Multicast to deliver HDTV & digital radio at higher quality than DAB+

Multicast, or IP Multicast to give it its full name, is a disitribution technology for live TV and radio Internet streams, and although it may not sound very exciting, it's the key technology that makes it feasible to broadcast live TV channels and radio stations at high quality to broadcast-sized audiences over the Internet.

The key feature of multicast is that it is as bandwidth-efficient as it is possible to get, whereas the exact opposite can be said of unicast, which is the current distribution technology used to deliver live Internet streams -- see the unicast vs multicast section below for more details. And whereas on unicast the cost of bandwidth is a significant issue to the broadcasters, on multicast the bandwidth costs are almost zero, which means that the broadcasters can use high bit rates for TV and radio streams, and the quality can be far higher than has been the case until now. A good example of this is that Virgin Media subscribers will be distributing an HDTV channel called VoomHD via multicast later this year when Virgin launches its 50 Mbps cable broadband package, and it certainly isn't feasible to distribute HDTV channels to large audiences at the moment using unicast, because HDTV channels use bit rate levels in the region of 8 - 15 Mbps. And considering that HDTV channel bit rates are about 60 - 120 times higher than the 128 kbps bit rate needed to provide good audio quality on radio streams when using modern audio codecs like AAC, it is effectively guaranteed that multicast radio streams will always provide higher audio quality than DAB+.

BBC's multicast trial

Although the multicast radio streams are still technically in trial status at the moment, people who're on one of the few ISPs that are participating on the BBC's multicast trial can already receive them today. But as far as them "going live" is concerned, they were meant to be launched last year alongside the iPlayer, but they were then held back. But James Cridland, the Head of Audio & Music at BBC Radio, has recently said the following on a BBC Radio Labs blog:

So although I very much doubt there will be any fanfare, I would interpret that to mean that the multicast radio streams will be available to Virgin Media subscribers once their 50 Mbps broadband package has been launched -- whether new equipment will be needed in order to receive them remains to be seen, though.

ISP support

The thing that's been holding multicast back until now has been the lack of big ISPs supporting it. However, as mentioned above, Virgin Media will be supporting multicast, Tiscali's network is already configured to support it, as is Sky's network. But by far the biggest development is that BT has demonstrated delivering video over multicast on their 21CN (21st Century Network) Next-Generation Network (NGN) to a group of assembled journalists recently, and Sam Crawford, who's the person behind the Samknows Broadband website, said that he would "be massively surprised if [BT] didn't release [multicast] or at least announce it this year".

The reason why BT supporting multicast is such a big deal is that 61% of broadband users get their broadband via BT, even though most of them pay a different ISP for the privilege, because as well as the 27% of broadband users who actually subscribe to BT, there's also the "retail" or non-LLU ISPs which pay BT to provide ADSL to the ISP's customers, and the LLU (local-loop unbundling) ISPs that want to provide a national service who pay BT to provide ADSL to customers on exchanges that the ISP hasn't unbundled.

And the problem up to now has been that because BT's current "20CN" network doesn't support multicast, any multicast streams that BT would handle for the other ISPs would have to be split up into unicast streams prior to being sent to BT, which defeats the whole point of using multicast, because it means that the ISPs wouldn't save any bandwidth on the streams they send to their customers via BT. Therefore, BT supporting multicast would be the final piece of the puzzle needed to allow widespread support of multicast by all of the ISPs -- it doesn't necessarily mean that they will support it, but it does at least remove the major obstacle in the way of them supporting it, and as TV and video traffic over the Internet is expected to approximately double in volume year-on-year between now and 2011 they would be wise to support it because it would ease the flood of TV traffic they're going to have to handle anyway.

Also, with the addition of BT's 21CN, four of the biggest six broadband providers in the UK would have networks that are configured to support multicast, all of which are already or are planning to provide IPTV services, and Orange is also planning on launching an IPTV service, so it will probably support multicast as well, in which case 74% of all broadband users could be enabled to receive multicast by the time BT has rolled out its 21CN nationally between now and 2010.

