I’m going to keep bringing this up until people start to take note and understand that with an electronic camera ISO is NOT sensitivity.
With an electronic camera ISO is a guide to the required shutter speed and aperture needed to get the correct exposure. This is different to sensitivity. The ISO rating of a video camera and it’s sensitivity are closely related, but they are not quite the same thing. Because different gamma curves require different exposures the ISO rating for each gamma curve will be different even though the gain and actual sensitivity of the camera may be exactly the same.
Lets take the Sony PXW-FS5 as an example.
If you shoot using the standard camera settings you should expose white at 90%, middle grey will be around 42% and skin tones typically around 70%. At 0dB gain the camera the camera will display an ISO equivalent rating of 1000 ISO. So let’s say you are using a light meter. You set it to 1000 ISO and it tells you you need an aperture of f5.6 to get the right exposure.
Now you change to S-Log2. If you do nothing else your white card will now be at around 75% and middle grey will be around 40%. At 0dB gain the camera will show an equivalent ISO of 3200 ISO.
But hang on – The camera is still at 0dB gain, so there is no change in sensitivity. .But the camera is over exposed, S-Log2 is supposed to be exposed with white at 59% and middle grey at 32%.
So we go to our light meter and change the ISO on the light meter from 1000 ISO to 3200 ISO. Because the light meter now “thinks” the camera is more sensitive by almost 2 stops it will tell us to close the aperture by nearly 2 stops. So we go to the camera and stop down to f10 and bingo, the image is exposed correctly.
But here’s the important thing – The camera hasn’t become any more sensitive. We haven’t replaced the sensor with a different, more sensitive one (as you would do with a film camera where you actually change the film stock). We are still at 0dB gain (even though the camera tells us this is the equivalent to a higher ISO).
The only reason that ISO number changes is so that if we were using an external light meter we would get the recommended exposure levels for the gamma curve we are using. In this example closing the aperture increase the highlight range that the camera would be able to cope with and this helps us get that full 14 stop range from the camera, although closing the aperture means less light on the sensor so the pictures end up a little noisier as a result – That is unless you choose to rate the camera at a different ISO by over exposing the log a bit.
ISO is useful, but you need to understand that it isn’t really sensitivity. After all we can’t change the sensors on our video cameras and that would be the only way to truly change the sensitivity. Any “sensitivity” change is really nothing more than a gain or amplification change. Useful but not the same as changing the actual sensitivity. Gain will make a dark picture brighter but it won’t allow you to see something that the sensor can’t detect.
It is much easier to understand dB gain with an electronic camera as it actually tells you exactly what the camera is doing and it is actually my recommendation that people use gain rather than ISO for all of the above reasons. The use of ISO on electronic cameras is very badly understood, in part because it’s a largely meaningless term because it doesn’t tell us how sensitive the sensor is, how much gain we are using or how much noise we are adding. Give any experienced camera operator a camera and ask them how noisy will it be a 18dB gain and they will have a pretty good idea of what the pictures will look like. Give them the same camera and ask them how noisy will it be at 8000 ISO and they won’t have a clue.
The problem is ISO is trendy and fashionable as that’s what “cinematographers” use. But lets be honest with ourselves – we are using electronic video cameras, whether that’s a Red, Alexa or FS5 so really we should be using the correct terminology for an electronic camera which is gain. It would eliminate an aweful lot of confusion and tell us how much noise and grain our pictures will have. It’s noise and grain will levels will determine how good a clip looks and how much we can grade it, so we need to clearly understand how much gain is being added in camera and dB gian tells us this. ISO does not.
Side Note: Modern film stocks will often have 2 ratings, the ISO or actual measured sensitivity of the film stock plus the EI or Exposure Index which is the recommended setting for the light meter to get the best exposure. In some respects the ISO rating of a video camera is closer to the EI rating of a film stock. Perhaps we should stop calling it ISO and use the term EI instead, this would be me appropriate and signify that it is a reference for best exposure rather than true sensitivity.
UPDATE: A comment on facebook was why not display both ISO and Gain side by side. This is an obvious solution really. Why do camera manufacturers force us to choose either ISO or gain? Why can’t we use a hybrid of the 2? I see no technical reason why cameras can’t show both the gain and ISO at the same time – Problem solved.
12 thoughts on “ISO Confusion Once Again!”
Brilliant spot on artical. Yes, some of us think it is trendy to talk ISO and EI.
I have a sensor related question for you. When out in the field and more ND is needed on the FS7 and I don’t have a mattebox on does under rating the sensor add noise to the image when shooting HG3 (800 ISO or -6db) ? We use to do this a lot years ago but I suspect I am reducing the dynamic range. Just a run and gun way of dealing with bright sunny day and reducing depth of field.
Using negative gain or an ISO lower than the base ISO will reduce the dynamic range. You will clip the shadows early and artificially limit the peak recording level, so you will never reach 100%. don’t do it, much better to add ND.
What is the “base ISO”? I always thought that the base ISO was the minimum ISO/ best image performance in a lovely sunny day. This is what FILM do at say ISO100 as a norm. (or 50,64). Now I hear that a base ISO would be 1000ISO? What? So IMO this value should rather be 0, or 100 to keep things consistent with film and start working from that.
