UHDTV – What are you waiting for?

Submitted by Jim DeFilippis

You just got back from the NAB. You are recovering from all the information that poured into your head regarding the latest technology, industry insights and solutions to the everyday problems of your station or facility. Let’s try to sort out all the different messages you heard and make some sense of the state of UHDTV routing and distribution.

One nagging problem is the core of the facility, the plant router. The current router is maxed out; or perhaps it’s a ‘green field’ situation but you are stuck with conflicting information and requirements on new designs. There’s pressure to expand or replace the plant router but what technology should you consider? Is SDI still viable or is it time to look at video over IP? If video over IP seems to be the right choice, which of several approaches to IP will be viable over the lifetime of your facility (or your career!)?

?

A further factor to consider is, while there may not be an immediate need to upgrade to UHDTV (4k) or HDR(PQ, HLG) or wider color (ITU Rec 2020) it’s not a certainty that in the not too distant future the new router will have to be modified or retrofitted to handle one or more of these new formats.

You have been hearing about video over IP. You probably know that there are several largely incompatible approaches each with their own supporters such as the Sony Networked Media Interface (NMI), the Alliance for IP Media Solutions (AIMS), and the Adaptive Sample Picture Encapsulation (ASPEN) Community to name but three. Then there are the standardized interfaces such as SMPTE ST 2022 and VSF TR03 and TR04. Each support transport of 3G/s uncompressed or compressed video signals over 10G Ethernet infrastructure.  To increase capacity and to ‘future proof’ installations today, vendors are recommending the use of mezzanine compression (JPEG 2k LL, TICO, NMI, VC 2) over 1G/10G Ethernet fabric.  This is a lot of new technologies to absorb, analyze and make a decision for the next 5 to 10 years.

Before jumping into the deep end of the IP pool, let’s take stock of the current state of SDI. A short history lesson first. Serial Digital Interface was developed in the early 90’s to fit a need; the ability to move digital video overcoax, in real time, between production equipment and monitors. At this point in time SD digital video required 270Mb/s (component ITU Rec 601). The design spec was to reliably interconnect video ‘sources’ to ‘sinks’ overcoax cable at distances that were typical for analog video (100m/300ft).

Why coax? Well the existing video infrastructure was coax cable and since digital video was a new and not a widespread format, it was being implemented in stages or islands. With coax already in place, it made sense to build a digital transport method that used coax.

Over time, SDI expanded to include more functionality. Need to transport audio with the video? The SDI standards were amended to embed audio in the horizontal ancillary data space (hanc) {which used to be the horizontal blanking interval}. Need to transport other program related data such as time code or captioning? Again SDI standards were amended to embed data in the vertical ancillary data space (vanc){which used to be the vertical blanking interval}.

In the late 90s HD TV became the new challenge. Component analog HD was a non-starter, too difficult to manage and keep pristine except for short distances. So SDI got a fork lift upgrade and moved up to 1.485 Gb/s (5.5 x SD-SDI). Again, the design spec was to re-use existing infrastructure, coax cable up to 100m.

In the last 10 years SDI moved up to 3 Gb/s , then 6Gb/s and now (2015) 12Gb/s. 3Gb/s was in response to applications such as 1080p50/60 and/or stereoscopic HD (3D). 6 Gb/s applications include dual link 3Gb/s (D-Cinema production and 1080p60 stereoscopic), while 12 Gb/s can support quad link 3G or 2160p/10bit at 60fps(UHD-1). All these SDI formats support synchronous real time transport of video over 100m distance, including embedding of audio channels and program related data (TC, captioning, etc.), in real time (no intrinsic latency other than the speed of light).

Further, 3Gb/s, 6Gb/s and 12Gb/s SDI standards include the ability to transport multiple lower SDI speeds when combined and mapped (channelized) into the higher speed SDI. Thus 12 Gb/s can transport (2) 6Gb/s or (4) 3Gb/s or (8) 1.5Gb/s signals.

