Cisco to Demonstrate Full Duplex DOCSIS 3.1 (“FDX”)
By John Chapman, Cisco Fellow, CTO, Cable Access, Cisco
It occurred to me recently that it was in 1997 that the first DOCSIS-based cable modems first entered the market in a meaningful way. Five years ago, in 2012, we celebrated the 15-year birthday of DOCSIS (with cake, even!); this year, at age 20, I’m gratified and excited to share what I think is one of the biggest developments since the original DOCSIS 1.0 specification — if not the biggest.
I’m talking about a proof of concept demonstration we’re hosting with Intel in our booth at ANGA COM this week, of the first Full Duplex (abbreviated “FDX”) DOCSIS setup. It’s one of the biggest developments since the original DOCSIS because it positions the industry to achieve fiber-equivalent speeds — without replacing the “last mile” of coaxial plant.
It’s also big because it enables the upstream path to eventually achieve speeds of 5 Gbps! Today, four ATDMA carriers in the upstream provide about 100 Mbps of capacity. So this would be a 50x increase in upstream capacity! And this is done without removing spectrum from the downstream path. Instead, the spectrum is shared between the upstream and downstream paths.
Now. Let me level-set right here, right now: This is not a ‘this year’ thing. It’s not a next year thing. It’s probably not even a 2019 thing. FDX DOCSIS builds upon several other technology and business trends in the cable industry that are just starting to unfold:
- The first trend is the deployment of DOCSIS 3.1 with OFDM. FDX DOCSIS is an extension to the DOCSIS 3.1 architecture and only works with OFDM/OFDMA channels.
- The second trend is the deep fiber architecture that migrates the HFC plant from N+5 (optical node plus 5 amplifiers) to an N+0, passive plant architecture. FDX DOCSIS assumes a passive plant between the optical node and the CM.
- The third trend is the use of Remote PHY technology where the CMTS PHY is placed in the node.
- The fourth trend will be FDX DOCSIS that places an echo canceller in the Remote PHY node, which is located in an N+0 plant, along with more upstream channels.
With an all-DOCSIS downstream plant, and with DOCSIS 3.1, service providers can ultimately achieve speeds of 10 Gigabits per second in the downstream. Now, with FDX DOCSIS, the upstream can be extended to 5 Gbps. If a node were to contain one forward path and two reverse paths, then that optical node would have 10 Gbps capacity in both the downstream and upstream. This would match the 10 Gbps wavelength used to feed the digital optical node. This means that a 1×2 DOCSIS MAC domain with DOCSIS 3.1 and FDX would have symmetrical 10 Gbps capacity, and be equivalent to a fiber wavelength, and is a legitimate extension to fiber — without actually running fiber.
Why a proof of concept that “only” does 96 MHz symmetrical, when the end game is 576 MHz of shared spectrum? Because by necessity, proofs of concept must rely on current technology — and today’s cable modems “top out” at an upstream spectral location of 204 MHz. The 96 MHz shared spectrum in the demo will have a downstream rate of 890 Mbps (with 4K QAM) and an upstream rate of 680 Mbps (with 1K QAM). (Ultimately, we can expand the upstream return path from 42 MHz where it is now to a top frequency of 684 MHz, with a downstream start at 108 MHz, and no I’m not kidding — but that’s fodder for another blog.)
Here’s what you’ll see, should you be in Cologne at ANGA COM: a 96 MHz chunk of spectrum, located between 108 MHz and 204 MHz, with traffic moving simultaneously up and downstream. The intent is to show that a node can transmit and receive at the same time, on the same spectrum. This was heretofore impossible, because of the way amplifiers, taps, and diplex filters were “notched” to transmit, in one direction only at a given frequency.
Intel’s contribution (thanks again!) is its FPGA (Field Programmable Gate Array) silicon, valuable for its role in rapidly responding to marketplace developments — like FDX. Only a fully programmable, DOCSIS 3.1 Remote PHY system on chip (SoC) platform, connected across an HFC plant to Intel’s latest “Puma” cable modem silicon, enabled us to create the FDX demo. The Remote PHY Device (RPD) with FDX in the optical node is running OpenRPD open source software.
The magic of it is our (substantial) FDX work on echo cancellation. One way to envision this is to think about the noise-canceling headphones people tend to wear on airplanes. The headphones listen to the ambient noise of the plane, and create an inverse signal that cancels it out. In the case of FDX DOCSIS, what one modem transmits, the other sees as noise. An FDX echo canceller, like we’re demonstrating in Cologne, removes multiples of such “echos.”
Why is this happening? Ultimately, echo cancellation is an example of what can be achieved — when you have a whole lot of gates, and a whole lot of DSP horsepower.
FDX ultimately puts to rest — both logically and economically — the question of whether and when the time will come to sunset DOCSIS, and replace it with fiber to the premise.
The answer is probably never, because FDX creates what is essentially a fiber equivalent, throughput-wise, with coax. Moving 10 Gbps downstream, toward nodes, is already within reach; two upstream node ports running FDX yields 10 Gigs upstream. Why dig up yards to take fiber to the side of the house, when the bandwidth is already there?
Believe me when I say: FDX is big. Milestone big.