The astonishing impact of a home wiring "bridged tap"

I'm currently decorating my study where the VDSL router is installed and have had to, temporarily move it to the hall where the master socket is installed. The master socket is equipped with a VDSL MK2 faceplate. From that I run a (filtered) phone extension and a single pair (unfiltered) VDSL feed to the study using CAT5 cabling. The length of this cable is less than 10 metres. I had been used to getting a reliable DS sync speed of about 58-59mbps, despite being 650 metres from the cabinet (as measured by me using GPS).

I figured when plugging in the router into the hall master VDSL unfiltered socket I'd suffer some loss of speed as the study extension forms what's called a "bridged tap" - that is an unterminated pair as a T off the main path. What I had not expected was that it would lose me fully 25mbps downstream and almost 3mbps upstream (out of 10mbps). That is simply astonishing and if anything demonstrates how sensitive VDSL is to the quality of domestic extension wiring it is this. I figure if I put a balanced termination on the extension it might help, but I simply disconnected the unfiltered extension pair and got back my original speed (in fact, as you'd expect, a fraction more as I'd shortened the cable run to the cabinet by about 10m).

These are the stats from the master without the extension (similar in the study extension).


6. Data rate: 10470 / 59987
7. Maximum data rate: 10496 / 60012
8. Noise margin: 6.2 / 6.0
9. Line attenuation: 30.1 / 21.3
10. Signal attenuation: 29.9 / 20.5

This is what happens when operating from the master with that <10m of CAT5 pair extension to the study.

6. Data rate: 7558 / 33144
7. Maximum data rate: 7562 / 33922
8. Noise margin: 5.7 / 6.4
9. Line attenuation: 32.9 / 21.4
10. Signal attenuation: 21.0 / 18.4

That's almost a 9dB increase in downstream signal attenuation from less than 10m of CAT5 pair leading to a 45% drop in DS speed. OK, I know some of the theory of signal reflected off of unterminated transmission lines, but even so...

If you run DSL stats with the router in both location and look at the Tones graphs for each you may well see a large notch or series of notches - probably around 7.5MHz and 15 MHz. 1 metres of cable is a 1/4 wave at 7.5 Mhz give or take and will slightly different as it is copper and not free space. The signal goes down the cable and reflected back to form a full invesrion and thus cancel the signal. Away from those frequencies it will appear as "noise".

What do the graphs actually show?

Unfortunately it's an HH5 so it's not possible. I was aware of the theory (i recall solving equations for transmission lines in my physics studies almost 50 years ago). However, it's one thing doing the theory and a second seeing just what a huge impact it had. It does make me wonder just how many people's VDSL performance is hit by this sort of thing as public awareness is not exactly high. I've always been surprised that ISPs (and especially BT Retail) don't include a leaflet on optimising home wiring in an easy to understand form. It's not so difficult - you want your VDSL modem at the end of the unfiltered loop with no unfiltered extensions.

I did have an idea that you may already know about it however decided to post to give simplistic explanation to others. The problem was always there on ADSL although the tap needed to be nearer 50m to 100m and also on ISDN2 services.

I agree that it is likely that there are plenty of installs with which are not optimum, however what would be the cost of actually improving them and who would pay? Quite a few are down to the users' own installation. I have managed to spend several hours with the result that one connection is now performing at four times the original speed - that cost a friend a few pints! However, would a normal residential customer be prepared to pay £400 for that ? probably not.

The problem is there from DC at "Zero Hz" and upwards.

Consider a simple DC circuit such as a battery feeding a load.

Maximum power transfer occurs when the load equals the internal resistance of the battery; and the potential voltage of the battery is shared equally over that internal resistance and the load.

With a typical battery potential of 1.5V, at maximum power transfer, the load has 0.75 Volts across it, the other 0.75 Volts being apparently "lost" across the internal resistance of the battery.


Now disconnect the load totally, no power out, so the terminal voltage of the battery rises to "equal" the potential voltage of 1.5V


WARNING - DO NOT DO THIS AT HOME.
Now short-circuit the battery, say by replacing the original load with a simple piece of wire.

The voltage across that short-circuit wire is 0 V (zero volts), with the potential voltage of the battery being totally lost inside the battery on its internal resistance.

------

Analyse the three conditions, assuming the DC to be an extremely low frequency AC.

The pattern fits, as does testing at intermediate loads, from open circuit down through matching load to short-circuit., similar to partial reflections.

-----

Given the lengths/distances of the National Grid, similar can happen on that; and is also part of the reason for the Import/Export Interconnectors handling DC only, as well as the more obvious Frequency and Phasing aspects.