Understanding ISO with the RED ONE

[Dan Hudgins]

Think Crome not Neg

Think Crome not Neg

Well, it would be easier to forget about all this, shot 320ISO and underexpose 1 or 2 stops. But then the image in the LCD would appear too dark. I think that´s the reason for all this. XD

The monotor should LOOK darker when you stop down 1/2 to a max of 2 stops. You then need to ADD light to the dark areas to light the image back up, if you use the gamma or curvers to boost the lower tones, from 18% gray to "black" you will just see more noise that is always there but too dark to matter much. If you make the monitor lighter, then you get a false reading on the noise in the shadows because on a larger monitor AFTER recording compression the noise will look worse with the digital push. Better maybe to keep the ISO and monitor even, and stop down and use fill light back up to get the image on the monitor looking good ON THE SET, "in the camera" not in post.

I have this problem all the time working with by brother, he wants to "fix it in the lab" or in the case with the RED ONE (tm) fix it with the Color Correction or download-transform utilities. You cannot make bad lighting on film look like it should later, nor can you with digital, they both have noise and grain in the shadow areas. You can make the gray card lighter, just like you could when you push process film, but the result ALWAYS looks worse than just having enough light on the set, or fill outdoors.

If you look a the daylight parts of "The Train (1964)" and Planet of the Apes (1968 film)" you will see them using Arc lamps or 10Ks in daylight to fill the shadows. Reflectors look bad sometimes, but we have found that flat white Styrofoam sheets about 3/8" thick and about 2x3 foot held at or slightly above head level give a soft fill that works well without the glare that some of the glossy or metal film covered reflectors can give.

The older negative films had a more limited contrast range more like 'Crome reversal films, the RED ONE (tm) sensor has a contrast range and break through highlights like reversal film, more forgiving of under exposure then over exposure. But with the disadvantage of the RED CODE (tm) blocking more in the shadows since the size of the binary steps gets BIGGER in the shadow parts relative to how the eye sees things, if you give the image a log curve you are compressing the highlights and expanding the shadows, so the sensor records linear at 12 bit rather than log at 8 bit, you then need to crush the highlights to get the tones to look more like film, otherwise it looks like a sharp camcorder with little shadow detail and blasted highlights.

It would be nice if the sensor had less noise and the camera recorded RAW sensor data, but that is not the case, so there is some loss in recording just like any video recorder using compressed format.

Many fine movies have been made on film that looks intrinsically worse on a test chart than the RED ONE (tm), and having 14 stops rather than 11 stops will not make up for bad lighting on the set and not having it look good on the viewfinder. I keep telling my brother, "it needs to look good IN the viewfinder since it only get worse further on." Yes you can filter and things, but a clean good image will look better, for the most part Color Correction spreads the image tones by raising contrast in some part of the image curves, and so can only REDUCE the amount of actual image data in the end result frames.

See this example image from The Garden of Allah (1936),

http://www.widescreenmuseum.com/oldcolor/technicolor10.htm

As an example of how good an image you can get with a camera with limited tonal range by carefully controlling the light and not pushing the film too much. Later they pushed the film more, and it never looked as good, better to fry the actors then overrate the ISO for quality results.

Color negative Film has about 2 stops advantage over the RED ONE (tm) in the highlights before it gets too off to save, so that is why I think from the test images EI 1280 (320 -2 stops) is the "Bottom" exposure for 18% gray, you cannot light much more than that up in the dark areas unless you also light up the mid-tones, you would need to add +2 stops of light in the shadow and +1 stops of light near midtone). Because of the way the sensor is sensitive in "powers of two" bringing up the shadows two or even three stops will only bring the highlights up less than 1/4 stop. It will not look that way on the set, it will look like the shadows are way over light, but that is where a linear (not log) display mode on the monitor could help, if the monitor shows all 12 sensor bits as linear, then you can see that the shadow areas look dark to the sensor even when they look bright to your eye.

The RED ONE (tm) is the first Digital Cinema camera that will make a real impact on how movies are shot worldwide, it is changing motion picture production forever. It was possable because sensors are just becoming good enough to use for Digital Cinema in place of film. In the future we should expect better sensors, but no sensor can make up for a lack of art direction, it can help, but is not a replacement if you are going to show images and not just use the image as a template for computer generated "images".

I wonder if there will be a need for cameras in the future since more and more of the frame is being used for computer generated images, i.e. not getting it "right in the camera" with lights and filters like so many great films of the "past"...

 

[Blair S. Paulsen]

Zonal flow

Zonal flow

After a couple months with the RedOne I added soft edge ND grads and a 2 stop Attenuator (provides an even gradient from no ND to 2 stops across the entire filter) to my kit. When time is tight or the grip package is too small for the scene I am able to quickly roll off the most problematic areas of the image.

