Newbie MoCA questions

[gneuner2]

Hi all,

A little background:
I've been working with Ethernet for ~30 years, but I'm a newbie with MoCA, so please forgive me if my questions (below) are naïve.


I'm on Xfinity cable, which enters (and splits) in the basement in a closet at one end of the house. The closet is shrouded by a metal beam, so there are WiFi access points connected to a wired business/workgroup router and cable modem in the closet.

The house is an open plan with no good location for a central WiFi router, so there are two SOHO WiFi routers configured as access points, one at each end of the house, which also serve wired clients.

I was using Powerline plugs to connect the "other end" AP to the router - but when one of the plugs died, I decided to try MoCA because there was existing coax available. So I got a pair of ScreenBeam ECB6250 (bonded 1Gbps) to experiment with.

My setup:

​

The results have been mixed: the MoCA bridge is more stable and faster than the Powerline, but testing with iPerf shows that a single TCP connection can achieve only 100..120 Mbps. I can get up to ~930 Mbps between wired clients across the MoCA bridge, but only in aggregate by using multiple TCP connections.

After the first few connections, the bandwidth of individual connections drops off slowly, but aggregate bandwidth keeps rising until the bridge is saturated with 15..18 simultaneous connections.

This is in contrast to Ethernet where a single TCP connection can use all the bandwidth.

Is this to be expected? I know MoCA creates multiple frequency channels, but I would have thought that the controller would spread Ethernet traffic across the channels to try to use all the bandwidth. Or is it just the case that the particular equipment I have doesn't do this?

Thanks,
George

 

[krkaufman]

testing with iPerf shows that a single TCP connection can achieve only 100..120 Mbps. I can get up to ~930 Mbps between wired clients across the MoCA bridge, but only in aggregate by using multiple TCP connections.

This is common with iPerf testing of MoCA; 4+ parallel streams are required when using iPerf. Maybe @gocoax can shed some light on why.

I'm on Xfinity cable

Given your cabling and device locations, it would be preferable to split the cable modem signal off prior to the amplifier using a 2-way splitter, then installing a 70 dB “PoE” MoCA filter on the input port of the amplifier (as you currently have) to isolate the ISP/modem feed from your MoCA-infused coax. Also, you might get better MoCA performance/efficiency using a “designed for MoCA” amp, like the equivalent PPC model# PPC-5M-U/U.

This won’t alter the iPerf behavior but it might improve the MoCA PHY rates or power usage; and it will prepare your setup for DOCSIS 3.1. Related to:

• DOCSIS encroachment on the MoCA [Band D] frequency range

 

[gneuner2]

This is common with iPerf testing of MoCA; 4+ parallel streams are required when using iPerf. Maybe @gocoax can shed some light on why.

If I understand correctly, you are saying that MoCA will limit a single TCP connection to the bandwidth of a single coax channel and *not* try to spread over all available channels.

That's a little disappointing.

Internet-only or do you also subscribe to cable TV, or might you?

Given your cabling and device locations, it would be preferable to split the cable modem signal off prior to the amplifier using a 2-way splitter, then installing a 70 dB “PoE” MoCA filter on the input port of the amplifier (as you currently have) to isolate the ISP/modem feed from your MoCA-infused coax. Also, you might get better MoCA performance/efficiency using a “designed for MoCA” amp, like the equivalent PPC model# PPC-5M-U/U.

The X1 boxes are TV (not game consoles).

I can try isolating the modem, but I'm not sure what that really accomplishes. The modem is *not* MoCA and should ignore it completely.

WRT the amplifier: the X1 boxes are MoCA and they seem to have to talk to one another to work. Xfinity called one of them the "master" and the others "slaves". Also, Xfinity installed that amplifier, so I figured it was ok.

This won’t alter the iPerf behavior but it might improve the MoCA PHY rates or power usage; and it will prepare your setup for DOCSIS 3.1. Related to:

• DOCSIS encroachment on the MoCA [Band D] frequency range

The modem is DOCSIS 3.1. Don't know about the X1 boxes or the underlying service.

 

[gneuner2]

This is common with iPerf testing of MoCA; 4+ parallel streams are required when using iPerf. Maybe @gocoax can shed some light on why.

On one of the wired clients of the MoCA bridge I have two separate instances of Ookla Speedtest which can be run simultaneously. This is the desktop app (not in browser) and it uses just one TCP connection to the test server.

Running either Speedtest singly, I get 110..120 Mbps. Running both simultaneously, I get 110..120 Mbps from both.

So it's not just an iPerf artifact ... it seems to be a real situation.

 

[tgl]

I have a pair of ScreenBeam ECB7250 adapters, which are claimed to support 2.5Gbps, and I have actually measured single-threaded iperf3 rates of 2.32Gbps through them. So I don't believe the claim that there's something inherent in MoCA that creates a 120Mbps-per-connection limit. Perhaps those ECB6250 units behave differently, or maybe the bottleneck is somewhere else.

