Inconsistent upload with DOCSIS 3.1 modems



Figured I would post here, since I have also posted on the DSLReports Community about this with my Gigabit Internet service, as seen here:


it seems for the past few months, the upload on Gig service has gotten a bit inconsistent with TCP traffic. I can still get 42Mbps if I saturate the upstream by performing an upload and also running a download that can run at a Gigabit. If I have a standalone upload happening, such as if i am uploading a YouTube video, streaming, or in a video call, the upload will jump between 12Mbps and 42Mbps. Packet captures show a high amount of TCP Re-transmits and TCP Duplicate Acks when uploading.


The interesting bit is if I were to perform an upload using the UDP Protocol. Such as what games use, and some VPNs. Uploads over UDP will run at a full 42Mbps. Also, if I run an iperf3 test over the connection at 41Mbps, I will be able to transfer packets without an issue at the full 41Mbps rate, with no out of order packets or packet loss observed.


This almost looks like an issue with the DOCSIS-PIE Configuration (For Active Queue Management) or a problem with the CMTS. i've had a few techs out to check signals as well as change out modems. At this point I've used four different modems (Hitron E31H2V1, two of these, Technicolor E31T2V1, and now my current Ubee E31U2V1) and have been visited by plant maintenance close to three times. All of the coax between my modem to the cable tap has been changed. My grounded coupler (as I do not have splitters) has been checked and replaced. Fittings have also been changed out. A supervisor even stopped by on my last tech visit to see what was going on, and agreed something was wrong when I demonstrated the problem, and would escalate to engineering at the headend. My last visit was a few months ago, and I have not seen much action since. A lot of the techs continue to blame the tilt on my node, although plant maintenance keeps saying that's not a problem.


For the record, my service is also Residential. Others on DSLReports seem to be reporting similar behavior, even on lower speed tiers like the 400Mbps/20Mbps Spectrum Ultra plan, and that switching to a DOCSIS 3.0 modem resolves the issues with upstream. I also have a Spectrum Business Gig with Static IP connection I'm monitoring, and it, too, has similar behavior with the upload. Rebooting the modem does seem to allow the upload to occasionally remain consistent and work correctly, but this often doesn't last more than an hour. At times, the upload will correct itself for a few minutes, then return back to being inconsistent and broken, again with TCP traffic.


I've not been able to correlate this issue with any source/destination issue, with any configuration issue on my end, with IPv4 vs IPv6, with time of day, or weather. I use a UniFi Security Gateway 3P which reports active bandiwdth usage, and I can confirm it is not some other device on my network causing issues.




This is what the upload looks like when it is misbehaving.




This is what the upload looks like when it's working correctly.



This is what the upload looks like when it switches from  "working correctly" to "broken"



The only other consideration for this problem is that, prior to having the upload issue, I was seeing a lot of download issues with the Gigabit service. The download would have a hard time getting above 340Mbps for again.... TCP Traffic. UDP traffic could saturate the Gigabit connection without issue every time. This was fixed after some overnight maintenance, but I don't know if the upload broke after the last maintenance.


These are my signals:

CM Downstream Channel Info
Channel Lock Status Channel Type    Channel ID  Frequency   Width   Power   SNR Modulation Profile ID   Correctables    Uncorrectables
1   Locked  OFDM Downstream 25  314000000 Hz    94000 kHz   8.1 dBmV    42.5 dB 0 1 2   0   0
2   Locked  SC-QAM Downstream   1   507000000 Hz    6000000 Hz  5.2 dBmV    43.6 dB QAM256  0   0
3   Locked  SC-QAM Downstream   2   513000000 Hz    6000000 Hz  5 dBmV  43.5 dB QAM256  0   0
4   Locked  SC-QAM Downstream   3   519000000 Hz    6000000 Hz  4.6 dBmV    43.3 dB QAM256  0   0
5   Locked  SC-QAM Downstream   4   525000000 Hz    6000000 Hz  4.1 dBmV    43.1 dB QAM256  0   0
6   Locked  SC-QAM Downstream   5   633000000 Hz    6000000 Hz  -0.7 dBmV   40.2 dB QAM256  0   0
7   Locked  SC-QAM Downstream   6   639000000 Hz    6000000 Hz  -0.1 dBmV   40.6 dB QAM256  0   0
8   Locked  SC-QAM Downstream   7   645000000 Hz    6000000 Hz  -0.5 dBmV   40.2 dB QAM256  0   0
9   Locked  SC-QAM Downstream   8   651000000 Hz    6000000 Hz  -0.1 dBmV   40.5 dB QAM256  0   0
10  Locked  SC-QAM Downstream   9   657000000 Hz    6000000 Hz  -0.8 dBmV   40.2 dB QAM256  0   0
11  Locked  SC-QAM Downstream   10  663000000 Hz    6000000 Hz  -0.8 dBmV   40.1 dB QAM256  0   0
12  Locked  SC-QAM Downstream   11  669000000 Hz    6000000 Hz  -1 dBmV 40.1 dB QAM256  0   0
13  Locked  SC-QAM Downstream   12  675000000 Hz    6000000 Hz  -1 dBmV 39 dB   QAM256  0   0
14  Locked  SC-QAM Downstream   13  681000000 Hz    6000000 Hz  -0.8 dBmV   40.2 dB QAM256  0   0
15  Locked  SC-QAM Downstream   14  687000000 Hz    6000000 Hz  -0.8 dBmV   40 dB   QAM256  0   0
16  Locked  SC-QAM Downstream   15  693000000 Hz    6000000 Hz  -0.7 dBmV   40.1 dB QAM256  0   0
17  Locked  SC-QAM Downstream   16  699000000 Hz    6000000 Hz  -1 dBmV 40 dB   QAM256  0   0
18  Locked  SC-QAM Downstream   17  705000000 Hz    6000000 Hz  -0.5 dBmV   40.4 dB QAM256  0   0
19  Locked  SC-QAM Downstream   18  711000000 Hz    6000000 Hz  -0.4 dBmV   40.3 dB QAM256  0   0
20  Locked  SC-QAM Downstream   19  717000000 Hz    6000000 Hz  -0.3 dBmV   40.4 dB QAM256  0   0
21  Locked  SC-QAM Downstream   20  723000000 Hz    6000000 Hz  -0.3 dBmV   40.4 dB QAM256  0   0
22  Locked  SC-QAM Downstream   21  729000000 Hz    6000000 Hz  -0.6 dBmV   40.1 dB QAM256  0   0
23  Locked  SC-QAM Downstream   22  735000000 Hz    6000000 Hz  -0.5 dBmV   40.2 dB QAM256  0   0
24  Locked  SC-QAM Downstream   23  741000000 Hz    6000000 Hz  -0.8 dBmV   40.1 dB QAM256  0   0
25  Locked  SC-QAM Downstream   24  747000000 Hz    6000000 Hz  -0.8 dBmV   39.9 dB QAM256  0   0
CM Upstream Channel Info
Channel Lock Status Channel Type    Channel ID  Frequency   Width   Power   Modulation/Profile ID
1   Locked  ATDMA   25  19400000 Hz 3200000 Hz  36.3 dBmV   1
2   Locked  ATDMA   26  24200000 Hz 6400000 Hz  37 dBmV 2
3   Locked  ATDMA   27  30600000 Hz 6400000 Hz  37.5 dBmV   2
4   Locked  ATDMA   28  37000000 Hz 6400000 Hz  37.8 dBmV   2

For signal junkies out there - I am on a Value 12 two-port tap, and between me and the node there is a single post-node amp. The majority of the plant is underground, and only goes arial for 800 feet before it reaches the node. My cable plant is 860Mhz but operates at 750Mhz.

Proven Sharer

Re: Inconsistent upload with DOCSIS 3.1 modems

It sure looks like you are the victim of poor network maintenance or training.  You've already taken a careful look at the DS signal levels to identify the tilt.  If the node is set up properly, the difference in received signal levels from lowest to highest should not exceed 3 dB.  In your case the lowest reported level is -1.0 dBmV (on CH11, 12, 16)  and highest is 5.2 (CH2), so the difference is 6.2 dB, which is far beyond the limit the Gigabit modem can handle.  Those numbers almost look like you've got either RG59 or waterlogged RG6 cable for the drop or in the walls, instead of clean and dry RG6. 


Re: Inconsistent upload with DOCSIS 3.1 modems

It would be at least an 8dB tilt if we also include the OFDM channel in there. The funny thing about the OFDM channel is that, itself, it has a 4dB tilt if analyzed from a spectrum analyzer. The tilt is literally in steps. Which almost looks to be an incorrect bias set on power output somewhere. I don't have my old SurfBoard modem connected right now to get a picture of that, but I do remember seeing that on the techs' meters.


The techs did perform checks at my tap as well and saw the tilt there as well, with power levels being around +2 for the lowest channels.. The drop into my house is direct-bury and is a few months old (also the really thick RG-11 with silicon filling stuff). In the walls is all RG-6.


I will mention that other modems on other nodes serviced out of the same headend also see the same sort of tilt. I've asked the techs about that, and some have cited old infrastructure (although my neighborhood's plant is from 2000, and the node was installed in 2010 to split the neighborhood off of the surrounding area, due to serious congestion issues). Others have cited amps which are shot. And a business tech I asked about the Gigabit issues has mentioned that the plant really wasn't supposed to be ready for Gigabit as soon as it was, so that makes me think there is a lot of work that needs to be done still.


I am still curious about what other DSLReports posters are mentioning. Many have better signals (far less tilt) than I do and they see the same issue with upstream, in a slightly different form.


Re: Inconsistent upload with DOCSIS 3.1 modems

It looks like others are starting to notice over at DSLReports:

Re: Inconsistent upload with DOCSIS 3.1 modems

I've got quite a bit of tilt on my downstream as well, and seem to see a lot of correctables (the stats I am posting are about 12 hours after a reboot).  I used to have trouble with getting uncorrectables at 705 and 711 MHz due to LTE ingress but replaced the run of RG6 from the modem to the demarc with quad shield and eliminated that.  My upstreams are on the low side as well.  I've actually got a 3 dB attenuator in line because the modem is the only thing in the house hooked up, we do all our TV via streaming and antenna, and the tech who installed my service a few years ago used a splitter to pad it down.  My modem is a SB8200 on the 400/20 tier.


