AV Phoneline Performance
Performance of the AV Phoneline product was more problematic. To make a long story short, it appears that CopperGate still has some tuning to do to accommodate the curve balls thrown at their technology by the wide range of devices that can sit on a pair of phone wires. The TCP throughput plots in Figures 7 and 8 illustrate what I mean.
Both plots show throughput dropping to essentially zero around the 30 second point, continuing for a bit, then completely stopping, when I connected my Westell 327W DSL router to the same phoneline that was connecting the pair of AV Phoneline bridges. Apparently, there's something about an active G.DMT DSL connection or - according to Corinex - the phone line filtering in the Westell router (or lack thereof) that pretty much kills the AV Phoneline data flow. But things started right up again when I unplugged the modem at around the 1 minute mark.
But DSL isn't the only thing that can put a crimp in the AV Phoneline's data flow. The throughput drop around 1:30 in Figure 7 and 1:20 in Figure 8 happened when I took the speakerphone in my Panasonic KX-T3910 cordless phone off-hook (the phone was connected to the filtered "phone" jack on one of the AV Phonelines). In this case, at least data kept flowing, albeit at a significantly reduced rate, until I put the speakerphone back on hook.
The good news, however, is that as long as I didn't use the phone, or my DSL connection, I could get average throughput around 40Mbps! What's interesting is that this is twice the maximum throughput that I measured with the AV CableLAN, which had a nice, quiet, shielded and controlled impedance environment to work in.
Ok, so what's a little throughput drop among friends? Unfortunately, throughput isn't the only thing that's dropped if you take your phone off-hook. Figures 9 and 10 show the results of the same UDP streaming test that I ran on the AV CableLAN. Figure 9 shows a slight throughput disturbance caused by taking the speakerphone off-hook between the 20 and 35 second marks, which doesn't seem to be much of a biggie.
But Figure 10 shows that while throughput stays up, there is about 3% data loss on the receive side (you can't tell from the plot - but I verified this with the original file) and about 1% loss for transmit. So you might as well put that video you were streaming on hold while you yak on the phone, because it wouldn't be pretty to watch.
The Corinex technical contact I spoke with generally attributed these problems to "noise", either from the phones themselves, or in the case of DSL equipment, inferior microfilters on phones or inferior (or missing) anti-alias filters on the DSL modem's phone connection.
He also said that CopperGate is working on hardening their technology against these problems, but said that it would probably be around the end of this year (2005) before a solution robust enough to handle most home networking situations was available.