal gain to a given base sensitivity). In terms of the information though, you’d be surprised. This isn’t exactly a complicated part of image sensor technology nor is this a general audience. The information in technology white papers are exactly what Directors of Photography want to know so that they are informed about what they’re actually operating with, rather than get misled by clever marketing. It’s certainly not just something relevant only to engineers.
An increase in amout of stops usable is not an increase in sensitivity. It's an increase in sensor dynamic. A real increase in sensor sensitvity just would just still have an equal distribution of usable stops below and above the 0ev line.
@@PierrePotonnier Technically speaking 'sensitivity' is how low of a photon count the diode wells can deliver before introducing signal error (which appears visually as spatial/chroma noise). As sensors improve they often extend the minimum usable signal and more importantly /how/ the image behaves as it approaches that minimum signal level. Same applies for the maximum photon count the diode can handle before it just sends a clipped signal. So sensitivity and dynamic range are interlinked concepts. In theory, if you had a magic imaging sensor capable of capturing scene light levels all the way from 0.001lux to 50,000 lux and encode those values in a 32bit float format, you wouldn't need to change your ISO rating at all, it would become irrelevant. But for now, it is necessary when recording to 8 and 10 bit codecs as some sort of OETF needs to be used to decide on how much of the signal will be used to transform the recorded light down into the electrical signal and then down to the much smaller (in range) encoded value.
I shot a video about this exact same topic that I just haven't got around to finishing. What you're talking about here is SO crucial for run and gun type filmmakers and I'm not sure why it's not talked about more on RUclips. This adds to a handful of videos I've seen discussing it. RUclipsrs always say shoot everything at the camera's native ISO for the 'most dynamic range' and that is just so wrong. I would have saved myself a ton of unuseable footage if I had just known this earlier. I shot a short documentary last year and I got some of my best looking footage ever. Deep shadows with Zero noise. Completely blew my mind
The one negative of noise is digital compression. With photography I actually love the look of noise in digital photos but then I get to print them with that noise untouched or view them as uncompressed image formats. Video sadly rarely gets to do that and how most people end up watching it is with highly compressed formats. I once read that Netflix has a compression process where they remove the noise, stream the video and the decoder adds noise back into the image as a way to get around how digital compression destroys noise and has to work much harder to compress images with noise. Ingenious way to get around that problem.
I was expecting the usual 'the right iso' video, but instead you got a new subscriber! Very well put I would say! I would also add that there's also people shooting things outside of well lit sets, I find myself very often shooting wildlife, so very long zooms, high F stops, on a Micro 4/3 sensor because of the reach and the convenience, most likely while being in a dark forest 😅. I have to use different shutter speeds and iso than the 'correct' ones because otherwise it would be better if I stayed at home 😂
You are correct on so many points but most non pros do not have time or the ability to add lighting to a scene. Even pros sometimes don’t have time to add light to a scene.
I can think of two additional reasons as to why you want noise added back to your image 1. It reduces H.264 compression artifacts in places that have gradients, for example the wall in your left(screen right),. 2. It reduces stepping in color gradients that occur when your high bitrate image gets inevitably crushed to 8-bit, causing color steps to become more visible. DIgital video only looks clean when the bitrate is absurdly high.
If you want a grain look in post you can get a better look by adding the noise or grain as an overlaid layer then editing the blur and opacity to those layers. Using that method you can get a grain look that will be distributed throughout the full tonal range from low to high. A much better look.
Great video Mitchell! More and more I find I'm having to rework my mind when it comes to how I should observe/handle ISO. This is not only a great reminder but an addition to how I think I'll view ISO moving forward. Grain can be good!
With Sony cameras in EI mode ,EI ( exposure index ) its not ISO , your recording will always be in one of the two base ISO,s. EI is just like rating a film stock different than whats on the can , that is setting a light meter to 250 ISO on 500ISO rate film stock . Sony EI works like this too. If your base ISO is set to 800 , and you set the EI to 400, viewing with a MLUT , your image will get 1 stop darker from the correct exposure , so add light or open up 1 stop, you are intentionally setting a uniformed 1 stop over exposure off set , because your recorded image will still be 800ISO, and will look stop over exposed in Slog play back. So it's a bit tricky and counter intuitive , raise the EI , and you will get way under exposed recorded footage . Viewing without a LUT you will see no change in the image at all ,changing the EI.
Even in the days when we were shooting 16mm film, using an ASA 100 stock would be better at making the blacks look black. Shooting with ASA 400 Tri-X tended to cause blacks to look gray because it was drawing in more light.
I agree with the concept of adding noise but you explained the exposure physics right. I'm a computer engineer, it's more technical. The ISO is how you set the range of the data captured by the bit depth of the camera, for example 8-bit, 10-bit, HLG, SLog3, etc. When you are adjusting the gain, you are setting the minimum and maximum range of the light you're capturing, which you can refer to as clipping the light on the sensor and normalizing (i.e. setting from 0-100%) to the bit depth. You have have heard the terms Expose to the Left/for the shadows, and Expose to the Right/for the highlights, and expose for the center/shadow and highlights. The light that is coming in is in a certain number range, and you need to scale that range so that you capture the dynamic range that you need. You do this with a histogram. You will see in the histogram that the light coming in at the different frequency ranges, and you need to get all of the information into the histogram that you want to capture. You can adjust your exposure compensation, and this will shift the light pixel data to the left or right of the histogram, hence the name expose to the left or right. If you are using 8-bit SLog3 in sunlight, you will either have to expose to the left or right with SLog3, or turn of log and expose to the center, but you will not be able to recover shadows or highlights. You are using the gain to set the range of the 8-bit and 10-bit color data, which means each pixel is either 24-bits (8-bits times 3) or 30-bits (10 bits times 3), but they pack that up in YUV format to get 4:2:2 or 4:2:0 subsampling, which is confusing, to do the Hybrid Log Gamma and SLog2/SLog3/CLog/FLog/etc All of them are variations on HLG, which is part of the HDMI standard but none of the other brands log curves are HDMI standards.
Hi, Cale, can I bother you with a technical question given that you seem to know about this stuff? Do histograms we see in the camera display the RAW data (the RAW histogram?) or the histogram of a JPEG? Once we get to processing software, does that change?
@@davidturner5 I would assume it's coming in raw. The white balance only adds and subtracts rbg values (usually HSV to RGB but Mokose uses RGB). The sensor is RGB, and the value is 0 to 100%.
@@CaleMcCollough Groovy, thanks. As I recall, the image on the back of the camera is a JPG, which makes sense given that RAW data can't be displayed "raw" and *some* sort of display/output decisions must be made in order to show it there. I suppose I wondered whether the histogram reflected that in-camera JPG conversion, but I can see now where my intuitions might have misled me.
@@davidturner5 don't take my word for it, but back many years ago I remember that it was JPEG in fact (for the on-camera viewing), which is why you could always push the histogram a bit more. That does change when you work with RAW files in whatever your choice of software is. It differs even from RAW software to other RAW software because there is always a default tone curve and color profile applied, affecting the image and its histogram. That may have changed though as a lot of time has passed since. I am too lazy to google that out right now, but I might do that some of these days.
I shoot on Fuji cameras at 640 ISO. The resulting image is clear and has a gritty filmic look to it. No one can usually tell it's a pro-sumer grade camera vs like an Alexa and some people even think I shot things on film. I started out in 35mm, and really didn't like the newer clean clinical look of digital. It has its place like in nature films and other areas for sure.
When I color for people, I often get questions of whether they should target a certain ISO and over/under exposure level. I've simplified it to "your camera has one ISO its always recording at no matter the setting (or two if its dual native) and one white balance. Everything other ISO is a level on the volume nob. If it's not well exposed at native, it's not well exposed at all."
@@Lathrop_Media what do you mean? Outside of native ADC reading, everything else is gain functions. Even internally, the camera will just perform analog or digital gains on the sensor info when shifting ISOs or white balance. Peak pixel well saturation is fixed and sensor read noise level is fixed, so your floor and ceiling is static, assuming the ADC bit depth doesn’t cut off information. Outside of noise reduction (a lot of Sony cameras do pre and post ADC noise reduction), nothing changes that floor or ceiling.
I really enjoyed this video. As a Videgrapher I’m trying to level myself up and just purchased the FX3. I’m really obsessed with light and how that plays an integral role in my story telling. I did this after my RUclips video colours and lighting were just waaaay off. I’ve been practicing and this knowledge you shared helped someone like me more than you know. It might be common knowledge but that you. I had no idea.
I just got an FX3 myself . Am I crazy but the HD in this camera really rocks I can blow it up 200% and the eyes and face in the picture are still crisp and in focus 😮. I’m finding using the base ISO is not always the best option.
