How BT get faster speeds than any other ISP?

Seems to be obvious.

Sine my line was fixed I could sustain a really nice 5Mbps at around
6- 8dB SNR.. However despite having zero issues at this speed, the SNR
margin has been progressively raised until its now at 15dB.

Possibly connected with the fact that I spent some serious time this
weekend watching the BBC HD F1 stream..and listening to test match
special over the internet? This consuming lots of bandwidth..

Arte BT using SNR margins to control traffic? And if they are, how do I
prove it, and who can do something about it?

I am fairly sure its outside the ISP's powers to fiddle with this,
beyond asking if it can be reset..

Comments welcome...

In message <j19215$vvd$(E-Mail Removed)>, The Natural Philosopher
<(E-Mail Removed)> writes
>Seems to be obvious.
>
>Sine my line was fixed I could sustain a really nice 5Mbps at around
>6- 8dB SNR.. However despite having zero issues at this speed, the SNR
>margin has been progressively raised until its now at 15dB.
>
>Possibly connected with the fact that I spent some serious time this
>weekend watching the BBC HD F1 stream..and listening to test match
>special over the internet? This consuming lots of bandwidth..
>
>Arte BT using SNR margins to control traffic? And if they are, how do I
>prove it, and who can do something about it?
>
>I am fairly sure its outside the ISP's powers to fiddle with this,
>beyond asking if it can be reset..
>
>Comments welcome...

Possibly - would make a fairly good conspiracy theory, but the
relationship between BT Retail (the ISP) and Openreach who own the ADSL
kit that controls the SNR is officially the same as between any other
ISP and Openreach. I doubt that Openreach give BT any more (or less)
control than they give to other ISP's.

A more likely explanation is that your line has high(ish) error rates.
The things the router reports as things like

SF (CRC) Errors:
RS Corrected:
RS Un-Corrected:
HEC:
Errored Seconds:
Severe ES:
etc.

I understand that the DSLAM in the exchange monitors these things and if
the error rate rises "too high" then it increases the SNR to reduce
them. The theory being that high error rates mean lots of retransmission
and hence slow(er) real speeds and inefficient network use.

[I seem to remember reading somewhere that because of the design of the
ATM protocol if the ADSL error is uncorrectable by the router then
packet retransmission takes place at the TCP level from the IP's kit not
from the exchange - so its the load this places on their backbone
network they are trying to avoid; as well as trying to give you the
fastest possible download experience ]

If you have got an error prone line, then a very heavy usage weekend
would have increased the numbers of packets passed and consequently the
number of errors recorded by the DSLAM. So it reacted.

If you can monitor the error rates (Routerstats does a good job if you
have the right router) you might get some idea as to the problem.

--
Invalid

Invalid wrote:
> In message <j19215$vvd$(E-Mail Removed)>, The Natural Philosopher
> <(E-Mail Removed)> writes
>> Seems to be obvious.
>>
>> Sine my line was fixed I could sustain a really nice 5Mbps at around
>> 6- 8dB SNR.. However despite having zero issues at this speed, the
>> SNR margin has been progressively raised until its now at 15dB.
>>
>> Possibly connected with the fact that I spent some serious time this
>> weekend watching the BBC HD F1 stream..and listening to test match
>> special over the internet? This consuming lots of bandwidth..
>>
>> Arte BT using SNR margins to control traffic? And if they are, how do
>> I prove it, and who can do something about it?
>>
>> I am fairly sure its outside the ISP's powers to fiddle with this,
>> beyond asking if it can be reset..
>>
>> Comments welcome...
>
> Possibly - would make a fairly good conspiracy theory, but the
> relationship between BT Retail (the ISP) and Openreach who own the ADSL
> kit that controls the SNR is officially the same as between any other
> ISP and Openreach. I doubt that Openreach give BT any more (or less)
> control than they give to other ISP's.
>
> A more likely explanation is that your line has high(ish) error rates.
> The things the router reports as things like
>
> SF (CRC) Errors:
> RS Corrected:
> RS Un-Corrected:
> HEC:
> Errored Seconds:
> Severe ES:
> etc.
>
> I understand that the DSLAM in the exchange monitors these things and if
> the error rate rises "too high" then it increases the SNR to reduce
> them. The theory being that high error rates mean lots of retransmission
> and hence slow(er) real speeds and inefficient network use.
>
> [I seem to remember reading somewhere that because of the design of the
> ATM protocol if the ADSL error is uncorrectable by the router then
> packet retransmission takes place at the TCP level from the IP's kit not
> from the exchange - so its the load this places on their backbone
> network they are trying to avoid; as well as trying to give you the
> fastest possible download experience ]
>
> If you have got an error prone line, then a very heavy usage weekend
> would have increased the numbers of packets passed and consequently the
> number of errors recorded by the DSLAM. So it reacted.
>
> If you can monitor the error rates (Routerstats does a good job if you
> have the right router) you might get some idea as to the problem.
>


well further to this I said 'what the hell' and phoned up the ISP.

Who said something very strange..

"Your UPSTREAM error rate is very high. That's why it's reset the noise
margin."

I was seeing no errors at all reported on the router going upstream though.

And precious few downstream either. But I suppose that correctable CRC
errors that the DSLAM detected would not necessarily be reported back to
the router..only completely damaged packets that needed retransmits..not
sure how the ADSL works at that level.

Anyway after mulling that one over, I asked for a noise margin reset and
currently the line is holding a shade under 6Mbps at a noise margin of
6dB. And no interleave, which is a first EVER. I doubt it will hold that
through the night-time MW transmissions, but it should settle out over
5Mbps anyway.

The tech droid was insistent - no idea why - that this was something
injecting noise at my end. He didn't seem to grasp the basic fact that
the UPSTREAM signals were weakest at the DSLAM end, and therefore more
likely to be affected by noise further up the line... a lecture on
communication theory being of no profit I simply asked that the line be
reset.

Which happened LESS THAN A MINUTE after I put the phone down.

I am really puzzled by the whole affair. If what he was saying was
correct, that could only be LF trash of some sort right up at the
exchange... some kid with a poorly suppressesd motorbike? someone
fiddling with wires?


Mind you, its does suggest a way to bugger up your neighbourhood DSL.
switch your router OFF and then get a spark transmitter going. When they
are totally narfed at 24dB margin and crawling, you can reboot your
router and get a clean almost uncontended connection :-)

Any other thoughts welcome.

