ISO - Useless while shooting and in post

[Scot Yount]

Yes, I agree with that. But this is a workflow related convenience, and not exactly a technical "must" that will make or brake the final image. Just imagine that you are a one man band, shooting a ridiculous short film (just one clip, for the sake or argument). How will the selection of a particular ISO value while shooting impact the final footage in a way that it could not be done latter in post? Remember that in this scenario you will be the DOP, DIT and Colorist.

So, do you really need to change the ISO while shooting, and why?

Changing the ISO while shooting may reveal problems on set that you might not notice if you set ISO and forget it. Like noise that you won't see until you bring up the levels in post when it is too late to fix it. If you see it on set...the idea is that you can fix it...but changing things that are baked...like your T-stop, shutter angle, and the scene....not enough light in the shadows? Add light. It helps you make the right creative choices while you can still make changes.

 

[Mark A. Jaeger]

Rui, Phil gave you some things to think about. Here are my comments regarding your first post:

- ISO adjustment has the same effect in camera or in post.
- The full dynamic range is always recorded without influence by ISO
- One must keep the left goal post clean and the histogram, at least slightly, away from the left side to avoid noise.
- Changes in aperture, shutter angle, lighting or ND filters will each change image brightness. So will the choice of ISO.
- On my computer Premiere Pro, Resolve 17 and REDCineX Pro each display the ISO selected in camera and the brightness in each app can be adjusted with ISO changes.
- ISO definitely has effect on the image brightness.

I suggest you watch a tutorial on RED ISO I made at https://youtu.be/x1uAMYq7yQk.

Here's some further thoughts on your later posts:

You ask "what's the point of changing ISO?". You could set your ISO to 800 and never change it again.
You can set the monitor brightness at any level. However, that could skew your visual connection to the image described by the GIO Scopes. I advise setting the monitor brightness so that what you see with the GIO Scopes makes sense to you. When something is 6 stops below middle gray – How bright or dark do you want it in the monitor?
With your fixed ISO, choose physical exposure parameters to achieve a balanced exposure with no stoplights and avoiding goal post activation. Then, if you need to adjust the brightness in post, dial in an ISO change and you will see corresponding adjustment of the image. You can also leave the ISO alone and change luminance in the NLE.

 

[Rui Guerra]

Changing the ISO while shooting may reveal problems on set that you might not notice if you set ISO and forget it. Like noise that you won't see until you bring up the levels in post when it is too late to fix it. If you see it on set...the idea is that you can fix it...but changing things that are baked...like your T-stop, shutter angle, and the scene....not enough light in the shadows? Add light. It helps you make the right creative choices while you can still make changes.

If in a low light situation if you lower the ISO, then the image will be darker and any noise will be much less visible. Lowering the ISO will not help you see it on set.
In fact, my understanding is that, it may indeed be an helpful tool but in a little more indirect way, meaning that the darker image resultant from the lower ISO, hopefully reminds you (ring the bells!) that if everything is so dark, something must be done.

But then, if you recognize in the first place that it's a low light situation (and because of that you lower the ISO), then you can just skip that step and just solve the lack of light problem.

 

[Rui Guerra]

I suggest you watch a tutorial on RED ISO I made at https://youtu.be/x1uAMYq7yQk.

Hello Mark,

Let me start by congratulating you for the GREAT video you've made about ISO! Indeed it's the definitive video about the subject! You should be hired by RED to make more such great vídeos!
THANK YOU very much for the time and effort spend in doing it and by the clear an direct speech used!

It makes very clear the answers to most of the questions I've made.

Taking all this into perspective, I believe that most of the general misinterpretation about ISO, comes from the fact that some reality/concepts gets mixed and that leads to some not so good explanations, for example:

1 - RAW DATA vs VISIBLE IMAGE - ISO do not affect in any way the first, only the second.

2 - ISOLATED CLIP vs SEVERAL CLIPS ("film") - In the realm of film shooting (several clips that will be edited together) the advantage of choosing some ISO value, and keep it consistent, becomes more evident, so we have a reference, and in relation to it we can make other exposure decisions, scene after scene, so the ones that should apear darker will look indeed darker (and vice-versa). If we are talking about just one isolated clip, then it's usefulness can be more like a "post-it" note the rest of the film team, down the workflow chain.

