KOMODO....

[Phil Holland]

Doesn't film stack most of its dynamic range above middle gray? Not really being a low light performer, there can be like twice as much DR above middle gray as below, can there not?

I know digital cameras have surpassed film in terms of total captured dynamic range. However, if you grab some ASA 500 film stock, rate any RED at ISO 500, and shoot the same daylight scene, you're going to get much more highlight detail from the film.

If the goal is to have the equivalent latitude or more above and below middle gray than film, because film and digital are different mediums, you would rate digital to the approximate ISO that you would expose film to a similar scale.

You decide what middle gray is, I'm talking mainly about measured dynamic range. ISO 500 Digital versus ASA 500 isn't going to give you the best from digital for instance.

That said, due to digital being cleaner at higher sensitives you get away with a higher ISO value when exposing in comparison to film.

My advice if you are shooting Monstro, swing around ISO 800-1280, again maybe ISO 1600 on the Standard OLPF.

Oddly, at 400-500 ISO on Monstro you do have what you are looking for anyways if metering correctly.

Film doesn't win overexposure tests. People like the way film looks when it clips, but once data (i.e. the medium isn't capturing anything) is crushed or clipped, it's gone. There's no magical fairly dust in there to reveal anything found in the noise floor or clipped highlights other than a phantom ghost of what channel clipped last when it did. What happens to color rolling off into highlight mainly is a property of color science and you have to make a general choice of what things look like. Since most of us like somewhat of a print film aesthetic I think this largely influenced the look as the medium adheres to that theme, similar to even printing on paper optically. Short of LSD, you don't want one level of saturation at all times, that's super weird. I think it's interesting because my last chat with Yedlin surrounded throwing away standard workflows and developing your own from the ground up, which provides a deeper understanding of how you can use the captured image information. Tuning every aspect of the image to produce a result through a unique pipeline you create. It's a fun process, time consuming, but a nice exercise in what I would call full digital development of a given capture system.

I think I'll write something up likely for the last time with the side by sides to make this clearer, but film was what I primarily worked with intimately for the first decade of my career as digital was poking into the field. The 2nd decade of my career is where digital really made strides which is why it was so hot to test it's viability in comparison to film's expense, technical merit, and processing time related issues. From about 2012 on, it's been a landslide of people navigating away from film and RED played a fairly big role in this around that time as Dragon was one of the main cameras to test against. You'll notice that a whole bunch of "film peeps" also navigated to shooting on RED for the first time around then as well. It really was that stark once you saw the comparisons in a theater or suite. i have a few pressing matters, but I even got some stock in the fridge leftover, might fun to bang on this old dead horse one last time.

 

[Sabovic Adis]

Once again, you misunderstand. We're talking about the amount of highlight latitude the medium can capture before clipping. Middle gray is always (for most of our purposes, anyway) middle gray but that doesn't mean your camera can capture a luminance value 10 stops brighter than that.

Phil, this comment is EXACTLY why I got away from using the terms "stops" and "middle gray" in favor of "brightness values" and "luminance values."

With all due respect, with total dynamics of 20 stops it means exactly that.

 

[Brian Boyer]

With all due respect, with total dynamics of 20 stops it means exactly that.

ONLY IF your camera is capable of that. To date, no RED camera is, which is what we're talking about. Besides, most modern digital cinema cameras redistribute their dynamic range above or below middle gray depending on what ISO/ASA you rate the camera (presumably with a light meter).

 

[Sabovic Adis]

ONLY IF your camera is capable of that. To date, no RED camera is, which is what we're talking about. Besides, most modern digital cinema cameras redistribute their dynamic range above or below middle gray depending on what ISO/ASA you rate the camera (presumably with a light meter).

Well, it's how middle gray value is derived - you measure total dynamics, then you cut that in half. Put an ISO/ASA number next to it and there you go, film or sensor. :smile5:

 

[Phil Holland]

I assure you a stop of light is a fairly standard unit of measurement within the world of photography.