@@wotiluv really funny you say that! I am working on a video and I was saying out load how insane the zoom capabilities are! Ill upload it around tomorrow so stay on the look out. See if you can pick out the really zoomed shots. I was astonished! But I did that at 4k. I am really happy with this camera. I love it soooo sooo much!
Noise used to be a huge problem back when cameras were mostly using compressed codecs. But now with less compression and even raw internally in some hybrid cameras now I don't think it's as much of an issue.
ISO is like the volume control on audio amplifier which boosts whatever voltage the light hitting the photo sensor sites creates. Those sensor sites are like buckets with have finite volume and the dilemma of digital exposure under high contrast lighting is that the ‘buckets’ in the highlights get filled to the brim and overflow (start to clip) before those in the deepest shadows get enough light to create voltage (signal). When you don’t have signal recording it the shadow tones and then amplify the signal in post processing to lift them all that is there to amplify is the noise, because their is no signal. The solution is simple LEARN TO USE FILL LIGHT to match the dynamic range of the scene to the dynamic range range of the sensor. What most still and video photographer do is start with the desired aperture for DOF, then adjust key light for properly exposed highlights on white objects and skin based on the clipping warning in the camera and then add fill as the last step. It is actually better to add the Fill source first so what is being Filled to the point of recording a signal in the darkest content can be easily seen. I’m primarily a still photographer who started shooting portraits in 1970 guided by a Kodak book on Portrait lighting which suggested placing a fill source directly behind and above the camera where it would reached everywhere the camera lens saw and cast it’s shadow down and directly behind the subject out of sight. Following that advice I never took a portrait that didn’t have a full range of tone and detail in the darkest clothing. In 1972 I got a job assisting top wedding photographer Monte Zucker who I discovered use the same centered fill technique for every flash photo he took, keeping a flash on a bracket centered over the lens at all times for Fill for a second flash used off camera to create the highlight pattern. When using film I used both incident and a 1° spot meter for setting lights and exposure but starting in 2000 when I bought my first digital camera I realized it was much simpler to just drape white and black wash cloths over a gray card (for WB reference) and set the lights visually based on detail seen in the playback and the clipping warning: 1) Set aperture for desired DOF, shutter for sync, ISO low (sensor baseline) 2) Turn on centered fill and raise until detail is seen in the black target. 3) Turn on back RIM light(s) and raise until it is 1/3 stop below clipping where it hits the white towel target. 4) Turn on frontal Key light and raised until the highlights it creates on the front of the white towel target are slightly darker than the rim-lit non specular white highlights (i.e., 2/3 below triggering clipping warning). 5) Turn on background lighting and raise until black target placed there has detail to avoid noise there due to fill light inverse-square fall off. If you use centered fill with that workflow you will never have a problem with noise in the shadows. If fill isn’t centered it will created shadows the camera sees which because they are not fill will always have no signal and only noise when amplified to try to lighten them. If the lighting pattern starts with a foundation of even fill there is no need to try to lift shadows which have no signal because signal above noise will be recorded everywhere because scene range matches range of camera sensor perfectly 😊
cameras like arri alexa been made to imitate the film look from the very beginning afaik, but it's a quite different story for median mirrorless consumer cameras that more often than not doesn't have any 'organic' noise to it (except maybe for a few models with a huge asterisk)
Some good stuff here. Important to note though that how ISO changes above/below middleG distribution changes depending on the camera/brand, etc. you're using. Clog2 on C200 for instance caps the distribution at ISO800 and allows you to go below 800 and distribute down into the shadows but going above 800 doesn't actually shift into the highlights but it remains flat all the way up, only increasing the gain. So you only have 800 and below on that camera to redistribute if you want vs. other cams which tend to move the distribution above/below middleG up or down. Anyways, always good to test this stuff and understand how one cam responds under what circumstances. Noise doesn't bug me at all really, compression making noise look worse is sometimes the culprit or why I think "yuck." A little chroma noise reduction can go a long way too vs. trying to scrub the noise out.
@@mitchellbrinkerdp Yea it really depends on the cam. The gain situation on C200 was also why most people found themselves frustrated by the noise at the supposed base 800, because it already had 12db of gain, while ISO200/400 was 0db and 6db. Google "clog2 iso eoshd" in Google Images and you should find the chart.
All this information is correct. Nice work. ISO is not about how good or bad a camera is … it’s about how good or bad a cinematographer is. The original ARRI Alexa at HD is still the best image for that film look.
Isn't it nice how you talked about that dark scene in the end which was the peoples favourite? Now imagine if you included footage or a photo of it in the video so we could see better what you meant with that. (: Definitely something I love on camera related videos 😵💫
In the end if you have no extra light you only can bump up the ISO and usually better to over expose that higher ISO in order to crush it in post. If you even try to overexpose in post slightly the grain goes crazy.
EI (Exposure Index) and ISO are different things! Changing ISO by gain up or down lowers the DR. Cameras with EI (shoots at base iso) or shot in RAW format (amplification done in post, in camera only metadata changes), such as arri or venice, dislocates the DR. You are talking about "shoring to the left" that is super valid, but be careful on how to do that on mirrorless style cameras. You can create fake EI with luts with lower stops or in this case more exposure.
Correct me if I’m wrong but 500T and 50D are referencing the color temperature and lighting, I.E 50 Day light with the ASA being around 250 or near 250-300 ISO in equivalent
best iso video ive ever seen! You got a follower more! Now i finallyy understand iso... photography (and hopefully cinematocgraphy in the future) will step up now. Thanks man!
I love the look of this video specifically. I always try to replicate this kind of lighting and feel but I can't quite get it down, my footage always ends up looking either too cold or too warm.
Hopefully, I can shed some light on what's actually going on with that ARRI chart Mitchell is referencing (haha, sorry for the pun, I couldn't resist!) The "shifting" in dynamic range that you are seeing when the ISO is changed in the ARRI chart is absolutely, 100% NOT what happens when ISO is adjusted in normal camera operation. ARRI is actually applying very special parameters in-camera that do not occur in normal camera operations. The point is, you don't have to worry about using your ISO the wrong way, unless you are shooting video with an ARRI Alexa, or other certain camera models in certain shooting modes. Nothing has changed in how we understand ISO and how it affects image quality. For most of us, changing your ISO will not cause your camera's dynamic range to start drastically shifting up and down in multiple stops. SHORT VERSION: This is a technology that a lot of camera manufacturers are using nowadays mainly for videographers that require maximum image quality, flexibility in post-processing and tonal range. Also some photography applications using a menu setting to activate. What you're seeing is the result of corrections happening in-camera in real time during shooting using sensor technology and in-camera software from manufacturers. The bottom line is there will not be these large shifts in dynamic range performance by simply using different ISO's in typical shooting for most photographers; unless you are using particular cameras and particular shooting modes for specific applications. LONG VERSION: There is a very important reason why the dynamic range "shifts" when the ISO changes. Notice how ARRI's total dynamic range is always 14.5 stops of light regardless of which ISO is selected, it is simply "shifting" up or down. It has to do with what is actually happening in-camera before raw files are even written to a card. Canon also does this with their HTP (Highlight Tone Priority) as well as Sony's DRO (Dynamic Range Optimizer). Which they offer as a menu feature you can turn on/off and was even in the old Canon Rebel's and Sony's A57 and NEX-7's from many years ago. The modern A7S series also uses a similar feature for "log" when shooting video, as well as many other manufacturers that are now using a "log" setting for videographers. ARRI also uses this method to help directors/cinematographers acheive maximum image quality in any lighting situation and give more flexibility in post-processing. They are all changing/moving the "base" ISO (i.e. removing the ability to select certain ISO's) in order to provide more stops of tonal range for a given lighting scenario. What is happening is in-camera parameters are applied to the exposure and curves, pre-RAW stage. Basically, when you want more stops of highlights and make a selection accordingly in-camera, the camera then intentionally underexposes the image during capture, then digitally "gains" the brightness back up to original exposure before sending the file to be written to a card. This is all happening in-camera during the imaging chain before the photographer ever sees the file. The underexposure is performed automatically by reducing sensor amplification (i.e. lowering ISO) while maintaining shutter and aperture selections. For example, if your camera's normal base ISO is 100, but your camera is using this mode, it will increase your minimum available ISO to 200. So if you are shooting at ISO200 @ 1/30sec. & f/2.8 in this mode, the camera will automaticaly reduce sensor amplification by 1-stop to ISO100 @ 1/30sec. & f/2.8. This will give you 1-stop of extra highlights. Then software in-camera gains the brightness back up exactly 1-stop to match your original selected exposure. Also, some curves are applied to help the highlights "roll-off" better. The downside is that underexposure causes a loss in detail in the shadows and more noise (less light = more noise). This is why the dynamic range appears to "shift" with less and less shadow information as you gain more and more highlight information. If you want better/cleaner shadows you can inversely lower or "pull" the base ISO below it's minimum by forcing the camera to overexpose, then pulling the exposure back down to it's original exposure. This gives more tonal range in the shadows with better detail and less noise, but at the cost of less detail in the highlights. Overexposure provides more shadow detail, and underexposure provides more highlight detail. It's always a trade off. This is why ARRI's chart @ 1:59 is always equal to exactly 14.5 stops and the dynamic range is only "shifting" up or down. DPReview gives an excellent explaination of this technology in one of their camera reviews: "Highlight Tone Priority option (Custom Function II.6) is a method for capturing more information in the brightest parts of the scene. It does this by applying less amplification to the signal coming from the sensor, then compensating for it by using a different tone curve to ensure the correct brightness in the final image." *reference: www.dpreview.com/reviews/canoneos600d/12 Canon's HTP, Sony's DRO, of course ARRI, and pretty much any company that offers a "log" video shooting mode with a "base" ISO of like 800 or even 3200 minimum ISO are basically all doing the same thing: overriding exposure by adjusting sensor amplification at image capture, then compensating back to original exposure with in-camera software corrections using curves to preserve tonal range. All of this is done in the blink of an eye, out-of-sight and out-of-mind of the photographer. This is the technology being using today. Shooting at high ISO's during the day and low ISO's at night is a great idea if you are shooting video with an ARRI Alexa, based on how they designed it to operate. But it is important to understand that this is NOT a representation of how ISO fundamentally works. This is a technology available in certain cameras, for certains applications, in certain shooting modes. What matters is the specific camera you are using and understanding how it works to get the best image quality possible out of your gear.
