Our home server machine, a Koolu, is located in the office, while the media center box (MythTV) is in the living room near the television. Wireless networking was unsatisfactory, at least with the NetGear WG111v2 USB NIC on Windows XP and Vista, so we removed that and ran a wire around the edge of the rooms. But this is tacky. So we considered power line networking.
Characteristics and features of power line networking:
You connect a wired Ethernet NIC on your computer, network switch, or whatever networked device, to a RJ45 port on the channel station, which plugs into the 120 volt power line. Elsewhere in the house you have some other equipment similarly connected. Packets sent from one station pop out of all the others. The attached computer (etc.) accepts it or not, according to the packet's MAC address, same as for wired Ethernet.
There may be several channel stations. Probably there is no absolute limit on their quantity, though there are practical limits since they all share the same bandwidth.
There are reports of communication between nearby apartments or even houses, but this is not an advertised feature.
The protocol includes encryption using AES-128, and there is a procedure to set all your channel stations to the same randomly generated key.
The devices are not exactly power misers; mine is noticeably warm to the touch. Typical ratings are 4 watts each.
Where the line voltage is 120 volts, such as the USA and Canada, the outlets are in two groups (phases). Channel stations on the same phase line communicate much better than across phases.
Among the reviews I read, the most useful was this one:
Review 5 Power
Line Devices That Take You Online Where Ethernet or Wi-Fi Can't
in Computerworld by Bill O'Brien, 2009-02-18. Here's a summary:
He reviews devices from Belkin, D-Link, Linksys, Netgear and Zyxel, plus comparisons with Wi-Fi (802.11n) and wired gigabit Ethernet. The data speed varies with the vendor, the distance between channel stations, and the number of streams sharing the connection. His main test was to copy between Windows boxes a folder containing 4000 files, 8.05Gbytes total. I'm reporting million bits per second here (divide by 8 for Mbytes per second). Copying under the most favorable conditions ran at 22 to 35 Mbit/sec depending on vendor, in this order: Zyxel (35), Linksys (34), Netgear (33), Belkin (32), D-Link (22). For comparison, the same test on 802.11n did 72 Mbit/sec and on wired Gbit Ethernet it was 164 Mbit/sec. With five similar competing streams the rates declined only to 28 to 17 Mbit/sec. It wasn't quite clear what the test was here: I think there were multiple copy operations in progress at the same time between the same pair of machines, since he had only two channel stations from each vendor. The reported speed would be for all streams combined.
He describes each of the products in detail; I won't reproduce that here. There are several interesting comments appended to the article. In particular, surge suppressors severely attenuate the signal (Belkin has one that is specially designed not to), and my experience is that even non-suppressing plug strips attenuate the signal significantly, so plan on connecting directly to the wall outlet.
Products from different vendors are supposed to interoperate, but this was not tested explicitly in the review. Some comment posters say they succeeded in interoperation, while others failed.
Based on vendor reputation (as experienced by jimc), price, availability
from preferred resellers, and physical configuration, I picked the
Belkin 200 Mbps Powerline 3-Port Adapter, US $150 from Amazon.com.
This kit includes one three-port channel station and one one-port one.
A pair of one-port stations is available for $20 less.
The box includes two channel stations, two one-meter Ethernet cables, two power cords about 1.5 meters long, two wall plug adapters, and a CD with the users manual as a PDF file. You can attach the plug adapter for plugging the station directly into a wall outlet, or you can attach the cord if the station needs to be distant from the outlet. The plugs are USA standard NEMA 1-15P, that is, two parallel blades and no grounding pin. Presumably if this item is sold outside NEMA territory the provided plugs will be suited to the locale.
For initially checking if the devices work, the ideal is to plug both of them into the same paired outlet, or at least on the same wall. Later, move them to the production location.
Make sure to use an unswitched power outlet (or turn on the switch).
Make sure to plug the Ethernet cable into a working port. Avoid the conveniently unoccupied uplink port of a network switch.
It's best to check out the devices, then randomize the crypto key, so there's no question that the crypto setup procedure might have been messed up.
Residential wiring in the USA and Canada is generally split phase. In other words, you get balanced 240 volts with a ground wire. Big appliances like ovens and air conditioners connect across both phase lines, but small equipment like lamps, computers and network channel stations connect between ground and just one of the phases, getting 120 volts. But an approximately equal number of outlets are connected to each of the two phases, generally being split up by rooms or by walls. This means that if the two channel stations are on the same phase they will get a strong signal, but if they are on opposite phases the signal has to cross over in the air conditioner wiring or the house feeder cable, and so the signal is reduced. A lower signal means slower data transmission. Cross phasing is the biggest reason that users (in Amazon.com rating posts) complain that their power line networking is unsatisfactory.
Residential wiring in the UK, much of Europe, and many (but not all) other countries provides balanced 240 volts to everything, so there is only one phase and all locations will get a strong signal.
Here is the result of my speed tests. The task was to transfer a large compressed file, 129Mb, using HTTP on Jacinth, the home server, to Xena, the laptop, in various rooms. Calibration tests were done to demonstrate that the bottleneck was the network, not the endpoints. Speeds are reported in million bytes per second; multiply by 8 to get bits per second.
|Internal||--||207||Xena to Xena|
|Internal||--||32.4||Jacinth to Jacinth|
|Office||A||6.14||Other side of room|
|In the following, the link light was red|
|Studio||B||0.22||On plug strip|
First, the devices degrade gracefully: they can get data through, even if slowly, in unfavorable conditions. But cross-phase operation is very noticeably slower than communication on the same phase. The advertised data rate is 200 Mbit/sec (or 25 Mbyte/sec); the actual speed under the most favorable possible condition is 1/3 of that. The link light is supposed to turn red when a poor signal makes the station reduce its speed below 80 Mbit/sec (10 Mbyte/sec); again, the actual data rate for this indication is about 1/3 what is advertised.
So I conclude that my goal for powerline networking is met: to replace the wired link on the floor that connects the media center box to the rest of the house net.
An additional idea is to make music available in the bedrooms by power line networking. This is definitely feasible, since compressed audio requires a relatively low bandwidth. Another Koolu would make a nice low power playback node.
Another goal is to put a media center box in the studio, displaying content from the backend near the television. The data rate of 440 Kbyte/sec or 3.5 Mbit/sec is marginal for this service and it would have to be tested carefully.