In message <j1a5l1$62t$(E-Mail Removed)>, The Natural Philosopher
<(E-Mail Removed)> writes
>Invalid wrote:
>> In message <j19215$vvd$(E-Mail Removed)>, The Natural Philosopher
>><(E-Mail Removed)> writes
>>> Seems to be obvious.
>>>
>>> Sine my line was fixed I could sustain a really nice 5Mbps at
>>>around 6- 8dB SNR.. However despite having zero issues at this
>>>speed, the SNR margin has been progressively raised until its now
>>>
>>>
>>> Possibly connected with the fact that I spent some serious time this
>>>weekend watching the BBC HD F1 stream..and listening to test match
>>>special over the internet? This consuming lots of bandwidth..
>>>
>>> Arte BT using SNR margins to control traffic? And if they are, how
>>>do I prove it, and who can do something about it?
>>>
>>> I am fairly sure its outside the ISP's powers to fiddle with this,
>>>beyond asking if it can be reset..
>>>
>>> Comments welcome...
>> Possibly - would make a fairly good conspiracy theory, but the
>>relationship between BT Retail (the ISP) and Openreach who own the
>>ADSL kit that controls the SNR is officially the same as between any
>>other ISP and Openreach. I doubt that Openreach give BT any more (or
>>less) control than they give to other ISP's.
>> A more likely explanation is that your line has high(ish) error
>>rates. The things the router reports as things like
>> SF (CRC) Errors:
>> RS Corrected:
>> RS Un-Corrected:
>> HEC:
>> Errored Seconds:
>> Severe ES:
>> etc.
>> I understand that the DSLAM in the exchange monitors these things
>>and if the error rate rises "too high" then it increases the SNR to
>>reduce them. The theory being that high error rates mean lots of
>>retransmission and hence slow(er) real speeds and inefficient network use.
>> [I seem to remember reading somewhere that because of the design of
>>the ATM protocol if the ADSL error is uncorrectable by the router
>>then packet retransmission takes place at the TCP level from the IP's
>>kit not from the exchange - so its the load this places on their
>>backbone network they are trying to avoid; as well as trying to give
>>you the fastest possible download experience ]
>> If you have got an error prone line, then a very heavy usage weekend
>>would have increased the numbers of packets passed and consequently
>>the number of errors recorded by the DSLAM. So it reacted.
>> If you can monitor the error rates (Routerstats does a good job if
>>you have the right router) you might get some idea as to the problem.
>>
>
>
>well further to this I said 'what the hell' and phoned up the ISP.
>
>Who said something very strange..
>
>"Your UPSTREAM error rate is very high. That's why it's reset the noise
>margin."
>
>I was seeing no errors at all reported on the router going upstream though.
>
>And precious few downstream either. But I suppose that correctable CRC
>errors that the DSLAM detected would not necessarily be reported back
>to the router..only completely damaged packets that needed
>retransmits..not sure how the ADSL works at that level.
>
>Anyway after mulling that one over, I asked for a noise margin reset
>and currently the line is holding a shade under 6Mbps at a noise margin
>of 6dB. And no interleave, which is a first EVER. I doubt it will hold
>that through the night-time MW transmissions, but it should settle out
>over 5Mbps anyway.
>
>The tech droid was insistent - no idea why - that this was something
>injecting noise at my end. He didn't seem to grasp the basic fact that
>the UPSTREAM signals were weakest at the DSLAM end, and therefore more
>likely to be affected by noise further up the line... a lecture on
>communication theory being of no profit I simply asked that the line be
>reset.
>
>Which happened LESS THAN A MINUTE after I put the phone down.
>
>I am really puzzled by the whole affair. If what he was saying was
>correct, that could only be LF trash of some sort right up at the
>exchange... some kid with a poorly suppressesd motorbike? someone
>fiddling with wires?
>
>
>Mind you, its does suggest a way to bugger up your neighbourhood DSL.
>switch your router OFF and then get a spark transmitter going. When
>they are totally narfed at 24dB margin and crawling, you can reboot
>your router and get a clean almost uncontended connection :-)
>
>Any other thoughts welcome.
>
If your line is clean apart from the overnight drop in SNR caused by MW
interference, you could try my approach.

I use an Ex-Sky DG834GT router (£10 off e-bay) re-flashed with DGTeam's
firmware. This allows you to tweak the SNR target, and the tweak
survives a router restart.

My line (3.5km all overhead 40db attenuation ) varies about 2dB from
best (mornings) to worst (late evening). So a daytime reboot at a 6dB
target gets me 6500 ish in sync speed. The router will normally hold on
down to about 3-3.5db, but any spikes significantly below that will
cause a resync. A restart or line drop in the late evening when MW is
at its worst gets me 4500-5200 ish.

I have tweaked the router to restart at 66% of SNR (so it aims for about
4dB). A spike at night that drives SNR below 3db causes a recovery to
4dB and 6400-6500 which the router can easily hold at night in the
absence of spikes. (current WAN uptime is approaching 1000 hours)

The downside is that if I (or the power company!) restart the router
during the day I get 4dB and see sync speeds in excess of 7000 - knowing
full well that a dropout will happen when it goes dark.

But it all conspires to maintain my BRAS profile at 5500 or 6000 ( a
significant improvement on my neighbours 4000-4500).
--
Invalid

Invalid wrote:
> In message <j1a5l1$62t$(E-Mail Removed)>, The Natural Philosopher
> <(E-Mail Removed)> writes
>> Invalid wrote:
>>> In message <j19215$vvd$(E-Mail Removed)>, The Natural Philosopher
>>> <(E-Mail Removed)> writes
>>>> Seems to be obvious.
>>>>
>>>> Sine my line was fixed I could sustain a really nice 5Mbps at
>>>> around 6- 8dB SNR.. However despite having zero issues at this
>>>> speed, the SNR margin has been progressively raised until its now
>>>>
>>>> Possibly connected with the fact that I spent some serious time this
>>>> weekend watching the BBC HD F1 stream..and listening to test match
>>>> special over the internet? This consuming lots of bandwidth..
>>>>
>>>> Arte BT using SNR margins to control traffic? And if they are, how
>>>> do I prove it, and who can do something about it?
>>>>
>>>> I am fairly sure its outside the ISP's powers to fiddle with this,
>>>> beyond asking if it can be reset..
>>>>
>>>> Comments welcome...
>>> Possibly - would make a fairly good conspiracy theory, but the
>>> relationship between BT Retail (the ISP) and Openreach who own the
>>> ADSL kit that controls the SNR is officially the same as between any
>>> other ISP and Openreach. I doubt that Openreach give BT any more (or
>>> less) control than they give to other ISP's.
>>> A more likely explanation is that your line has high(ish) error
>>> rates. The things the router reports as things like
>>> SF (CRC) Errors:
>>> RS Corrected:
>>> RS Un-Corrected:
>>> HEC:
>>> Errored Seconds:
>>> Severe ES:
>>> etc.
>>> I understand that the DSLAM in the exchange monitors these things
>>> and if the error rate rises "too high" then it increases the SNR to
>>> reduce them. The theory being that high error rates mean lots of
>>> retransmission and hence slow(er) real speeds and inefficient
>>> network use.
>>> [I seem to remember reading somewhere that because of the design of
>>> the ATM protocol if the ADSL error is uncorrectable by the router
>>> then packet retransmission takes place at the TCP level from the
>>> IP's kit not from the exchange - so its the load this places on
>>> their backbone network they are trying to avoid; as well as trying
>>> to give you the fastest possible download experience ]
>>> If you have got an error prone line, then a very heavy usage weekend
>>> would have increased the numbers of packets passed and consequently
>>> the number of errors recorded by the DSLAM. So it reacted.
>>> If you can monitor the error rates (Routerstats does a good job if
>>> you have the right router) you might get some idea as to the problem.
>>>
>>
>>
>> well further to this I said 'what the hell' and phoned up the ISP.
>>
>> Who said something very strange..
>>
>> "Your UPSTREAM error rate is very high. That's why it's reset the
>> noise margin."
>>
>> I was seeing no errors at all reported on the router going upstream
>> though.
>>
>> And precious few downstream either. But I suppose that correctable CRC
>> errors that the DSLAM detected would not necessarily be reported back
>> to the router..only completely damaged packets that needed
>> retransmits..not sure how the ADSL works at that level.
>>
>> Anyway after mulling that one over, I asked for a noise margin reset
>> and currently the line is holding a shade under 6Mbps at a noise
>> margin of 6dB. And no interleave, which is a first EVER. I doubt it
>> will hold that through the night-time MW transmissions, but it should
>> settle out over 5Mbps anyway.
>>
>> The tech droid was insistent - no idea why - that this was something
>> injecting noise at my end. He didn't seem to grasp the basic fact that
>> the UPSTREAM signals were weakest at the DSLAM end, and therefore
>> more likely to be affected by noise further up the line... a lecture
>> on communication theory being of no profit I simply asked that the
>> line be reset.
>>
>> Which happened LESS THAN A MINUTE after I put the phone down.
>>
>> I am really puzzled by the whole affair. If what he was saying was
>> correct, that could only be LF trash of some sort right up at the
>> exchange... some kid with a poorly suppressesd motorbike? someone
>> fiddling with wires?
>>
>>
>> Mind you, its does suggest a way to bugger up your neighbourhood DSL.
>> switch your router OFF and then get a spark transmitter going. When
>> they are totally narfed at 24dB margin and crawling, you can reboot
>> your router and get a clean almost uncontended connection :-)
>>
>> Any other thoughts welcome.
>>
> If your line is clean apart from the overnight drop in SNR caused by MW
> interference, you could try my approach.
>
> I use an Ex-Sky DG834GT router (£10 off e-bay) re-flashed with DGTeam's
> firmware. This allows you to tweak the SNR target, and the tweak
> survives a router restart.
>
> My line (3.5km all overhead 40db attenuation ) varies about 2dB from
> best (mornings) to worst (late evening). So a daytime reboot at a 6dB
> target gets me 6500 ish in sync speed. The router will normally hold on
> down to about 3-3.5db, but any spikes significantly below that will
> cause a resync. A restart or line drop in the late evening when MW is
> at its worst gets me 4500-5200 ish.
>
> I have tweaked the router to restart at 66% of SNR (so it aims for about
> 4dB). A spike at night that drives SNR below 3db causes a recovery to
> 4dB and 6400-6500 which the router can easily hold at night in the
> absence of spikes. (current WAN uptime is approaching 1000 hours)
>
> The downside is that if I (or the power company!) restart the router
> during the day I get 4dB and see sync speeds in excess of 7000 - knowing
> full well that a dropout will happen when it goes dark.
>
> But it all conspires to maintain my BRAS profile at 5500 or 6000 ( a
> significant improvement on my neighbours 4000-4500).

I am doing that sort of thing already. Actually all I want is 4000 BRAS
so I can watch more F1 in HD. Brilliant!


Burt if what I think has happened did happen, then its basically a bum
algo in the BT DSLAM stuff..you shouldn't lose downstream noise margin
for weeks just because you get a bit of corrupted upstream packets..

...which then vanishes..

On Wed, 03 Aug 2011 11:55:59 +0100, The Natural Philosopher
<(E-Mail Removed)> wrote:

>Invalid wrote:
>> In message <j1a5l1$62t$(E-Mail Removed)>, The Natural Philosopher
>> <(E-Mail Removed)> writes
>>> Invalid wrote:
>>>> In message <j19215$vvd$(E-Mail Removed)>, The Natural Philosopher
>>>> <(E-Mail Removed)> writes
>>>>> Seems to be obvious.
>>>>>
>>>>> Sine my line was fixed I could sustain a really nice 5Mbps at
>>>>> around 6- 8dB SNR.. However despite having zero issues at this
>>>>> speed, the SNR margin has been progressively raised until its now
>>>>>
>>>>> Possibly connected with the fact that I spent some serious time this
>>>>> weekend watching the BBC HD F1 stream..and listening to test match
>>>>> special over the internet? This consuming lots of bandwidth..
>>>>>
>>>>> Arte BT using SNR margins to control traffic? And if they are, how
>>>>> do I prove it, and who can do something about it?
>>>>>
>>>>> I am fairly sure its outside the ISP's powers to fiddle with this,
>>>>> beyond asking if it can be reset..
>>>>>
>>>>> Comments welcome...
>>>> Possibly - would make a fairly good conspiracy theory, but the
>>>> relationship between BT Retail (the ISP) and Openreach who own the
>>>> ADSL kit that controls the SNR is officially the same as between any
>>>> other ISP and Openreach. I doubt that Openreach give BT any more (or
>>>> less) control than they give to other ISP's.
>>>> A more likely explanation is that your line has high(ish) error
>>>> rates. The things the router reports as things like
>>>> SF (CRC) Errors:
>>>> RS Corrected:
>>>> RS Un-Corrected:
>>>> HEC:
>>>> Errored Seconds:
>>>> Severe ES:
>>>> etc.
>>>> I understand that the DSLAM in the exchange monitors these things
>>>> and if the error rate rises "too high" then it increases the SNR to
>>>> reduce them. The theory being that high error rates mean lots of
>>>> retransmission and hence slow(er) real speeds and inefficient
>>>> network use.
>>>> [I seem to remember reading somewhere that because of the design of
>>>> the ATM protocol if the ADSL error is uncorrectable by the router
>>>> then packet retransmission takes place at the TCP level from the
>>>> IP's kit not from the exchange - so its the load this places on
>>>> their backbone network they are trying to avoid; as well as trying
>>>> to give you the fastest possible download experience ]
>>>> If you have got an error prone line, then a very heavy usage weekend
>>>> would have increased the numbers of packets passed and consequently
>>>> the number of errors recorded by the DSLAM. So it reacted.
>>>> If you can monitor the error rates (Routerstats does a good job if
>>>> you have the right router) you might get some idea as to the problem.
>>>>
>>>
>>>
>>> well further to this I said 'what the hell' and phoned up the ISP.
>>>
>>> Who said something very strange..
>>>
>>> "Your UPSTREAM error rate is very high. That's why it's reset the
>>> noise margin."
>>>
>>> I was seeing no errors at all reported on the router going upstream
>>> though.
>>>
>>> And precious few downstream either. But I suppose that correctable CRC
>>> errors that the DSLAM detected would not necessarily be reported back
>>> to the router..only completely damaged packets that needed
>>> retransmits..not sure how the ADSL works at that level.
>>>
>>> Anyway after mulling that one over, I asked for a noise margin reset
>>> and currently the line is holding a shade under 6Mbps at a noise
>>> margin of 6dB. And no interleave, which is a first EVER. I doubt it
>>> will hold that through the night-time MW transmissions, but it should
>>> settle out over 5Mbps anyway.
>>>
>>> The tech droid was insistent - no idea why - that this was something
>>> injecting noise at my end. He didn't seem to grasp the basic fact that
>>> the UPSTREAM signals were weakest at the DSLAM end, and therefore
>>> more likely to be affected by noise further up the line... a lecture
>>> on communication theory being of no profit I simply asked that the
>>> line be reset.
>>>
>>> Which happened LESS THAN A MINUTE after I put the phone down.
>>>
>>> I am really puzzled by the whole affair. If what he was saying was
>>> correct, that could only be LF trash of some sort right up at the
>>> exchange... some kid with a poorly suppressesd motorbike? someone
>>> fiddling with wires?
>>>
>>>
>>> Mind you, its does suggest a way to bugger up your neighbourhood DSL.
>>> switch your router OFF and then get a spark transmitter going. When
>>> they are totally narfed at 24dB margin and crawling, you can reboot
>>> your router and get a clean almost uncontended connection :-)
>>>
>>> Any other thoughts welcome.
>>>
>> If your line is clean apart from the overnight drop in SNR caused by MW
>> interference, you could try my approach.
>>
>> I use an Ex-Sky DG834GT router (£10 off e-bay) re-flashed with DGTeam's
>> firmware. This allows you to tweak the SNR target, and the tweak
>> survives a router restart.
>>
>> My line (3.5km all overhead 40db attenuation ) varies about 2dB from
>> best (mornings) to worst (late evening). So a daytime reboot at a 6dB
>> target gets me 6500 ish in sync speed. The router will normally hold on
>> down to about 3-3.5db, but any spikes significantly below that will
>> cause a resync. A restart or line drop in the late evening when MW is
>> at its worst gets me 4500-5200 ish.
>>
>> I have tweaked the router to restart at 66% of SNR (so it aims for about
>> 4dB). A spike at night that drives SNR below 3db causes a recovery to
>> 4dB and 6400-6500 which the router can easily hold at night in the
>> absence of spikes. (current WAN uptime is approaching 1000 hours)
>>
>> The downside is that if I (or the power company!) restart the router
>> during the day I get 4dB and see sync speeds in excess of 7000 - knowing
>> full well that a dropout will happen when it goes dark.
>>
>> But it all conspires to maintain my BRAS profile at 5500 or 6000 ( a
>> significant improvement on my neighbours 4000-4500).
>
>I am doing that sort of thing already. Actually all I want is 4000 BRAS
>so I can watch more F1 in HD. Brilliant!
>
>
>Burt if what I think has happened did happen, then its basically a bum
>algo in the BT DSLAM stuff..you shouldn't lose downstream noise margin
>for weeks just because you get a bit of corrupted upstream packets..
>
>..which then vanishes..

FWIW I've seen target margins rise when the line is apparently
(almost) error free. Many people will scream I am wrong but I have
observed this too. For years my line was stuck at 9 and 12dB but
recently has dropped back to 6.

The algorithms that control the noise margin are secret and I've never
met anyone who knows what they are.

I've not had a huge amount of success tweaking the noise margin with
the dgteam firmware. It just causes instability for me but YMMV.

You might be better off with interleaving on since forward error
correction will then be enabled too.

And I dream of an IP profile of 4M. The best I've got in recent years
is 2.5M.
--
(\__/) M.
(='.'=) Due to the amount of spam posted via googlegroups and
(")_(") their inaction to the problem. I am blocking some articles
posted from there. If you wish your postings to be seen by
everyone you will need use a different method of posting.

