It was standing room only at the ACM 2012 panel on “Digital Non-Linear Workflows and Centralized Transcoding”. The geeky subject title belies the lively exchange of information between panelists Chris Perry, TelVue’s Senior Systems Engineer, and Daniell Krawczyk of LiveU, and a roomful of access station operators with many format questions to be answered.
Here’s a summary of what was covered in the session:
First question, is which format to use, and when?
Here’s what’s out there:
– Acquisition formats: P2, XD-Cam, AVCHD, DV, DVcam, MPEG2
– Intermediate formats: MOV, AVI, DV, ProRes
– Broadcast & Distribution formats:
- Broadcast: MPEG2, CableLabs MPEG2, HD 19mbps, SD 7-9mbps, SD-3.75mbps
- Web: HD 1.5mbps, SD 500-700kbps
- Archive: mpeg2, H.264
A quick show of hands showed only one person in the entire room was using something other than mpeg2 for broadcast. Intermediate formats like .mov, .avi or ProRes are lossless, but not that broadcastable. So that focused the first part of the discussion on mpeg2 and h.264, the two formats most in play.
“Mpeg2 is a fickle creature”, says Perry, “So you need to make good quality mpeg2, or you’ll have problems down the line broadcast, archiving, or transcoding.” There are so many versions of mpeg2, he added, and an improperly formatted mpeg2 file might work in one place and not the next.
That led to a discussion of what tools people in the audience were using to create their mpeg2 files. It turns out a lot of those present were using Compressor for conversion, and some of them had problems playing that out on their broadcast server.
In case anyone was thinking of switching over to H.264 for broadcast, Perry cautions that at least one major cable operator has made it clear they will be staying with mpeg2 for at least a decade. It is the least expensive and most developed of available broadcast codecs.
H.264 is primarily a webcasting medium, often encapsulated in rtmp, as LiveU does, and used for online VoD. H.264 packs data together, but takes a long time to encode and decode, taking 4x more computing power to encode and decode than mpeg2.
There is an H.265 codec on the horizon. A revision of H.264, it will take 5x more computing power to encode and decode. At that rate, we will soon reach a point where we can’t keep up with our ability to transcode.
So what is the proper balance of formats? Perry recommends thinking of mpeg2 for “near-line” archive and broadcast, and H.264 for online material and long-term archive.
A word about switching back and forth between formats. “The primary goal is how to integrate all formats into an easy-to-use layer”, Perry explains. “There is a point where you can exchange between formats and not lose quality.”
CableLabs mpeg2, for example, can do that: it serves HD at 19 mbps and SD in 3.75 mbps. There is an even quality exchange between the two mediums, so you can shrink a file to H.264 by factor of four for storage and keep video quality.
Turning to the question of how to set up your operation for most efficient transfer of video files: investing in quality network infrastructure makes possible everything else talked about here.
Perry described an ideal setup as being like an hourglass, where everything goes through single point. In some ways this might be counterintuitive, because we generally think of that kind of structure as being vulnerable to a single point of failure. But as the format explosion continues, the hourglass model begins to make more sense.
The best example of this is gigabit switches either stacked or trunked (4 gb lines together). A stack is two switches on rack with hdmi or other hard connection, which means no latency between switches. Perry’s advice is to buy the biggest most advanced switch you can afford. Wireless will never replace throughput on copper. Copper is more reliable and robust. This starts to matter when doing realtime IP video.
Perry also recommends a big NAS array. In response to a question from the audience, the alternative, SAN, is fantastic, higher speed, but harder to manage – like a house that is constantly being added on to. NAS is like a single house that will never get bigger. But it can handle connectivity better over the public Internet.
For transcoding solutions, Perry noted how Telestream Episode is emerging as a leader in the pack of new transcoding solutions. He walked the audience through its three-step workflow, and how you can automate several output channels from a single source file. Highly effective.
The session closed out with a discussion of Aspect Ratio conversion, which is a headache for many stations, currently operating in a hybrid SD/HD environment.
Most of audience, it turns out, is shooting more than half 16:9 footage, but few are broadcasting 16:9. A lot of people end up crushing the anamorphic video to fit their 4:3 feed. The question in this case is consistency: if we shoot in anamorphic, do we want to change everything to anamorphic? What if you have to integrate a 4:3 satellite source into a 16:9 workflow? TelVue servers can fix that problem automatically, if all you do is broadcast. Or you can choose a transcoding solution that creates a uniform output. More on that in the TelVue User Group tomorrow, at the ACM National conference.