I've been researching the native ISO of RED cameras (primarily the Dragon and MX). What's been confusing isn't the notion of a native ISO, but rather that in the debate between 320 and 800, people are referring to shooting at ISO 800 as underexposing when compared to ISO 320.
From one such individual:
"At 320 (which may be closer to the sensors "native" sensitivity- or the point where the raw matches the redcolor as far as middle grey) you have much less range in the highlights. By rating 800, you are underexposing the sensor a bit and protecting the highlights."
Bear with me as I'm more of a lighting guy, but my question is: how is it that by increasing ISO by 1⅓ stops are you underexposing the sensor? What am I missing?
Increasing ISO doesn't make the sensor more sensitive.. its just amplifying the signal.. and giving you more noise.. this is often miss understood.. the SN signal to noise ratio of a sensor wont change.. like turning up the volume of a radio.. more volume and more hiss..
Whether you need more or less lights wont make footage under/over exposed .. you just have to expose it correctly..
I don't know RED.. but they could be referring to the camera in EI..Sony and Arri have this so i guess RED does too.. exposure index..mode.. where by you can "rate' the sensor like you could a film stock.. 320 to 800 you have to stop down as your seeing a brighter image in the VF,but the sensor is still recording at the lower native ISO of the sensor.. so you are under exposing the recorded footage .. like changing your light meter to a different ISO than the film can says.. your shifting the grey level down.. giving more highlight protection.. although mostly people go the other way.. over expose a bit.. lose a bit of head room..theres usually way more than you need.. but get less noise..when you apply a LUT/grade the footage back to "normal" exposure..
This could be totally wrong regards RED.. but thats the only way I can think of to under expose by raising the ISO ..
By rating the ISO higher you are using less light to shoot at the same f-stop, hence why it is considered underexposing compared to using the lower "native" ISO. Same as using 200 ISO film stock but rating it at 400 ISO, for example.
Jeff, you should think of cameras like the Red Epic, Alexa, and Sony F5/55/65 as having a fixed (or native, if you like) sensitivity when recording in raw mode. If you give the sensor less light in order to protect your highlights, then you have to boost the image later which reveals more noise. Changing the ISO or EI in camera under those circumstances doesn't add exposure, it only shows you what the image will look like later should you choose to increase the gain in post.
Think of it like changing the headphone volume when recording sound; it doesn't change the recording at all. But if you turn the volume up high, it may affect you to turn the recording level down to compensate which would lead to a muddy "underexposed" recording. That is what's happening when you choose an ISO above the recommended rating on a Red camera. Ideally, you would test each camera to find how much noise you find acceptable at a given ISO and then light to that level.
when you overexpose via ISO settings on the camera you are boosting the signals gain. so in effect you have more highlight latitude because the sensor itself is not clipping. You can really see this happen if you want to better understand by pushing the ISO really far... if you light a shot for say 1600iso on a red and have some part of the frame just below the clip point as you raise your ISO you will notice that the highlights arn't moving much if at all and you may see at even 6400 iso 2 stops up that they are in the same place ....instead you are just introducing more and more noise in the bottom end of the picture. The inverse of this is also true, although harder to actually see happening.
its also important to note that more modern cameras seem to be moving away from native iso to a certain extent....RED doesn't really give native ISO for the Dragon and some cameras there are two native isos or the native iso is so high that its crazy 2000-3200 on the Sony cameras.
Edited by Albion Hockney, 02 October 2015 - 04:34 PM.
Look at the Alexa -- in Log-C mode, you don't change DR if you shoot at different ISO ratings, just the clip and crush point, so clearly it must be remapping that 14-stop DR into a log gamma mode in different ways for the changing midpoint to retain a total of 14-stops.
Sure when a camera is in EI mode.. But e.g. F55 in "custom mode".. REC709.. WYSIWYG .. if you change the ISO you are burning in the change.. from 320 to 800 .. something near clip will be over.. but your not gaining anything in shadows.. just boosting the noise.. so losing DR.. and in this mode your not protecting high lights but actually getting less of them..
There is an idea that adding gain will make the sensor more sensitive,or capture more highlight latitude .. which it doesnt.. Sensor aways stays the same..
Edited by Robin R Probyn, 03 October 2015 - 08:52 AM.
Very interesting. Thanks for all the feedback! So if I understand correctly, in most cases, if I'm recording RAW, the ISO will always be the same (ie., 320 for the RED M-X sensor). But if I light the image by monitoring/rating the sensor at 800, I'm effectively underexposing by 1⅓ stops.
What I don't get is, why is there more noise introduced in the raw image by boosting the ISO if the S/N ratio isn't actually changing, rather it's being artificially changed for monitoring and then (during post) processing? Is it simply the noise that occurs as a result of underexposing and then lifting the noise floor in post?
There isn't more noise being introduced into raw at higher ISO ratings UNTIL you convert the raw image to RGB to look correct in brightness for those higher ISO. Now the fact that the sensor was underexposed and you have a darker raw image to convert from will lead to noise in the brighter RGB image.
It's just like underexposing film -- if you decide that a normal print is, let's say, printing at 25-25-25, and then underexpose the image by three stops but print it at 25-25-25, it isn't going to look grainier anymore than the shadow areas of a normally exposed image look grainy. It's just going to look dark. But as soon as you print it "up" by three stops to look normal in brightness, let's say at 13-13-13, it's now going to look grainy.
So yes, the noise is due to the fact that you've placed more of your picture information nearer the noise fioor of the signal and now you are lifting that signal up to look normal in brightness for display or post purposes.
Underexposure alone isn't enough to cause grain in film or noise in digital -- otherwise we couldn't have a scene where the room lights go out or someone covers the lens with a black cape or the camera pans through a black shadow. It's really more a matter of the inherent noise or grain of the device/format/system and what we do with that signal or recording in terms of setting the blacks and midtones.
I mean, with film, I might find, by developing some unexposed film (no exposure at all) at normal development time and then printing at various printer lights that I get really deep blacks at 40-40-40, let's say. Well, then, how a subject looks printed at 40-40-40 then depends on how I expose it, I know what the blacks will look like already at those printer lights. I could underexpose the subject by three stops and it 40-40-40 it will look very dark.
Same goes with digital -- I could expose a raw file with the camera set to 30,000 ISO and then process it at 100 ISO for very clean, deep blacks with low noise... but the subject is going to look very dark probably.