HD Testchart: tiff to video

Belle Nuit publishes an interesting looking HD Testchart here. This tiff file must be transformed to video for use. I wanted to make 8-bit uncompressed 4:2:2 and tried to use FCP to do it.

The RGB values in the tiff must be understood as R'G'B' values or else the demarcations at 16 and 235 in the Testchart make no sense. A grey patch in the tiff with R'G'B' values (x,x,x) should become a grey patch in the video with Y'CbCr values (x,128,128). Colored patches in the tiff should become colored patches in the video with Y'CbCr values according to the BT.709 equations. BT.709 actually reserves values 0 and 255 for synchronization signals, so there must be tiny deviations -- truncation or shrinking -- in going from the R'G'B' values to the Y'CbCr values. Let's not get hung up on these.

I imported the tiff into an HD 8-bit uncompressed 4:2:2 sequence. I found that for FCP to make anything like accurate video from the tiff, gamma must be set to 2.22 under FCP User Preferences>Editing>Imported Still and also processing of maximum white must be set to Super-White in the sequence settings. Nevertheless, the video loses all the sub-blacks of the Belle Nuit Testchart. FCP makes no Y' less than 16 and it linearly scales the expected 0-255 range into 16-254. Level 16 black in the tiff becomes Y'=31 black in the video. Etc. Consequently almost every R'G'B' patch of the tiff is wrongly represented in the Y'CbCr video from FCP. Belle Nuit included the sub-blacks and super-whites with reason. Apple's own white paper "Uncompressed Y'CbCr Video in QuickTime Files" recognizes the use of values 1-254 in the codec.

Even making setup negative in the Proc Amp filter doesn't get the sub-blacks. FCP doggedly protects its 8-bit uncompressed 4:2:2 from Y' values less than 16. That is, FCP 7 does. Does freshly written FCP X do it better?

Dennis Couzin
Berlin Germany

NOTE ADDED LATER: Oops! The BT.709 equations preclude Y' values below 16 (because E'Y can't be less than 0). They much less strongly preclude Y' values above 235 (because the gamma-like correction in section 1.2 can easily be extended to accommodate L>1 which then can yield E'Y>1). So it's understandable that FCP countenances super-whites but not sub-blacks. Belle Nuit must be contemplating non-BT.709, wide-range video with their HD Testchart.