After my PowerBook G4's CPU destroyed itself, I replaced it with a much cheaper, but less portable, Mac mini in my living room. When I made this change my wireless network became just a way to bridge the computers in my home office with the computers in the living room.
Although 802.11g wireless networking works adequately for this purpose, I always have my eye out for something faster. My iTunes library is stored on a machine in my computer room and is almost 100 GB in size. Synchronizing my video iPod from the Mac mini in my living room could do with a bit of a speed boost.
Upgrading to 802.11n, with its significantly faster performance, is not really an option. Apple doesn't offer an upgrade for older Macs that lack the 802.11n hardware. USB adapters may be available for other companies' 802.11n implementations, but I'd need to purchase an adapter for each device I wanted to add to the network.
While wandering around the computer store recently (a habit I really should try to stop) I found that HomePlug adapters had recently received a boost in throughput. For many years the HomePlug Powerline Alliance has promulgated a standard for Ethernet over electrical wires called HomePlug 1.0. Unfortunately this standard maxes out at around 14 Mbps, which is hardly fast enough for home users accustomed to (theoretical) 100 Mbps wired connections and 54 Mbps wireless connections.
Two competing standards have arisen to provide the next generation in Ethernet over home electrical systems. The first is the produced by the HomePlug Powerline Alliance. The second, called , is from the Universal Powerline Association.
The two standards offer similar functionality. Both claim a theoretical maximum throughput of 200 Mbps. Both offer improved encryption and quality-of-service (QoS) options over the older specification. The specifics of the standards differ enough that you can't use a Powerline HD adapter to connect to a HomePlug AV adapter. Unfortunately, like Betamax versus VHS and Blu-ray versus HD-DVD, having multiple competing standards will hurt adoption of the technology by consumers.
At the store, there were no HomePlug AV adapters in stock, but Powerline HD adapters from both and  were available. I chose the D-Link Powerline HD adapters based on price ($180 for the two-adapter starter kit versus $200 for the Netgear starter kit).
Installation -- Installation was extremely simple; all I had to do was plug the Powerline adapters straight into a wall jack (it's important to plug directly into a wall outlet, and not to use a surge protector or UPS). Neither HomePlug nor Powerline signals will cross a transformer, which is seldom an issue in residential construction.
You may wish to use to verify your plugs are wired correctly. In my testing, Powerline adapters worked plugged in either way (the adapters aren't polarized), which means they'll work with incorrectly wired circuits, but circuits with reversed wiring are a safety hazard, so it wouldn't hurt to test anyway. A weird effect appeared in my testing, however. I saw the fastest file transfers with one adapter plugged in normally and the other plugged in reversed.
The D-Link Powerline adapter has three lights: one shows that it is receiving power, another indicates it has detected another Powerline adapter, and a final light indicates when a live Ethernet connection has been established. The adapters will work in the default configuration, but on the default D-Link network and without encryption. If you live in an apartment building, or if your house shares a transformer with other houses, your signal could be available at their power outlets. (And I don't know about you, but I don't want my neighbor's blender to be able to snoop on my email!)
Configuration -- Setting up your own network name and turning on encryption requires that you configure the adapters, which, unfortunately, requires a Windows computer. (Every HomePlug AV and Powerline HD adapter I've found has this same requirement.) Configuration certainly should work from Windows running under Boot Camp and will probably also work from Windows running under Parallels Desktop or VMware Fusion. However, I've not tested any of these options myself.
To get started with configuration, you plug the adapters into the wall jacks (preferably in their final locations, but that's not necessary). Both adapters should indicate they see another Powerline adapter.
Next, using an Ethernet cable, plug your Windows computer directly into one of the adapters and launch the configuration software that comes with the adapters. It automatically detects both adapters, showing one as ETH and the other as PLC. The ETH (short for Ethernet?) adapter is the one directly connected to the computer and the PLC ("Powerline Connected" perhaps?) adapter is the remote one.
First configure the remote (PLC) node. If you first set up the node you are plugged into (ETH), the adapters will no longer be able to see each other and the configuration software will not be able to talk to the remote adapter.
Basic configuration consists of assigning a Net-ID (similar to an SSID on a wireless network) and encryption key (just like the encryption key on a wireless network). Plus, you can give each adapter a more identifiable name - I named mine for their physical locations - and you can set a configuration password to prevent other users from changing your configuration. Once you've done the PLC adapter, repeat the process with the ETH adapter.
Advanced configuration lets you set two rules for quality-of-service, which enables you to prioritize your traffic. For example, you can give Skype packets a higher priority than other packets. This would help keep the voice quality of your Skype communications high, while an FTP file transfer going on at the same time would be slowed down. Gamers may wish to set certain game traffic higher than regular traffic to keep their games from lagging behind other players. Since I have no need for QoS on my setup, however, I stuck with the basic configuration.
Real World File Transfer Tests -- To get a feel for how fast the adapters operate in the real world, I ran several tests, first over the wireless network, next over the Powerline connection, and then comparing to wired Ethernet. To keep the configurations as similar as possible, both networks used encryption (WPA for the wireless) and had no QoS rules in place. I repeated the tests over a couple of days to verify the reliability of the numbers.