Well back to the here and now, the question to be resolved is what are the choices when upgrading a current SDI routing system or designing a completely new facility? As I’ve outlined above, SDI has grown and continues to be developed in response to the needs of video production and distribution. It is optimized to work over the existing infrastructure of coax cable while providing for the synchronous transport of video, audio and associated data .

So how do the ST 2022/TR 03 and 04 standards stack up? There are some definite advantages to distributing and routing video over IP but there are also limitations and tradeoffs. First it’ll require a whole new physical infrastructure as ST 2022 is not specified for transport over coax. 100m distance over copper, yes for 1Gb/s but although possible not so common at 10Gb/s. That’s ok because optical fibers can certainly go 100m or further at 10Gb/s, but which fibre type and which fiber optic connectors to use?. Current video over IP standards do not specify the type of fiber cable or recommend optical fiber connectors. The physical layer is up to the adopter to decide and figure out whereas SDI is fully standardized for use of common 75 ohm cable and BNC connectors.

Which IP format to adopt? Is the ASPEN approach using MPEG2 TS packets over IP the right choice? Or is the AIMS approach, which uses either ST 2022-6 encapsulating SDI into IP packets or the VSF TR03 / TR04 separate essence packetization approach? What about the control layer, how will these packets be routed over the IP switch fabric? Not simple as a ‘crosspoint’ per source per output, but the need for router flow control that can insure packets of video arrive at the right port at the right time, in order, without loss or excessive delay. Each vendor has their own ‘secret sauce’ to orchestrate the transport of the video packets over the IP fabric, so a key decision will be which one to adopt and embrace and will this solution be supported over the lifetime of the facility?

Another consideration is the quality of the video switch? With SDI, we have choices of switches. Synchronous SDI switches have been developed and are in use every day in your facility. In less sensitive applications, a SDI non-synchronous switch is acceptable with a minimum recovery time (typically less than 1 TV line). The proposed video over IP switch protocols require coordination between the source and destination to establish the round-trip transit delay and to pre-synch the sources prior to the ‘clean switch’. The clean video IP switch may also require twice the bandwidth during the ‘synch’ period. A video IP non-synchronous switch is possible but the recovery maybe several TV lines or perhaps cause a frame repeat/skip.

Timing and synchronization.  SMPTE has developed a new timing and synchronization approach using IEEE 1588 Precision Time Protocols over IP. This new standard (ST 2059) supports legacy synchronization signals (black burst, DARS,) so is applicable for both SDI plants as well as video of IP facilities. However if the choice is video over IP, only the new synchronization approach is applicable. So in addition to replacing all the coax cable, all the timing and synchronization equipment needs to be upgraded.

Equipment interfaces. While all modern broadcast production equipment has an Ethernet connection (RJ45) few if any have a 1G Ethernet interface let alone a 10G Ethernet connection. So today to implement an IP switch fabric in a facility, there has to be SDI coax (BNC) to IP converters. There are such devices, called SFP’s (Small Form-factor Pluggable) transceivers that can convert to/from SDI to IP (for example Embrionix makes several SDI types of SFP’s, BNC, F/O, 3G/6G/12G).

However these converters add to the overall cost of a large router, given that there are few broadcast equipment types ready for direct IP connections. Where there is gear that can interface directly to IP, the connection type is not standardized for video over IP use. In fact most of the equipment supporting 10G is via a SFP auxiliary port. 10G Ethernet physical layer cable and connectors include:

10GBaseT: Cat 6a or 7 copper cable with RJ45 connectors.