3 - HIGHLIGHTS PROTECTED VS LOOKING BETTER - As can be seen in your video (at 7:37​ - ISO effect on clipping), using a higher ISO value do not protect against clipping, just make it look better, with a more soft highlight roll off. ISO do not affect RAW data in any way.

4 - CHANGING THE WHOLE IMAGE vs CHANGING SELECTIVE PARTS - Here is where things get really interesting. As it can also be seen at your video (at the same timecode), ISO changes is a linear one, affecting all the image at the same time. Rising it will make highlights better looking but makes noise more visible in the shadows, and vice-versa.

Maybe the "holly grail" will be better looking highlights and near invisible noise in the shadows, meaning a non-linear change. But, hey, we have that already, it's named "Curves"...!

Then, if you need to adjust the brightness in post, dial in an ISO change and you will see corresponding adjustment of the image. You can also leave the ISO alone and change luminance in the NLE.

If we agree that (maybe most of the time) what we need to achieve the "holly-grail" is a non linear change (making better highlights and less visible noise in the shadows) and we already have it in curves (to name just a very common one), is there any particular circumstances where there is an obvious advantage of changing ISO in post and not simply relying in the NLE tools?

If using Resolve, for example, is there a real advantage in going to the camera RAW tab and change ISO in a clip basis before continue the color correction/exposure in the color wheels? OR will the image quality be the same if we just rely on the info of the flat/log base image and go strait to the color wheels? We know that the first option is a different RAW interpretation and for that reason could be a best route, but since the IPP2/RedWideGamma/Log3G10 have in it all the possible information recorded by the sensor, what would be the difference?

 

[Mark A. Jaeger]

Rui,
First - thanks for the compliments on the video. I believe the content of the tutorial is fully correct as I had the script reviewed by the most imminent authority before I recorded. He later affirmed this when the availability of the tutorial was first posted on REDUser

The process of getting from a scene to a final product is a spectrum and not a prescription. Said differently: different strokes for different folks. I perceive that you understand these:
- There are a lot of choices for physical and in-camera controls (like ISO) when the scene is recorded. BUT, best results avoid limits (no stoplights, red goal posts, etc)
- What choices does the editor/colorist make. Do they do anything in REDCineX Pro at the raw level before passing the R3D to the NLE? I used to go direct to Premier Pro and adjust raw parameters there but have recently been mixing adjustments between RCXP and Resolve 17.
- Once in the NLE you can adjust ISO or does one ignore that and simply adjust the luminance in the primary or HDR color grading tools.

For me, the 3 considerations just listed are a "work in progress". Like I said, I used to use Premiere Pro but after Rand directed me to the dark side (kidding, of course) I have been exclusively using Resolve 17 and liking it much, much more.
My choice for ISO is to largely leave it at a conservative value (800 for most things) and not adjust in post. Typically I have hit the exposure close enough that 0.25 to 0.50 adjustments of luminance during the color grade will get me where I want to be.
As they say, your results may be different but that's just part of the path to discovery. I experiment a lot and feel that I am homing in on an SOP that works.
I wish you sincere "good luck".

 

[Rui Guerra]

Thanks Mark, for your precious input and shared points of view and workflow.

All the best to you.

 

[Nick Morrison]

Yes, I agree with that. But this is a workflow related convenience, and not exactly a technical "must" that will make or brake the final image. Just imagine that you are a one man band, shooting a ridiculous short film (just one clip, for the sake or argument). How will the selection of a particular ISO value while shooting impact the final footage in a way that it could not be done latter in post? Remember that in this scenario you will be the DOP, DIT and Colorist.

So, do you really need to change the ISO while shooting, and why?

Generally speaking, when we shoot, we don't change the ISO ever as that's just metadata. If we need to add or remove light we do that physically -- either open or close the iris, or adjust lighting on set.

 

[Rui Guerra]

Generally speaking, when we shoot, we don't change the ISO ever as that's just metadata. If we need to add or remove light we do that physically -- either open or close the iris, or adjust lighting on set.

I agree with that. That's what I also do.

I wonder why even in RED's official videos we can find a somehow misleading information about this.
An example is here: https://www.youtube.com/watch?v=2y95tFuRUiY

At 1:00:16, one can see a side by side comparation shoot that's supposed to show the added clip protection and increased DR in highlights.

- The left side shoot was done with ISO 400 and then in post corrected to ISO 50 (a 3 stops decrease). We can clearly see the skin highlights burned out in both shoots, but not due to the ISO choice, since ISO do not change RAW information. So in my point of view its a case of a shoot that simply was overexposed from the beginning (I'm almost sure that those traffic lights were on...) and the problem remains after a change in ISO, as one would expect.

