There’s More To The Story (There Always Is!)

A couple weeks ago, I posted a story about a particularly irritating problem I was having with my high-speed Comcast internet service dropping out. After lots of troubleshooting, I thought I had cornered the culprit – a Samsung-manufactured set-top box that Comcast was using for delivery basic Xfinity services (no DVR).

When I connected the cable TV input connection to a spectrum analyzer, I saw some pretty nasty burst of spectral noise that ranged from 11 MHz all the way up to 400 MHz. I figured this might have been the cause of the dropouts and promptly returned it to the local Comcast Store, only to find out that particular model wasn’t in use any more and that I’d have a much smaller, flatter version to take home – one that would link automatically to my main X1 DVR.

Coincidentally, the dropout problem stopped, so I did what any other reasonable person would do: I assumed that was the end of it.

Except it wasn’t. A few days after I replaced the box, the WAN connection started dropping out again. Some days it dropped only a couple of times, but on April 13, it dropped out almost fourteen times in four hours. Out came the test equipment (and plenty of expletives) as I started testing every line in the house, taking more and more things of-line.

At one point, the only thing connected to the Comcast drop was my wireless gateway and my spectrum analyzer, through a brand-new 2-way splitter good to 1.5 GHz. Sure enough, the WAN connection dropped again – but this time, I caught something on the analyzer I hadn’t seen before.

Figure 1. The entire system (two cable boxes and a wireless gateway) are working just fine with these signals.

Figure 1 shows the ‘normal’ levels of QAM carriers coming through the drop. There’s a little up-and-down there, but the entire system – in particular, the downstream QAM carriers above 650 MHz – all measured at least 32 dB above the noise floor (about -87 to -88 dBm). In this condition, the wireless gateway was chugging along just fine and broadband speeds were pretty fast.

Just ten minutes later – while I was watching the analyzer screen – the QAM carriers from 50 MHz through 400 MHz dropped precipitously, as seen in Figure 2. Right on schedule, the WAN connection stopped working! Yet, I hadn’t touched, changed, or re-wired anything. This was starting to look like a classic ghost in the machine, and likely an issue outside my house. (Yes, the Samsung box did need to be replaced in any case – it was quite dirty, RF-wise.)

Figure 2. Ten minutes later, KABOOM – the WAN connection dropped and my analyzer showed some nasty QAM waveforms below 400 MHz. What was causing this?

Well, after escalating this problem to Comcast’s Special Operations unit (yes Virginia, they do have Special Ops guys), I was visited by Jason Litton and Fredrick Finger of Comcast. I asked them to replace the DC block/ground block outside the house and also to sweep the underground cable coming back to the house. I had previously gone out to check the ground block and discovered (a) it was grounded – the wire had come loose) and (b) there was a tiny bit of play in the connections at either end, which I tightened up before they arrived.

Long story short; the block was eventually replaced, re-grounded, and new connectors were installed at either end of the underground drop. During testing, Jason spotted noise coming from a neighbor’s coaxial drop and proceeded to install several more new connectors. I also took the opportunity to have them put in two brand-new splitters in my basement (overkill, but what the heck) and run a new coaxial line to my workbench.

And that finally did the trick. Whatever phantom was haunting my system had finally been exorcised for good. Using Comcast’s brodband speed test to New Castle, Delaware and Secaucus, New Jersey, I saw wired LAN and 5 GHz 802.11ac download speeds hitting 100 MB/s. Using the popular server in Dallas, Texas; I measured download speeds around 30 – 48 Mb/s. Upload speeds to all servers were in the range of 10 – 12 Mb/s.

Figure 3. The final setup with all QAM levels where they should be!


Figure 4. Best of all, there’s no noise below 50 MHz on the upstream channels. FINALLY!

So what was the culprit? Most likely the cheapest thing in the system – the DC block. Noise from the Samsung STB didn’t help, and apparently neither did the noise coming from my neighbor’s cable drop. But the block probably had an intermittent connection and was creating some nasty standing waves, causing tilt on the lower QAM carriers and noise at the uplink frequencies around 30 MHz.

I’ll have more details on this unfortunate series of events during my RF/Wireless class at InfoComm in June. Until then, things are working well (knock on wood, or metal, or coax, or modem…)