I should mention that my setup has a dedicated coax line between the two adapters, nothing else connected to them at all. So maybe there is something about your rather complex setup that is creating a bottleneck on the coax line. I know next to nothing about proper MoCA installs, but I do know that if you're actually trying to put both cable TV and MoCA signals on the same wire, you're supposed to install filters to keep those signals from leaking to the wrong places. Your diagram mentions no filters, so I wonder if there's more you need to do there.

 

[krkaufman]

If I understand correctly, you are saying that MoCA will limit a single TCP connection to the bandwidth of a single coax channel and *not* try to spread over all available channels.

No idea. I was just confirming that iPerf needed parallel streams to max the throughput; couldn’t suggest why.

And it would seem to be particular to something about iPerf testing, else people would have real issues with MoCA, generally.

 

[gneuner2]

I have a pair of ScreenBeam ECB7250 adapters, which are claimed to support 2.5Gbps, and I have actually measured single-threaded iperf3 rates of 2.32Gbps through them. So I don't believe the claim that there's something inherent in MoCA that creates a 120Mbps-per-connection limit. Perhaps those ECB6250 units behave differently, or maybe the bottleneck is somewhere else.

I should mention that my setup has a dedicated coax line between the two adapters, nothing else connected to them at all. So maybe there is something about your rather complex setup that is creating a bottleneck on the coax line. I know next to nothing about proper MoCA installs, but I do know that if you're actually trying to put both cable TV and MoCA signals on the same wire, you're supposed to install filters to keep those signals from leaking to the wrong places. Your diagram mentions no filters, so I wonder if there's more you need to do there.

There are no filters because all the lines are carrying MoCA. The X1 TV boxes form a private network, logically separate from the ECB6250 network but not physically separate: unfortunately, one line is shared between them. There is no way around that.

I don't know what kind of filters could separate the traffic - according to the ECB6250s, it's all MoCA 2.0 and 2.5. AFAIK, they use the same frequencies.

 

[gneuner2]

No idea. I was just confirming that iPerf needed parallel streams to max the throughput; couldn’t suggest why.

And it would seem to be particular to something about iPerf testing, else people would have real issues with MoCA, generally.

Well, I don't see it only with iPerf ... I see it with nearly everything. The only activity that really seems to benefit is web browsing, because most pages pull from multiple sources with separate connections.

E.g., I use Fastcopy which reports its speed. Copying one large file across the bridge maxes out at ~11MBps (110Mbps). Copy two files simultaneously both max out at ~11MBps. With three or more, the speed of the individual copies starts to drop.

I see the same with backups to the NAS. The backup can drive ~70+MBps (which I see on the ethernet wired clients), but across the MoCA bridge it gets only ~11 ... better than across the old Powerline bridge, but still disappointing.

 

[krkaufman]

View attachment 55447

The PHY rates for the MoCA 2.5 link are abysmal; should be up near 3500 Mbps for a 5-channel connection. You might try experimenting with the adapters direct connected, or some other means allowing them to connect bypassing the amplifier. (Then upgrading to an amp designed for MoCA.)

 

[gneuner2]

The PHY rates for the MoCA 2.5 link are abysmal; should be up near 3500 Mbps for a 5-channel connection. You might try experimenting with the adapters direct connected, or some other means allowing them to connect bypassing the amplifier. (Then upgrading to an amp designed for MoCA.)

I don't know what the PHY rates should be. The ECB6250 only claim 1Gbps. I believe that because I can get ~900+Mbps aggregate (multiple connections), but I can't get it from any single connection.

I can bypass the amplifier, but I can't isolate the line because it is shared with an X1 box and the X1s have to be able to talk to one another.

 

[kokishin]

This is common with iPerf testing of MoCA; 4+ parallel streams are required when using iPerf. Maybe @gocoax can shed some light on why.



Given your cabling and device locations, it would be preferable to split the cable modem signal off prior to the amplifier using a 2-way splitter, then installing a 70 dB “PoE” MoCA filter on the input port of the amplifier (as you currently have) to isolate the ISP/modem feed from your MoCA-infused coax. Also, you might get better MoCA performance/efficiency using a “designed for MoCA” amp, like the equivalent PPC model# PPC-5M-U/U.

This won’t alter the iPerf behavior but it might improve the MoCA PHY rates or power usage; and it will prepare your setup for DOCSIS 3.1. Related to:

• DOCSIS encroachment on the MoCA [Band D] frequency range

Hi krkaufman,

As you indicated, it appears he needs a MoCA 2.x rated amp. The amp he's currently using has a 0dB forward gain up to 1002MHz and worse a 0dB reverse gain up to 42MHz.

What do you think of this amp?
https://www.amazon.com/dp/B09C2R9VJC?tag=snbforums-20

 

[krkaufman]

What do you think of this amp?
https://www.amazon.com/dp/B09C2R9VJC?tag=snbforums-20

Its specs document maps to MoCA requirements, or at least fully documenting performance/characteristics for MoCA frequencies. Like most amps, it has just 35+ dB attenuation of MoCA signals on the input, so an additional MoCA filter on the input line would still be recommended to fully snuff MoCA signals. And specs indicate from 25-35 max attenuation between all output ports. (This amp’s specs get a little more detailed on that front.) Top of my head, can’t say how these figures compare to other amp’s. Comparable to an Antronix 8-way MMC1008H-B passive splitter w/ a “PoE” MoCA filter on its input port, so looks good.