- In the right side an ND 0.6 which of coarse will block 2 stops of light, protecting the light from burning the skin tone. Then the ISO was lowered in post to 400 (only 2 stops decrease this time) so we can see better shadow detail (and the expected skin tone protection due to the ND filter).

It looks like this example can't show any difference in DR distribution by changing ISO. On the other hand it clearly shows the use of an ND filter, which was not the intention at all.

It's indeed said in the video that any ISO change must always be accompanied by a physical light change (obviously), which may lead to the question of the necessity of changing it in the first place, besides to have a specific level of brightness in the monitor, which of course can be useful.

In my point of view, the terms used are not the best. It's been repeatedly said that the "exposure was matching" and that was not the case (exposure is the light amount reaching the sensor, which was 2 stops lower with the ND). Maybe a better term would be the "monitor's brightness was matching".

Then again at 1:02:08, two shoots made at ISO 400 and another at 1600, both pushed 1 stop in post. It's said that the one with a starting lower ISO (400) is cleaner because of that. But the ISO 400 shoot was exposed at T2.8, while the 1600 ISO shot was exposed at T5.6. That's less 2 stops of light for the 1600 ISO shoot. No wonder why is not that clean, it just received less 2 whole stops of light!

It's the basic principle of any simple "scientific experiment" that only one variable change at the time, so that results can prove the intent.

 

[Nick Morrison]

I agree with that. That's what I also do.

I wonder why even in RED's official videos we can find a somehow misleading information about this.
An example is here: https://www.youtube.com/watch?v=2y95tFuRUiY

At 1:00:16, one can see a side by side comparation shoot that's supposed to show the added clip protection and increased DR in highlights.

- The left side shoot was done with ISO 400 and then in post corrected to ISO 50 (a 3 stops decrease). We can clearly see the skin highlights burned out in both shoots, but not due to the ISO choice, since ISO do not change RAW information. So in my point of view its a case of a shoot that simply was overexposed from the beginning (I'm almost sure that those traffic lights were on...) and the problem remains after a change in ISO, as one would expect.

- In the right side an ND 0.6 which of coarse will block 2 stops of light, protecting the light from burning the skin tone. Then the ISO was lowered in post to 400 (only 2 stops decrease this time) so we can see better shadow detail (and the expected skin tone protection due to the ND filter).

It looks like this example can't show any difference in DR distribution by changing ISO. On the other hand it clearly shows the use of an ND filter, which was not the intention at all.

It's indeed said in the video that any ISO change must always be accompanied by a physical light change (obviously), which may lead to the question of the necessity of changing it in the first place, besides to have a specific level of brightness in the monitor, which of course can be useful.

In my point of view, the terms used are not the best. It's been repeatedly said that the "exposure was matching" and that was not the case (exposure is the light amount reaching the sensor, which was 2 stops lower with the ND). Maybe a better term would be the "monitor's brightness was matching".

Then again at 1:02:08, two shoots made at ISO 400 and another at 1600, both pushed 1 stop in post. It's said that the one with a starting lower ISO (400) is cleaner because of that. But the ISO 400 shoot was exposed at T2.8, while the 1600 ISO shot was exposed at T5.6. That's less 2 stops of light for the 1600 ISO shoot. No wonder why is not that clean, it just received less 2 whole stops of light!

It's the basic principle of any simple "scientific experiment" that only one variable change at the time, so that results can prove the intent.

Yes changing ISO will not change the fact you may be under or over exposing your image - esp when shooting RAW.

Its still a useful baseline to have on set. For example, it can be useful when you're working on a dark scene (say at night) to ironically drop the ISO to say 400 to get the camera and lighting teams primed to ADD more light (as opposed to tempting them to take away light if you set your ISO at say 1600).

ISO is an interesting starting point. It's a directive of how you plan to expose the image that everyone can follow.

Another interesting choice is for example to shoot outdoors at 1600 ISO - thereby encouraging your team to NOT over light, especially when it's very bright, and there are not too many shadows to protect, etc.

