Exchange to cabinet distance - impact on VDSL2 speeds

Have just run some initial figures, to look at how the cabinet distance impacts the throughput speeds people experience with VDSL2.

There is a potential difference due to how the PSD masks operator to ensure that VDSL2 does not cause problems for ADSL/ADSL2+ signals. Did try and find a straight forward description of the distances involved, but seems to be buried under lots of mathematics and problems about not knowing the gauge of cu or al involved i.e. how low will the ADSL/ADSL2+ signal be that needs protecting. Therefore have split cabinets into three categories

SHORT - 1000m or less to the exchange
MEDIUM - 1001 to 3000m to the exchange
LONG - over 3000m to the exchange

The results are:

SHORT median download 31.9 Mbps, median upload 6.9 Mbps (48% of sample)
MEDIUM median download 30.2 Mbps, median upload 6.7 Mbps (38% of sample)
LONG median download 29.6 Mbps, median upload 6.2 Mbps (14% of sample)

One problem with this analysis is that I suspect there is a tendency for long cabinets from the exchange to also be cabinets that serve more spread out areas.

A solution to this would be to add a further filter and only include speed tests that are with 500m of the cabinet too, so there should be a fairly similar distance profile on the VDSL2 component.

I think your assumption about cabs with long E sides implying long D sides too is incorrect.

This just being based on my experience, I guess we�ll have to agree to differ.

Also conductor size and type can vary between different E side cable ranges to the cab. So for a more valid result you would need to assess a few users off the same PCP and also know what E side pair they were on too, and what the make up of that E side cable was .....

A toothsome problem is what my Father would have called that.

The figures make no assumption on long E sides with long D sides

I said "One problem with this analysis is that I suspect there is a tendency for long cabinets from the exchange to also be cabinets that serve more spread out areas."

The key word was suspect.

As you have pointed out ultimately this requires masses more data, and verification of that data, hence a high cost if you want to prove beyond any doubt, but am simply trying to get an idea of the level of impact the different power masks may have.

If the consensus is don't bother because being 90% confident is no good, you need to be 99.9% confident then I won't bother and am happy to leave people to speak about their feelings on the matter.

I'm not convinced that the small variations you have recorded are statistically significant. The other thread implied that distance from cabinet to exchange had a significant effect on upstream connection speed. As I also said on that thread, I am firmly in the 3000m+ bracket for distance to the exchange, about 450m from the cabinet and have never had less than 20Mbps upstream sync and less than 18-19Mbps on speed tests.

That was my thoughts too on the not a major difference, and just attempting to bring some numbers to the table on the subject.

Considering doing some more filtering

i.e. splitting ECI and Huawei into the lengths and excluding the vectored cabinets. We know ECI vs Huawei has a difference.

Its just a few more lines of code.

How do you know the lengths from the cabinets?

Also not all people close to cabinets will be on 80/20 connections. That will lower the near-to-cabinet averages.

Then there are quite a lot further than your suggested 500m from the cabinet, such as myself. 600m walked years ago and 0.6km on the subsequent installing engineer's JDSU. I was rather pleased with my measurement!

Wired test a few moments ago.

Already stated that the figures in the other thread were greatly exaggerated to make a point. I'll edit that seeing as some people only read half the thread

Also it will only affect those on cabinets a long distance from the exchange AND a distance from the cabinet.

The fact you have full sync means nothing. Nobody ever suggested this affects every single line on cabinets that are far from the exchanges.

Anyone who wants to look further in to this read up on Spectral Power Management, PSD masks and upbo/dpbo
Openreach use Annex B Plan 998ADE17 aka Plan B8-11

Further reading up on this it affects the bands D1, U1 & U2.

It looks (not sure) like only D1 uses PSD masks that vary by cab distance from the exchange.
The U1 & U2 PSD masks look like they (again not sure) vary by line length only.

This would match with Andrews figures above.

If that's the case then I got it backwards and it's the downstream that would be impacted by PSD masks for those on cabinets a large distance from the exchange.

Edited by j0hn83 (Sat 25-Aug-18 17:36:57)

Andrew

The analysis looks good, the Cab distances may differ a bit as a Cab further from the Exchange will tend to have a heavier weight gauge.(Up to 0.9 is common on distant cabs)

I once had to do a write up on why voice service stopped working when a customer had FTTC provided due to the ties being 0.5cu between the cabs putting the voice out of distance. ( the Cab -Exchange was the biggest BT uses just to get voice to work!)
The write up included the recommendation to planners that cab-cab ties should NEVER be less than the gauge of the Exchange to Cab gauge. ( However being in voice TSO not OR and not being able to affect OR policy meant I never found out if this was adhered to!).

Never come across 0.9 FTTC ties ...

The 500m was muted not actually used in these figures.

Don't know the measured open reach line lengths (not always sure Openreach does), but we do have our model for line lengths

On the product choice yes am aware, very aware in fact from all the product split analysis we do