Welcome to our community

Be a part of something great, join today!

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

Understanding ISO with the RED ONE

Miguel "Macgregor" De Olaso

Miguel "Macgregor" De Olaso

Well-known member
Joined
Dec 29, 2006
Messages
796
Reaction score
0
Points
16
Location
Los Angeles
Yestarday i found some time to finally do some tests i´ve been wanting to perform since the red one was out. And i´m talking about ISO, dynamic range, zone system and exposure.

Now, in short, the RED ONE has an ISO setting (well, it says ASA, but lets say ISO to get an international agreetment rather than just American, allthoug they are the same). This ISO setting does work like if you were increasing the sensitivity of the sensor, meaning, being able to see in darker places. However this is not really what is happening in camera and definately we are not recording a brighter image at all.

In order to understand ISO setting we must have in mind what a zone system is. That´s all. Now according to my tests the camera´s dynamic range is something like this (at build12):

ZONES.jpg


This means that at 320ISO at camera we are getting up to +4 stops above 18% grey and -6 stops under. I can see a bit more detail below -6 but it is so full of noise that we would never want to use it, so we will set our black (0,0,0) at that point.

The most important thing here is to understand that changing the ISO will change the position of the 18% grey in the system zone. It does not affect the image itself. It doesn´t make the image brighter, it just looks brighter (call it LUT if you want). By setting the 18% grey one position to the right of our system zone we are seeing previously darker areas brighter now. And also we are affecting somehow the dynamic range inside the image (more precisely the stops over and under the mid gray), since we are trading highlights vs shadows.

So at 640ISO we should have +5 stops obove mid gray and -5 stops under it.

Now, what is the trade off? Well, it seems that the less light that hits the sensor the more noise those areas will have. Meaning that dark greys in our zone system are going to have some noise. IF we leave those dark greys just as "dark grey", we get a pretty clean image (250 or 320ISO). But if we bring those pixels up to a higher zone (ie: from I to III) by changing ISO (and therefore affecting the position of 18% grey), the noise is enhanced and very visible (the noise was there since the beggining but we were not able to see it).

So does the ISO setting affect the exposure of the image? No. It just plays within the dynamic range of the image to make darker areas brighter. Ok, ok, ok, this is not new for many people. Let me go on to the really interesting part.

A couple months ago, Stu Maschwitz (prolost.com) made a very interesting description of what was the best way of exposing in cinema digital cameras in order to get the best dynamic range out of them. It was all academic since i think he had not time to play with a RED ONE as much as he would have liked.
 
Again, by changing the ISO we just move the mid gray accross the zone system of our image, making our 709 image displayed in the LCD appear correctly (but affecting dynamic range). But since we are exposing accordingly to the ISO, the amount of light that will hit the sensor will not be the same.

Now, how can we get extra dynamic range in the highlights (usually the most dangerous areas in digital images)? Chaning the ISO, of course.
We don´t want clipped images, but we want our image as more on the right of the histogram as possible (since highlights also get most of the information stored in our digital files, just remember how bits and stops are stored).
So the higher the ISO, the more stops we get in the highlights, but also, more noise we will get on our shadows. Let´s see how much:

We shot the same scene at different ISO settings, but always having our 18% grey card as 18% grey (use a lightmeter or the built in false color). We are shooting 4K 2:1 5000ºK and we will compensate exposure by closing the iris of the lense (in this case a cooke 20-100 T3).

THis is what we see in our LCD and also what we get in REDCINE when just opening our images (Camera RGB, gamma rec709):

709.jpg



They more or less look the same (redcine seems not to be very accurate setting the exposure from the metadata). The 800iso will look grainier if we click play. But most important, can you see the loss of highlights? The higher the ISO the less we see out of the window!!! Less dynamic range? What the...!!!

Then we disable all color grading (press F or Color), bringing the image to Linear mode. THis shows us what we really have from our sensor:

lin.jpg


Big difference, isn´t it? The 800ISO shows a lot more detail in the highlights basically because the image recorded was darker than the 320 ISO one. Remember we just closed the iris on the lense! The ISO setting had no real effect on the raw image. But why is Redcine clipping information that has been recorded (remember those highlights lost in the 800ISO photo? Yeah, that same one that is the one with more highlight information!!!)??