Mark wrote:
> On Wed, 03 Aug 2011 11:55:59 +0100, The Natural Philosopher
> <(E-Mail Removed)> wrote:
>
>> Invalid wrote:
>>> In message <j1a5l1$62t$(E-Mail Removed)>, The Natural Philosopher
>>> <(E-Mail Removed)> writes
>>>> Invalid wrote:
>>>>> In message <j19215$vvd$(E-Mail Removed)>, The Natural Philosopher
>>>>> <(E-Mail Removed)> writes
>>>>>> Seems to be obvious.
>>>>>>
>>>>>> Sine my line was fixed I could sustain a really nice 5Mbps at
>>>>>> around 6- 8dB SNR.. However despite having zero issues at this
>>>>>> speed, the SNR margin has been progressively raised until its now
>>>>>>
>>>>>> Possibly connected with the fact that I spent some serious time this
>>>>>> weekend watching the BBC HD F1 stream..and listening to test match
>>>>>> special over the internet? This consuming lots of bandwidth..
>>>>>>
>>>>>> Arte BT using SNR margins to control traffic? And if they are, how
>>>>>> do I prove it, and who can do something about it?
>>>>>>
>>>>>> I am fairly sure its outside the ISP's powers to fiddle with this,
>>>>>> beyond asking if it can be reset..
>>>>>>
>>>>>> Comments welcome...
>>>>> Possibly - would make a fairly good conspiracy theory, but the
>>>>> relationship between BT Retail (the ISP) and Openreach who own the
>>>>> ADSL kit that controls the SNR is officially the same as between any
>>>>> other ISP and Openreach. I doubt that Openreach give BT any more (or
>>>>> less) control than they give to other ISP's.
>>>>> A more likely explanation is that your line has high(ish) error
>>>>> rates. The things the router reports as things like
>>>>> SF (CRC) Errors:
>>>>> RS Corrected:
>>>>> RS Un-Corrected:
>>>>> HEC:
>>>>> Errored Seconds:
>>>>> Severe ES:
>>>>> etc.
>>>>> I understand that the DSLAM in the exchange monitors these things
>>>>> and if the error rate rises "too high" then it increases the SNR to
>>>>> reduce them. The theory being that high error rates mean lots of
>>>>> retransmission and hence slow(er) real speeds and inefficient
>>>>> network use.
>>>>> [I seem to remember reading somewhere that because of the design of
>>>>> the ATM protocol if the ADSL error is uncorrectable by the router
>>>>> then packet retransmission takes place at the TCP level from the
>>>>> IP's kit not from the exchange - so its the load this places on
>>>>> their backbone network they are trying to avoid; as well as trying
>>>>> to give you the fastest possible download experience ]
>>>>> If you have got an error prone line, then a very heavy usage weekend
>>>>> would have increased the numbers of packets passed and consequently
>>>>> the number of errors recorded by the DSLAM. So it reacted.
>>>>> If you can monitor the error rates (Routerstats does a good job if
>>>>> you have the right router) you might get some idea as to the problem.
>>>>>
>>>>
>>>> well further to this I said 'what the hell' and phoned up the ISP.
>>>>
>>>> Who said something very strange..
>>>>
>>>> "Your UPSTREAM error rate is very high. That's why it's reset the
>>>> noise margin."
>>>>
>>>> I was seeing no errors at all reported on the router going upstream
>>>> though.
>>>>
>>>> And precious few downstream either. But I suppose that correctable CRC
>>>> errors that the DSLAM detected would not necessarily be reported back
>>>> to the router..only completely damaged packets that needed
>>>> retransmits..not sure how the ADSL works at that level.
>>>>
>>>> Anyway after mulling that one over, I asked for a noise margin reset
>>>> and currently the line is holding a shade under 6Mbps at a noise
>>>> margin of 6dB. And no interleave, which is a first EVER. I doubt it
>>>> will hold that through the night-time MW transmissions, but it should
>>>> settle out over 5Mbps anyway.
>>>>
>>>> The tech droid was insistent - no idea why - that this was something
>>>> injecting noise at my end. He didn't seem to grasp the basic fact that
>>>> the UPSTREAM signals were weakest at the DSLAM end, and therefore
>>>> more likely to be affected by noise further up the line... a lecture
>>>> on communication theory being of no profit I simply asked that the
>>>> line be reset.
>>>>
>>>> Which happened LESS THAN A MINUTE after I put the phone down.
>>>>
>>>> I am really puzzled by the whole affair. If what he was saying was
>>>> correct, that could only be LF trash of some sort right up at the
>>>> exchange... some kid with a poorly suppressesd motorbike? someone
>>>> fiddling with wires?
>>>>
>>>>
>>>> Mind you, its does suggest a way to bugger up your neighbourhood DSL.
>>>> switch your router OFF and then get a spark transmitter going. When
>>>> they are totally narfed at 24dB margin and crawling, you can reboot
>>>> your router and get a clean almost uncontended connection :-)
>>>>
>>>> Any other thoughts welcome.
>>>>
>>> If your line is clean apart from the overnight drop in SNR caused by MW
>>> interference, you could try my approach.
>>>
>>> I use an Ex-Sky DG834GT router (£10 off e-bay) re-flashed with DGTeam's
>>> firmware. This allows you to tweak the SNR target, and the tweak
>>> survives a router restart.
>>>
>>> My line (3.5km all overhead 40db attenuation ) varies about 2dB from
>>> best (mornings) to worst (late evening). So a daytime reboot at a 6dB
>>> target gets me 6500 ish in sync speed. The router will normally hold on
>>> down to about 3-3.5db, but any spikes significantly below that will
>>> cause a resync. A restart or line drop in the late evening when MW is
>>> at its worst gets me 4500-5200 ish.
>>>
>>> I have tweaked the router to restart at 66% of SNR (so it aims for about
>>> 4dB). A spike at night that drives SNR below 3db causes a recovery to
>>> 4dB and 6400-6500 which the router can easily hold at night in the
>>> absence of spikes. (current WAN uptime is approaching 1000 hours)
>>>
>>> The downside is that if I (or the power company!) restart the router
>>> during the day I get 4dB and see sync speeds in excess of 7000 - knowing
>>> full well that a dropout will happen when it goes dark.
>>>
>>> But it all conspires to maintain my BRAS profile at 5500 or 6000 ( a
>>> significant improvement on my neighbours 4000-4500).
>> I am doing that sort of thing already. Actually all I want is 4000 BRAS
>> so I can watch more F1 in HD. Brilliant!
>>
>>
>> Burt if what I think has happened did happen, then its basically a bum
>> algo in the BT DSLAM stuff..you shouldn't lose downstream noise margin
>> for weeks just because you get a bit of corrupted upstream packets..
>>
>> ..which then vanishes..
>
> FWIW I've seen target margins rise when the line is apparently
> (almost) error free. Many people will scream I am wrong but I have
> observed this too. For years my line was stuck at 9 and 12dB but
> recently has dropped back to 6.
>
> The algorithms that control the noise margin are secret and I've never
> met anyone who knows what they are.
>
> I've not had a huge amount of success tweaking the noise margin with
> the dgteam firmware. It just causes instability for me but YMMV.
>
> You might be better off with interleaving on since forward error
> correction will then be enabled too.
>
> And I dream of an IP profile of 4M. The best I've got in recent years
> is 2.5M.

Well Mark, at 5:30am Monday the carrier went COMPLETELY.

