Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[asbokid]

Hello Tristan,

The RF interference from AM broadcast radio looks dramatic, but when quantified by the number of tones affected by the ingress, it probably doesn't impact the overall DSL performance that badly.

Channel bandwidth in Europe for AM broadcast radio is only 9kHz, meaning that ingress from an AM radio station should only impact two or three DSL subcarriers.

Using a re-purposed wavelet-based peak-det algorithm:  [1]

Code: [Select]

$ R
library(MassSpecWavelet)
# Loading required package: waveslim
qlnDAT <- read.table("qln.txt", colClasses = c(rep("NULL",1), rep("numeric",1)), header=TRUE)
qlnMAT <- data.matrix(qlnDAT)
wCoefs <- cwt(qlnMAT, scales=seq(1,31,1), wavelet='mexh')
localMax <- getLocalMaximumCWT(wCoefs)
ridgeList <- getRidge(localMax)
majorPeakInfo <- identifyMajorPeaks(qlnMAT, nearbyPeak=TRUE, ridgeList, wCoefs, SNR.Th=1)
peakIndex <- majorPeakInfo$peakIndex
#

The tones notably affected by AM broadcast ingress (and other RFI) are the following.  From a total of some 4000 tones, it's not that many, really.   The line attenuation is the real service killer...

Code: [Select]

peakIndex

   127    167    188    211    238    253    282    311    451    562    763
   777    858   1216   1324   1391   1405   1430   1506   1604   1687   1704
  1736   1791   1885   1962   2799   2819   2913   2970   3006   3099   3192
  3286   3324   3380   3433   3472   3537   3567   3763   3959

And plotted..

Code: [Select]

plotPeak(qlnMAT, peakIndex, main=paste("identified QLN peaks"))


Suggesting that the real problem, as Paul has already said, is the high attenuation on the line. A consequence, as you mentioned before, of suffering a loop of considerable length. An unfortunate property of the twisted pair. As such, even if every joint was re-made between your home and the cabinet, it still might not result in an improvement.

What can be done?  It's optimistic to hope that any more tweaking by Openreach is going to improve the performance - short of running fibre to your doorstep. As such, I would be tempted to ditch DSL and look for another solution.

What about establishing a private point-to-point microwave link from your community to the nearest town?  It sounds very expensive or complicated but it's neither.  Just a few hundred pounds and an afternoon's work can result in a surprisingly good job.

There is a specialist firm, Stella Doradus, in Waterford, [2] who sell all the kit needed. Helpful people too   With a pair of high gain (22dBi) parabolic dishes and good quality transceivers, it is possible to form a decent microwave link of some miles. With the right antennas, a point-to-point wi-fi bridge of 25 miles even using domestic 802.11 kit, isn't unheard of 

cheers, a


[1] http://forum.kitz.co.uk/index.php/topic,11503.msg221918.html#msg221918
[2] http://www.stelladoradus.com/

[burakkucat]

As far back as I can remember, we have always been known professionally as 'Technicians'.

It goes back to the days before Black Sheep even existed as a little Black Lamb . . . back to the days when the General Post Office provided the nation's telephone service. I first heard the descriptor "GPO Telephone Engineer" used in the 1950s.

Before all-figure numbering was introduced, when one wished to report a fault with the telephone service one would dial ENG and the call would then be answered at the fault desk of the local exchange -- usually with the one word: "Engineers". Immediately after all-figure numbering had been introduced, the number 151 was used as the replacement for the ENG code.

[Black Sheep]

Interesting stuff, BK. Wasn't aware of that.

[NewtronStar]

Would good coaxial sheilding help provent AM interference on the BT copper wire infrustucture ? 

I say this as the 1970's & 80's phone lines was not designed for high frequencys (Fibre)

[burakkucat]

In theory a screened twisted pair (with the screening earthed at both ends) would be a great improvement. 

In practice such a configuration would never be installed. 

[c6em]

It would indeed be interesting to know how much of the typical MW radio ingress into the line is due to the general line length E sides & D sides (underground) and how much is due to the last 50 yards of unshielded overhead drop wire.