I do see some odd fluctuations on upload speed but I don't really see it on speed tests, so I thought it might have been the VPN connection I was using or a peering issue with the receiving end since that location uses a different ISP.  I haven't done extensive analysis to see what's going on, however.


I'm not sure if the levels or tilt are worth a service call since things seem to work correctly otherwise and I typically see 480/23 on a wired computer speed test.



Downstream Bonded Channels
Channel IDLock StatusModulationFrequencyPowerSNR/MERCorrectedUncorrectables
1LockedOther338000000 Hz6.5 dBmV41.2 dB331937080
2LockedQAM256579000000 Hz2.1 dBmV40.4 dB1300
3LockedQAM256585000000 Hz2.3 dBmV40.4 dB1170
4LockedQAM256591000000 Hz2.7 dBmV40.7 dB1290
5LockedQAM256603000000 Hz2.1 dBmV40.2 dB1750
6LockedQAM256609000000 Hz1.8 dBmV40.0 dB1470
7LockedQAM256615000000 Hz2.0 dBmV40.1 dB2270
8LockedQAM256621000000 Hz2.0 dBmV40.0 dB2680
9LockedQAM256627000000 Hz1.3 dBmV39.6 dB3200
10LockedQAM256633000000 Hz1.2 dBmV39.4 dB3050
11LockedQAM256639000000 Hz1.4 dBmV39.8 dB2740
12LockedQAM256645000000 Hz0.8 dBmV39.2 dB2590
13LockedQAM256651000000 Hz1.4 dBmV39.6 dB2290
14LockedQAM256657000000 Hz0.7 dBmV39.2 dB3200
15LockedQAM256663000000 Hz0.8 dBmV39.2 dB2530
16LockedQAM256669000000 Hz0.4 dBmV39.1 dB2700
17LockedQAM256675000000 Hz0.4 dBmV38.2 dB2630
18LockedQAM256681000000 Hz-0.7 dBmV38.3 dB2320
19LockedQAM256687000000 Hz0.3 dBmV38.8 dB1880
20LockedQAM256693000000 Hz-0.2 dBmV38.6 dB1590
21LockedQAM256699000000 Hz0.6 dBmV39.2 dB1640
22LockedQAM256705000000 Hz0.1 dBmV38.8 dB1630
23LockedQAM256711000000 Hz0.9 dBmV39.3 dB1340
24LockedQAM256717000000 Hz0.8 dBmV39.2 dB1210
25LockedQAM256723000000 Hz0.4 dBmV38.9 dB1410
26LockedQAM256729000000 Hz-0.4 dBmV38.3 dB1440
27LockedQAM256735000000 Hz0.6 dBmV39.0 dB1110
28LockedQAM256741000000 Hz-0.3 dBmV38.4 dB920
29LockedQAM256747000000 Hz0.1 dBmV38.6 dB690
30LockedQAM256753000000 Hz-0.2 dBmV38.3 dB800
31LockedQAM256759000000 Hz-0.2 dBmV38.3 dB840
33LockedQAM256597000000 Hz2.7 dBmV40.6 dB1510


Upstream Bonded Channels
ChannelChannel IDLock StatusUS Channel TypeFrequencyWidthPower
19LockedSC-QAM19400000 Hz3200000 Hz33.0 dBmV
210LockedSC-QAM24200000 Hz6400000 Hz34.0 dBmV
311LockedSC-QAM30600000 Hz6400000 Hz35.0 dBmV
412LockedSC-QAM37000000 Hz6400000 Hz36.0 dBmV




Re: Inconsistent upload with DOCSIS 3.1 modems

It's funny that you mention the speed tests not showing the problem. That is true in my case, although with a slight variation. If the upload is working correctly, the speed test will get to 42Mbps and stay there steady. With the way my upload is behaving right now, the speed test reports 32-38Mbps and is inconsistent. The only speed tests that do report the full upload speed are those that use more than one TCP connection for the test.

I see more people on DSLReports starting to report upload behavior problems. This can't be exclusive to just a few people...

Re: Inconsistent upload with DOCSIS 3.1 modems

Spectrum recently replaced my TM1602 modem with a DOCSIS 3.1 E31N2V1 (Hitron) modem.


My upload speeds went from a consistent 23mbps to being inconsistent, usually around 15mbps.    This behavior reminded me a lot of how an old DSL line I had performed when traffic shaping was missing on the ISP side, so I did some experiments:

Traffic shaping OFF

Screen Shot 2019-04-30 at 9.05.05 PM.png

Traffic shaping ON :  Codelq @ 24500
Screen Shot 2019-04-30 at 9.02.17 PM.png


With traffic shaping on I now consistently get 22-23mbps up.    This suggests that at least in my case, the inconsistent upload rates are a result of incorrect provisioning of some sort, or an issue of some kind with the modem's built in rate limiting.