HD grading – scale or clip to legal LUT?

[Sascha Haber]

Yeah, well, there is no wrong or right setting for this question.
It strongly depends what you got and where you want to go with it then.
So, lets assume you capture from a HDSR tape. then you need to know if the media was scaled or full on the tape.
If you capture scaled media into full range, you have lifted black levels and lowered white points.
If you capture full into scaled you clip it off.
Same goes for the other way round.
If you have full color levels in your source media and are supposed to deliver a digital product then scaling it is wrong.
But if you are planing to make a tape, then you need to do it.
And excuse my french, FCP is “la merde” for finalizing stuff because its really hard to convince it not to “help” you.
You either want to limit the FCP use to offline or audio fusion and either use Resolve or Smoke for the final product if color levels, integrity and speed are important.

A slice of color…

DaVinci 7.1 OSX 10.6.6
Dual Xeon 2,4 RAM 24 GB
RAID0 8TB eSata 6TB
GTX 285 / GT 120
Extreme 3D+ WAVE

[Gabriele Turchi]

sascha it totally right .

But : if you grade using scaled , than output files without lut , and import to FCP :
over the FCP SDI you will have the same scaled signal …
unless you use a particular codec that FCP mess the gamma ….

g

[Kevin Cannon]

So to be sure that I’m following this correctly:

There are two places to set scaled/unscaled, in the Deck Capture and Playback config and in the Monitoring config. DaVinci internally uses 0-1023 regardless of these settings (the manual calls this “computer levels”) so selecting scaled in the Deck Capture and Playback makes sure it converts 64-1023 to 0-1023 on input and converts 0-1023 to 64-940 on output, and unscaled leaves it alone.

Whereas the monitoring scaled/unscaled only affects monitoring, so if your display device is expecting a 64-940 range, you select scaled so it presents your internally 0-1023 image correctly. But you’re not really “grading” in scaled, just monitoring. Because it’s still 0-1023 internally and I believe exporting a 10-bit DPX without any LUTs will still give you a 0-1023 image… at least that what Nuke tells me when I spot-meter DPX SMPTE bars.

Does that sound right? For simplicity I have all my monitors at scaled to 16-235/64-1023 even though I output DPX and specify 0-1023 (presuming that the i/o facility will scale as necessary).

But I’d be curious if there was a function somewhere (rather than a LUT) that would use the same math for the DPX output to go out scaled…

KC

prehistoricdigital.com
hardworkingpixels.com

[Gabriele Turchi]

that it is correct ,

when set on scaled in monitoring , is just the preview on monitor that scale black and white point (it compress the histogram) and internally is full range .

but it is also true that if you layoff to a deck you will layoff that scaled image (that you used for grading ) so it’s ok .

if you deliver a DPX sequence , that sequence will be full , so you have to make sure that who received will play it back on the same condition out of their SDI (for example flame and smoke are scaled by default )

Even FCP is scaled by default .

if you end up with QT file , QT it scale the image as well …

for example : try to grade using scaled and than render out a QT , that grade using full range and render out a QT … you will see that opening the image on teh computer monitor in QT (except for the little difference between you r comp monitor and your SDI monitor ) the QT graded using scaled is closer to what you saw on the SDI

So in my workflow , i always had continuity grading using scaled , but if you use unscaled , QT , FCP , Flame will show a different image (darker blacks )

i guess that full range SDI need to be used only when going out to film printer ..

g

[Andi Winter]

thank you guys for all the replies!

it is now more transparent for me…