Welcome to our community

Be a part of something great, join today!

  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Best Practices. Obtaining Proper Exposure. Share your methods please.

ISO and Analog Gain

ISO and Analog Gain

Most CMOS digital cameras apply analog gain to photosite readings if their sensor prior to the analog to digital conversion. Up to an optimum ISO, the amount of the gain increases with the ISO. Above that ISO, analog gain is no longer applied, and the sensor data is multiplied digitally.

While gain applied at the analog stage can result in more accurate analog to digital conversions, further digital multiplication produces no benefit to RAW data. It can be applied as, or even more accurately later. But up to the point where the analog gain tops out, there may be a benefit to accurate ISO settings.

The trick is to determine precisely where the analog gain stops.

Cheers.
 
Oh and . . . buy fast F1.X glass
Like one of these ? ;-)
scarlet2_up.JPG
 
Let me see what I can find at 2500ISO. In the meantime, most of the night stuff in this piece was 1600ISO or 2000ISO...

http://vimeo.com/29281836



Hey Joseph,

Very nicely done. What glass were you using on this? Also, how did you use your histogram here? Sorry, I've just received my Scarlet so a bit of a newby here. Read this thread and feel much better informed now.
 
@Sascha

Yup . . . my plan exactly with Canon primes. The Epic likie the light - F2.8 just not enough for interior available light or nighttime
 
Incredible thread. I'm still not entirely sure I understand it all. Correction: I am sure I don't understand it all. I have two questions. Is there disagreement about the native ISO of the sensor? Everyone keeps saying 800 but an Epic owner told me 320.

Secondly, I have my first shoot on Monday and there will be both bright-ish outdoor scenes (talent in shade with sunlit background) and a dark scene in a basement. Based on past, non-Red experience, I'm going to really want to be at 800 for that basement. Is this a mistake? If I just watch my histogram and stoplights am I fine no matter where I am?

Also, what if I want to stop it down for greater depth of field? Can I go to 800 or higher if I'm still careful about clipping on either end?
 
Last edited:
Is there disagreement about the native ISO of the sensor? Everyone keeps saying 800 but an Epic owner told me 320.

Yes there will always be disagreement about this because there are so many different definitions of ISO, most of them relating specifically to film or to gamma-corrected images, and with arbitrary manufacturer ratings being one of the most common definition. ISO is only useful if you know what it means, and it usually does not mean the same thing for very different camera systems. Some propose that for RAW cameras, it should be determined by equal over and under exposure latitude, which is reasonable but that is only meaningful to talk about in that way if all parties understand that is the way it is being used and understand the limitations of that definition, and that is rarely the case. In general I think the idea of "native" ISO is far more limited in its usefulness than many people give it credit for. Many people assume it means lowest noise, or assume it means greatest latitude, but that depends on so much more than simply your ISO which in most cases either tells you only about either your middle grey video level or about your middle range raw level, and usually both are important.
 
Hej Arnold
REDucation says: Use ISO800 (to protect highlights)
But the Native ISO actually is 320. (You can easily test this, by switching RAW-Mode on and off in the Display ... with ISO320 the Image doesn't change when you switch to RAW-Mode) So if you got your highlights under control, you light and shoot with ISO320.
Then, Expose to the right of your Histogram, stay between the goalposts keep the traffic-lights off ... then your all good and safe. Even in the Basement i'd stay on 320 and do all the lighting ans Exposure like this. If you stop down for more DOF you gotta pump more light into the scene ... if you ain't got more light ... raising ISO in Post still is an option of course ... but be sure to get your License of NEAT VIDEO Denoiser ;-) Couldn't live without it anoymore.
 