My rationale is that even if the RedOne with the original Mysterium sensor is capable of 11 stops, I am better off capturing a range of 8 stops and redistributing the information into my intended zones in the grading. YMMV.

 

[Daniel Reichenbach]

Great thread Mc Gregor: Ansel Adams would be proud of you: The Zone system, that simple and helpfull it is. When you understand it you will always have your lights where you want them.

 

[Emmanuel Cambier]

Do I get banned if I say Digicon (Schneider), I mean, isn't this the best solution to this very problem.
If there is any drawback to this filter, could someone shine in?

Emmanuel

 

[Miguel "Macgregor" De Olaso]

The monotor should LOOK darker when you stop down 1/2 to a max of 2 stops. You then need to ADD light to the dark areas to light the image back up, if you use the gamma or curvers to boost the lower tones, from 18% gray to "black" you will just see more noise that is always there but too dark to matter much. .

That´s not correct, sorry. You are talking about a different thing, filling shadows with light.

 

[Gregor Hagey]

learn to expose to the right

learn to expose to the right

If i not remember wrong, Stu said that it would be great if we could underxpose by 2 stops our cine digital images in order to have the transition from highlights to pure white as a soft curve (again, filmlike). Well, this is just what we did. Is it worth the extra noise? Maybe for some. For me 500ISO seems the best compromise at this point. Will this change with bulid16? We´ll see.

I think that you won't have any issues with clipping if you move on from the Zone system and where middle grey should be and use the RGB histogram and Expose to the Right. I just finished a feature with Red where I consistently got pristine results by using this method. The ISO settings are merely LUTs so the director can see on set what you have in mind when you process the RAW image later. For dark night scenes setting the ISO to 100-160 yields great results. For bright high contrast exterior scenes (ie. an unlit actor in a car with full sun outside) 640-800 ISO yields great results (digicon helps too!). For average scenes stick to 320 ISO. Of course use your incident meter to set the contrast ratio of your lighting.

 

[Thor Wixom]

Is this Ansel Adams revisited?
Very enlightening.

See my tag below...

 

[Antoine Fabi]

Macgregor,

I'd be curious to see the result if you could raise the black level a little in the 800 ISO clip. I suspect it would look a little smoother. Just curious.

 

[Bill Anderson]

Man, I understand the "zone system" inside out and all this seems counter intuitive to me. i.e "use a slower iso at night and a much faster one in bright sun". Is it me or the terminology being used?

So, if I translated this correctly: expose to place optimum highlight detail, and (should the shadows fall out of bounds and you can't fill light) change the "ISO" to raise your shadow details (and noise)??????? And produce a more "film like highlight seperation."


Wouldn't it have been more helpful (accurate even) to have placed the window highlight values exactly, and then shown the effect of various ISO settings on the shadows (and dynamic range)? Since we're all agreed it's generally better to expose for the highlights.
And this example seems (to me) to infer that ISO choice is associated with incorrect highlight exposure: Mac says the 320 iso image highlights are overexposed when compared to the 500 and 800 samples. Why did you not expose the 320 iso highlights correcly, Mac? sounds like you were exposing for the shadows.????? or placing the 18% card at the expense of highlight seperation/detail. Does this reflect real world behaviour? Perhaps in some instances; so many variables boil down to taste .

What am I missing here (apart from a RED!)????

 

[Lauri Kettunen]

Macgregor, a very nice and informative thread, thanks.

Up to build 15 the light meters have been relative, and as you seem to say the ASA/ISO setting just moves the mid point corresponding to 18% gray to left or right. Now, this raises a question, why the light meters are relative? In case of a film camera the relative ISO setting is obvious, but in case of a digital camera the sensitivity of the sensor is absolute. Consequently, the new technology could also have absolute light meters which are more intuitive than the readings of the old tradition. The situation is like the cameraman had to convert Farenheits to Celsius, dollars to euros and so on.

The absolute scale could be established on stops; 0, -1, -2, -3, -4 ... stops from the maximum. In this sense increasing the ISO from the nominal value corresponds to decreasing the maximum value from 0 to smaller number of stops. Decreasing the ISO is the other way around about increasing the maximum from zero.

 

[Dan Hudgins]

Changing the camera ISO does not give more shadow data?

Changing the camera ISO does not give more shadow data?

That´s not correct, sorry. You are talking about a different thing, filling shadows with light.

I think these tests have been valuable.

If I understand them, when you change the ISO on the camera setting from 320 to 800 the image of the 18% gray card gets lighter on the monitor with the iris unchanged?

But the amount of 12bit signal compressed stays the same no mater what ISO the camera is set to with the iris unchanged?

