TBB FTTC Speed Estimate exchange-cabinet distance

I was wondering if the TBB FTTC speed estimation process for superfast coverage takes into account the exchange-to-cabinet distance as, despite what is commonly quoted it's not just the cabinet-to-property distance that matters (at least in the UK).

This may seem odd, but the reason is in the detail of the ANFP (Access Network Frequency Plan).

The purpose of the ANFP is to permit the co-existence of multiple services over the phone network so that they don't unduly interfere with each other. One important part of the ANFP is to allow VDSL and ADSL2+ to co-exist on the same network. This is achieved by controlling the power of signals at the cabinet from "swamping" the weaker ADSL signals from the exchange by excessive cross-talk using what's called a PSD mask. Of course the simplest way to achieve this would be to eliminate any frequency overlap at all. However, what the ANFP does is much more subtle. It allows for increasing of ADSL2+ frequencies as they become attenuated to the point of being useless according to the cabinet's distance from the exchange. Interestingly this means that cabinets further from an exchange in rural areas should be able to offer service over a longer range than would otherwise be the case as it's precisely those lower ADSL2+ spectrum frequencies (in the 0.1-2.2 Mhz region) that travel furthest.

See the ANFP and particularly the section on PSD and what's called the Cabinet Assigned Loss or CAL, which is a characterisation of each cabinet position according to the electrical characteristics of it's position with relation the the exchange.

http://www.niccstandards.org.uk/files/current/ND1405...

I noticed this on moving house recently. From my previous cabinet about 4km from the exchange and 640m distant from hone, I could get 58mb downstream and just below 10mb upstream. At the new house, about 900m from the exchange and a cabinet about half way, the maximum data rate downstream is again about 58mb, yet upstream the maximum data rate is 22mbps (although capped to 20mbps of course). The reason for this can only be down to the virtual elimination of any ADSL2+ frequencies from the downstream PSD in the latter case as, at 400m from the exchange, all those frequencies are usable so must be protected. In contrast, at my previous cabinet, much further from the cabinet, some of those ADSL2+ frequencies were so attenuated as to be useless and were hence reused by the VDSL service.

Unfortunately I don't have a router which can give me the bit loading profile, but it's wholly consistent with the theory.

What this also means is that it's rather difficult to build in maps of expected speed by using solely cabinet to consumer distances. The distance from the exchange must also be taken into account. I wonder if ThinkBroadband's assessments have taken this into acount?

The CAL value is not soley dependant on the distance of the pcp but specifically the e side cable loss which is dependant on cable length and guage.
Openreach are required to provide a CAL to CP's for SLU, and is one particular reason why multiple pcp's cannot be linked to single FTTC cabinets.
As you appear to have discovered, this has more of an impact for cabinets that are furthest from the exchange.

Edited by deleted (Mon 19-Oct-15 10:32:10)

It is not coded in, but a near worst case crosstalk scenario with tweaks from observing a collection of FTTC speed tests where users provided location data too helped to shape the chart originally, so unless we picked cabinets of a single profile by some random fluke it is built in.

When comparing sampled areas against observed speed tests and what BT Wholesale reckon I am more than happy with what we estimate, which generally falls on the low side, i.e. much more in line with what the general consensus from campaigners would like to be shown.

In short the effort to work out the ANFP mask for 71,000 cabinets is actually better spent on trying to identify things like infill going on in Northern Ireland.

If there was a team of 20 people working on the data, then I could afford to set 3 people aside on the task, but with a team of 1 that is impossible.

I did hint at that in the fourth paragraph where I mentioned the electrical characteristics of the line to the exchange. It's just difficult to fit in a summary.

That was what I suspected. It's clearly a complicated area and would hugely increase the complexity of the algorithm. Also, as is pointed out by somebody else, the CAL doesn't just depend on distance but on other line characteristics on the E side. Presumably different gauge copper might make a difference for instance.

However, I think it's important to note as it could go a long way to explaining some anomalous results which some people experience (like the fact my current house can do 58/10 and the new one is saying 58/22). Some appear to be getting much lower download speeds than might be expected from the attenuation and possibly this is due to the cabinet being close to the exchange. Conversely some in more remote locations may get better reach than expected if the cabinet/exchange distance is long.

Perhaps of more relevance when dealing with forum queries than anything else. I think I might pick up a router that can give me the bit loadings so I can at least check the accuracy of this speculation.

In any event, it's clear that it's not only the electrical characteristics of the line connecting a property to the cabinet that determines achievable speed.

I think I worked out that the worst impact is for medium CAL values, compared to either extreme.

I'm intrigued about the major difference to the upstream, when there has been little difference in the downstream.

Do you have any "--stats" output?