Powerline Followup: Preliminary Results and Next Steps

Continued from 'Powerline Followup: Is DS2 For You?'

And here are the performance results, first with the interference-suppressing notch filters disabled, when writing to (I'm eyeballing the speed at ~35 Mbps sustained, ~45 Mbps peak) and reading back from (~35 Mbps peak, ~25 Mbps sustained) the ReadyNAS. Note the variability in the bandwidth, and that (as predicted by the diagnostics utilities) the bandwidth flowing from the hut back to the office is slightly lower than the bandwidth going the opposite direction.

I repeated the results with the notch filters enabled, and got the following results: ~30 Mbps sustained when writing to the ReadyNAS and ~20 Mbps when reading back, in line with the 20-25% performance degradation that DS2 had previously suggested I'd see with notching enabled. The write waveform, with notch filters on, is much more stable than its non-notch counterpart, wouldn't you agree? Note, I did not yet test the adapters with TripleDES encryption enabled, although I plan to; DS2 has already informed me not to expect a performance impact, which I'd like to confirm.

Granted, the DS2 adapters are faster than the HomePlug v1.0 Turbo units (before you ask, I didn't test HomePlug AV adapters because they're not available yet). I admit, though, that I'm a bit underwhelmed by the read and write speed that the DS2 powerline link is delivering, especially considering the '200 Mbps PHY, 100 Mbps real-life' marketing claims that motivated my evaluation of it. DS2's documentation suggests a number of configuration options, such as output power, that a user can manipulate in search of higher speed. I suspect I'll get some feedback from DS2 to guide me on which of these might help in my particular situation, and I'll try them out as time allows.

DS2 has also already suggested upfront that I might want to tweak the TCP Window Size variable in my Windows configuration in order to account for the potentially higher latency of a powerline network versus one based on CAT5. To date I've resisted this for several reasons; 1) it's beyond the capabilities of the average consumer (although it'd be simple to create a program which'd automate this adjustment), 2) doing so might degrade the performance (LAN, WAN, or both) through other network media, a no-no in a mixed-media (wired, wireless and powerline) network such as mine and 3) some network clients (such as my NASs) don't have user-configurable TCP/IP stacks, so altering the TCP Window Size setting isn't feasible for them.

My next step will be to search for interference caused by the powerline adapters in the amateur and shortwave radio broadcast bands, and whether or not the notch filters adequately suppress this interference. DS2's Vice President of Technology informed me this morning that the DS2 notch filters operate in the following frequency bands (kHz):
1800-2000
3500-4000
7000-7300
10100-10150
14000-14350
18068-18168
21000-21450
24890-24990
28000-29700
30000-32000

Ed Hare from the American Radio Relay League, a prolific poster on my past powerline networking writeup, has generously provided me with a portable shortwave radio that I'll use to accomplish this task.

p.s….it never fails; yesterday I did my testing, and this morning I coincidentally got an email from DS2 indicating that an updated firmware version was available. In the words of Chano Gómez, the company's Vice President of Technology, 'amongst other things (higher speed, etc)' it 'includes improved notches (reducing power levels even further in the notched bands), making our system even more friendly with other radio services.' I'm dropping off the adapters Friday at DS2's Santa Clara office for upgrade and, after they're back in my hands, I'll repeat my performance testing and will then do interference testing.