Matching Red One in Autodesk Softimage

[Julian Johnson]

I'm trying to match some Red One footage to a camera in Autodesk Softimage. Unlike Maya, Softimage has no option for Film Gate or Resolution Gate - it simply has a Projection Plane (Film Back) and Focal Length. Normally, to match a camera in Softimage I would get the physical sensor size, use that as my film back and then input the supplied focal length. It usually works! If I need to effectively gate that image then I can either alter the focal length or the field of view from that point onwards.

In Red, I'm not sure that the process is exactly the same. Although the sensor is, according to the website 24.4mm x 13.7mm, the captured image is 4096 x 2034 i.e. a subset of the overall sensor's maximum pixel array (4900 x 2580).

If I overlay my plate at 4096 x 2034 and centre it in an image of 4900 x 2580 have I reconstructed the image as projected onto the sensor as a whole?

If so, then if I construct my camera in Softimage to have a film back of 24.4 x 13.7 and input the focal length of the actual shot (i.e. 20mm) and then project my overlaid image through the camera I should have an exact match with scene geometry - i.e. a 1m cube placed 6m from the camera will line up exactly (provided the orientations and positions of the camera/cube are all precisely measured on set).

I guess the question is fundamentally: in Red One is the relationship between focal length and CCD size as it is with a normal camera or is the focal length possibly related to the captured area as opposed to the total CCD size? At the moment I can't get a match between the set measurements and my scene geometry - either the measurements are wrong or my camera setup is wrong!

Cheers,
Julian

[Juhan Malmstein]

If you just calculate the difference (4096px/4900px)x24,4mm you get that the area the image is recorded on should actually be 20,39*10,8mm. I might be totally wrong, but it's definitely worth a try if the difference is just in changing 2 numbers...

[Deanan]

Multiply 5.4u by the number of pixels and divide by 1000 to get mm.

[Erich Ocean]

Multiply 5.4u by the number of pixels and divide by 1000 to get mm.

Is that for the M sensor, or the MX sensor (or both)? Also, how accurate is 5.4u? Is there a more exact number (e.g. 5.39678u)?

[Jeff Kilgroe]

Focal length is just that and focal lengths are not altered from one type of camera to another, what changes is the image plane or imager size, film back, sensor, etc.. You just need to use the proper dimensions. Same within Maya, the film gate or resolution presets there are sometimes a problem as they don't always match specifically with the format, so you have to be careful. In SI, you just need the dimensions of your recorded image area.

The RED One, both the original Mysterium and Mysterium-X sensors, have square pixels and they are 5.4um in size, right on the nose 0.0054mm. So to elaborate on what Deanan posted, you just multiply that dimension by the number of pixels.

For example, RED One 4K 16:9 would be:

4096 * 0.0054mm = 22.12mm
2304 * 0.0054mm = 12.44mm

So 22.12 x 12.44 mm is the dimension you would use if the source imagery was recorded at 4K 16:9 on the RED One.

[Erich Ocean]

The RED One, both the original Mysterium and Mysterium-X sensors, have square pixels and they are 5.4um in size, right on the nose 0.0054mm.

Thanks Jeff! And I'll add, since the Epic also uses the MX sensor, the same measurements apply there too.

[Julian Johnson]

Thanks for the detailed explanations and suggestions - it's been super helpful and informative. I'll let you know how I get on.

[Gavin Greenwalt]

h/v-aperture in the r3d metadata would be very welcome. Especially once the 2/3", S35 MX, REDOne-MX, REDOne, FF35, 645, FF35 -2k, 3k, 4k etc etc etc... start coming out.

It's always annoying to have to go scrounging around the RED.com website to look it up when match-moving.

[Erich Ocean]

h/v-aperture in the r3d metadata would be very welcome.

Great idea!