Noise margins cycle down - strange pattern ????

Hi

I'll try to describe what I have captured using routerstats from my netgear.

Most of the day the noise margin jitters around 13db - so you see small spikes up to 15db and down to 12 db ...but quite often it's a flat line for hours. But ... come around 18:30, the differences get a little larger, and this is what I would expect at night due to the many factors which we have had described many times ...

But ... now ... this jitter is superimposed onto a large "cycle" (looks like a sine wave!), with a period of around 30 mins.

Once in this cycle, the noise margin drops by 4 or 5 db to the trough, but only recovers up to about 3db down, then another half hour cycle, and so on until it goes negative.

So, if you drew a straight line average though it all, then you are seeing a gradual drop over a few hours from say 15db down to negative noise margin.

Through the night it will recover and then stick around the 13-15 db for the rest of the day until evening comes around again.

So, in simple terms, imagine a sine wave of period 30 mins from peak to peak, and an amplitude of 5 db peak to peak, sitting with the "zero crossing" at 15 db.

Now tilt this so the zero crossing point goes negative after about 3 hours (ie 6 cycles), this is what I see (with the normal couple of db jitter as well, but this is more or less random).

JohnR here suggests that perhaps something is charging and discharging ... but why does it start at 6:30 to 7 pm?

It does this with everything in the house switched OFF and nothing else plugged into the phone line.

[BT OR have chased this line to the exchange, fitted NTE faceplate, checked all the internal wiring, losses to green cab, etc etc, lift and shift done. Also, the various params for the rate adaptive line had been incorrectly set, so this has been fixed. BT OR tests say it's a VERY GOOD line and should support a decent speed.]

Line stats are 57 db loop loss, noise margins .. well see above ...

I get syncs from 3776k down to 160k. I get IP Profiles from 130k to 2.5meg.

I get thoughputs from 127 k (sometimes weeks of it) up to 2.5 megs.

Of course, all this still fits the "up to 8 megs" product, so my ISP and BT are being very good at trying to solve the problem which has been around now since the move from a 512k fixed line to the new product on 7th July.

I have no complaints at all about how both parties are trying to get to the bottom of this issue. I get phoned updates every few days.

Other people in the same area who's cable also take the same route in the same ducts, get around 3.5 megs throughput, and syncs of 4.5 Megs. They also have the same 57db loop loss AND noise margins of around 12 db.

So, anyone have any bright ideas about what is causing my noise margins to oscillate as they do in the evening?

This behaviour somehow is upsetting the algorithms. I am not seeing any HARD resets or disconnects, just changes in sync and throughput.

And, finally, anyone else seen anything like this and found a solution?

Thanks

Rob

I'm afraid a solution to this isn't going to be quick. Over the next few months, you will be able to see if the start of this phenomenon coincides with a definite time of day, or whether it coincides with the appearance of dusk. You are obviously aware of the interference that gets stronger at night, and if this is what we are looking at, then it's possible that you have a radio signal that fades or is cancelled out by another, with a period of 30 minutes. If this phenomenon starts to take place at an earlier time every evening, then we are definitely looking at radio interference (or could feasibly be a street lamp coming on). If the start of the interference stays at 6:30PM right through the winter, then you may have to look at other sources of interference (industrial heaters, interference from local domestic users, etc).

@Zebedee

Thanks for the reply. Pretty sure it's getting earlier ... so as you say probably either radio or street lights. Neither of which can be fixed. I am pretty sure that the cyclical effect is getting stronger too ... a month ago is was simply a gradual drop over the 3 hours or so from the 15db to negative value, now the sine wave is very evident in the plots.

So, the next question would be ... how do I stop this having such a bad effect on my BB via the rate adaptive software, which simply sees the line as very unstable, some of the time (spose it is, in reality!). From the green cab (which has the sealed pressurised cable from the exchange, so I am told), to my house shows a loop loss from 51db to 57db, which seems to be the same for those who live next to the green box (my friend) and me quite a bit further down the road! He gets the 3.5 meg with a 4.5 meg sync, fully stable.

What I would like of course is this, but I would settle for something a bit more stable, and not the sudden drops to 135k IP Profiles (even with high syncs!)..

Like I said, all parties so far have been very good, and have kept me informed on a very regular basis where we are in the saga.

The lift and shift to a new cab in the exchange made no difference, but it was done mainly to get the line parameters fixed up (which unfortunately, they were not and the old incorrect ones were simply transferred over, rather than starting afresh!). So getting that changed took quite some time ISP talking to BT.

The good thing I suppose is that whatever sync I get and whatever IP Profile I am given, then a BT speedtest will ALWAYS show more or less the maximum throughput for that IP Profile!

Just that the IP Profile has been varying from 135k to 2.5 meg and just about all values in between (I could write the table for you!!!).

I'll continue to monitor. The graph over the next month will be interesting to look at (once I find and replace all those silly values from the netgear which cannot display negative numbers - oh why didn't the routerstats author build in a fix in the sofware given that this is so well known!)

Rob

I assume you know, (as long as I have my arithmetic right), that if you subtract 2,147,483,648 from the huge numbers you get the real negative value?

Strange, I'm sure with RouterStats Lite my graphs (from a Netgear 834PN) showed negative margin correctly.

Also in RouterStats and Lite you can set it to log to a file as well as graph, giving you one file rather than all the mini-graphs it saves. You could import that to a spreadsheet and do a simple correction then graph the result.

@Robertos

Hi yes, thanks I know this one!

Aha ... I grabbed the full version, not the Lite version ...

I am logging, so yes, will import to XL and correct as appropriate!

Such strange behaviour. Just peeking at the stats page now and then of course did not reveal the whole "horror" of what's going on ...

Of course, finding the source and stopping it may be difficult ... and of course, removing the influence of the ionosphere etc may prove somewhat impossible.

It is odd that the cycles continue for some hours. This to me would suggest that it's not the street lights as they would all be on, fired up and burning bright after that time.

Radio waves phasing in and out, maybe ... but I would have thought that this would affect everyone in my area ... and does not seem to.

I am at the end of a cul-de-sac, with good gaps between modern houses (25yrs max). The GPO hole in the pavement is just outside the house, and some distance from others. All underground cables too. Not a pole in sight.

I guess I need to see how to shield the cables perhaps from RF?

Mind you, don't want to live in a Faraday Cage! Did 3 years research inside one of those!

Rob

In reply to a post by RobertoS:

I assume you know, (as long as I have my arithmetic right), that if you subtract 2,147,483,648 from the huge numbers you get the real negative value?

It's called two's complement arithmetic.

I assume you've done tests at the master socket, with extension wiring disconnected ?

Errr - yes? I probably knew that before you were born .

@Z

Yup, and so did the OR engineer who worked for a few hours at my home and chasing the line along its path back to the exchange. All sorts of equipment etc was trundled out of the van and used to test bits and bobs!

The conclusion was that it is a VERY GOOD LINE, and should support around 3 megs without much trouble.

But, as I said the funny noise cycles early evening onwards seem to confuse the hell out of the exchange s/w and end up with inconsistent BB throughputs.

Rob

@Robertos

Me too! Programmed the old Elliot Automation 803's in 1965/1966 - see wikipedia for info on the 39-bit word and instruction set! 5-hole punch tape too ... so a restricted character set

All a bit too much in the dim and distant past now ... says he dribbling in the corner of the room

Rob