Hej Arnold
REDucation says: Use ISO800 (to protect highlights)
But the Native ISO actually is 320. (You can easily test this, by switching RAW-Mode on and off in the Display ... with ISO320 the Image doesn't change when you switch to RAW-Mode) So if you got your highlights under control, you light and shoot with ISO320.
Then, Expose to the right of your Histogram, stay between the goalposts keep the traffic-lights off ... then your all good and safe. Even in the Basement i'd stay on 320 and do all the lighting ans Exposure like this. If you stop down for more DOF you gotta pump more light into the scene ... if you ain't got more light ... raising ISO in Post still is an option of course ... but be sure to get your License of NEAT VIDEO Denoiser ;-) Couldn't live without it anoymore.

ISO320 is the arbitrary curve they used for RAW view mode, and ISO800 is what RED publicly recommends for MX, but it could be misleading to describe either as "native" without describing what that means to you. MX color science was built from a ISO800 starting point from what I understand, and ISO320 is what raw view shows you, but is that any reason to call either "native"? People always talk about "native" ISO as if it is some secret they have discovered that is inherent to some part of the system, but for the most part it is meaningless unless you define it every time you talk about it.
 
Even at 800 I doubt I'll come close to lighting the basement enough to avoid clipping on the low end. There is going to be a lot of black. I'll get my talent and immediate surroundings up but some areas will be dead black. Do I need to worry about this?


Hej Arnold
REDucation says: Use ISO800 (to protect highlights)
But the Native ISO actually is 320. (You can easily test this, by switching RAW-Mode on and off in the Display ... with ISO320 the Image doesn't change when you switch to RAW-Mode) So if you got your highlights under control, you light and shoot with ISO320.
Then, Expose to the right of your Histogram, stay between the goalposts keep the traffic-lights off ... then your all good and safe. Even in the Basement i'd stay on 320 and do all the lighting ans Exposure like this. If you stop down for more DOF you gotta pump more light into the scene ... if you ain't got more light ... raising ISO in Post still is an option of course ... but be sure to get your License of NEAT VIDEO Denoiser ;-) Couldn't live without it anoymore.
 
d'accord ... i definitely don't want to argue about a single word, since as a swiss guy, my english doesn't allow for these subtle gradations anyway ;-) i just wanted to give Arnold a few useful answer to his questions. So let's not use "Native" and instead i'd put it that way: At ISO320 the monitor shows just about a pretty true representation of the RAW-Signal and therefore is a pretty good place to be if you need to judge the image you are recording. Whereas at ISO 800 the Monitor "lies" to you because it actually shows a brighter picture than recorded. (Of course this is exactly to make you stop down a little which actually makes you protect the highlights)
It is also worth knowing that the Histogram represents the image with your ISO setting included. But The Goalposts and stoplights don't get affected by the ISO-Setting. So if you want to "Expose to the right" i think its best to judge the histogram that represents the RAW-Image as closely as possible i.e. @ISO320.
 
Even at 800 I doubt I'll come close to lighting the basement enough to avoid clipping on the low end. There is going to be a lot of black. I'll get my talent and immediate surroundings up but some areas will be dead black. Do I need to worry about this?

That's an artistic choice. Normally you want to have at least some detail in a good portion of the picture, even if its just some gradations of darkness (;-) because dead black just looks ... well ... black and therefore pretty flat. so if you manage to break up the dead black with some source of light (not lighting the whole basement) but maybe take a 650 and shoot i through some blinds or some leafs on a wall in the back to give some streaks of dull light this would add detail and if put in the right place even add depth to the shot. So short answer. The camera won't complain if you have 80% of the frame pitch black (as long as you don't try to bring back some details in the black areain post) ... but usually big flat areas don't look very cool (no mather if its black, white, grey ...)
 
d'accord ... i definitely don't want to argue about a single word, since as a swiss guy, my english doesn't allow for these subtle gradations anyway ;-) i just wanted to give Arnold a few useful answer to his questions. So let's not use "Native" and instead i'd put it that way: At ISO320 the monitor shows just about a pretty true representation of the RAW-Signal ...

