I've been routinely importing XDCAM EX files, then transcoding to DNxHD so that I can output from MC via Quicktime Reference. (XDCAM EX files won't export with QTRef)

This weekend, I decided to compare the original XDCAM EX file with the transcoded DNxHD file. The comparison was made by putting each clip on a different track, applying a chroma and luma inversion to the bottom clip, then applying a 50% Blend to the top clip. I was surprised to see a difference.

This begs the question, am I losing some quality by transcoding to DNxHD? Would the final result be of higher quality if I exported directly from the XDCAM EX track to my final format(QT Movie)? Obviously, if I'm exporting to a compressed format, like MPEG2, I'll be giving up quality anyway, but, I beleive the final result will be of higher quality when coming from a higher quality source.

And a follow on question: When importing XDCAM EX files, does the importer care what color space is selected, RGB or 709? I believe XDCAM EX is imported independently of these options.

Quality degradation has not been experienced at all here when transcoding XDCAM EX files ro DNxHD - tested extensively early on w/ wipes, etc. Not sure if your method is going to show degradation - and what specifically is it that you see? What flavor DNxHD are you transcoding to?

There are no parameters to set for the import of XDCAM files, only once you transcode them w/in Avid.

- Keith 

Attached is a frame grab of the difference image. The original image is from some tree branches and leaves. I transcoded to both DNx145 and DNx220x. Note: this isn't a grey image...look closely. The difference looks much less significant as an exported JPEG than it did on the timeline record window.

Would you explain why a difference 'blend' would show degradation? Could it not just show the 'difference' when applying effects to different media?

bonang:

Would you explain why a difference 'blend' would show degradation? Could it not just show the 'difference' when applying effects to different media?

Because I applied a chroma and luma "INVERT" to one track. This, essentially, creates a negative that, when summed with the positive, will cancel itself out....except for the differences.

cuervo:

bonang:

Would you explain why a difference 'blend' would show degradation? Could it not just show the 'difference' when applying effects to different media?

Because I applied a chroma and luma "INVERT" to one track. This, essentially, creates a negative that, when summed with the positive, will cancel itself out....except for the differences.

If you have scopes, try a split screen w/ native XDCAM 35 and a transcode - here, at least, there is nothing whatsoever visible.

The difference check is valuable, but I wouldn't perform it within Avid - I suspect the way it processes effects and works with media of different formats could play against it. I'd export uncompessed still TIFFs of of a frame of each codec and then do the comparrasion in Photoshop.

I did something very similar a few years ago with Standard Definition - 1:1 against various Avid resolutions.

Avid Resolutions Compared

But this is a mathmatical comparision - the splitscreen or and A/B test on a good quality monitor is possibly more beneficial in many ways.

DNxHD 120 (or 145) is obviously a lower bitrate, so it will be more lossy than higher bitrate versions, but the perceptibility of that difference is important. To my eye, I've not been able to identify a difference between 120 and 185 - and will choose the codec based on my needs. For video that will be ending up on another highly compressed medium (HDV, XDCAM HD or even HDCAM) I will choose a higher bitrate codec to maintain the highest quality I can for further compression (and avoid compounding compression artifacting). Otherwise I may find that DNxHD 120 is sufficient for my needs.

The actual measurable difference is small and the visual difference is virtually non-existant as far I can tell.

Thanx, guys. In practice, the visual differences are not easily discernible.

Bonang...no FX applied other then the "inevert". There may well be inherent differences incured because of this.

Sycophant...

I repeated the diff test and could see no diff between DNx145 and DNx220x. Regardless, as you say, "the diff is small and, visually, non-existent.