Incidentally the specification for lay of the twists in std BT black sheathed drop wire is one twist every 89mm in the case of the orange/white pair and every 95mm in the case of the green/black pair.

[Black Sheep]

Nice one c6em, I didn't know that. May I ask where you get that info from, for my own personal use ?

[c6em]

I first noted it from an un-recorded source and added to my personal knowlege store!
Then I saw it listed more recently with the same figures in a pdf document of a cable spec from B3 Cables Solutions - External 2 pair dropwire - complying with BT spec CW 1411
Finally I have taken to pieces a decades old gash/scrap section of dropwire and confirmed the figures as being correct.  I was doing this just to check whether the very early black dropwire stuff was indeed twisted pair.

[Black Sheep]

Thanks c6em, and  at you disecting an old DW to prove it !! Like yer style.

[tristan]

I have been looking through your latest stats & a few of them look "strange" when compared against any other VDSL2 stats that I have looked at over almost a year now.

Thanks Ballk_Eagle1. This is most useful. I'm going to pass this on to the chap who's coming out tomorrow morning to take yet another look.

It really is quite frustrating - a new person comes each time, re-runs the same tests, says they can't understand what's going on, then tells us it needs to be escalated. When he leaves the max download will be ~5MB and the actual >4MB (due to the reset) so he goes away thinking they have made an improvement. Then overnight/the next few days it drops back down necessitating a more urgent next visit!

I'll will try and run those scripts you mentioned on a Mac that I have remote access to (how I run vdslcmd from here in London) and see if there's anything interesting.

What a saga!

[tristan]

Tristan quoted ...................

"I get the impression the people called out are 'installers' rather than engineers".

As a qualified Electrical Engineer, I can understand your comments fully. But for pedants sake, the term 'Engineer' has been banded about for eon's when referring to 'us' Openreach guys. God only knows who or how it became so, but I can tell you that it wasn't from BT. As far back as I can remember, we have always been known professionally as 'Technicians'.

Only when I was on the 'Power & BES' group was I referred to as a 'Power engineer'. On this side of the fence, dependant on skill set, we are called Technician 2B (T2B), Technician 2A (T2A), and the following are grades that are still tagged to individuals, but are now obsolete and will be lost as and when the individuals retire. They are Technician 1 (T1), Senior Technician (ST) and Technical Officer (TO).

As you will see, nowhere in our descrip do we refer to ourselves as 'engineers'. So the term is protected as far as BT are concerned.

That's interesting. Thanks for the input. Who would be the best person to get this escalated to? A TO? I'm not sure who has been coming out but as they are on their own they can't check any individual stretch of line for us (they need 3 people I think to operate the cherry picker?) - each says it needs escalated so I guess we are gradually working our way up the ranks!

[tristan]

The RF interference from AM broadcast radio looks dramatic, but when quantified by the number of tones affected by the ingress, it probably doesn't impact the overall DSL performance that badly.

Channel bandwidth in Europe for AM broadcast radio is only 9kHz, meaning that ingress from an AM radio station should only impact two or three DSL subcarriers.

Using a re-purposed wavelet-based peak-det algorithm:
The tones notably affected by AM broadcast ingress (and other RFI) are the following.  From a total of some 4000 tones, it's not that many, really.   The line attenuation is the real service killer...

Suggesting that the real problem, as Paul has already said, is the high attenuation on the line. A consequence, as you mentioned before, of suffering a loop of considerable length. An unfortunate property of the twisted pair. As such, even if every joint was re-made between your home and the cabinet, it still might not result in an improvement.

What can be done?  It's optimistic to hope that any more tweaking by Openreach is going to improve the performance - short of running fibre to your doorstep. As such, I would be tempted to ditch DSL and look for another solution.

What about establishing a private point-to-point microwave link from your community to the nearest town?  It sounds very expensive or complicated but it's neither.  Just a few hundred pounds and an afternoon's work can result in a surprisingly good job.

