Kitz Forum
Broadband Related => Router Monitoring Software => Topic started by: AArdvark on September 23, 2014, 11:07:20 PM
-
I have noticed a strange blip on my FTTC line at 22:35 23/09/2014.
My max attainable and SNRM jumped up by a large amount for 3 mins then dropped again.
Very odd.
I only noticed when I looked at a 24hr graph by random luck.
Any comments explaining what might be happening would be appreciated.
-
Maybe a neighbour rebooted or switched off their modem for 3 minutes, allowing you to experience a glorious 3 minutes worth of relatively crosstalk-free connection.
-
Maybe a neighbour rebooted or switched off their modem for 3 minutes, allowing you to experience a glorious 3 minutes worth of relatively crosstalk-free connection.
Would this be conclusive proof that I am suffering quite heavy crosstalk ?
My neighbours are only on ADSL2 at the most, would it improve if they were on VDSL2 as well or get worse ?
I am trying to anticipate the future as some of my neighbours are looking to get 'Superfast Broadband' so more is yet to come. :(
-
Maybe a neighbour rebooted or switched off their modem for 3 minutes, allowing you to experience a glorious 3 minutes worth of relatively crosstalk-free connection.
Would this be conclusive proof that I am suffering quite heavy crosstalk ?
I think it would take a number of similar results that could be intentionally initiated & repeated at will by switching off/on a neighbour's connection before any 'real' conclusions could be drawn.
It might just have been an unexplainable one-off event.
My neighbours are only on ADSL2 at the most, would it improve if they were on VDSL2 as well or get worse ?
ADSL crosstalk only affects the lower of the available VDSL2 frequencies, so your higher VDSL2 frequencies are currently probably not being 'interfered' with.
As an increasing number of neighbours sign up for VDSL2 connections, I can only imagine that crosstalk will increase rather than decrease.
I am trying to anticipate the future as some of my neighbours are looking to get 'Superfast Broadband' so more is yet to come. :(
I don't think anything can be done to avoid increasing crosstalk, at least until vectoring is introduced.
Just keep an eye open for smallish, but sudden & sustained reductions in SNRM, probably followed a few days later by a resync at lower speeds with SNRM back to 'normal' levels.
That's exactly what I have seen on my connection over 12 months or so, causing me to gradually lose around 1/3 of my DS sync speed.
-
I can only concur with what BE has already said... and also add that cross-talk doesnt have to come from an immediate neighbour, it could be in a different street.
-
a large disturber probably temporarily went offline.
-
In Ireland, the reports of changes to SNRM when vectoring was turned on were of the order of 10-12dB, suggesting that was the kind of crosstalk being suffered.
The change you saw could well have been from one of your significant disturbers turning off.
-
thats huge.
Its a real shame BT appear to not be too bothered about crosstalk.
-
Ive no doubt they are bothered.. but in their usual manner, they will be looking at existing technology then coming up with their own way of doing things.
10-12dB may seem a lot.. but then again look how much some of us on this forum have seen our speeds reduced by. BS has seen 25Mb, Ive seen 20Mb and I wasnt even first on the cab. Aardvark hasnt been enabled that long and already hes seen some pretty steep losses :(
-
10-12 db on my line is about 28mbit attainable, its a lot. I would go from a 73 sync to full 80 with about 101 attainable so about 10db or so snrm with max sync.
By the way I do agree BT do seem to waste money and time redesigning the wheel.
-
By the way I do agree BT do seem to waste money and time redesigning the wheel.
It's not funny when you see are UK Broadband is years behind other countrys, who's to blame here BT or the government :-\
-
Broadband Forum has produced this document as a description of vectoring:
http://www.broadband-forum.org/marketing/download/mktgdocs/MR-257.pdf
Figure 6 shows the effects of many competing VDSL2 lines, and the impact of crosstalk on speeds - and the improvement possible by vectoring. Note: I think it refers to 0.4mm copper, so isn't directly equivalent to UK lines.
For example, at 400 metres, in a heavily used system, the majority of lines will be running between 65 and 90Mbps, while the absolute worst case is around 45Mbps, and the vectored speeds are around 120Mbps.
That shows the huge variance - huge randomness - in the way crosstalk can affect lines. Plus, of course, the huge amount of speed that can be lost.
-
both.
ofcom are anti progress, and the gov control ofcom. e.g. FTTP would be an issue for ofcom as it sort of breaks LLU.
BT are extremely cautious with spending money which seems to have been because of their past failures in the late 1990s.
Also remember the gov refused BT's deal to do a nationwide FTTP as they wanted competition instead. Which came in the form of the cable tv companies and sky.
BT's obsession with very long testing period's and wanting to use things not the way they originally designed is entirely their own fault tho. I guess one of the obstacles for vectoring is that it is something entirely for the purpose of improving the customer experience but without extra revenue for BT, so there is arguably no business case for it other than higher customer satisfaction. There is the possibility vectoring will reduce fault complaints which is a business case as engineer callouts cost money.
Dont under estimate the problem of lack of competition, although we suffering from the lack of wholesale competition it is there at the retail level, in america where it isnt there, they have issues with congested peering, netflix and co having to pay isp's to uncongest routes and lower speeds.
Also the UK is only behind in cities, in rural areas the UK is probably more advanced. So I do respect its not all bad.
I also think without ofcom wholesale BT line rental would be sky rocketing but thats regulated, so BT do it at the retail level instead, and of course other telcos are mirroring BT on the line rental rises effectively making a competition failure on line rental.
-
Thanks for all the comments.
I did not reply as there was little I could say in response.
You have all confirmed what I already know BUT I still try to find an alternative answer 'just in case' I am wrong !!! ;D
I will not hold my breath for vectoring as it took me many years to get past 4Mbps.
['Tongue in cheek' mode on]
What is needed is for some 'cheap compatible hardware' to become available to provide the Vectoring Infrastructure which may encourage BT to spend some money.
['Tongue in cheek' mode off]
(Although I really expect another 'BDUK type event' will happen and we (Joe Public) will end up paying for BT to upgrade their network again ;D :D ::) )
If you think the situation in the UK is bad, see Australia and their 'NBN'
http://en.wikipedia.org/wiki/National_Broadband_Network (http://en.wikipedia.org/wiki/National_Broadband_Network)
(An alternative model to do a 'National Broadband rollout' the Hard/Expensive way .....[Telstra is the Australian BT]) ;D ;D :no:
-
10-12 db on my line is about 28mbit attainable, its a lot. I would go from a 73 sync to full 80 with about 101 attainable so about 10db or so snrm with max sync.
Ive just had a quick look at some of my old stats - a very approx figure is 1dB of SNRm = 3 Mbps of bandwidth. Obviously this figure wont be the same for everyone as it will depend on how many tones can be bit loaded throughout the full spectrum
-
Broadband Forum has produced this document as a description of vectoring:
http://www.broadband-forum.org/marketing/download/mktgdocs/MR-257.pdf
Figure 6 shows the effects of many competing VDSL2 lines, and the impact of crosstalk on speeds - and the improvement possible by vectoring. Note: I think it refers to 0.4mm copper, so isn't directly equivalent to UK lines.
For example, at 400 metres, in a heavily used system, the majority of lines will be running between 65 and 90Mbps, while the absolute worst case is around 45Mbps, and the vectored speeds are around 120Mbps.
That shows the huge variance - huge randomness - in the way crosstalk can affect lines. Plus, of course, the huge amount of speed that can be lost.
Thank you for that link - difference in gauge noted, but the figures are indeed still going to be large.
It shows though that what weve always said that the shortest lines are impacted by far the most by crosstalk.. and according to that chart Vectoring is going to make very little difference for those with long d-sides. Those with long d-sides often have a better quality gauge copper... as may some of those with what would have previously been a long e-side, so that may cause some variance too, if they are now close to a cab.
