Dismiss Notice

Welcome To SNBForums

SNBForums is a community for anyone who wants to learn about or discuss the latest in wireless routers, network storage and the ins and outs of building and maintaining a small network.

If you'd like to post a question, simply register and have at it!

While you're at it, please check out SmallNetBuilder for product reviews and our famous Router Charts, Ranker and plenty more!

Test results, HomePlug 'round the house

Discussion in 'MoCA, HomePlug, HPNA' started by stevech, Aug 3, 2011.

  1. stevech

    stevech Part of the Furniture

    Mar 13, 2010
    San Diego
    EDIT: You should probably ignore the individual test results below where the "office power strip was used".

    Let's discuss these informal tests using HomePlug AV pair of devices around my 1800 sq. ft. townhome.
    The devices are TRENDnet TPL-303E. These use an Atheros chipset that produces up to 200Mbps in the physical layer (PHY).
    Geek diversion: The 200Mbps is the frame burst rate including bits for error correction and management that add overhead, as in any signal carrier communications - wired or wireless. The "modulation order" is the choice of how many information bits/sec to attempt to send in the channel's PHY. With weaker signals or interference, the PHY might still be at 200Mbps, but the information bit rate may have been reduced to cope with the poor channel.​
    So reducing the information rate to cope is like talking slower on a noisy/echoey phone connection to get the information through. Morse code (dit dit dah dah) is the epitome of this.

    Other/newer HomePlug AV products increase the PHY bit rate to 500Mbps. But the transmission media stays the same, so to reap the benefit of this higher speed, the medium needs to have fewer impairments (as discussed below). In wireless, like WiFi, the same applies, but the impairments in wireless are far greater because the ether is very encumbered!

    How this compares to MoCA (signalling over TV coax) is discussed later.

    Test Method:
    • Two rather fast PCs, each with a static LAN IP address
    • PC #1 is a laptop connected to HomePlug AV device via ethernet
    • PC #2 is a desktop connected to a gigE switch and to that switch is also connected a HomePlug AV device.
    • Thus, the PCs and HomePlug devices communicate via the switch.
    • Test throughput using "jperf", this being a iperf with graphs/charts.
    • iperf in server mode on laptop that moves around; jperf on desktop to see througput versus elapsed time
    First a baseline, connecting both PCs to the switch, no HomePlug devices in use. Just PC to PC via the gigE switch...
    • One IP stream, one-way, 300KB TCP window size: about 900Mbps.
    With the default smaller window size, (about 50KB) the throughput is of course much less. As expected, using a smaller window size, with many simultaneous streams from iperf, the aggregate speed approaches the 900Mbps capacity of the switch and TCP header inefficiency and Windows' TCP stack software overhead (which makes it hard to get to Gbps with real TCP traffic).

    The next baseline:

    Use the two HomePlug AV devices, but plug both into same power strip. Use a large TCP window size so that one stream can fill the capacity of the link channel.
    • One IP stream, one-way, large 300KB TCP window size: about 70Mbps.
    This seems to be the max; adjusting window size, etc, the speed didn't get above this number when using TCP.

    Now to the tests with impaired transmission paths on the power line. The impairments are distance (wire length) and signal attenuation due to items plugged into AC outlets that have internal signal "sinks" for the HomePlug AV transmission frequencies. In lay terms, this is like pinching the water hose.
    So we connect PC 1 and PC 2 to various AC outlets and measure each. Each pair of outlets "sees" different impairments to the other AC outlet. The impairing devices might include: GFI interrupter outlets, UPSes, so-called surge-protecting plug strips, major appliances, and so on. The closer the "sink" device is to the HomePlug device, in terms of in-wall routes of wires, the more of an impairment it is.

    Another impairment is when the modem pair are plugged into different AC power phases (in a home with split-phase as in No. America). So outlet one runs to the breaker panel to a breaker tied to phase A, and the other outlet wires to phase B. In a home with 220VAC appliances, phase A and B run in the same jacket cable (romex if you will) and the high frequency signals tend to couple by accident from one phase to another. Some houses have no such cable (no 220V appliances). Some coupling might occur inside a 220V appliance). Ideally, and these exist, the breaker box has a phase coupler which connects the phases together with low impedance at the high frequency, but has no affect on the 60Hz power signal.

