Server: Cb offset on Decklink output

[CasparCGImport]

Converted from SourceForge issue 3603756, submitted by avanier

YCbCr values are offset by -4Cb for all non-black pixels that should be white on SDI output of Decklink SDI.

Bug found on system running
Windows 7 x64
Core i3-540
GeForce GTX 460 (Drv. V. 8.17.13.142)
Blackmagic Design Decklink SDI (Drv. V. 9.7.0.0) (It was confirmed other applications such as Adobe Premiere output correct levels for same source files)
Tested on CasparCG 2.0.3.3590 / 2.0.3.3760 / 2.1.0.3385 with identical results

Bug exhibited on material originating from original material provided in TGA, flash templates, and pictured alpha masks part of the CasparCG demo file library.

[avanier]

During testing of a project I found evidence that the alpha could have the same offset as well. It was not possible to get full opacity. The bug was not observable on our reference monitors, but cheap tv sets were plugged in and exhibited artifacts suggesting of modified alpha compared the the designed material.

I'll try to test the key output with a pixel analyzer to confirm, but it'll be a few weeks before our production Caspar build swings by the shop.

[avanier]

[hummelstrand]

Thank you for reporting!

[ronag]

Please try this with the BlackMagic application included in the Decklink drivers. I think this is an issue outside of CasparCG.

[TomKaltz]

He has stated he used Photoshop to output directly without problem.
On May 15, 2013 10:56 AM, "Robert Nagy" [email protected] wrote:

Please try this with the BlackMagic application included in the Decklink
drivers. I think this is an issue outside of CasparCG.

—
Reply to this email directly or view it on GitHubhttps://github.com/CasparCG/Server/issues/44#issuecomment-17943735
.

[avanier]

Indeed, other applications did not exhibit the behaviour.

[ronag]

Was it output with key? The key and non-key outputs work quite differently internally.

[moschopsuk]

Any follow up on this issue? We have noticed this on various displays we use.

[avanier]

Unfortunately not on my side. I've just gotten hands back on a decent
scope, but haven't found the time to collect reasonable and up to date
evidence yet.

On Tue, Jun 18, 2013 at 7:30 AM, moschopsuk [email protected]:

Any follow up on this issue? We have noticed this on various displays we
use.

—
Reply to this email directly or view it on GitHubhttps://github.com/CasparCG/Server/issues/44#issuecomment-19605468
.

[ronag]

Fixed in 2.2