It's possible that Tiscali TV already uses multicast to deliver live TV to its subscribers, although there's no information to prove this, but there is one company which is already using multicast to deliver TV channels is called Inuk Networks, which is multicasting the Freeview TV channels over the JANET university network to students living in halls of residence, and 40,000 students at forty different universities have signed up for the service since it was launched in September 2007.

HDTV, SDTV & digital radio at better quality than DAB+

Digital radio at higher quality than DAB+

The first channels on the BBC's multicast trial to be made widely available will be the radio stations, and the following table shows the bit rates and audio codecs that the BBC and the commercial radio groups have been using for their streams on the multicast trial up to now:

1 - The BBC was using 128 kbps AAC for its trial multicast streams until recently, and the Real streams may actually be using 128 kbps AAC (because AAC is supported by Real) so the BBC is likely to use 128 kbps AAC when the multicast streams launch

All of those formats (possibly with the exception of the Real streams, because I don't know which of the codecs that Real supports is being used) provide as high or higher quality tha you would ever expect DAB+ to provide, and the bit rates and therefore the quality of the multicast streams are only likely to increase over time. The reason why the bit rates are only likely to increase over time is that multicast is inherently so bandwidth-efficient, and also because the cost of Internet bandwidth for content providers like the BBC has fallen in-line with Moore's Law since the mid 1980s, so you would expect it to continue getting exponentially cheaper and exponentially more plentiful in the future as well -- the cost of Internet bandwidth to the ISPs doesn't go down in-line with Moore's Law, though, because there are other things to consider than just things like routers and servers, such as the cost of laying extra fibre optic cable, and paying BT largely consists of contributing towards the upkeep of BT's network, which includes 6,000 or so telephone exchanges and so on.

DAB+, on the other hand, has the same capacity limitations that DAB does, because DAB+ stations will transmit on exactly the same DAB multiplexes that are transmitting today, and the only reason why DAB+ will provide higher quality than DAB is due to the fact that it uses the AAC/AAC+ audio codec, which performs far better than the highly inefficient MP2 codec that DAB uses. The bit rate levels I would expect the broadcasters to use for stereo stations on DAB+ are as follows:

1 - AAC provides higher quality than AAC+ at bit rate levels above approximately 85 kbps, and vice versa

The reasons why I would expect the broadcasters to use the above bit rate levels on DAB+ are that the BBC has already launched five new digital stations, and it isn't expected to launch any further stations, so once it switches to using DAB+ it is likely to use similar bit rate levels to what it is using now, and it will use the switch from using MP2 on DAB to AAC/AAC+ on DAB+ to provide higher audio quality than it's currently providing, albeit that it will probably launch additional data services. The commercial radio broadcasters, on the other hand, will want to use lower bit rate levels in order to reduce their transmission costs, because the transmission costs have been a major problem on DAB so far -- the high transmission costs on DAB were one of the main reasons why the decision was taken to close Planet Rock, theJazz, Oneword and Core -- and the only reason the commercial radio stations will still sound significantly better than they do on DAB is due to the fact that the AAC+ audio codec is so efficient at low bit rates and the MP2 audio codec performs so badly.

SDTV

On the TV side, the following table shows the bit rates and formats used for the BBC and ITV TV channel streams on the multicast trial:

The BBC's TV channels using 1 Mbps H.264 video / 128 kbps AAC audio will provide a similar level of quality to many of the commercial TV channels on the 'traditional' digital TV platforms do at present due to the H.264 codec performing so much better than the MPEG-2 video codec -- MPEG-4 H.264 is about twice as efficient as MPEG-2.

HDTV

The BBC first tested an 8 Mbps HDTV stream over multicast back in January 2006, which was possibly a world-first, and they've carried out further successful trials since then, so it's possible that the BBC might start multicasting its BBC HD channel in the not-too-distant future.