If so, all I was saying was that in other words was, do not use the camera ISO setting, since the signal is poor in the shadows if you think the ISO is above 320, just add enough light to see a balance on the monitor after setting the iris for the highlights. Having the monitor get lighter when you change the "ISO" in the camera could give a false view of the noise in the 4K images later? Having the monitor show the absolute linear sensor output as a constant lets you light for the absolute limits of the sensor with a visual constant reference... Yes you can use an absolute histogram, but it may be hard to tell how much of the histogram TO clip on the top if there is a window in the shot like the example images, all of the outdoor part of the histogram (that part past clipping) is not in those example images (at the high end something bright in the frame is almost always clipped), so being able to see the monitor in linear display absolute rather gamma adjusted to mock up ISO change might help avoid optimistic exposure.

Does anyone know if the signal to the monitor comes from the sensor BEFORE or AFTER the RED CODE (tm) compressor?

If the monitor signal comes before the RED CODE (tm) compressor then it may show image detail that is lost in the RED CODE (tm) compression. Otherwise the camera would compress then de-compress the image before it gets to the monitor and resize it to make the shadow noise look less (or more?) than in the 4K image.

When you resize a RAW 4K image down to 2K you average 4 pixels to 1 pixel, when you do that you get a noise reduction of about 1/4th. The reason that you will probably not gain so much in pixel to pixel noise lowering with RED CODE (tm) is that the RED CODE (tm) pre-processing has already averaged the adjacent pixels of the same color (with a threshold) to act as a niose filter to reduce the bandwidth, making it hard to tell if the noise would be better or worse in the 2K images from 4K images with or without the RED CODE (tm).

 

[Dj Joofa]

When you resize a RAW 4K image down to 2K you average 4 pixels to 1 pixel, when you do that you get a noise reduction of about 1/4th.

I think, assuming certain conditions, the typical reduction is by a factor of 2 and not 4 when you average 4 pixels.

 

[Miguel "Macgregor" De Olaso]

What i still don´t understand is why the RED is rated 320ISO in rec709 when lightmeters and false color confirms it is something like 125-160. So,

- either RED engineers were drunk that day
- there is something missing here

Of course it would make sense to set 320 in your lightmeter and set 320 in the camera, frame with a dark image in the LCD and sdi outputs and have a 1 extra stop in the highlights as discussed before.
But dark screens are not usually liked by agency creatives (clients). XD

 

[Miguel "Macgregor" De Olaso]

Macgregor,

I'd be curious to see the result if you could raise the black level a little in the 800 ISO clip. I suspect it would look a little smoother. Just curious.

Yep, let me play tomorrow and i will upload.:w00t:

 

[Andrew M.]

After a couple months with the RedOne I added soft edge ND grads and a 2 stop Attenuator (provides an even gradient from no ND to 2 stops across the entire filter) to my kit. When time is tight or the grip package is too small for the scene I am able to quickly roll off the most problematic areas of the image.

My rationale is that even if the RedOne with the original Mysterium sensor is capable of 11 stops, I am better off capturing a range of 8 stops and redistributing the information into my intended zones in the grading. YMMV.

Redistribution that you mentioned could be done in many creative ways.
Here is one way that I like to add extra two stops to the scene.
http://www.gimpguru.org/Tutorials/NDFilter/

I use it also in tungsten light when I do not have time to compensate for red and blue with color filters (pain in the neck)

Basically you can correct shadows in the separate version of the same clip and highlights in yet the other separate version of the same clip and then to merge it in AE together.
Also if scene allows for it, you can use above mentioned method, on top of two way (shadows/highlight) process.

 

[Júlio Taubkin]

Man, I understand the "zone system" inside out and all this seems counter intuitive to me. i.e "use a slower iso at night and a much faster one in bright sun". Is it me or the terminology being used?

So, if I translated this correctly: expose to place optimum highlight detail, and (should the shadows fall out of bounds and you can't fill light) change the "ISO" to raise your shadow details (and noise)??????? And produce a more "film like highlight seperation."


Wouldn't it have been more helpful (accurate even) to have placed the window highlight values exactly, and then shown the effect of various ISO settings on the shadows (and dynamic range)? Since we're all agreed it's generally better to expose for the highlights.
And this example seems (to me) to infer that ISO choice is associated with incorrect highlight exposure: Mac says the 320 iso image highlights are overexposed when compared to the 500 and 800 samples. Why did you not expose the 320 iso highlights correcly, Mac? sounds like you were exposing for the shadows.????? or placing the 18% card at the expense of highlight seperation/detail. Does this reflect real world behaviour? Perhaps in some instances; so many variables boil down to taste .

What am I missing here (apart from a RED!)????

It's just a way to compensate exposure using RED's REC709 monitoring options as a guide. Basically RED monitors are not true WYSIWYG because they can't display RAW, naturally. Changing ISO only changes the monitoring ISO and the exposure guides your camera will give you. So you compensate your monitoring backwards, and can "automatically" compensate exposure by following your monitor guidance. Even better, that allows you to see a "Preview" of what you could get in post when you correct the image, underexposing but monitoring with a higher ISO is like having a way to see if the noise levels will fall inside your tolerance.