-
I calculated 3db as around 7mbit for lines that can use the full frequency
but I may have got it wrong
for longer d-sides kitz the number may be smaller but I am guessing its still significant as a % of whole speed, e.g. BE reports he lost a lot.
e.g. 5mbit lost to crosstalk doesnt sound a lot, but it is if the entire sync is 15mbit.
your extra 2mbit per 3db may well be accounted for by that you can use all tones in D1 whilst I effectively cannot due to agressive adsl power cutback.
-
kitz we talked about BT redesigning the wheel and BS said BT are only doing NEXT mitigation in the testing?
yet this is quoted in the article.
NEXT coupling is generally more severe than FEXT coupling. However, since VDSL2 uses
separate frequency bands for upstream and downstream transmission, the upstream and
downstream bands on all the wire-pairs in the cable never overlap so there is no self near-end
crosstalk from similar signals in the cable. Hence, for VDSL2 self far-end crosstalk (SELF
FEXT) is the dominant crosstalk disturber
so either we hope BS is wrong, or BT do plan to do FEXT but are not doing it in this particular test.
That graph seems to have been sampled from cable better quality then BT, I had 110 attainable yet their reported FEXT free sync rates are way higher, for 100m is like 180 attainable O_o
-
What is needed is for some 'cheap compatible hardware' to become available to provide the Vectoring Infrastructure which may encourage BT to spend some money.
The vectoring tests that BT did previously were with a generic FPGA DSP - which *might* mean it would run on the DSP's already installed in the Huawei DSLAMs. It could be just a software change to enable vectoring.
The current field test is with ASIC hardware (ie specifically-designed hardware). I imagine they are testing to see if there is a difference in performance, to decide whether it is worth using fresh hardware.
I recall seeing, a while back, an estimate of the lifetime costs of FTTC; that suggested that the lifetime of the DSLAM electronics within the cabinet is about 8 years. That might mean that BT would be expecting to have a programme to refresh the hardware - which might then be the point that vectoring-specific hardware gets added.
Either way, I imagine that the contract that BT has with Huawei pays for the feature-set used; if BT turn on vectoring within existing hardware, I'd imagine that Huawei would be due some extra dosh.
(Although I really expect another 'BDUK type event' will happen and we (Joe Public) will end up paying for BT to upgrade their network again ;D :D ::) )
I imagine that this could well happen.
However...
What BDUK has paid for, in these rounds, seems to include dark fibre delivered to (the vicinity of) nearly every cabinet. The majority of properties served nationwide are within 1km of their cabinet - and this is true whether the cabinet was upgraded commercially or with BDUK. The real difference in viability was the number of properties served by the cabinet, plus the ease of getting fibre/power to the box.
The next round of upgrades are likely to be either FTTP/FTT-outside-wall or a hybrid with FTTdp. The decision about viability there is likely to be based around groups of 10-20 houses.
It strikes me that the FTTC-based outcome, with dark fibre, of this set of BDUK projects is likely to make "rural" groups-of-properties just as commercially-viable for future upgrades as "urban" groups-of-properties. The distance from "cabinet+dark fibre" to "group-of-properties" is likely to be similar (at least for a good chunk of "rural"), and it might even be easier/cheaper to feed the fibre in rural areas.
If you think the situation in the UK is bad, see Australia and their 'NBN'
Australia is a strange case - they do actually seem to have problems with the quality of the copper. I read that, for a while ten or twenty years ago, they'd been sealing copper joints with some gel of one sort or another. It turns out that this was a bad move - causing corrosion up to 3m either side of the joint. That makes it questionable as to whether their copper really is up to the job of carrying FTTC! http://www.abc.net.au/technology/articles/2013/09/19/3851924.htm (http://www.abc.net.au/technology/articles/2013/09/19/3851924.htm)
Other than that, they've done one of the hard jobs of a fibre rollout - getting together a single, nationwide plan, getting the infrastructure owners into that plan, and figuring out the funding. Politics & economics are the problems, not the technology...
OK. They've utterly changed the technology they're planning to use in the newer "mixed" model - but I tend to believe that we have the technology to meet whatever plan we want to go for. The hard part is getting an agreement on the plan, and funding it.
-
kitz we talked about BT redesigning the wheel and BS said BT are only doing NEXT mitigation in the testing?
I can't imagine that BT are trialling vectoring for NEXT purposes - the problem is indeed all about FEXT, in both direction.
The document I linked even tells us that the title of the ITU standard for vectoring is: "Self-FEXT Cancellation (Vectoring) for
use with VDSL2 transceivers"
The NEXT problem would rear its head if subscribers were allowed different bandplans, where the downstream-upstream bands split at different frequencies, but the ANFP kills this by demanding just one bandplan (which, of course, changed in 2011-2012)
That graph seems to have been sampled from cable better quality then BT, I had 110 attainable yet their reported FEXT free sync rates are way higher, for 100m is like 180 attainable O_o
It is hard to find examples that specifically use BT's gauge and noise model, however we should bear in mind that almost no-one will ever have experienced a true FEXT-free connection. The design of the DSLAM/PCP tie cables ensures that, once there are 2 or more subscribers, *everyone* gets some crosstalk from the first 5-100m of cable.
-
On the question of how many Mbps per dB of SNR (on which I can safely say I know little about), I did find this:
https://www.iol.unh.edu/services/testing/dsl/tools/UCP/ (https://www.iol.unh.edu/services/testing/dsl/tools/UCP/)
This is from the lab that runs the inter-operability tests for the vectoring hardware.
Their test setup inserts 10-20dB of crosstalk into the line; the result of the crosstalk is a 9dB drop in SNR reported by the modems and a 13Mbps drop in speed, both undone by vectoring.
-
Australia is a strange case - they do actually seem to have problems with the quality of the copper.
Last OT comment.
From friends in Australia, one of the complaints seems to be they have/are paid/paying to fix the copper only to now have to pay to rip out the same copper to upgrade to Fibre (NBN).
From what I think I understand there is a commitment to replace the copper by fibre and move services over to it as part of the NBN Rollout.
Also there is some debate about the real coverage possible as the intent was originally 90% but now is no where near that figure.
(Again this is being debated as to the real figure possible!!!)
Don't even go near the politics !!!
-
Thanks www.. I just took a stab based on one of my 'cliff edge' dips and how much I lost according to my stats.
-
kitz we talked about BT redesigning the wheel and BS said BT are only doing NEXT mitigation in the testing?
I can't imagine that BT are trialling vectoring for NEXT purposes - the problem is indeed all about FEXT, in both direction.
The document I linked even tells us that the title of the ITU standard for vectoring is: "Self-FEXT Cancellation (Vectoring) for
use with VDSL2 transceivers"
The NEXT problem would rear its head if subscribers were allowed different bandplans, where the downstream-upstream bands split at different frequencies, but the ANFP kills this by demanding just one bandplan (which, of course, changed in 2011-2012)
That graph seems to have been sampled from cable better quality then BT, I had 110 attainable yet their reported FEXT free sync rates are way higher, for 100m is like 180 attainable O_o
It is hard to find examples that specifically use BT's gauge and noise model, however we should bear in mind that almost no-one will ever have experienced a true FEXT-free connection. The design of the DSLAM/PCP tie cables ensures that, once there are 2 or more subscribers, *everyone* gets some crosstalk from the first 5-100m of cable.
A small 'Cut & Paste' from one of our communiques.
The problem:
Crosstalk is due to coupling between pairs in a multi-pair copper cable and limits FTTC speeds
Increasing cable fill through customer take-up increases crosstalk and reduces line-rates
The solution:
By calculating a “pre-compensation” factor for each line, the crosstalk can be cancelled in the DSLAM
The technique for doing this is called “Vectoring”, and is conceptually similar to noise cancelling headphones
Vectoring is global standard with multiple suppliers and is in deployment by other operators
-
ok that wording to me just says if vectoring is enabled in the dslam it can cancel the crosstalk doesnt mention NEXT at all.