It came back at 1pm today after the usual 'it must be your router, there
is nothing wrong with the line' shenanigans from the ISP. And Me
insisting it wasn't ME. Like what sport of changes would I have made at
5:30 am?


I got a new router (waste of £30) and it said the same. 'No signal'...

Now sizzling at 6.52Mbps..which its never ever achieved before..(and
probably wont hold for long, either)

I am guessing that something was failing at the exchange, and FINALLY
they have swapped it out..

The Natural Philosopher <(E-Mail Removed)> considered Tue, 09 Aug
2011 17:02:23 +0100 the perfect time to write:

>Mark wrote:
>> On Wed, 03 Aug 2011 11:55:59 +0100, The Natural Philosopher
>> <(E-Mail Removed)> wrote:
>>
>>> Invalid wrote:
>>>> In message <j1a5l1$62t$(E-Mail Removed)>, The Natural Philosopher
>>>> <(E-Mail Removed)> writes
>>>>> Invalid wrote:
>>>>>> In message <j19215$vvd$(E-Mail Removed)>, The Natural Philosopher
>>>>>> <(E-Mail Removed)> writes
>>>>>>> Seems to be obvious.
>>>>>>>
>>>>>>> Sine my line was fixed I could sustain a really nice 5Mbps at
>>>>>>> around 6- 8dB SNR.. However despite having zero issues at this
>>>>>>> speed, the SNR margin has been progressively raised until its now
>>>>>>>
>>>>>>> Possibly connected with the fact that I spent some serious time this
>>>>>>> weekend watching the BBC HD F1 stream..and listening to test match
>>>>>>> special over the internet? This consuming lots of bandwidth..
>>>>>>>
>>>>>>> Arte BT using SNR margins to control traffic? And if they are, how
>>>>>>> do I prove it, and who can do something about it?
>>>>>>>
>>>>>>> I am fairly sure its outside the ISP's powers to fiddle with this,
>>>>>>> beyond asking if it can be reset..
>>>>>>>
>>>>>>> Comments welcome...
>>>>>> Possibly - would make a fairly good conspiracy theory, but the
>>>>>> relationship between BT Retail (the ISP) and Openreach who own the
>>>>>> ADSL kit that controls the SNR is officially the same as between any
>>>>>> other ISP and Openreach. I doubt that Openreach give BT any more (or
>>>>>> less) control than they give to other ISP's.
>>>>>> A more likely explanation is that your line has high(ish) error
>>>>>> rates. The things the router reports as things like
>>>>>> SF (CRC) Errors:
>>>>>> RS Corrected:
>>>>>> RS Un-Corrected:
>>>>>> HEC:
>>>>>> Errored Seconds:
>>>>>> Severe ES:
>>>>>> etc.
>>>>>> I understand that the DSLAM in the exchange monitors these things
>>>>>> and if the error rate rises "too high" then it increases the SNR to
>>>>>> reduce them. The theory being that high error rates mean lots of
>>>>>> retransmission and hence slow(er) real speeds and inefficient
>>>>>> network use.
>>>>>> [I seem to remember reading somewhere that because of the design of
>>>>>> the ATM protocol if the ADSL error is uncorrectable by the router
>>>>>> then packet retransmission takes place at the TCP level from the
>>>>>> IP's kit not from the exchange - so its the load this places on
>>>>>> their backbone network they are trying to avoid; as well as trying
>>>>>> to give you the fastest possible download experience ]
>>>>>> If you have got an error prone line, then a very heavy usage weekend
>>>>>> would have increased the numbers of packets passed and consequently
>>>>>> the number of errors recorded by the DSLAM. So it reacted.
>>>>>> If you can monitor the error rates (Routerstats does a good job if
>>>>>> you have the right router) you might get some idea as to the problem.
>>>>>>
>>>>>
>>>>> well further to this I said 'what the hell' and phoned up the ISP.
>>>>>
>>>>> Who said something very strange..
>>>>>
>>>>> "Your UPSTREAM error rate is very high. That's why it's reset the
>>>>> noise margin."
>>>>>
>>>>> I was seeing no errors at all reported on the router going upstream
>>>>> though.
>>>>>
>>>>> And precious few downstream either. But I suppose that correctable CRC
>>>>> errors that the DSLAM detected would not necessarily be reported back
>>>>> to the router..only completely damaged packets that needed
>>>>> retransmits..not sure how the ADSL works at that level.
>>>>>
>>>>> Anyway after mulling that one over, I asked for a noise margin reset
>>>>> and currently the line is holding a shade under 6Mbps at a noise
>>>>> margin of 6dB. And no interleave, which is a first EVER. I doubt it
>>>>> will hold that through the night-time MW transmissions, but it should
>>>>> settle out over 5Mbps anyway.
>>>>>
>>>>> The tech droid was insistent - no idea why - that this was something
>>>>> injecting noise at my end. He didn't seem to grasp the basic fact that
>>>>> the UPSTREAM signals were weakest at the DSLAM end, and therefore
>>>>> more likely to be affected by noise further up the line... a lecture
>>>>> on communication theory being of no profit I simply asked that the
>>>>> line be reset.
>>>>>
>>>>> Which happened LESS THAN A MINUTE after I put the phone down.
>>>>>
>>>>> I am really puzzled by the whole affair. If what he was saying was
>>>>> correct, that could only be LF trash of some sort right up at the
>>>>> exchange... some kid with a poorly suppressesd motorbike? someone
>>>>> fiddling with wires?
>>>>>
>>>>>
>>>>> Mind you, its does suggest a way to bugger up your neighbourhood DSL.
>>>>> switch your router OFF and then get a spark transmitter going. When
>>>>> they are totally narfed at 24dB margin and crawling, you can reboot
>>>>> your router and get a clean almost uncontended connection :-)
>>>>>
>>>>> Any other thoughts welcome.
>>>>>
>>>> If your line is clean apart from the overnight drop in SNR caused by MW
>>>> interference, you could try my approach.
>>>>
>>>> I use an Ex-Sky DG834GT router (£10 off e-bay) re-flashed with DGTeam's
>>>> firmware. This allows you to tweak the SNR target, and the tweak
>>>> survives a router restart.
>>>>
>>>> My line (3.5km all overhead 40db attenuation ) varies about 2dB from
>>>> best (mornings) to worst (late evening). So a daytime reboot at a 6dB
>>>> target gets me 6500 ish in sync speed. The router will normally hold on
>>>> down to about 3-3.5db, but any spikes significantly below that will
>>>> cause a resync. A restart or line drop in the late evening when MW is
>>>> at its worst gets me 4500-5200 ish.
>>>>
>>>> I have tweaked the router to restart at 66% of SNR (so it aims for about
>>>> 4dB). A spike at night that drives SNR below 3db causes a recovery to
>>>> 4dB and 6400-6500 which the router can easily hold at night in the
>>>> absence of spikes. (current WAN uptime is approaching 1000 hours)
>>>>
>>>> The downside is that if I (or the power company!) restart the router
>>>> during the day I get 4dB and see sync speeds in excess of 7000 - knowing
>>>> full well that a dropout will happen when it goes dark.
>>>>
>>>> But it all conspires to maintain my BRAS profile at 5500 or 6000 ( a
>>>> significant improvement on my neighbours 4000-4500).
>>> I am doing that sort of thing already. Actually all I want is 4000 BRAS
>>> so I can watch more F1 in HD. Brilliant!
>>>
>>>
>>> Burt if what I think has happened did happen, then its basically a bum
>>> algo in the BT DSLAM stuff..you shouldn't lose downstream noise margin
>>> for weeks just because you get a bit of corrupted upstream packets..
>>>
>>> ..which then vanishes..
>>
>> FWIW I've seen target margins rise when the line is apparently
>> (almost) error free. Many people will scream I am wrong but I have
>> observed this too. For years my line was stuck at 9 and 12dB but
>> recently has dropped back to 6.
>>
>> The algorithms that control the noise margin are secret and I've never
>> met anyone who knows what they are.
>>
>> I've not had a huge amount of success tweaking the noise margin with
>> the dgteam firmware. It just causes instability for me but YMMV.
>>
>> You might be better off with interleaving on since forward error
>> correction will then be enabled too.
>>
>> And I dream of an IP profile of 4M. The best I've got in recent years
>> is 2.5M.
>
>Well Mark, at 5:30am Monday the carrier went COMPLETELY.
>
>It came back at 1pm today after the usual 'it must be your router, there
>is nothing wrong with the line' shenanigans from the ISP. And Me
>insisting it wasn't ME. Like what sport of changes would I have made at
>5:30 am?
>
>
>I got a new router (waste of £30) and it said the same. 'No signal'...
>
>Now sizzling at 6.52Mbps..which its never ever achieved before..(and
>probably wont hold for long, either)
>
>I am guessing that something was failing at the exchange, and FINALLY
>they have swapped it out..
>
If you're really really lucky, the something that was failing at the
exchange will be the same something that was restricting your speed,
and the 6.52Mbps will stick (or at least only reduce a little).

