Exposure Techniques for Komodo

[Bill Ravens]

There's been multiple web videos posted dealing with proper techniques for exposing a Komodo. Additionally, RED has published a document describing their recommended procedure... https://www.red.com/red-101/iso-speed-revisited et al.

My question regards shooting, specifically, in RAW.

The basic recommendation, as I understand it, is to ETTR, with the caveat that some headroom should be allowed for high key scenes, to protect the highlights, while a bit more exposure be allowed for low key scenes, to protect the shadows in post.

The suggestion is made that ISO be adjusted during the shoot to accomplish the exposure goals I summarized above, with the entire tactic aimed at best exposure when in post, low noise/no clipped hilites. That is to say, setting ISO below 800 for low key scenes, and above 800 for hi key scenes.

So, my question regards whether ISO manipulation is strictly a means of encouraging a better exposure(since it only affects the display when shooting in RedRaw), in other words, rather than manipulating the ISO, simply set the aperture(lighting, shutter speed) lower or higher than metered; .......or, does manipulating the ISO affect the grey point in post in some way other than strictly aperture(lighting, shutter speed) changes.

Clearly, adjusting ISO during the shoot carries over into the Clip Properties in Resolve.

 

[Leonardo Newman]

As you said ISO only affects what you see on the display. It does not affect the recorded Raw data. However, changing apeture of shutter speed is a physical change to the light hitting the sensor. Changing apeture or shutter speed or ND affects the recorded raw data and truly determines if you underexpose or clip. Exposing at a higher ISO, meaning you raise ISO above 800, see that the image seems overexposed and you then normalize exposure by NDing, stopping down the lens or adjusting shutter speed (though you shouldn't view shutter speed as an exposure tool for video) is essentially just underexposing the sensor and then digitally making the image brighter so that it looks correct. It's literally the same thing as shooting at 800, underexposing by 3 stops and then raising the ISO or exposure by 3 stops in post only difference is if you exposed at 6400iso in camera you wouldn't have to normalize the exposure in post. Hope this clarifies it a bit more for you.

 

[Bill Ravens]

Thanks, Leonardo. That's my thinking, as well. Several video posts imply that changing the ISO during the shoot affects the RAW recording, in some way, shifting the dynamic range up or down, reallocating where the data is recorded on the gamma curve... I just couldn't see how that was possible, other than the way you've described..

 

[Leonardo Newman]

Yh, a lot of people explain it wrongly. The idea of it shifting your DR is based on an illusion. You don’t actually shift your DR, you just trick yourself into under or overexposing. If you exposed middle grey at 800 iso, but then raised the iso to 1600, your image is going to look 1 stop brighter, so intuitively you’ll want to reduce exposure, if you reduce the exposure by physically reducing the light hitting the sensor, eg stopping down the lens, you’ll have properly exposed your image again, with the added Benefit of giving yourself an extra stop of protection in the highlights because you technically just underexposed the image. But because the bump in iso raised the brightness and you then physically adjusted your exposure, essentially you’ve remapped your middle grey.

Every time people explain it, they tend to make it seem as if ISO is actually doing something, the only thing iso does is makes the image brighter and darker, it’s what the user does in response to the raising or lowering of ISO, that actually introduces a greater protection for highlights or a greater protection for shadows.

ETTR might actually be the opposite of this process though, yes you want to capture as much light as possible without clipping, but when in doubt it is far better not to clip than to underexpose. As long as your subject is not too underexposed, you can always clean up a noisy shot, you can never get back clipped highlights though. And so in scenarios where you have traffic lights on both ends, I’d always allow for more clipping in the shadows.

Here's a practical example. I was bored and didn't want to clip the sun. So I had to expose the image at 6400, as you can see if I had just shot it at 800, the image would have been so dark, no DP no matter what they know can be recovered wants to send an image looking like this to post. And so, exposing at 6400 iso, gives me an image that looks normal, but protects all the information that I wanted in the highlights, so now when post sees the image, their eyes won't be raised. And in the final image, we see the flexibility that the image has for grading. So while 800 and 6400 is exactly the same image and contain the same raw date, one looks correct.... and one looks like I should be fired lol.

Image at 800 iso = Underexposed


Image at 6400 Iso = Correct Exposure


Final Image

 

[Bill Ravens]

Likewise, if you're shooting a flat, low key image, ETTR to protect the shadows, then bring the image diwn in post.

 

[Leonardo Newman]

You don’t even have to bring it down in post, just let as much physical light hit the sensor as possible and drop the iso right there while shooting. And make any further adjustments accordingly.

 

[Les Hillis]

Watching the histogram on the camera while adjusting the ISO, the numbers change and the image gets lighter or darker on the screen, but the histogram (showing the RAW data) doesn't move, showing there's no change to the actual exposure.

Kind of trippy if you're looking at a highlight that's about to clip and whatever ISO you use doesn't do anything to change it, even when you know actual exposure changes are the only thing that effects it.

ETTR is a good way to get a cleaner, less noisy image, but there's a limit to it before you start intoducing colour-shifts, just like when underexposing, but maybe not so obvious. It's also not always easy to see where some of the highlights actually start to skew towards clipping before it triggers the camera's indicators.

It's not the Komodo, but here are three crops from the Helium sensor showing some noise at three exposure levels (exposure adjusted via the light level), then a fourth exposure with the light pushed almost to clipping then pulled back down in post (about three stops worth).






There's a fair bit of compression in the stills there, but I can see the benefits of using a lower ISO and ETTR.

Here are some crops showing three levels of overexposure (ETTR) with exposure adjusted via shutter angle and pulled back down in post -






Always good to re-examine where things are at, thanks for the reminder.

 

[Josh Rainey]

I’ve been using gioscope to double check what the histogram is telling me.

For example shooting toward the sun with some cloud/fog diffusing it slightly, I’ll scan the whole scene and make sure only the sun and what’s immediately around it is clipped.

The histogram is great but doesn’t allow for the same precision.

 

[Bill Ravens]

I really dislke the term "ETTR". It's an acronym that has its origins in the photographic world; and, it's ripe with the potential for exposure errors. Nine times out of ten, when one does the ETTR, they blow out one , or all, of the RGB channels. That's unrecoverable in post. In fact, the gist of Red's paper on proper exposure, especially for high key scenes, they recommend underexposing the image . That's the point of adjusting the ISO, to give an visual hint to underexpose. In my mind, it's much simpler to set the exposure, then close 1 stop or less, or apply an ND. If DOF is a concern, using ISO to hint at reducing exposure, ignores any DOF choices. The converse is true for low key scenes, where the exposure strategy changes. In this case, one wants to ETTR as much as possible, to protect shadows.

The entire point of my post is to clarify whether adjusting ISO in camera has any effect in post. Since ISO is nothing but metadata, it has absolutely NO EFFECT in post. It's just as easy to open or close a stop, without resetting ISO.

So, concluding all this, ETTR is not a term I would suggest using in video, especially for a RED. Sometimes that works, but, more often than not, it leads to unrecoverable exposure errors. Having said all this, the goal posts are the most effective way to protect the highlights. If there is a primary subject in the scene, like a human face, then, I would resort to using a histogram, be it gioscope or video. Keeping in mind that if you're using an in camera LUT, that will affect both Video false color modes.Exposing for the key light is primary, keeping my eye on the goal posts to be sure I'm not blowing a channel. Underexposing by 1/2-1 stop, when shooting RAW, is good insurance to prevent blowing highlights. I'd always choose to err on the side of unerexposure than overexposure.

 

[Jon Pais]

Yh, a lot of people explain it wrongly. The idea of it shifting your DR is based on an illusion. You don’t actually shift your DR, you just trick yourself into under or overexposing. If you exposed middle grey at 800 iso, but then raised the iso to 1600, your image is going to look 1 stop brighter, so intuitively you’ll want to reduce exposure, if you reduce the exposure by physically reducing the light hitting the sensor, eg stopping down the lens, you’ll have properly exposed your image again, with the added Benefit of giving yourself an extra stop of protection in the highlights because you technically just underexposed the image. But because the bump in iso raised the brightness and you then physically adjusted your exposure, essentially you’ve remapped your middle grey.

Every time people explain it, they tend to make it seem as if ISO is actually doing something, the only thing iso does is makes the image brighter and darker, it’s what the user does in response to the raising or lowering of ISO, that actually introduces a greater protection for highlights or a greater protection for shadows.

There is a difference, though. With cameras like RED and ARRI, when changing ISO (or exposure index), the signal values are redistributed, with the result that overall dynamic range is uncompromised, whereas with some cameras (e.g.the BM Pocket 6K), dynamic range falls off a cliff when increasing ISO by two stops.