Film also has it's own unique "redistributed" tone curve, negative and print stock each for that matter. Whenever I publish measurements and charts you'll find I'll often put up linear response as well as whatever flavor of log is considered standard within the current workflow of whatever system as that has changed over time for a few companies. Studio side, Rec.709 and Rec.2020 standards are pushed out, then we also make some "other stuff" to see how we like it when projected or pushed through a pipeline. Similar to what Light Iron Color is really in terms of the curve aspect of it.

 

[Gregory Karydis]

I think I'll write something up likely for the last time with the side by sides to make this clearer, but film was what I primarily worked with intimately for the first decade of my career as digital was poking into the field. The 2nd decade of my career is where digital really made strides which is why it was so hot to test it's viability in comparison to film's expense, technical merit, and processing time related issues. From about 2012 on, it's been a landslide of people navigating away from film and RED played a fairly big role in this around that time as Dragon was one of the main cameras to test against. You'll notice that a whole bunch of "film peeps" also navigated to shooting on RED for the first time around then as well. It really was that stark once you saw the comparisons in a theater or suite. i have a few pressing matters, but I even got some stock in the fridge leftover, might fun to bang on this old dead horse one last time.

I've gotta say I really enjoy your more scholarly articles and these things sure provide an interesting read that tickles my intellect!
Though, yeah, the poor creature has been beaten to a pulp, but there's always someone who questions digital (and RED sensors in particular), so on occasion I'll pull up your posts and explain what's what to them in layman's terms.

As for me, shooting for 800 is working out beautifully for VFX and I love how clean my plates look!
Film to digital scanning is not always done right and that inconsistency frustrates me.
One of the reasons I trust in RED! :wink:

 

[Brian Boyer]

To be clear, I'm not trying to make the old "film is better than digital" argument, here.

I am saying film is still better than digital in two notable ways that I hope can be overcome sooner than later. One is highlight latitude at equivalent ISO/ASA settings and the other is the way film handles things (saturates the colors first before clipping) like traffic and tail/brake lights versus digital's tendency to clip them.

If the goal is to have the equivalent latitude or more above and below middle gray than film, because film and digital are different mediums, you would rate digital to the approximate ISO that you would expose film to a similar scale.

You decide what middle gray is, I'm talking mainly about measured dynamic range.

I'm also saying, despite digital and film being different beasts accomplishing the same thing in vastly different ways, there's still an expectation of the same (or better) behavior at equivalent settings.

This is what it seems everyone is after (in addition to being able to do what film can't), given the way digital cinema has adopted the film workflow.

Being a musician, you know that there are many digital audio plugins that very closely mimic the behavior and functionality of analog gear with nary a tube or circuit board to be found. I guess I always expect digital cinema to get to the place where it does the same, regardless of how it has to get there.

It's close, but falls short in these two areas. Well, you say film doesn't win overexposure tests, but I'll get to that later.

Okay, for the sake of clarity and all our sanity, let's forget film for a second. The way I'm thinking about it is this. Maybe I'm off base. If I am, someone please tell me how and why.

Take a Monstro, a Gemini and Komodo. Rate the Komodo sensor at ISO 800 at T4.5, let's say. Rate the Monstro and Gemini at ISO 3200 at the same T4.5 aperture, ND/adjust shutter accordingly, point them all at a bright daylight scene and see what you get.

If the Komodo sensor captures as much or more highlight detail as the other two, you've effectively got a sensor that has at least two stops better highlight performance, amirite? Who wouldn't want that? I honestly don't know why people debate me on it every time I bring it up.

ISO 500 Digital versus ASA 500 isn't going to give you the best from digital for instance.

Agreed, and this is relevant to my point about how people test - and design, for that matter - digital cinema cameras. Manufacturers of digital cams have adopted terminology, methodology and a lot of functionality that are holdovers from the film days of photography.