Of course it also requires modem compatibility which I would like to think BT understand that as well.
-
I tend to agree.
By calculating a “pre-compensation” factor for each line, the crosstalk can be cancelled in the DSLAM
Ive said before in another post (http://forum.kitz.co.uk/index.php?topic=14448.msg270097#msg270097) that its the above type of phrase could be what is confusing people, only last time I used the word at rather than in and Im not sure if anyone picked up what I was getting at.
ie
1) Crosstalk can be cancelled [by vectoring equipment] IN the DSLAM
or
2) Crosstalk can be cancelled [that originates] AT the DSLAM
See how those 2 sentences mean entirely different things?
Judging from the above quote I'd say that 1) is the most likely
-
I have always assumed that 1) is the case. :)
-
aha... it doesnt make sense for them to ignore FeXT because thats where most of the issues are said to be.
NeXT can to some extent be lessened by use of PSD masks and Power Cut Back. FeXT affects VDSL2 badly particularly the shorter lines.
The vectoring engines are supposedly designed to eliminate FeXT
-
The technique for doing this is called “Vectoring”,
I know wwwW has already posted some good links.. but heres one to the Huawei equipment that is supposedly being used during the phase1 trials.
http://www.huawei.com/en/about-huawei/publications/communicate/hw-145796.htm
As you can see they address the issue of correcting crosstalk which occurs in the copper pairs sheathed in the cable with other pairs and not x-talk which occurs at the dslam.
Ive mentioned elsewhere that a lil birdie told me the BT Phase1 trials were done using the Huawei vectoring equipment. Phase 2 (on which theyve gone very quiet) is supposedly about them designing their own hybrid solution [which rumour and I can only stress rumour, has it may also work with the ECI cabs]. Hence the earlier comments triggering 're-inventing the wheel'.
-
My understanding matches (1) rather than (2).
The way that I understand how vectoring works is as follows:
- During the sync process, the CPE modem measures how much error exists in the QAM constellation as the received signal gets decoded from the analogue; this error-data is what is sent back to the DSLAM. This is the bulk of the contribution that the CPE modem has to make to the vectoring process.
- The error-data gathered allows the DSLAM to figure out which other lines are the major FEXT disturbers of this line.
- During showtime, the DSLAM uses the error data to calculate & apply a pre-compensation prior to transmission, in real-time, such that when received at the CPE modem, the received error (predominantly caused by self-FEXT) is cancelled out by the calculated pre-compensation.
Thus the vectoring equipment sits in the DSLAM, where the bulk of its work is at showtime. However, the work it does is to calculate an "anti-noise" compensation that allows the CPE modem (doing business as normal) to perceive the received signal to be error-free.
Note that vectoring is also applied to the upstream, but my understanding is that it works entirely within the DSLAM this time - using the same process as downstream, but performing the sync measurements within the DSLAM, and applying a post-compensation at the receiver (instead of pre-compensation in the transmitter) within the DSLAM.
Ive mentioned elsewhere that a lil birdie told me the BT Phase1 trials were done using the Huawei vectoring equipment. Phase 2 (on which theyve gone very quiet) is supposedly about them designing their own hybrid solution [which rumour and I can only stress rumour, has it may also work with the ECI cabs]. Hence the earlier comments triggering 're-inventing the wheel'.
It wouldn't surprise me for BT to have applied something of their own research into the pot. However, the point at which the pre-compensation gets applied to the transmitter is right in the heart of the DSP - at the point that the digital stream is converted into an analogue signal. I'm not a DSP engineer, but I'd imagine it would be hard to put something that is BT-specific into that point in the DSLAM's DSP.
An interesting rumour though...
I *can* believe that they'd have their own way of managing the vectoring process, and determining how to form vector groups. This kind of stuff might be seen as an output of an enhanced DLM process.
-
Me thinks that Kitz, WWW*t and b*cat basically share the same understanding -- to a greater or lesser degree! ;)
-
.. and very educational for the rest of us.