In reply to a post by WWWombat:

Beautiful looking graphs and numbers there.

Is it normal to have that bizarre-looking rolloff in S/N ratio and bits/carrier between 1 and 2MHz?

I see the same on my line and it wasn't there using ADSL2

If it's normal, does anyone have an explanation about what causes it? As a RF-tech that kind of frequency response makes me distinctly twitchy.

Yes, if the effect you mentioned is what I think it is, compare to my graph at http://81.138.100.210/images/2013-6-5/bit_tones/bit_... then yes it's normal. It's to try and minimise the impact of VDSL2 crosstalk for ADSL2+ users.

Edited by Ixel (Wed 05-Jun-13 19:23:06)

Thanks. That makes sense. If only BTOR could provide this information when asked..

It looked like some kind of unterminated spur about 90 metres long on the line, which was making me twitchy.

Peripherally related: With a 12db SNR line syncing at 80/20 would you expect 30-50 uncorrectable DSL errors per hour? (It was a LOT worse on ADSL2 - upwards of 200/hour with 6dB SNR and syncing at 16.5/1)

That notch in SNR (and bit-loading) is the result of downstream PSD shaping.
This shaping is performed to preserve, as Ixel has said, the ADSL connections served by exchange DSLAM kit.
The Profile 17a VDSL2 class mask used by BT to define that shaping is ADE17-M2x-A (short name: B8-11). See page 21 of BT SIN 498 (Issue 4.3) [1]
The PSD mask itself is defined in ITU-T VDSL2 Recommendations G993.2. See page 306 of the 12/2011 Recommendations at [2].

cheers, a

[1] http://www.sinet.bt.com/498v4p3.pdf
[2] http://www.itu.int/rec/dologin_pub.asp?lang=e&id=T-R...

Edited by deleted (Wed 05-Jun-13 19:54:02)

Thanks. That makes sense. If only BTOR could provide this information when asked..

It's actually a branch of Ofcom known as NICC that set the standards for connecting to Openreach's network - the "ANFP".

The PSD has that notch somewhere between 200kHz and 2.1MHz - the location actually depends on the distance that the cab is from the exchange.

You can find the ANFP at the NICC Standards website. Page 30 shows the 6 different PSDs that apply in the UK.

There is also an old VDSL2 primer from Openreach, that dates back to the beginnings of the trial (so only uses up to 7MHz rather than the current 17MHz).

It looked like some kind of unterminated spur about 90 metres long on the line, which was making me twitchy.

If that were the case, you'd presumably see something in the HLog or QLN graphs as well as the SNR graph.

With a 12db SNR line syncing at 80/20 would you expect 30-50 uncorrectable DSL errors per hour? (It was a LOT worse on ADSL2 - upwards of 200/hour with 6dB SNR and syncing at 16.5/1)

The error rate isn't always dependent on the SNR that you are seeing - it really depends on your local environment and the noise seen there. One factor is the crosstalk seen from your neighbouring lines in the cable, so the situation changes as more people take up FTTC services.

My current line runs at 80/20 with about 5dB of SNRM at the moment. There is no FEC or interleaving turned on, so all errors are seen as "OHFErr" counts, rather than "RSUncorr".

It gets around 50-150 OHFErr errors per hour.

This time last year, when there was more leeway, and it ran about 7dB, it ran at between 400 and 1000 OHFErr per hour. There was a step change in error rate sometime between September and November that I didn't catch in my statistics.

Ye Gods! I forgot all about the ANFP complicating things even further! Possibly if you studied the SNR profile closely enough you could maybe work out which of the six downstream PSD masks is being utilised?

cheers, a

Oh yes.

The troughs would be at
- CAL 50 = 450 kHz
- CAL 40 = 600 - 800 kHz
- CAL 30 = 1.1 - 1.2 MHz
- CAL 20 = 1.8 - 2.0MHz
- CAL 10 = 2.2 MHz (but not as deep as the previous troughs)
- CAL 0 = No trough (flat from 200kHz to 3MHz)

I've not seen a graph for CAL-0 though. I guess you need to be on a cabinet outside the exchange for that one.

On my old line, the trough was around tone 410-480, or 1.7-2.0 MHz. That'd be "CAL 20". The cabinet on that line was about 1.2km from the exchange building, following the obvious road.

On my current line, the trough is at tone 300, or about 1.2MHz. I guess that is the "CAL 30" mask. This cabinet is definitely further from the exchange, but I don't know the distance. Probably about 1.5-1.6km.

Hi,
Thanks for the responses,

My ISP is BT, I have done a speedtest at this link: http://speedtest.btwholesale.com/ and it confirms that my IP profile is 77.43 Mbps.

As this looks like a congestion issue with the network would moving to an alternative provider resolve this issue? Or do all IPS's route through the same BT equipment?

The connection the router is wired and all speedtests are being done wired as well.

Edited by deleted (Thu 06-Jun-13 08:43:31)

the Committed Information Rate on the cabinet to exchange link from Openreach is 30 Mbits/s if you're on 80/20 product. So there's one potential bottleneck. That would apply to any FTTC using Openreach cabs, inc Talk Talk and Sky.

After that it's the BT Wholesale network vs either TT or Sky.

I can't recall what BT Wholesale's ambitions for minimum service are - something like 12M minimum at the busiest 3 hour period ?

For End-User Accesses selecting the "Standard" traffic option, BT Wholesale will
provide a throughput level of at least 12 Mbit/s for 90% of the busy 3 hour period.

For End-User Accesses selecting the "Elevated" traffic option, BT Wholesale will
provide a throughput level of at least 16Mbit/s for 90% of the busy 3 hour period.

Edited by yarwell (Thu 06-Jun-13 08:54:10)

Hi,
Thanks for the responses,

So just to confirm if I did move to an alternative provider this problem would most likely continue as all providers route traffic through the BT network?

Thanks,