Only true uncompressed raw behive like this. Worked with many cameras. Only Alexa cameras, sigma fp and kinefinity cameras and cameras that shot true raw behive like this. They all shot 12 or 14bit raw in 4:4:4 color subsampling. On my sigma fp if I switch to 8bit h264 rules doesn't apply. Even in 10bit 4:2:2 gets tricky 😊
you also need to consider camera's that are effectively "ISOless" with iso invariance. effectively meaning shooting at lower iso's will preserve the highlights while also letting you boost the shadows without adding any more noise than if you shot for the shadows
Wow I had no idea about the more details in the shadows and highlights part of ISO. Thank YOU! Now I’m even less afraid of making a video in my room at 3200. Some people do that on purpose! 😂
Nice video man😃I have the fx30 with dual iso and “should” I still use the higher base ISO for outdoor and vice versa or does it not matter if they have 6 stops each direction either way? Also would’ve been cool to see the last clip you were talking about🙂
Another great video, anytime I see someone sharing the gospel on ISO I get feels. Only one note I would add... ISO does NOT change the sensitivity of a digital sensor, native or dual native sensor are fixed.... In general these "native" ISO's are where the camera manufacturer has best tuned for dynamic range and noise patterns... they are not actually more or less sensitive to light... Hence Gain... It's still the same process, ISO just became adopted after the DSLR revolution. Also The ISO is going to change from manufacturer to manufacture... DO NOT TRUST ISO to be accurate to film equivalents. Other then those points I totally agree, RUclips has had a massive impact on the way DP's (especially in the last decade and a half)... the obsession with clinical images seems largely of a byproduct of test footage/review culture... Film was never clean and honestly go back and look at something like Heroes/Lie To Me from the late 2000’s would be considered by many "unuseable"... an at the same time the photo chemical process had reached the ability to deliver film stock and printing techniques that would be considered sterile by many now days chasing the "film look". I think most most great DP's will tell you its about your choices, your process and if you are working with a DP, its about what the two of you are looking for... not about "The right way"... This also goes for the current ongoing obsession with "perfect soft light".... but that's another topic.
@@Vitaphone don’t even get me started on soft light, mostly because I heavily disagree with the over reliance on big soft light. it’s good to know how to achieve it, but it’s robbing you of the creative choices a DP has to make if you use it exclusively in every scenario. Either way, thanks for sharing. That’s important info. Anyone in these comments please reference above in addition to the video!
@@mitchellbrinkerdp would love to start seeing more working DP’s on the RUclips space opening up on the soft light/ “perfect exposure” topic… it’s been wild to see commercial and commercial photography lighting ideas dominate a lot of the conversation in the post RUclips era… Most DP’s I work with with years under their belt skoff at this, but newer generations seem obsessed with the RUclips meta.
Awesome video! Glad the algorithm brought me here. Thanks. Since you mentioned the FX3: do you have "favorite" ISO levels or recommendations for that sensor? Obviously this depends on the amount and distribution of light in each scene and the look one wants to achieve, but as you mentioned it for another camera, there might be a value that generally introduces a nice amount of noise in most scenarios. Like 640, 2500, or 10000 or 16000. I am basically always exclusively using the base ISOs when i'm working... I might need to change that. Surely I can just do some experimenting myself. Curious about your opinion though.
Why don‘t you just ETTR? If the DR shifts from more stops below average grey to more stops above but the total amount stays roughly the same, it shouldn‘t matter too which ISO you pick but rather how you expose, right? Maybe I am lacking the practical examples here.
Great video. Thank you. I love the way your video looks, by the way. Do you use some sort of cinema filter? (I’m a beginner, so I’m sorry if that’s not the question to ask.)
Came to this video late, so sorry for the delayed comment. ISO is absolutely not 'how sensitive your camera sensor is to light'. Your comment that you can change the sensor's sensitivity to light is mistaken (except, arguably, for sensors with dual conversion gain, but there it only changes at one point in the ISO range). The ISO standard for ISO speed ratings for digital cameras, ISO 12232, does not even mention the sensitivity of the sensor to light - so cannot define what you say it does. Nor does it define the gain, which wouldn't change sensitivity anyway. It's not 'technically true' that the higher you set the ISO the more noise it introduces. In fact, it's generally false. On most cameras, the higher you set the ISO the less noise is introduced. The noise is because a high ISO causes a low exposure (ISO is inversely proportional to exposure), which means the image is composed of fewer photons which means a lower shot noise SNR. I won't go through the rest of the video in detail, but my guess would be that when the fundamentals are wrong, the rest of it will likely be too. Sorry to be so critical, but when putting information out there you need to do your background research first and get it right.
As I’m watching this video I had a moment where I thought “this is exactly what RUclips was made for.” Simple but high-quality video filled with amazing info, re-watchable, and not even 10 mins long. Great work. Instant subscribe
I leave my A7C on Auto ISO. I've shot over 100 videos for 7 clients this year and have made $80k so far. Guess how many of my clients or the people commenting on the videos have brought up noise. (Hint: 0).
Does anybody have links to ISO distribution graphs for Sony mirrorless series? I'm using A7iii. Side note, your framing in this video is incredibly detailed! Very cool
Isn't the optimal approach to use one of your camera base ISO's (though the ridiculous 12,800 on the FX3 is actually noisier just because of quantum fluctuations- physics!) and use ND filters and/or lighting (or light control) to make that base ISO 'correct' ?
@@muuuuusic6269 definitely does, especially since the A7S3 and FX3 have the same sensor. If you go to the 12800 base iso, you won’t have anymore information in the highlights. You’ll have the same 6/8 in the highlights and shadows as you do at 800. If you raise the iso from either base, you will get those stops in the highlights.
this was an amazing video, my R6 has amazing dynamic range so I can shoot at 10000 ISO and theres only a little bit of grain. I did not know about shooting high iso in sunlight to get better highlights, thank you so much!
I’m sorry, but the part about using filters in a bright setting and augmenting ISOs is plain wrong. This will NOT give you more details. The point of ISOs is to "turn on the volume" on signal. So yes, in a dark setting, indeed increasing it won’t improve the ability of your camera to capture details in low light, hence the noise, but in bright light, the signal (as in meaningful color and brightness data) is already crispy clear and the signal-to-noise ratio is therefore optimal (as long of course as the whites aren’t burned). Using a filter and augmenting ISOs will achieve NOTHING for your signal-to-noise ratio, it can only dampen your signal and make you lose precious meaningful data... ISOs aren’t necessarily evil, they can be useful in a reasonably dark setting and the noise can be used in a artsy kind of way but if you’re looking for sharpness and clean signal, it should always be the last thing you turn up. There are literally zero scenario where ISOs augment signal-to-noise ratio, from a purely physics point of view. They just can’t do that.
@@r.c8756 nah man, i showed the graphs on how dynamic range is distributed across the EI range. this is true for alexa, sony venice, etc. the same principle can be used on other cameras such as the fx6 and fx3, as well as blackmagic. next time you are out, check false color in the highlights at 400 and go to 3200 and compensate. you will see the red become yellow and yellow become green.
