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 :)

B16: DR and latitude test

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
RED ONE BUILD 16 DYNAMIC RANGE AND LATITUDE TEST

Hi all! First of all i´d like to say that this is a personal test (not a universal truth), therefore while posting this thread i´m open and more than happy for discussion. ;D
If you wish to watch any jpg in 1,4K resolution just remove "small" from the url.

001n.jpg

N take. Image exposed for an 18% grey card with a 500ASA setting. Global tungsten light to achieve a T4 iris in all parts of the frame. WB set at 5000ºK with an 80C Tiffen filter. 4k 2:1 RC36 25fps 1/50sec. Build 16. Arri/zeiss Master Prime 50mm. Images processed with Scratch 4.0. Camera RGB color space. Gamma Rec709.

002.jpg

+1 stop

003.jpg

+2 stops

004.jpg

+3 stops
 
009.jpg

-4 stops

0092.jpg

-5 stops

0093.jpg

-6 stops


DYNAMIC RANGE

Starting from a 500ASA setting it´s clear that we´ve got more room for underexposure than overexposure. Even at -6 stops we are still getting details (we did not reached the limit at -6, therefore there should be at least 1 more stop of dynamic range in the shadows that we couldn´t check. However it might show too much noise).
So by analyzing this test we can draw this chart:

isos.jpg




For these tests we´ve started at 500ASA, but as you can see on the chart the results would have been the same at any other given ASA/ISO camera setting (we are developing a RAW image, not an RGB processed image). We just have to choose the ASA setting that we like the most and count how many stops of over/underexposure we can have. In film we usually have stocks with the same amount of stops over and under mid grey. And that is the easiest way of working (that´s why we choose an 18% grey card as a measurement of exposition). However with the RED ONE, the scale is not balanced if we start with the factory settings (ASA320). Ideally we´d like to work with higher ASA/ISOs in order to have a more balanced dynamic range. But this would produce images with more noise than we´d like, and that is why we must search for the perfect balance depending on the scene we are shooting. This is the work of the DoP (or sometimes the DIT), to evaluate the scene contrast and choose the ASA setting that produced the best results (noise/DR).

NOTE: Caution, many people get confused by the option of being able to change the exposure setting (ASA) later in post. The amount of light that hits the sensor is given by a phisical fact (shutter speed/iris) and that will mark the exposure.

So the question is: What is the native ISO/ASA of the RED ONE camera? Well, it´s a tricky question and we might find different answers. It will all depend where we want to put our mid gray (this is not new something with the RED, remember how the Varicam or the little HVX200 achieve much better DR results in film modes while giving more noise?). With Build16 i prefer to work between ASA500 and ASA640. This creates images with an acceptable noise levels while we can work without fear of burning the highlights.


Let´s analyze now the latitued of these images. According to wikipedia, exposure latitude is the extent to which a light-sensitive material (our sensor) can be over or underexposed and still achieve an acceptable result. Since the acceptability of the result is dependent on both personal aesthetics and artistic intentions, the measurement of exposure latitude is by definition somewhat subjective. To achieve best results we´ll be using Scratch 4.0, which is the only grading solution that can work natively with Redcode RAW R3D files from the RED. We don´t need to transcode files to any other image file format and therefore we avoid the possibility of loosing data or color depth during conversion.
"Greater exposure latitude allows one to compensate for errors in exposure while retaining quality. In the advancement of digital cinematography, expanding the exposure latitude is a critical ingredient which has often been a central part of professional critiques of the new camera systems because digital systems have a much narrower latitude than film."

In order to evaluate these tests i have processed each image trying to match it with our N take, the correclty exposed image. The key is to maintain the same level of brightness while preserving the mid grey at the same level, and adjusting black and white input levels to look as similar as possible as the N take. I must say that i haven´t tried to correct the color or saturation of the images. I thought it was interesting no to do this in order to show the color shift that happens when we process the footage from different zones (when we recover details from highlights or shadows). Generally speaking we can say that while our mid grey might be perfectly calibrated, highlights tend to go to yellowish while shadows adquire a cold shift. This is why most of the times, when we color correct red one footage we end up constantly "warming" the shadows.
 
latitud3.jpg


By analyzing the results the first thing that amazes me is the capacity of recovering shadow information (and have in mind that we stopped at -6 stops, there might be even more information down there!).
With an underexposure error of -2 stops, we can recover the image with almost no degradation. With -3 and -4 stops, we get too much noise in the blacks but the mid tones, skin, and highlights remain excellent. Beyond -3 we can even recover details in the shadows, but i find the noise excessive (i wouldn´t cut that shot together with others correctly exposed in the same timeline). However its good to know that even a -6 stops (remember we started at ASA500) we could even recover an image equivalent to an ASA 32,000!!!
I must point too that when noise is magnified we can see a vertical pattern, even when the noise is random. This pattern starts to be seen at -4 stops.

