Kitz ADSL Broadband Information
adsl spacer  
Support this site
Home Broadband ISPs Tech Routers Wiki Forum
 
     
   Compare ISP   Rate your ISP
   Glossary   Glossary
 
Please login or register.

Login with username, password and session length
Advanced search  

News:

Author Topic: Anyone know what would "knock out" the lowest 66% of D1?  (Read 2624 times)

HughPH

  • Just arrived
  • *
  • Posts: 8
Anyone know what would "knock out" the lowest 66% of D1?
« on: February 25, 2016, 11:35:42 AM »

Using an unlocked HG612...

I've written some code to plot SNR over Time, and it shows 66% of U1 being relatively low, and then suddenly returning to normal. Obviously the inverse also happens - a very sudden cut of signal strength across 66% of U1.

When I say "suddenly" - the graph is updated once every second. So this graph being approx. 950px wide, and captured at 11:22, I guess it happened about 11:10.

There seems to be no correlation with electrical hardware in our house.

With the line in this condition, I can sync at ~27500Kbps (sometimes even up to ~28000Kbps) - when it's in the pre-11:10 condition, it will only sync at ~25500Kbps. If I sync at 27500Kbps, I can keep that sync and the SNR margin just reduced to about 4.2dB - the code was originally written to monitor Attainable Rate and resync if a great enough gain would be achieved by doing so.

Any ideas about this noise, though?

Many thanks.
Logged

WWWombat

  • Kitizen
  • ****
  • Posts: 1674
Re: Anyone know what would "knock out" the lowest 66% of D1?
« Reply #1 on: February 25, 2016, 11:40:43 AM »

We might need a better image than that one...
Logged

HughPH

  • Just arrived
  • *
  • Posts: 8
Re: Anyone know what would "knock out" the lowest 66% of D1?
« Reply #2 on: February 25, 2016, 11:56:05 AM »

Unfortunately I only have 1080 pixels to plot on the screen, and I compressed it so I could put it alongside the SNR Margin plot.

I could dump out a bigger image to disk at the same time - and perhaps improve on the palette to make smaller variations more clear.

0 is at the top, n is at the bottom where n is the highest used tone from pbParams.
Logged

HughPH

  • Just arrived
  • *
  • Posts: 8
Re: Anyone know what would "knock out" the lowest 66% of D1?
« Reply #3 on: February 25, 2016, 12:04:18 PM »

Because of that scaling it can look a bit weird if I reconnect and more tones get used...

I'll get the larger plot put together.
Logged

WWWombat

  • Kitizen
  • ****
  • Posts: 1674
Re: Anyone know what would "knock out" the lowest 66% of D1?
« Reply #4 on: February 25, 2016, 12:18:09 PM »

OK ... that images looks almost the same, and equally uninterpretable.

What are we supposed to be seeing? I just see blocks of red/green colour, with little variation horizontally, but lots vertically. But I have no idea what it means.

It might be helpful if you could capture snapshots of the more standard SNR/tone graphs at "normal" and "low" times too.
Logged

HughPH

  • Just arrived
  • *
  • Posts: 8
Re: Anyone know what would "knock out" the lowest 66% of D1?
« Reply #5 on: February 25, 2016, 12:29:11 PM »

If you imagine a normal columnal plot (or line graph plot) of SNR, color each point with the magnitude (black being 0, red being 1, green being 60), rotate it on the X-axis, as if you're looking down on it, then rotate it on the Z-axis, so it appears vertical, then the right-most column is the current graph and everything to the left is history. Tone 1 is at the top of the graph.

In the first image posted, the lower 66% of D1 takes up about 40% of the top of the graph - ignoring U0 which was disabled in an attempt to minimise any possible harmonic cross-talk from the highest amplitude signal. Reading from left to right, it's evident that D1 becomes brighter about 400 seconds (pixels) in.

Unless you want to look at specific tones, a higher resolution version probably won't help you.
Logged

gt94sss2

  • Kitizen
  • ****
  • Posts: 1281
Re: Anyone know what would "knock out" the lowest 66% of D1?
« Reply #6 on: February 25, 2016, 12:34:28 PM »

Using an unlocked HG612...