 

[Jon Pais]

There's been multiple web videos posted dealing with proper techniques for exposing a Komodo. Additionally, RED has published a document describing their recommended procedure... https://www.red.com/red-101/iso-speed-revisited et al.

My question regards shooting, specifically, in RAW.

The basic recommendation, as I understand it, is to ETTR, with the caveat that some headroom should be allowed for high key scenes, to protect the highlights, while a bit more exposure be allowed for low key scenes, to protect the shadows in post.

The suggestion is made that ISO be adjusted during the shoot to accomplish the exposure goals I summarized above, with the entire tactic aimed at best exposure when in post, low noise/no clipped hilites. That is to say, setting ISO below 800 for low key scenes, and above 800 for hi key scenes.

So, my question regards whether ISO manipulation is strictly a means of encouraging a better exposure(since it only affects the display when shooting in RedRaw), in other words, rather than manipulating the ISO, simply set the aperture(lighting, shutter speed) lower or higher than metered; .......or, does manipulating the ISO affect the grey point in post in some way other than strictly aperture(lighting, shutter speed) changes.

Clearly, adjusting ISO during the shoot carries over into the Clip Properties in Resolve.

Something to keep in mind is that constantly changing ISO during filming will introduce changes in texture, something which you'll have to decide for yourself whether it is acceptable or not.

 

[Bill Ravens]

Something to keep in mind is that constantly changing ISO during filming will introduce changes in texture, something which you'll have to decide for yourself whether it is acceptable or not.

Precisely, Jon. When one uses the ETTR strategy, the tendency is to set it and forget it. If the scene lighting changes, such as in a pan, the original setting is no longer valid. I do a lot of outdoor scenes. If the lighting isn't consistent across a pan, it's very easy to blow out highlights. Rather than playing around with ISO, it's so much easier to set the exposure with ND or aperture, leaving headroom for those changes in key light.

 

[Leonardo Newman]

Reds exposure false color tool is the best tool to show clipping in my experience. Not to be confused with the video false color modes. This is the false color that only reads the raw data. Sometimes you’ll find the traffic lights not lit but there might be a little bit of clipping in the scene. That traffic lights don’t light up until 2% of the pixels are clipped. So I’m the examples I posted above. My traffic lights were not clipped, but with the false color the middle of the sun was clipped. Obviously I’d never need that data so it makes since for the traffic light not to light up in that scenario, but there were a few test that I’ve shot where multiple clouds in the sky were clipped even though the traffic lights we not lit.

as a side note, if I had used the video false color, it would have shown the whole sky as blown out.

My images were shot on Raptor btw, I wouldn’t recommended exposing Komodo at 6400 iso. Raptor is kind of a different animal when it comes to underexposure, pretty wild.

 

[Leonardo Newman]

Something to keep in mind is that constantly changing ISO during filming will introduce changes in texture, something which you'll have to decide for yourself whether it is acceptable or not.

This is really not much of a concern on Raptor, up to 2500, your not going to experience a real texture difference. Going up to 6400, you’ll experience a bit, but if your one of the many people that adds grain in post or uses NR, then there would be no difference to be seen.

Also, You are right, Arri, Red iso doesn’t effect actual recorded data, but my pocket 6k pro for instance, once I go above 1000iso, all the highlights go nuclear

 

[Les Hillis]

It all comes down to whether or not you're shooting in a controlled/controllable or available/limited lighting situation.

Until such a time as sensors have a dynamic range capable of capturing 'everything', allowing the end-user to choose where to place the highs and lows in post, the best way to control exposure is to control the actual light in the scene being captured, with all the camera and lens based methods of lighting control coming second.

If and when the sensors are capable of capturing 'everything', then under controlled lighting situations, the creative control returns to how and what kind of physical light you apply to the scene, and in uncontrolled lighting situations, the creative control then gets applied when manipulating in post the available light you captured.

At the moment we're still dealing with sensors that basically impose themselves upon the scene, and so we have to use in-camera and lens tools to make the captured scene fit within those technical impositions, just like they did with motion picture and still film and with video.

So, first and foremost, control the light in the scene.

But that's not what we're talking about here, so...

Secondly, control how the light is recorded by the the camera and lens, taking into account their unique combinations of characteristics and capabilities.

It could be said the in-camera and lens adjustments are also a way to attenuate and creatively control the light and shadow being recoreded, but...is it the instrument or the tune itself that makes the music sound good? In my opinion, as long as the instrument can hit the notes and isn't out of tune, what matters more is the tune (the scene) itself.