Distances up to 100m. (FYI, distance dependent on quality of installation, shielding and quality of the final terminations)

10GBase-SR: Short range multi-mode fiber with SC or PC type connectors. Distances up to 300m

10GBase-LR: Long range single mode fiber with SC or PC type connectors. Distances up to 10km

UHD-TV and the future. How does each approach, SDI v. IP, compare when needing to move up to UHD?  Let’s start out with what I’ll call UHD 1.1: 3840×2160 p50/60Hz, 10bit, ITU Rec 2020, 4:2:2, SDR. That’s a base data rate of 12Gb/s (actually 11.88Gb/s).  SMPTE ST 2082-10 single link 12G SDI handles that or SMPTE ST 425-5 for QUAD 3G SDI can also be used. On the IP side, 10G Ethernet is not able, to carry uncompressed UHD 1.1 (2160p50/60Hz, 10bit, 4:2:2). To carry UHD 1.1 on an IP infrastructure would require a step up to 25G Ethernet when it becomes available or use quad 10G Ethernet, which is known as 40G Ethernet.

It should be noted however that currently the only standardized and therefore interoperable way to connect any two pieces of UHD 1.1 equipment together are the SMPTE SDI standards – there are no standardized IP interfaces or protocols available for the carriage of UHD signals.

The next level UHD 1.2, supports either an increase to 12bit per sample (4:2:2 or 4:4:4) or increase the frame rate to 120Hz (10bit 4:2:2) or both (UHD 1.3). UHD 1.2 needs double the bandwidth of 1.1 or 24Gb/s (23.76Gb/s). For SDI there is dual link 12G SDI (SMPTE ST 2082-11). Unfortunately 25G Ethernet is not quite able to support UHD 1.2 as IP switches generally cannot handle data rates more than 80% of the line rate so 40G Ethernet is required (quad 10G Ethernet). Finally if we consider UHD 1.3 for 120Hz at 12 bit, 4:2:2/4:4:4, this requires up to 48 Gb/s. Quad 12G SDI (SMPTE ST 2082-12) can be employed today while the equivalent IP connectivity required is 100G Ethernet, which is only just starting to appear in data centers today so is still very expensive and not so widely deployed.  IP is definitely a ‘future’ format when considering 4k production ecosystem choices today.

How about 8k? Also known as UHD 2, with 7680×4320 pixels at 50/60Hz, this is also do-able as quad link 12G-SDI today, but as with UHD 1.3, will require the adoption of a future 100G Ethernet standard.

Here is a summary table:

To be complete, let’s consider the use of mezzanine compression to help reduce the raw data rate. Assuming mezzanine compression rather than base-band UHDTV routing is acceptable, and the associated cost and end-to-end processing latency (of mezzanine compression), can be accommodated, then you have to choose from amongst several candidate compression techniques that have now been proposed and are in the process of being documented in SMPTE.

While the mezzanine solutions are standardized at this time interoperability is not guaranteed. The JPEG XS Working Group under ISO/IEC has generated a call for technology and expects to standardize a mezzanine compression by 2017, the goal of which is to have a common interoperable mezzanine compression standard that can be used for such applications.

There are paths to the future with either SDI or IP, however the SDI roadmap is better defined with interface technology that is available today for use at any of the UHD levels that we have discussed.

Conclusions.

While IP is a developing solution for transport of real time video over IP, there are many parts that need to resolved and standardized.  Currently available SDI interfaces (3G/6G/12G) have been standardized and can cover up to 48Gb/s applications today and into tomorrow.

Should one consider hybrid approaches? Yes, where SDI is not practical, such as for transport of video to a data center, which is all IP or where the telecom infrastructure only supports IP protocols (MPLS). It may also make sense to have IP for applications which are more computer based or for large multi-viewer installations.

But for video production, especially beyond HDTV rates, SDI has been developed from the ground up to support live video production, routing and interconnection. It has evolved to the latest version at 12Gb/s, more than sufficient for all current production formats and extensible to even 8k! While there are good arguments for use of IP technologies, IP is not a universal panacea. As the Brits say, “Horses for courses”, meaning each technology (horse) has its sweet spot of use (courses).

Our advice: listen, learn and educate yourself about the pros and cons of each approach. Match up your requirements with a ‘best fit’ today and ‘best guess’ as to where the future may lie. Your solution will be that technology with provides the best coverage at the least total cost of implementation, including any ‘learning curve/teething pain’ during the rollout.

Let’s compare SDI v. IP in terms of desirable traits: