In reply to a post by mixt:

For a line that must be ~400m long, that's pretty impressive stats (I think the Huawei is reporting rubbish).

If you were the only line on the cabinet, so that no crosstalk existed, you ought to expect that kind of result.

The introduction of vectoring has the aim of making every line behave like it is the only one on the cabinet. Trials of vectoring suggest that lines of 500m can get 100Mbps.

Unfortunately, such speeds may be restricted in the UK,as we employ power reduction masks to ensure that the VDSL2 signal from the cabinets doesn't smother the ADSL signals coming from the exchange. Those masks might cause us to get 5-10Mbps less than the theoretical best.

Interestingly, latency has increased (ever so slightly) and there is an overall higher max latency.

http://www.thinkbroadband.com/ping/share/f4f16f8a6b6...

I presume that's because of G.INP - it's able to handle more errors allowing the line to hold a higher sync, with the side effect of a minor latency increase.

Very interesting graph - I don't think I have seen the green area change pixel-by-pixel before. Obviously I've seen jumps at resyncs, but within a sync - rarely.

But you were getting that behaviour before the resync!

Anyway, with you running at a higher speed by about 5mbps, I'd expect your SNRM to be reduce by 1.5dB or so, and you to be getting a few more errors.

The errors ought to be enough to be handled by G.INP retransmission, but there will be extra latency for *some* packets. That seems to be borne out by the yellow area (the max latency incurred by any packet) being notably higher - by some 2-3ms - while the blue area (the average latency) not really shifting above the green to any greater extent.

Unexplained is the fact that the green area becomes perhaps 1 pixel higher.

Can you see your G.INP counters? The rtx_* counters, and the LEFTRS ones? If so, you might want to track them for a bit.

Tiny changes like that on the minimum line (the green boundary) occur on Plusnet when moving between certain gateways. Others add a few ms rather than 1.

It's quite possible that's what is happening with Zen.

Edited by RobertoS (Thu 30-Apr-15 22:24:45)