I've written some code to plot SNR over Time, and it shows 66% of U1 being relatively low, and then suddenly returning to normal.

Like others, I am having difficulty following the attachments.

While writing code isn't something I can do - have you seen the following which I think do something similar to your code:

DSLStats
HG612 Modem Stats

While both programs allow you to monitor your connection locally, you can also upload the data to https://www.mydslwebstats.co.uk/

I note in another thread you offer to share your code, and I guess the authors of these programs and others might be interested - they can both be found in the Router Monitoring Software board on this forum.

Apologies, if you already know all this! :)
Logged

WWWombat

  • Kitizen
  • ****
  • Posts: 1674
Re: Anyone know what would "knock out" the lowest 66% of D1?
« Reply #7 on: February 25, 2016, 02:54:31 PM »

If you imagine a normal columnal plot (or line graph plot) of SNR, color each point with the magnitude (black being 0, red being 1, green being 60), rotate it on the X-axis, as if you're looking down on it, then rotate it on the Z-axis, so it appears vertical, then the right-most column is the current graph and everything to the left is history. Tone 1 is at the top of the graph.

I think I got this part

Quote
In the first image posted, the lower 66% of D1 takes up about 40% of the top of the graph - ignoring U0 which was disabled in an attempt to minimise any possible harmonic cross-talk from the highest amplitude signal. Reading from left to right, it's evident that D1 becomes brighter about 400 seconds (pixels) in.

This is the hard part to interpret. The first problem was just figuring (easily) where the different bands were. Some horizontal line in a different colour (white?) would help identify the bands, and make it easy to see the affected spectrum.

The second problem is harder ... the colour represents 60 SNR values, so it can be hard to judge whether an apparent change from "light green" to a "lightish green" signifies a big change in SNR. Or, perhaps the change in colour is an apparent one because one or two rows of pixels change to "light green", but were already surrounded by "light green" - making the colour more solid. Each pixel represents 10 tones, so the aggregation technique becomes important too.

The third problem is consistency. In the original post, you mention U1. In the title and this post you mention D1. I guess you mean D1?

IMO, your graph is really good at telling me there was a change (and in this case, just the one change), and how widespread it was across the spectrum. But it doesn't give me a very good feel for how big the change was (in dB) nor how individual tones were affected.

Having seen that, I would then move on to look at the "normal" graphs - one that plots the summary SNR over time, over the same time period. And one that plots the SNR/tone - and see one of those from just before the change to one just after.

Without knowing any of those, though, I will take a guess:
- Your problem is one of crosstalk, caused by other FTTC subscribers. In this case, it is likely to be one single subscriber.
- If this was a one-off occurrence, then the subscriber was just activated.
- If this happens regularly, then they have a habit of turning their modem off when not using the internet
- The scale of difference is of the order of 1-5dB across a whole chunk of spectrum
- The impacted spectrum is likely to be the overlap between the spectrum your modem can use and the spectrum their modem can use.

That is an uninformed guess, mind!

I've attached one of the logs of SNR over time from my line, from just a few days ago. That was likely caused by crosstalk from a new subscriber. Unfortunately, I wasn't logging the SNR/tone graphs in a way that lets me check them.
Logged

HughPH

  • Just arrived
  • *
  • Posts: 8
Re: Anyone know what would "knock out" the lowest 66% of D1?
« Reply #8 on: February 25, 2016, 04:03:38 PM »

Many thanks for your input WWWombat - if you've ever seen an FFT, this is a similar sort of thing:
X is Time
Y is Frequency
Z is Magnitude/Amplitude/etc

You're right, I meant D1. Upload speed is of little concern because I'm on Plusnet's 38x2 package - my upstream sync could be as high as 4Mbps but it's capped at 2Mbps. And it's not worth paying for the Extra package just to get 2Mbps extra upstream bandwidth.