 

[Mark A. Jaeger]

Nick,
I think we can all agree that camera capability is not infinite. As such, it is vital to adjust the physical exposure parameters, to the extent practical, to keep the image clear of stoplights and goal posts=sensor optimum. Setting ISO IMO in this case to something close to 800 is a good baseline. If you run out of range for the physical parameters my first choice is to add or subtract light. But, there are many situations where the light is not controllable. Then, say in very low lights, you need to use all the tricks (T1.5@1.5, LL OLPF, 360 shutter, etc) and this includes raising ISO even though this may expose the noise. It’s a balancing act.
With respect to inciting the crew to light as needed I suggest skewing the display (ISO or monitor brightness) may not be the best managerial choice. Who’s the boss?

 

[Karim D. Ghantous]

Rui, I'm glad you brought this up. I think we need to reacquaint ourselves with these kinds of fundamentals now and then. I take all this knowledge quite seriously, so it's good to be reminded to think about it.

My take on RAW ISO: you have to set your light meter to something. ;-)

It seems that, unlike stills cameras, RED RAW does not 'bake in' your ISO setting. It just records what hits the sensor, and that's it. ProRes is obviously a different thing.

Mark, I look forward to watching your video.

 

[Rui Guerra]

Rui, I'm glad you brought this up. I think we need to reacquaint ourselves with these kinds of fundamentals now and then. I take all this knowledge quite seriously, so it's good to be reminded to think about it.

Karim, I could not agree more that we all need to have a deep understanding of this and many other aspects of filming. That's why it is so important to have these kinds of "debates" and brainstorm. And no better place like this forum to do it.

Mark, I look forward to watching your video.

Don't know if you saw the link, but here it is again: https://youtu.be/x1uAMYq7yQk

 

[Mike P.]

This is one of those things that, on paper, seems like it should work fine, but in practice falls apart/isn't realistic. Ironically, RED's are weaker at this than others because their over/under is low-end weighted at 800ISO (which is what the raw tools/meters are based off of). It also makes ETTR somewhat less effective too; you're better off setting the ISO higher, and keeping exposure balanced (rather than ETTR), as it'll be more accurate and easier to finely tune based off of WYSIWYG (which ETTR does *not* do well, especially in log).

Changing ISO in-camera vs in-post yields the same brightness adjustment steps. However, the in-camera ISO setting, when used to help dictate your physical exposure on set, obviously *does* have an huge impact on the quality of your capture (and gives a better idea of how usable the image is). As mentioned, I'd argue that the WYSIWYG of an ISO applied image is more functional nowadays (more so in conjunction with the goal posts/stoplights as raw floor/clip guides).

Again, all REDs raw exposure tools use "ISO800" as their base (which is brighter than the native "brightness" of the sensor) and is already being converted/mapped from linear. So either way, you're using an "ISO" like function/representation (sans normalization in the raw tools). The nomenclature - using ISO in fixed steps - is used merely because it's a known quantity/industry standard (or should've been... took RED until DSMC2 to recalibrate to a more meter friendly ISO scale).

 

[Mike P.]

CONT'D (for some reason, I'm getting an error msg when I try to post this all in the same post):

What else? OH, Resolve defaults to 320 likely as a hold over from Mysterium, which was safely clean at that ISO (aka it's the lowest common denominator of all RED sensors), but it doesn't matter since you can change that setting to whatever you want as a default (in other words, it's moot).

End of the day if you think exposing using a Log image with just the raw exposure tools is less dangerous than using an ISO with gamma/colourspace (and/or LUT) applied reference image, you're welcome to shoot that way! But you'll inevitably run into unforeseen/unexpected problems, where important image info is too deep in the noise-floor and/or too hot to be recovered accurately (more likely). I think this is exactly what was happening with CineD's recent Komodo DR test; they did what you're thinking -- expose via log using ISO independent metering -- but because of RED's low-end weighted DR distribution it caused their skin tones to be an unrecoverable clip at only +1 from their base exposure (by comparison the p6k was +4, even though it has less DR).

More practically, when using log/raw tools, you can't accurately see unwanted distractions on screen because the log is so flat there's hardly any differentiation by colour/brightness. Similarly, you'll have a difficult time trying to convince clients/producers at video village that your Log/GioScope image is "captured correctly"... but I digress.

 

[Nick Morrison]

Nick,
I think we can all agree that camera capability is not infinite. As such, it is vital to adjust the physical exposure parameters, to the extent practical, to keep the image clear of stoplights and goal posts=sensor optimum. Setting ISO IMO in this case to something close to 800 is a good baseline. If you run out of range for the physical parameters my first choice is to add or subtract light. But, there are many situations where the light is not controllable. Then, say in very low lights, you need to use all the tricks (T1.5@1.5, LL OLPF, 360 shutter, etc) and this includes raising ISO even though this may expose the noise. It’s a balancing act.
With respect to inciting the crew to light as needed I suggest skewing the display (ISO or monitor brightness) may not be the best managerial choice. Who’s the boss?

