Your original speeds were reasonable for a 16dB attenuation line. At one time I was seeing high 80s downstream. As others came online it slowly dropped ontil one particular neighbour was enabled and I saw a drop of well over 8Mbps ... Looking at your graphs it is likely that somewhere the noise floor has risen by 8 to 10dB and probably not much you can do about it.

NOTE

Before people panic no exhanges are closing for voice yet. Trial in Salisbury and Mildenhall is due later this year.

Result of trial is not a foregone conclusion

So suggesting people should/might worry about being on a small exchange they want to clsoe is over reacting at this time.

I didn't notice any panic about voice closure.

However, closure of ADSL means that that people will be forced onto VDSL, which for some properties will be worse than ADSL.

Once ADSL is removed can the VDSL profile be changed to include the now vacant ADSL frequencies to boost data rates?

Only once LLU ADSL2+ is also removed can the PSD masks be updated

You should try to deliberately slow down the line (see: https://forum.kitz.co.uk/index.php/topic,16427.0.html), and then slowly increment speed as to find the best characteristics whereby there are no Errored Seconds (ES or SES) being reported. As soon as you see even a single ES/SES slow it down one notch, and then DLM will keep the line on FastPath.

For me, despite having a cab literally glaring at me outside my window, I'm on a 600m line with 17.4dB attenuation that can only just support 56mbps down and 16.5mbps up on FastPath while maintaining the prized 5ms latency that's of utmost importance.

Why DLM can't do this automatically, even if we chose to be on the 'super stable' IP profile, is solely down to Openreach's ignominious incompetence.

Thank you, i'll have a look into this and see what it does.

What exchange in Essex? I noticed in your last stats that it�s targeting the 6db whereas before it was going down to 34.

In reply to a post by Danman7200:

What exchange in Essex? I noticed in your last stats that it�s targeting the 6db whereas before it was going down to 34.

My exchange code is EAWDF.

Yes, that's one of the main things I have noticed the SNR seems to target 6db rather than 3db. So I am hoping the limits on the line brings back the 3db target. Looking at the errors on the line today, its very clean, so I can only hope DLM is feeling generous.

Prior to reducing my speed, my stats looked like this:

adsl info --stats
adsl: ADSL driver and PHY status
Status: Showtime
Last Retrain Reason: 0
Last initialization procedure status: 0
Max: Upstream rate = 19846 Kbps, Downstream rate = 77153 Kbps
Bearer: 0, Upstream rate = 18994 Kbps, Downstream rate = 61081 Kbps

Link Power State: L0
Mode: VDSL2 Annex B
VDSL2 Profile: Profile 17a
TPS-TC: PTM Mode(0x0)
Trellis: U:ON /D:ON
Line Status: No Defect
Training Status: Showtime
Down Up
SNR (dB): 5.1 5.5
Attn(dB): 17.4 0.0
Pwr(dBm): 6.7 6.7

VDSL2 framing
Bearer 0
MSGc: 14 68
B: 39 59
M: 1 1
T: 64 3
R: 14 16
S: 0.0208 0.1001
L: 20736 6072
D: 1495 253
I: 54 76
N: 54 76

While I was getting more speed through the line, my SNR was struggling to get to 6dB and DLM was interleaving quite heavily. This resulted in quite a fast download, but terrible latency. (The 'D' is interleave depth). In further stats you will also probably see INP applied (INP and G.INP are different things), and a delay, but I've omitted for brevity. They were 4 and 8ms respectively.

Today, my stats are like this:

adsl info --stats
adsl: ADSL driver and PHY status
Status: Showtime
Last Retrain Reason: 0
Last initialization procedure status: 0
Max: Upstream rate = 22222 Kbps, Downstream rate = 68559 Kbps
Bearer: 0, Upstream rate = 15047 Kbps, Downstream rate = 56494 Kbps

Link Power State: L0
Mode: VDSL2 Annex B
VDSL2 Profile: Profile 17a
TPS-TC: PTM Mode(0x0)
Trellis: U:ON /D:ON
Line Status: No Defect
Training Status: Showtime
Down Up
SNR (dB): 9.1 9.6
Attn(dB): 17.4 0.0
Pwr(dBm): 5.0 5.0

VDSL2 framing
Bearer 0
MSGc: 17 159
B: 239 236
M: 1 1
T: 64 2
R: 0 16
S: 0.1352 0.5010
L: 14200 4072
D: 1 1
I: 240 255
N: 240 255

By deliberately slowing down my line, I sacrifice some speed, but get rid of interleaving and increase my SNR to 9dB. On balance I now have a much more stable line. As you can see, I've limited speeds to 56500 down and 15050 up, but I have FastPath back, and my latency dropped from 21-24ms to 5-7ms. These figures coincidentally to the largest number that rounds up to the 'Low' estimate of an impacted line on BT Availability Checker (56.6 mbps down and 15.1 mbps up for me), although I'm only pointing this out without empirical evidence.

DLM will not remove interleaving if you have an average of 10 or more Errored Seconds (ES) per hour. On a troublesome line it's highly unlikely you'll get ES down to zero (I'm in the 2-3 ES/hr average), but as long as the MTBE is below the threshold, then DLM will become forgiving (read: https://kitz.co.uk/adsl/DLM.htm for everything you need to know about DLM and MTBEs).

For the record, I'm on an ECI cabinet without 3dB nor G.INP enabled in Shepherd's Bush (LWSHE). The latter would probably help me massively - I've asked my ISP, IDnet to opt in, but so far nothing.

For some 24ms latency is no big deal and they need fast downloads. Peddle to the metal, my line can go quite fast, but that's not important. For me, losing 1MB/s on downloads is annoying but not nearly as annoying as the delays of high latency for DNS, SSH and the general 'snappiness' of everything, so it's each to their own.

Please note, to slow down a line, it has to be done on the modem itself (see: https://forum.kitz.co.uk/index.php/topic,16427.0.html) as part of the DSL training process when it syncs with the DSLAM, and deliberately increasing the SNR on the line in the process - slowing down a line on the router via QoS or some other limiter won't affect the line's quality.

Incidentally for QoS, you would want the QoS max line speed to be precisely 96.68% of the sync speed of your modem on a FastPath line. On an interleaved line, I haven't quite worked out the maths unfortunately, but people say you have to consider the Reed Solomon parity overhead (in order for FECs to work properly), and that it's related to R / N. However that would mean a 25.9% penalty on my original stats, and I could hit 60mbps on speed tests, which is far higher than such a penalty would suggest (43mbps), so I don't quite know.

Edit: grammar

Edited by zzing123 (Wed 01-Apr-20 12:53:45)