@@mitchellbrinkerdp I just looked deeper into it. The thing is, it’s a feature that’s pretty much specific to high-end cinema cameras. I come from a photographic background so what I said is true for digital cameras but does not necessarily apply to cinema cameras. I apologize, I should have checked that before commenting.
*you are explaining ISO wrong.* Raising your ISO does *NOT* change the "sensitivity" of your sensor. When capturing an image, light enters the lens and hits your digital sensor. The sensor is made up of photodiodes that convert the photons (light) to an electrical signal. ISO amplifies the signal, similar to turning up the volume on a radio. You are digitally amplifying the signal, unlike film stocks where are actually more or less sensitive to light. You almost had it when you mentioned Gain. There is ALWAYS noise in digital imagery. It's about how what noise to signal ratio is acceptable. It's important to understand this so we're not spreading the same myths around the creator community. You may say, "Yes, either way, the answer at the end of the day is all about the light." But the reason WHY is very different. Raising your ISO on a theoretical sensor that is "sensitive" to light would actually "get you more light" without having to add light to the scene. But since it's an electrical amplification happening, that's where the noise comes from. If we have little to no light, and multiply it times 10, we still have dark pixels. And, THIS is where we come full circle and now know WHY we need to increase the *light* and not our *ISO.*
@@RunNGunPhoto The nitty gritty nuanced science of the sensor is great, but 99% of working DPs aren’t concerned with it. They’re concerned with how to get the image they want, and what the settings are doing to achieve that at the most streamlined level. Everything mentioned here about dynamic range in relation to the EI is the meat that everyone should know. And whether or not it’s actually “increasing sensitivity” or not, you can change your sensor to 12800 ISO and use a matchstick to get exposure with no loss of information if you want. Call it more sensitive or don’t, but that’s what’s important and if the easiest explanation for someone is that it makes it more sensitive then that’s fine. As long as there is an understanding of what that does to your image vs opening up your aperture, etc. The goal is to simplify for people, which is why I’m not getting into the deep inner working of science of what’s happening at the sensor level. There are plenty of DPs in the ASC who would simply say, “who cares, just know how to get the image you want.” It’s art, and you’re painting. It’s a lot about feel and understanding and taste. That’s my spiel to all the “uhm actually the sensor isn’t” comments.
@@mitchellbrinkerdp That was a really drawn out way to 1+1= whatever you feel like. I concisely explained *WHY* it's important to know how your sensor works. Please do your research before spreading misinformation. Making up your own "facts" for clicks on RUclips is pretty low-effort.
@@RunNGunPhoto i can promise you there are no made up facts or misinformation here. we are in completely different industries, and there are very big differences in how things are approached. whether or not “sensitivity” is the technically correct word doesn’t matter. if you pick up an Arri Alexa or Sony Venice down to an FX3 what I explained about how ISO affects your image is the case and all you need to know. ISO higher = more range in the highlights, less range in the shadows hence higher noise floor. ISO lower = more range in the shadows, less in the highlights. If you bump your ISO to 6400 in a dark room, of course there will be noise, it’s clearly not “creating” light. I said it doesn’t matter, if that’s what you want to do for the shot, then go for it. There are no rules.
@@mitchellbrinkerdp No this is an extremely convoluted and backwards way of saying "expose for the highlights." Which is also poor advice. I am a photographer and cinematographer. "Industry" isn't relevant. It's still bad advice.
@@RunNGunPhoto it’s not a convoluted way of saying expose for the highlights. i’m not saying anything about exposing for the highlights. that is completely different. i am talking about the way exposure index dictates the distribution of the dynamic range across the ASA range. i have also mentioned how the noise floor is affected with those changes in ASA and why. there is absolutely a difference in industry here. how things are done in the photography and videography world is not equal to how things are in the cinematography world. unless you have 9+ years of experience shooting on and working with arri cameras, i don’t know why you’re telling me what i am right and wrong about. this isn’t top 10 iphone cinematic tricks here, this is stuff you would be told by people at arri. im simply taking this and saying you can apply it to other cameras because they all fundamentally do the same thing in their own way.
So you are telling me if I want a very high key image, rather than shooting at iso 100 I should instead shoot at 3200 and add an ND and that will produce a better quality image?
@@tobiasyoder not necessarily. whether the image is high key or low key is going to come down to your lighting. changing the iso won’t change your lighting. but if you are in a very bright environment or somewhere with a lot of bright highlights, raising your iso and ND’ing down can help retain more information in the highlights.
I think you have to be careful when telling this to a beginner because I have mentored interns and co workers who don’t understand ISO and just wing it half of the time and their footage looks bland, over exposed or underexposed. I think it’s better teaching someone new that base ISO’s are a good starting point and should always be where you start and if you can always stay on those base ISO’s. I think what you’re teaching is an advance technique which is really cool and I think good but I hope it doesn’t misinform people who are just learning
Кино
at the end.. "im really runnning out of light"
just bump the iso bro xD
Speaking of organic noise from digital cameras, I was blown away when I learned that The Holdovers was shot digitally. Apparently, they shot the Alexa at 1280 ISO to add natural texture in camera, before hitting it with heavy film emulation in post (with added grain/halation/gate weave). I was pretty much convinced that it was shot on specially made film stocks from the 70s, but it wasn't!
I did the same thing with a short I shot on the Mini LF. I pushed it to 1600. It’s Black & White, super high contrast. The noise pattern looks great. 3200 on the Mini LF looks like 16mm
Love it when actual DP's who know what they're talking about chime in on youtube. There is so so so much misinformation coming from these video gurus
I have definitely learned to embrace camera noise over the years.
Bro, holy shit! I've been taking pictures for a year with always 200 ISO. Just tried this on my GX80 and my images are not only more sharp but has a little more life. Thank you so much!
This is crazy. I’ve refused to watch this video for a few weeks now but I thought for sure this was clickbait. But you’re right, I’ve actually been doing this wrong 😂😂😂 thanks
Great insight on the importance of intentionally using your tools. The BMPCC 4k is very interesting with its ISO chart
The bit about 'gain' you moved through quite quickly is actually critical as it explains much more literally what's happening at the sensor level. This is because each camera actually only has one ISO rating, you mention it as the 'base' iso, anything above or below that is gain added to the sensor signal /after/ the photo diode well, so it doesn't actually make the sensor more sensitive to light. It just takes that same sensor signal and add signal gain down the image pipeline. Think of it like gain in audio. This is why the 'rated' iso for a camera system often has an equal distribution of usable stops above below the 0EV line, and 'iso' ratings above and below that native sensitivity don't incrase the total amount of usable stops (which WOULD be an increase in sensitivity) but instead just shift where those stops of range are assigned, giving you more available stops in the highlights or more in the shadows, but not both. This is just like adding 'gain' in audio (or in colour finishing for that matter). You can't add gain to find sound that wasn't sensed in the first place by the recorder, you can only move the whole measurement up (and the noise floor with it), same thing with light. The reality is that our hardware digital sensors can actually record way more light than a single ADC can capture and convert into signal, that's why ARRI run dual 14bit ADCs, to increase the range of signal the recorded file is able to hold at any one time without clipping off the top/bottom values.
while i do agree with most of your points, I do want to point out certain architectures like the Exmor in the Sony Venice line do have two sets of metal wirings for each of its native EI, which effectively shifts the curve at no cost to signal to noise ratio. This leads to my disagreement with your approach of the information, because I don't think it is necessary to go into electrical engineering under a video that is meant for general consumption. there is an old training video for engineer students where the presenter is just saying technical sounding nonsense and is expecting the audience to follow. sure, what you are saying might sound sane to your peers, but that belongs in a white paper, not in an educational youtube video.
Different gain circuits certainly do give the camera a second or even third base ‘rating’, not to mention DGO sensors like the Alexa ALEV mentioned. My comment was more trying to cover the fundamental idea of what iso adjustments are actually doing (adding or subtracting signal gain to a given base sensitivity). In terms of the information though, you’d be surprised. This isn’t exactly a complicated part of image sensor technology nor is this a general audience. The information in technology white papers are exactly what Directors of Photography want to know so that they are informed about what they’re actually operating with, rather than get misled by clever marketing. It’s certainly not just something relevant only to engineers.
An increase in amout of stops usable is not an increase in sensitivity. It's an increase in sensor dynamic.
A real increase in sensor sensitvity just would just still have an equal distribution of usable stops below and above the 0ev line.
@@PierrePotonnier Technically speaking 'sensitivity' is how low of a photon count the diode wells can deliver before introducing signal error (which appears visually as spatial/chroma noise). As sensors improve they often extend the minimum usable signal and more importantly /how/ the image behaves as it approaches that minimum signal level. Same applies for the maximum photon count the diode can handle before it just sends a clipped signal. So sensitivity and dynamic range are interlinked concepts.