The amp he's currently using has a 0dB forward gain up to 1002MHz and worse

Worst is who knows how it’s messing with MoCA signals. It certainly doesn’t seem to be doing them any good.

 

[gneuner2]

kokishin said

As you indicated, it appears he needs a MoCA 2.x rated amp. The amp he's currently using has a 0dB forward gain up to 1002MHz and worse a 0dB reverse gain up to 42MHz.

What do you think of this amp?
https://www.amazon.com/dp/B09C2R9VJC?tag=snbforums-20

krkaufman said

Its specs document maps to MoCA requirements, or at least fully documenting performance/characteristics for MoCA frequencies.

Worst is who knows how it’s messing with MoCA signals. It certainly doesn’t seem to be doing them any good.

I'm willing to change out the amplifier if you think that will help.
Xfinity installed the one I have, and since their own TV boxes are using MoCA I figured it was ok. [And maybe it is if they just pass small messages.]

 

[kokishin]

I'm willing to change out the amplifier if you think that will help.
Xfinity installed the one I have, and since their own TV boxes are using MoCA I figured it was ok. [And maybe it is if they just pass small messages.]

That amp is bandwidth limiting your full MoCA capability, especially the capability of the ECB6250. Comcast is probably using MoCA 1.x.

 

[gneuner2]

Update: I tried bypassing the amplifier, injecting the ECB6250 signal directly into the shared cable using a 2-way splitter (5-2300 MHz -3.5db power pass). The TV end is/was split out the same way. Taking the amplifier out of line made no difference in performance across the bridge.

I do have a MoCA filter which I bought with the ECB6250s and didn't use because Xfinity already had one on the input. Would putting that filter between the shared cable and amplifier help? Or would it screw up whatever Xfinity is doing?

Guessing I still have to get a better amplifier.

 

[gneuner2]

Oh, and I did cap the unused port on the amplifier.

 

[degrub]

Take your Xfinity moca nodes offline. The moca modems will try to sync with each other, but may end up syncing with the xfinity gear instead. This will rule out cable issues. There are several threads here discussing boot order of moca nodes leading to working/not working, reduced bandwidth specifically with xfinity type nodes.

 

[gneuner2]

Take your Xfinity moca nodes offline. The moca modems will try to sync with each other, but may end up syncing with the xfinity gear instead. This will rule out cable issues. There are several threads here discussing boot order of moca nodes leading to working/not working, reduced bandwidth specifically with xfinity type nodes.

I can try that - and a few other things that have been suggested.

Unfortunately, in this house it's tough to find a block of time to work on things. I work during the day, the wife teaches online most evenings (including weekends), and my mother who lives with us has progressive dementia - watching shows she likes helps keeps her calm. There's only a few hours in the (very) early morning when she is asleep to take TV offline without causing chaos. And I like to sleep too.

It may take a little while to try some of these things and report back. But I will report back and I thank everyone for helping out.

 

[gneuner2]

Ok, I had a chance to experiment a bit.

I bypassed the amplifier injecting the ECB6250 signal directly into the shared cable. The PHY rates are much higher [shown below], but *single* connection throughput remained stuck at ~11 MB/s.

Then I inserted a MoCA filter inline between the injection splitter and the amplifier. PHY rates remained about the same, but single connection throughput rose to ~50MB/s. However, the TV on the shared cable didn't work. Clearly the X1s are communicating with MoCA 2.x despite the amplifier cutting off at 1002 MHz.

I have left the ECB6250 injecting into the shared cable (bypassing the amp) - but I have removed the filter so the TV works. The PHY rates remain good, but I am still stuck at ~11MB/s.

Didn't have enough time to try letting the ECB6250s sync with no X1s online ... they don't have off-switches (have to unplug them) and that's a lot of bother and running stairs.

 

[degrub]

There was a case here a few years back where turning off all of the tv boxes and the moca modems, then only powering up the moca modems first allowed them to link first. Data transfer and sync rates were normal for the moca spec. Then the TV boxes were powered on. Everything continued to work with the moca nodes at expected rates.

If there was an electrical outage, on restore, the moca modems and tv devices would power up in a random fashion. Then the user saw much the same as you are describing. Apparently, the moca "Master" xfinity node was controlling/limiting all devices. Not sure how that works. It may be that they have started using the higher bands since that is a doccis 3.1 modem and the two ECB devices cannot get enough bandwidth.

Power on order mattered for some reason in the case above..

At lease the cable is showing correct sync rates which means nothing wrong with the cable. The amp/splitters are clearly interfering with the higher 3 moca bands. There may be enough power budget in the xfinity modems that are using moca2 and the lower two bands to still communicate.

You could try setting up a private moca network by isolating the two moca modems on a short piece of coax and setting up security codes for just the two modems. Then put them back in the network as you have it now and see what happens.