    Bear in mind: Outlets in the same room can be on different circuit breakers. So the wire runs from outlet A and B may both go back to the breaker panel even though the outlets are in the same room. Or not. Perhaps they tie together in the attic in a J-box. Hard to say without a lot of dirty work.

    Below, the outlet "Office plug strip" has a LOT of electronics devices plugged in: PC, router, switch, many wall warts, LCD monitor, etc. The data suggests that this is NOT the right place to plug in the HomePlug AV modem.

    1. PC 1 at Office plug strip. PC 2 at LR wall outlet shared with home theater stack of things. 1.5 - 2Mbps wow.
    2. As above, but PC 2 at LR outlet #2, on a different wall: 1.5 - 2Mbps
    3. As above, but PC 2 at kitchen counter outlet #1 4 - 5Mbps
    4. PC 1 at LR outlet 1; PC 2 at kitchen counter outlet #1 20-22Mbps
    5. PC 1 at LR outlet #3; PC 2 at kitchen counter outlet #1 20-22Mbps
    6. PC 1 at office power strip; PC 2 at LR outlet #3 15-17Mbps
    7. PC 1 downstairs bath wall; PC 2 at LR outlet #3 50-60Mbps
    8. PC 1 downstairs bath wall; PC 2 at Kitchen counter #1 42-45Mbps
    9. PC 1 downstairs bath wall; PC 2 at Kitchen counter #2 25-26Mbps
    10. PC 1 at office power strip; PC 2 at Kitchen counter #2 0.0Mbps, the modem's link light stayed dark, power cycle, stayed dark.
    11. PC 1 at LR wall #3; PC 2 at upstairs BR 1 40-43Mbps
    12. PC 1 at kitchen counter #1; PC 2 at upstairs BR 1 43-45Mbps
    13. PC 1 at kitchen counter #2; PC 2 at upstairs BR 1 40-43Mbps
    14. PC 1 LR wall #3; PC 2 at upstairs BR 2 17-18Mbps
    15. PC 1 office power strip; PC 2 at upstairs BR 2 4-5Mbps
    16. PC 1 LR wall at home theater; PC 2 at upstairs BR 2 10-12Mbps
    17. PC 1 LR wall at home theater; PC 2 at upstairs BR 2 at analog TV outlet 20-22Mbps
    18. PC 1 LR wall outlet #3; PC 2 at upstairs BR 2 at analog TV outlet 38-40Mbps
    19. PC 1 garage washer/dryer (120VAC) mach outlet; PC 2 at upstairs BR 2 at analog TV outlet 43-45Mbps
    20. PC 1 office power strip; PC 2 at upstairs BR 2 at analog TV outlet 28-30Mbps
    Some pondering points:
    • Office power strip plugged into outlet is on wall common to breaker panel in garage
    • Home Theater outlet also has many signal sinks
    • BR 2 is above office; short run to breaker panel
    • One breaker in panel runs to all outlets near sinks/bath (builder put a single GFI outlet in garage, daisy-chained all wet area outlets to it. I guess GFI outlets were expensive in the 1980's when home was built.

    So this is a bit anecdotal. But it's clear than the office power strip and/or the electronics items plugged in there attenuate significantly. I wanted to correlate lower throughput with where outlets are on opposite phases. But to do so, I have to look at the vague markings in the breaker box and/or flip breakers to try to determine what's on what phase. Need to do that on a rainy day when I'm home alone and don't mind resetting clocks and stuff.

    Let's consider all this versus MoCA.

    Of course, MoCA is viable only when there's TV coax at the two places to be IP connected. Like a home theater rig in a room where it's too hard to run CAT5. The other end would be on a TV coax outlet somewhat near the home LAN switch or router, or close enough to run cat5 to the MoCA device on that end.