In theory, if you had a magic imaging sensor capable of capturing scene light levels all the way from 0.001lux to 50,000 lux and encode those values in a 32bit float format, you wouldn't need to change your ISO rating at all, it would become irrelevant. But for now, it is necessary when recording to 8 and 10 bit codecs as some sort of OETF needs to be used to decide on how much of the signal will be used to transform the recorded light down into the electrical signal and then down to the much smaller (in range) encoded value.
@@_burd yes and?
I shot a video about this exact same topic that I just haven't got around to finishing. What you're talking about here is SO crucial for run and gun type filmmakers and I'm not sure why it's not talked about more on RUclips. This adds to a handful of videos I've seen discussing it. RUclipsrs always say shoot everything at the camera's native ISO for the 'most dynamic range' and that is just so wrong. I would have saved myself a ton of unuseable footage if I had just known this earlier. I shot a short documentary last year and I got some of my best looking footage ever. Deep shadows with Zero noise. Completely blew my mind
Hurry up and get it done! You need to post some videos, bro!
Ya please post soon!
The one negative of noise is digital compression. With photography I actually love the look of noise in digital photos but then I get to print them with that noise untouched or view them as uncompressed image formats.
Video sadly rarely gets to do that and how most people end up watching it is with highly compressed formats. I once read that Netflix has a compression process where they remove the noise, stream the video and the decoder adds noise back into the image as a way to get around how digital compression destroys noise and has to work much harder to compress images with noise. Ingenious way to get around that problem.
I was expecting the usual 'the right iso' video, but instead you got a new subscriber! Very well put I would say!
I would also add that there's also people shooting things outside of well lit sets, I find myself very often shooting wildlife, so very long zooms, high F stops, on a Micro 4/3 sensor because of the reach and the convenience, most likely while being in a dark forest 😅. I have to use different shutter speeds and iso than the 'correct' ones because otherwise it would be better if I stayed at home 😂
You are correct on so many points but most non pros do not have time or the ability to add lighting to a scene. Even pros sometimes don’t have time to add light to a scene.
I can think of two additional reasons as to why you want noise added back to your image
1. It reduces H.264 compression artifacts in places that have gradients, for example the wall in your left(screen right),.
2. It reduces stepping in color gradients that occur when your high bitrate image gets inevitably crushed to 8-bit, causing color steps to become more visible. DIgital video only looks clean when the bitrate is absurdly high.
Some might use their ISO wrong, but you are using your lavalier microphone wrong.
@@DutchDiederik nah I prefer holding it
I'm not even a photographer nor filmographer, but this video was super interesting and very well done, good job man.
If you want a grain look in post you can get a better look by adding the noise or grain as an overlaid layer then editing the blur and opacity to those layers. Using that method you can get a grain look that will be distributed throughout the full tonal range from low to high. A much better look.
I loveee your channel I’m learning so much!! Thank you for creating content to share your knowledge!!!
Great video Mitchell! More and more I find I'm having to rework my mind when it comes to how I should observe/handle ISO. This is not only a great reminder but an addition to how I think I'll view ISO moving forward. Grain can be good!
thanks for the clear, concise, informative video without a ton of added fluff / filler! learned a lot
With Sony cameras in EI mode ,EI ( exposure index ) its not ISO , your recording will always be in one of the two base ISO,s. EI is just like rating a film stock different than whats on the can , that is setting a light meter to 250 ISO on 500ISO rate film stock . Sony EI works like this too. If your base ISO is set to 800 , and you set the EI to 400, viewing with a MLUT , your image will get 1 stop darker from the correct exposure , so add light or open up 1 stop, you are intentionally setting a uniformed 1 stop over exposure off set , because your recorded image will still be 800ISO, and will look stop over exposed in Slog play back. So it's a bit tricky and counter intuitive , raise the EI , and you will get way under exposed recorded footage . Viewing without a LUT you will see no change in the image at all ,changing the EI.
Even in the days when we were shooting 16mm film, using an ASA 100 stock would be better at making the blacks look black. Shooting with ASA 400 Tri-X tended to cause blacks to look gray because it was drawing in more light.
Related to texture, I thought that Anatomy of a Fall had a cool and pretty simple process where they used super 35 lenses on the mini LF sensor.
I agree with the concept of adding noise but you explained the exposure physics right. I'm a computer engineer, it's more technical. The ISO is how you set the range of the data captured by the bit depth of the camera, for example 8-bit, 10-bit, HLG, SLog3, etc. When you are adjusting the gain, you are setting the minimum and maximum range of the light you're capturing, which you can refer to as clipping the light on the sensor and normalizing (i.e. setting from 0-100%) to the bit depth. You have have heard the terms Expose to the Left/for the shadows, and Expose to the Right/for the highlights, and expose for the center/shadow and highlights. The light that is coming in is in a certain number range, and you need to scale that range so that you capture the dynamic range that you need. You do this with a histogram. You will see in the histogram that the light coming in at the different frequency ranges, and you need to get all of the information into the histogram that you want to capture. You can adjust your exposure compensation, and this will shift the light pixel data to the left or right of the histogram, hence the name expose to the left or right. If you are using 8-bit SLog3 in sunlight, you will either have to expose to the left or right with SLog3, or turn of log and expose to the center, but you will not be able to recover shadows or highlights. You are using the gain to set the range of the 8-bit and 10-bit color data, which means each pixel is either 24-bits (8-bits times 3) or 30-bits (10 bits times 3), but they pack that up in YUV format to get 4:2:2 or 4:2:0 subsampling, which is confusing, to do the Hybrid Log Gamma and SLog2/SLog3/CLog/FLog/etc All of them are variations on HLG, which is part of the HDMI standard but none of the other brands log curves are HDMI standards.
Hi, Cale, can I bother you with a technical question given that you seem to know about this stuff?
Do histograms we see in the camera display the RAW data (the RAW histogram?) or the histogram of a JPEG? Once we get to processing software, does that change?
@@davidturner5 I would assume it's coming in raw. The white balance only adds and subtracts rbg values (usually HSV to RGB but Mokose uses RGB). The sensor is RGB, and the value is 0 to 100%.
@@CaleMcCollough Groovy, thanks. As I recall, the image on the back of the camera is a JPG, which makes sense given that RAW data can't be displayed "raw" and *some* sort of display/output decisions must be made in order to show it there. I suppose I wondered whether the histogram reflected that in-camera JPG conversion, but I can see now where my intuitions might have misled me.
@@davidturner5 don't take my word for it, but back many years ago I remember that it was JPEG in fact (for the on-camera viewing), which is why you could always push the histogram a bit more. That does change when you work with RAW files in whatever your choice of software is. It differs even from RAW software to other RAW software because there is always a default tone curve and color profile applied, affecting the image and its histogram.
That may have changed though as a lot of time has passed since. I am too lazy to google that out right now, but I might do that some of these days.
I shoot on Fuji cameras at 640 ISO. The resulting image is clear and has a gritty filmic look to it. No one can usually tell it's a pro-sumer grade camera vs like an Alexa and some people even think I shot things on film. I started out in 35mm, and really didn't like the newer clean clinical look of digital. It has its place like in nature films and other areas for sure.
When I color for people, I often get questions of whether they should target a certain ISO and over/under exposure level.
I've simplified it to "your camera has one ISO its always recording at no matter the setting (or two if its dual native) and one white balance. Everything other ISO is a level on the volume nob. If it's not well exposed at native, it's not well exposed at all."
Not exactly, but better than overthinking it!
@@Lathrop_Media what do you mean? Outside of native ADC reading, everything else is gain functions. Even internally, the camera will just perform analog or digital gains on the sensor info when shifting ISOs or white balance. Peak pixel well saturation is fixed and sensor read noise level is fixed, so your floor and ceiling is static, assuming the ADC bit depth doesn’t cut off information. Outside of noise reduction (a lot of Sony cameras do pre and post ADC noise reduction), nothing changes that floor or ceiling.
I really enjoyed this video. As a Videgrapher I’m trying to level myself up and just purchased the FX3. I’m really obsessed with light and how that plays an integral role in my story telling. I did this after my RUclips video colours and lighting were just waaaay off. I’ve been practicing and this knowledge you shared helped someone like me more than you know. It might be common knowledge but that you. I had no idea.
I just got an FX3 myself . Am I crazy but the HD in this camera really rocks I can blow it up 200% and the eyes and face in the picture are still crisp and in focus 😮. I’m finding using the base ISO is not always the best option.
@@wotiluv really funny you say that! I am working on a video and I was saying out load how insane the zoom capabilities are! Ill upload it around tomorrow so stay on the look out. See if you can pick out the really zoomed shots. I was astonished! But I did that at 4k. I am really happy with this camera. I love it soooo sooo much!