And just to give you a taster of what multicast could deliver in future, someone at the BBC told me that when the London Olympics is on in 2012, he expects that we'll see a dozen HDTV streams -- each one covering a different sporting event -- all being distributed over the Internet in parallel via multicast. If each of these HDTV streams used a bit rate level of 8 Mbps (the BBC HD channel on satellite is currently using a bit rate of 16 Mbps on satellite), which would work out as being a total bandwidth of 96 Mbps, whereas the total capacity on the Freeview system is only 120 Mbps!

Unicast vs multicast

The problem with unicast, which is the distribution method that's currently used for live streams, is that every user receives their own individual stream, which is true even though thousands of people will be listening to the same BBC radio station at any one time, for example, like in the diagram on the left below. This means that the broadcasters and the ISPs have to carry large numbers of duplicated streams of the same channel in parallel on the same network links, and the bandwidth required for this is as follows:
 

Unicast bandwidth required  =  stream bit rate  x  number of users
 

The problem with unicast is therefore that it scales very badly, because when the number of users becomes very high the bandwidth required is huge. For example, if there were 1 million viewers watching a 10 Mbps HDTV channel, the total bandwidth the BBC would require just for that channel alone would be 10 Tbps, or 10 million Mbps, which isn't feasible.

Multicast, on the other hand, only requires a single copy of any given stream on any network link -- like in the diagram on the right above -- and the Internet routers that direct traffic on the Internet deal with forwarding the stream onto the required network links.

Multicast therefore eliminates all of the duplicated streams that unicast has to carry, and using the example above of there being 1 million people watching an HDTV channel, the BBC would theoretically only require 1 millionth of the bandwidth that would be require if unicast were used. In practice, the ISPs "peer" with the BBC at Internet Exchange Points, which means that the ISPs receive the multicast streams directly from the BBC, which slightly reduces the amount of bandwidth the BBC saves, but in reality the reduction is insignificant in percentage terms.

Multicast percentage bandwidth saving for broadcasters

The percentage bandwidth saving relative to unicast for broadcasters can be calculated using the following formula:

Multicast percentage bandwidth saving  =  (((N_users  /  N_ISPs) - 1) / (N_users  /  N_ISPs))  x 100

where N_users is the number of users of a stream, and N_ISPs is the number of ISPs peering with the BBC. The following graphs show the percentage bandwidth saving for the BBC assuming there are 50 ISPs peering with the BBC (both graphs start at 1,000 users):

Normal graph	Zoomed in graph

The following table shows some of the percentage bandwidth saving values for different numbers of users using the above formula:

Multicast percentage bandwidth saving for ISPs

The percentage bandwidth saving for ISPs on multicast is even more impressive than it is for the BBC, because you don't have to divide the number of users by the number of ISPs like you do in the above formula, so the percentage bandwidth saving gets very high very quickly as the number of users increases (albeit that the number of users will obviously be lower for each of the ISPs than it will be for the BBC). The equation for the percentage bandwidth saving for ISPs is as follows:

Multicast percentage bandwidth saving for ISPs  =  ((N_users - 1)  /  N_users)  x  100

The following table shows some values of the percentage bandwidth saving for ISPs:

Put simply, multicast solves the bandwidth problem for live streams.

BT's 21CN Next-Generation Network

There's an excellent overview of the 21CN on the Samknows Broadband website, but the basic facts about the roll out of the 21CN regarding broadband are as follows:

• 21CN will enable telephone exchanges for "up to 24 Mbps" ADSL2+ broadband
• 21CN roll-out has already started
• 21CN is scheduled to launch at 84 exchanges in May 2008, providing coverage to 5% of the population
• 21CN is scheduled to be rolled out to 55% of the population by May 2009
• 21CN should be rolled out to the vast majority of the population by 2010, with the last few exchanges being covered in 2011

Bookmark with:
Slashdot	Digg	del.icio.us	Reddit