But because i am interested in testing how much dynamic range i can gain in the highlights by playing with ISO, what i am going to do is export the 3 clips at the best quality we can get right now from REDCINE. This is exporting 10bits REDlog DPXs process FULL. BUT before i do this, it is very important to bring manually all clips down to 200ISO. WHy? Because REDCINE clips the highlights once you go over 200 ISO. WHy is this. I don´t know. But if you shot your entire feature at 800ISO, and you bring all the footage down to 200 for later grading it back to 800 in a decent color grading application... you will not regret since you´ll have a LOT more detail in the highlights.

log.jpg



This is what it looks in log. Don´t take too much time looking at these. This is log. It doesn´t look right for 99% of humanity. There is a difference of about 2 stops in exposure between the 1st and 3rd image. How will this translate at the end?

What i am going to do is very simple. I will make (or at least try) the 3 images to look as near as possible but i will preserve as more detail in highlights and shadows as i can. This scene has a variety of details to look at:

IMAGE0B.JPG


Using Speedgrade DI, i can easily match the grey card in all three shots, as well as the clipped white outside and the darkest part of the pelican case:

LINK TO WMV VIDEO
(WMV, not the best codec in the world, but good enough for proving this test).

Watch the video. The 500ISO and 800ISO image retain a LOT more information in the highlights than the 320ISO, making them more filmlike. Also the contrast between different zones is softer (more similar to film than the 320ISO clip). On the contrary, the 800 one has more noise in the shadows and also presents a bit less detail in the darkest areas (check the pelican case, at least 1 stop is lost). In the end, it´s a trade off.

More things to look at: check for the window reflection in the floor and walls. The 320ISO tastes more like video, while the higher the ISO the more resambles to what we usually are acostumed when watching films.

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.
 
Very nice post. I agree with you. And you will be happy with Build 16.

Jim
 
Welcome back, Jim.

-Thor
 
So far I've stuck to the received wisdom that the camera is a 320ASA system, and to change this is unnecessary and dangerous. However I've been caught out a couple of times with very noisy shadows in low key, low light interiors as a result, and reading this test makes me feel I should adapt my understanding of Red ASA.

So if the following is true:
Higher ASA = more detail in highlights but more noise in shadows.
Lower ASA = less noise in shadows but less detail / more sharply clipped highlights.

would it therefore be fair to say that you might change camera ASA according to the situation? i.e.
For a high key exterior you might shoot at 800ASA
For a low key interior you might shoot at 100ASA

If so this is a major counterintuitive idea, but one I'm willing to play with. I mean imagine choosing a 100ASA daylight film stock to shoot a low light tungsten lit interior!
Ideas?
 
That´s correct. Usually the work of a DIT or a nice DP is to measure the contrast of the scene and therefore choose the best "ISO setting" of the camera.

Imagine you have a scene with only 3 stops of contrast (ie: a brick wall in a cloudy day), you´d like to set the camera at a very low ISO, lets say 250.
Now if you have a high contrast scene (sunny day with clouds and shadows) you might want to set your mid grey higher in the zone system (usually in the V) and therefore you do this by increasing the ISO setting.
Basically this is old information, since this is similar to the way slide films should be exposed using the zone system (exposing to the highlights).
 
Hi Macgregor,

IMHO, there is no such thing as ISO (ASA) setting in the camera. All you do is changing a LUT for monitoring. RED's sensor is said to have a sensitivity of 320 ASA at 5000K. As with any filmstock its up to the DP to rate REDs sensitivity.

In most cases I find RED more to 250 ASA at 5000K. However, in high contrast scenes 500 ASA might be applicable. In this case there is a trade off towards more noise. If significant oversampling takes place (downconverting from 4K to 1080p or even PAL/NTSC, for instance) the slightly higher noise can be "swallowed" by the oversampling.

