Print

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

[art37]

Supervectoring 35b seems to be the path that Germany, for example, is now going down to improve FTTC speeds.  Is this a route that the UK will follow or is all our money now on G.Fast and FTTP? I ask because I am looking to replace my modem/router, and supervectoring keeps appearing as a 'future proof' option.

[Ixel]

I don't think this will happen for FTTC in the UK for a few reasons.

Primary reasons:
- Vectoring would be needed for all cabinets to support this? ECI is out of luck, unless they just roll this out to all of their Huawei estate and let their ECI estate fall even further behind than it currently is.
- G.fast is starting at 19 MHz currently I believe, until if and when they figure out how to 'share' the current VDSL2 17a spectrum.

I'd say money is now on G.fast.

[art37]

I don't think this will happen for FTTC in the UK for a few reasons.

Primary reasons:
- Vectoring would be needed for all cabinets to support this? ECI is out of luck, unless they just roll this out to all of their Huawei estate and let their ECI estate fall even further behind than it currently is.
- G.fast is starting at 19 MHz currently I believe, until if and when they figure out how to 'share' the current VDSL2 17a spectrum.

I'd say money is now on G.fast.

That's my take on the situation. Thanks.

[kitz]

As an aside Duetche Telekom used ECI for their NGA network.  They used a large amount of M41's in their cabs.   They were experiencing some similar issues to Openreach when it came to G.INP. - link
Not sure whats happened since then as I havent seen much mention...  but they could have upgraded to V41s ?

[WWWombat]

Same answer as @Ixel.

My emphasis would be less on the need for vectoring; the "supervectoring" name doesn't indicate any superior need for vectoring, but came about really because the 35b profile was all about keeping vectoring compatibility with the 17a profile (as 30a isn't compatible). That allows some lines to be on 17a, and some lines on 35b.

My emphasis would be on the choice of spectrum compatibility; whatever spectrum VDSL2 uses, G.Fast cannot use the same spectrum in the same cabinet area (NEXT problems). Limiting G.Fast to above 35-40MHz would then likely make it too slow and/or too restricted a range to do the job they currently want it to (300Mbps at 300m) and would prevent its use nearer the home (eg at the DP) in the future. The choice of 35b or G.Fast appears to be a single-shot choice of one or the other for a cabinet area.

[WWWombat]

They were experiencing some similar issues to Openreach when it came to G.INP. - link

Oooh. This translation caught my eye:

The fact that individual users measure too low a bandwidth is explained by Telekom: "Due to the inconsistent definition of international standards in the transmission of intermediate information between the MSAN and the downstream BRAS, however, certain constellations also result in a reduced speed Value. "

I wonder if this links to the behaviour we see with the combination of "Retransmission High" and "XdB"?

[kitz]

Hmm  good point wombat.   

I originally found and used that link when Openreach first introduced G.INP and when it was discovered that those users who had modems which were only capable of G.INP in the downstream direction were losing ~10Mbps from their downstream speed if g.inp was enabled on upstream.
Why upstream g.inp should decrease downstream speeds by 10Mbps if the modem was incapable of upstream g.inp was a bit of a mystery, but whatever the technical reason,  its why Openreach did the botch job of turning off default upstream g.inp.

Back then we didn't have 'x'dB, so we wouldn't have been aware of the fact that it could also affect that too. 

[ejs]

I think regardless of the type of cabinet, supervectoring / profile 35b would need new line cards and a new vectoring engine. I thought the crosstalk problem gets worse with higher frequencies, so it did need more powerful vectoring. Certainly the vectoring engine has more processing to do to handle the increased bandwidth of the signals.

Probably supervectoring is less "future proof", because there's nothing beyond it with VDSL2, whereas G.fast continues to be developed.

[KIAB]

Will the time come when they give up on the ECI cabinets & start replacing with them with Huawei cabinets.

[WWWombat]

I thought the crosstalk problem gets worse with higher frequencies, so it did need more powerful vectoring.
Certainly the vectoring engine has more processing to do to handle the increased bandwidth of the signals.

The latter is true; I suspect the amount of work is in proportion to the number of tones, or maybe even its square.

Crosstalk does get worse at higher frequencies, but not in a way that affects how many calculations you need to do. Just the weighting you apply within the calculation.

However, the G.Fast research found that crosstalk get more worse at really higher frequencies than the old models predicted, so came up with new models. The old models were valid to around 30MHz, the new models to 500MHz. See page 43 in
http://www.joepeesoft.com/Public/DSL_Corner/Docs/Presentations/PUB_2014_05_20_GFAST_Summit_4GBB_Brink.pdf

I presume that standardisation of vectoring for VDSL2 only took into account the old models, so gets the calculations wrong when the second-order fext effect kicks in. So I suspect that vectoring for VDSL2 becomes less effective at the higher frequencies, as the second order effect increases in magnitude.