@@32SIX I can’t wait to see it lmk when it’s up
Great video Mitchell! It would interesting to see you do a noise pattern comparison video.
Tysm for this. I often second guess whether I should change my ISO or aperture for my shots
I like the way you covered this without being condescending!
Noise used to be a huge problem back when cameras were mostly using compressed codecs. But now with less compression and even raw internally in some hybrid cameras now I don't think it's as much of an issue.
ISO is like the volume control on audio amplifier which boosts whatever voltage the light hitting the photo sensor sites creates. Those sensor sites are like buckets with have finite volume and the dilemma of digital exposure under high contrast lighting is that the ‘buckets’ in the highlights get filled to the brim and overflow (start to clip) before those in the deepest shadows get enough light to create voltage (signal).
When you don’t have signal recording it the shadow tones and then amplify the signal in post processing to lift them all that is there to amplify is the noise, because their is no signal. The solution is simple LEARN TO USE FILL LIGHT to match the dynamic range of the scene to the dynamic range range of the sensor.
What most still and video photographer do is start with the desired aperture for DOF, then adjust key light for properly exposed highlights on white objects and skin based on the clipping warning in the camera and then add fill as the last step.
It is actually better to add the Fill source first so what is being Filled to the point of recording a signal in the darkest content can be easily seen. I’m primarily a still photographer who started shooting portraits in 1970 guided by a Kodak book on Portrait lighting which suggested placing a fill source directly behind and above the camera where it would reached everywhere the camera lens saw and cast it’s shadow down and directly behind the subject out of sight. Following that advice I never took a portrait that didn’t have a full range of tone and detail in the darkest clothing. In 1972 I got a job assisting top wedding photographer Monte Zucker who I discovered use the same centered fill technique for every flash photo he took, keeping a flash on a bracket centered over the lens at all times for Fill for a second flash used off camera to create the highlight pattern.
When using film I used both incident and a 1° spot meter for setting lights and exposure but starting in 2000 when I bought my first digital camera I realized it was much simpler to just drape white and black wash cloths over a gray card (for WB reference) and set the lights visually based on detail seen in the playback and the clipping warning:
1) Set aperture for desired DOF, shutter for sync, ISO low (sensor baseline)
2) Turn on centered fill and raise until detail is seen in the black target.
3) Turn on back RIM light(s) and raise until it is 1/3 stop below clipping where it hits the white towel target.
4) Turn on frontal Key light and raised until the highlights it creates on the front of the white towel target are slightly darker than the rim-lit non specular white highlights (i.e., 2/3 below triggering clipping warning).
5) Turn on background lighting and raise until black target placed there has detail to avoid noise there due to fill light inverse-square fall off.
If you use centered fill with that workflow you will never have a problem with noise in the shadows. If fill isn’t centered it will created shadows the camera sees which because they are not fill will always have no signal and only noise when amplified to try to lighten them. If the lighting pattern starts with a foundation of even fill there is no need to try to lift shadows which have no signal because signal above noise will be recorded everywhere because scene range matches range of camera sensor perfectly 😊
It did! also that Peanuts short is so rad. Would love to see a bts on that. As always, thanks for the good shit!
cameras like arri alexa been made to imitate the film look from the very beginning afaik, but it's a quite different story for median mirrorless consumer cameras that more often than not doesn't have any 'organic' noise to it (except maybe for a few models with a huge asterisk)
As of now 430 views, I hope it get's plenty of traction. Very valuable information that would help lot's of people.
Some good stuff here. Important to note though that how ISO changes above/below middleG distribution changes depending on the camera/brand, etc. you're using. Clog2 on C200 for instance caps the distribution at ISO800 and allows you to go below 800 and distribute down into the shadows but going above 800 doesn't actually shift into the highlights but it remains flat all the way up, only increasing the gain. So you only have 800 and below on that camera to redistribute if you want vs. other cams which tend to move the distribution above/below middleG up or down. Anyways, always good to test this stuff and understand how one cam responds under what circumstances. Noise doesn't bug me at all really, compression making noise look worse is sometimes the culprit or why I think "yuck." A little chroma noise reduction can go a long way too vs. trying to scrub the noise out.
@@KyleProhaska this is interesting. I didn’t know that. Canons always out here being…. Fine
@@mitchellbrinkerdp Yea it really depends on the cam. The gain situation on C200 was also why most people found themselves frustrated by the noise at the supposed base 800, because it already had 12db of gain, while ISO200/400 was 0db and 6db. Google "clog2 iso eoshd" in Google Images and you should find the chart.
All this information is correct. Nice work. ISO is not about how good or bad a camera is … it’s about how good or bad a cinematographer is. The original ARRI Alexa at HD is still the best image for that film look.
This was really insightful! I have some testing to do. Thx brother!
my life changed when I discovered shooting pictures in bright daylight at 800 iso using the dr400 setting on my X100F
Isn't it nice how you talked about that dark scene in the end which was the peoples favourite? Now imagine if you included footage or a photo of it in the video so we could see better what you meant with that. (:
Definitely something I love on camera related videos 😵💫
@@StefanCH1999 yeah I just don’t have the footage anymore
In the end if you have no extra light you only can bump up the ISO and usually better to over expose that higher ISO in order to crush it in post. If you even try to overexpose in post slightly the grain goes crazy.
Accurate however, t's important to note that this is NOT true for raw photos
Good points. I never thought of using the ISO to emulate a film look. Sounds worth experimenting with. Thanks. New subscriber.
EI (Exposure Index) and ISO are different things! Changing ISO by gain up or down lowers the DR. Cameras with EI (shoots at base iso) or shot in RAW format (amplification done in post, in camera only metadata changes), such as arri or venice, dislocates the DR. You are talking about "shoring to the left" that is super valid, but be careful on how to do that on mirrorless style cameras. You can create fake EI with luts with lower stops or in this case more exposure.
Correct me if I’m wrong but 500T and 50D are referencing the color temperature and lighting, I.E 50 Day light with the ASA being around 250 or near 250-300 ISO in equivalent
You're correct.
man this makes so much sense haha, thanks for sharing!
best iso video ive ever seen! You got a follower more! Now i finallyy understand iso... photography (and hopefully cinematocgraphy in the future) will step up now. Thanks man!
Except it incorrectly explains ISO...
Wow, brilliant post. Thank you, Mitchell.
I love the look of this video specifically. I always try to replicate this kind of lighting and feel but I can't quite get it down, my footage always ends up looking either too cold or too warm.
Hopefully, I can shed some light on what's actually going on with that ARRI chart Mitchell is referencing (haha, sorry for the pun, I couldn't resist!)
The "shifting" in dynamic range that you are seeing when the ISO is changed in the ARRI chart is absolutely, 100% NOT what happens when ISO is adjusted in normal camera operation. ARRI is actually applying very special parameters in-camera that do not occur in normal camera operations.
The point is, you don't have to worry about using your ISO the wrong way, unless you are shooting video with an ARRI Alexa, or other certain camera models in certain shooting modes. Nothing has changed in how we understand ISO and how it affects image quality.
For most of us, changing your ISO will not cause your camera's dynamic range to start drastically shifting up and down in multiple stops.
SHORT VERSION:
This is a technology that a lot of camera manufacturers are using nowadays mainly for videographers that require maximum image quality, flexibility in post-processing and tonal range. Also some photography applications using a menu setting to activate. What you're seeing is the result of corrections happening in-camera in real time during shooting using sensor technology and in-camera software from manufacturers. The bottom line is there will not be these large shifts in dynamic range performance by simply using different ISO's in typical shooting for most photographers; unless you are using particular cameras and particular shooting modes for specific applications.
LONG VERSION:
There is a very important reason why the dynamic range "shifts" when the ISO changes. Notice how ARRI's total dynamic range is always 14.5 stops of light regardless of which ISO is selected, it is simply "shifting" up or down. It has to do with what is actually happening in-camera before raw files are even written to a card. Canon also does this with their HTP (Highlight Tone Priority) as well as Sony's DRO (Dynamic Range Optimizer). Which they offer as a menu feature you can turn on/off and was even in the old Canon Rebel's and Sony's A57 and NEX-7's from many years ago. The modern A7S series also uses a similar feature for "log" when shooting video, as well as many other manufacturers that are now using a "log" setting for videographers. ARRI also uses this method to help directors/cinematographers acheive maximum image quality in any lighting situation and give more flexibility in post-processing. They are all changing/moving the "base" ISO (i.e. removing the ability to select certain ISO's) in order to provide more stops of tonal range for a given lighting scenario.
