Have a look at the link in an earlier post: http://forums.thinkbroadband.com/fibre/t/4434799-re-...

Hm yes... Well, what I can tell you is that with a corded phone connected to the test socket on master, I get an unbelievable amount of static during quiet test. I can actually hear it during the dial tone as well, it's so loud.

Thankfully, Zen/BTO have identified a fault somewhere along my line and are on it. Strangely, the Billion router has been syncing at higher speeds that the Technicolor router - it's now at 39500kbit vs 38500 over the past few days.

Let's hope that once they fix the noisy line (on voice level anyway), BB issues will go away. That said, the BB itself is quite stable with not drops.

In reply to a post by MHC:

There are a few nasties in the Tone 211 to 350 region. 211 is 909kHz - a BBC radio frequency! and the other notches are too. They appear on others but yours is quite drastic. Do you live close to a radio transmitter?

What I can never understand is why the top end cuts off ... you are still seeing 6 bits per tone and even 1 bit is usable. I wonder if there is a lot of noise at the top end?

I broadly agree with both of these points being sources for slower speeds than expected.

My previous line had an attenuation of 16.7dB, and had bit loading like this, with usable SNR continuing up to near tone 4000:
http://postimg.org/image/xuzqz9p0n/

The matching bitloading:
http://postimg.org/image/o3093jt9z/

These graphs only really show us the "effect", not the "cause". For that, we might get some ideas from seeing Hlog and QLN graphs:
http://postimg.org/image/g3g9m6qax/
http://postimg.org/image/5bait1t8f/

Hlog and/or QLN might suggest why the top-end tones are being cut off prematurely.

Of course, they might be perfectly explained by the recently-discovered audible fault...

I'd already look at that. It tells me that there's a fairly broad band about 600Khz wide ending at about 1.5Mhz where the SNR is considerably reduced. You cannot tell from that chart alone whether that reduction in SNR is due to an increase in noise or a reduction in the power density. However, given the ANFP it seems obvious that the great majority of it is due to the PSD. There are some very distinct "notches" in that area too. I'm not privy to the details, but I think it far more likely that it's power notching and not narrow band interference.

It's also clear that the SNR is measured up to a bit over 7MHz. I know that VDSL2 can go up to 30MHz, but it was my understanding that the maximum in the UK is around that 7Mhz signal. I'll make a guess now that it was cutoff early to allow room for g.fast. gfast in it's first version goes up to 106Mhz but, presumably BT will want it to extend reasonably far down the spectrum to extend its useful reach.

Hm, well, not sure why it's reporting 0 SNR above 2MHz - maybe I am missing some setting in DSL stats?

Zen came back today to say that there's a battery contact fault on my line, so hopefully with that being cleared, I'll get my crackle free line back and decent net speeds.

In reply to a post by TheEulerID:

If the cabinet is close to the exchange, then that power roll-off can extend all the way up to about 1.5Mhz. (The result of this is, perhaps perversely you might think, the worse you ADSL2 service the better you might expect your VDSL to be for any given subloop length).

I think there is a balance... the further the cabinet is from the exchange, the smaller the band of affected frequencies is, but the deeper the power cutback. The closer the cabinet to the exchange, the broader the band, but the shallower the power reduction.

At one extreme, you'll notice that, for CAL=0 (cabinet is *really* close to the exchange, or within 100m), there is no power reduction at all.

We saw some graphs on here recently, with a very tiny notch at 2.2MHz. That may have been a CAL of 2.

I reckon that around CAL=20 is the worst case, which I guess is for cabinets around 2km from the exchange. The ANFP graph suggests a broad and deep disruption.

A proper mathematical integration of the formulae would probably give us the answer...

In reply to a post by TheEulerID:

It's also clear that the SNR is measured up to a bit over 7MHz. I know that VDSL2 can go up to 30MHz, but it was my understanding that the maximum in the UK is around that 7Mhz signal. I'll make a guess now that it was cutoff early to allow room for g.fast. gfast in it's first version goes up to 106Mhz but, presumably BT will want it to extend reasonably far down the spectrum to extend its useful reach.

Back in the days of profile 8c, the UK limit used to be 7.05MHz, even though the profile allowed up to 8MHz. But with the current ANFP allowing profile 17a since 2011, we ought to be able to use up to 17.664MHz. I think your link goes to the 2007 version.

This might be better:
http://www.niccstandards.org.uk/files/current/ND1602...

The Huawei cabs seem to top out at tone 3970, which is 17.1MHz

That's not 2Mhz. Those are tone numbers and they are approximately 4K buckets so multiply by about 4 to get the rough frequency.

I suspected my link was going to be a bit out of date, and the cut-off does seem to be closer to 8Mhz than 7MHz.

As far as extending beyond that it concerned, then if the ANFP allows for it, then that's fine but I would still maintain that OpenReach might have chosen not to exploit that as it does allow more space and longer reach for g.fast. That might take preference over allowing higher speeds n VDSL2 lines.

Of course the ANFP will need reviewing for g.fast anyway and, in theory at least, that's not something BT can dictate. However, given that the overwhelming majority of VDSL2 in the UK is provided by BT, then they will be in a better position to argue g.fast can start lower down the frequency range than would be the case if the VDSL2 frequency plans are extended higher up the range.

Of course, some of this might just be down to limitations in line cards at particular locations.

One thing is very clear, and that is the simple minded modelling of possible VDSL2 speeds based solely on the project attenuation of the sub-loop via the proxy of cabinet distance is far too simple. It looks very much like the speed that might be expected will also depend heavily on the details of the frequency plan and psd at any one cabinet location.

It would be wonderful to have stats on all this available, but unfortunately I think such information will only be available to OpenReach, assuming they can collect things like bit loading profiles, attenuation and so on by line. It would, of course, be a huge amount of data but would surely be very useful in working out coverage and even might what be projected into g.fast coverage (from the existing cabinet locations). I suspect that this is information BT would consider highly confidential though.

Welp, the BT engineer has been - unannounced (why do they never say they are coming?) - but the problem is unresolved. He spent a while looking for the other end of my cable - he isn't a local - and after two hours said that the problem is with the underground cable between my house and the next connection point. He tried swapping the wire-pair to the unused one in my cable but that didn't work.

So, apparently, someone else will need to come out and find where the fault actually lies along the length of the cable, dig down and correct the fault.

Anyone have any experience with this kind of underground wire fault?