    On Impairments

    Interference: A proper TV coax system, and choice of MoCA frequencies, would say there's negligible impairment from interference. The Power Lines for HomePlug may be interference free at the high frequencies, but there are notorious things like AC conducting- noisy switching regulators in electronics and wall wart power supplies, and the infamous light dimmers. Of course, one cannot operate the MoCA pair on the same frequencies used by, say, a satellite TV set top box to dish link, nor the same freqs used by a Cable TV set top box with its own MoCA for whole-house DVRs. A proper MoCA product finds unused frequencies automatically and allows the installer to designate the desired frequencies.

    Attenuation: With MoCA, the attenuation is predictable: number of splitters n the path (beware splitter quality and high freq. ability); length of coax. Once planned, the attenuation won't change simply due to changing what's plugged in to various AC wall outlets as is the case with HomePlug AV.

    Reflections: (Geek warning) The high frequency (OFDM) signals used by both HomePlug AV and MoCA are less subject to time-delayed reflections in the transmission media than non-OFDM. To be sure, there are certain conditions in either that cause a reduction in net yield throughput due to reflections. In HomePlug AV, this would depend on what outlet each is plugged into, and perhaps changing conditions due to what's plugged where in the AC outlets. With MoCA, reflections can occur due to crummy splitters or incorrect coax cable connectors, or coaxes left dangling (un-terminated). But these coax issues are fewer in number and can be viewed.

    Conclusion: The test numbers say that HomePlug AV can provide the 30Mbps or more needed for HD TV 1080i streaming with margins, but it's not a willy-nilly plug and go. A test should be run to find if the throughput is too low and if so, one or both outlet sites need to be changed and re-tried. Then in the future, and this is the hard part, one needs to be mindful of plugging in signal sinks that may cause a speed reduction.

    The MoCA case might be too complicated for the non-techie user, as planning is needed, whereas HomePlug is plug and go, lady-luck permitting.
    Last edited: Aug 4, 2011
  2. thiggins

    thiggins Mr. Easy Staff Member

    May 18, 2008
    Thanks for the report, Steve. Nicely done. Just a few comments / questions:

    - Every powerline adapter comes with instructions that say to not plug it into an outlet strip, but directly into the wall outlet. Even if the outlet strip doesn't have surge protection / filtering (does yours?), the line cord will usually impose some throughput loss.

    - What do you get for MoCA throughput using the same measurement technique?

    - GFI outlets are generally not a problem for powerline, but AFCI breakers can be. See #4 here. It's also common practice to daisy-chain outlets off GFI outlets. GFI's are still much more expensive than plain ol' duplexes.

    - I'm not sure your signal "sink" assertion is valid. The throughput losses I've seen are due to noise generators plugged into the same / nearby outlets. Internal line filters / surge suppressors should have no effect.

    It would seem that even without a 220 appliance plugged in to provide additional coupling, that there is sufficient coupling in the breaker panel to enable powerline to work across phases.
  3. stevech

    stevech Part of the Furniture

    Mar 13, 2010
    San Diego
    I was remiss in not acknowledging Tim for providing the HomePlug AV devices for these tests. Thanks!

    MoCA: I'll measure it. But mine are fairly old D-Link DXN220 models. As I read it, these use a 270Mbps PHY, a bit faster than the 200Mbps PHY of the HomePlug AV devices I tested.

    One section of the manual for these says, on automatic/manual MoCA frequency selection:

    Frequency Planning
    The DXN-220 is capable of operating between frequency bands of 800MHz and 1500MHz. In order to avoid interference
    between the DXN-220 device and other devices on the same cable network, DXN-220 device must operate at a
    frequency spectrum that is different from other devices.
    • 5-42MHz (High-speed Internet Access Upstream)
    • 42-850MHz (High-speed Internet Access downstream)
    • 50-806MHz (Off-Air TV)
    • 50-860MHz (Cable TV)
    • 950-2150MHz (Satellite L-Band)
    One common scenario is where DXN-220, cable television and high-speed Internet service need to coexist. The cable television in United States operates between 50-860MHz while high-speed Internet service upstream
    and downstream operate between 5-42MHz and 42-850 MHz respectively. Hence, the logical choice for DXN-220 frequency would be between 40-1500MHz. Similar frequency planning approach should be taken into
    consideration when dealing with Satellite and off-air antenna.

