Internet speeds have fallen off a cliff in the last 48 hours

jrmayfield

Modem: Motorola MB8600 (firmware 8600-18.2.12)

Location: Lexington, KY

Connection: Wired

Starting about 48 hours ago, my normally stable speeds (as tested automatically by speedtest every few hours) have fallen off a cliff. I'm on the ultimate 500 package so my speeds normally are in the 480Mbps range. According to my logs, starting on the morning of 12 Nov, speeds began dropping until they bottomed-out at around 30Mbps around 6am on 13 Nov. Since then, they've oscillated between 30Mbps and 150Mbps (using wired ethernet). It's interesting to note that only downstream speeds have been affected. Upstream speeds have been rock-solid.

Of note is the weather: we had light rain starting early Friday morning through Saturday afternoon. The rain might not be related since the issue didn't appear until after the rain had mostly tapered off.

Here's a plot of the speedtest results (taken every 3 hours) where you can clearly see something happened starting Saturday morning.

Here's the corresponding ping loss statistics for that time period. This measures packet loss to the first hop upstream from me in Spectrum's network.

(Ignore the Sunday evening spike...that's an artifact of me rebooting the modem).

Of note from the ping loss plot is that we see very sporadic loss (1 packet per 60 sec) up until Saturday afternoon where we see a fairly frequent, though still minor, packet loss (1-2 packets per 60 sec). This maps pretty well to when the drop in speeds began...

Here is the corresponding downstream SNR plot:

And the corresponding downstream power plot:

Note: The SNR plot does not show the SNR values for the OFDM PLC channel since it's much lower than I anticipated when I created the logging database. I'll attach an image of the actual signal stats below. For the power plot, you can see all the QAM channels are grouped together and the OFDM channel is much weaker.

(Also, ignore the Sunday evening spike on the power plot. That's just an artifact of me rebooting the cable modem; the job that queries the modem's stats retrieved garbage for that interval). Neither the SNR nor Power plots appear to show any sudden events on Saturday that correspond to the gradual decrease in speeds.

Finally, here's a snapshot of the modem's signal level page:

The only thing of note is the OFDM PLC channel. In my experience, it has always had a much lower power level and SNR ratio than the QAM channels though in the past the OFDM power level has hovered in the range of 1-2dB. Now it's at 0.3dB. It's not clear to me whether this OFDM behavior is expected or not. The uncorrected error count could be the smoking gun, though.

On a whim, this morning I decided to connect my modem via an RG6 cable directly to the grounding block on the side of the house. So the modem is connected directly to Spectrum's cable outside and one of my computers is connected directly to the modem so I'm bypassing all the wiring and network gear inside the house. I then re-ran a wired speedtest. Results: 44.84Mbps down / 23.27 Mbps up. This seems to suggest that either the problem is the cable modem or something upstream from me.

Things I've tried:

• Rebooted the modem. No change.
• Rebooted the firewall. No change.
• Bypassed the house wiring and connected directly to the Spectrum cable outside. No change.

James_M

Thanks for the detailed post. The uncorrected errors on channel 29 are red flags, are you able to do a factory reset of the modem and test again?

jrmayfield

Hi. Thanks for the response. I'd do the factory reset[1] but according to my logs, speedtest results returned to more-or-less normal around 1pm with no actions on my part. Something else I noticed that the OFDM channel that had such a low power level in the plots above has now jumped to 1.3dB.

Perhaps Spectrum fixed a problem on their side. Or perhaps it's related to temperature. The downstream power plot seems to show the OFDM channel in particular has a temperature dependence, peaking in the afternoon. On Friday where temperatures fell throughout the day and temps over the weekend were more-or-less steady around 32F. Today, we reached mid-40s and OFDM signal strength correspondingly increased. Perhaps a 9dB difference between the QAM channels and OFDM channel is a sufficient gap to cause problems? We're supposed to have much colder temperatures later this week so that ought to push the OFDM strength back down to 0dB or below so I guess I'll find out.

Should I be worried to see OFDM signal strength so much lower than the QAM channels? I'm not sure that I can convince tech support to send someone out just because one channel weaker than all the others since the first thing they're going to ask is whether or not it's leading to connectivity problems (so far it hasn't...just slowness).

[1] I presume a factory reset would possibly revert the firmware to an older version and Spectrum would need to re-download the latest firmware?

James_M

Thanks for the update. Glad to hear the issue appears resolved. I did send a refresh signal yesterday that sometimes fixes issues like power levels (assuming they are not the result of damage somewhere). The issue could be weather related, so contnue to monitor to see if the issue returns. Seems like it was more likely just ingress on the channel that needed to be cleared.

The factory reset clears out any settings that can be interfering, and any firmware updates are automatically downloaded by the network. However, it soulds like that step is not necessary at this time.

Let us know if the issue returns or if we can assist further with this or a different issue. Thanks!