16 BIT REDRAW

[Bob Gundu]

The mystery of REDRAW reminds me of the scene from Spinal Tap as they discuss drummers dying mysteriously.
http://www.youtube.com/watch?v=tZrqC5LL_oo

[Peter Moretti]

I think anyone who takes a few minutes to do a little bit of maths realises the files are not 16bit linear (in iany traditional sense)... If any of it matters at all ...it is the question of 'are redraw files linear?'... And as far as I can tell nobody at Red has every wanted to confirm this.

Michael, I distinctly remember hearing when the R1 came out that Red files are 12-bit linear. I can't point you to quotable source at the present time, but I might be able to find a reference somewhere. And I certainly don't know what changes may have been made to the .R3D structure to accomodate the Epic and Scarlet cameras.

[Les Dittert]

I'd love to see how they get 16 bits out of a 5.4 micron pixel ;)

[Jon MIchael Puntervold]

Sebastian,

I've just spent some time working on this, and I believe these are the uncompressed rates for RAW 5K (5,120 X 2,700):

8-bit: 18.5 GB/min.
10-bit: 23.2 GB/min.
12-bit: 27.8 GB/min.
14-bit: 32.4 GB/min.
16-bit: 37.1 GB/min.

So I have to agree with you that at 3:1 compression (as I understand it) the bit depth of your .R3D file would seem to be closest to 12-bit.

Are these number based on raw sensor data, or finished uncompressed images at 5,120 X 2,700 resolution. If you calculate from raw sensor data, how do you find the data rate from the MX-sensor?

[Michael Lindsay]

Yes I am pretty sure the Red one encoded and created a 12bit linear file... But I don't see how that precludes epic creating a nonlinear file. Unless you know of a way to get to the picture in a epic file without the red sdk or redcineX?

Might be wrong but my guess is 12bit nonlinear otherwise Red wouldn't have made such a big deal about increased precision...

Pictures look good either way ;-)

[Peter Moretti]

Are these number based on raw sensor data, or finished uncompressed images at 5,120 X 2,700 resolution. If you calculate from raw sensor data, how do you find the data rate from the MX-sensor?

Those were calculated assuming RAW sensor data at 5,120 X 2,700. By that I mean it assumes a single color value per pixel. If it were uncompressed 4:4:4, the rates would have to be multiplied by three.

These are the rates that I calculated for the MX sensor's 4K (4,096 X 2,304) uncompressed RAW data assuming various bit depths:

8-bit: 12.7 GB/min.
10-bit: 15.8 GB/min.
12-bit: 19.0 GB/min.
14-bit: 22.1 GB/min.
16-bit: 25.3 GB/min.

Hope that helps, and that my math is correct :).

[Peter Moretti]

Yes I am pretty sure the Red one encoded and created a 12bit linear file... But I don't see how that precludes epic creating a nonlinear file. Unless you know of a way to get to the picture in a epic file without the red sdk or redcineX?

Might be wrong but my guess is 12bit nonlinear otherwise Red wouldn't have made such a big deal about increased precision...

Pictures look good either way ;-)

Michael, your suspicion may of course be confirmed. But I think the increased precision comes from internally processing the sensor data at 16-bits, instead of 12-bits like I believe the Red One does. But you definitely said it best:

"Pictures look good either way ;-)"