But, again, that's not what we're talking about here, so...

What are the RED-specific exposure practices and methods, based on the functionality of built-in exposure tools available across the different camera bodies and sensors, that can be used to tune into and hit the visual notes required of the given images being played and recorded?

 

[Les Hillis]

About the use of ISO values in general...

I've mentioned this before, but imo the use of ISO for RAW-capturing digital video should have been ditched ten years ago, superceded by something like SVT (Standardized Viewing Transform).

The use of ISO made sense in terms of making it easier to transition from film to it's digital successor, but it also caused a lot of confusion and was basically a misapplication of the term.

When trying to shoe-horn modern digital sensor capabilities into the old ISO standard gives you things like ISO 128000 and lets manufacturers fudge the numbers on their camera's capabilities, and when the use of the old analogue technology that the standard was actually good for has been largely superceded, the standards should and should have been updated.

With modern computing and measurement capabilities, a new standard like SVT could give a much more precise way of measuring and representing exposure values of all the image-capturing systems, now and into the future.

Anyway...

Just mentioning it because the original post reminded me that what we've got with ISO values in-camera is effectively an N-SVT (Non-Standardized Viewing Transform) that bears little practical relation to the reason it was invented (ie. an exposure value that when adjusted has no effect on the actual exposure, but just makes it look like it does).

 

[Les Hillis]

Some other opinions an observations...

All the in-camera tools read reflectance values of light hitting the sensor, but if you don't always want to be adjusting light intensities standing at a monitor and trying to see and do everything from there, use an incident light meter to literally get into the scene and see the relative incident light values, while looking at the subject-matter itself and taking into consideration it's reflectance values under that incident light.

Conversely, use the in-camera tools to read specific reflectance values instead of using a spot-meter, as you'll be reading those reflectance values in-camera from an angle closer (literally the same) as what the sensor will be recording.

As others have said, the 'Stop Lights' that indicate highlight clipping or lowlight underexposure are useful but not ultra-precise, so look closer with the other tools or your light meter to see exactly where the limit really is when pushing towards it.

When it comes to actual exposure; flags, scrims, partial diffusion and bounce are a real thing. They can't always be replicated in post with power-windows etc., let alone in-camera (yet, hmmm...) and are a crucial part of not just settling for one overall camera-based exposure.

Exposure strategy in regards to changing the ISO shot to shot or scene to scene and pushing or pulling the image in post (or not) should be a choice that is made, not something just done randomly for other varying reasons. Or not. You can still deliberately ignore it, but you shouldn't be deliberately ignorant of it, imo.

Just my opinion, but keep your overall exposure within a stop or two of your chosen ISO, by three stops off things start to...change.

And generally speaking, as part of lighting (and exposing) for the intent of the scene, light for the intended grade so things don't have to be pushed so far in post (or in the worst case worked against).

Exposure and post-processing can't be separated in the digital medium, so it's difficult not to factor in all that can be done in post when making exposure choices in-camera. In a way, the camera's nowadays are just the data-gathering extension into the real world of the creative post manipulation process, so even for something as simple as determining exposure, some thought needs to be given to how the image captured could or will be changed in post. Which requires attaining at least some degree of knowledge in that area.

Have to say too, nothing wrong with clipping highlights and crushing shadows in-camera, if it's deliberate and serves your creative purpose.

 

[Josh Rainey]

I like to evaluate, by eye, what highlights I’m willing to clip, and then just commit. That shot with the raptor is something else, but in practical usage I’m happy with how the Komodo blows out, as long as it’s where I want ‘bright white’.

 

[Les Hillis]

With enough experience and familiarity, some kinds of exposure choices can be made accurately enough more quickly and easily by eye. At that point though, it's even more useful having the tools available to measure what can't be readily discerned.

And going off on that tangent again, imo one of the aims of ongoing sensor development should be to make clipping and crushing a post-processing adjustment only, not something unavoidably hard-baked into the image in-camera (unless you want to do that for some reason, with a LUT or some other as yet undeveloped in-camera method of constricting the image).

 

[Bill Ravens]

All very interesting discussion. As a landscape cinematographer, I abhor blown out skies/clouds/highlites. The "picture" changes, significantly with human faces, or other objects of attention in which exposure is critical. In these cases, I've found false color invaluable, not only for defining contrast, but, in opting to choose exceeding the goal post limits in order to put skin tones in a favorable place.