We came away from our testing and discussions with the strong belief that HomePlug could play a major role in consumer networking. Our tests confirmed that HomePlug would be a good solution for connecting multiple PCs to a cable or DSL modem. Indeed, we think HomePlug is a good fit for additional applications, including telephony and audio distribution (see Problem Solving for use in audio networking).
While HomePlug falls short for video distribution, that's still several years away as a mass-market requirement, and vendors say that the technology has the potential for higher speeds.
We ran into several problems during our testing and interacted with many vendors about them. The vendor responses suggested that most should be resolved in the next generation of chips that are already being integrated into products for later this year. The most serious problem - assuring seamless interoperability between "HomePlug"-branded products - is being studied by the HomePlug organization and we hope it will be resolved soon.
Network Speed and Applications
HomePlug worked well in our house.
While our measured speed averaged about 3.8 Mbps -- far less than the claimed "14 Mbps" -- it would work well for sharing a broadband modem between several PCs. While slower than Ethernet over structured wiring, it worked in some places in our home where our current Wi-Fi equipment did not.
HomePlug somewhat reduced the speed of Internet downloads through our cable modem. The speed through HomePlug was slower than through Ethernet; we consistently measured somewhat slower Internet download speeds than the internal file transfer speed which we used for comparison.
Networks using only ST&T adapters performed 21% faster than the others. This was a surprise since several of the ST&T adapters were based on the same chips and appeared nearly identical to other units we tested. It seemed to indicate that manufacturers will need to pay considerable attention to the details to get the maximum performance from HomePlug.
What Makes Outlets "Good" or "Bad"?
We spent a lot of effort trying to understand why some outlets worked much better than others. We read the product manuals, discussed this with many vendors, and tried to decipher our test results for clues. (We wanted to read the HomePlug specifications, but they are considered proprietary and were not available to us. In the absence of understanding just how the technology works, it's pretty hard to deduce why some outlets were particularly bad.)
It's especially important to find a "good" outlet for the "master" adapter. The most obvious outlet -- where you've already plugged in a broadband modem, the primary PC and its peripherals, and perhaps a cable/DSL router -- is probably one of the worst outlets for HomePlug. Manufacturers provide little help in identifying a good outlet for the master.
Circuits seemed to be most impaired by the type of equipment hard-wired to the circuit or plugged in to outlets. "Brick" transformers of the type that power many small electronic devices appeared to be among the worst. Several of these are typically plugged in with the primary PC - that's why that outlet and probably nearby outlets on the same circuit are so bad for the master adapter.
Manufacturers recommend not plugging HomePlug adapters into surge protectors. We found that it's very important to follow this recommendation -- the better the surge protector, the worse the HomePlug performance. And a UPS completely blocks HomePlug.
We saw some impairment from vacuum cleaners and hair dryers plugged into the same outlet, especially if the outlet was already impaired (as indicated by a low data transfer rate). Just plugging in one vacuum cleaner reduced the HomePlug data transfer rate substantially, and turning it on effectively stopped data transfer.
Outlet distance from the main electrical panel had much less impact on measured speed than we expected; the attached equipment and the wire gauge and routing through the walls had a much larger impact. Some of the longest circuits - including one that passed through a secondary electrical panel at least 75 feet from the main panel - performed much better than short circuits with the "wrong" type of equipment.
HomePlug crossed AC phases very well. We saw little difference in performance between circuits on different phases.
We saw little effect from fluorescent lights plugged into the same outlet.
We encountered several problems when running these tests. Many were inevitable when testing first-generation products. We expect most will be resolved before these products enter mass distribution.
We found installing USB adapters to be more difficult than Ethernet adapters. We were on the telephone several times with vendors to figure out how to get USB adapters working properly in our network. Even when a PC seemed to be networking properly through a USB adapter, we encountered problems reading files from it. Our installation and operating environment may be atypical: (1) Our test portable did not have a built-in CD-ROM drive, so we needed to use a network drive for software installation. (2) Our network is more complex than that in a typical home -- we operate with a Windows NT domain; have an existing network with lots of PCs; and many PCs have multiple network adapters. (3) Our test portable is still using Windows 98SE. We plan to test again with Windows XP to see if installation is easier.
We observed a very odd interaction with Internet Connection Sharing under Windows 98SE. When ICS was enabled to share an Ethernet-based HomePlug connection by wireless, it appeared to cut the HomePlug file transfer speed in half.
As mentioned earlier (see Mixed-Vendor Networks), we encountered two interoperability problems in multi-vendor networks. One problem involved networks with more than two Ethernet Bridge adapters; this should be resolved in products based on the next generation of Intellon chips. The other involved different vendor approaches to setting the encryption keys for "advanced" level of network security. These are indicative of a more general issue: the lack of comprehensive interoperability testing between different implementations of HomePlug. We believe that the HomePlug group is well aware of the need to resolve these issues to fulfill its promise of product interoperability.