Kitz Forum
Broadband Related => Broadband Technology => Topic started by: kitz on August 21, 2015, 05:28:36 PM
-
Split from the G.INP thread to discuss crosstalk
I'm not sure. If they did advertise it then, I think it would make those (1/3rd?) of users on ECI's a tad unhappy.
There's no fast rule about how much speed can be regained by the use of vectoring, but all I do know is that crosstalk has affected my line by 30Mbps so there are some big improvements that it could make.
Despite being on an ECI cab myself, I dont think that BT should use that as any excuse to stop roll-out to the Huaweis.. and although I'd be green if I couldnt get it, Im not so selfish as to realise that it could be of benefit to others.
BDUK is an area where it looks like they could utilise it effectively, but saying that BDUK is strange and its not all out in the sticks.
Look at this PCP near me - On the left is an ECI cab... on the right is a BDUK Huawei.
(https://forum.kitz.co.uk/proxy.php?request=http%3A%2F%2Fwww.kitz.co.uk%2Fadsl%2Fimages%2Ffttc%2FECI_PCP_Huawei_small.jpg&hash=cb6dc266cd6a6f529e3cb61d2cd369e89138eaab)
I just wish that they could be a bit more open on whats going on with vectoring and let us know if there is something that can be done with the ECIs or not.
-----
ETA my reply was to ktz392837 - as Alec's reply wasnt there when I started to post.
-
There's no fast rule about how much speed can be regained by the use of vectoring, but all I do know is that crosstalk has affected my line by 30Mbps so there are some big improvements that it could make.
FWIW, I have gradually lost around 10Mbps or so DS since May 2012 on my 1100m line (from 30+Mbps to 20+Mbps) & around 2 Mbps US, but in percentage terms, that equates to around 33% loss.
That 10Mbps or so loss of speed is very noticeable here & anything that vectoring can do to improve matters would be most welcome indeed.
I'm on a legacy PlusNet product (40/10) & if I didn't need the US speed for transferring files over VPN when I work from home, I'd revert to the 40/2 entry level product that new users have to accept.
The reason for that is that when testing with PlusNet a few years ago, it was confirmed that lower US sync speeds (i.e. capped at 2Mbps) always resulted in higher DS sync speeds on my sub-40Mbps connection & vice-versa.
I more recently experimented by capping US sync speed directly via my HG612 modem, but that didn't have the same effect.
It also appears that removal of G.INP on US on my connection had a slightly negative effect on DS sync speed, but I don't have any other empirical evidence to state that as fact.
It could just be coincidental with a further increase of crosstalk.
-
likewise I dont want it to stop on the hauwei's as thats just been selfish, but like you said I would be annoyed if ECI didnt get it.
But unless BT raise the max speed, people like me and you kitz arent luckily losing too much from lack of vectoring, as I sync in the 70s and you at max speed, g.inp will make both our lines almost immune to DLM given we already have low ES counts. The fear really with no vectoring is possible future crosstalk.
Plus to be fair I think you are the only person who has asked the question, everyone else tbb,elreg etc. dont ask or not published the answer.
-
FWIW, I have gradually lost around 10Mbps or so DS since May 2012 on my 1100m line (from 30+Mbps to 20+Mbps) & around 2 Mbps US, but in percentage terms, that equates to around 33% loss.
That 10Mbps or so loss of speed is very noticeable here & anything that vectoring can do to improve matters would be most welcome indeed.
I'm almost at a 1/3rd loss too, and Ive noticed it on a few other lines, so I wonder if that 33% is of any relevance or just co-incidence? I think even BS is heading for a third as last time he mentioned it he said he'd lost about 20Mb.
The reason for that is that when testing with PlusNet a few years ago, it was confirmed that lower US sync speeds (i.e. capped at 2Mbps) always resulted in higher DS sync speeds on my sub-40Mbps connection & vice-versa.
I more recently experimented by capping US sync speed directly via my HG612 modem, but that didn't have the same effect.
I wonder if that's anything to do with them making some changes to the bandplans last year - in particular the U0 band. Previously that band could contain more of a mix. Since they made those changes then Ive noticed on several short lines that it leaves less for the upstream. U0 has always been very heavily masked and theres not that much available for bit loading in that particular band now :(
The other bands ie D1, D2, D3, U1 & U2 are fixed and can only carry the relevant up or down tones.. so that means even if you did transfer to 40/2 it cant shift tones in or out of designated bands.... in exactly the same way that with adsl capping upstream wont affect downstream. (obviously unless you change annex)
-
If i take my line it has crosstalk of 4dB and if vectoring removed it by 100% i would only gain 5Mbps that would take my sync rate to 40Mbps which would be just right for a 1000m line and G.INP gave 3-4 Mbps back, and on Bald_Eagles1 line G.INP made no difference.
