Topics

The Making of the CML NFA Tests 2020 - part 1


Adam Wilt
 

This is a quick review of how the 2020 camera & lighting tests were done, to explain why things look the way they do, what mistakes we made, and what we’d do differently next time.

Pre-production started months in advance. Getting fifteen different lights, eight different cameras, and a space to test them all in is a nontrivial bit of work, and booking a crew adds to the effort. Geoff contacted me last July about helping with tests, initially planned around Cine Gear Expo in Los Angeles; in October he changed the location to the Netherlands Film Academy in Amsterdam as this made the entire process much less expensive. 

In late December Geoff posted detailed test specifications on Google Docs. Here’s an excerpt:

—-

The object of these evaluations is solely to establish the differences in the colour response of 8 cameras when matched with 12 different lights in both tungsten and daylight.

The same lens will be used on all cameras. The lens for the lighting test will be a Cooke 65mm S5i and we will use a Zeiss Supreme 50mm for the C500-2 and FX9 stress test. Cameras will be set (when possible) to whatever resolution records a frame closest in size to 24mm x 13mm, larger when necessary.

The cameras will be set to the manufacturers’ preset settings in both tungsten and daylight. If the camera manufacturers’ preset values are different from the standard values of 3200K and 5600K with zero offsets in the G/M axis we will use the manufacturer’s settings whilst noting the non-standard value used by the manufacturer.

All lights will be set to their manufacturer’s preset values for 3200K and 5600K and, where the option is available, they will be set to the appropriate camera.

Light level will be set to 100FC measured in the middle of the Kodak Grey Scale card using the Sekonic C800 pointing directly at the camera.

The ISO of all cameras will be set to 800, this is the base setting of all of the cameras except one which is 500. The aperture will be set to T5.6, this will result in all cameras being exposed to a stop more than the manufacturer’s recommendation, this is normally the best balance between Dynamic Range and noise. The film will be exposed at 1/3rd stop over the rated ISO with daylight stock the lens will be set to T4 rather than T5.6 as with all other tests.

Where possible we will record RAW but in the case of some cameras, this may not be possible in which case we will try to use ProRes444 or an equivalent as recommended by the manufacturer. 

A lighting diagram is attached but due to limitations in the software, the 8 * 4 vertical trace frame and 8 * 4 poly sheets either side of it have had to be represented by a softbox.

pastedGraphic.png

We will be using an 8 * 4 trace frame 1.2 metres from the subjects. The light will be used on full flood at a distance that completely fills the trace side to side. The height of the light will be adjusted to keep the vertical angle of the light constant.

pastedGraphic_1.png

This lighting technique is used to provide a consistent look from all the different styles of light.

In both of the examples above I have used an additional backlight which is switchable between daylight and tungsten. We will use BB&S Pipelines for this in the current evaluations.

Light levels and colour will be established with a Sekonic C800 but checked with a number of other light meters including Cine Meter II, Lighting Passport Spectrometer, Minolta VI, Minolta IV and more :-)

As all the manufacturers claim that their dimming is accurate we will use the lights built-in dimming to establish the correct levels.

—-

This document was accompanied by files containing the camera idents, Netflix’s best-practice setup guides for the cameras under test, and the like.

Three weeks prior to the test Geoff announced that he was going to be undergoing a complex medical operation and likely would not be in attendance, so all this pre-production work let the rest of us jump in and get to work without making too many mistakes.


Adam Wilt
technical services
Vancouver WA USA (no, not that Vancouver, the other one)


Jesse Harris
 

Adam,

Thank you so much for this very valuable overview - it adds great context and detail to an extremely helpful (this is an understatement) resource!  Great job, I haven’t had the chance to review everything on my grading monitor, but I already feel and see the consistency that is the only basis for any true test!

Many thanks to you and of course Geoff,

Jesse David Harris

Cinematographer | Editor - NYC


On Mar 5, 2020, at 9:48 PM, Adam Wilt <adam@...> wrote:

This is a quick review of how the 2020 camera & lighting tests were done, to explain why things look the way they do, what mistakes we made, and what we’d do differently next time.

Pre-production started months in advance. Getting fifteen different lights, eight different cameras, and a space to test them all in is a nontrivial bit of work, and booking a crew adds to the effort. Geoff contacted me last July about helping with tests, initially planned around Cine Gear Expo in Los Angeles; in October he changed the location to the Netherlands Film Academy in Amsterdam as this made the entire process much less expensive. 

In late December Geoff posted detailed test specifications on Google Docs. Here’s an excerpt:

—-

The object of these evaluations is solely to establish the differences in the colour response of 8 cameras when matched with 12 different lights in both tungsten and daylight.

The same lens will be used on all cameras. The lens for the lighting test will be a Cooke 65mm S5i and we will use a Zeiss Supreme 50mm for the C500-2 and FX9 stress test. Cameras will be set (when possible) to whatever resolution records a frame closest in size to 24mm x 13mm, larger when necessary.

The cameras will be set to the manufacturers’ preset settings in both tungsten and daylight. If the camera manufacturers’ preset values are different from the standard values of 3200K and 5600K with zero offsets in the G/M axis we will use the manufacturer’s settings whilst noting the non-standard value used by the manufacturer.

All lights will be set to their manufacturer’s preset values for 3200K and 5600K and, where the option is available, they will be set to the appropriate camera.

Light level will be set to 100FC measured in the middle of the Kodak Grey Scale card using the Sekonic C800 pointing directly at the camera.

The ISO of all cameras will be set to 800, this is the base setting of all of the cameras except one which is 500. The aperture will be set to T5.6, this will result in all cameras being exposed to a stop more than the manufacturer’s recommendation, this is normally the best balance between Dynamic Range and noise. The film will be exposed at 1/3rd stop over the rated ISO with daylight stock the lens will be set to T4 rather than T5.6 as with all other tests.

Where possible we will record RAW but in the case of some cameras, this may not be possible in which case we will try to use ProRes444 or an equivalent as recommended by the manufacturer. 

A lighting diagram is attached but due to limitations in the software, the 8 * 4 vertical trace frame and 8 * 4 poly sheets either side of it have had to be represented by a softbox.

<pastedGraphic.png>

We will be using an 8 * 4 trace frame 1.2 metres from the subjects. The light will be used on full flood at a distance that completely fills the trace side to side. The height of the light will be adjusted to keep the vertical angle of the light constant.

<pastedGraphic_1.png>

This lighting technique is used to provide a consistent look from all the different styles of light.

In both of the examples above I have used an additional backlight which is switchable between daylight and tungsten. We will use BB&S Pipelines for this in the current evaluations.

Light levels and colour will be established with a Sekonic C800 but checked with a number of other light meters including Cine Meter II, Lighting Passport Spectrometer, Minolta VI, Minolta IV and more :-)

As all the manufacturers claim that their dimming is accurate we will use the lights built-in dimming to establish the correct levels.

—-

This document was accompanied by files containing the camera idents, Netflix’s best-practice setup guides for the cameras under test, and the like.

Three weeks prior to the test Geoff announced that he was going to be undergoing a complex medical operation and likely would not be in attendance, so all this pre-production work let the rest of us jump in and get to work without making too many mistakes.


Adam Wilt
technical services
Vancouver WA USA (no, not that Vancouver, the other one)