It has been established, though, that it is not particularly indicative of anything that "RAW view" happens to use the ISO320 curve, it is not somehow a more "true" representation of the RAW signal. Graeme confirmed this when I asked him about it. The RAW signal is linear light and would look nothing like anything you are used to seeing and would look extremely dark if shown to you un-gamma-corrected. Even ISO250 as you say "lies" to you because it shows you a brighter picture than recorded, but that's because you aren't recording pictures, you are recording linear sensor data that doesn't look like much until a curve is applied, whether its RED's ISO320 or ISO800 curve.

My theory about why RAW view happens to use the ISO320 curve is that when RAW view was first designed, redspace was the gamma curve in use and was designed around the ISO320 curve. Back then, RAW view was more important to keep track of because redspace did not hold the whole DR of the sensor for other ISO curves, and increasing ISO would sometimes make highlights blow out that were not blown in RAW or in ISO320, so it was important to be able to switch quickly back to that initial ISO320 5000k setting since it would show you what was really blowing out and be reminded how hard you are pushing your B channel when you are shooting at 3200k for example. Now this is less important because new REDGAMMAs are designed mostly around ISO800 and scale much better throughout various ISO settings, keeping the whole range in there, so now it is not so useful to have a quick switch back to ISO320.
 
Noah and Sascha, thank you for reviving this thread for me. Let me replay in my own words what I think I'm hearing and please tell me if I have it wrong:

The sensor has one sensitivity, 320 ISO, and all others can be thought of as "virtual" ISO's. No real gain adjustments can be made in-camera like they can in my HVX or 5D. Regardless of what "ISO" I shoot at the sensor accepts it at 320. If I choose 1600 I am, to a degree, pre-choosing that gain/ISO in post because all other exposure parameters (shutter speed, aperture, etc.) were chosen based on that ISO. ISO is "chosen" on set and "confirmed" (with some latitude available) in post. The histogram will change with ISO changes based on the assumption that that ISO will be chosen in post. However, what remains the same is that a gain increase, although now in post, still results in added noise like it always has.

Is all of that correct?

It's hard to imagine that all the very dim, subtly lit shots I've seen made with the Epic and Scarlet were lit brightly enough to expose at 320 ISO. Take, for example, that one Jim posted in the Recon thread of Leonardo DiCaprio. I know I don't have the skills to throw 3K of light onto a scene (in order to keep it at 320) and make it look like it's lit by candle-light. In addition, I don't want to shoot wide open. I don't want the compromises in sharpness many lenses make wide open and I don't yet have the skills to work with the micro-thin of depth of field that results.



It has been established, though, that it is not particularly indicative of anything that "RAW view" happens to use the ISO320 curve, it is not somehow a more "true" representation of the RAW signal. Graeme confirmed this when I asked him about it. The RAW signal is linear light and would look nothing like anything you are used to seeing and would look extremely dark if shown to you un-gamma-corrected. Even ISO250 as you say "lies" to you because it shows you a brighter picture than recorded, but that's because you aren't recording pictures, you are recording linear sensor data that doesn't look like much until a curve is applied, whether its RED's ISO320 or ISO800 curve.

My theory about why RAW view happens to use the ISO320 curve is that when RAW view was first designed, redspace was the gamma curve in use and was designed around the ISO320 curve. Back then, RAW view was more important to keep track of because redspace did not hold the whole DR of the sensor for other ISO curves, and increasing ISO would sometimes make highlights blow out that were not blown in RAW or in ISO320, so it was important to be able to switch quickly back to that initial ISO320 5000k setting since it would show you what was really blowing out and be reminded how hard you are pushing your B channel when you are shooting at 3200k for example. Now this is less important because new REDGAMMAs are designed mostly around ISO800 and scale much better throughout various ISO settings, keeping the whole range in there, so now it is not so useful to have a quick switch back to ISO320.
 
Last edited:
The sensor has one sensitivity, 320 ISO, and all others can be thought of as "virtual" ISO's. No real gain adjustments can be made in-camera like they can in my HVX or 5D.

More like all ISOs including ISO320 can be thought of as virtual ISOs, each with its benefits and tradeoffs, but all with essentially the same dynamic range.
 
