DSL stats graphs microanalysis

[#1]

Quote from: Anotherone

As I'm sure you appreciate ADSL2/2+ has no direct bearing on the upstream, the resync is more likely to have resulted in what you've seen.

If you were not referring to ADSL2/2+, then what other less likely causes were you referring to then? What was the resync more likely than?
It's going to be a pretty pointless debate if you turn around and say you didn't mean what you last said in every other post.

[#2]

I have also heard an explanation given for ADSL2+ performing worse than ADSL2 on long lines is said to be due to the higher frequency signals of ADSL2+ somehow turning into additional noise on the lower frequencies. I'm not really convinced that that explanation is correct for two reasons. Firstly, I think that there'll be very little or no signal transmitted onto the higher frequencies that have zero bits allocated to them. Secondly, looking at the bit loading graphs, this extra noise would somehow have to impact the upstream and lowest portion of the downstream the most, and have little or no impact on the higher part of the ADSL2 downstream band. Yet the lower frequencies should have the strongest signal so they should be less susceptible to noise.
The bit loading graphs from the 582n on ADSL2+ vs ADSL2 have the same different shapes as graphs from my netgear, with reduced bit loading on each end of the upstream band and the lowest end of the downstream band.

[#3]

You have a particularly nasty way of putting things when you think someone doesn't agree with every detail of what you have posted, and so you may wonder why people don't engage with you in threads very frequently. I have neither stated nor implied I didn't mean what I said, and if I change my mind about something or decide I have something wrong, I will clearly say so. In the meantime, I'll get back to this later when I have time and can be bothered.

[#4]

Well perhaps you'd like to explain what you feel is "particularly nasty" about whatever I said or how I said it, so that I can try to avoid doing the same thing again in future. I explained how I interpreted what you said because you often claim I'm twisting your words. Since you've accused me of being particular nasty in this public forum, I'd prefer your explanation to also be in public rather than in a private message.
Personally I really don't see there's any point in stating you'll reply more later, I expect we've all got other things to do.
This thread wasn't really soliciting any input from anyone. Although there do happen to be more than one other thread with people on very long ADSL lines on ADSL2+. I don't think there's any benefit whatsoever from them continuing to operate on ADSL2+.

[#5]

That's your opinion, and your rarely put it in a pleasant way, and that's my opinion

Quote from: ejs

It's going to be a pretty pointless debate if you turn around and say you didn't mean what you last said in every other post.

You have implied something that you think I'm going to do - so you are a mind reader
If I'm going to change my mind about something I've said, I'll clearly say so and where I disagree with you, I'll clearly say so.
In the meantime, whilst I am here, there are other things deserving my attention. I'll get back to you in due course - your continued sniping will not help matters.

[#6]

Some DSL stats graphs with some non-zero error rates on. Are they causing any particular problem? Not really.

[#7]

There is, of course, a level where the CRC error rate does become more problematic. The 3 disconnections around 7am / 8am on those 2 graphs were trigger by me. Things are back to more usual levels today though.

[#8]

I did finally lose the 3db profile on 9^th April, which was not surprising considering the vast amount of errors leading up to that.
On 18^th June I decided to add a bit more stability in an attempt to eventually get the 3db profile back. After 2 days, I see this morning that the DLM decided to switch interleaving off. In my case, I got more bandwidth with interleaving on, presumably the coding gain from the FEC resulted in more bandwidth than was used up to carry the FEC data. I was not really aiming for the ping time to go from 20ms to 11ms, since the 20ms ping time with interleaving on didn't seem too bad anyway.

[#9]

Daily graph for Monday 12 October - Before
Old style LJU master socket and one extension socket, star wired including a section of non twisted-pair cable.
Daily graph for Monday 26 October - After
NTE5A master socket plus SSFP, extension socket with section of non twisted-pair cable now filtered off at SSFP. ADSL lead also replaced with a bit of CAT5 cable with RJ11 plug on one end.
So quite a few changes, which resulted in ... almost no difference whatsoever.

[#10]

There are literally no prizes for guessing what if anything happened next.

[#11]

Quote from: ejs

...........what if anything happened next.

Your line went dead?

[#12]

Are you any closer to identifying the sources of interference?

[#13]

Nope, line not dead. As for sources of interference, no, I'm not currently doing any investigations, there hasn't really been many signs of anything around recently except perhaps on Mondays late afternoon to early evening.
Yesterday the DLM decided to reduce the downstream target SNRM to 3, probably due to that high upstream SNRM over the last 4 days, which only appeared to be due to the low upstream speed. 
This resulted in me having to enable the +2 dB SNRM tweak, but still, I suppose the net result is a very slight increase in bandwidth.

[#14]

DLM on 21CN does some stupid things on occasions. Making adjustments to the Downstream simply because of the Upstream stats is one of them, and this is the 2nd instance that I'm aware of, although the other case was where it raised the Target SNRM on the DS because things were a bit iffy on the Upstream. That connection had been perfectly stable both before and after!