Israeli broadband chip firm Sckipio, this week upgraded its G.fast chip to the wider 212 MHz bandwidth, the second generation standard. It has been in a race against Broadcom for mind share and market share and Broadcom already has out its 2nd generation chip. The Sckipio solution is actually a 3 chip system called the 23000, though it’s not clear what discrete workloads each of the chips have.
Sckipio was first in the market to come out with first generation chips and it may look like it is slipping somewhat against Broadcom, but it came out with added features which it is convinced will continue to keep it ahead of its far larger rival. It is fairly vital that Sckipio maintains its lead and challenge against Broadcom, because despite an investment by Intel, it does not have the brand and long term relationships that Broadcom can muster. It needs the edge.
But that’s what makes this fight interesting, and Sckipio has once again leapfrogged Broadcom, coming up with bonding off the bat, so that two pairs of twisted pair can be used for one service at roughly 3 Gbps PHY rate, a little slower in reality, and its continued use of distributed vectoring – meaning that as long as the vectoring boards on each DPU can talk directly to one another, they can cancel cross talk from all lines in any single binder, up to 96 ports.
This 96 ports has been a target for quite a while, from operators including British Telecom, which is believed to have mandated it. What Sckipio has come out with is a 48 port device distribution point unit (DPU) made up of two 24 port systems, which can be connected together to support 96 ports. Sckipio makes the point that reverse power feeding cannot work with one customer powering 96 or 48 lines, so claims this as a bit of a strategic win, whereby they have reverse power feeding on 24 lines, and then have 4 of these talk to one another. Doing it with a single cross talk cancellation server, makes that harder to do, and usually it offloads the cancellation to another single device which will likely add to the expense.
Sckipio is widely rumored to have all or part of the AT&T contract, which is known to be connecting its fiber directly to the coax which DirecTV customers in MDUs use to take DirecTV – usually known as the coax connecting the channel stacking switch. This means it has to get its fiber up the side of buildings to the roof above apartments, and not to the basement.
Recently the Broadband Forum did device certification on products from Adtran, Exfo and Viavi, to go with Arris, Calix, Huawei, Metanoia, Nokia and Technicolor, who had already passed certification. The chips are all from either Broadcom, Metanoia or Sckipio.
Michael Weissman, VP Marketing for Sckipio told Faultline this week, “Cross DPU vectoring, needs a superhighway to talk to one another on, but it allows you to start small, get systems to pay for themselves before growing as your revenue grows. You can start with a 16 node installation then install another alongside it and plug the two together.
The other feature that Sckipio has been tantalizing telcos with has been cDTA. The key to cDTA is two-fold. First, it uses the fact that G.fast uses Time Division instead of Frequency Division, to allow a different ratio of upstream to downstream. Because this is dynamically switchable, it can be allocated on the fly using policy algorithms to adapt to changing combinations of upstream and downstream traffic.
This can mean that operators can reach longer distances for G.fast by allocating more or all of the capacity at any point in time, in one direction to pass a service level test. The ability to bond was used in the previous generation of AT&T’s broadband, using ADSL2+ bonding back in the 2007 timeframe, and in many instances it is already configured to work with two twisted pairs. Bonding will also likely mean that 1 Gbps can be introduced further away from the DPU as well. Broadcom’s move.