Yes we set 800 as a baseline, but that's not set in stone. An Epic Dragon is noisy at 800 ISO, so we set it at 400 ISO. Also situations vary. If you are shooting Tungsten, you can sometimes get a noisier picture at 800 ISO. Some cameras do better if you ISO down with Tungsten (the blue channel can get starved when you have too much warm light). I say this from experience.

And yes sometimes you get cleaner blacks if you rate down. Just having black in the frame doesn't mean it won't be noisy. It's horses for courses.

In general not a fan of absolutes.

 

[Phil Holland]

Nick makes an interesting point regarding Color Temperature and Illuminant Source being an issue occasionally. This manifests squarely in how much apparent visual noise and even color stability when working under different color sources. Most modern digital sensors obviously work well within a general range of 2800-6500K, but you certainly see some photon starving sources that dip into 1800K or lower as well as above 6500K.

This is actually why I have to first test how cameras work under both of the main hero temperatures used in this industry. Tungsten and Daylight.

Very much pertaining to this conversation is my last RED Monstro and Arri Alexa test from 2019, I'm going to link to the ISO Ramps that showcase the differences in dynamic range, color stability, as well as apparent image noise for an equally exposed image:

Timecode: 6:04
https://youtu.be/XZPa3yp7lsQ?t=364

This was done around the last time I had to test all the professional cameras targeting a 4K finish. i.e. RED, Arri, Sony, Canon top of the line bodies. The goal here was to understand deeply the strengths and weaknesses of each system, discover the yield and stop signs when it comes to a clean-ish out of camera image and when a noise reduction workflow would be needed when it comes to rating the ISO and exposure practices. We do some additional testing including custom tone maps, experiment with some look development, and a few other things. Eventually opening into testing a lot of colorful lighting conditions. Generally the footage from these tests get looked at for some time and it would be tricky to show every bit of information learned in an edited video online, but here I'm showing the most useful and practical information.

 

[Mike P.]

Yes we set 800 as a baseline, but that's not set in stone. An Epic Dragon is noisy at 800 ISO, so we set it at 400 ISO. Also situations vary. If you are shooting Tungsten, you can sometimes get a noisier picture at 800 ISO. Some cameras do better if you ISO down with Tungsten (the blue channel can get starved when you have too much warm light). I say this from experience.

But totally dependent on OLPF. Like LLO at 1600, is as clean as ISO400 of STH on Epic Dragon (especially after downscale to 4k). As it pertains to the OP's 'ISO relevance', this is a prime example of where, outside of the goalposts/stoplights, RED's raw tools aren't as effective (since they're locked at "ISO800" but the noise floor isn't a concern for another ISO stop or two).

And, yeah, tungsten be starving the blues, which is also another prime example of where the raw tools wouldn't register it properly. I was under the impression that's why a lot of cinematographers don't change the temp in low-light tungsten shots (e.g. a lot of Soderberg's RED shot night scenes are left at 5000k regardless of the light source temp.)

 

[Karim D. Ghantous]

If you are using tungsten lights with a digital sensor, are you not better off by using a filter? That way, you subtract the appropriate amount from your nominal ISO (the distance between daylight and tungsten, I believe, is 2/3 stop?). So your chosen grey point is exactly the same.

 

[Phil Holland]

If you are using tungsten lights with a digital sensor, are you not better off by using a filter? That way, you subtract the appropriate amount from your nominal ISO (the distance between daylight and tungsten, I believe, is 2/3 stop?). So your chosen grey point is exactly the same.

Depends on what you are trying to do. Back in the MX days we certainly used blue filters to assist with pulling keys off of green screen in particular, but these days, i.e. the last decade, that's been pretty much something that's not needed anymore.

We're in a pretty colorful and vibrant world now given what the lights can do and how they are used. People often who shoot tungsten, daylight, and other things nearby are often tying it to the look they are going for and most often how it impacts the skin tones and subsequent surrounding colors you build upon.

 

[Simon Dunne]

If ISO is just metadata, then why do we need HDRx?

:driving: :auto: :emote_couch: