Kitz ADSL Broadband Information
adsl spacer  
Support this site
Home Broadband ISPs Tech Routers Wiki Forum
 
     
   Compare ISP   Rate your ISP
   Glossary   Glossary
 
Please login or register.

Login with username, password and session length
Advanced search  

News:

Pages: 1 2 [3] 4

Author Topic: How long should I expect to be banded for?  (Read 13896 times)

WWWombat

  • Kitizen
  • ****
  • Posts: 1674
Re: How long should I expect to be banded for?
« Reply #30 on: May 13, 2017, 12:31:21 AM »

My observations on this, from looking at MDWS...

- The power reduction is on the upstream only, and kicked in sometime between May 7th and May 9th;
- The upstream ES rate approximately tripled
- The upstream noise margin hasn't changed much
- The change in power is roughly from +3.9 to -0.9dBm. Does that make the new power about one third of the original?

I'll throw in a reminder that UPBO will be in play here; we can expect it to adjust things to try to keep power lower in the lower frequencies, depending on the line length.

I wonder why, if the transmission power dropped so much, didn't the SNRM drop much?

My further observations on surrounding line conditions
- Hlog is similar on both 7th and 9th (but D3 looks iffy)
- QLN looks similar on both 7th and 9th
- SNR/tone looks similar downstream. Upstream shows data on the 7th (mostly U1, but a few tones in U2), but nothing on the 9th (or after the resyncs on the 10th).
- Bits/tone looks similar on both upstream & downstream.

I can't get sufficient data on the framing parameters to know if they changed at all (eg changing FEC settings), which could affect the way in which quality is determined.

So, ultimately, I can't fathom why the drop in power is not seen with a drop in SNR per tone, and a consequent drop in bits per tone.

In which case, I wonder whether there really has been a drop in power. Instead, I wonder whether we are seeing a bug in the presentation of power data; the only other anomaly I can see is that, in the "after" case we also see no tones measured in U2. I therefore wonder if this is a fault in calculating (or displaying) the aggregate power, when that aggregate no longer includes U2.

  • Perhaps WWWombat may be able to make use of the concept of the reduction in power as another empirical indicator of a circuit that is banded?

As I see it, the line isn't banded in the upstream. Or it isn't hitting that limit.
Logged

Chrysalis

  • Content Team
  • Addicted Kitizen
  • *
  • Posts: 7388
  • VM Gig1 - AAISP L2TP
Re: How long should I expect to be banded for?
« Reply #31 on: May 13, 2017, 12:35:10 AM »

wombat I do believe US power does vary from sync to sync, as I see from one sync to the next changes on my US snrm, however the changes are minor in terms of the overall snr, I do see huge differences in US ES rate from one sync to the next as well.

I have observed after power cuts I tend to get very favourable treatment from the dslam on the US and have extremely low US ES rates, dslstats is currently reporting 0.2 ES/hour on my US for the previous day.  However I suspect if I were to do a resync now, it would increase to over 5 ES/hour.
« Last Edit: May 13, 2017, 12:41:34 AM by Chrysalis »
Logged

burakkucat

  • Respected
  • Senior Kitizen
  • *
  • Posts: 38300
  • Over the Rainbow Bridge
    • The ELRepo Project
Re: How long should I expect to be banded for?
« Reply #32 on: May 13, 2017, 12:47:11 AM »

- The change in power is roughly from +3.9 to -0.9dBm. Does that make the new power about one third of the original?

Doing the arithmetic --

10+3.9/10 ~ 2.46 mW

10-0.9/10 ~ 0.81 mW

2.46 / 0.81 ~ 3

Confirmed!  ;)

Quote
As I see it, the line isn't banded in the upstream. Or it isn't hitting that limit.

Ah, I see. So let's scratch that idea.  :-[
Logged
:cat:  100% Linux and, previously, Unix. Co-founder of the ELRepo Project.

Please consider making a donation to support the running of this site.

WWWombat

  • Kitizen
  • ****
  • Posts: 1674
Re: How long should I expect to be banded for?
« Reply #33 on: May 13, 2017, 12:59:51 AM »

I think it's just wishful thinking that you've got a contractual right to the highest sync speed free of whatever things restricting it that you don't like.

[rant]
I think BT have subtly changed the goalposts over time.

In ADSL and ADSL2+ days, BT/Openreach set their thresholds WRT engineer involvement, but ISPs had some flexibility over the line relating to DLM.

In original FTTC days, Openreach worked to similar thresholds WRT engineer involvement, and gave ISPs no flexibility relating to DLM. The attitude to the former says that BT are willing to investigate and fix physical faults, but they are forcing (some) ISPs to ignore unprovable faults. The attitude to the latter says that BT consider DLM to be vitally important, *and* that it is perfect in the decisions it makes.

[As a software engineer, I can say this: it is rather hard to make software perfect. If you can't make it perfect, then make it self-correcting]

The older variants of DLM weren't perfect. But they did relent, slowly. So, even if DLM made a bad choice sometimes, and got involved when it didn't need to, it would eventually remove itself. Thankfully.

The ability to self-correct meant that DLM's failure to be perfect was not a fatal flaw.

In current FTTC design, Openreach continue to make (some) ISPs ignore those unprovable physical faults. But as a consequence, they are also forced to ignore faults that DLM has latched on to and not let go.

The current variant of DLM seems to make bad choices - including being far too prone to using banding. It compounds this problem by having removed (or, at least, reduced) the self-correction. Banding, sometimes, is sticky.