It's only natural for people to treat them both the same, pursuant to how they've been designed and intended to be used. And familiarity with those tools and processes serve to reinforce that traditional design and usage.

Digital may not fare so well in overexposure latitude compared to film at ISO/ASA 500 but that's how they're going to be compared (as well as digital versus digital) 99 times out of 100 when testing cameras, from what I've seen. That 1 time they're not is you, by the way.

So, did digital cinema camera manufacturers shoot themselves in the foot by carrying over analog concepts that may cause people to underutilize digital by not playing to its strengths?

Should there be a paradigm shift in how we think and go about testing and using digital on an industry-wide scale, similar to what you discussed with Yedlin?

If that's the case, is the concept of a standard outdated as it pertains to digital cinema? Or, is everything close enough for rock 'n' roll and a few adjustments in mentality and methodology should handle the rest?

Oddly, at 400-500 ISO on Monstro you do have what you are looking for anyways if metering correctly.

Interesting.

Film doesn't win overexposure tests.

This runs contrary to every film vs. digital overexposure test I've ever seen. Of course, all the tests I've seen set an equivalent ISO/ASA.

People like the way film looks when it clips, but once data (i.e. the medium isn't capturing anything) is crushed or clipped, it's gone. There's no magical fairly dust in there to reveal anything found in the noise floor or clipped highlights other than a phantom ghost of what channel clipped last when it did. What happens to color rolling off into highlight mainly is a property of color science and you have to make a general choice of what things look like.

So what must digital cameras do to overcome the traffic/tail/brake lights clipping problem, particularly at dusk/night/dawn, that most film stocks don't have? Is that more a function of color science than how bright of an object a sensor can capture before clipping?

Or, is it a combo of both, where color science is hindered by the amount of brightness the sensor can capture before clipping. Does grading for and subsequent viewing of HDR reveal more of the color saturation I'm looking for versus SDR?

One thing I routinely do is crank the ISO all the way down in RCX or Resolve to see what's actually clipped and what simply got rolled off that can be recovered with some work.

Whatever the fix is, I want that behavior gone. Since I always get pushback on the notion of increasing overexposure latitude in RED's sensors, I guess I don't understand what has to happen to make capturing those types of lights accurately a reality. Doesn't that look bother anyone else or am I just about the only one?

And before people think I'm obsessing over minutiae, the ability to capture lights like this accurately will have a positive impact on capturing things like faces and skin tone in bright light as well.

I think I'll write something up likely for the last time with the side by sides to make this clearer, but film was what I primarily worked with intimately for the first decade of my career as digital was poking into the field. The 2nd decade of my career is where digital really made strides which is why it was so hot to test it's viability in comparison to film's expense, technical merit, and processing time related issues. From about 2012 on, it's been a landslide of people navigating away from film and RED played a fairly big role in this around that time as Dragon was one of the main cameras to test against. You'll notice that a whole bunch of "film peeps" also navigated to shooting on RED for the first time around then as well. It really was that stark once you saw the comparisons in a theater or suite. i have a few pressing matters, but I even got some stock in the fridge leftover, might fun to bang on this old dead horse one last time.

If you ever have the time or inclination, I'd love to see tests you conduct that support your position on film's overexposure prowess vs. digital's. Maybe you've done them already and I missed it.

Of all the posts of yours that I've seen (and there are many I haven't), the only ones I can recall having actual examples from film were the noise ones. I've only ever seen you do side-by-side comparisons in a controlled, lit setting with digital images.

Well, it's how middle gray value is derived - you measure total dynamics, then you cut that in half. Put an ISO/ASA number next to it and there you go, film or sensor. :smile5:

Um, unless I've totally misunderstood what middle gray is and how to determine it, this is wrong, at least in terms of photography. If this were true, middle gray would be different for every camera and light meters would be useless because there would be no standard frame of reference.

