DOCSIS 3.1 and the Bandwidth Bender

Here’s a shocker: Bandwidth usage is on a complete bender.

The average amount of traffic people used each month was 15 Gigabytes in 2013; that doubles to 30 Gb by 2018. A per-household snapshot puts the average household at 150 Gigabytes per month. Globally, IP traffic is expected to triple between now and 2018.

That’s why the arrival of the latest in the cable-modem specification family, DOCSIS 3.1, can’t come soon enough, panelists at Monday’s DOCSIS 3.1 and Wireless Symposium noted.

“One reason why 3.1 is so useful to us is that we can get a 50% capacity increase upstream and downstream, from the same spectrum we’re using today,” Jorge Salinger, vice president of access architecture for Comcast, said. “Most of the work is already done — the silicon design is started, and at least three system-on-a-chip (SoC) vendors are working on headend silicon.”

And as bandwidth balms go, DOCSIS 3.1 is impressively packed with relief. Spectrally, the spec enables a 149% increase in extended upstream frequency (by potentially widening from a limit of 42 Megahertz, up to 85 or 204 MHz); in the downstream, it brings an 86% improvement, pushing the upper edge to as high as 1794 MHz (or 1.7 GHz).

Meanwhile, the shift to Orthogonal Frequency Division Multiplexing (OFDM), a modulation new to cable but broadly used in mobile and telecommunications networks, yields spectral efficiency improvements, too.

While it essentially makes vestigial the decades-old unit of measure that is the 6 MHz channel, OFDM widens channel widths from 24 MHz to 192 MHz, filled with tiny (25 kHz to 50 kHz) subcarriers. The upshot: Much more efficient spectral utilization.

Plus, when something bad happens — impulse noise, ingress noise or interference — only the impacted subcarriers impacted go down, not the entire channel, panelists noted.

And then there’s LDPC — Low Density Parity Check — again, new to cable, but not new-new. As a replacement forward error-correction technique to Reed Solomon encoding, LDPC brings with it both time and frequency interleaving, which will make 3.1-based gear bear up better when hit with burst noise, Belal Hamzeh, director of network technologies for CableLabs and a self-described “PHY guy,” said.

To assess the capacity gains of 3.1, Hamzeh and his colleagues built a model based on 20 million (3.0- and 3.1-based) modems, normalized to 6-MHz channel widths, and running on 256-QAM plant. The findings: DOCSIS 3.1 will deliver an estimated 46% capacity increase compared to DOCSIS 3.0, and if signal-to-noise (SNR) ratios were to continue to improve, every 3-decibel gain will yield another 10% in network capacity, Hamzeh said.

“It means that about 70% of your cable modems will be able to achieve 1 Gig QAM without doing any changes in the network,” Hamzeh added.