Thanks for that Bob.

You know a great deal more about exchange kit than I do (well actually I know nothing ) but would 21CN conservatively apply a lower sync in the interests of stability compared to LLU?

Would I, for instance, expect much of a drop in downstream sync from currently 13107K with an attenuation of 37dB, a SNRM of 6.5dB and interleaved if changing from LLU to 21CN?

There's a few "ifs" involved in answering that .

On initial connection to WBC, (which is the proper name for 21CN broadband, to distinguish it from 21CN PSTN - phone), I would expect about the same sync. But you would possibly get a small drop in throughput due to the IP Profile.

As well as the IP Profile you also lose a little throughput speed due to the ATM overheads between you and the exchange. You lose that on LLU as well as it's simply the way the TCP/IP data packets are encapsulated on non-digital (analogue) circuits. I can't speak for Sky LLU, but on O2/Be LLU the throughput is around 84% of sync.

On WBC there is the possibility that the DLM will raise the sync-time noise margin if the line proves unstable or has high error rates. That will lower the sync. I don't think the Sky one does, eben if it isn't switched off after a few days. O2/Be DLM is switched of after about 3 days from migration in, but the WBC DLM runs all the time you are connected.

On the other hand, if the line proves to be stable, the WBC DLM wil lower the sync-time noise margin to 3dB. That gives a big increase in sync speed. 1Mbps or more.

I think that covers most of it .

Edited by RobertoS (Sat 14-Jul-12 00:09:58)

In reply to a post by RobertoS:

On WBC there is the possibility that the DLM will raise the sync-time noise margin if the line proves unstable or has high error rates. I don't think the Sky one does, eben if it isn't switched off after a few days. O2/Be DLM is switched of after about 3 days from migration in, but the WBC DLM runs all the time you are connected.

On the other hand, if the line proves to be stable, the WBC DLM wil lower the sync-time noise margin to 3dB. That gives a big increase in sync speed. 1Mbps or more.

Just popped in for a peek. Had a couple of whiskys watching the Last Night of the Proms...

Egads, tis Friday night!

Sorry.

I watched most of the prom. Didn't think it was up to much. I shall tell Mark next time I'm talking to him.

I did expect the conductors to pas the baton, given the hype about it in the runup over the last few days .

That certainly does cover it

But what about this: http://www.kitz.co.uk/adsl/max_speed_calc.php - if I put in say a downstream attenuation of 34dB then BT 21CN WBC syncs at 12352K with an IP Profile of 10894K yet LLU syncs at the same rate but throughput is 10932K, wouldn't throughput with 21CN WBC be less than the IP Profile?

Certainly I always noticed that with adsl1 throughput was generally ~500K less than the IP Profile which may be significant on a longish line or is that an irrelevance with adsl2+?

Eeeek!

Two things there.

First, I agree with you in a way. In fact I think you understate the losses after the IP Profile is taken into account. But I feel you are possibly still slightly confused - not surprising, as it's a complex subject to pick up.

On any given attenuation I would expect higher throughput from an LLU provider who is using a 6dB or 7dB target noise margin, than on BT Wholesale circuits. At one time some used 9dB, which would be slower.

The IP Profile is specific to BT Wholesale circuits. It is used on ADSL Max/IPStream (20CN), IPSC, and on WBC/WBMC (21CN). I'd be very surprised to see reliable speed tests of throughput, (so I'm excluding speedtest.net which is inherently biased to reporting speeds towards the peak during a very short test), showing more than 90% of the IP Profile.

I always thought Newnet a few years ago were very good. I expected 70+% of profile. Zen, IDNet and AAISP were probably better.

Even more contentious is the difference between speed tests and real-life file and burst data transfer. The BT speed test for instance, (when it isn't having a hissy fit of total instability as seen over the last few weeks), seems to set up an uncontended path from the BT server to the ISP's router. Thus eliminating the ISP's links to the rest of the world. I might add that I think that is the right thing to do, as it tests the link between the ISP and you. No other speed test can do that.

LLU is in general ADSL2+, but is occasionally capped at 8Mbps sync. The relationship between sync and throughput is independent of whether ADSL or ADSL2+ is in use, on either BTW or LLU, but on BTW is of course affected by the IP Profile. The IP Profile tries to reflect a data speed, in that it absorbs a lot of the overheads that cause the drop on LLU lines to about 84% of sync.

Second:-

If I summon up the will-power I shall contact kitz about that page. There are a number of things that aren't obvious on it which to me make it a dangerous reference. That rather takes me aback as I have always regarded kitz and samknows as far ahead of me in technical knowledge, and on my own website recommend them for more in-depth information than I provide there. It's the content of the page that is dodgy, not the underlying expertise of kitz.

It looks as though it has been hurriedly produced from an earlier version, that earlier version being designed to estimate the speed increase people may get on moving from ADSL to ADSL2+.

In it's current incarnation the stand-alone statement "Rate adaptive DSL also makes use of any spare SNR Margin therefore the higher your SNR Margin, then the better your speeds will be" is simply untrue. In fact on a third or fourth reading becomes gibberish. It is true only in the rare circumstance of an ADSL(1) user on 8128kbps Fast Path or 8128/7616kbps Interleaved who is moving to BT WB(M)C or LLU.

It also assumes Interleaving is on for ADSL Max.

The graph linked to at the bottom is a very similar concept to the one I gave you, and one of the mainstays of his calculator. Also see this alternative, (MrSaffron), which I often use as a backup to the graph.

I'm on 20CN. speedtest.net results vary between 6.72 and 6.98 so I'm seeing less than 500 drop from my 7150 profile.

In reply to a post by RobertoS:

In it's current incarnation the stand-alone statement "Rate adaptive DSL also makes use of any spare SNR Margin therefore the higher your SNR Margin, then the better your speeds will be" is simply untrue. In fact on a third or fourth reading becomes gibberish. It is true only in the rare circumstance of an ADSL(1) user on 8128kbps Fast Path or 8128/7616kbps Interleaved who is moving to BT WB(M)C or LLU.

It also assumes Interleaving is on for ADSL Max.

My interpretation of that statement (& what I have sort of seen in practice) is that when SNRM is high, a forced resync at that time usually produces higher sync speeds (unless capped by DLM or an ISP/BT), but SNRM is immediately lowered as the "spare" margin has then been used up to deliver the higher sync speed.

My observations have been on my FTTC/VDSL2 connection, but it does use much of the technology of ADSL2+.
Indeed it is supposedly backwardly compatible with ADSL2+.

I have also seen this effect in graphs from ADSL2 & ADSL2+ users' connections.

Edited by deleted (Sat 14-Jul-12 10:06:15)

In reply to a post by Bald_Eagle1:

My interpretation of that statement (& what I have sort of seen in practice) is that when SNRM is high, a forced resync at that time usually produces higher sync speeds (unless capped by DLM or an ISP/BT), but SNRM is immediately lowered as the "spare" margin has then been used up to deliver the higher sync speed.

I've also seen that on my FTTC connection (not with Plusnet).

But OR's DLM on FTTC seems more benign than BTw's on ADSL, and a period (10-14 days?) of high SNRM after some event that caused the line to drop will result in the DLM initiating a resync on its own initiative.

You are misinterpreting what is going on. You are just talking about the effect on speed of re-sync'ing when for some reason you have an abnormally high noise margin.

You are thinking of cases (1) and (2) on this page. The kitz quote is specifically about case (3), which I was trying to explain in the context of that post.