Can you comment on my other interpretations?

Is there any reason why I can't set my ISO to 1600 tomorrow and just try to watch clipping on both ends? How will the result be different than shooting at 320. All of my lenses would have to be f1.2 or faster to shoot in a dark room at 320.

More like all ISOs including ISO320 can be thought of as virtual ISOs, each with its benefits and tradeoffs, but all with essentially the same dynamic range.
 
My take

My take

The in camera ISO setting changes what you see in the monitor path and what the image looks like when first opened in post, therefore:
If the levels on the monitor and in the RGB derived histogram lead you to change your exposure then they have indirectly changed your results.
If your only practical way to communicate your intentions to post is via the metadata (like ISO) then it can influence post choices and by extension the final look.

Exposing to the right while staying out of the clip is a solid formula.
The notes about the way a linear light container assigns available bits is a critical bit of understanding.
Adding a bit of fill light to dark areas and then crushing them back down in post is an excellent way to reduce noise in the shadows.
Setting the ISO at 800 has pitfalls as noted, but IMO is the best option when working on a collaborative set with clients/producers - especially if the monitor they are viewing is in an environment with high ambient light!

For digital sensors instead of ISO, I prefer to look at the signal to noise ratio (S/N) and keep an eye on "feeding the blue beast". 99% of the time, as long as there is a decent amount of exposure in the blue channel, noise on the MX chip is minimal.

TIP: Even though it is an "artificial" representation of what you are putting in the RAW data set, I use the RGB derived histogram as a critical tool in choosing exposure. After doing all the obvious things to calibrate my exposure choice, I spin the iris and focus on the histogram. What I am looking for is a distribution of values that makes sense when I consider the scene in front of me and that spreads out the histogram. My goal is to get lots of discrete code values into the data set so there is more to play with in the grade. (Yes Virginia, that does require that you trust the colorist or are planning to do at least a primary "one light" grade yourself.) YMMV.

Cheers - #19
 
Arnold. Thats why i mentioned NEAT VIDEO ;-) We all shoot underexposed from time to time (we don't always have a lighting-truck at hand, do we?) and push the ISO in post to get a reasonable exposure. Killing noise with a decent plug-in is all thats left to do ;-) The above theories is how you SOULD do it in a perfect world - But there is no perfect world (with lightingtrucks fittin into a backpack) and nobody forces you to stay @ 320 in post ;-)
 
It's hard to imagine that all the very dim, subtly lit shots I've seen made with the Epic and Scarlet were lit brightly enough to expose at 320 ISO. Take, for example, that one Jim posted in the Recon thread of Leonardo DiCaprio. I know I don't have the skills to throw 3K of light onto a scene (in order to keep it at 320) and make it look like it's lit by candle-light. In addition, I don't want to shoot wide open. I don't want the compromises in sharpness many lenses make wide open and I don't yet have the skills to work with the micro-thin of depth of field that results.

Actually that thing with Leo was shot at 2000 ISO if I remember right. Granted I believe they were using a fast Master Prime wide open but still the ISO was way up there if I remember right. And I don't believe they did any post processing on it either because they were specifically testing it for low light.
 
What I don't understand about the "always shoot RED at ISO800 because it is native" is the following:

Let's say I shoot THE SAME scene using the two approaches below...

Case #1: Shoot at ISO320, and compensate exposure with a wider aperture (faster f-stop), while ensuring that RAW histograms are ETTR and highlights NOT clipped.

Case #2: Shoot at ISO800, compensate the exposure with slower f-stop, ensure that RAW histos are ETTR and highlights unclipped.

In Both cases, one could argue the image is "properly exposed" ... however, to me it seems that case #1 produces a better image because you have less noise due to the lower ISO!
(i.e. higher ISO brings more noise)

Therefore, if you know how to monitor your image and exposure properly, why wouldnt you prefer to shoot at ISO320 whenever possible ??
 
Back
Top