Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Weaver]

This table shows the downstream and upstream sync rates for my lines currently, and with the values of the fractional deviations from the geometric mean are given for both directions.

Line	Downstream				Upstream
	kbps		dev		kbps	dev
1	3037		4.27%		515	5.77%
2	2803		-3.76%		555	13.99%
3	2891		-0.74%		435	-10.66%
4	2924		0.39%		452	-7.17%
Geo mean:	2913				487

* I am sure I have asked this question before, but - why the differences?

The lines are supposed to be identical apart from line 2 which probably has a different story E-side now. (I think now we have established I actually have a D-side and an E-side now, but I take it that I had no such thing before - Is that correct? Because the PCP cab now shown in photos of the main road at Harapul is not there in Google StreetView earlier.)

Could it be down to different joints, of better or worse quality?

Pairs 1+3 are in the older drop cable into my house and 4+2 are in a second drop cable. They are in that order chronologically, ie. 1 being the oldest, then 3, then 4 and 2 the most recently made live.

I can’t see how invoking crosstalk really helps us explain the precise results. Otherwise I am stumped.

[andyfitter]

How can they *ever* be identical? Nothing in the world is. They might be nearly identical lengths, but that’s about it. Different places in the cable bundle, different levels of interference from different crosstalkers, minute differences in attenuation from a whole load of factors. If everything were identical we’d all get the same sync rate every time we did a resync.

On lines as long as yours all of these factors probably play a far bigger part, as you’re at the extremes of the technology.

I’d be more surprised if they were exactly the same.

[jelv]

I'd agree with you in respect of the downstream, the variation Weaver is seeing is about what I'd expect.

But around a 25% difference between the best and worst on the upstream?

[dee.jay]

Whilst I agree, the upload figures are very low anyway, so a small variance will be a bigger percentage, I think?

[burakkucat]

The lines are supposed to be identical apart from line 2 which probably has a different story E-side now. (I think now we have established I actually have a D-side and an E-side now, but I take it that I had no such thing before - Is that correct? Because the PCP cab now shown in photos of the main road at Harapul is not there in Google StreetView earlier.)

Correct. Certainly for line number two. What is still not certain is whether lines one, three and four now have what can be considered as E- & D-sides.

[Black Sheep]

Whilst I agree, the upload figures are very low anyway, so a small variance will be a bigger percentage, I think?

I too would agree with that thought process, dee.jay.

[dee.jay]

Yeah, if we were talking about 4 x 100Mb circuits, the percentage in variance would be tiny...

However, ~500kbps upload, dang.

[Weaver]

@burakkacat Thank you for sorting out my confusion

@blacksheep @dee.jay - Do you think the fraction / percentage should stay the same for slow lines compared with faster ones?

Mind you there is also the fact that here slow means long, so perhaps with a long line there are more opportunities for variation.


I still don’t understand why lines are not absolutely identical in my case, within statistical error that is. We should perhaps leave out line 2 because it could now have a physically different story. But the others are not the same even on the downstream, never mind the upstream.

You would expect some variation. but there still has to be an actual why. Unless it is just measurement error, or time variation, or difference between resynchs, or DLM messing up the results. I wonder if some bogus variation could be produced by simply resynching at different times, when noise levels are better or worse in the day. Perhaps modems should be forced to resynch to make things fairer. I should also check what the SNRMs are.

[Weaver]

Does anyone have candidate theories as to what the contributions could be to the differences? Some were mentioned earlier in the thread.

I am wondering about different crosstalkers. Maybe some are near to noisy problem crosstalk sources and others are further away from those sources.

[Weaver]

Crosstalk decreases with increasing trace separation.
The figures for this web page show just how fast crosstalk varies with decreasing distance.

Trace pitch (mil)	Crosstalk (mV)
13	220
15	124
17	69
19	38
21	22
23	12
25	7

[j0hn]

That page is discussing "crosstalk" on a multi layer, Printed Circuit Board. I used to make circuit boards for a living.
I spent all day loading prepreg in to a giant heated press to bond the multiple layers.

I am wondering about different crosstalkers. Maybe some are near to noisy problem crosstalk sources and others are further away from those sources.

Yes, that is likely. Remember that most crosstalk comes from other pairs within the same multi pair cable.
Pairs which are directly adjacent to each other for long distances within the multi pair cable will cause increasing levels of crosstalk.

You have 2 pairs sharing the same drop cable, twice.
A 20m run in your drop cable is nothing when you consider that your line is some 7km long. That's why you see little to no change between the lines when resyncing the surrounding lines.

With FTTC crosstalk can be much more noticeable, particularly 2 lines on a single drop cable.
I've seen someone with 80Mb sync who took out a 2nd line. This 2nd line knocks nearly 20Mb off the 1st line when synced.
The 2 pairs in question will share more than just the drop cable though and may well be in close contact within the multi pair cable all the way back to the cabinet.

[sevenlayermuddle]

Strikes me you need to consider the effect of these lines’ crosstalk on one another.   To illustrate, let’s imagine an installation with just two lines, A and B.   Imagine the two lines are abosulutely identical electrically, and suffer identical amounts of crosstalk from external sources...

..Now, from a situation where both lines are out of service, line A becomes active and syncs up.   Line B is idle and generating no crosstalk, so line A is able to negotiate a good speed.   But when B syncs up, it finds itself impacted by A’s crosstalk, so B gets a lower speed.

From that point on, if either line resyncs in isolation, A will always do better as B is generating less crosstalk, whilst B will fare worse because A is doing better.

[Weaver]

The point about drop cables is very much correct, clearly. So maybe it is all about who is near to whom ? But then why would some be noisier sources than others ? Perhaps some neighbours are not active, I don’t know if there are so many pairs in the bundle that many are inactive, so some lucky pairs might have few active neighbours.