Initially, I transferred a 1.4 GB file via FTP from my Linux server using pure-ftpd to the Mac running Interarchy. According to Interarchy's transcript, the wireless transfer of 1,472,156.9 kilobytes took 1,797 seconds (just under 30 minutes) or 6.4 megabits per second (Mbps). (Remember that 802.11g has a theoretical bandwidth maximum of 54 Mbps.)
The Powerline adapters demonstrated quite a bit of variation in transfer rates. The worst transfer took 1,210 seconds (a hair over 20 minutes) or 9.52 Mbps. The best clocked in at 948 seconds (almost 16 minutes), or 12.16 Mbps. Most transfers ran at around 10 Mbps, putting the Powerline adapters at 50 to 100 percent faster than my 802.11g wireless network, if nowhere near the 200 Mbps theoretical maximum. I can't really account for variation in the rates, although noise on the power lines, such as from an air conditioner, compact fluorescent lights, or washing machines, may be involved.
Next, I decided to compare the results to a direct Ethernet connection. I don't have a long enough network cable to test this in my current setup, so I moved the Mac mini to my computer room and connected it via 100 Mbps Ethernet. With this setup, transferring the 1.4 GB file via Interarchy showed a significantly faster result of nearly 87 Mbps, or almost 9 times better than the Powerline adapters. Clearly, if performance is key and you can run real Ethernet cable, it's the preferred way to bridge two locations.
Network and Ideal Situation Performance Tests -- In addition to testing with Interarchy, I used the network performance tool  to measure pure network throughput. Testing with file transfers involves the operating system, file system, and hard drives. It's a good benchmark for expected real world operations but doesn't isolate just the network component. Iperf measures just the network.
Since I had already moved the Mac mini to my computer room for the Ethernet test, I also decided to do some additional tests of wireless and Powerline connections in an ideal setup. My wireless router is in the computer room, so I tested a wireless connection with the Mac mini just a few feet from the router. Iperf recorded a throughput of 12.8 Mbps, but the Interarchy test dropped back down to 7.0 Mbps, just a touch faster than when the Mac mini was in the living room.
For the ideal Powerline test, I plugged the two adapters into the same electrical outlet. It doesn't get more ideal than that for Powerline adapters - no circuit breakers or fuses to cross, and less noise on the line from other devices. In this ideal configuration, Iperf reported a throughput of 69 Mbps, and Interarchy reported 65 Mbps, nearly 7 times faster than my real world configuration. This result suggests that I should try to find two outlets that share the same circuit breaker. Unfortunately, the builder of my house neglected to label my circuit panel, so until I can get a wire tracer I'll have to just move the adapters around and see if I get better results with different outlets.
For comparison, with the 100 Mbps wired Ethernet connection, Iperf reported a throughput of 91.1 Mbps, which is a huge increase over both the wireless and Powerline connections. Ethernet really is the way to go for maximum performance, assuming it's feasible to run cable to the necessary locations.
Latency Tests -- I made one other test of the adapters: latency. Network latency is a measure of how long it takes packets to travel through a network. Although the file transfer measurements above are typically referred to as "speeds," they're really a measurement of bandwidth. Think of a pipe with water coming out at a rate of one liter per minute. To get water faster - say, two liters per minute - I have two options: I could push the water molecules through the pipe faster, or I could make the pipe larger. Although individual water molecules in the larger pipe still take the same amount of time to travel, more of them are traveling in the same amount of time, making the rate appear faster. Networks can do the same thing. Bits can travel through the system faster, or more of them can be transferred at the same time.
To measure the round-trip latency of each connection, I used the ping command, which sends an "Are you there?" packet to a device. The device, if present and active, responds with a "Yes." The sending computer can time how long it takes to receive the answering packet, which is the round-trip latency of the network.
I used Mac OS X's built-in ping command in Terminal to ping my Linux server 100 times. Over the 802.11g wireless network the ping command reported a minimum transit time of 1.5 ms (milliseconds), an average of 1.8 ms, and a maximum time of 4.8 ms.
Like the transfer rates, the latency when using the Powerline adapters changed over several days. The worst was a minimum time of 1.5 ms, an average of 9.1 ms, and a maximum of 121.7 ms. Most of the tests came out to a minimum of 1.5 ms, with an average of 3.5 ms and a max of 27 ms. Although they push a lot of packets through, the D-Link Powerline adapters seem to have a very high (and variable) latency. This means that, even with the QoS features, the Powerline adapters may be less suitable than other network types for VoIP (Voice over IP) applications such as Skype and iChat.
Summary -- On price it doesn't appear the D-Link adapters can be beat. Both other Powerline HD and HomePlug AV adapters appear to be more expensive than the D-Link adapters, and they're also cheaper than outfitting multiple Macs with new wireless gear, and are likely cheaper (or at least easier) than running Ethernet cable inside walls. The downside is that because of the competing Powerline HD and HomePlug AV standards, I may be locked into purchasing D-Link adapters to expand my network.
In terms of performance, the D-Link adapters are good, but not stunning. They seem to have a high network latency, but since my network needs to bridge only two locations, and since my use is mainly downloading files, either through Web browsing or moving TiVo movie or iPod music files around, the 50 to 100 percent increase in throughput over 802.11g is certainly welcome.
PayBITS: If Kevin's comparative testing was helpful to you, consider thanking him with a few bucks via PayPal! <https://firstname.lastname@example.org> Read more about PayBITS: <http://www.tidbits.com/paybits/>