usefulness of upgrade from actiontec 2.0 -> 2.5?

[astrorob]

Hi All,

i've currently got a 4-node actiontec moca 2.0 network going in the house. with all the WFH going on, i started noticing intermittent problems with wifi. long story short, these problems seem to really be moca backhaul issues.

first off, i'm seeing asymmetric bandwidth between certain nodes; some run at 800mbit in one direction and 100 in the other.
secondly, i've got some nodes that run only at ~100mbit in both directions. both of these measurements were taken with a mostly quiescent network.

i'm pretty sure that when i commissioned the network 2 years ago, i was getting ~800mbit everywhere bidirectionally. at this time i'm not sure what, if anything, changed.

most of the coax in my house is old, and buried in the walls, so if this is a signal integrity issue, there's not much i can do about it. all of my splitters are 2.5GHz so i think those are probably OK. i need to run iperf3 exhaustively between all the nodes to really understand if there's a common element here. i also need to check the status pages of all the nodes to see what they thing their margins are.

but, assuming i can't find any physical problems, the question is - are moca 2.5 adapters more resilient to signal integrity issues? if i could get to 500mbit bidirectional everywhere, i'd be happy.

beyond that i'm starting to wonder if something like the orbi pro wifi6 ax6000 system would be a better way to connect the two floors of my house. i do have a lot of "always on" security camera traffic so i'm somewhat suspicious that this would be a good solution in general.

thanks for any advice!

 

[astrorob]

from reading more threads here, i'm wondering if either my cablemodem is stepping on something, or i have some very fundamental problem in the plant. one thing i can't remember is when i upgraded my modem, but i believe it was before installing the ECB6200s. the 6200s are set to extended D band with a center frequency of 1100Mhz. i'm thinking of moving that to D-high to see what happens.

one thing i see is that apparently the TX power is maxed out on all adapters, but not consistently. still, they are all +1 - +3 dB which sounds like a problem.

having said all this i don't have any problems with the internet as far as i can tell, so the modem itself is happy. but i am going to try powering down the modem and seeing if anything changes on the moca network, and vice versa.

 

[degrub]

re-seat and tighten all of your coax connections. Any open ends ? cap them.

what other changes have occurred over time ?

Check the diagnostic page on the moca modems. That will give you the link sync rate. You may have to start with modem disconnected to the interior cable plant and run testing on each modem pair (pwer off the others) to find the issue. If they are multidrop, they share the entire bandwidth among them, so all you may be seeing is that effect. However, one pair at 100 mbit sounds off.
Is your cable modem using MOCA ? DOCCiS 3 or 3.1 ? Check with your ISP about what bands it is using if you cannot determine locally.
Do you have to share cable with the modem ? if not, put in a MOCA block filter between them.

All MOCA 2.5 does is use 5 of the available bands in parallel. If your modem is using 1 or more you will loose that portion anyway, so less advantage, even if going to D-high band (you would loose one anyway, but avoid interference perhaps).

 

[astrorob]

thanks for the repiy.

i reseated everything yesterday; all open connections are (and were) terminated with 75 ohm caps.

8 channel modem - no moca. currently using channels 40-47 so 645Mhz-687Mhz. upstream channels 5-8. i don't actually think there's any conflict here. i did disconnect the modem and observed no change on the moca network.

part of the 100Mbit problem was that a single NIC on one machine autonegotiated to 100MBit for some reason. i happen to have an airport extreme hanging off that moca node and 802.11ac clients show much higher bandwidth at that node. however, there were other moca endpoints in the system connected to machines with gigabit nics which see only 30MB/sec sometimes, including the very asymmetrical one which of course has a gigabit NIC.

for fun i switched to Band D High from Band D extended and at first things seemed to be working better. however after a couple of hours the network returned to its old slowness. it may have just been power-cycling all of the moca nodes that improved things for a while. the bit loading diagrams and powers looked pretty bad both before and after this switch.

after all the WFH is done today i might try going down to Band D low and see if there are any changes... in theory the transmission line losses should be lower at lower frequencies, right?

