Originally Posted by Joe Bloggs
I thought they used uncompressed connections (uncompressed video) for the UHDTV tests (in trucks or wherever prior to conversion to HEVC or AVC) but you were saying connecting them was taking too many connections (eg. 4x 3G or using 12G).
I was suggesting they could temporarily use those same connections to send compressed video at 100 to 240 fps instead of uncompressed video at 50/60 fps and it may look better, including at the consumer side - on the consumer's UHDTV.
Uncompressed SD-SDI, HD-SDI, 3G-SDI and now 12G-SDI are the standard interconnects for SD, HD, 3D or 1080/50-60p and 2160/50-60p respectively. However for 3D you could use 2 x HD-SDI instead of 1 x 3G-SDI, and for 4K you could use 4x3G-SDI rather than 1x12G-SDI.
Mezzanine encoding could in theory help, but you'd need a standard, and one of the reasons that you use 4x3G-SDI isn't because of a lack of interconnects - it's because the kit you are connecting to is modified HD/3G-SDI stuff so is implementing 4K by re-configuring itself internally to use the 4xHD feeds. So whilst you could bolt on external mezzanine compression boxes to reduce your cabling and routing issues, the core, current, issue of needing 4 channels of, say an EVS, would not be solved, as you'd still currently need to decode back to the original uncompressed formats to feed the inputs, and if you wanted a higher frame rate you'd need even more inputs.
This is because at the moment 4K is experimental, and rather than build and entire range of 4K production gear (as we did for HD), they are (as they did for 3D - which often used 2xHD-SDI inputs and outputs and halved your processing on standard HD-SDI gear), instead modifying existing HD production gear to work in 4K, but by using 4x the I/O and processing capability you effectively end up with 1/4 the amount of inputs, outputs and processing available to you. So you need 4x the kit to do it. If you wanted to double the frame rate the core processing would still increase to 8x (i.e. you'd have 1/8th the capacity), though if you could agree a compatible mezzanine IO standard (as the BBC have internally with Dirac Pro for HD over SD-SDI) you could simplify your routing. Mixers/Switchers and servers are unlikely to be benefit in processing terms though.
Originally Posted by Joe Bloggs
edit: In "SMPTE UHDTV Ecosystem Study Group Report (28/March/2014)" they say something like it (page 21) but in their example they're talking about converting 12 Gbps (I assume for 60 fps) uncompressed to 3 Gbps. I was thinking you could have higher fps uncompressed (100-240) and convert to eg. 12 Gbps compressed.
Yes - you could for interconnects potentially, but processing of images is still baseband (i.e. you need to uncompress the mezzanine compressed content before you process it) and still requires the increased processing. The current requirement for more IO is purely a reflection of this - and the fact that currently HD kit is being repurposed rather than dedicated 4K kit being used. So even if you could use a single 12G-SDI feed to carry 2160/120p content (rather than 2x12G-SDI feeds as a pair), you'd still need twice the amount of 12G processing, or 8x the 3G processing, so an 8 input EVS would drop to a single input (and I'm not sure EVS make an 8 input 3G-SDI box - so you may need 2 x 4 input models slaved together some how)
And the key thing is ensuring that all the different manufacturers agree on a single mezzanine standard and it is cost effective to implement. Some companies will see mezzanine coding as a way of delaying implementing full baseband quality, and being very much a temporary thing, seen as 'second best' once non-mezzanine interconnects kick in, so that could mitigate against it every being implemented. It's a lot of R&D and development costs which may not actually last for very long as it is a stop-gap, and a more expensive stop-gap in some ways than cable bundling.
Unlike HD, 4K specific production kit is still not being produced (with a few exceptions like BlackMagic Design who do offer 4K and 12G-SDI stuff - but at the very low-end), even the camera chains are using HD/3G-SDI Camera Control Units with bolt on pre-processors to split off the 4K video (whilst letting stuff like production talkback, reverse vision, prompter, camera colour balance and exposure control data carry on as if an HD system camera were being used), and the cameras being used are 4K D-Cinema cameras with a bolt on live-broadcasting fibre back, rather than being 4K broadcast system cameras. Sony and GrassValley don't currently have a 4K system camera. (BlackMagic would appear to - but nobody would seriously use the BlackMagic 4K Studio camera in this environment - it has almost non-existent real-world production facilities.)
As I say, the BBC do use Dirac Pro mezzanine compression extensively internally. This is to allow existing 270Mbs uncompressed SD-SDI circuits within buildings and between buildings to carry HD-SDI content at tolerable quality without requiring a total upgrade to the circuits in question. However it isn't perfect, and there is a quality loss (albeit small), particularly since you want very low-delay in mezzanine compression (it has to compress across only a few lines - you can't use frame-based compression as the additional frame delays they introduce are a nightmare in production chains) It's a useful stopgap IF you can implement it.