Performance - Noise Immunity
I ran noise immunity (or susceptibility) tests with both adapters plugged into the outlet strip. I used two noise sources: a compact travel hairdryer (broadband noise) ; and a halogen floor lamp with dimmer (impulse noise). For a worst-case test, the noise source was plugged into the same power strip.
In Figure 10, the points where throughput spikes downward are where the dimmer was turned on to a low to moderate setting. Throughput reduction is immediate, with recovery taking 10 - 20 seconds as the powerline PHY algorithms find better carriers to use (recall Figure 1).
Figure 10: Line noise effect - light dimmer
Motor noise from the hairdryer had less of an effect, as shown in Figure 11. You can again see drops where I turn the hairdryer on and off. But the drops aren't as large and recovery takes only a few seconds.
Figure 11: Line noise effect - hair dryer
I've found that noise effects, especially impulse-type, drop off very quickly with distance. When I moved the noise source to the wall outlet where the unfiltered power strip that held the two adapters was plugged in and repeated the tests, I could see no effect from either the hair dryer or light dimmer. All it took was six feet of line cord! So before you go spending money on a powerline filter, try just plugging your noise source into an extension cord.
The HomePlug AV spec provides for both coexistence and interoperability with HomePlug 1.0 (original 14 Mbps products) and HomePlug 1.0 plus Turbo (85 Mbps devices). Coexistence is mandatory, but interoperability is not. So HomePlug AV—and Atheros' HomePlug AV compatible "500 Mbps" technology—won't talk to earlier generation devices, but they won't interfere with them either (and vice versa).
I tested both coexistence and interoperability with HomePlug AV and HomePlug Turbo products that I had on hand. The HomePlug AV devices were the TRENDnet TPL-303Es tested in the HomePlug AV round up and the Turbo devices were ZyXEL PL-100's I used for interoperability testing with ZyXEL's PLA-400's a few years back.
For interoperability, I simply checked to see if a TPL-401E would connect and pass traffic to the other devices. The results were as expected: HomePlug AV, yes; HomePlug Turbo, no. Figure 12 shows the combination of TPL-401E "500 Mbps" and TPL-303E HomePlug AV devices produced 55 Mbps best base transmit throughput. This is about 25% lower than the 73 Mbps best case throughput found using two TPL-401E's.
Figure 12: Atheros 500 Mbps to HomePlug AV transmit throughput
For coexistence, I used the same method described here, basically running a "500 Mbps" powerline and HomePlug AV then HomePlug Turbo networks in parallel. Since the "500 Mbps" and HomePlug AV networks do interoperate, I had to change the security code on one of the pairs of adapters (I chose the "500 Mbps" pair) so that they didn't all end up in the same network and cause a broadcast storm.
Figure 13 shows the "500 Mbps" / HomePlug AV coexistence test with the "500 Mbps" pair starting first and the HomePlug AV pair ending last. Sharp-eyed readers will notice that the "500 Mbps" pair exhibits significantly higher speed than shown in the previous testing, with throughput starting at 80 Mbps and climbing to peaks of 112 Mbps!
Figure 13: Atheros 500 Mbps / HomePlug AV coexistence test
I thought this was an anomaly, but it was very repeatable. But when I went back and ran just the 500 Mbps pair, the throughput plots looked like my previous testing, i.e. ~ 70 Mbps. After carefully retracing my steps, the higher throughput appears to be real, but shows up only when there is another active HomePlug network present.
Figure 14 shows the "500 Mbps " / HomePlug Turbo coexistence test, where the "500 Mbps" pair once again exhibits 100+ Mbps throughput!
Figure 14: Atheros 500 Mbps / HomePlug Turbo coexistence test
It seems that the handshake that occurs at the beginning of the IxChariot test with all test endpoints is sufficient enough to kick the "500 Mbps" adapters into the higher throughput. I found that I could even unplug the HomePlug AV / Turbo adapters after starting and quickly quitting the coexistence test, then start just the "500 Mbps" test pair and the 100+ Mbps throughput would remain in effect until the adpaters were power cycled.
For a sanity check, I contacted Atheros, who lined up someone from their PLC applications engineering team to discuss these findings. The upshot of our conversation is that Atheros said what I was seeing could be caused by the less-then-current firmware in the TRENDnet adapters (Atheros INT7400-MAC-5-0-5010-01-650-20100818-FINAL-B) and noise coupling caused by the adapter board design. Since I was able to "trick" the adapters into 100+ Mbps throughput, though, it's possible that the TPL-401Es may just need more current firmware.
Atheros said that their testing showed best case TCP/IP throughput of around 140 Mbps for their reference adapters, which did not vary significantly across locations in their home testing. They also said the 70 Mbps throughput I got for most of my testing indicated that the extended frequencies used in their 500 Mbps technology were not being used.
Atheros is sending me a pair of their reference design adapters so that I can see what their 500 Mbps technology can do when properly implemented. Once I get them tested, I'll post the results in another article.
I thought a gain of 30 Mbps or so from the mid- 40 Mbps throughput provided by the current crop of HomePlug AV adapters wasn't too bad. But I was very disappointed to see that throughput drop off significantly in the same test locations where HomePlug AV's speeds remained rock-solid.
But if Atheros' AMP 500 Mbps technology can really produce steady 100+ Mbps throughput in typical domestic use, I predict that it's going to make life a whole lot easier for folks who have been tearing their hair out trying to wirelessly stream HD!
At this point, however, TRENDnet's TPL-401E doesn't appear to represent the best that "500 Mbps" powerline can do. It's the only "500 Mbps" game in town for the moment, however, at least until NETGEAR starts shipping its XAVB5001 "AV 500" adapters. Maybe now we know why NETGEAR's adapters have been delayed so long.