What is happening is in-camera parameters are applied to the exposure and curves, pre-RAW stage. Basically, when you want more stops of highlights and make a selection accordingly in-camera, the camera then intentionally underexposes the image during capture, then digitally "gains" the brightness back up to original exposure before sending the file to be written to a card. This is all happening in-camera during the imaging chain before the photographer ever sees the file. The underexposure is performed automatically by reducing sensor amplification (i.e. lowering ISO) while maintaining shutter and aperture selections. For example, if your camera's normal base ISO is 100, but your camera is using this mode, it will increase your minimum available ISO to 200. So if you are shooting at ISO200 @ 1/30sec. & f/2.8 in this mode, the camera will automaticaly reduce sensor amplification by 1-stop to ISO100 @ 1/30sec. & f/2.8. This will give you 1-stop of extra highlights. Then software in-camera gains the brightness back up exactly 1-stop to match your original selected exposure. Also, some curves are applied to help the highlights "roll-off" better. The downside is that underexposure causes a loss in detail in the shadows and more noise (less light = more noise). This is why the dynamic range appears to "shift" with less and less shadow information as you gain more and more highlight information.
If you want better/cleaner shadows you can inversely lower or "pull" the base ISO below it's minimum by forcing the camera to overexpose, then pulling the exposure back down to it's original exposure. This gives more tonal range in the shadows with better detail and less noise, but at the cost of less detail in the highlights. Overexposure provides more shadow detail, and underexposure provides more highlight detail. It's always a trade off. This is why ARRI's chart @ 1:59 is always equal to exactly 14.5 stops and the dynamic range is only "shifting" up or down.
DPReview gives an excellent explaination of this technology in one of their camera reviews:
"Highlight Tone Priority option (Custom Function II.6) is a method for capturing more information in the brightest parts of the scene. It does this by applying less amplification to the signal coming from the sensor, then compensating for it by using a different tone curve to ensure the correct brightness in the final image."
*reference: www.dpreview.com/reviews/canoneos600d/12
Canon's HTP, Sony's DRO, of course ARRI, and pretty much any company that offers a "log" video shooting mode with a "base" ISO of like 800 or even 3200 minimum ISO are basically all doing the same thing: overriding exposure by adjusting sensor amplification at image capture, then compensating back to original exposure with in-camera software corrections using curves to preserve tonal range. All of this is done in the blink of an eye, out-of-sight and out-of-mind of the photographer. This is the technology being using today.
Shooting at high ISO's during the day and low ISO's at night is a great idea if you are shooting video with an ARRI Alexa, based on how they designed it to operate. But it is important to understand that this is NOT a representation of how ISO fundamentally works. This is a technology available in certain cameras, for certains applications, in certain shooting modes. What matters is the specific camera you are using and understanding how it works to get the best image quality possible out of your gear.
Thanks, Tom Cruise and Nicholas Hoult's son.
Insanely helpful, great video
So i didn't used it wrong, good
Great video! Thanks for this Mitchell!
this was such helpful insight, thank you so much 🙏🏻
Only true uncompressed raw behive like this. Worked with many cameras. Only Alexa cameras, sigma fp and kinefinity cameras and cameras that shot true raw behive like this. They all shot 12 or 14bit raw in 4:4:4 color subsampling. On my sigma fp if I switch to 8bit h264 rules doesn't apply. Even in 10bit 4:2:2 gets tricky 😊
you also need to consider camera's that are effectively "ISOless" with iso invariance. effectively meaning shooting at lower iso's will preserve the highlights while also letting you boost the shadows without adding any more noise than if you shot for the shadows
@@dyrwolv which cameras?
@@mitchellbrinkerdp off the top of my head I know the A73 and the Nikon z6 z7 are.
It’s a rather interesting rabbit hole to go down.
Wow I had no idea about the more details in the shadows and highlights part of ISO.
Thank YOU!
Now I’m even less afraid of making a video in my room at 3200. Some people do that on purpose! 😂
Actually it's the job of the gaffer.
@@edjefferson9175 what?
Nice video man😃I have the fx30 with dual iso and “should” I still use the higher base ISO for outdoor and vice versa or does it not matter if they have 6 stops each direction either way? Also would’ve been cool to see the last clip you were talking about🙂
Next up: crop factors vs. "light sensitivity" myths and legends, or why multiplying your f number by a crop factor doesn't make sense
Another great video, anytime I see someone sharing the gospel on ISO I get feels.
Only one note I would add... ISO does NOT change the sensitivity of a digital sensor, native or dual native sensor are fixed.... In general these "native" ISO's are where the camera manufacturer has best tuned for dynamic range and noise patterns... they are not actually more or less sensitive to light... Hence Gain... It's still the same process, ISO just became adopted after the DSLR revolution.
Also The ISO is going to change from manufacturer to manufacture... DO NOT TRUST ISO to be accurate to film equivalents.
Other then those points I totally agree, RUclips has had a massive impact on the way DP's (especially in the last decade and a half)... the obsession with clinical images seems largely of a byproduct of test footage/review culture... Film was never clean and honestly go back and look at something like Heroes/Lie To Me from the late 2000’s would be considered by many "unuseable"... an at the same time the photo chemical process had reached the ability to deliver film stock and printing techniques that would be considered sterile by many now days chasing the "film look".
I think most most great DP's will tell you its about your choices, your process and if you are working with a DP, its about what the two of you are looking for... not about "The right way"... This also goes for the current ongoing obsession with "perfect soft light".... but that's another topic.
@@Vitaphone don’t even get me started on soft light, mostly because I heavily disagree with the over reliance on big soft light. it’s good to know how to achieve it, but it’s robbing you of the creative choices a DP has to make if you use it exclusively in every scenario.
Either way, thanks for sharing. That’s important info. Anyone in these comments please reference above in addition to the video!
@@mitchellbrinkerdp would love to start seeing more working DP’s on the RUclips space opening up on the soft light/ “perfect exposure” topic… it’s been wild to see commercial and commercial photography lighting ideas dominate a lot of the conversation in the post RUclips era… Most DP’s I work with with years under their belt skoff at this, but newer generations seem obsessed with the RUclips meta.
Sir, you are SOOOOOO correct. I am getting much better performance from my #sonyFX30 by using this understanding. Much appreciated.
Awesome video! Glad the algorithm brought me here. Thanks. Since you mentioned the FX3: do you have "favorite" ISO levels or recommendations for that sensor? Obviously this depends on the amount and distribution of light in each scene and the look one wants to achieve, but as you mentioned it for another camera, there might be a value that generally introduces a nice amount of noise in most scenarios. Like 640, 2500, or 10000 or 16000. I am basically always exclusively using the base ISOs when i'm working... I might need to change that. Surely I can just do some experimenting myself. Curious about your opinion though.
@@SCHAUDENLUCAS I tend to shoot it just about always at 12800 unless I’m out in the sun, then it’s 1600.
Great video … thanks dude really helped my understanding
Why don‘t you just ETTR? If the DR shifts from more stops below average grey to more stops above but the total amount stays roughly the same, it shouldn‘t matter too which ISO you pick but rather how you expose, right?
Maybe I am lacking the practical examples here.
I’m getting a BMPCC OG specifically for the noise pattern.
My dumbass though he was gonna talk about iso from valorant
im swapping to a valorant channel now
Great video. Thank you. I love the way your video looks, by the way. Do you use some sort of cinema filter? (I’m a beginner, so I’m sorry if that’s not the question to ask.)
@@0xggbrnr I grade in Davinci. This was one of my LUTs with some tweaking. I may post them in the future if that’s something people end up wanting.
Came to this video late, so sorry for the delayed comment. ISO is absolutely not 'how sensitive your camera sensor is to light'. Your comment that you can change the sensor's sensitivity to light is mistaken (except, arguably, for sensors with dual conversion gain, but there it only changes at one point in the ISO range). The ISO standard for ISO speed ratings for digital cameras, ISO 12232, does not even mention the sensitivity of the sensor to light - so cannot define what you say it does. Nor does it define the gain, which wouldn't change sensitivity anyway. It's not 'technically true' that the higher you set the ISO the more noise it introduces. In fact, it's generally false. On most cameras, the higher you set the ISO the less noise is introduced. The noise is because a high ISO causes a low exposure (ISO is inversely proportional to exposure), which means the image is composed of fewer photons which means a lower shot noise SNR. I won't go through the rest of the video in detail, but my guess would be that when the fundamentals are wrong, the rest of it will likely be too. Sorry to be so critical, but when putting information out there you need to do your background research first and get it right.
As I’m watching this video I had a moment where I thought “this is exactly what RUclips was made for.” Simple but high-quality video filled with amazing info, re-watchable, and not even 10 mins long. Great work. Instant subscribe
Thanks for the insight 💡
this video might be top 5 videography lessons I’ve ever got
I leave my A7C on Auto ISO. I've shot over 100 videos for 7 clients this year and have made $80k so far. Guess how many of my clients or the people commenting on the videos have brought up noise. (Hint: 0).
