Topics

This is a 70 second video comparing the results of using the 5D Mark III's internal codec (ALL-I) with the video from an external recorder via ML's clean HDMI out. The external recorder is the HyperDeck Shuttle and ML is a1ex's Alpha 2 version for the 5D Mark III

As is discussed in the notes on the Vimeo, the current challenges in getting good results this configuration are: 1) the video via HDMI out is 1620x1080, requiring a zoom to get the same 16x9 image as what the 5D records; 2) The HDMI output is at 1080i, requiring deinterlacing; and 3) the output of the HDMI is 8-bit and might be 4:2:2, but could be 4:2:0.

Here's my bit rate experiment with my 600D/T3i taken on a windy day in order to get plenty of movement in the frame. CBR 1.0x and CBR 2.4x are used. The difference in the 500% zooms is very clear, but less apparent at 100%. Still, the CBR multiplier is clearly reducing compression artifacts and, therefore, preserving/recording more of the detail that is there. The avg (max) bit rate at 2.4x was about 105 Mb/s (115 Mb/s).

Vimeo users can download the 400MB file, which is the best way to see this experiment.

Please post any thoughts or observations.

I'm curious as to which setting gives the potential for the "best" (least artifacts, smoothest gradations) recording: QScale or a high CBR.

Having experimented with CBR, I understand that at high values, say 3.0, it might cause the camera to stop recording, especially in sunlight. And, with some material, I've seen improvements with my 600D. But, I'm about to install on a 5D Mark II and am curious if others who might have experimented more than me or understand the CBR vs. VBR issue better than me might weigh in.

It seems that QScale set to -16 might offer the best quality, but I don't quite understand the User Guide's comment "(don't use it!)" Does this mean not to use QScale? If there is a feature, why say not to use it? Would QScale at -16 be better than CBR 3.0?

Thanks!