Kitz Forum
Broadband Related => Broadband Technology => Topic started by: William Grimsley on January 29, 2016, 08:02:29 PM
-
Hi,
Does anyone know what a typical maximum data rate drop per FTTC connection is? Is it about 200 - 500 kbps?
Thanks.
-
What do you mean? Are you talking about crosstalk?
-
if you mean crosstalk over the life of the connection it can be a whole lot more than 500kbps, but like eric said you need to explain what you mean.
-
Yes, sorry I am talking about crosstalk.
-
only BT will have the figures for their copper network, but in published trial results, average speed drop can be measured in 10s of mbit/sec rather than a few hundred kbit/sec.
-
only BT will have the figures for their copper network, but in published trial results, average speed drop can be measured in 10s of mbit/sec rather than a few hundred kbit/sec.
For one customer? If that was for one customer, I'd be back on ADSL speeds by now!
-
There's a document from the Broadband Forum (that puts together standards for how DSL systems work) that describes vectoring.
Inside, there's a great graph that shows just how bad crosstalk can get, in the absolute worst case. And how good things can be if vectoring is implemented.
The downside is that it doesn't show FTTC in quite the same way it works in the UK - it doesn't include the power masking we use to keep ADSL working, and it is based on 0.4mm copper, where we are more likely to be on 0.5mm. However, it gives a good overview...
https://www.broadband-forum.org/marketing/download/mktgdocs/MR-257.pdf
Figures 6 and 7 are most appropriate.
I can't explain it in more detail right now, but ask questions for when I'm back again...
only BT will have the figures for their copper network, but in published trial results, average speed drop can be measured in 10s of mbit/sec rather than a few hundred kbit/sec.
For one customer? If that was for one customer, I'd be back on ADSL speeds by now!
In crosstalk, the pairs that run closest to yours are always going to be the biggest disturber. If those don't carry VDSL, then you'll see little. If they do, you'll see a lot. Other pairs sitting a few mm further away in the bundle will affect your line less.
You can find heatmaps from lab results, showing a 2D map of how line X affects line Y.
-
only BT will have the figures for their copper network, but in published trial results, average speed drop can be measured in 10s of mbit/sec rather than a few hundred kbit/sec.
For one customer? If that was for one customer, I'd be back on ADSL speeds by now!
yeah but vdsl still tonks adsl.
would rather have a 110mbit line down to 70mbit from crosstalk than a 6mbit line
-
There's a document from the Broadband Forum (that puts together standards for how DSL systems work) that describes vectoring.
Inside, there's a great graph that shows just how bad crosstalk can get, in the absolute worst case. And how good things can be if vectoring is implemented.
The downside is that it doesn't show FTTC in quite the same way it works in the UK - it doesn't include the power masking we use to keep ADSL working, and it is based on 0.4mm copper, where we are more likely to be on 0.5mm. However, it gives a good overview...
https://www.broadband-forum.org/marketing/download/mktgdocs/MR-257.pdf
Figures 6 and 7 are most appropriate.
I can't explain it in more detail right now, but ask questions for when I'm back again...
only BT will have the figures for their copper network, but in published trial results, average speed drop can be measured in 10s of mbit/sec rather than a few hundred kbit/sec.
For one customer? If that was for one customer, I'd be back on ADSL speeds by now!
In crosstalk, the pairs that run closest to yours are always going to be the biggest disturber. If those don't carry VDSL, then you'll see little. If they do, you'll see a lot. Other pairs sitting a few mm further away in the bundle will affect your line less.
You can find heatmaps from lab results, showing a 2D map of how line X affects line Y.
at the same time was no ali tho which the uk has a lot off, leics is a mix of sub 0.3 copper and ali
-
There's a document from the Broadband Forum (that puts together standards for how DSL systems work) that describes vectoring.
Inside, there's a great graph that shows just how bad crosstalk can get, in the absolute worst case. And how good things can be if vectoring is implemented.
The downside is that it doesn't show FTTC in quite the same way it works in the UK - it doesn't include the power masking we use to keep ADSL working, and it is based on 0.4mm copper, where we are more likely to be on 0.5mm. However, it gives a good overview...
https://www.broadband-forum.org/marketing/download/mktgdocs/MR-257.pdf
Figures 6 and 7 are most appropriate.
