The main issue with powerline networking is that it is heavily dependent on the usage environment. The other issue with powerline networking products is that most of them are based on the same chipset, and offer similar performance. It is not the case for a consumer that one powerline networking product performs better than the other. Ideally, we would have liked to benchmark the WD Livewire against a product based on a chipset from Gigle Networks. Unfortunately, we are not in possession of a Belkin Gigabit unit. On the other hand, we did manage to lay our hands on a couple of other Atheros/Intellon based powerline adapters. As expected, they gave similar sort of benchmark figures as the WD Livewire.

We decided to do nothing fancy in our benchmarking methodology for the WD Livewire. The kits were connected to the two power outlets to be tested and two notebooks with GbE ports were connected to them to set up a P2P network. Both machines were running Windows 7 and the hard disks had more than enough performance to surpass the 100 Mbps rate at which the interfaces were capped by the powerline kit. It is quite well known that nothing can be as inefficient as the Windows SMB network sharing protocol. We set up 5 files of varying sizes (5 MB, 700 MB, 1.4 GB, 4.37 GB and 16 GB), and transferred them from one computer to the other four times. The average of the time taken to transfer each file was recorded. The experiments were repeated across one outlet in each room of the house, and transfer rates corresponding to the worst performance are reproduced in the table below. The experiments were conducted in two houses, the first of which (Location 1 in the table) was a 23 year old single bedroom apartment of approximately 700 sq. ft. The other house (Location 2 in the table) was a 10 year old five bedroom single family dwelling spread across two floors, with an approximate area of 2800 sq. ft.

SMB Transfer Rates with the WD Livewire (Mbps) - Higher is Better
File Size Worst Case @ Location 1 Worst Case @ Location 2
5 MB 23.98 28.19
700 MB 32.57 38.62
1.4 GB 31.99 38.07
4.37 GB 31.79 37.95
16 GB 31.86 38.12

Surprisingly, better results were obtained in the bigger house, but this can probably be explained away by the fact that there is probably too much noise in the power lines in an apartment setting. Also, note that the figures quoted above are probably the worst case bandwidth measurements. That said, it is probably what an user would encounter while trying to use the powerline network for file transfers within the home network. As the figures indicate, it is not quite possible to stream Blu-Ray bitrate movies from a PC or NAS to a media streamer with this generation of powerline networking products. Compressed 720p copies work reliably (and this was confirmed to be without issues in both our test locations). Netflix and YouTube streaming, which used to stutter on a Wireless-N network in Location 2, was managed without breaking a sweat, once the WD Livewire units were set up. In this respect, the WD Livewire achieves what it was intended/advertised to do by Western Digital.

Readers contemplating powerline networking must remember that the usage scenario heavily influences perceived performance. There are a number of good resources on the Internet about how to make the best of your powerline network. SmallNetBuilder has one such excellent piece. Also, it must be remembered that capacitive loads close to the routing unit result in reduced performance, while inductive loads improve performance. Also, the topology of the electrical wiring in the house heavily influences effective bandwidth. In places such as the UK, where the electrical wiring has a ring topology, the wires almost act equivalent to running an Ethernet cable around the house. In places such as the US, a ring / mesh topology is often seen. The performance does go down in such cases. Physical separation between the locations of the units often doesn't matter. It is often characteristics such as electrical distance, and whether the units are on opposite sides of a circuit breaker which matter more. All in all, it is best to try out a powerline adapter kit and see whether it fits the intended requirements in one's own place of usage, rather than relying on the advertising.

Livewire Internals Analyzing Western Digital's Strategy
Comments Locked

31 Comments

View All Comments

  • jkostans - Tuesday, August 24, 2010 - link

    Ethernet is not twisted to prevent radiation, it's to reduce noise in a differential transmission line. (The fact that ethernet signaling is differential reduces radiation and loss but not the twists) In theory by keeping the wires as close as possible any noise that influences one wire will equally influence the other wire in the same manner. When the subtraction is done to find the voltage, the net result of the noise will have no effect. It's like the equation (5+X) - (0+X) = 5. No matter what value of the noise (X) is, the equation will always equal 5.
  • chromatix - Tuesday, August 24, 2010 - link

    The twists have both effects, and effectively isolate the electromagnetic fields of the wires and their surroundings up to a certain frequency and down to a certain distance. It is the isolation which eliminates both emissions and received noise.
  • flgt - Tuesday, August 24, 2010 - link

    I think to some extent you're both right. The reduction in emissions comes from the fact that the signal is driven differentially (equal amplitude, opposite polarity) and the control of the electric fields which are generated. In a non-perfect transmission line there is a finite amount of cancellation between the two fields which can be improved with cable construction (Cat 6 > Cat 5 >> Romex). Twisting improves performance but a shield is required optimal emission performance. However the increased costs and installation time of shielded cables make them a less desirable solution as long as regulatory requirements can be met.