based on the bit loading diagram and moca spec it seems that moca 2.0 is using 2 bonded channels, is that correct? 2.5 would increase that 2.5x to 5 channels of the same bandwidth? seems like the channels need to be narrower given the ~500MHz total bandwidth available.

anyway, i can't point to anything that has changed. we did have our electric service upgraded a couple of months ago but aside from being near where the cable service enters the house from underground, that doesn't seem related. my cable plant is pretty typical i think - home run from comcast entry to the attic, where a splitter feeds all the rooms. currently i have a 4-1 moca capable splitter there. there are 3 loads from that splitter. where the service enters the house there is a 2-1 moca splitter with one output going to the attic and the other going to a moca endpoint on the first floor. there's a POE filter in front of that first 2-1 splitter.

here is the node map about a half-hour after rebooting all the nodes (after the switch to D-high)

here is one this morning, looks quite different:

the node this was run on is the one that has asymmetrical BW problems when talking to a9:b4 at least. it is exhibiting that problem currently.

 

[degrub]

so you have analog cable service or DOCCIS 3.0 ? RG6 or RG59 cable plant ?
do you have a moca POE filter at the head end of the moca network to prevent signal going into the modem or out to the cable demarc ?

A sketch of your actual hookup cable paths, devices, would help.

try just two modems active at a time and see what the bit rates do without the cable modem connected on the house side. power off the other moca modems.
If multiple nodes are are trying to communicate on a multidrop cable arrangement, they will share the same bandwidth. A star arrangement, while using more moca adapters in paris will provide the highest throughput.

 

[astrorob]

DOCSIS 3.0 8 channel modem. RG6 plant (house built in the late 90s). no analog cable, no digital TV service, just internet.

the POE filter is right at the cable demarc. the modem is connected to the TV/STB port of one of the 6200 boxes. i will draw the network properly and upload it.

when i'm testing the network is mostly quiescent. there are a lot of devices in the network and they are always sending traffic but i've shut off the heavier loads (security cameras) when testing...

 

[degrub]

can you physically partition the camera and recorder network from the regular internet users ?

or instead partition the bandwidth sensitive internet users away from the rest ?

 

[astrorob]

it turns out the cameras are all connected to a single moca endpoint (via a switch with POE ethernet connections) and there happens to be a wifi access point there too, but that one is lightly used. i don't think the security camera traffic is what's borking the network as i've stopped it to run the iperf3 tests. maybe i should stop it before looking at the bit loading diagrams to see if it has something to do with that. but all nodes report phy rates in the 600mbit and if they actually ran at that speed in both directions i'd be happy.

in total the cameras account for 40mbit/sec of traffic. it flows to the node which sometimes shows asymmetric bandwidth. as it turns out in the camera->computer direction the bandwidth on that node is always good. it's the opposite direction which sometimes slows way down. so maybe it's not that important but it is weird.

maybe another experiment would be to move the moca adapters around and see if the trouble follows the adapter or the physical location.

the real problem is the main downstairs wifi access point, where the moca network sometimes only gives 20Mb/sec in both directions to the node connected to the router. that really messes up zoom, etc.

part of the problem while debugging is the variable nature of the bandwidth issues. i was fooled last night into thinking the problem was fixed but as mentioned it returned eventually.

 

[astrorob]

i still owe the network diagram - haven't had time to work on this too much.

for fun i just tried to switch all the moca endpoints to band D low, and the behavior was very strange. not only could the endpoints not synchronize with one another but they started rebooting themselves continuously. i had to disconnect all the coax from them (which stopped the rebooting) and set them all back to band D extended.

one more puzzlement is that the "slow" node has now migrated to a different place (security cameras, but there is still enough BW). very strange. however, as noted before the phase of the moon seems to affect this network so that may not mean too much. i havent had time to selectively take them offline in turn and test the performance... just too many users wanting to connect to the internet.

i feel like one or more of these 6200s has become marginal and might be borking the network. but that's just a hunch for now.