I can't explain it in more detail right now, but ask questions for when I'm back again...
only BT will have the figures for their copper network, but in published trial results, average speed drop can be measured in 10s of mbit/sec rather than a few hundred kbit/sec.
For one customer? If that was for one customer, I'd be back on ADSL speeds by now!
In crosstalk, the pairs that run closest to yours are always going to be the biggest disturber. If those don't carry VDSL, then you'll see little. If they do, you'll see a lot. Other pairs sitting a few mm further away in the bundle will affect your line less.
You can find heatmaps from lab results, showing a 2D map of how line X affects line Y.
That's really interesting! Just shows that it all depends on the type of connection that other pairs are running on which makes a lot of sense as if they're running VDSL like I am then I'm going to see the biggest drop in speed.
I didn't see a graph showing how bad crosstalk can get. Is that in the linked document or is that somewhere else?
-
Yes - figure 6 in that document.
Let me see if I can annotate it a bit...
-
There is a newer edition of that document:
https://www.broadband-forum.org/marketing/download/mktgdocs/MR-257_Issue-2.pdf
-
Here's a copy of the figure, annotated.
Remember that it isn't very indicative of UK behaviour - but is indicative of a flavour of VDSL2 somewhere, and gives a reasonable picture of how badly crosstalk *can* impact things.
The figure depicts 80 VDSL2 users in a cable with 100 pairs. That is very high usage.
The upper blue triangles represent perfect behaviour. You might see this if you were the only VDSL line on the cab.
The red circles represent the speeds that could be achieved for vectored lines.
The blue x-shaped crosses represent the range of speeds for the non-vectored lines.
The lowest blue triangle represent a theoretical worst-case speed.
In The UK, we run without vectoring, so could expect speeds to be somewhat like the blue x-shaped crosses.
Of course, take-up probably hasn't reached 80% on many cables yet, so things might not be as bad as that.
However, the "tie pairs" that carry VDSL2 signal back from the FTTC cabinet into the PCP, before distribution around the neighbourhood *will* end up completely full of VDSL2 users. We will see full crosstalk there - but thankfully only for a relatively short distance.
I've added 2 explanations to the figure
- Blue text for behaviour at 200m
- Purple text for behaviour at 500m
-
Thanks @ejs. I'll take a look later.
-
I think that particular diagram is a strange one, because it's the performance with some of the lines using vectoring, and some not, at the same time.
-
Yes, that graph has its strange points. I like it because it makes you realise that there is a range of effects, and some randomness. Nothing about crosstalk is perfectly predictable.
There are others around, and can usually be found in articles and presentations on vectoring - because that is the common "solution" to crosstalk. A shame we're not getting it widely deployed here.
By the way, there is another article here, that talks about the impact of crosstalk:
http://www.increasebroadbandspeed.co.uk/2014/vectoring-crosstalk-crisis
And I found a "heatmap" in this presentation on Vectoring trials in Denmark, page 8:
https://marketingwholesale.tdc.dk/assets/tdcws/1411340050.pdf
That presentations has a graph showing speed improvements in the trial, on page 10. The gains aren't necessarily as high as in the BBF document - presumably with only 20 real customers, spread out geographically, there wasn't as much crosstalk as you see in some of the lab trials.
-
I was surprised fig. 6 in MR-257 showed vectoring working fairly well, despite only slightly over half the lines were using vectoring. I thought vectoring worked best if all the lines are vectored, but the vectored lines seemed to be doing pretty well despite plenty of non-vectored lines.
The next graph in MR-257, shows a drop of about 10 Mb, from 15 other lines to 49, no vectoring.
-
Last year I was quoting that the average speed lost by members of this forum through crosstalk appeared to be in the region of 15-20Mbps. This is a rough ball park figure based purely on the stats of those who were monitoring their lines.
Bear in mind that crosstalk affects the shorter lines the most and these are the ones who tend to lose the most speed. In the few years since Ive had fttc Ive lost just slightly over 25Mbps from my headline rate due to crosstalk, some have lost more than that.
-
Yeah when I said averages are in 10s of mbits second thats also taking into account on the trial average initial speeds were high also. A vdsl line syncing at say 20mbit without crosstalk isnt going to lose 10s of mbits to crosstalk.