Print

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

[sheddyian]

Now, here's a thing.

Since around 4pm yesterday I've had Routerstats running, and I've just checked it.  The tx and rx noise margin stayed the same throughout the entire run, not one glimmer of change!  Is that normal?

I wondered if it might be stuck, so I unplugged the RJ11 lead and replaced it with the one that gives me the higher S/N margin, and the graph showed this improvement once synch was restored.  I didn't restart the modem nor did I do anything to routerstats, just swapped the cables.

Attached are plots - they don't show the whole of the run (I hadn't set it to log it all) but I scrolled the graph all the way to the start to look at it, and take my word for it, from 16:00 yesterday until I changed the cable at around 13:20 just now, the plots were entirely flat line.  And you can see the small improvement in noise margin and synch speed when swapping from the fancy twisted pair screened RJ11 to the 26AWG flat untwisted RJ11 cable.

Ian

[burakkucat]

Now, here's a thing.

Since around 4pm yesterday I've had Routerstats running, and I've just checked it.  The tx and rx noise margin stayed the same throughout the entire run, not one glimmer of change!  Is that normal?

No, definitely not. I suspect a glitch in the recording software (Routerstats). 

[sheddyian]

I just paused routerstats, looked at the the s/n figure on the modem web interface, and it reported a higher figure than routerstats had been plotting.

Resumed routerstats and it started plotting the higher figure.

So it does look like a bug, or something is getting cached somewhere.

Ian

[burakkucat]

Are you using the latest version of Routerstats? Do you think it would be worth mentioning to John Owen, creator and maintainer of the package?

[c6em]

I've had some difficulty on occassions like this
Only on R.Stats using telnet - and never on R.Statslite.
Either it won't record and runs at zero, or it records and although the data seems to be getting to R.Stats in the terminal window on telnet - its not being correctly plotted so I get the straight line syndrome...seemingly from the last point some time before just re-plotted.

HOWEVER in my case after only a bit it rights itself and starts behaving normally.  If it was running at zero I see a sudden jump up to the correct value of SNRm, sync etc.
It is possible its linked to a router or computer reboot.  If I switch off R.Stats and put the computer on standby ovenight I never get any problems, but if I've switched computer totally off then sometimes I get the hassle.

I did a router re-boot today (computer on standby) and sure enough I got the zero plot to start with.....

[les-70]

 Maybe I am missing something but I am a bit puzzled by all this.  When ever you change anything such as a cable that means a reconnect and resync. The modem will aim at its set SNR margins for upstream and downstream.  Usually it connect close to but not exactly at these margins.  To evaluate a component I think you have to take it in a out at least a few times in succession at a time of day when the SNR seems be stable and then look at mainly the Sync rates and  not so much the SNR.  Generally the better component will give a higher Sync with the SNR's being roughly fixed by the target values.  If the SNR varies a bit after resyncs it would be better to estimate what the sync should have been to give the target of perhaps 6db or 9db (depending on your DSLAM settings).  This can be done with a rough ratio of sync speed increment to an SNR increment of order 0.5Mhz per db of SNR.

[sheddyian]

Are you using the latest version of Routerstats? Do you think it would be worth mentioning to John Owen, creator and maintainer of the package?

I was using the previous version, but have just downloaded the latest.  Also downloaded Routerstats lite, which I've just been running, but it doesn't plot the upstream S/N margin, which is the one that seems to vary more then the downstream.  The upstream is also the figure that changes when I use different cables, and which I can roughly predict what it'll be, based on which cable I use 

And yes, I will have a look on the author's site and see about contacting them.

Ian

[sheddyian]

Maybe I am missing something but I am a bit puzzled by all this.

So am I 

My S/N margin is pretty steady (regardless of routerstats showing it as a completely flat line, from looking at the figures, it doesn't change that much)

I've just been trying another cable, an RJ11 twisted pair lead I made myself a while back.  This consistently gives me around 8.3db noise margin UP. 

The "magic" flat RJ11 cable consistently gives me around 9.5db noise margin UP.  Having just put that cable back again,. I'm currently getting 9.9db UP.

My up synch speed is always 1020, never higher, never lower.

My down synch speed fluctuates slightly at every resynch, but never by much. 

It does seem odd that one particular cable gives me better figures than others, especially given it's relatively short length compared with everything else. 

I shall stick with this "magic" basic flat RJ11 cable, since it's clearly what the modem likes

Ian

[c6em]

On SOME modems RSlite will plot the US SNR
Go to the SNR graph and with mouse over graph area - right click
Drop down menu appears - tick plot US noise.
However it may not actually work on some router model - it plots a zero line.

[les-70]

  That clarifies things a bit.  I assume the upstream sync is pretty well at its max assuming you are on an ADSL2+ connection.  In that case as you say, the sync is fixed and the SNR varies and for Upstream a bigger SNR should give less  errors.

  What happens down stream?  Again assuming you are on ADSL2+ your sync is excellent but does that Sync and the matching SNR vary?

  I also by the way always do a little better with an old flat cable (it came with a D-Link). I suspect it is impedance matching that is making one better than another.

[sheddyian]

On SOME modems RSlite will plot the US SNR
Go to the SNR graph and with mouse over graph area - right click
Drop down menu appears - tick plot US noise.
However it may not actually work on some router model - it plots a zero line.

Thank you, I hadn't spotted that!  It does work on my modem (Dlink DSL-2780) and I'm now successfully plotting a slightly fluctuating U/S SN margin, along with a constantly steady D/S!

I shall watch that with interest 

Ian

[sheddyian]

  What happens down stream?  Again assuming you are on ADSL2+ your sync is excellent but does that Sync and the matching SNR vary?

  I also by the way always do a little better with an old flat cable (it came with a D-Link). I suspect it is impedance matching that is making one better than another.

The downstream is pretty stable on 6db S/N margin.  It occasionally drops to around 3.5, but not for long.  I thought I'd discovered why - it seemed to be when the sun was heating the drop cable on the west facing wall in the afternoon, but it's not consistant, so    But it doesn't drop often enough to bother me.  I think I've only ever had 1 resynch since November 2011!

Impedance matching seems a very plausible answer to why the S/N margin is better with one cable over another.

Ian

[sheddyian]

Just for interest, here's a plot from Routerstats lite, which seems to work correctly on my modem.  Couldn't get full version Routerstats to plot anything other than the first sample value for noise - just gave me straight lines.

Top trace is upstream, bottom trace is downstream.  The resynchs are where I unplugged the RJ11 cable to try different ones, the final cable being my preferred "magic" RJ11 cable that gives best performance.

Having sung the praises of my Dlink DSL-2780 modem as being reliable, Routerstats Lite broke it (temporarily).  After around 4 hours of monitoring, the stats all dropped to zero, and the web interface on the modem wouldn't respond.  Modem and router function carried on working though.  A reboot cleared the problem.

Ian

[sevenlayermuddle]

Interesting, Ian.

Just to confirm...are the various cables exactly the same length, and follow exactly the same route (ie, does each exactly overlay the other)?

If not, it introduces another uncertainty.  The cable acts an antenna picking up interference, and the slightest variation in position or orientation can have a significant effect on that antenna's efficiency in picking up interfering signals.

I'm probably not alone in remembering early 60s VHF TV reception, when the only way to get a decent picture was to persuade Gran to stand at some very specific point in the corner of the lounge with the Aerial held aloft, and a teapot in her hand.  Actually the teapot may not have made much difference but it suited her image, so we never let on.

[sheddyian]

Ok, out of curiosity, I've done a few more tests.

Firstly, the cable runs : the modem is on a shelf about 80 cm above the NTE/5 with a filtered faceplate.  The cables do differ a bit in length, see the table below.  Each cable followed the same route, namely straight up from socket to modem, although any excess was left to dangle downwards.

Moving the cables about, or even winding them through a ferrite core whilst the line was synched didn't seem to make any difference to the data - I watched the graph as moved the wire or wound it, then left it a bit.


Cable   Length    Description

  1      203cm    "Magic" RJ11 flat untwisted cable that gives me best S/N ratio. 26AWG
  2      100cm    ADSL Nation Screened Twisted pairs RJ11 cable
  3      124cm    Home-made RJ11 single twisted pair (from old solid core CAT5 cable)
  4      100cm    Original flimsy RJ11 flat pair supplied with modem, 28AWG
  5      184.5cm  RJ11 to BT plug from old 56K modem. (Plugged into NTE/5 test socket)

 (Lengths are measured plug tip to plug tip)


View the attached graph in conjuction with the cable descriptions above.

You will see that there isn't a significant synch speed change, but that cable 1 consistently gives the highest upstream S/N ratio.  The unlabelled start of the plot was from previous days synch, still connected using cable 1.

I was also interested in the amount of ADSL noise radiated from the different cables.  I used a MW radio, held at a distance from the vertical RJ11 cable, and moved it around to determine the distance where I could no longer detect ADSL noise from the background hiss.


cable                         distance
  1                      21cm
  2                       5cm
  2 (earthed screen)      1cm
  3                       5cm
  4                      10cm
  5                      omitted test - oops!


Unsurprisingly, the screened cable (no.2) radiated the least noise, which was even further reduced when the screen was earthed at one end.  Even at 0cm proximity in contact with the MW radio, the ADSL noise was noticeably muted compared with other cables, and I could instead detect some processor noise from the modem.

Make of that what you will, I know which cable is my favourite, and which sees to give me fairly consistent synch speeds

Ian