What format is most requested?

[Brian Drysdale]

My understanding is that, to date. the problem is gathering that dollar from each of those viewers.

[Galen Rath]

Alister, good statement about the future.

Please enlighten me. Is 50Mbps 10 bit codec better straight out of the camera, or is it better because you can produce something better with it in post?

[David Heath]

Originally Posted by Alister Chapman

It's also in part why Mpeg 2 in the form of XDCAM HD422 continues to hang on as a broadcast codec ......... even though it's "only" 8 bit and a somewhat dated codec by todays standards, it looks like XDCAM will still be around in another 5 years, even though by then it really will be old technology.

Firstly, fully agree with all the early points. For a big organisation, such as a camera re-equipment is a big investment and they will want to get the return on the assets. Just because something isn't the newest or most advanced doesn't mean it's not capable of doing the job - though it may mean it wouldn't be what they would buy now. Let's think ourselves lucky we don't work in such as the rail industry - rolling stock is expected to carry on working for many more years than camera equipment!

But I'm less sure about calling MPEG2 a "dated codec", and it all goes back to what a codec really is. A means of coding a signal to maintain quality whilst trying to reduce datarate. And three things can be traded against each other - quality, bitrate and complexity (or computing power).

And the latter often gets ignored in debates. Keeping up quality whilst reducing bitrate is nearly always seen as a good thing, discussions often ignoring that it must involve extra processing. There is no denying that very often a highly efficient (in terms solely of bitrate versus quality) codec is indeed worthwhile - H264 is well worth the effort in such as broadcast TV where it's bandwidth that is at the real premium. But in many other cases then using a lot of extra complexity to squeeze a relatively small extra bitrate reduction is pointless - and that may well apply often to acquisition. If you were buying a car and I offered you a £5,000 addon to save 20% of your fuel costs, then would you take it? Answer will depend on your annual mileage, if you don't do that many miles it's probably more cost effective to pay for the extra fuel, but not spend the £5,000.

I recently linked to this - XAVC Sonys implementation of Advanced Video Coding - Sony Community - 36571 - in another thread, and it's very relevant here. (It shows many of the graphs about codecs that Sony were showing at a recent F5/55 roadshow.)

Scroll down to figure 5 and it shows how quickly a given computer will process various codecs. It's pretty obvious that in this respect MPEG2 based codecs (XDCAM422 here) are the clear winners in terms of speed of processing, compared to all the H264/AVC ones. I have to say the amount came as a surprise to me - I'd have previously guessed that the workload to decode AVC in an I-frame only codec would be roughly comparable to the interframe compression in XDCAM. The graphs show that not to be the case, interframe will need more work than I-frame only, but nowhere near as much as going to something AVC based.

Those graphs come from a presentation to introduce the new Sony codec coming with the F5/55 - XAVC. They were at pains to point out that XAVC was seen in no way as a replacement for XDCAM and XDCAM422 for very much this reason. XAVC enabled extra possibilities - such as 10 bit and 4k resolutions - but at the expense of a lot of extra processing. As such, XDCAM and XAVC should not be seen as rivals as much as different horses for different courses. Just as RAW should be seen as a third horse.

To Galen - once you start talking about 10 bit, then (all else equal) don't expect to see any big difference in basic video as recorded. Any advantage is likely to be that it will allow greater manipulation later - assuming the fundamental quality of the image from the front end makes it worth it.

I suspect Alister is referring to a hypothetical extension to the AVC-HD spec - which would allow for 10 bit, 4:2:2, and better than standard AVC-HD compression. Theoretically, what he says would be correct - but the downside will be interframe *AND* AVC processing is likely to really up the computing power it will need.

Would it be worth it? It may be argued that it would hit a sweet spot - personally, I think it falls between the stools of XDCAM and XAVC. Once 10 bit and 4:2:2 start to really matter to a user, then it's probably best to stick with I-frame only. Yes, that will obviously mean a higher bitrate (100Mbs, so twice as much) but if 10 bit etc is really important, then the doubling of file size is less likely to matter.

And for a user who just wants basic broadcast standard and isn't worried about 10 bit, but is more worried about file size etc - that's where XDCAM is likely to remain optimal for a considerable time to come. My personal feeling is that a 50Mbs AVC-HD is more likely to give you the worst of both worlds rather than the best, but who knows what the manufacturers are cooking up at the moment?!

Obviously both memory cost is going down and computing power is going up. The former leads to file sizes being less of a problem, the latter to more complex codecs becoming ever more viable - taken together they are likely to see a steady growth in such as XAVC, and maybe less need for 50Mbs AND AVC? We'll see.