Colour temperature below or above 5000K changes the DR and noise level of RED considerably, not to the better.

All around, I fully agree with you.

Hans
 
In my book talking about ASA is confusing to many and so may not the way forward to understanding how to get the best from a digital sensor.

A CCD or CMOS chip is a different beast to film. There is bugger all exposure latitude (maybe +- 1/3 stop) at highest signal to noise ratio.
At one end the photo site gets saturated at the other it is registering noise.

A DPs job is to fit the scene contrast within the two bars.

As has been stated, if the dynamic range of the scene is beyond the range of the sensor your choice is to expose for the highlights and then recover (selectivly amplify) shadows in post which will also amplify camera noise. This is not increasing dynamic range unless you like moving the goal posts.

The other option is to alter the dynamic range of the scene.


If the shadows are noisey it is because, for the chosen exposure, they need more light!


Sure making noisey pictures is at times appropriate but otherwise underexposing (altering ASA) is second choice to lighting!




Mike Brennan
 
Understood.
But surely the ASA setting is only metadata anyway. All we're doing is using it as an exposure guide for a given lighting situation.

So shooting Red is like shooting a 320ASA daylight balanced reversal stock?
 
Clip notes

Clip notes

Back in the early days Jim & Jarred preached a philosophy for setting ASA (ISO) that Mac's tests seem to support. To paraphrase, they suggested setting the ASA to 160 for scenes with limited DR, 320 for typical scenes, and setting it at 500 for contrasty scenes like open daylight with no fill. I ran with that concept and when out on my first half dozen shoots I would routinely reset the ASA to 500 for day exteriors to keep the DP out of the harder clip, after closely examining a lot of footage I retreated to 400 as a compromise that addressed the noise issue while still avoiding the worst of the clip.

Since then I have become ever more concerned with managing noise levels and started using a ETTR philosophy. Doing product shots with splashy specular highlights I have learned even more about the nature of the results of clipping on the RedOne and no longer fear it as I once did. The Old Navy spots in the Shot on Red compilation, particularly the rooftop scenes, pushed the envelope for overexposed skin tones as a creative choice. While I can only guess at the color grading intent and its effect on the images screened it was an interesting look at managing the right side.

FWIW I have started increasing fill (and even adding 1/4 CTB to feed the blue beast) to create a flatter image to take into post. This is the basic expose for highlights, light for shadows mantra historically used for shooting reversal film and I recommend it as a starting point for the RedOne. The great thing is that you can easily restore a higher key to fill ratio by manipulating the exposure curve in RC/RA/Scratch. With careful management of the shape of the curve in the toe I attempt to strike the best balance between noise and shadow detail. YMMV.
 
Fantastic post and tests Macgregor!

This is pretty much the same principle that the newer Canon DSLRs are using (they call it Highlight Tone Priority) in order to give a softer highlight roll-off (at the expense of more shadow noise as they point out).

Blair is right on the money in adding fill light for the shadows in order to keep the scene within the 11 stop range of Red. The mindset for DPs needs to be that colour correction with this camera is mandatory, so given that you can always crush the blacks to taste later, why not preserve the best possible RAW image to give you the most flexibility once you hit the post process?

Great stuff!

Paul
 
Blair S. Paulsen said:
Doing product shots with splashy specular highlights I have learned even more about the nature of the results of clipping on the RedOne and no longer fear it as I once did. YMMV.
Click to expand...

Yes, the roll-off in the whites is astounding good presumably you shot/filtered in a daylight environment. However, it does also depend how you develope your footage in REDCine I find the roll-off in REDlog much better than in Rec709. Over all RED delivers more room in the whites as one would expect.

Hans
 
visceralpsyche said:
Blair is right on the money in adding fill light for the shadows in order to keep the scene within the 11 stop range of Red.

Paul
Click to expand...

Paul,

I'm not sure where you find the 11 stops other than in a testchart. I never shot a testchart but many real world settings and found max. 9 stops. Which is not bad at all.

Hans
 
You must log in or register to reply here.
Back
Top