Phil W Lee wrote:
> The Natural Philosopher <(E-Mail Removed)> considered Tue, 09 Aug
> 2011 17:02:23 +0100 the perfect time to write:
>
>> Mark wrote:
>>> On Wed, 03 Aug 2011 11:55:59 +0100, The Natural Philosopher
>>> <(E-Mail Removed)> wrote:
>>>
>>>> Invalid wrote:
>>>>> In message <j1a5l1$62t$(E-Mail Removed)>, The Natural Philosopher
>>>>> <(E-Mail Removed)> writes
>>>>>> Invalid wrote:
>>>>>>> In message <j19215$vvd$(E-Mail Removed)>, The Natural Philosopher
>>>>>>> <(E-Mail Removed)> writes
>>>>>>>> Seems to be obvious.
>>>>>>>>
>>>>>>>> Sine my line was fixed I could sustain a really nice 5Mbps at
>>>>>>>> around 6- 8dB SNR.. However despite having zero issues at this
>>>>>>>> speed, the SNR margin has been progressively raised until its now
>>>>>>>>
>>>>>>>> Possibly connected with the fact that I spent some serious time this
>>>>>>>> weekend watching the BBC HD F1 stream..and listening to test match
>>>>>>>> special over the internet? This consuming lots of bandwidth..
>>>>>>>>
>>>>>>>> Arte BT using SNR margins to control traffic? And if they are, how
>>>>>>>> do I prove it, and who can do something about it?
>>>>>>>>
>>>>>>>> I am fairly sure its outside the ISP's powers to fiddle with this,
>>>>>>>> beyond asking if it can be reset..
>>>>>>>>
>>>>>>>> Comments welcome...
>>>>>>> Possibly - would make a fairly good conspiracy theory, but the
>>>>>>> relationship between BT Retail (the ISP) and Openreach who own the
>>>>>>> ADSL kit that controls the SNR is officially the same as between any
>>>>>>> other ISP and Openreach. I doubt that Openreach give BT any more (or
>>>>>>> less) control than they give to other ISP's.
>>>>>>> A more likely explanation is that your line has high(ish) error
>>>>>>> rates. The things the router reports as things like
>>>>>>> SF (CRC) Errors:
>>>>>>> RS Corrected:
>>>>>>> RS Un-Corrected:
>>>>>>> HEC:
>>>>>>> Errored Seconds:
>>>>>>> Severe ES:
>>>>>>> etc.
>>>>>>> I understand that the DSLAM in the exchange monitors these things
>>>>>>> and if the error rate rises "too high" then it increases the SNR to
>>>>>>> reduce them. The theory being that high error rates mean lots of
>>>>>>> retransmission and hence slow(er) real speeds and inefficient
>>>>>>> network use.
>>>>>>> [I seem to remember reading somewhere that because of the design of
>>>>>>> the ATM protocol if the ADSL error is uncorrectable by the router
>>>>>>> then packet retransmission takes place at the TCP level from the
>>>>>>> IP's kit not from the exchange - so its the load this places on
>>>>>>> their backbone network they are trying to avoid; as well as trying
>>>>>>> to give you the fastest possible download experience ]
>>>>>>> If you have got an error prone line, then a very heavy usage weekend
>>>>>>> would have increased the numbers of packets passed and consequently
>>>>>>> the number of errors recorded by the DSLAM. So it reacted.
>>>>>>> If you can monitor the error rates (Routerstats does a good job if
>>>>>>> you have the right router) you might get some idea as to the problem.
>>>>>>>
>>>>>> well further to this I said 'what the hell' and phoned up the ISP.
>>>>>>
>>>>>> Who said something very strange..
>>>>>>
>>>>>> "Your UPSTREAM error rate is very high. That's why it's reset the
>>>>>> noise margin."
>>>>>>
>>>>>> I was seeing no errors at all reported on the router going upstream
>>>>>> though.
>>>>>>
>>>>>> And precious few downstream either. But I suppose that correctable CRC
>>>>>> errors that the DSLAM detected would not necessarily be reported back
>>>>>> to the router..only completely damaged packets that needed
>>>>>> retransmits..not sure how the ADSL works at that level.
>>>>>>
>>>>>> Anyway after mulling that one over, I asked for a noise margin reset
>>>>>> and currently the line is holding a shade under 6Mbps at a noise
>>>>>> margin of 6dB. And no interleave, which is a first EVER. I doubt it
>>>>>> will hold that through the night-time MW transmissions, but it should
>>>>>> settle out over 5Mbps anyway.
>>>>>>
>>>>>> The tech droid was insistent - no idea why - that this was something
>>>>>> injecting noise at my end. He didn't seem to grasp the basic fact that
>>>>>> the UPSTREAM signals were weakest at the DSLAM end, and therefore
>>>>>> more likely to be affected by noise further up the line... a lecture
>>>>>> on communication theory being of no profit I simply asked that the
>>>>>> line be reset.
>>>>>>
>>>>>> Which happened LESS THAN A MINUTE after I put the phone down.
>>>>>>
>>>>>> I am really puzzled by the whole affair. If what he was saying was
>>>>>> correct, that could only be LF trash of some sort right up at the
>>>>>> exchange... some kid with a poorly suppressesd motorbike? someone
>>>>>> fiddling with wires?
>>>>>>
>>>>>>
>>>>>> Mind you, its does suggest a way to bugger up your neighbourhood DSL.
>>>>>> switch your router OFF and then get a spark transmitter going. When
