Locating noise in cable's upstream can be like finding a needle in a haystack. That problem is compounded for field technicians because the needle itself can appear, vanish and then reappear. Using traditional, manually operated field monitors, techs can find themselves moving about the plant for long lengths of time as they attempt to hunt down and pinpoint the source of that noise, sometimes to no avail.

The need for cable operators to zero in on upstream noise that can result in packet loss and spawn other impediments that reduce the performance of the upstream has only become more pronounced during a pandemic that has caused usage patterns to rise. According to the NCTA's latest COVID-19 Dashboard, a tool that collects network data from several US cable operators, peak upstream usage has surged 23.1% since March 1, compared to a rise of just 8.4% in the downstream.

Figure 1: Please click here for a full-sized version of the chart.

Although the pandemic has shined the spotlight on cable's upstream, Promptlink Communications was focusing on the issue even before COVID-19 hit with the Network NoiseHawk, a tool introduced last year that pairs network data with artificial intelligence and machine language techniques to help operators keep an ear out for noise and then pinpoint its source in an automated fashion.

Noise is one of the most challenging network-facing issues that cable operators face, because all of the modems in a nearby grouping can be impacted negatively even if just one modem is injecting noise into the network, Dr. Foad Towfiq, Promptlink's president and CEO, explains.

Figure 2: Dr. Foad Towfiq, president and CEO, Promptlink Communications

And that noise is not easy to find using more traditional spectrum analyzers and network sweeping techniques, claims Towfiq, an exec who hails from the aerospace and telecommunications industries, including a spell at McDonnell Douglas/Boeing.

Yet another issue is that the noise must be present when the technicians are actually listening for it. Network noise can fluctuate, appearing and then vanishing, only to resurface later, making it a very elusive phantom for operators to pin down, he adds.

Rather than requiring operators to manually move about and attempt to detect noise at the fiber nodes or at different branches of the network, Promptlink's software-based platform accesses the DOCSIS network and collects telemetry data from the cable modem termination system (CMTS) and the individual cable modems.

From there, the system uses AI and ML techniques to listen for various types of noise "signatures" that might be caused, for example, by a faulty amplifier or a broken or loose connector. It then determines where that noise is being generated on the network. The resulting analysis comes up with a profile of not only what kind of noise is being generated but also from where it is emanating.

Figure 3: Promptlink's system supports dashboard views that map out sources of network noise and how persistent that noise is.

"The software … analyzes this information," Towfiq says. "Different modems behave differently depending where they are located relative to the source of the noise."

By accessing and analyzing that data, "we can pinpoint and find the common points where the modems are affected the same way" and provide a more accurate view of the network noise situation than what operators can typically get with manually operated spectrum analyzers, he adds.

Promptlink introduced Network NoiseHawk – a Leading Lights 2020 finalist – last year and claims to have trials and deployments underway with unnamed cable operators in the US, Canada and Latin America. The company recently issued free 90-day access to the platform's managed, remote feature so operators could kick the tires on NoiseHawk during the pandemic without taking away from their existing resources.

While Promptlink's approach competes with manually focused network analyzers, it's not the only cable entity out there using AI and ML to get a grip on network conditions that could impair the upstream.

Comcast, for example, has been fast-tracking a new system called Octave. That AI-powered platform is designed to boost the performance of the cable access network by rapidly pinpointing network anomalies and noise-induced issues by polling and analyzing anonymized data from gateways and modems.

Comcast is also using that tool to dynamically set the maximum modulation/throughput profiles of individual cable modems based on current network conditions. The MSO has recently added an upstream focus to Octave to complement an initial rollout focused on the downstream.

Related posts:

— Jeff Baumgartner, Senior Editor, Light Reading