Highlight banding issue with HDV

[Peter Barrett]

Hi Graeme, congrats on everything you’ve been up to lately. Yes, I think it’s ok in both a native HDV and AIC timeline, and transcoded to Uncompressed 10-bit HD (although that introduces its own fine mosaic). The material is native in the timeline, ingested via XDCAM Transfer. The primary delivery format for this is SD on Digibeta, so it’s a 10-bit Uncompressed SD sequence I’m using.

At first I blamed the fact that the offline editor had inadvertently set the Video Processing to render only 10-bit at floating point, ignoring HDV’s 8-bit – I selected the “Render all…” option and the problem disappeared – til I rendered.

I think it’s an FCP scaling issue – we’ve found that hardware downconverts from Uncompressed 1080i to PAL tend to be better than either FCP’s time;line renders (which go a bit soft) and Compressor (which often introduces moir

[Peter Barrett]

Ah! Motion filtering! It’s been set to “Normal”, “Best” is no better, but “Fastest (linear)” fixed the problem, presumably at the cost of some theoretical but invisible accuracy. I can live with that!

Thanks Mattdawgpro for the heads-up on that – sometimes the solutions for things are utterly counterintuitive.

Quad G5/NVidia Quadro/Kona LHe/Infortrend RAID/Final Touch HD 2.6.1/FCP 5.1.2/OS X 10.4.8

[Ron James]

[Mattdawgpro] “I haven’t reported this to apple, do you know where/who I would report it to?”

https://www.apple.com/feedback/finalcutpro.html

[Peter Barrett]

I found that RGB rendering clipped anything above 100% – not so good for grading.

Quad G5/NVidia Quadro/Kona LHe/Infortrend RAID/Final Touch HD 2.6.1/FCP 5.1.2/OS X 10.4.8

[Graeme Nattress]

Ah, a bug in the scaling with 10bit. Great find! I’ll have to see if I can dupe the bug here.

Graeme

– http://www.nattress.com – Film Effects and Standards Conversion for FCP

[Jeremy Garchow]

[Graeme Nattress] “Ah, a bug in the scaling with 10bit. “

Yes, it’s either in HD or SD and only happens in white areas of the image hovering at or just below 100ire. The weird pattern does not show up in the canvas, it doesn’t show up in an exported quicktime, only on display out through a capture card. The best educated guess that was explained to me, was that in order to display, FCP converts from YUV/interlaced video to RGB/progressive for display (in essence a real time render). SO there’s something going on with the real time render. If you have a 3 way cc on the clip, you can eliminate the banding by bringing the whites way down (more than I would if color correcting, below a white value of 235 according to the filter). The other way is to force another render by adding the desaturation or Image Control filter, and adjusting them so they aren’t visually doing anything to the video but they cause another render somehow canceling the banding. It works, but it’s a kind of a pain. I have a desat filter favorite now just to cope with it.

If you find it let me know.

Jeremy

[Graeme Nattress]

Output to the card is direct from rendered YCbCr data, and should not have gone through the FCP RGB pathway whichis 8bit only. When you render in 10bit mode, it goes through a 32float YCbCr pathway, and that’s the data that goes to the render file on disc, and also, as 10bit YCbCr out to the card.

Graeme

– http://www.nattress.com – Film Effects and Standards Conversion for FCP

[Jeremy Garchow]

[Graeme Nattress] “Output to the card is direct from rendered YCbCr data, and should not have gone through the FCP RGB pathway whichis 8bit only. When you render in 10bit mode, it goes through a 32float YCbCr pathway, and that’s the data that goes to the render file on disc, and also, as 10bit YCbCr out to the card.”

I understand, but on order for it to display on a computer monitor (i.e. the canvas and quicktime) it’s a real time render from the above mentioned path to RGB progressive. This is why it doesn’t show up in a quicktime movie when you look at it on your computer monitor. I have a sample file if you want me to dig it up. I also shot video of what the pattern looked like when I queried AJA about the problem.

Jeremy

[Graeme Nattress]

To display on screen, yes, it goes through a different pathway that will certainly involve RGB at some point. I think the playback of rendered material or video is different to if there’s an effect and RT gets enabled, but at that point I often get lost off with the complexity.

Graeme

– http://www.nattress.com – Film Effects and Standards Conversion for FCP

[Jeremy Garchow]

Yes, it seems complicated.