Vectoring looks like it will help more on shorter lines than on longer lines.
-
I wonder if that's anything to do with them making some changes to the bandplans last year - in particular the U0 band.
My experiments with PlusNet were back in 2012, so perhaps not related.
I'm possibly being too cynical, but having seen the effect for myself of US sync speed being reduced at source rather than at an EU's modem, it immediately made me wonder if reducing service speeds to 40/2 is simply a ploy to increase headline DS speeds slightly as it is DS speed that is quoted when making claims of faster broadband, not the US speed.
-
Wait, so Kitz your PCP has two cabinets - one BDUK and one commercial?
-
If i take my line it has crosstalk of 4dB
Not sure what you mean there, NS.
How have you measured the level of crosstalk?
and if vectoring removed it by 100% i would only gain 5Mbps that would take my sync rate to 40Mbps which would be just right for a 1000m line and G.INP gave 3-4 Mbps back, and on Bald_Eagles1 line G.INP made no difference.
Vectoring looks like it will help more on shorter lines than on longer lines.
Bear in mind that my line is only 100m longer than yours.
If all my speed loss over the years is down to crosstalk & if vectoring removed its effect by 100%, I would regain around 10Mbps sync speed.
The activation of G.INP on my connection didn't increase speeds, but it did more or less completely eradicate errors, thus making it more stable.
I have been told by PlusNet that I am on the fastest speed profile available & I suppose the closeness of sync speeds/attainable rates confirm that.
So for my own connection, I have to hope that vectoring will increase sync speeds back to what they were & that G.INP helps it to remain stable & thus avoid unnecessary DLM intervention.
-
Wait, so Kitz your PCP has two cabinets - one BDUK and one commercial?
The 2 FTTC cabs arent connected to the same PCP. The PCP in the photo has an ECI cab.
The Huawei cab belongs to a PCP which is on the other side of the road.
-
So they have two PCPs for one street?
-
Not sure what you mean there, NS.
How have you measured the level of crosstalk?
Yes i have if my crosstalk disturber is off my SNRM of 6dB will increase to 10-11dB and if i resync at that stage the sync rate will increase by 5Mbps.
-
my attainable was 110mbit when I first unlocked modem.
I currently sync at 72mbit
It was in mid to high 60s for several monthsm before dropping to under 50mbit which triggered the pair swap that got me back up to 72.
So in my case I am over 33% loss.
-
My attainable was about 60 Mbs at day one when I was one of the first to be connected to my cabinet, it is now about 42/44Mbs so about a 25% drop. Purely academic in my case as Infinity1 is sufficient for my needs.
-
my attainable was 110mbit when I first unlocked modem.
When did the attainable rate become more important than the actual sync rate think we discussed that the attainable rate is an unreliable target
-
As Newtronstar says, attainable rate is just a load of BT nonsense really. It doesn't say anything about your performance.
-
So they have two PCPs for one street?
Suppose that road in Kitz' photograph runs due east - west, for example, and suppose Kitz was facing due south when taking the photograph, then it is perfectly feasible that the PCP, which is not visible, the one on the north pavement, along with its Huawei equipped cabinet serves an area to the north of that road and the other PCP, which is visible, the one on the south pavement, along with its ECI equipped cabinet serve an area to the south of that road.
Every area served by a PCP & Fibre cabinet pair has to have a boundary and in this particular case the PCPs & Fibre cabinets for two adjacent areas are situated on the road which delineates the two areas.
Thinking about it further, it makes even more sense that the second Fibre cabinet to be installed (the Huawei equipped, under the BDUK scheme) was placed near to the first Fibre cabinet that was first installed (the ECI equipped, under BTs own commercial roll-out scheme) for easy access to the fibre-feed to/from the Fibre head-end exchange.
-
Yes i have if my crosstalk disturber is off my SNRM of 6dB will increase to 10-11dB and if i resync at that stage the sync rate will increase by 5Mbps.
Ah. I understand now.
My increases in crosstalk have been in measurable, but smaller steps with no single disturber appearing to be a main culprit.
-
As Newtronstar says, attainable rate is just a load of BT nonsense really. It doesn't say anything about your performance.