>>>>>> they are totally narfed at 24dB margin and crawling, you can reboot
>>>>>> your router and get a clean almost uncontended connection :-)
>>>>>>
>>>>>> Any other thoughts welcome.
>>>>>>
>>>>> If your line is clean apart from the overnight drop in SNR caused by MW
>>>>> interference, you could try my approach.
>>>>>
>>>>> I use an Ex-Sky DG834GT router (£10 off e-bay) re-flashed with DGTeam's
>>>>> firmware. This allows you to tweak the SNR target, and the tweak
>>>>> survives a router restart.
>>>>>
>>>>> My line (3.5km all overhead 40db attenuation ) varies about 2dB from
>>>>> best (mornings) to worst (late evening). So a daytime reboot at a 6dB
>>>>> target gets me 6500 ish in sync speed. The router will normally hold on
>>>>> down to about 3-3.5db, but any spikes significantly below that will
>>>>> cause a resync. A restart or line drop in the late evening when MW is
>>>>> at its worst gets me 4500-5200 ish.
>>>>>
>>>>> I have tweaked the router to restart at 66% of SNR (so it aims for about
>>>>> 4dB). A spike at night that drives SNR below 3db causes a recovery to
>>>>> 4dB and 6400-6500 which the router can easily hold at night in the
>>>>> absence of spikes. (current WAN uptime is approaching 1000 hours)
>>>>>
>>>>> The downside is that if I (or the power company!) restart the router
>>>>> during the day I get 4dB and see sync speeds in excess of 7000 - knowing
>>>>> full well that a dropout will happen when it goes dark.
>>>>>
>>>>> But it all conspires to maintain my BRAS profile at 5500 or 6000 ( a
>>>>> significant improvement on my neighbours 4000-4500).
>>>> I am doing that sort of thing already. Actually all I want is 4000 BRAS
>>>> so I can watch more F1 in HD. Brilliant!
>>>>
>>>>
>>>> Burt if what I think has happened did happen, then its basically a bum
>>>> algo in the BT DSLAM stuff..you shouldn't lose downstream noise margin
>>>> for weeks just because you get a bit of corrupted upstream packets..
>>>>
>>>> ..which then vanishes..
>>> FWIW I've seen target margins rise when the line is apparently
>>> (almost) error free. Many people will scream I am wrong but I have
>>> observed this too. For years my line was stuck at 9 and 12dB but
>>> recently has dropped back to 6.
>>>
>>> The algorithms that control the noise margin are secret and I've never
>>> met anyone who knows what they are.
>>>
>>> I've not had a huge amount of success tweaking the noise margin with
>>> the dgteam firmware. It just causes instability for me but YMMV.
>>>
>>> You might be better off with interleaving on since forward error
>>> correction will then be enabled too.
>>>
>>> And I dream of an IP profile of 4M. The best I've got in recent years
>>> is 2.5M.
>> Well Mark, at 5:30am Monday the carrier went COMPLETELY.
>>
>> It came back at 1pm today after the usual 'it must be your router, there
>> is nothing wrong with the line' shenanigans from the ISP. And Me
>> insisting it wasn't ME. Like what sport of changes would I have made at
>> 5:30 am?
>>
>>
>> I got a new router (waste of £30) and it said the same. 'No signal'...
>>
>> Now sizzling at 6.52Mbps..which its never ever achieved before..(and
>> probably wont hold for long, either)
>>
>> I am guessing that something was failing at the exchange, and FINALLY
>> they have swapped it out..
>>
> If you're really really lucky, the something that was failing at the
> exchange will be the same something that was restricting your speed,
> and the 6.52Mbps will stick (or at least only reduce a little).

It is remarkable that I get that.

I've had a slew of issues that led to a complete line switch and much
less attenuation, but all the final resolutions have been 'at the
exchange' and with no word by Openreach as to what they consisted of.

I probably won't hold 6.2 Mbps through the night: CRC errors are piling
up a bit.

"Mark" <(E-Mail Removed)> wrote in message
news:(E-Mail Removed)...
> FWIW I've seen target margins rise when the line is apparently
> (almost) error free. Many people will scream I am wrong but I have
> observed this too. For years my line was stuck at 9 and 12dB but
> recently has dropped back to 6.
>
> The algorithms that control the noise margin are secret and I've never
> met anyone who knows what they are.

I sometimes wonder *if* anyone knows what they are.

My normal 6dB margin went up to 12 after a line fault a couple of years ago,
and stayed there for several months. Then I lost service because of a MUX
fault in the exchange, and whaddya know, on restoration I was suddenly back
to 6dB. Coincidence - surely not.

Earlier this year the margin went up to 9dB for no apparent reason. A few
weeks later, I had occasion to try a mate's router on my line to prove it
faulty or otherwise, and it connected at 6dB, where it's remained since.
Again, I can't believe it was coincidence.

Conclusion: if the DSLAM is playing silly buggers, try surprising it.