Here's a handy formula I picked up somewhere. In order to determine "proper" exposure, or middle gray: 100 ASA @ 24fps @ 180 degree shutter @ 100 fc = f2.8. You can derive every other exposure by halving and doubling the various values. This has nothing to do with the midpoint of a camera's individual dynamic range.

I assure you a stop of light is a fairly standard unit of measurement within the world of photography.

I assume it's me you're assuring. There's no need. I know that. But, when I start using that terminology in the context of sensors having more overexposure latitude, someone (usually Šabović Adis) comes along and starts saying things like the quote above and I can't shake him.

It distracts from what I've actually said, right or wrong, and causes me to waste time and effort trying to clear things up. As long as I said "luminance value" I was safe from his musings.

 

[Hrvoje Simic]

Yes, rating the camera for a higher ISO will gain you more highlight latitude.

Rating the camera higher by "digital gain" (equivalent) only changes the image representation of captured data.

It does nothing whatsoever to highlight lattitude.

So what must digital cameras do to overcome the traffic/tail/brake lights clipping problem, particularly at dusk/night/dawn, that most film stocks don't have?

Capture light of higher intensity.

Is that more a function of color science than how bright of an object a sensor can capture before clipping?

It is a property of a sensor.

Does grading for and subsequent viewing of HDR reveal more of the color saturation I'm looking for versus SDR?

No.

It cannot reveal what is not captured.

I don't understand what has to happen to make capturing those types of lights accurately a reality.

Increase od DR. Upwards.
I.e. sensor has to be able to capture significantly brighter light.

 

[Brian Boyer]

Rating the camera higher by "digital gain" (equivalent) only changes the image representation of captured data.

It does nothing whatsoever to highlight lattitude.

When I refer to rating the camera at a higher ISO, I know this also means subsequently stopping down or adding ND to compensate for the "increased sensitivity" (by digitally raising the gain) a light meter thinks you have when you set it for that higher ISO.

It's a 2-step process. Those two things together do, in fact, let you capture more highlight detail. Since Phil already knows this, there was no need to go into exposition.

With you, on the other hand, there's a tendency to pounce on details that were left out intentionally because they were believed to be understood. It's what the phrase "it goes without saying" means.

I can't seem to ever do that with you because you tend to ignore context in favor of strict, literal interpretations.

Capture light of higher intensity.

It is a property of a sensor.

Increase od DR. Upwards.
I.e. sensor has to be able to capture significantly brighter light.

That's what I thought.

No. It cannot reveal what is not captured.

I'm talking about before clipping occurs. Sometimes tail lights, for example, are only clipped at the bulb (or not not at all, until the brakes are applied) but the rest of the light appears pink instead of the saturated red it actually is.

Will HDR show a less steep roll-off to white (more color saturation) up until the point it clips at the bulb position (if it does at all, before the brakes are applied) than when viewing in SDR?

 

[Blair S. Paulsen]

Rolling log

Rolling log

I would venture to say that the key difference is the chemical response to photons in film stock is natively log vs digital values that are inherently linear off the sensor. There are many ways, both in camera and in post, to mimic log response curves - but, as noted, they can't utilize values that weren't captured to begin with. The dual gain architecture in the Alexa does a pretty good job of capturing highlight detail at lower gain, while still providing a relatively normal middle grey via the boosted signal. Arguably, it's one of the reasons DPs raised on film feel more at home with an Alexa.

Personally, when I'm looking to get more highlight information captured - including retaining chroma in things like taillights, I like to take advantage of the excellent S/N ratio in modern RED sensor tech. In other words, ETTL to keep the highlights well below clip during capture, then put middle grey back where I want it in RAW development. Yes, I realize it's not really the same thing - but with some finesse in grading it can look rather film-like. Warning - underexposure risks need to be calculated/mitigated with intension to get optimum results. YMMV.

