Crosstalk or something else?

Well, my connection is now in sync at only 20764/4859 Kbps.

It has reduced in 3 steps to that level since December 2012.

In December, sync speeds were 29430/5704 Kbps, where they had more or less been for the 7 months since my connection was finally repaired near the end of May 2012.

DS Attainable rates have gone from 34Mb to 21.5Mb




Here's a before & after set of snapshot graphs:-


http://i.imgbox.com/ackiwNN4.gif


The really obvious differences are Bitloading, QLN & SNR.
However, I can also see slight deterioration in the Hlog graph, along with slightly increased attenuation as reported via the pbParams data.

I also notice that TX Power has reduced slightly.

Right up to May 1st 2013, Interleaving depth had been around 400-500 DS & 1 US since it was applied a few days after DLM was reset back in May 2012.


When the connection resynced on the fly at 04:54 May 1st at speeds of 24802/5172 Kbps, DS Interleaving was switched off completely.
INP & delay were also set to zero at the same time.

This was quite surprising as my 1000m connection does usually seem to need Interleaving, INP & delay to be applied at reasonably low levels.

There had been quite a few resyncs since 1st May, until appearing to settle at such low sync speeds.

I presume such low speeds no longer require Interleaving, INP & delay setting.


I disconnected all internal extension telephony wiring, switched the mains power off in my house & actually ran the connection for a while via a car battery & a battery powered laptop.
This made absolutely no difference to the stats obtained when fully mains powered, so I assume nothing in my own house can be causing such deterioration in speeds & stability.



Now, my question is, can this really all be down to increased crosstalk or is it possible that either something else has gradually caused QLN to deteriorate or is some sort of a line 'fault' brewing?


I have no idea how many users between the cabinet & my house have broadband connections or even which flavour(s) of DSL they are using.
I also have no idea how to discover if there has been an increase in the number of broadband users, coinciding with the dates I noticed a deterioration.


Does anyone have any ideas/suggestions?

I don't believe requesting DLM to be reset would have any mileage (even in the unlikely event that such a request would be granted), particularly as Attainable rates are so low & there doesn't really appear to be much in the way of spare SNRM.

I assume I'm talking about how to suck eggs if I ask about the test socket, replacement of the socket >> modem cable, and replacement of the modem/PS for it?

Yep.

Done all that.

I even turned off all the mains power to the house & ran the HG612 from a car battery for a while, using a battery powered laptop to grab the stats.

I have also used a spare unlocked HG612 & a HG622 on the connection to no avail.

Yep, I read that . But it's very rare for people to swap out cables - often re-using them when changing the kit at each end.

I agree with you it looks more like a specific cause such as deterioration somewhere, rather than just increased cross-talk due to additional connections.

Smart meters?

Battery-powered weather stations/radio-sync'ed clocks?

Hope for a power cut and run the hardware up via batteries and everyone elses is off. If speed returns its cross talk

a big problem is one data is fast path and the other is interleaving.

interleaving usually bumps up the attainable rate so that alone stops it been comparable properly and of course also reduces the actual sync rate.

you not got both fast path comparisons or both interleaving?

To me it does look most likely crosstalk tho.

In reply to a post by Chrysalis:

a big problem is one data is fast path and the other is interleaving.

interleaving usually bumps up the attainable rate so that alone stops it been comparable properly and of course also reduces the actual sync rate.

I posted the comparison graphs with/without interleaving just to show the differences since December when my connection was quite stable with 'reasonable' speeds.

you not got both fast path comparisons or both interleaving?

I do have relevant stats from as far back as November 2011, but as my connection can't usually sustain Fastpath at its usual higher speed, I didn't think it relevant for comparison purposes.
It has previously achieved over 30Mb sync speed with interleaving at a depth of around 400 or so.

My theory is that it only recently went back to Fastpath due to DLM seeing far fewer errors at such low sync speeds.

I'm trying to confirm why sync speeds are currently so low, without much success so far.


Plusnet have confirmed their own tests do actually show a fault, but I don't yet know what those tests were.

To me it does look most likely crosstalk tho.

Are you basing that on the deteriorated QLN graph or some other factor(s)?

I live in a relatively sparsely populated area & many residents are quite elderly so possibly less likely to be broadband users.

In reply to a post by Bald_Eagle1:

In December, sync speeds were 29430/5704 Kbps, where they had more or less been for the 7 months since my connection was finally repaired near the end of May 2012.

What was the original estimate for your line?

DS Attainable rates have gone from 34Mb to 21.5Mb

As Chrysalis points out, the attainable figures when interleaving/FEC is turned on turn out to be artificially high.

So the likely realistic drop is from around 31Mbps to 21Mbps - so you've lost about a third of the line's throughput.

I think this is still within the bounds of "expected behaviour", even if it isn't very nice to experience.

My line, over the last 17 months, has gone from around 88Mbps attainable to 72Mbps, but most is in the last 6 months. It has survived that reduction since December, and is still synchronised at 80Mbps but with around 3dB SNRM.

The best graph I've seen that gives a hint about the random nature of the effect of crosstalk can be seen in a description of vectoring. Look at the graph on page 12 of this document...

The blue crosses all represent the speed seen on various lines at different distances, and the variation between the best and worst is remarkable. For example, at 600 metres the best can get 65Mbps while the worst gets 45Mbps.

Your line is longer than 1km isn't it, so I guess you need to extrapolate to get comparable results.

Here's a before & after set of snapshot graphs:-
http://i.imgbox.com/ackiwNN4.gif

How did you make those? I'd like to try the same!

The really obvious differences are Bitloading, QLN & SNR.
However, I can also see slight deterioration in the Hlog graph, along with slightly increased attenuation as reported via the pbParams data.

The Hlog has changed slightly. As this says something about the underlying copper, it'd be interesting to see if this change has anything to do with time-of-day, weather or seasonal effects. Unfortunately, I guess it is only produced at each resync.

QLN seems to show more noise throughout, which would cause reduced SNRM, and reduced bit-loading. These seem to be consistent, without any notable singularities.

I also notice that TX Power has reduced slightly.

D1 has the same power, but D2 has reduced. Perhaps this is because the modem is using fewer frequencies in D2.

I presume such low speeds no longer require Interleaving, INP & delay setting.

If the lower speeds are caused by increased noise (as they seem to be), then you'd expect wideband interference to carry on in the same way.

If, however, DLM intervened to restrict your speed by banding, then interference can be reduced. However, you'd also see a higher SNRM than 6dB - which you don't.

What might have happened is that all your interference was being caused in tones that are no longer used. Without adverse effects there, DLM can be turned off. But I'm not sure that is the order in which you saw speed reductions happen - DLM turned off before you stopped using the tones.

Now, my question is, can this really all be down to increased crosstalk or is it possible that either something else has gradually caused QLN to deteriorate or is some sort of a line 'fault' brewing?

Wouldn't this be more visible in the Hlog graph? Or by QLN changes at specific frequencies?

I have no idea how many users between the cabinet & my house have broadband connections or even which flavour(s) of DSL they are using.
I also have no idea how to discover if there has been an increase in the number of broadband users, coinciding with the dates I noticed a deterioration.

The interferers can live further away too - their lines just need to run parallel and close to yours before yours branches out.

They're probably not ADSL users (of either flavour) - the QLN graph is being affected way above 2MHz.

Does anyone have any ideas/suggestions?

I don't believe requesting DLM to be reset would have any mileage (even in the unlikely event that such a request would be granted), particularly as Attainable rates are so low & there doesn't really appear to be much in the way of spare SNRM.

I don't think it would make much difference.

Are you still using the new firmware? Perhaps swapping back would say something (or onto it again)

In reply to a post by WWWombat:

What was the original estimate for your line?

Back in May 2011 (8c profile) it was 14.6Mb. I achieved around 33Mb sync speed & 32Mb actual throughput for the first month (before 'someone' disconnected my telephone & broadband at the PCP).

When 17a was introduced it went to 24Mb estimated & it is currently 'up to' 30Mb.

DS Attainable rates have gone from 34Mb to 21.5Mb

As Chrysalis points out, the attainable figures when interleaving/FEC is turned on turn out to be artificially high.

So the likely realistic drop is from around 31Mbps to 21Mbps - so you've lost about a third of the line's throughput.

Yes, I've seen that effect, but when DLM has been reset on Fastpath, Attainable rates have been as high as 35Mb & sync speed at around 33Mb.

I think this is still within the bounds of "expected behaviour", even if it isn't very nice to experience.

My line, over the last 17 months, has gone from around 88Mbps attainable to 72Mbps, but most is in the last 6 months. It has survived that reduction since December, and is still synchronised at 80Mbps but with around 3dB SNRM.

The best graph I've seen that gives a hint about the random nature of the effect of crosstalk can be seen in a description of vectoring. Look at the graph on page 12 of this document...

The blue crosses all represent the speed seen on various lines at different distances, and the variation between the best and worst is remarkable. For example, at 600 metres the best can get 65Mbps while the worst gets 45Mbps.

Your line is longer than 1km isn't it, so I guess you need to extrapolate to get comparable results.

I'll have a look at that document in due course. Thanks for the link.
My line is somewhere between 1000m & 1100m from the PCP.

Here's a before & after set of snapshot graphs:-
http://i.imgbox.com/ackiwNN4.gif

How did you make those? I'd like to try the same!

Paintshop Pro (very, very old version 5) came with anim.exe for creating/editing animated GIFs.
One simply adds another graphic file into the next frame, confirming the display time & that's it.

The really obvious differences are Bitloading, QLN & SNR.
However, I can also see slight deterioration in the Hlog graph, along with slightly increased attenuation as reported via the pbParams data.

The Hlog has changed slightly. As this says something about the underlying copper, it'd be interesting to see if this change has anything to do with time-of-day, weather or seasonal effects. Unfortunately, I guess it is only produced at each resync.

QLN seems to show more noise throughout, which would cause reduced SNRM, and reduced bit-loading. These seem to be consistent, without any notable singularities.

I also notice that TX Power has reduced slightly.

D1 has the same power, but D2 has reduced. Perhaps this is because the modem is using fewer frequencies in D2.

I presume such low speeds no longer require Interleaving, INP & delay setting.

If the lower speeds are caused by increased noise (as they seem to be), then you'd expect wideband interference to carry on in the same way.

If, however, DLM intervened to restrict your speed by banding, then interference can be reduced. However, you'd also see a higher SNRM than 6dB - which you don't.

What might have happened is that all your interference was being caused in tones that are no longer used. Without adverse effects there, DLM can be turned off. But I'm not sure that is the order in which you saw speed reductions happen - DLM turned off before you stopped using the tones.

Now, my question is, can this really all be down to increased crosstalk or is it possible that either something else has gradually caused QLN to deteriorate or is some sort of a line 'fault' brewing?

Wouldn't this be more visible in the Hlog graph? Or by QLN changes at specific frequencies?

I have no idea how many users between the cabinet & my house have broadband connections or even which flavour(s) of DSL they are using.
I also have no idea how to discover if there has been an increase in the number of broadband users, coinciding with the dates I noticed a deterioration.

The interferers can live further away too - their lines just need to run parallel and close to yours before yours branches out.

They're probably not ADSL users (of either flavour) - the QLN graph is being affected way above 2MHz.

Does anyone have any ideas/suggestions?

I don't believe requesting DLM to be reset would have any mileage (even in the unlikely event that such a request would be granted), particularly as Attainable rates are so low & there doesn't really appear to be much in the way of spare SNRM.

I don't think it would make much difference.

Thanks for your other comments which I will digest shortly.

Are you still using the new firmware? Perhaps swapping back would say something (or onto it again)

No, I had already reverted to the original BLOB in order to compare like for like.

Basis for crosstalk is it doesnt look like anything else and yes the QLN.

Remember we on here only have the graph data not JDSU diagnosis etc. On kits forum I even found out yesterday even if a JDSU pass there is other tests that can show failure (you took part in that thread also). But those tests typically arent reported to the consumer, and engineers dont even have to run them on openreach policy.

Whilst its hard to accept I think the most likely thing at the moment is crosstalk.

On my line BT gave me a dedicated dropwire a couple of weeks back and I got all my attainable back, I shot up to 97/39 from the current 68/27 attainable.but they wouldnt let me keep it even tho they had spent time putting it in place, they were just testing and apparently policy only allows 1 dropwire to my building (which has 3 flats). One of my neighbours does have FTTC so now I have pretty much found my cause, both me and my neighbour crosstalking each other. I even had an argument with the openreach manager for my area over the engineers phone for 20 mins to try and keep the new dropwire in place but he wouldnt have it, to him 65mbit and 80mbit no difference.

Since then I have given up and just waiting for vectoring or for my neighbour to cancel/move. My line no longer jumps up and down tho, I supect I am/was affected by other lines also which stopped happening after my first engineer visit. After that first visit the attainable has been on 68/27 is with no variances apart from when they changed the dropwire.

In reply to a post by WWWombat:

Here's a before & after set of snapshot graphs:-
http://i.imgbox.com/ackiwNN4.gif

How did you make those? I'd like to try the same!

If you're on Windows 7 you can do a very similar thing to easily step through your graphs day by day.

Navigate to the Current Stats or On-going Stats folder, whichever contains the graphs you want to view, so if you wanted hlog then naviagate to Current_Stats folder, so you can see all the sub folders.

Now type Hlog in the search box, and then save the search.

Then just double click the first picture and you can move through them, just like that Gif. There are plenty of online places you can make animated gifs, or I think there may be a plugin for paint.net.

My searches get saved to my Favorites, so it's just case of clicking the search, making sure they are sorted by filename and then viewing.

This doesn't work so well with the Bits graph as we have two for each one, normal and high res.

When my connection changed to fast path recently, my attainable went from around 50 to 47, I'm still on the new firmware.

Hi Bald_Eagle,
Do you think yours is a similar problem to the one we discussed here? - http://forums.thinkbroadband.com/fibre/t/4218198-int...

In reply to a post by smouty:

Do you think yours is a similar problem to the one we discussed here?

Yes, no, maybe - I don't know


I did notice that mains power in the area where I live 'blipped' this morning, causing the house lights to go off & the PC's amplified speakers made a noise for less than 1 second.

The PC, electronic clocks, TV & the HG612 didn't switch off, but the Netgear router did disconnect/reconnect.

From the HG612's own log:-

2013-5-14 7:19:13 Debug 104500 LAN1 up
2013-5-14 7:19:10 Debug 104500 LAN2 up
2013-5-14 7:19:8 Debug 104500 LAN1 down
2013-5-14 7:19:8 Debug 104500 LAN2 down

By pure chance I had the HG612's GUI open at the time & noticed for just a few seconds that DS SNRM shot up to 10.3dB & Attainable Rate was almost 28Mb before it returned to 6.3dB & 21128Kbps.

Unfortunately, as this happened between scheduled data harvests I don't have the relevant stats & it was that brief I didn't get change to grab a screenshot or force a snapshot data harvest.

I can only assume that other users' routers/modems also disconnected & I was able to see no or vastly reduced interference on my own connection before they all reconnected.

I really would have loved to have seen a QLN graph at that time.

On my way to work I did also see some contractors carrying out some work to a road crossing not far from the cabinet. Electrical company lorries were parked at the roadside at the time.

Coincidence? - Maybe.

It would be interesting to see what happens during a power cut, mines all hooked up to a UPS so will carry on logging for about half an hour until it all gracefully shuts down. Trouble is (luckily) we don't have power cuts very often now - must be years since the last one here.

Although not directly related, we had a power cut here 5 days back that took out a wide area. Like you I'm on a UPS so RouterStats keeps running for a while. Very surprised to see below which I guess is less the ambient noise + crosstalk? Line is ADSL2+, 2.9km/39dB from exchange and normally runs consistently at 9-10Mb/s at 3dB or so depending on how I have the SNR Margin set.

RouterStats with Power Cut

A bit longer ago, I came across a fallen power line (insulated) across a footpath and reported it. Local energy co cut the power to just this part of the village area for it to be sorted which gave an improvement of around 4dB in the SNR margin.

Unfortunately, I had been out for this one and got back to find the UPS announcing it was about to run out of juice, managed to save the current RS graph above but no time to reconnect and see what fantastic connection speed I could manage with another 9dB! The route between myself and the exchange is pretty well all residential plus shops, no heavy industry.

Edit: there is about 0.3dB noise in the house from various domestic things, established during major problems about a year ago. The equipment in my workroom adds about another 0.2dB.

Edited by tbailey2 (Wed 15-May-13 10:31:59)

In reply to a post by tbailey2:

Although not directly related, we had a power cut here 5 days back that took out a wide area. Like you I'm on a UPS so RouterStats keeps running for a while. Very surprised to see below which I guess is less the ambient noise + crosstalk? Line is ADSL2+, 2.9km/39dB from exchange and normally runs consistently at 9-10Mb/s at 3dB or so depending on how I have the SNR Margin set.

RouterStats with Power Cut

That certainly shows how much noise there can be in the vicinity. However, knowing how noisy some UPSs can be - radiated and transmitted, I wonder how much of that noise is from the UPS. There is the conversion from 230v down to the battery charging voltage which will be running continually and then when the battery is supplying power the up-conversion.

Have you tried running with the UPS input turned off? With the modem connected to a normal mains supply with UPS running? and with UPS off?

you can maybe guess what 9db would have gave you by looking at attainable stats.

When my line last jumped to 85+ attainable I did a force resync to get a 79999 sync, when the crosstalk returned I had the 79999 sync with a 3db snrm, sadly it didnt last long due to DLM fiddling with my upstream power.

So on my line 3db of snrm is around 7mbit sync speed. I thought was 14-15 before but calculated wrong as I based it on the attainable jump from 72 to 85-86 but forgot to account that on the 79999 sync there was also a jump of snrm to 9db. So when the attainable reverted with the crosstalk returned it went from around 9db to 3db. 6 db drop for loss of approx 14mbit attainable.

MHC
Yes, but a long time ago and couldn't see much difference then and the UPS has been changed since (APC 1500W). If that was mostly UPS noise, then I wonder what the connection speed would be? The current attainable is 13308 @ 3.3dB with actual d/l 11303. Over the past few weeks looking at RS graphs each 2dB of extra noise margin is worth about 1Mb/s d/l increase (if I reset the SNR Margin via the router to what it was) so 7dB could mean an extra 3.5Mb/s !! I think not...

However, I'll try that UPS test again when I get a moment.

Edit: corrected figs!!

Edited by tbailey2 (Wed 15-May-13 13:54:40)

In reply to a post by MHC:

Have you tried running with the UPS input turned off? With the modem connected to a normal mains supply with UPS running? and with UPS off?

Well doesn't seem to be much contribution from the UPS under normal conditions.

Went through the combinations and, allowing for the effects of the router rebooting (disconnected the sessions first though) the only detrimental effect I could see was when pulling the mains on the UPS so it was running on actively upconverted battery power. There was then a drop of about 0.3dB in the SNR Margin which came back as soon as it was on mains power again.

Running everything entirely on mains really was no different. After a few tests, I did end up with ~300k more d/l than when I started though! Probably because the SNR Margin was higher when I did this than the last time I played around.

Only downside was my Latency has gone up by about 5ms as the interleave is back on upload, it was on fastpath. DLM removed it a couple of days back and that dropped the ping times.

When I started the speeds were 11303/1072 @ 3.4dB with an attainable of 13308
On mains only 11460/1082 @ 3.1dB with an attainable of 13612
After tests back on UPS (but not active) power, 11585/1080 with an attainable of 13724

So somewhere out there is another ~2000kb/s if I could find it.

Doesn't really matter too much though as there is now a shiny new FTTC cab next to my current cab that I gather from an OpenReach guy has power and fibre but isn't active yet.

STREET SIDE DSLAM CAB NNAHWF, COLLEGE LANE, HURSTPIERPOINT, HASSOCKS, BN6 9AB on Exchange HURSTPIERPOINT is served by Cabinet 4

Sounds like a nice quiet UPS ...

There is something close to you that is not very quiet though.

Let's hope it is, sounds like the same model as mine APC Smart UPS 1500.

In reply to a post by R0NSKI:

Let's hope it is, sounds like the same model as mine APC Smart UPS 1500.

Yes, it's the same. Battery is about 3.5 years old but still tests okay.

Talking about UPSs, I see mentions of FTTC Modems having a small UPS supplied with them but few photos of installations show anything. Does anyone have one?

UPS and Openreach is referring to the FTTP installs, where Fibre Voice Access is supplied, and as part of the telephone USO it has to support phone calls during a power outage.

No requirement for FTTC to operate during a power outage, at least not one that is defined in a USO

In reply to a post by MrSaffron:

UPS and Openreach is referring to the FTTP installs, where Fibre Voice Access is supplied, and as part of the telephone USO it has to support phone calls during a power outage.

No requirement for FTTC to operate during a power outage, at least not one that is defined in a USO

Thanks Andrew, that makes sense of course. For FTTC, I believe the cabs themselves have backup batteries though that provide local power for a limited time (hour or so) for the broadband side?

The cabinets do, I believe it was around 12 hours If I remember...

Correct, so if you have a UPS at home/work you can keep your broadband running for as long as you can generate power.

Believe if a power outage was extended a system of swapping out batteries could be employed by Openreach.