Lies, damned lies and router statistics.

This might be sub-titled SWMBO technophobia. A friend who is not very
technically minded asked me to recommend a new ADSL router. He was running a
2Wire 2700HGV because he is on a rural exchange and has a long line. This
has given excellent service almost non stop for three years. My friend's
SWMBO does not like wires and is very fussy about the colour of domestic
items. The arrival of SWMBO's shiny new laptop with "N" wireless capability
meant that the means this had to accommodate. I had a suitable "N" wireless
access point but his was rejected because "there are too many wires". Having
successfully used a Netgear D834G with DG Team firmware I decided a Netgear
DGN2000 with DG team firmware might fit the bill. To prove the concept I
took a spare DG834G along and fitted to see how it performed. SWMBO did not
like the white colour but agreed that a black one would be acceptable.



After installing the DG834G I checked the statistics and found that the
reported line attenuation was 8dB more than the 2Wire reported. Despite this
the DG834G synced at the same speed as the 2Wire and remained connected for
more than ten days non stop. Having satisfied myself that the Netgear would
work I obtained a DGN2000 for my friend. When I installed it and it worked
exactly as I had expected. The only puzzling thing is that the DGN2000
reported a line attenuation 8dB less than the DG834G. This has also run non
stop for ten days. So my question is why should two routers from the same
manufacturer with, presumably very ADSL circuitry and software, report such
a different line attenuation? Does the router measure the line attenuation
or does it get measured by the DSLAM and get passed to the router? In the
past I have done comparative tests on two or three router modems on the same
line and the difference between the reported line attenuation has never been
more than a dB or so. Should we really believe what the hardware tells us or
regard it as fiction?



Peter Crosland

"Peter Crosland" <(E-Mail Removed)> wrote in message
news:(E-Mail Removed) o.uk...
> This might be sub-titled SWMBO technophobia. A friend who is not very
> technically minded asked me to recommend a new ADSL router. He was running
> a 2Wire 2700HGV because he is on a rural exchange and has a long line.
> This has given excellent service almost non stop for three years. My
> friend's SWMBO does not like wires and is very fussy about the colour of
> domestic items. The arrival of SWMBO's shiny new laptop with "N" wireless
> capability meant that the means this had to accommodate. I had a suitable
> "N" wireless access point but his was rejected because "there are too many
> wires". Having successfully used a Netgear D834G with DG Team firmware I
> decided a Netgear DGN2000 with DG team firmware might fit the bill. To
> prove the concept I took a spare DG834G along and fitted to see how it
> performed. SWMBO did not like the white colour but agreed that a black one
> would be acceptable.
>
>
>
> After installing the DG834G I checked the statistics and found that the
> reported line attenuation was 8dB more than the 2Wire reported. Despite
> this the DG834G synced at the same speed as the 2Wire and remained
> connected for more than ten days non stop. Having satisfied myself that
> the Netgear would work I obtained a DGN2000 for my friend. When I
> installed it and it worked exactly as I had expected. The only puzzling
> thing is that the DGN2000 reported a line attenuation 8dB less than the
> DG834G. This has also run non stop for ten days. So my question is why
> should two routers from the same manufacturer with, presumably very ADSL
> circuitry and software, report such a different line attenuation? Does the
> router measure the line attenuation or does it get measured by the DSLAM
> and get passed to the router? In the past I have done comparative tests on
> two or three router modems on the same line and the difference between the
> reported line attenuation has never been more than a dB or so. Should we
> really believe what the hardware tells us or regard it as fiction?
>
>
>
> Peter Crosland
>

Hi Peter,

One of the reasons I switched from a DG834Gv3 to a 2700 - apart from your
recommendation - was the crazy way in which Netgear compute the noise
margin. If they've changed their methodology in their newer products, then
it is about time, too!

PA

"Peter Able" <stuck@home> wrote in message
newsM-dnYzVZo-(E-Mail Removed)...
>
> "Peter Crosland" <(E-Mail Removed)> wrote in message
> news:(E-Mail Removed) o.uk...
>> This might be sub-titled SWMBO technophobia. A friend who is not very
>> technically minded asked me to recommend a new ADSL router. He was
>> running a 2Wire 2700HGV because he is on a rural exchange and has a long
>> line. This has given excellent service almost non stop for three years.
>> My friend's SWMBO does not like wires and is very fussy about the colour
>> of domestic items. The arrival of SWMBO's shiny new laptop with "N"
>> wireless capability meant that the means this had to accommodate. I had a
>> suitable "N" wireless access point but his was rejected because "there
>> are too many wires". Having successfully used a Netgear D834G with DG
>> Team firmware I decided a Netgear DGN2000 with DG team firmware might fit
>> the bill. To prove the concept I took a spare DG834G along and fitted to
>> see how it performed. SWMBO did not like the white colour but agreed that
>> a black one would be acceptable.
>>
>>
>>
>> After installing the DG834G I checked the statistics and found that the
>> reported line attenuation was 8dB more than the 2Wire reported. Despite
>> this the DG834G synced at the same speed as the 2Wire and remained
>> connected for more than ten days non stop. Having satisfied myself that
>> the Netgear would work I obtained a DGN2000 for my friend. When I
>> installed it and it worked exactly as I had expected. The only puzzling
>> thing is that the DGN2000 reported a line attenuation 8dB less than the
>> DG834G. This has also run non stop for ten days. So my question is why
>> should two routers from the same manufacturer with, presumably very ADSL
>> circuitry and software, report such a different line attenuation? Does
>> the router measure the line attenuation or does it get measured by the
>> DSLAM and get passed to the router? In the past I have done comparative
>> tests on two or three router modems on the same line and the difference
>> between the reported line attenuation has never been more than a dB or
>> so. Should we really believe what the hardware tells us or regard it as
>> fiction?
>
> One of the reasons I switched from a DG834Gv3 to a 2700 - apart from your
> recommendation - was the crazy way in which Netgear compute the noise
> margin. If they've changed their methodology in their newer products,
> then it is about time, too!

Agreed. It may be more complex than that. The DG834G I used was a V4 version
that is relatively recent. Netgear presumably had a good reason for
producing four different iterations of their design but I have not had the
opportunity to test all the others. Having said that I am told that the
earlier ones suffered from capacitor degrading that could have a serious
effect on performance and indeed an early DG834G would not stay connected on
my line for more than a few hours before it locked up. I still hold the 2700
in high regard but there are two drawbacks i.e. no "N" capability and no
capability to tweak the SNR ratio. It was the latter that made me switch
because of the ludicrous way the BT DLM system works. What has pleasantly
surprised me is how stable the DG834G V4 and DGN 2000 have been because
previously the only router that would stay connected for a reasonable time
was the 2700. Of course there have been some changes to my line because
three times in the last two months the underground cables serving my village
have been stolen.

Peter Crosland

On Tue, 30 Aug 2011 10:11:13 +0100, Peter Able <stuck@home> wrote:

> One of the reasons I switched from a DG834Gv3 to a 2700 - apart from your
> recommendation - was the crazy way in which Netgear compute the noise
> margin. If they've changed their methodology in their newer products, then
> it is about time, too!

My V3 showed similar average figures to my V4, although the V3's
granularity was 1dB compared to V4's 0.1 and the excursions were somewhat
larger (don't know if I really believed the V3).
I bought the V4 to get control of the SNR margin.

"Paul Ratcliffe" <(E-Mail Removed)78> wrote in message
news:(E-Mail Removed)...
> On Tue, 30 Aug 2011 10:11:13 +0100, Peter Able <stuck@home> wrote:
>
>> One of the reasons I switched from a DG834Gv3 to a 2700 - apart from your
>> recommendation - was the crazy way in which Netgear compute the noise
>> margin. If they've changed their methodology in their newer products,
>> then
>> it is about time, too!
>
> My V3 showed similar average figures to my V4, although the V3's
> granularity was 1dB compared to V4's 0.1 and the excursions were somewhat
> larger (don't know if I really believed the V3).
> I bought the V4 to get control of the SNR margin.

The problem I had with the v3 was that the reported SNR appeared to be the
worst detected smoothed SNR plus the short term current variance. So after a
reset the SNR would be +6dB plus/minus one or two, and then would degenerate
to, say, +2 plus/minus one or two by about 11pm - but would then STAY at +2
plus/minus one or two. Now this "locking to worst mean case" means that
there is a significantly greater probability of the line being failed -
which it does, again and again! Sooner or later the DLM gets fed up and
raises the SNR inital value to 9dB and your line-speed degrades. The DGTeam
fixed this - but what a poor choice of algorithm by Netgear. OK in theory,
no doubt - OK on short lines - but hopeless on longer lines in the real
world.

Then PlusNet made some change (ADSL2?) and the v3 modems, whether with
Netgear or DGTeam firmware, collapsed. Hopelessly unreliable.

"Peter Crosland" <(E-Mail Removed)> wrote in message
news:(E-Mail Removed) o.uk...
> "Peter Able" <stuck@home> wrote in message
> newsM-dnYzVZo-(E-Mail Removed)...
>>
>> "Peter Crosland" <(E-Mail Removed)> wrote in message
>> news:(E-Mail Removed) o.uk...
>>> This might be sub-titled SWMBO technophobia. A friend who is not very
>>> technically minded asked me to recommend a new ADSL router. He was
>>> running a 2Wire 2700HGV because he is on a rural exchange and has a long
>>> line. This has given excellent service almost non stop for three years.
>>> My friend's SWMBO does not like wires and is very fussy about the colour
>>> of domestic items. The arrival of SWMBO's shiny new laptop with "N"
>>> wireless capability meant that the means this had to accommodate. I had
>>> a suitable "N" wireless access point but his was rejected because "there
>>> are too many wires". Having successfully used a Netgear D834G with DG
>>> Team firmware I decided a Netgear DGN2000 with DG team firmware might
>>> fit the bill. To prove the concept I took a spare DG834G along and
>>> fitted to see how it performed. SWMBO did not like the white colour but
>>> agreed that a black one would be acceptable.
>>>
>>>
>>>
>>> After installing the DG834G I checked the statistics and found that the
>>> reported line attenuation was 8dB more than the 2Wire reported. Despite
>>> this the DG834G synced at the same speed as the 2Wire and remained
>>> connected for more than ten days non stop. Having satisfied myself that
>>> the Netgear would work I obtained a DGN2000 for my friend. When I
>>> installed it and it worked exactly as I had expected. The only puzzling
>>> thing is that the DGN2000 reported a line attenuation 8dB less than the
>>> DG834G. This has also run non stop for ten days. So my question is why
>>> should two routers from the same manufacturer with, presumably very ADSL
>>> circuitry and software, report such a different line attenuation? Does
>>> the router measure the line attenuation or does it get measured by the
>>> DSLAM and get passed to the router? In the past I have done comparative
>>> tests on two or three router modems on the same line and the difference
>>> between the reported line attenuation has never been more than a dB or
>>> so. Should we really believe what the hardware tells us or regard it as
>>> fiction?
>>
>> One of the reasons I switched from a DG834Gv3 to a 2700 - apart from your
>> recommendation - was the crazy way in which Netgear compute the noise
>> margin. If they've changed their methodology in their newer products,
>> then it is about time, too!
>
> Agreed. It may be more complex than that. The DG834G I used was a V4
> version that is relatively recent. Netgear presumably had a good reason
> for producing four different iterations of their design but I have not had
> the opportunity to test all the others. Having said that I am told that
> the earlier ones suffered from capacitor degrading that could have a
> serious effect on performance and indeed an early DG834G would not stay
> connected on my line for more than a few hours before it locked up. I
> still hold the 2700 in high regard but there are two drawbacks i.e. no "N"
> capability and no capability to tweak the SNR ratio. It was the latter
> that made me switch because of the ludicrous way the BT DLM system works.
> What has pleasantly surprised me is how stable the DG834G V4 and DGN 2000
> have been because previously the only router that would stay connected for
> a reasonable time was the 2700. Of course there have been some changes to
> my line because three times in the last two months the underground cables
> serving my village have been stolen.
>
> Peter Crosland
>
>
>

Just to address your original point, I've found DG834Gv3, 2-wire 2700 and BT
Home Hubs to report the same attenuation.

I think that there was quite a significant redesign inbetween DG834Gv3 and
v4. The problem I had with the v3 was that the reported SNR appeared to be
the worst detected smoothed SNR plus the short term current variance. So
after a DLM reset the reported SNR would be +6dB plus/minus one or two, and
then it would degenerate to, say, +2 plus/minus one or two by about 11pm -
but would then STAY at +2 plus/minus one or two. Now this "locking to worst
mean case" means that there is a significantly greater probability of the
line being failed - which it does, again and again! Sooner or later the DLM
gets fed up and raises the SNR inital value to 9dB and your line-speed
degrades. The DGTeam fixed this - but what a poor choice of algorithm by
Netgear. OK in theory, no doubt - OK on short lines - but hopeless on
longer lines in the real world.

Then PlusNet made some change (ADSL2?) and the v3 modems, whether with
Netgear or DGTeam firmware, collapsed. Hopelessly unreliable. Our bain here
is power-cuts, not cable theft (yet!) - but both BT modems hold onto the
line very reliably inbetween such events. I agree that the wireless is no
great shakes and that an SNR tweak would be attractive but, on a -53dB line
the 2700 delivers a 3.5M profile, translating to about 25MByte/minute
delivery rate, day in, day out. Except, that is, about every fortnight when
the DLM has a try at a 4M profile (approx 30MByte/minute) - which last for
about 3 days before a retreat to 3M5!

PA

"Peter Able" <stuck@home> wrote in message
news:(E-Mail Removed) o.uk...
>
> "Paul Ratcliffe" <(E-Mail Removed)78> wrote in message
> news:(E-Mail Removed)...
>> On Tue, 30 Aug 2011 10:11:13 +0100, Peter Able <stuck@home> wrote:
>>
>>> One of the reasons I switched from a DG834Gv3 to a 2700 - apart from
>>> your
>>> recommendation - was the crazy way in which Netgear compute the noise
>>> margin. If they've changed their methodology in their newer products,
>>> then
>>> it is about time, too!
>>
>> My V3 showed similar average figures to my V4, although the V3's
>> granularity was 1dB compared to V4's 0.1 and the excursions were somewhat
>> larger (don't know if I really believed the V3).
>> I bought the V4 to get control of the SNR margin.
>
> The problem I had with the v3 was that the reported SNR appeared to be the
> worst detected smoothed SNR plus the short term current variance. So after
> a reset the SNR would be +6dB plus/minus one or two, and then would
> degenerate to, say, +2 plus/minus one or two by about 11pm - but would
> then STAY at +2 plus/minus one or two. Now this "locking to worst mean
> case" means that there is a significantly greater probability of the line
> being failed - which it does, again and again! Sooner or later the DLM
> gets fed up and raises the SNR inital value to 9dB and your line-speed
> degrades. The DGTeam fixed this - but what a poor choice of algorithm by
> Netgear. OK in theory, no doubt - OK on short lines - but hopeless on
> longer lines in the real world.
>
> Then PlusNet made some change (ADSL2?) and the v3 modems, whether with
> Netgear or DGTeam firmware, collapsed. Hopelessly unreliable.


Noted. I don't think Plusnet would have changed anything. What may have
happened is that BT made some changes to their DSLAM software. It is also
possible that the capacitors in the V3 had degraded or failed. I am told
this is a common problem with them. The symtoms are repeated lockups.

Peter Crosland

"Peter Able" <stuck@home> wrote in message
news:(E-Mail Removed) o.uk...
>
> "Peter Crosland" <(E-Mail Removed)> wrote in message
> news:(E-Mail Removed) o.uk...
>> "Peter Able" <stuck@home> wrote in message
>> newsM-dnYzVZo-(E-Mail Removed)...
>>>
>>> "Peter Crosland" <(E-Mail Removed)> wrote in message
>>> news:(E-Mail Removed) o.uk...
>>>> This might be sub-titled SWMBO technophobia. A friend who is not very
>>>> technically minded asked me to recommend a new ADSL router. He was
>>>> running a 2Wire 2700HGV because he is on a rural exchange and has a
>>>> long line. This has given excellent service almost non stop for three
>>>> years. My friend's SWMBO does not like wires and is very fussy about
>>>> the colour of domestic items. The arrival of SWMBO's shiny new laptop
>>>> with "N" wireless capability meant that the means this had to
>>>> accommodate. I had a suitable "N" wireless access point but his was
>>>> rejected because "there are too many wires". Having successfully used a
>>>> Netgear D834G with DG Team firmware I decided a Netgear DGN2000 with DG
>>>> team firmware might fit the bill. To prove the concept I took a spare
>>>> DG834G along and fitted to see how it performed. SWMBO did not like the
>>>> white colour but agreed that a black one would be acceptable.
>>>>
>>>>
>>>>
>>>> After installing the DG834G I checked the statistics and found that the
>>>> reported line attenuation was 8dB more than the 2Wire reported. Despite
>>>> this the DG834G synced at the same speed as the 2Wire and remained
>>>> connected for more than ten days non stop. Having satisfied myself that
>>>> the Netgear would work I obtained a DGN2000 for my friend. When I
>>>> installed it and it worked exactly as I had expected. The only puzzling
>>>> thing is that the DGN2000 reported a line attenuation 8dB less than the
>>>> DG834G. This has also run non stop for ten days. So my question is why
>>>> should two routers from the same manufacturer with, presumably very
>>>> ADSL circuitry and software, report such a different line attenuation?
>>>> Does the router measure the line attenuation or does it get measured by
>>>> the DSLAM and get passed to the router? In the past I have done
>>>> comparative tests on two or three router modems on the same line and
>>>> the difference between the reported line attenuation has never been
>>>> more than a dB or so. Should we really believe what the hardware tells
>>>> us or regard it as fiction?
>>>
>>> One of the reasons I switched from a DG834Gv3 to a 2700 - apart from
>>> your recommendation - was the crazy way in which Netgear compute the
>>> noise margin. If they've changed their methodology in their newer
>>> products, then it is about time, too!
>>
>> Agreed. It may be more complex than that. The DG834G I used was a V4
>> version that is relatively recent. Netgear presumably had a good reason
>> for producing four different iterations of their design but I have not
>> had the opportunity to test all the others. Having said that I am told
>> that the earlier ones suffered from capacitor degrading that could have
>> a serious effect on performance and indeed an early DG834G would not stay
>> connected on my line for more than a few hours before it locked up. I
>> still hold the 2700 in high regard but there are two drawbacks i.e. no
>> "N" capability and no capability to tweak the SNR ratio. It was the
>> latter that made me switch because of the ludicrous way the BT DLM system
>> works. What has pleasantly surprised me is how stable the DG834G V4 and
>> DGN 2000 have been because previously the only router that would stay
>> connected for a reasonable time was the 2700. Of course there have been
>> some changes to my line because three times in the last two months the
>> underground cables serving my village have been stolen.
>>
>> Peter Crosland
>>
>>
>>
>
> Just to address your original point, I've found DG834Gv3, 2-wire 2700 and
> BT Home Hubs to report the same attenuation.
>
> I think that there was quite a significant redesign inbetween DG834Gv3 and
> v4. The problem I had with the v3 was that the reported SNR appeared to
> be the worst detected smoothed SNR plus the short term current variance.
> So after a DLM reset the reported SNR would be +6dB plus/minus one or two,
> and then it would degenerate to, say, +2 plus/minus one or two by about
> 11pm - but would then STAY at +2 plus/minus one or two. Now this "locking
> to worst mean case" means that there is a significantly greater
> probability of the line being failed - which it does, again and again!
> Sooner or later the DLM gets fed up and raises the SNR inital value to 9dB
> and your line-speed degrades. The DGTeam fixed this - but what a poor
> choice of algorithm by Netgear. OK in theory, no doubt - OK on short
> lines - but hopeless on longer lines in the real world.
>
> Then PlusNet made some change (ADSL2?) and the v3 modems, whether with
> Netgear or DGTeam firmware, collapsed. Hopelessly unreliable. Our bain
> here is power-cuts, not cable theft (yet!) - but both BT modems hold onto
> the line very reliably inbetween such events. I agree that the wireless
> is no great shakes and that an SNR tweak would be attractive but, on
> a -53dB line the 2700 delivers a 3.5M profile, translating to about
> 25MByte/minute delivery rate, day in, day out. Except, that is, about
> every fortnight when the DLM has a try at a 4M profile (approx
> 30MByte/minute) - which last for about 3 days before a retreat to 3M5!

Noted. The problem is that we are not using a fixed and stable reference
because line characteristics can, and indeed do, vary quite a lot. My theory
is that the 2700 is particularly good at dealing with these fluctuations
without resorting to decreasing the sync speed which a number of other
routers do. I found that Draytek range that were otherwise excellent were
dreadful at dealing with unstable lines. The other variable is the chipset
used by the manufacturer. Finally BT do use DSLAMs from a number of
different manufacturers which also muddies the waters

Peter Crosland

"Peter Crosland" <(E-Mail Removed)> wrote in message
news:(E-Mail Removed) o.uk...

>> Then PlusNet made some change (ADSL2?) and the v3 modems, whether with
>> Netgear or DGTeam firmware, collapsed. Hopelessly unreliable.
>
>
> Noted. I don't think Plusnet would have changed anything. What may have
> happened is that BT made some changes to their DSLAM software. It is also
> possible that the capacitors in the V3 had degraded or failed. I am told
> this is a common problem with them. The symtoms are repeated lockups.
>
> Peter Crosland
>

Oh, they notified me, Peter, and when I reported the problems I was having
their reply was "DG834Gv3 don't work reliably with ADSL2. We'll send you
one of ours for a fiver and a commitment to stay with us for at least
another year".

Well, I spent the fiver on the 2700, and as that turned out to be the latest
version 5 type, I considered that I'd fallen well and truly on my feet!

PA