Just upgraded from the Nortel 1-meg modem to the Alcatel 3mbps modem with Bell Canada's Sympatico high speed consumer service. Was initially unimpressed with the 1.1 mbps download speed tests (cf 800 kbps on the old platform). Became singularly depressed with slowdowns
- sometimes dropping to below 15kbps - and virtual disconnects. Suspected some issue with my SMC barricade g wireless router as it has been flaky on the wireless front. Substituting my old 1-port Linksys router initially showed speed tests up to 1.6 mbps but became similarly sporadic. Re-booting modem or router after an extreme slowdown might restore speed temporarily and might not.
The modems were both located right at the line entrance. One filter (upgraded with the 3-meg) branched off to feed a couple of phones. Another couple of lines branched off to phones with filter at the phone (in support of my pre-router "Access" dial-up). Disconnecting the local filter (and its phones) resulted in pretty consistent 2.6 mbps speed tests, without the fatal slowdowns. Great DSL but no kitchen phone! Restoring the connection with a different filter degraded back to the slow, unreliable connection so not a defective filter. Relocated the modem an extra 30 ft or so to be adjacent to router to see what was going on. With the local filter disconnected as before, 2.6 mbps. Now here's the surprise: re-connecting the local filter at the line entrance did not degrade performance. Now I've got great DSL and a kitchen phone!
The only change is that the modem is now about 30 ft of phone wire more distant from the line entrance and the filter that branches from there. Why would adding cable improve performance when little change or slight degradation might be reasonably expected?
I think transmission line theory provides an explanation. The 3 mbps DSL spectrum must extend into the radio frequency range of at least a few MegaHertz and maybe even above 10 MHz. The phone cable has a characteristic impedance at these frequencies and behaves as a transmission line. The filters and phones present a complex impedance that is not a perfect match to the transmission line. At impedance mismatches, a fraction of the energy proportional to the magnitude of the mismatch is reflected. Such reflections also occur at the junction of two different kinds of cable having different characteristic impedances and wired in series. These reflections bounce back and forth getting progressively weaker but the time delays of the echoes distort the shape of the pulse train at the receiver and can cause (jitter) errors in distinguishing between 1's and 0's. I suppose that displacing the modem an extra 30 ft or so changed the timing relationship of these echoes in a favourable way for me. The opposite might be the case in another house, or another part of this house. As they say, your mileage might vary.
Knowing more about the DSL spectrum and the transmission line properties of household phone wiring might suggest optimal lengths for modem or filter displacement that may be tried for possible improvement.
Does anyone else share my theory and have more information on optimising household phone wiring for DSL performance?
Tom