How hard is that to judge from a REC 709 HD display is something I cannot say.

But, anyway, if you want to shoot blind, expose with your ligtmeter only, and trust your memory of noise in latitude tests and your spotmeter values (pretty much what we do with film), all this certainly is irrelevant, since we have the first chart (Macgregors grayscale) and can determine the absolute sensor values for iso and latitude tolerances. Basically 320 underexposed 2/3rds to be on the safe side (or 500 ISO) + or - 5. - Actually, just realized that is a lot!

 

[Gregor Hagey]

What i still don´t understand is why the RED is rated 320ISO in rec709 when lightmeters and false color confirms it is something like 125-160.

Typically DSLR's true ISO is the lowest ISO you can set. Every ISO above that requires gain. It becomes a matter of choice how much noise is acceptable at a given exposure. Red is no different. When you process the .r3d you can carefully place middle grey where you want it - exposing RAW on set is about getting as much of the picture in the brightest portion of the image. Also the way the Red specifically monitors at ISOs below 320 you won't get a true indication when you are clipping, so monitoring at 320 is the safest way to go right now.

I would love to see a feature where I can quickly toggle between a linear light gamma at the native ISO with no white balance correction and the normal REC 709 gamma at whatever ISO and other settings you have for monitoring to see where the picture is truly clipping.

 

[Hans von Sonntag]

What i still don´t understand is why the RED is rated 320ISO in rec709 when lightmeters and false color confirms it is something like 125-160. So,

- either RED engineers were drunk that day
- there is something missing here

Of course it would make sense to set 320 in your lightmeter and set 320 in the camera, frame with a dark image in the LCD and sdi outputs and have a 1 extra stop in the highlights as discussed before.
But dark screens are not usually liked by agency creatives (clients). XD

Hi Mac

I'm totally with you.

I find Rec709 in REDCine useless almost dangerous, it developes the whites' roll-off in a inaccetable way (fringing). I only use REDlog which is much more pleasing, concerning the whites and moving the histo in a much more "right" position.

According to Graeme Nattress REDlog is desinged to render the best basis for 10 Bit files to grade off. I'm not sure what Rec709 is for.

I also find the HDSDI-Out too dark and the Ref Movies need some tweaking when beeing transcoded to ProRes to make them acceptabel looking for post.

If you address this issue in the camera by changing the Mointor-LUT with the ASA slider to 120 ASA you have to make sure that no one is using this metadata in REDCine. We did it this way (I use a lightmeter for the right exposer) and found it in the end too misleading.

I tell my clients this is a videoassist, a bit on dark side but we can raise the brightness on the monitor...

Not a final solution. Build 16 must fix this.

Hans

 

[Mark K.]

I'm confused, this all seems quite counter-intuitive to me. As I'm understanding this - the ISO setting on the Red One is simply metadata, and it's not actually possible to adjust the sensitivity of the Mysterium sensor, is that correct?

Therefore does adjusting the "ISO" on the camera simply affect the brightness of the monitor you've attached to the camera? i.e. you underexpose by a stop or two to preserve highlight detail, and raise the ISO/brightness simply to make monitoring the camera easier?

Should you therefore essentially just shoot for the histogram (i.e. just keep channels from clipping) and worry about the rest later?

 

[Hans von Sonntag]

I'm confused, this all seems quite counter-intuitive to me. As I'm understanding this - the ISO setting on the Red One is simply metadata, and it's not actually possible to adjust the sensitivity of the Mysterium sensor, is that correct?

Therefore does adjusting the "ISO" on the camera simply affect the brightness of the monitor you've attached to the camera? i.e. you underexpose by a stop or two to preserve highlight detail, and raise the ISO/brightness simply to make monitoring the camera easier?

Should you therefore essentially just shoot for the histogram (i.e. just keep channels from clipping) and worry about the rest later?

Yes, you are right.

Please bare in mind that the histogram (with the colour channel traffic lights) does show the values of the LUT you are using.

All possible ASA settings in RED do not reflect what the sensor feels. There is always a REC709 gamma applied with that infamous "headroom" in the whites and some crushing in the blacks.

I find it meaninful to use a light meter but I'm sure that using a monitor LUT (say 320 ASA) and the histo will bring you some useful results when you know how interpret it. Although it would be much better if the RED camera would not use REC709 but something like REDLog.

For now the colour traffic lights are helpful because they tell you when which channel clips (there is more in the RAW though). Again, also the traffic lights depend on the ASA you set. I would not care about the histos "mountain" since this is artificially moved to the left by the REC709 gamma. Ah well....

Hans