Print

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

[Weaver]

I just somehow missed that point that Black Sheep raised.

I'm assuming that aluminium has a bad reputation because of its nonlinearity when it corrodes.

The number of twists per unit length is presumably linked to the wavelength of the disturbed signal. This wavelength is huge for DSL. I presume BT has no problem easily meeting the requirement here. [ But I'm rapidly getting out of my depth here and will defer to others who have some expertise of what's actually true as opposed to what sounds like it ought to be true. :-) ]

[general disquiet]

Some really helpful responses here, that are giving me much food for thought.  It seems that my original questions may not be as straightforward to answer as I had imagined!

[general disquiet]

@ licquorice - your mention of dodgy joints in poor aluminium may be the key to the problem.  Conventional wisdom, at least as it applies to good quality copper lines,  seems to be saying that there shouldn't be any crosstalk between VDSL and ADSL that causes problems to the ADSL signal (though there can be issues the other way round).

Now, if you have a long aluminium line with one or more poor joints acting as diodes, then there is a chance that each of these diodes/joints is going to act as a simple, unbalanced frequency mixer - producing output frequencies that are the sum and the difference between the ADSL and VDSL input frequencies in addition to the two original frequencies.  The more diodes acting on the circuit, then the more horrible the mix of frequencies is going to become and the more likely that the ADSL signal will be adversely impacted.

[Weaver]

@general disquiet  - I think your logic is sound about IM distortion from a nonlinearity. But...

We should do the actual arithmetic, to check the difference frequency and see where that lies. I don't know the numbers off-hand.

[general disquiet]

ADSL2 Downstream runs from 0.14MHz up to 1.1 MHz.  ADSL2+ runs from 1.1MHz to 2.2MHz and VDSL runs from 2.2MHz all the way up to 12MHz. Simplistically (I cant do more than simple!), the passive intermodulation caused by the grotty aluminium will produce additive and subtractive frequencies.  So, if we focus on frequencies at the top of the ADSL frequency range (1.1MHz) and the bottom of the VDSL frequency range (2.2MHz) then the new lower frequency will be at 1.1MHz. To me, that looks as if even a single dodgy join on an aluminium cable might result in detrimental crosstalk from VDSL to downstream ADSL.  Add in a second jointfault (ie two diodes) and the effects could be far worse - in fact, you might get a range of new frequencies generated between 0MHz and 2.2MHz, causing VDSL crosstalk to not only the whole of the ADSL2+ and ADSL2 downstream frequencies, but also to ADSL2 upstream and even POTS.

[Weaver]

With just one non-linearity we could get 2f₂ - f₁ for example, and all the other combinations involving harmonics.

[WWWombat]

ADSL2 Downstream runs from 0.14MHz up to 1.1 MHz.  ADSL2+ runs from 1.1MHz to 2.2MHz and VDSL runs from 2.2MHz all the way up to 12MHz.

All three start at 0.128MHzso they fully overlap in the lower spectrum. VDSL2 on the 17a profile, as used by BT, goes up to 17.6MHz. VDSL2 could go up to 30MHz (but BT doesn't use that) and a new version is coming out that goes up to 35MHz (there's no indication that BT are considering it).

To me, that looks as if even a single dodgy join on an aluminium cable might result in detrimental crosstalk from VDSL to downstream ADSL.

I'm a simple digital creature, not analogue. I can see the arguments that a dodgy joint can cause detrimental signals on that particular cable; with effects on the capacity of that pair.

However, an effect on crosstalk requires some further steps, doesn't it? Some things related to the power of the detrimental signal, a question about which part of the line this detrimental signal occurs on, and the proximity of other pairs in the cable.

I don't think you'll find a nice linear relationship that leads from a dodgy aluminium joint into speed drops (to one third) on all lines in the village.

Has anyone verified whether the speed drops are actual changes in sync speed, or merely in throughput?

[Weaver]

> However, an effect on crosstalk requires some further steps, doesn't it? Some things related to the power of the detrimental signal, a question about which part of the line this detrimental signal occurs on, and the proximity of other pairs in the cable.

I absolutely agree. There are a lot of ifs. The IM products would (surely?) have quite a lot lower amplitude, so that they would be less significant as disturbers in any possible crosstalk.

[general disquiet]

@WWWombat - I would love to be able to answer your question regarding whether there was an issue with falling IP profiles or merely throughput.  Unfortunately, the BT Speedester hasn't worked on any line from the village, as far as I know, for months.  It hasn't even been registering downstream throughput - often returning a download speed of 0 Kbps - let alone getting to the second stage of the test that returns the IP profile.

I will speculate that the IP profile on my own line dropped from 2 to 1Mbps in October and has returned to 2Mbps in the last few weeks.  Massive TCP pauses both downstream and upstream (see http://www.connectionanalyzer.com/home.php reduced the throughput to around a third of this.  Perhaps 50% of lines in the village experienced something similar.  The other 50% seemed to take an even more massive hit to the IP profile, with a few poor unfortunates losing their service completely - their routers simply can no longer "see" the exchange and sync with it.