There is a specialist firm, Stella Doradus, in Waterford, [2] who sell all the kit needed. Helpful people too   With a pair of high gain (22dBi) parabolic dishes and good quality transceivers, it is possible to form a decent microwave link of some miles. With the right antennas, a point-to-point wi-fi bridge of 25 miles even using domestic 802.11 kit, isn't unheard of 

Thanks so much for this. I might try and replicate your work, just out of interest, and fire up R - not used it much and more accustomed to C/Fortran/Matlab.

As I wrote above another 'technician' (;-) ) is coming out tomorrow. I'll report all the help you guys have provided and hope that the guy sent out is suitably technical and willing to have a quick read to see if he agrees; the last guy was the best so far but maybe didn't have the level of knowledge contained within this forum. He did have some tales of broadband not working for folk and it taking weeks to work out that it was due to an electric fence being nailed to a cross-field pole. This has a high likelihood of occurring where we are!

As I see it, we have three issues that require checking/explaining:

1) The attenuation, if it can't be lowered by the discovery of any obscure issue along the line, is as you say the crux of it. OpenReach tell us it's >3.5km long, but our measurement makes it just over 2km. It is almost entirely above ground so we can follow it from the cabinet along the side of the road and to our house. Would there be any possibility of length reduction? Excess cable taken out? Checking that our line isn't doing any silly routing.

2) Why did the connection give a good stable 4mb (good enough for us) for a period of time before getting awful? Is it usual for the modem to initially sync at 4 or 5mb if the line can't sustain it? ie, if it did it once why can't it remain at that level?

3) The modem. This is the one bit of kit they haven't replaced in part, I think, due to the fact that the OpenReach diagnostic machine can't sync at all! Maybe a lightning strike frazzled something inside.

Regarding solutions, I agree - a microwave link would be a good one. We'd have to find a willing volunteer to host the base! I wonder if BT would allow us to nail it to the pole next to the cabinet! I'll ring those guys tomorrow and also google for suppliers closer by to see if any of the farms near us already have such a setup we could get involved with (we have one of the most advanced milking parlours in the UK nearby and I guess they would have a decent internet connection...).

The other, albeit slim, option might be to switch to an alternative line we had activated for ISDN some years ago (for a "civic" job posting, so we had no dealings with the management of it) that goes to a different exchange - the line goes left at the bottom of the drive rather than right... This would still be a long way from the cabinet but maybe just good enough. Were there similar length/attenuation specs for ISDN? This could be why the installation was done in that manner... We'll ask tomorrow for the guy to maybe check it if possible.

Thanks again,

Tristan

[tristan]

It would indeed be interesting to know how much of the typical MW radio ingress into the line is due to the general line length E sides & D sides (underground) and how much is due to the last 50 yards of unshielded overhead drop wire.

Incidentally the specification for lay of the twists in std BT black sheathed drop wire is one twist every 89mm in the case of the orange/white pair and every 95mm in the case of the green/black pair.

So for the shortest possible path from cabinet -> home you would want green/black used? Although the difference would be small it could be significant in my case!

[tristan]

While I'm on a splurge of posting, attached is the temporal variation of our reported speeds since ~10th July.

Each point in green above 5000 bps is when an engineer came out and requested a reset on the line. The purple line shows a somewhat consistent decrease over time after each visit. Our upload is rock solid! Always seeking a positive, me. :-)

[Black Sheep]

It would indeed be interesting to know how much of the typical MW radio ingress into the line is due to the general line length E sides & D sides (underground) and how much is due to the last 50 yards of unshielded overhead drop wire.

Incidentally the specification for lay of the twists in std BT black sheathed drop wire is one twist every 89mm in the case of the orange/white pair and every 95mm in the case of the green/black pair.

So for the shortest possible path from cabinet -> home you would want green/black used? Although the difference would be small it could be significant in my case!

Unfortunately, it wouldn't make any noticeable difference to place you on the 'Green/Black', as it would only apply to the span from your house to the telegraph pole. The 'Green/Black' doesn't go all the way back to the PCP I'm afraid.