    As stated, differential signaling provides inherent common-mode noise immunity which is aided by the tight coupling of the twisted wires. The twisting helps ensure the radiated noise excites each conductor in the pair equally (again to some finite amount).

    I don’t see this being a major issue for radio operators for a number of reasons:
    1) The noise immunity provided by differential signaling allows the line to be driven at lower voltage levels, thereby further reducing emissions.
    2) The use of OFDM allows each individual frequency channel to be operated at very low signal-to-noise ratios due to the low-order modulation scheme being used. This is the only way they can operate over such a horrible channel in the first place (your power line). They most likely operate each carrier at a very low signal level to make the composite spectrum look more like broadband noise. DSL essentially operates in the same way.
    3) The signal will quickly attenuate since the channel is so bad.
  • Per Hansson - Wednesday, August 25, 2010 - link

    I can confirm your suspicions
    I recently bought the Belkin Gigabit Powerline HD networking kit

    It managed to do 5Mbps over my powerlines, an apartement built 1990
    When I turned on my FM radio there was an amazing ammount of noise in the reception when I was transmitting data, less so when I was not transmitting data but still some pops and cracks in the reception
    When I unplugged the adapter the reception became perfect

    Next test was to run the powerline adapter from the apartement out in to my garage, the speed now dropped to 1Mbps
    To my suprise though it managed to even interfere with the FM reception in my car!
    The signal got way harder to recieve, it did not crack and pop in the audio tho, but I think that may be simply due to the fact that the stereo in my car is much better at receiving a signal, and the fact that the car itself acts like a faraday cage...

    To say the least I returned this "Gigabit" junk
  • yottabit - Tuesday, August 24, 2010 - link

    There's a typo in the second sentence of the summary article on the anandtech home page... :(
  • reckert - Tuesday, August 24, 2010 - link

    In most homes, the electric coming in is 240v, and the circuits are split to 2 sets of 120v. I know from past experience power line protocols such as x10 couln't get their signals to jump between the two sides of the wiring without a special bridge.
  • ganeshts - Tuesday, August 24, 2010 - link

    reckert, as you mention, x10 needed something to plug in to bridge the legs of the circuits. However, as all Homeplug stuff is on high frequency, they bridge due to the capacitive effect of the bus bars in the circuit breakers. Therefore, you will not have a problem with the configuration you mention as long as you are using HomePlug certified devices.
  • Souka - Tuesday, August 24, 2010 - link

    My house build 1984...2 story, 2200 sq/ft

    Cable modem on second floor, tv on 1st with BriteView media player.

    Netgear XE-104 (up to 85Mbs claim)... on same curcuit directly plugged into wall I got barely 15Mbps. Different circuit in adjoining rooms 5Mbps. Different circuit on different floors...under 1Mbps and often dropped connection.

    BELKIN F5D4073 (up to 85Mbps) Powerline Turbo similar results...

    Also tried the kits in my parents 1960's 2 story 3500 sq'ft home....worse results...

    I really like the idea of powerline tech, just doesn't deliver for me
  • ganeshts - Tuesday, August 24, 2010 - link

    Souka,

    Please take a look at the SmallNetBuilder guide here, as it may be of some help to you: http://www.smallnetbuilder.com/lanwan/lanwan-basic...

    It is best to buy these type of products from stores which have a good return policy / no restocking fee. Unless the consumer actually tries it out, it is not possible to know in advance as to how well the technology would work, as is evident in your case.
  • Souka - Tuesday, August 24, 2010 - link

    I might give the XAVB5501 or XAVB5001 a shot when they come out....none of the listed retailers (or froogle.com) came up with any units available.

    I do wish they had dual ports... or at least in kits one of the units had dual ethernet ports.

    Yes I can easily hook a switch to it, but rather not.

Log in

Don't have an account? Sign up now