Cheers - #19

 

[Hrvoje Simic]

When I refer to rating the camera at a higher ISO, I know this also means subsequently stopping down or adding ND to compensate for the "increased sensitivity" (by digitally raising the gain) a light meter thinks you have when you set it for that higher ISO.

It's a 2-step process. Those two things together do, in fact, let you capture more highlight detail.

It's a two step process unless you just stop down on same ISO rating. In which case it's a one step process.

Which doesn't do anything whatsoever to the property of highlight lattitude in either case. You just stopped down. That's all.

I can't seem to ever do that with you because you tend to ignore context in favor of strict, literal interpretations.

Since the internets is filled with nonsense such as "higher ISO gives you more DR in highlights", "downscaling increases DR", etc. combined with wishful thinking and copy/paste mentality and as a result some get to grade & fix underexposed imagery coming from high end ISO 800 cameras in 2019, cameras otherwise capable of outstanding imagery, accurate, non-romanticised interpretations of technical matters, although understandably frustrating to go through (and repeat), are a rational thing to do, in order to reduce and hopefully avoid the ever-emerging misinterpretations of imaging tools which lead to their missuse.

That's a long sentence. Sorry.

I'm talking about before clipping occurs. Sometimes tail lights, for example, are only clipped at the bulb (or not not at all, until the brakes are applied) but the rest of the light appears pink instead of the saturated red it actually is.

Tonal consistency depends on sensor, base colour science and transformation and can be improved in latter stages, to an extent dependent on properties of the preceeding ones.

For the most part, that shouldn't be a problem today.

Will HDR show a less steep roll-off to white (more color saturation) up until the point it clips at the bulb position (if it does at all, before the brakes are applied) than when viewing in SDR?

Imagine a tail light red orb with bright red ring on the outside and white middle where the sensor clipped and there is no tonal data.

Provided that transformations to SDR and HDR don't deviate tonality nor clip or clamp gamut, SDR will show the whole signal extreme at a lower luminance and saturation level, HDR at higher. I.e. with HDR the same harsh clip will just slam you in the face harder.

To make the transition to clip softer you cannot add the data which is not there to fill in the gap. Display max luminance and saturation doesn't affect that. That only shows you what you already have at different light values.
You can only take out from the "top", expanding the area from nothing to something inwards through image manipulation, which has to go down, not up, so consequently will reduce the saturation of the tail light to a wider range.

You can't have both the saturation on that signal peak and a smooth transition, it is either/or. To keep both the saturation on that luminance level and smoothness to clip, clipping point has to be "further", which is achievable only if you have more DR captured on the top. So, you need more DR not just to prevent the clipping at some point , you need it for a transition up to that point as well . Captured DR, not display DR.

The same thing goes for shadows, and some stops have to be reserved for the toe, like they do for the shoulder in highlights, to have that natural transition and proper density left for contrast distribution. The more you force the image manipulation towards the extremes the reduced the smoothness of a rolloff.

Rolloff smothness and tonal consistency are a solvable part, how far you get to keep the colour and detail depends primarily on the sensor.

 

[Sabovic Adis]

(...)

Um, unless I've totally misunderstood what middle gray is and how to determine it, this is wrong, at least in terms of photography. If this were true, middle gray would be different for every camera and light meters would be useless because there would be no standard frame of reference.

Here's a handy formula I picked up somewhere. In order to determine "proper" exposure, or middle gray: 100 ASA @ 24fps @ 180 degree shutter @ 100 fc = f2.8. You can derive every other exposure by halving and doubling the various values. This has nothing to do with the midpoint of a camera's individual dynamic range.(...)

