In reply to a post by David_W:

That's not necessarily how crosstalk works. You can have many connections that do nothing to your speed, then one further line is connected that interferes badly with yours and drops your speed significantly. There was a recent thread on here where someone lost 20% of his downstream sync speed in one go.

That is close, but not quite right - it leaves out the "threshold" nature of DLM.

I'd re-write your answer as this:

You can have many VDSL connections that do not interfere with your line, some connections that interfere a little, and a few that have the potential to interfere a lot - the effect appears random, so the line that causes most interference might not be one of your direct neighbours.

Each small bit of interference can cause a small loss of speed, or can cause additional errors on your line.

BUT... when the total interference adds up, and causes the errors to hit a threshold, DLM will intervene, and add protection mechanisms to your data to reduce the error rate.

When DLM does intervene, you can expect it to drop your speed anywhere between 10% and 20%.

In reply to a post by Andrue:

I must admit I was thinking any drops would be a percentage rather than in terms of absolute figures but I could be wrong.

It does tend to be in percentage terms - especially when DLM steals some bandwidth to turn on FEC. Interleaving alone doesn't drop the speed, and has been seen to happen when you have extra capacity to spare (especially upstream).

I'd rather hope as Mr Saffron wrote that it being an established cabinet means he's already dealing with crosstalk but we've no way of knowing. Hopefully they'll come back with a status report in a few months

That's the problem. An established cabinet means that crosstalk is already happening, but you will never know if you are already suffering, or whether your biggest disturber is going to sign up next week.

So if one can only "see" the IP Profile then the IP Profile could drop proportionately to a 20% drop in sync speed?

In reply to a post by 4M2:

All this talk regarding cross talk: say one syncs at ~16Mbps on ADSL2+ then one moves to FTTC and all 288 lines eventually are in use; at a distance of ~250 metres from the cabinet would a one see significantly higher speeds, i.e. 32Mbps or more, than the previous ~16Mbps - in other words would it really be worthwhile moving to fibre except for the increased upstream speed?

There is a document from the Broadband Forum that describes Vectoring, and includes a graph of behaviour at different distances.

That graph shows the optimal behaviour (no crosstalk), plus average and a "99% worst case" figure too. It then also shows the values when vectoring is turned on.

For a 250 metre line, it shows the following:
- Theoretical top speed: 160Mbps
- Range of speeds with crosstalk present: 60-95Mbps
- 99% worst case: 50Mbps
- Range of speeds with vectoring present: 140-160Mbps

So yes, you would see significantly higher speeds even if you encountered the (almost) worst possible case, but you are highly likely to get into the 70-80Mbps region.

The graph is on page 12 here.

In reply to a post by 4M2:

So if one can only "see" the IP Profile then the IP Profile could drop proportionately to a 20% drop in sync speed?

Yes.

IIRC, It is quite an aggressive FEC setting to get a 20% drop.

Thanks for the link - I will have to study it in more detail later

Interesting though that with a 99% worst case, with cross talk present, one has to be ~1000 metres from the cabinet before the speeds approximate to the max 24Mbps sync speed possible with ADSL2+ - that's if I'm reading the graph correctly?

Edited by 4M2 (Tue 22-Oct-13 15:50:40)

well DLM is an artificial change added by BT thats not part of VDSL spec.

Thats why we didnt mention it.

So DLM is a possible indirect affect of crosstalk. But not direct.

In reply to a post by 4M2:

Thanks for the link - I will have to study it in more detail later

Remember that it really describes a modelled situation (ie based on what wiring ought to be out there), so doesn't really take account for the cases where faults & problems do exist, such as corrosion.

It is also based on 0.4mm wires, so the UK ought to be in a slightly better situation.

Interesting though that with a 99% worst case, with cross talk present, one has to be ~1000 metres from the cabinet before the speeds approximate to the max 24Mbps sync speed possible with ADSL2+ - that's if I'm reading the graph correctly?

That's about the right reading of the graph - I'd say somewhere in the 900-1000 metre region.

But we hit the same proviso as above, that we use 0.5mm wires, so the distance ought to be longer - perhaps by as much as 20-30%.

True, it is an indirect effect.

However, the existence of a DLM is built into the VDSL2 spec, as all the facilities necessary for that DLM to work are specified. VDSL2 is written with an understanding that some form of DLM is going to exist and drive it.

But the specifics of BT's implementation of DLM isn't directly specified by VDSL2.

With luck, DLM will change alongside vectoring - and replace the use of FEC/interleaving with PHYR.

Perhaps that also really emphasises the need for 24SWG (0.5mm), solid core, 100% copper CAT5e for anybody wishing to use a vdsl "data extension", of up to 30 metres in length, for their modem particularly if they are 500 metres or more from the cabinet?

Edited by 4M2 (Wed 23-Oct-13 01:16:45)