    Returning to the HomePlug discussion...

    Signal attenuation by power strips/modern TV/Appliances: From an RF engineer's viewpoint ... Inside many power strips, and also inside many newer electronics like HD TVs and the like there are components like "MOV" varistors. These have relatively low inductance at high frequencies, i.e., they shunt. That inductance affects the connecting cable to the wall outlet, and onward in the power wiring. The power wires are an inductor, so at some distance, the attenuation of the varistors and like components is minimized.

    So a HomePlug device's signals in/out would be attenuated when plugged into a power strip, or into an outlet in which there are also TV/appliances, etc., that have their own internal MOVs/MOV-like devices. I observed that in fact, in that speeds are higher when using an outlet with tens of feet of wire separating it from an outlet with many attenuating devices (signal sinks, if you will). Years of experience with X10 shows the same affects to it - though X10 signals are about 150KHz.

    To isolate the negative affects of devices that sink/attenuate, one can use a (plug-in) filter. Effectively, this passes the 60Hz power but blocks incoming signals (from HomePlug frequencies). Conversely, the filter blocks the attenuating affects of the TV/appliance from affecting the outlet into which it is plugged. There are X10 filters used for this purpose, at X10's lower frequency (so an X10 filter won't apply to the HomePlug signal).

    But my test results show certain outlet-pairs that yield fairly low speeds, assumed to be due to weak signals, not interference. I believe that this is from a lack of phase coupling in my no-220V appliances home, and less so, attenuation from items plugged into nearby outlets.

    Signal coupling in the breaker box: Most boxes I've seen have the heavy service drop wires enter the box then go to bus bars on opposite sides of the box, with two rows of breakers. I suspect that all boxes have a mechanical arrangement where every other breaker is on the same phase and a 220V breaker occupies two slots to grab both phases. So absent any 220V breakers/wiring in the home, the coupling is an air gap of inches and this is a very poor coupling. Indeed, the coupling might be greater in the first few feet of the service drop that feeds the box! At the other end of that service drop is probably what amounts to a big attenuator for HomePlug signals, but the inductance due to the length of that line is probably high and offsets.
    Last edited: Aug 3, 2011
  4. thiggins

    thiggins Mr. Easy Staff Member

    May 18, 2008
    That is all true.

    My point is, however, that even with your extreme case of no 220V appliances to provide phase-to-phase coupling, you were able to get powerline networking on all but one of your outlets. And, discounting the results taken with one of the adapters in a power strip, you got very serviceable throughput.
  5. stevech

    stevech Part of the Furniture

    Mar 13, 2010
    San Diego
    Yes, excluding the ones where the Office Power Strip was used, the others would support streaming video. I think 1080i w/H.264 takes a solid 15Mbps net IP yield, and one should almost double that to assure no jitter/lost frames. I stream 1080i from my garage PC to the TV via MoCA. Just one splitter. Works fine.

    I do need to try a HomePlug test using the same wall outlet as the office power strip is plugged into. That would put just 4 ft. of wire from the wall outlet to the power strip - this wire being an inductor to hopefully diminish the attenuation introduced by the power strip and its connected gizmos. I assume that cheap power strips with transient suppressors don't have a series inductor inside- as it would have to bear the full current in-flow and thus be costly and a safety hazard. The series inductor internally would isolate the attenuation from the wall outlet.

    I think, in general, a HomePlug modem should be plugged into an outlet far removed from gizmos and power strips. Over the years with X10, I've found that to be true. My home automation setup, in the garage, to get adequate signal strength all 'round the house, has its main modem on a 20' extension cord to go from the PC area to an outlet without gizmos plugged in. Makes a big difference. Also, I and most other X10 serious users install a phase bridge in breaker box. X10 may be more difficult than HomePlug because X10 is 150KHz or so, less apt to couple between adjacent wires due to the lower frequency.

    I suppose power strips internally differ a lot as to how they use MOVs or other transient suppressors. Also, I've read about consumer UPSes - both attenuate and tend to generate a lot of noise.
    Last edited: Aug 3, 2011

Share This Page