Your browser is not supported. Please update it.

27 February 2014

Marvell throws G.hn at the last mile for 500 Mbps Korean broadband

US chipmaker Marvell is this week parading a massive design win for its G.now technology at Mobile World Congress, built around its existing G.hn chips, working with a sub-contractor for Korea Telecom, which it claims will revolutionize fixed line broadband access, particularly in MDUs.

G.hn has split analysts and vendors alike since its name was first bandied around circa 2008. The standard was cut late 2009 and since then nothing but vitriol has been poured on the three wire technology by players who were not part of it, saying it would never work, it would not achieve the speeds it claimed and that anyhow, no-one needs it, and we can do the same with HomePlug and WiFi.

One analyst group promised G.hn would be in 42 million homes by 2013, while others said it would crash and burn and make no headway and both have proved wide of the mark ‘ our own take, given its supporters and the Chinese appreciation for technology that comes without a huge IPR bill, was that G,hn would make headway in China, but it was unlikely to rival the combined might of MoCA, WiFi and HomePlug and that the rest of the world remained outside its reach.

But now Marvell this week has pretty much tipped over the broadband apple cart, by aiming a 500 Mbps broadband service over only the twisted pair part of the equation (it also works on coax and powerlines) in conjunction with Korea equipment supplier Ubiquoss, a leading supplier to KT Corporation (formerly Korea Telecom). KT has some 18 million captive phone homes and many of them are based in MDUs. It has some 8 million broadband connections in these homes and 4.7 million IPTV customers, almost as many as AT&T.

Already South Korea is famous for offering 100 Mbps broadband to most homes, and has been a global leader in broadband deployment for over a decade. Where South Korean operators go, other telcos follow, and this installation will have observers fascinated given there are some 35% of global homes in MDUs, around 25% in the US, and concentrations as high as 50% in parts of Eastern Europe and parts of Asia Pacific, including Chinese suburbs and South Korea.

The MDU side of the broadband equation has been a complex one for years. Fiber can get to the basement of the building okay, but Fiber to the Home is really tough as a wiring challenge inside the MDUs, involving taking fiber around lots of corners, adding expense and weakening the throughout. In Japan the trick has been to bring fiber to the basement and attach it to VDSL copper which had previously been installed inside the building for voice connections. Cable technologies have seen the MDU as a major opportunity for bringing broadband to the home and Ethernet over Coax or specifically over MoCA, as well as inexpensive versions of cable DOCSIS, have all been tried as a route to bring broadband into MDUs. Many MDUs also have coax installed from apartment to apartment due to analog cable TV being around when they were built. However most ducting is impossible to re-wire without drilling into the reinforced concrete which holds up the MDU, so broadband will always be cheapest if it can come in over one of the existing wires ‘ either twisted pair or Coax, with just some local amplifiers or relays or splicers.

As MDUs have stretched the capability of VDSL2, this problem has left many telcos wondering if they have to simply give up the effort for MDU broadband and let coax do the job.

But this new approach by Marvell will mean that a 500 Mbps broadband signal can get into each MDU apartment over twisted pair, built around a core system that has 9.6 Gbps of total capacity in a 24 port configuration, and opening up the twisted pair all over again as a source for broadband. The system can be backhauled with either GPON or EPON, at anywhere up to 10 Gbps.

The Ubiquoss designed system has cross talk cancellation built in where the twisted pair come out as a bundle in the basement of the MDU, which in a high rise block of flats can be a significant hurdle. It also has IGMP (Internet Group Management Protocol) built into the system so that IPTV can be easily laid across it.

Marvell said that the architecture uses Time Division Duplexing and increased spectrum up to 100 MHz and OFDM. As we remember it the specification for G.hn was reduced to 80 MHz, so this looks like a variant chip that has been used, but the rest of this is identical to its G.hn chips.

This set up is perhaps the way that people have seen G.Fast being used when it arrives, and this looks like Marvel can deliver an interim step to G.Fast, perhaps 2 to 3 years before devices that adhere to that standard ship, giving telco broadband a big lift up. Can it be used outside of MDUs in the street cabinet? Well as long as it really does have some level of cross talk cancellation, that sounds feasible, although it is not yet clear over what kind of distance this 500 Mbps signal can reach ‘ there is no magic here, the higher levels of QAM that are used in G.hn may simply fall back to lower levels after some attenuation, but although this is really solving the same problems as G.Fast, the fact that a chip has been built and can be shipped in volume today, may make an interim step achievable throughout Telco driven broadband.

So when Marvell says that ‘G.now is poised to take on the leadership of the broadband access services for the last mile,’ we are inclined to think that’s a distinct possibility, much to the chagrin of chip vendors like Broadcom, which remains determinedly in the ‘other’ HomePlug and VDSL camp, and which is seen to be leading the VDSL2 vectoring charge in Europe.

While it is unlikely that vectoring will be overtaken among telcos that are already a long way down the trialing path, many other telcos may be seduced to use this for an MDU strategy and then subsequently adopt it for all broadband if they are impressed and if such a version emerges. Pricing of course will be very important and the total cost of ownership of vectoring is pretty complex, involving chip replacement at both ends as well as an ongoing server and maintenance costs.

Right now Marvell is pitching this plain and simple as a cheap way of getting FTTH speeds in MDU homes, and while we are not seeing wide deployments of FTTH in MDUs, it is a compelling way to reach volume deployments, especially given that Marvell is already active in FTTH and it has recently introduced its Avanta family of cheap ARM based optical network terminals, which can be used with this.

Marvell says that G.now is already being trialed in Asia by telcos looking to provide the most powerful access services to their users at a competitive price and still be able to reuse their most valuable asset, the existing copper infrastructure. In particular this is being seen as a way to get multiple streams of 4K video into the home.

G.Now can also bond 2 channels to help extend its reach, and provide extra protection against noise and uses the Marvell 88LX3142 G.hn digital baseband processor and the Marvell 88LX2718 G.hn analog front end.

The ITU group which is aiming to reach a standard for G.Fast met last week and said that the standard could be ratified as soon as the end of 2014, but that still means that it could be two years or longer before chips appear around that standard, and another 6 to 9 months before they appear in devices.

Other chip makers that are supporting G.hn including Sigma Designs and Chinese group Metanoia, while Lantiq, which was one of the original supports of G.hn seems to have fallen by the wayside and has put all its eggs into the VDSL2 vectoring basket.