Earlier this week, we took delivery of a pair of NewLink HomePlugAV (200Mbps) powerline Ethernet adapters, to replace the older HomePlug “Turbo” (85Mbps) pair which had faithfully sent the Internet upstairs to our computer room since 2006. They were still working as well as ever, but were beginning to struggle with the demands we were putting on them, so we decided to buy a pair of HomePlugAV adapters to boost the speed we get on the Mac upstairs.

OK, that was the theory.

When I plugged everything in and tried playing back a 720p HD video from our home server (an acid test if ever there was one), it “sort of” worked, in that it played at all—the old 85Mbps adapters would’ve choked completely on a file that size. However, there were still quite a few pauses, and I found myself thinking: isn’t HomePlugAV designed for handling traffic like HD video? True, the link was faster than the old setup, but not by much at all. What was going on?

Worse was to come. I’d kept the old 85Mbps adapters for use with another PC in a different room (sort of an “overflow” machine, in case the Mac was in use). Later in the evening, I found that on the Mac, the HomePlugAV link had slowed to a crawl—worse than even the 85s managed. Either I’d made a bad purchasing decision, or something screwy was happening between the HomePlugs.

I needed to test how fast the data link was, so I set up my Linux netbook next to the router (via an Ethernet connection), and prepared to use iperf—a cross-platform network speed-testing tool—with the netbook acting as the “server”, and the Mac the “client”. (I actually had to compile iperf on the Mac from source—hardcore! 🙂 )

Bearing in mind that HomePlugAV adapters tout a “200Mbps” figure (although every review will tell you that with interference in electrical wiring, etc., 150Mbps is probably the most you’ll get), the initial results were very disappointing.

The first iperf tests revealed that I was getting no more than an average of about 20Mbps—divide that by 8 to get the megabytes, and that means a file would transfer over the link at barely more than one megabyte per second. That’s more or less barely more than the speed of our ADSL broadband connection—certainly not sufficient for SD video (let alone HD) on our local network, and really quite a let-down. (The results were even worse, when I initiated some activity over the 85Mbps HomePlugs—it basically slowed the HomePlugAV link to almost nothing.)

To cut a long story short, I did some digging, and learned that according to a SmallNetBuilder article on boosting HomePlug speed, it is a bad idea to mix 85Mbps and 200Mbps HomePlugs on the same electrical system, as the 85s appear to cut the throughput of the 200s by as much as half. Further iperf tests confirmed that the 85s do appear to affect the 200s on our network (though not by as much as the article suggests), so I plan to replace the 85s with 200s as soon as my budget allows.

But that still begged the question: even without the 85s active, why was the throughput on the 200s so poor (if you recall, low- to mid-20s Mbps)? I then moved onto the next advice from the SmallNetBuilder article: find out what devices are injecting “noise” into your powerlines, and either filter, move or remove them. (In short: certain kinds of electrical equipment create interference in your mains wiring, which will lower the speed of data sent through a HomePlug connection.)

Nearly an hour of unplugging, swapping and trial/error in our computer room later (and there were plenty of sockets to try up there), iperf informed me I had only managed to nudge the top speed up to the high-20s Mbps. It was time to move the focus downstairs, to the other end of the HomePlugAV link: in the lounge, by the ADSL router.

To set the scene: the HomePlugAV adapter was plugged into one of the two wall sockets, and the other contained a Belkin surge protector, powering a four-way “gang socket” with four devices connected (almost all “mains adapters” or “wall-warts”). I had an inking that one or more of these factors was spraying interference into the system, but I couldn’t really turn any of them off.

Then it hit me: why not try moving the HomePlug to another socket, about 3-4 metres further along the wall, and run an Ethernet cable back to the router/switch? In fact, I’d already been doing that with the 85Mbps HomePlugs, so I swapped the locations of the 85 and 200 adapters, and fired up iperf once again.

My jaw almost dropped. In one fell swoop, I had just about doubled the throughput on the HomePlugAV link—iperf was now reporting speeds of around 45-50Mbps, just by moving the adapter away from a group of mains adapters on a surge protector. (Before you ask: no, I hadn’t put the HomePlug on the surge-protector—that’s one thing you never do with HomePlugs if you can help it, as the surge-protector ruins the network connection.)

So, I now have a reasonably-speedy HomePlugAV link running in our house—45-50Mbps is some way short of the 150Mbps “ideal”, but it’s just about fast enough for 720p HD video, so I can live with it for the moment. I plan to get rid of the 85Mbps HomePlugs as soon as I can scrape the cash together for another pair of 200s, and then I’ll probably look and see if I can squeeze any more Mbps out of our home wiring (hopefully without needing an electrician!).

If all goes well, it might even achieve an acceptable WAF (Wife Acceptance Factor)…