The concept of a Full Duplex DOCSIS (FDX) Amplifier has been discussed for years inside the cable industry. But revelations that emerged at this week's SCTE Cable-Tec Expo make it clear that such a product is not only in development, but could start to surface as early as next year.
Such a product could have major implications for operators – and for Comcast, in particular – that adopt the FDX flavor of DOCSIS 4.0, a platform that will put cable operators in position to deliver symmetrical multi-gigabit broadband services on their hybrid fiber/coax (HFC) networks.
Originally developed as an add-on to DOCSIS 3.1, FDX allows an operator to dynamically serve upstream and downstream traffic in the same block of spectrum, rather than having to keep that traffic separate. FDX originally envisioned a fiber-deep network with a so-called "node+0" architecture that has no amplifiers present between the node and the home. That made it a non-starter for some operators that had one or more amplifiers in the cascade, or had no plans to spend heavily on node+0 upgrades.
The FDX Amplifier could change all that. Rather than having echo cancellation – a technique that tamps down the interference caused when upstream and downstream traffic overlaps the same block of spectrum – that technology is instead packed inside a specially made amplifier that can pass through the FDX signals. Harmonic is one of the companies working with Comcast on the FDX Amplifier project.
"Obviously, it's not your grandfather's amplifier," Rob Howald, an engineering fellow at Comcast, explained in a "10G" session earlier this week. The development of FDX Amplifier technology has "accelerated our thinking" about FDX, as it "simplifies, architecturally, where we can put FDX," Howald said.
But there's been very little said about when an FDX Amplifier might see the light of day.
Jorge Salinger, VP of access engineering at Comcast, shed some light on that during a Wednesday panel focused on distributed access architecture (DAA), a requirement for DOCSIS 4.0 networks.
Stressing that the FDX Amplifier is a "longer term" initiative, Salinger said Comcast expects to have samples of an FDX Amplifier in-hand by the second half of 2022. That could be followed by trials later in 2022 and the start of deployments in 2023, he added.
Elsewhere on the FDX technology front, Salinger said work is underway on the company's implementation of FDX to support multiple interference groups, and that Comcast already has samples of an FDX remote PHY device that's now undergoing tests.
In the meantime, Comcast is making progress elsewhere with FDX, announcing this week that it had successfully tested a "network to modem" connection of DOCSIS 4.0 FDX technology in the lab.
DAA lessons learned, and putting 'Sherlock' to work
The FDX Amplifier update was a tidbit shared on a Cable-Tec Expo presentation focused on Comcast's deployment of DAA, which now spans tens of thousands of remote PHY devices (RPDs). Comcast also shared some of the lessons learned from that rollout so far.
Among them, "managing power interruptions becomes very critical" with DAA, Salinger explained. If, for example, the power to a leg of the network needs to be temporarily disconnected to help find the source of ingress noise, it will cause the RPD to reboot. While that's not a big deal for an older analog node that can be quickly restored when power is flipped back on, an RPD, which uses a processor, takes a much longer time to boot up.
"Powering becomes very impactful" on a DAA network, Salinger said.
Comcast is also learning about the scaling challenges with respect to both deployments of RPDs and managing the upgrade process of RPDs already deployed on the network. To help with that, Comcast has developed a new provisioning system that is undergoing constant improvements, Salinger said.
Salinger also highlighted the need for DAA to be backed by modern monitoring equipment and sophisticated analysis tools. A piece of that is being handled by "Sherlock," the codename for a relatively new DAA-focused data telemetry system that's being aided by automation and machine learning techniques.
Matthew Stehman, a data scientist at Comcast, explained that DAA and the use of RPDs enable the operator to receive rich telemetry across all components of the network and see everything at a micro-level. However, the company still needs a way to pull it all together and get an automated, "holistic view of DAA."
That led to the development of Sherlock, an internal application that can aggregate that data into a central location and help Comcast to rapidly analyze its full DAA footprint. Sherlock is currently able to detect 20 different kind of events, such as an RPD software upgrade or a RPD reboot.
Stehman said DAA is also ripe for machine learning technologies that can be applied to the Sherlock system. That initial work involves data "clustering" that can be fed into the model to help find patterns. From there, the idea is to teach the system to recognize those patterns and, eventually, to predict when those patterns might occur so Comcast can know about certain problems before they happen.
Stehman said a clustering case study identified 897 RPD upstream service events over a three-week period. "We're really pushing forward with this," Stehman said, noting that big data analytics require big data tools.
Salinger also offered a few other details about Comcast's DAA deployment, noting that it's a multivendor platform that required a bunch of interoperability work.
He estimates that it took a year for Comcast to set up its basic interop. "Interoperability is not easy … It takes time, it takes effort, it takes manpower," Salinger said.
Comcast is also facing a multi-generational challenge as it prepares to deploy a second generation of RPDs within the next month or so. That, Salinger said, will broaden the repertoire of RPDs that Comcast will need to track. That complexity will only become greater when Comcast later moves ahead with RPDs that support Full Duplex DOCSIS.
But the good news is that RPDs are not difficult to swap out in a node.
"It's not terribly expensive, and it's fairly easy to replace an RPD," especially when compared to having to replace a traditional Converged Cable Access Platform (CCAP) architecture or having to overhaul equipment in the headend, Salinger said.
- Cable network set to become 'predictive and more proactive,' Cox CTO says
- Comcast forges 'network to modem' DOCSIS 4.0 connection
- DOCSIS 4.0 puzzle starting to come together
- Remote MACPHY gaining steam with cable ops, CommScope says
- Comcast 'making a lot of good progress' with DOCSIS 4.0
- Full Duplex DOCSIS amplifier chatter heats up
- Harmonic: Let's put the MAC here, there... anywhere
— Jeff Baumgartner, Senior Editor, Light Reading