When a connection is able to sync at the maximum rate for a service, attainable rate doesn't matter so much, other than to flag up if conditions deteriorate.
However, when a connection syncs way below the service maximum, yet attainable rate suggests a much higher sync speed is theoretically possible, it can act as evidence that there is or has been a fault that caused DLM to take aggressive action or an indicator that crosstalk has increased.
e.g. Attainable rate & sync speed on my connection used to be 30Mbps+.
Now they are much closer to 20Mbps (due to increases in crosstalk).
However, when I saw big & sudden swings in attainable rates (accompanied by SNR/SNRM swings), usually shortly followed by lower sync speeds, it was part of the proof needed to pursue a line fault that was finally repaired back in 2012 after almost 11 months of intermittent disconnections.
Maybe G.INP, had it been active back then, would have masked some of the evidence of the fault, but I'm not sure it would have been able to deal with the sudden resyncs which were actually due to a corroded joint that performed better in wet, cold weather than it did in warm, dry weather.
-
my attainable was 110mbit when I first unlocked modem.
When did the attainable rate become more important than the actual sync rate think we discussed that the attainable rate is an unreliable target
given we are talking about line deterioration from crosstalk its pretty important. Newt as an active member here I am surprised at the comment.
We are talking about line detoriation not "service" deterioration which are two different things.
So obviously since I initially synced at the full 80 I have to use the attainable to calculate the affect of crosstalk on my line.
Also the attainable can remain important at other times, e.g. someone may take a crosstalk hit but manage to stay synced at a very low margin e.g. 3db, they may think ahh no effect as I still am at the same speed, but the attainable will be lower than the sync speed.
-
To further add to this, which is where I think confusion may occur. The important fact to remember is that the max attainable rate goes skewy on some routers when interleaving and error correction has been applied.
The max attainable on my VMG8324 has always fallen in line with any drop in SNRm that has been lost through crosstalk. So for me at least when using the Zyxel I take it as a fairly reliable measurement of speed lost from crosstalk.
Those who are able to sync in the high 70s and certainly those whose lines are capable of over 80Mbps are unlikely to have Interleaving & Error Correction applied. The parameter which is possibly confusing the max attainable is the INP value.
Although the INP value when set by DLM is a static figure (ie 3, 3.5, 4, etc) you cant say yeah that INP = 3 means 'x' Mbps, because that value is applied to the sync bit-rate to calculate the level of redundancy required. It changes each time a new sync speed is attained and afaik no router ever displays the bitrate value of those overheads. Although there will be a formula to calculate those overheads, all these values are always hidden.
I suspect that if we knew what that INP value meant in terms of bit rate overheads for that line.... and was able to subtract that bit rate value from what some routers display as the max attainable, then we would have a closer figure to what the true max attainable would be if Interleaving and Error correction wasn't applied.
I made a post (http://forum.kitz.co.uk/index.php/topic,16007.msg297542/topicseen.html#msg297542) the other day about this when talking about the HG612.
As others have said the max attainable figure goes skewy if Interleaving (or more correctly Error Correction) has been applied.
We are uncertain why this is, but my guess would be that its something to do with the overheads required for FEC.
FEC overheads aren't part of the sync speed, yet the modem still has to transmit those overheads on top of line speed. If we work on the basis that FEC overheads are circa 12%* and you are syncing at 40Mbps, then in all reality you are pushing 44.8 Mbps down the line. - the extra is hidden from view but theres a couple of different methods in how that overhead can be transmitted.... and how much is subtracted from sync or sent via a reserved channel or frequency (too long of an explanation for this post). Its that overhead which seems to confuse the max calculation on some routers.
*12% is an average, but it can be more depending upon the INP value (http://www.kitz.co.uk/adsl/interleaving.htm#INP) set by DLM
But basically if your line doesnt have Interleaving, Error Correction, INP and isn't one of those modems (such as the HG612 and several more where it does skew the results), then the max attainable rate is a very good indicator of line deterioration from crosstalk. Certainly on my line I'd take it to be a fairly accurate figure. Chrys's line doesn't have INP so it would also be reasonably accurate for him too.
-
I'm splitting off this topic - because its more moved over to discussion of Vectoring and crosstalk. :)
Bear with me whilst Im juggling some posts
-
Vectoring looks like it will help more on shorter lines than on longer lines.
Yep. Vectoring should actually help those on short lines the most as those are the ones most impacted by crosstalk.