Since I'm cutting the graph down to only the tones used, we're actually seeing (at the time of writing) ~1500 tones:
Medley Phase (Final) Band Plan
US: (871,1205)
DS: (33,859) (1216,1497)

So with a height of ~500px, we're seeing ~3 tones per pixel.

U0 is the black band at the top, D1 is the long section, then there's a gap before we see U1 which is the very gentle gradient followed by another gap and then a very "gritty" D2.
This was really built for me to look at, so it's just not designed for public consumption - if I were to dump a full 4096-pixel-high graph, I could document the banding.

I see what you mean about differences being difficult to see - although the one I'm pointing out here is fairly clear. I was thinking about changing the palette to run through black-red-orange-green-cyan-white. A legend would work nicely, too.

This could be the neighbours ADSL connection - unfortunately this isn't very clear, but it does seem to indicate ADSL overlaps with VDSL.
Logged

WWWombat

  • Kitizen
  • ****
  • Posts: 1674
Re: Anyone know what would "knock out" the lowest 66% of D1?
« Reply #9 on: February 25, 2016, 04:44:59 PM »

Right, I have something of a better feel, I think.

The bottom 2/3 of D1 is going to roughly run up to tone 500-550; ADSL2+ would  reach tone 512, if you are close enough to the exchange.

VDSL2 does indeed overlap ADSL (and ADSL2, and ADSL2+), but the FTTC cabinet is forced to drop power in the overlapping frequencies so as to not interfere with ADSL signals as they pass by.

In an SNR/tone chart, you can usually see a clear notch that is design specifically for that cabinet (it varies, based on the electrical distance from the exchange). I've attached a copy of mine, where the trough of the notch appears at tone 300-350.

Your notch looks to end around tone 375 (lower tones have the fuzzy red/green colouring; higher tones suddenly become a clear green block), so I'd guess your cabinet is about 1.5km from the exchange; Your speeds of ~25Mbps suggest you are 1.2km from the cabinet.

With a notch ending at tone 375, your DSLAM is located in a place where ADSL2+ will not be seeing tones as high as 512. As your interference picture suggests you are seeing interference up to around tone 500-550, your line is highly unlikely to be encountering that interference from an ADSL2+ subscriber. If crosstalk is the problem, it is much more likely to be an FTTC subscriber.

Note that the crosstalk is from one or more pairs in the cable neighbouring your pair for some of the cable length. It doesn't actually have to be someone living close to your house!
Logged

HughPH

  • Just arrived
  • *
  • Posts: 8
Re: Anyone know what would "knock out" the lowest 66% of D1?
« Reply #10 on: February 25, 2016, 08:37:32 PM »

Thanks again! I'm in a rural area with 6 houses in the immediate vicinity, 2 on the corner further away from the cabinet, 1 on the corner of the road that the cabinet is on, and another 5 or 6 on the road before the cabinet. Chances are it's one of my neighbours...

The cabinet is actually situated outside the front of the exchange - until recently we had an exchange only line, which insultingly meant we couldn't get VDSL - but they fitted a new cabinet and a fibre cabinet on the doorstep and hooked us up. The first I knew of it was junk mail from BT - I didn't believe it at first, so I checked everything again.

We are approx. 1.2km from the cabinet. Some of the houses served by the same cabinet are 4km away from it - sucks to be them!
Logged

WWWombat

  • Kitizen
  • ****
  • Posts: 1674
Re: Anyone know what would "knock out" the lowest 66% of D1?
« Reply #11 on: March 01, 2016, 07:02:30 PM »

Sorry, my brain has been too fuzzy for the last week, so I've not been up to much detective work...

Its good to see my estimate for your distance to the cabinet was good, but I'm disappointed about getting the other part wrong. If the cabinet is straight outside the exchange, then there should be no notch at all.

I would expect your image to show good solid green at the lowest frequencies, with a single "descent" to the red and black. No sudden improvement back to solid green around tone 375.

It would help me visualise things better with a copy of the SNR/tone graph and/or the bit-loading/tone graph from one of the packages in @gt94's post.
Logged
 

anything