Gibtelecom ends ADSL service

As an occasional Gibraltar visitor I was interested to see that Gibtelcom will soon be moving to VDSL only for broadband.

The interviewer makes the usual megabit/byte error;

http://www.gbc.gi/news/gibtelecom-upgrading-broadban...

If removing ADSL from the network totally then it allows VDSL to use the ADSL frequency without power mask restrictions, so you get the benefits at short distance, but at longer distance you get ADSL speeds

Running both together means a comprise, alas in the UK that would mean killing LLU ADSL2+ and the two big suppliers don't seem to be volunteering that

Edited by MrSaffron (Thu 26-Jan-17 09:51:49)

In reply to a post by MrSaffron:

... allows VDSL to use the ADSL frequency without restriction on power levels, ...

A minor correction as VDSL currently uses the full spectrum from 30kHz up to 17MHz albeit with power masks - which I am sure you know.

The improvement is welcomed

Looking at the speed people get at the longer line lengths, it does very much look as if very few bits in the bins at the low end where ADSL2+ lives get much use.

There are of course three different power masks, anyone know at what distance from the exchange these take effect? Asking as can then do something like look at the Huawei cabinets and split them into the three power masks to see if any difference.

Certainly removing the power masks will help those on the very long lines where most of the tones in use are within the masked area.

I have some friends with a line well in excess of 9km and they are around 2.8 or maybe 3km from the cabinet. VDSL speeds are around 1.5Mbps - which is way better than ADSL. I believe that all of their usable tones are within the masked area - up to around 1.1MHz and that the mask is very harsh. Remove the mask and they would see a possible 50% or maybe more, increase in overall speeds.

BT needs to look at locations such as theirs - which is just outside Maidenhead and consider a move similar to GibTel. Move all ADSL customers in that cabinet to VDSL even if it mean brining in a couple of packages such as 10/1 and 20/1.6 so customers can continue paying ADSL rates and then removing the power masks.

Higher power at that low end will all greater bits/tone and thus significant speed improvements.

LR-VDSL trials are essentially the work in progress on this, and trials of that have avoided exchanges with LLU at them for a reason

But, as far as I understand, ADSL will still be running through the cabinets and thus power masking will occur at the lower end. Increasing the power above that will benefit a large number, but not all and in the case I mention, I doubt if there would be a significant improvement.

Some of the LR-VDSL areas were overlapping ADSL, but at least one was an exchange that did not have ADSL at all.

In reply to a post by MrSaffron:

There are of course three different power masks, anyone know at what distance from the exchange these take effect? Asking as can then do something like look at the Huawei cabinets and split them into the three power masks to see if any difference.

Actually, there are more masks than that. There is one separate mask for every valid value of "CAL" (Cabinet assigned loss).

NICC's ND1602 (found here) specifies the mask in appendix B. Unfortunately, it uses a table and a few formulae to actually specify the mask, so it is hard to see.

It then goes on to say:

In reference to www.niccstandards.org.uk/files/current/ND1602V6.1.1.zip?type=zip:

The reason for the very long list of frequencies in column one is that the entered values ensure the functions are calculated at all the critical intersection points of the columns for a wide range of values of CAL (0 to 52 dB in steps of 2 dB).

I take that to mean there are 27 different masks.

The graph in the document shows what the mask looks like for 6 of the 27 masks.

CAL (Cabinet Assigned Loss) is the same as the insertion loss measured at 300kHz ... which is the same thing that is seen as attenuation back in plain ADSL.

For BT's standard 0.5mm copper, that IL is 10dB per km. So a CAL of 10 means the cabinet is roughly 1km from the exchange, line length. Stepping CAL by 2dB means there is a new mask for every 200m or so.

Of course, we know that cables tend to leave the exchange in a narrower gauge, to fit into the ducts. As they start to separate out in different directions, they are jointed to wider gauge cables. Physical distances, therefore, will tend to be a little less.

What impact does the mask have on speed?

I've always maintained that the mask has the biggest impact on the middle values of CAL, say values 20-30.

Looking at the graph above, if CAL=0 (ie cabinet within 100m of the exchange), there is no extra masking, so no slowdown. If CAL=50, the mask only applies to the 200-500kHz spectrum; the other 1.7MHz isn't masked ... so slowdown is minimised.

I recently found an old, but interesting, article on "The art of Spectral Management
Downstream power back-off for VDSL2". This describes what and why we have the PSD masks, and it also graphs the impact ... In this case, it has examples of cabs sited at 0km, 1.5km, 2.5km and 3.5km.
https://s30.postimg.org/ggfj9lknl/TNO_Effect_Of_PSDM...

We don't know that TNO used the same mask as BT, and they're unlikely to use the same cable, so the results aren't fully comparable, but give a flavour.

In reply to a post by MHC:

But, as far as I understand, ADSL will still be running through the cabinets and thus power masking will occur at the lower end.

No. The trials for LR-VDSL require the ADSL subscribers (fed via that cabinet) be converted to cabinet-fed services.

STIN522 says this:

In order to prevent exchange based ADSL or ADSL2plus customers served via a cabinet selected for LR-VDSL being adversely affected, these lines must be migrated to an FTTC GEA service (either VDSL or LR-VDSL, depending on D-side insertion loss measured at 300kHz). This means that LR-VDSL can be safely deployed without adversely impacting exchange based services.

Caveat: Elsewhere in the document, it waters this down slightly and says that SMPF services must be converted, and MPF services are allowed with the proviso that the expected degradation is acknowledged ... but still recommends GEA.

In reply to a post by MHC:

and thus power masking will occur at the lower end.

Power masking will actually take two forms...

For "close" subscribers who remain on the 17a profile, they will be masked as though the cabinet were set with CAL=0 ... which is "no extra masking at any ADSL/ADSL2+ frequency". But these lines do still follow that mask.

But for "LR" subscribers on the 8b profile, the entire PSD mask used by BT will be removed, and the "standard" one applied ("Annex B 998 (B8-4)"). Across the bulk of the ADSL2+ spectrum, this only limits power to -36.5dBm/Hz, compared to the -49.5dBm/Hz of the ANFP mask.

Because LR-VDSL lines are working to profile 8b, they are allowed an aggregate power of 20.5dBm, instead of the regular 14.5dBm. That means they can put more of the lower-frequency tones at a higher power.