Plus its the shorter lines that have the most PCB applied to prevent their lines interfering with other longer lines.
There's a graph here (https://techzine.alcatel-lucent.com/boosting-vdsl2-bit-rates-vectoring) from Alcatel which shows improvements that can be made by using vectoring.
(https://techzine.alcatel-lucent.com/sites/default/files/wp-content/uploads/2012/10/Boosting%20VDSL2%20Bit%20Rates%20with%20Vectoring/TZ-Typical-vectoring-gains-Fig-1.jpg)
As you can see vectoring makes massive gains for lines up to 400m and some decent gains for lines up to 800-900m.
There's still some not to be sneezed at gains to be made though at 1200m. If youre only getting 20Mbps, Im sure you'd welcome a possible increase to 30Mbps.
-
I don't understand that graph as it seems to suggest the maximum speed at 100m is only 60Mbps...
-
Its showing the impact how far crosstalk could reduce the line to.
Maximum speed without any interference is the green line.
-
I seem to be very lucky as my line isn't affected by crosstalk (much) at all :)
-
fair point on INP kitz, and as you said my 110 was taken with fastpath.
All the figures I put in the post were fastpath also, I didnt include any banded or interleaved sync speeds.
-
Can crosstalk happen in the cable from DP to NTE5 or is it normally just in the cab?
-
given we are talking about line deterioration from crosstalk its pretty important. Newt as an active member here I am surprised at the comment.
I was making reference to a line which has been interleaved with error correction and most of the time that's what my line was on and the attainable rate just gave me false readings, but since G.INP has been enabled the Sync and Attainable are evenly matched.
When looking at my attainable before and after G.INP i can't see any relationship that shows up deterioration from crosstalk on this line, all i can see from past events when disturber is off the SNRM will increase and the Attainable will increase and after a reboot the sync rate will increase but never matched the Attainable, but in saying that have not been able to test that out on G.INP.
-
Can crosstalk happen in the cable from DP to NTE5 or is it normally just in the cab?
It can happen . . . if it is Drop Cable 10 (containing two pairs) and both pairs are carrying separate xDSL services.
For the more normal users -- those only taking one VDSL2 service -- any cross-talk would occur in the multipair cable(s) between the (fibre) cabinet and the DP.
-
Its showing the impact how far crosstalk could reduce the line to.
Maximum speed without any interference is the green line.
In terms of giving you an idea of how much crosstalk can really affect a line, I like the graph from the Broadband Forum's MR-257
https://www.broadband-forum.org/marketing/download/mktgdocs/MR-257.pdf
(See figure 6 on page 12, attached below).
This shows a similar picture to Kitz' picture, but adds a few interesting details
- The upper and lower triangles show the most extreme range between "no other users" to "maximum crosstalk ever possible"
- The blue crosses show the range of actual speeds for 80 users causing crosstalk
- The red circles show the range of actual speeds when vectoring is activated.
At 350m, these details correspond to:
- Absolute best: 140Mbps
- Absolute worst: 45Mbps
- Actual results with crosstalk range between 65Mbps and 90Mbps
- Actual results with vectoring range between 130Mbps and 140Mbps.
These actual results don't apply in the UK (it is for 0.4mm cable, and likely doesn't include the power masks we use to keep compatibility with exchange-based services), but it gives you a clear idea of the potential impact of crosstalk if subscriber numbers get too high: losing two-thirds of the theoretical capacity!
Incidentally, I chose the distance for the example at 350m, because one property I had service at was at this length.
BT currently gives a range A prediction of 62-80Mbps; in reality our attainable started at around 90Mbps, and gradually dropped (along with the actual speed) to 78Mbps.
The picture, in the end, agrees that crosstalk can account for losses of around one-quarter to one-third of speed. And that doesn't include the effect of DLM adding error protection.
-
Thank you for the graph. Its quite scary just how badly crosstalk can affect a line by up to 2/3rds.
I think I made a post just the other day saying that from stats Ive seen on here over time most have lost 1/3rd.
I'll go with your 350m example. Im on what Im reliably informed is 385m of .5mm copper.
I wasnt first on the cab (port 25) as for various reasons I delayed a few weeks. At first my max attainable was 108 Mbps, but it soon dropped to the 90's. Now there's only 2 modems which allow me the full 80Mbps.
I have lost circa a third of headline so gawd knows how much it could have been if I'd gotten vdsl earlier. iirc initial prediction was "up to 62Mbps" but that was before they started doing the clean/impacted.