To me, that looks like DLM is fatally flawed software that has inherited an ego problem from its designers: it cannot cope with its own flaws. DLM believes itself to make perfect decisions, and to have no need for a human nudge that says "Are you sure?" DLM is now the all-powerful arbiter that, sometimes, stops you getting the speed you ought to, but cannot see it.

And, because ISPs like TalkTalk aren't willing to get an engineer involved, you've lost any manual backstop.

If the Openreach platform had a button labelled "Please ask DLM to try relenting", TalkTalk would hardly refuse to press it.
Logged

WWWombat

  • Kitizen
  • ****
  • Posts: 1674
Re: How long should I expect to be banded for?
« Reply #34 on: May 13, 2017, 01:13:40 AM »

It's all been documented before on this forum  ...... maybe someone else can provide links to said info, I haven't got the time or wherewithal to go looking.

We have this example of the opposite case, presented by Walter a year ago:
http://forum.kitz.co.uk/index.php/topic,16804.msg311500.html#msg311500

It involves a line that had been banded, and was recovering in 2.5Mbps steps every couple of days. Recovery then just stopped (at 27.4Mbps), with SNRM at 15dB, with no indication *why* it stopped. Certainly the error rates don't seem to have increased.

The post I link to is Walter reporting the freeze; two posts later we can see a graph of the line, from BaldEagle.

In that case, DLM ought to have continued to reduce banding (and increase the allowed speed).
Logged

WWWombat

  • Kitizen
  • ****
  • Posts: 1674
Re: How long should I expect to be banded for?
« Reply #35 on: May 13, 2017, 01:17:21 AM »

wombat I do believe US power does vary from sync to sync,

Yes. I think upstream handling is purposefully different, because it can make a difference to people on longer lines.

Here, you need something dynamic and flexible. DLM can be dynamic, but not always flexible.
Logged

WWWombat

  • Kitizen
  • ****
  • Posts: 1674
Re: How long should I expect to be banded for?
« Reply #36 on: May 14, 2017, 04:06:44 AM »

While re-connecting to MDWS, my screen updated @gbeth's data, and I noticed it had undergone a couple of resyncs earlier today (13th).

In particular, it was clear that the upstream power had, for a couple of hours, gone back up again. All with little change to the sync speed or much difference to the SNRM.

However, like I noticed here:

In which case, I wonder whether there really has been a drop in power. Instead, I wonder whether we are seeing a bug in the presentation of power data; the only other anomaly I can see is that, in the "after" case we also see no tones measured in U2. I therefore wonder if this is a fault in calculating (or displaying) the aggregate power, when that aggregate no longer includes U2.

The brief period today (between 2 resyncs) show measurements for U2 in the "SNR per tone" graph, which aren't present in the periods of low power before the first resync or after the second resync.

Note that while there is a difference in the upstream data in the "SNR per tone" graphs, there isn't really a difference in the consequent "bits per tone" graph.
Logged

GigabitEthernet

  • Kitizen
  • ****
  • Posts: 2243
Re: How long should I expect to be banded for?
« Reply #37 on: May 18, 2017, 03:57:16 PM »

Well my line has been green on MDWS for a couple of days now after I made some cabling changes. If it stays like that, in theory I should be looking at around 10 days before DLM does something I think?


Sent from my iPhone using Tapatalk
Logged

GigabitEthernet

  • Kitizen
  • ****
  • Posts: 2243
Re: How long should I expect to be banded for?
« Reply #38 on: June 01, 2017, 01:16:32 AM »

Something curious happened tonight. My line resynced and the sync speed is the same but the SNRM has dropped from 12 to 8? In one go, that's too much to put down to interference so what is going on here?
Logged

GigabitEthernet

  • Kitizen
  • ****
  • Posts: 2243
Re: How long should I expect to be banded for?
« Reply #39 on: June 01, 2017, 01:20:30 AM »

Ah! Upstream power has gone up significantly and I've gained almost 500Kbps of upload sync.
Logged

banger

  • Kitizen
  • ****
  • Posts: 1186
  • TTB 80/20
Re: How long should I expect to be banded for?
« Reply #40 on: June 01, 2017, 01:39:38 AM »

Is your DS still banded?
Logged
Tim
talktalkbusiness.net & freenetname
Asus RT-AC68U and ZyXEL VMG1312-B10A Bridge on 80 Meg TTB Fibre

https://www.thinkbroadband.com/speedtest/1502566996147131655

GigabitEthernet

  • Kitizen
  • ****
  • Posts: 2243
Re: How long should I expect to be banded for?
« Reply #41 on: June 01, 2017, 01:52:12 AM »

Logged

GigabitEthernet

  • Kitizen
  • ****
  • Posts: 2243
Re: How long should I expect to be banded for?
« Reply #42 on: June 15, 2017, 09:45:37 PM »

Would migrating to a slower product remove the banding?
Logged

GigabitEthernet

  • Kitizen
  • ****
  • Posts: 2243
Re: How long should I expect to be banded for?
« Reply #43 on: June 18, 2017, 01:03:27 AM »

I feel lonely in here :(
Logged

broadstairs

  • Kitizen
  • ****
  • Posts: 3700
Re: How long should I expect to be banded for?
« Reply #44 on: June 18, 2017, 07:56:28 AM »

Dont feel too lonely, my line was banded in August 2016 and it still is banded! My ISP refuses to do anything about it.

Stuart
Logged
ISP:Vodafone Router:Vodafone Wi-Fi hub FTTP
Pages: 1 2 [3] 4
 

anything