I’m really not seeing myself cranking the iso to 1000 and go filming without being extremely worried about the grain I’m going to generate. More than the values, I am interested in a 0 noise image to start with. Does all this means that sensors need to be electronically pushed to match Film sensitivity? I am also not ready to admit that filming at 100 ISO reduces my dynamic range. If it is the case, there is some sort of cheating, misleading thing from the manufacturers to explain a lack of true performance. In other words, it would be more beneficial to replace these wrongly called ISO units with something different and normalised to express the sensitivity of the sensor to the light. I use dB on my PMW200, and yes, I would not go negative as I understand that I’m burrying the dynamic range. but the same on say an A7SII at ISO100 or 200 and anything below 1000? By the way, someone could tell me what is the base ISO of the A7SII? I might have to reconsider a lot of things I believe then… Bit confused now. Thanks Alister.
Video sensors tend to be more sensitive than conventional film stocks, especially with large sensors that have very big pixels like the A7s (A7s II base ISO for S-log is 1600 and 800 ISO for standard gamma). PMW-200 is the equivalent to approx 340 ISO at 0dB. So these are the equivalent ISO’s that give the best performance, in particular best dynamic range. Reducing the ISO is exactly the same as using negative gain and will as a result in a reduction in dynamic range when shooting video (stills operation of the sensor is different as the A to D convertors can be normalised on a per image basis).
The problem with a video sensor is that changing the way the Analog to digital converters on the chip operate or changing gain settings effects how bright the image becomes. This is similar, but not quite the same as a change in sensitivity and there is nothing to stop the manufacturer choosing any setting they wish as the base sensitivity. Changing these parameters also has an impact on noise, but there is no clearly defined standard for how much noise is acceptable or to be considered normal. On top of that it’s normal to include noise reduction in the cameras processing. This means you could have two cameras that claim the same equivalent ISO but actually perform very, very differently depending on the manufacturers priorities. One of which is selling cameras and often bigger or higher is seen automatically as better by many consumers. So while ISO might help us get the correct exposure, it isn’t necessarily telling us how our sensor is performing.
The solution as I see it is to get camera manufacturers to declare the noise figure at the cameras base ISO. This would tell us a lot more about the cameras true performance. Then instead of just displaying ISO, display both ISO and gain side by side, for example 3200/+6 on an A7s II would indicate 3200 ISO or +6dB gain. Every time we add 6dB of gain we double the noise, so this would tell us the image will be twice as noisy as it could be.
There is just something I’d like to add. With an EX3 a few years ago I wanted to check if increasing the gain in camera results in a less noisy image than increasing the gain in post and that was totally the case.
At the time I wanted to get an explanation for that and I read someone say (I have no idea where anymore, sorry) that raising the gain adds voltage to the sensor, suggesting that it was indeed adding some sensitivity. Maybe he was wrong. But in the end, it still seems to look better when you add gain in camera.
Adding gain at the sensor is usually better than adding in post – not so sure these days with better post tools but the reason is obvious enough. You are adding noise before both the recording and any noise reduction that is going on in camera. So trying to boost gain in post means you are left doing it straight from a recording with any inherent noise in the recording being amplified as well.
The exception to this would be something like 16 bit raw – where you have all the data from the sensor available in post.
Adding gain in camera makes the picture brighter and also increases the noise. It doesn’t make the camera more sensitive. It’s not different to turning up the volume to make a quiet bit of music louder. Yes you will hear it more easily, but there won’t be any extra music.
It is typically beneficial to add gain in the camera rather than post as the majority of cameras and codecs incorporate a noise reduction process. In addition a bigger signal going into the codec encoder will result in a better signal than trying to encode a very small low level image. However modern post production tools that incorporate noise reduction and high quality processing where you can selectively amplify just the bits of the image that you need can do a really good job of adding gain later in the production chain.
Hi Alistair, great article. Could you please expand on it with a paragraph as to how this theory relates toS changing the base iso for S-log in firmware 4? It is less noisy now at same exposure levels. And i can’t quite square that with the nothing has changed in reality answer you allude to adove.
Shooting at base iso of 2000 makes no difference as to how one exposes log as the iso is just an indicative rating for the camera. Shooting at base iso 2000 in S-log will change how the image will look once captured Compared to S-log at 3200 iso, and it will have lees noise in the shadows, but as far as effecting whether or not to i should still expose brighter, it has no effect. One should still expose brighter in log on the FS5 so that the gamma curve is capturing more detail in the shadows and therefore less noise. Essentially by rating the ISO of the camera in S-log at 2000 as opposed to 3200 you are forcing it to require less light to achieve the same exposure. This will effect how the image is captured in the camera and helps protect the operator form underexposing s-log but shouldn’t necessarily change the way one approaches exposing for s-log.
Doe that sound about right?
Yes, about right. Putting more light on to the sensor will always lower the noise in the final image.
When the Sony A7S was first released, and i saw how high the ISO could go, it made me re-think the whole exposure process. People could essentially start using ISO to expose and image, instead of getting the aperture they wanted for light. Which left the aperture, just being a way to set what type of focus you needed. Im looking forward to working with the FS5 on 0db, and not messing with the ISO. If the camera is intended to work at the base, depending on the gamma curve, ill let it work its magic.