Welcome to the Spectrum Community

Get answers, share solutions, help your community.

Start here

Intermittent packet loss

cpjet64cpjet64 Posts: 3
in Connectivity Jul 07, 2020

Good day all!

400x25mbps service

SB6183 modem

HP Server serving a pfsense vm for routing/firewall

windows server 2019 for local dhcp/dns/domain services

ubiquiti 24 port switch

The issue i have been seeing a lot lately is random packet loss throughout the day. Sometimes it can be 8/10 good packet test and sometimes it can be as bad as 0/10. When it is bad it can last anywhere from 10 seconds to a full minute. About a year ago I was having the same issue and after many long phone conversations and plenty of tech visits replacing lines the final answer was a failing server in the headend. I am now having the same issue as before and cant imagine there could be another failing server. I think it might have to do with power levels but wanted to hear your opinions first. I have also always had a ton of critical errors in the event log on the modem and was hoping someone could help explain them and if they were to be a cause for concern. I am a senior systems and networking engineer by trade so dont be afraid to use the big words. Thank you all!


EDIT: To clarify there is only internet at this location and it is a direct connection from the drop to the modem. Total length of cable from drop to modem is approximately 10 ft. 








  • RAIST5150RAIST5150 Posts: 794 ✭✭✭✭
    Jul 07, 2020
    Looks like your line is boosted too much.

    Downstream signal levels are hitting 12 and 13db and upstream is staying low in the 37-39 range--may be on the order of around +10 gain/ boost on both forward and return paths.

    If there is any noise on the line, that will get amplified too and could actually make things worse if the SNR is too low on the upstream side. You have high SNR on the downstream side, so you don't need such high signal levels on that front.

    If you spot a drop amp on your line, may want to try bypassing it (if feasible) to see if things stabilize.
  • cpjet64cpjet64 Posts: 3
    Jul 07, 2020

    Unfortunately there is no drop amp or any other amps between the drop outside and the modem.

  • karlbeckmankarlbeckman Posts: 2,223 ✭✭✭✭
    Jul 08, 2020

    Since you are an "in the know" senior systems network engineer (me too, 43 yrs), there's no need to obscure the MAC address of the CMTS (Cable Modem Terminal Server) in all the status messages.  You only make it more difficult for Spectrum's networking staff to identify the server at their head end that you suspect of malfunctioning.  Masking of the CM (customer modem) MAC is justified for security protection from other users on your neighborhood's segment of the cable distribution network. 

    I'll agree with @RAIST5150 that the incoming downstream levels are about 10 dB higher than necessary (0 dBmV is nominal), suggesting an unneeded or misaligned line amp.  The upstream levels from your modem are quite acceptable, provided that there is no inbound RF leakage hitting the head end from upstream ingress locations. We can't tell because the error counters in the modem were reset not long before you took the screen shots.  Normally we like to see 24 to 48 hours of counter data. 

  • cpjet64cpjet64 Posts: 3
    Jul 08, 2020
    I will say I enjoyed your "in the know" comment, it actually made me laugh out loud!
    Anyone with a background in IS security knows public facing equipment can be spoofed with the right tools but that is besides the point.

    I honestly had no idea that there were actual Spectrum staff on here who used this information for production troubleshooting. That being said if any of these Spectrum employees would like the non-redacted logs, etc please message me and I will gladly send them.

    As for power levels what would your recommendation be as for a next step?

    I am unable to upload a screenshot of the power levels because the service is completely out right now but looking at them and comparing them to the screenshot i provided previously the error rates are identical, downstream power is no more than a .5 dBmv difference and upstream are the exact same as previously posted.
  • RAIST5150RAIST5150 Posts: 794 ✭✭✭✭
    Jul 09, 2020
    Generally speaking I have seen recommendations for cable modems to target downstream in the range of +/-10 and upstream between 38 and 50.

    While technically your values may be falling within the specs of the modem for it to operate, it is falling outside of such general norms.

    It really raises a red flag for me because of the expectations in our particular market.. moreso for the upstream level than the downstream. While they don't like to see downstream actually hitting 10, it does in fact manage just fine with many models in play (just not all of the Spectrum provided models). If they see it crossing 10 down, they will look at putting splitters/terminating caps in play when on the scene to see if things stabilize on that front in the short term before escalating to maintenance.

    As for the upstream side... power levels below 40 are a big WTF here. It is an immediate call to look for things like user installed Holland amps and such. They expect upstream to pretty much stick at 44-50 like clockwork... deviations below 42 or 43 just tends to trigger them. Don't know the exact reason why they prefer the upper range... perhaps there are known issues with the noise floor making it too easy to muddy things up if someone ramps up gains at the pedestal or puts in a drop amp with active return boost.

    If a tech were to see these levels in my market, they would first trace the line for a drop amp to remove, and failing that they would put in a splitter with a short lead on the unused port(s) with a termination cap on that lead in an attempt to bring the downstream numbers down and the upstream numbers up. They might even look at extending the length of the cable run a little as well to sort of "fine tune" it somewhat. If those measures did not show significant improvement then it would likely get escalated to maintenance.
This discussion has been closed.