I said I would compare the Mark3 with the BT NTE-2005 ADSl v1.0 faceplate made by Pressac. The NTE-2005 has from my looking a the circuit no "RF3" type component and connector A and B from the drop wire go straight to the adsl line.
Here are the Mark 3 stats.
Max: Upstream rate = 23663 Kbps, Downstream rate = 77788 Kbps
Bearer: 0, Upstream rate = 20000 Kbps, Downstream rate = 72030 Kbps
Down Up
SNR (dB): 7.9 6.6
Attn(dB): 17.5 0.0
Pwr(dBm): 13.3 5.1
Here are the NTE-2005 stats
Max: Upstream rate = 23936 Kbps, Downstream rate = 79268 Kbps
Bearer: 0, Upstream rate = 20000 Kbps, Downstream rate = 72030 Kbps
Down Up
SNR (dB): 8.3 6.7
Attn(dB): 17.5 0.0
Pwr(dBm): 13.3 4.9
With the absence of the "RF3" type component I was expecting a higher attainable and indeed it is 1.5 Mb/s higher
. My previous impression was that the Mark 2 had a more aggressive "RF3" than the Mark 3 so I guess that a Mark 2 to NTE-2005 swap should show a slightly bigger improvement. I attach the respective snrm and bitloadings per tone. After 45 mins the error rates (with the speed capped) look less
but a reliable error rate test will take a few days to do properly.
Very Tentative conclusion Both the Mark 2 and Mark 3 may have common mode filtering that has a negative impact impact on some (possibly many) lines. Almost all modems are pretty good at rejecting common mode and I have long wondered why BT thought they could do better. The RF2/3 may have had a purpose when people could hear radio channels on their phone calls but the adsl/vdsl filter in a face plate should stop RF on the drop wires reaching the phone (I think!).