Hm, you're talking about determining the proper exposure then all of a sudden you go with 100 ASA...
No, we don't know the ASA, we're yet to determine it. Why would one choose 100 ASA as a base (put your gray card here!) if everything gets scorched just two stops higher, and at the same time there's some 14 stops to play with down in the basement?
No, you go with 6400. Get yourself 8 stops over and under. :smile5:

 

[Brian Boyer]

Hm, you're talking about determining the proper exposure then all of a sudden you go with 100 ASA...
No, we don't know the ASA, we're yet to determine it. Why would one choose 100 ASA as a base (put your gray card here!) if everything gets scorched just two stops higher, and at the same time there's some 14 stops to play with down in the basement?
No, you go with 6400. Get yourself 8 stops over and under. :smile5:

WTF? Is it me or is it him? It’s him, isn’t it?

Dude, I didn’t come up with the formula. It is what it is. It’s helpful, say, if you don’t have a light meter on you but you know the photometric data of the lights you’re using at several distances.

Or, maybe you have your meter and took your readings but you don’t have it on you at the moment you decide to change settings or you don’t feel like pulling it out or the battery died.

This formula gets you to a proper exposure at f2.8. From there, you can figure out the rest.

Instead of trolling me, grab your camera and your meter and test it. It actually works.

 

[Phil Holland]

I'll make a new thread about this is as it's grossly off topic, but you'll love it. Or hate it. Are make pancakes with it.

 

[gabrielB]

Although the subject currently discussed always makes for an interesting read there's a part of me that would like to see part of the discussion circle back to Komodo.

Here are some thoughts. Komodo is a new camera design in a body significantly smaller than what RED has ever done before. People expect S35 or even hope for FF which as we know with current high quality sensor tech requires a lot of power and generates a lot of heat and needs beefy cooling.

This new camera design being smaller means smaller heatsink which has to mean less heat generated which then also means lower power draw (also nobody expects having to hook up V mounts for adequate run time on a camera of this size)

If RED can keep this sensor specs in line with the DSMC2s in terms of color science and DR, and be in line with expected resolution and frame rates compared to cameras such as the Zcam, it would then mean RED has developed new sensor tech and electronics design that can capture high quality cinema imageries in a package that draws less juice generates less heat and requires less cooling.

This to me is quite significant and pretty exciting. And it also makes me wonder about the future and how these new thermal & power improvements may be implemented in larger upcoming DSMC3 cameras/sensors that could draw less power run cooler and be quieter !!

 

[Toni Tur]

Hmmm... RED action cam???

 

[Blair S. Paulsen]

I'll make a new thread about this is as it's grossly off topic, but you'll love it. Or hate it. Are make pancakes with it.

Makes sense, just tell me the thread title and I'll bring the maple syrup...

Cheers - #19

 

[Sabovic Adis]

First off, I aint him.
Second off, I don't believe in trolls. Sometimes.

WTF? Is it me or is it him? It’s him, isn’t it?

Dude, I didn’t come up with the formula.

That's what I'm worried about.

It is what it is. It’s helpful, say, if you don’t have a light meter on you but you know the photometric data of the lights you’re using at several distances.

Or, maybe you have your meter and took your readings but you don’t have it on you at the moment you decide to change settings or you don’t feel like pulling it out or the battery died.

This formula gets you to a proper exposure at f2.8. From there, you can figure out the rest.

Instead of trolling me, grab your camera and your meter and test it. It actually works.

Dang...:rofl:
It sure works with 100 ASA film, no one ever said it wouldn't, but
what if you use film you don't know the ASA of? How do you determine if it's a 50, 100, 250, 500 ASA or in between, whatever? What do you do?

Phil, feel free to chime in! :biggrin:

 

[jamesbridges]

No HDMI, hopefully SDI
No XLR, hopefully stereo mini
Timecode in? Gosh I hope so...

 

[David Rasberry]

I don't think I ever wanted a camera as bad as I wanted the fixed 2/3" scarlet. To this day if RED had a garage sale and sold those i'd pick one up.

Me too. My favorite general purpose format is the 1” or S16mm range. I wound up with a Digital Bolex instead. Still use lenses I used on my film Bolexes.