 

[degrub]

The order that you power on the moca modems determines how they sync. Possibly not all modems were on the same band range when the sync went crazy ? Btw, i believe you have to power them off then back on to get the band adjustment to change.

are you sure your 20Mbit/s rate in the basement was on the moca coax side and not on the wireless side ? Was that a measured throughput rate or a link rate ?

you might try this - put a two way splitter just in front of the cable modem (isp side). Connect the cable modem to one leg. Connect a POE moca block to the other leg followed by the cable to the first moca modem. Connect the moca modem ethernet port to the lan port on the cable modem/router. I assume the cable modem is actually a modem/router combined unit.

 

[astrorob]

yes, i did have to do the sync changes piecemeal so there was a period where they might have been out of sync with each other, but at some point they were all set to the low band and powered off, then powered on. i did notice that they needed to be power cycled when switching to D-band high the other day. that transition was successful; not sure why i couldnt get the low band to work.

i am using iperf3 to measure everything. i'm using machines with 1000-baseT ethernet nics wherever possible. the one machine that autonegotiated to 100mbit does have a 1000-baseT port so i'm not sure what's wrong with it. anyway, throughout all of this the phy rates are generally in the 600MBit range.

i should have mentioned, the cablemodem is an arris standalone modem. my router is a separate box.

i'll have to buy another POE filter to try out your idea. in the past i did have the modem connected using a splitter, but at some point i realized the "CATV" output on the arris should do the same thing.

i drew the network on my daughter's ipad but it came out like crap - i don't know how to use notability. so i'm going to do it again using inkscape.

 

[degrub]

Swap out the cable on the device that went to 100mbit if you didn’t already.

if i remember correctly, 600 mbit is normal for bonded pairs. The communication is half duplex, so throughput can be half that at max (dedicated single pair). Then that is shared amongst all the drops. Really, anything 50 mbit/sec throughput or greater is all you really need. You would only see benefit if local transfers. What the moca 2.5 devices can give you is higher throughput if you really need it. For example, i have a pair of moca 2.5 modems serving as the link to my ATT ont device from the ATT remote router ( in a wiring closet). This allows for 1 Gbit full duplex to maintain the ATT gig fiber speed. My internal lan across other coax is all actiontec 6200 modem pairs in a star pattern from a central managed switch. I get up to 450 mbit/sec full duplex transfer rates where i have devices supporting it, but i don’t really require over 50 mbit/sec full duplex for any usage.

 

[astrorob]

well, this is kind of crazy. the security cam moca endpoint got really flaky, dropping the link for 10s every 2-3 minutes. on a whim i changed out the pigtail connecting it to the wall and things improved a lot. i think the whole plant must be teetering on a knife's edge. maybe this endpoint is going bad... i can't see how a 2 ft section of RG6 just goes bad after 2 years. it's not subject to any harsh conditions and has no bends/kinks.

 

[L&LD]

These are electrons, not earthmovers. We don't need to see it; we just need to fix it.

 

[degrub]

it is usually the attachment of the connector - cable, wall, face plate - that gets wonky over time. A single strand of the shield touching the central conductor or if the shield is not terminated through the crimp on connector can do it. crimp on connectors can work loose. Any place there is mechanical stress or a connection is a candidate. Sometimes a nail into the wall can hit a cable.

 

[astrorob]

this particular segment was factory-made, no crimps. i can't speak to the wall plate connection as i haven't taken it apart. what i really need here is a TDR/cable analyzer and then maybe i could get a handle on where the weak links are. the bit loading diagrams and RX/TX power seem to be unchanged.

one interesting thing is that i put an older YiTong endpoint into the network - there was one unused output on the attic splitter so i connected that to another bedroom. the actiontec node map claims that the yitong is running Tx power of -13dBm while the actiontecs are all at +3. unless this is a display/reporting bug it seems kind of suspicious, like maybe the actiontecs are putting out way too much output power.

 

[astrorob]

i replaced the problematic node with a goCoax bonded 2.5 and the link is now stable for 15+ hours. i guess the newer silicon has a bit more noise margin? anyway, we'll see how it goes. might replace all the nodes with the 2.5 adapters if it goes well.