DGND3700 re-connect DSL?

[Scubbie]

I can't comment about the 'false signal', but I am a little puzzle by your stats.

I have been advised by others that ADSL1 connection stats are always divisible by 32. Your upstream connection isn't divisible by 32.

Looking at your noise margins, it is clearly obvious that there is plenty of room for improvement too. I don't know what your ISP uses, by Sky (my ISP) uses a conservative 7dB target and O2 uses around 5dB as a target.

With this information in mind, I would be looking at your line quality and thinking that there may be something interfering with it. Things to consider are:

• REIN
• damaged line (trees or roots are often a good cause to check, perform a "quiet line test" by using a corded phone and dialling '1' and listen for any unsuitable noise)

Once you have identified the source of the damage or interference, providing there is any, getting your line cleaned up should be a lot easier.

[adl]

Originally Posted by Scubbie

I have been advised by others that ADSL1 connection stats are always divisible by 32. Your upstream connection isn't divisible by 32.

Typo. Should have posted 768.

Originally Posted by Scubbie

Looking at your noise margins, it is clearly obvious that there is plenty of room for improvement too. I don't know what your ISP uses, by Sky (my ISP) uses a conservative 7dB target and O2 uses around 5dB as a target.

I'll have to do some digging on that. I know the numbers are high but I don't have historical records. It has been a long while since I looked at the numbers.

[adl]

Double-checked my notes and see that I've only got line data since I connected 30 feet of flat telephone wire with 2 or 3 splices. It was coiled up in a couple of spots on either side of the printer until I strung it out in front of a staircase, across a doorway, around some furniture and out a window. I would not be surprised to find it prone to interference. I'll pull off stats after this testing is done to see what I get then.

As for the bypass-the-house-wiring test, so far I see no change at all -- 7 disconnects in 22 hours including two in the last 45 minutes. I did forget to isolate the house wiring overnight.

On to V19test2!

[Scubbie]

Seriously I would recommend against using an extension. All too often I have seen posts where some thought that things were fine, but they were proved otherwise.

Other sources of problems that I have seen are:

• damaged telephone sockets (even the copper contacts inside)
• burglar alarms wired on to the telephone line before the socket
• faulty junction boxes between the master socket and the line leaving the property

If you are just pulling my leg, then please remember that joke do not carry well on a text based forum. The use of smilies does help.

[adrenalize]

• REIN
• damaged line (trees or roots are often a good cause to check, perform a "quiet line test" by using a corded phone and dialling '1' and listen for any unsuitable noise)

A "quietline test" is a great method - I would suggest you try the quietline facility on a lot of BT UK exchanges dial 17070 and press option 2. A voice will say quietline test and then silence - listen for any pops crackles, buzz etc. Hang up when done.

"Breaking dialtone" by pressing a number gives you a short period of silence, but usually a loud tone a few seconds later as the number is invalid. It is often useful to listen for a few minutes especially for intermittent noise and things like heating boilers/thermostats cutting in and out.

Voice faults like lots of crackling are a lot easier to report as faults and get an engineer out to than slow ADSL speeds especially when they are within the wide BT tolerance for speed!

[adl]

I've done line-quality testing with this site but only
- when connected via Ethernet cable (i.e. NOT wireless)
- running through the in-house wiring (i.e. NOT under the current/temporary bypass testing)

Later today (or, more likely, over the weekend) I'm going to run again with the current/temporary bypass setup to see what I get then switch back to in-house wiring and do it again. I'll save the results for each.

[adl]

Fumbled a bit getting the house wiring isolated from the router but got that done -- and maybe can keep it that way until Monday morning. (No telephone service but who wants to hear telemarketers anyway?) No dropped Internet over the last 14 hours -- but I've gone up to 38 hours without any under V17.

So far, V19test2 is looking pretty robust by comparison.

[adl]

Almost 37 hours without a disconnection while running V19test2 and the router isolated from the in-house wiring & telephones. Approaching a new record for the DGND3700!

This configuration can run for another 22 hours but then I'll have to reconnect the telephones. No drops before then will NOT be conclusive evidence. What comes afterwards could be.

[Scubbie]

I think the fact that you have gone from several disconnects per hour to none in 37 hours can be used to confirm that your wiring needs to be looked at.

As mentioned previously, I frequently have seen instances where there has been the case.

Essentially I would suggest that you should plug the phone into the phone port beside the router, without any extensions. If this continues to work, then it is the wiring at fault, otherwise I would check the telephone itself.

I appreciate it is time consuming, but you need to eliminate different factors and slowly reintroduce things.

If your internal wiring has been knocked together and isn't using proper connections, then now would be a very good time to get a new extension cable. Especially with ADSL2+ any slightly damaged connections will interfere with the signal. A filtered faceplate at the master socket can isolate the internal wiring, but if there is something really bad, it can still cause problems.

[adl]

Originally Posted by Scubbie

I think the fact that you have gone from several disconnects per hour to none in 37 hours can be used to confirm that your wiring needs to be looked at.

You're jumping to a conclusion. It implies but confirms nothing. We've had 38 drop-free hours before, under V17. Also 34 hours and 28 hours. The distribution of disconnects follows no particular pattern that I can discern.

As stated here before, absence of evidence does NOT equal evidence of absence. If we end up with 60 solid hours by tomorrow morning then do the same thing again after I restore the "normal" configuration that will not confirm that V19test2 solves everything. If, however, the disconnects resume then I will have more testing to conduct.