On the contrary, it is really hard to recover information in overexposed frames. Just over +1 stop, we cannot recover information withouth heavy degradation. This is common to all digital sensors, once we clip the highlights there is no room to recover information. That is why in digital photography we always expose for lights.
 
This helps a great deal. We're about to start work on a month-long project and I will make it required reading for the camera dept. and dit. Thanks a great deal.

Nick.
 
Cripes! It sure doesn't look pretty, but to even have an image at a sensor rating of ISO 32000 is incredible.
 
Great as always Macgregor.

Why are 100% overexposed areas in your shots not 100% white?
Am I missing something?
 
Thanks for posting the test, as well as the detailed info regarding the test- much appreciated!

Can you- or someone else - explain to me: If the camera is rated at 320, then why change the ISO to 500, why not instead just under expose the image by 2/3 of a stop to protect the highlights? (The ISO is just metadata is it not?)

Also, from the results of this test, it would seem that the Red only has 4 stops of dynamic range / latitude. Am I reading that correctly?
 
The thing is that inside the frame ther are different zones like 4-5 stops away (the white cloth vs. the black cloth), so to get the dynamic range 2 or three more stops should be added in the top and in the low end of the latitude. That gives us the 9 to 11 stops ppl get in their tests.

All this discussion regarding the latitude/DR of the RED comes just from everybody using different definitions.

On thing that is clear is that unless you are photographing a very flat image, it's better to underexpose. Whether you do it by changing the ISO or underexposing 2/3 stops from what your light meter tells you, is irrelevant.
 
Concrete said:
On thing that is clear is that unless you are photographing a very flat image, it's better to underexpose. Whether you do it by changing the ISO or underexposing 2/3 stops from what your light meter tells you, is irrelevant.
Click to expand...

That's what I thought, thanks for reaffirming that. The only benefit I can see in changing the ISO instead of underexposing is in a case where you do not have enough light to get the T-stop you want to be shooting at.
 
I have to say that it all depends on the situation you are confronting. If it is a studio set up I would rather light very flat and over expose, so I can roll the blacks down in post and avoid the noise in low light levels. Keeping skin tone shadowed areas in zone VI. Keep highlights controlled at zone VIII will remain in post when you add contrast or shadow. And being able to crunch the shadows as black as possible for aesthetics concerns. In non controlled situations I'd say that it would be better to expose correctly or under expose no more than a stop so that you wont get noise in the talent. This would bring detail to highlight areas and bring up details from the shadowed areas.
 
Concrete said:
On thing that is clear is that unless you are photographing a very flat image, it's better to underexpose.
Click to expand...

A lot of movie set ups are relatively flat though so I don't get too caught up in the underexposing thing. I shot ISO 500 one day and though I had usable stuff it was clearly the worst looking stuff I've shot recently. It felt 1.5 to 2 stops under to me.

The camera really wants as much exposure as possible without clipping anything important. Sometimes nothing in the scene should be clipped and you underexpose. Other times I feel I'm better off letting the hot spots blow. So do your own tests every chance you can.

I really want an exposure mode that tells me what is actually going on with the sensor and I'd like to toggle between edge focus and false color so I know exactly what I'm letting blow out.
 
Thanks for the test. Came in handy for profiling the sekonic manually.
 
Exposing

Exposing

...it's better to underexpose. Whether you do it by changing the ISO or underexposing 2/3 stops from what your light meter tells you, is irrelevant.
Click to expand...

If you conclude you should underexpose by 2/3 stop you have two methods available to you - depends on how you prefer to work...

Leave ISO at 320 and dial in a 2/3 stop underexposure on your light meter. Expose to the meter.

Increase ISO to 500 and let the Histogram / False Color be your guide for exposure.

Increasing ISO is just a "push" on the RAW data set when its "developed" through your monitor path, so you are protecting the highlights.

The new RAW meter in Build 17 make the effect of ISO change more obvious.
 
On the best film stocks what are the amount of f/stops you can pull back with no issues. What is the exposure latitude (not dynamic range) of best film stocks?

Update: While researching this found out that Kodak Vision3 500T Negative Film 5219 / 7219 can pull back 2 stops.

"VISION 500T Film also features extended highlight latitude, so you can follow the action into bright light-in a single shot-without worrying about blown-out details. When the film is scanned and digitized, you'll find you can recover two stops of highlight detail. This technology is enabled by the use of sub-micron imaging sensors."
 
You must log in or register to reply here.
Back
Top