Oren Soffer said exactly this in his Sony FX3 tutorial on the Sony YT channel. High ISO and ND down.
love the way this looked.
Hi I love the info you lay out on the video. I would like to ask where can I find the camera ISO to dynamic distribution chart? Thanks.
Does anybody have links to ISO distribution graphs for Sony mirrorless series? I'm using A7iii.
Side note, your framing in this video is incredibly detailed! Very cool
Isn't the optimal approach to use one of your camera base ISO's (though the ridiculous 12,800 on the FX3 is actually noisier just because of quantum fluctuations- physics!) and use ND filters and/or lighting (or light control) to make that base ISO 'correct' ?
What lens did you use to film this video, looks great!
@@NineDekay sigma 24-70 art
agree 101%
Simple, clear and effectively communicated 👌🏾
Just factually incorrect on how ISO even works...
@@RunNGunPhoto 🫡
Thanks for this!
for someone who wants to get more serious about recording, what's a great intro camera to learn on?
Amazing insight. Does this apply to sony a7s3?? If i want to shoot during daylight at 12800 I will suffer
@@muuuuusic6269 definitely does, especially since the A7S3 and FX3 have the same sensor. If you go to the 12800 base iso, you won’t have anymore information in the highlights. You’ll have the same 6/8 in the highlights and shadows as you do at 800. If you raise the iso from either base, you will get those stops in the highlights.
@@mitchellbrinkerdpgreat tip! Thank you so much 🙏🏼
this was an amazing video, my R6 has amazing dynamic range so I can shoot at 10000 ISO and theres only a little bit of grain. I did not know about shooting high iso in sunlight to get better highlights, thank you so much!
I’m sorry, but the part about using filters in a bright setting and augmenting ISOs is plain wrong. This will NOT give you more details. The point of ISOs is to "turn on the volume" on signal. So yes, in a dark setting, indeed increasing it won’t improve the ability of your camera to capture details in low light, hence the noise, but in bright light, the signal (as in meaningful color and brightness data) is already crispy clear and the signal-to-noise ratio is therefore optimal (as long of course as the whites aren’t burned). Using a filter and augmenting ISOs will achieve NOTHING for your signal-to-noise ratio, it can only dampen your signal and make you lose precious meaningful data...
ISOs aren’t necessarily evil, they can be useful in a reasonably dark setting and the noise can be used in a artsy kind of way but if you’re looking for sharpness and clean signal, it should always be the last thing you turn up. There are literally zero scenario where ISOs augment signal-to-noise ratio, from a purely physics point of view. They just can’t do that.
@@r.c8756 nah man, i showed the graphs on how dynamic range is distributed across the EI range. this is true for alexa, sony venice, etc. the same principle can be used on other cameras such as the fx6 and fx3, as well as blackmagic. next time you are out, check false color in the highlights at 400 and go to 3200 and compensate. you will see the red become yellow and yellow become green.
@@mitchellbrinkerdp I just looked deeper into it. The thing is, it’s a feature that’s pretty much specific to high-end cinema cameras. I come from a photographic background so what I said is true for digital cameras but does not necessarily apply to cinema cameras. I apologize, I should have checked that before commenting.
great content
haha...I came up in the ASA generation. Ever heard of it?
@@checkeredflagfilms ASA forever
I don't mind noise as long as it's not color noise as it just looks ugly...
Really cool I learned a lot!
"You're the cinematographer, you tell me" You dropped fire one liners throughout the vid. good stuff lol
🙏🏾
*you are explaining ISO wrong.*
Raising your ISO does *NOT* change the "sensitivity" of your sensor. When capturing an image, light enters the lens and hits your digital sensor.
The sensor is made up of photodiodes that convert the photons (light) to an electrical signal. ISO amplifies the signal, similar to turning up the volume on a radio.
You are digitally amplifying the signal, unlike film stocks where are actually more or less sensitive to light. You almost had it when you mentioned Gain.
There is ALWAYS noise in digital imagery. It's about how what noise to signal ratio is acceptable.
It's important to understand this so we're not spreading the same myths around the creator community.
You may say, "Yes, either way, the answer at the end of the day is all about the light." But the reason WHY is very different.
Raising your ISO on a theoretical sensor that is "sensitive" to light would actually "get you more light" without having to add light to the scene.
But since it's an electrical amplification happening, that's where the noise comes from. If we have little to no light, and multiply it times 10, we still have dark pixels.
And, THIS is where we come full circle and now know WHY we need to increase the *light* and not our *ISO.*
@@RunNGunPhoto The nitty gritty nuanced science of the sensor is great, but 99% of working DPs aren’t concerned with it. They’re concerned with how to get the image they want, and what the settings are doing to achieve that at the most streamlined level. Everything mentioned here about dynamic range in relation to the EI is the meat that everyone should know. And whether or not it’s actually “increasing sensitivity” or not, you can change your sensor to 12800 ISO and use a matchstick to get exposure with no loss of information if you want. Call it more sensitive or don’t, but that’s what’s important and if the easiest explanation for someone is that it makes it more sensitive then that’s fine. As long as there is an understanding of what that does to your image vs opening up your aperture, etc. The goal is to simplify for people, which is why I’m not getting into the deep inner working of science of what’s happening at the sensor level. There are plenty of DPs in the ASC who would simply say, “who cares, just know how to get the image you want.” It’s art, and you’re painting. It’s a lot about feel and understanding and taste. That’s my spiel to all the “uhm actually the sensor isn’t” comments.
@@mitchellbrinkerdp That was a really drawn out way to 1+1= whatever you feel like.
I concisely explained *WHY* it's important to know how your sensor works. Please do your research before spreading misinformation.
Making up your own "facts" for clicks on RUclips is pretty low-effort.
@@RunNGunPhoto i can promise you there are no made up facts or misinformation here. we are in completely different industries, and there are very big differences in how things are approached. whether or not “sensitivity” is the technically correct word doesn’t matter. if you pick up an Arri Alexa or Sony Venice down to an FX3 what I explained about how ISO affects your image is the case and all you need to know. ISO higher = more range in the highlights, less range in the shadows hence higher noise floor. ISO lower = more range in the shadows, less in the highlights. If you bump your ISO to 6400 in a dark room, of course there will be noise, it’s clearly not “creating” light. I said it doesn’t matter, if that’s what you want to do for the shot, then go for it. There are no rules.
@@mitchellbrinkerdp No this is an extremely convoluted and backwards way of saying "expose for the highlights." Which is also poor advice.
I am a photographer and cinematographer. "Industry" isn't relevant. It's still bad advice.
@@RunNGunPhoto it’s not a convoluted way of saying expose for the highlights. i’m not saying anything about exposing for the highlights. that is completely different. i am talking about the way exposure index dictates the distribution of the dynamic range across the ASA range. i have also mentioned how the noise floor is affected with those changes in ASA and why. there is absolutely a difference in industry here. how things are done in the photography and videography world is not equal to how things are in the cinematography world. unless you have 9+ years of experience shooting on and working with arri cameras, i don’t know why you’re telling me what i am right and wrong about. this isn’t top 10 iphone cinematic tricks here, this is stuff you would be told by people at arri. im simply taking this and saying you can apply it to other cameras because they all fundamentally do the same thing in their own way.
Awesome video, very informative.
Very useful. Thanks.
So you are telling me if I want a very high key image, rather than shooting at iso 100 I should instead shoot at 3200 and add an ND and that will produce a better quality image?
@@tobiasyoder not necessarily. whether the image is high key or low key is going to come down to your lighting. changing the iso won’t change your lighting. but if you are in a very bright environment or somewhere with a lot of bright highlights, raising your iso and ND’ing down can help retain more information in the highlights.
@@mitchellbrinkerdp super interesting. Gonna have test it before I really accept it haha
what lens filter are using for this vid? looks so good
@@joshclark9700 just an ND on the lens. I do use scatter, so I may have that running in post. If I do it’s a 1/8 BPM
Using high ISO is like plugging a ukulele into a massive amplifier and expecting it to sound like an electric fender.
@@SHDEdits you’d be pretty surprised then at how many films and shows you’ve seen that have been shot at incredibly high ISOs.
😂
I think you have to be careful when telling this to a beginner because I have mentored interns and co workers who don’t understand ISO and just wing it half of the time and their footage looks bland, over exposed or underexposed. I think it’s better teaching someone new that base ISO’s are a good starting point and should always be where you start and if you can always stay on those base ISO’s. I think what you’re teaching is an advance technique which is really cool and I think good but I hope it doesn’t misinform people who are just learning
Can you tell me please where you got that thumbnail image from
@@theMouseHacker it’s from a short film I shot