Line Stats Analysis Results
This thread will contain the results of the analysis of everyone's line stats that have been posted in the Data/Stats Collection Thread.
My initial analysis is not focused so much on distance from the VRAD. The reason for this is that not too many people have distances that have been told to them by a tech using their specialized meters. Furthermore, different line conditions can cause inaccuracy in their measured distances.
As it turns out, distance from the VRAD isn't the most important thing regarding the type and quality of service that you get anyway. Instead, the wiring gauge, presence/absence of bridge taps, and the noise margin (signal-to-noise ratio) of the line are much more important.
To this end, the data that has been posted in the Data/Stats Collection Thread led to a rock-solid and extremely useful correlation that can be used to immediately judge the quality of your line and the service you can receive. That correlation is the Downstream Noise Margin and the Max Line Rate.
Using only these two parameters, you can tell if A) Your line is working properly (i.e.. no bridge taps), B) What gauge wire is running from the VRAD to your NID, and C) What line profile you should be able to get.
Please note that I arrived at the wire gauge conclusions here because there are two distinct groupings of data points. I am assuming the difference between them is due to wire gauge, because A) It is known that different neighborhoods around the country use two different wire gauges, 22 gauge, and 24 gauge, B) Thinner wire (24 gauge) would have higher attenuation in the high frequencies, which would reduce the max line rate for a given signal-to-noise ratio, which is exactly what the graph shows. It is possible that the two groupings of data are due to a different cause, but until I either gather more data or have references to other possible conclusions, I'm going to assume the difference is due to wire gauge even though I cannot currently prove that with certainty.
There is a 3rd grouping of data that shows some people with a drastically reduced max line rate. I'm fairly certain these people have a bridge tap on their line that is reducing the max line rate. If their line was conditioned by I&R to remove the bridge tap, their max line rate would jump up into one of the other two groupings. I have witnessed this first-hand, as my initial installation was on a line that had a bridge tap. When the bridge tap was removed, the max line rate shot up by over 15000 Kbps. I unfortunately cannot compare the numbers that I had at that point to the numbers I have now because those older numbers were using the VDSL1 protocol and at least 2 revisions earlier of the RG firmware. Those values cannot be reliably compared to today's values obtained with the VDSL2 protocol and the current RG firmware revision (5.29.135.47).
Here's the first fully analyzed chart:
You can plainly see how well the data groups are formed here. There are two distinct lines of people with properly working service, and a 3rd line of people with service that is probably not running right. In addition, you can see the very rare people who are so close to the VRAD that they are currently being capped to a maximum line rate of 64000 Kbps.
For those of you who have some background in statistics, the linear correlation coefficients (R^2) for each of the 3 trend lines on the chart was > 0.98. (This means that the computed lines fit the actual data very nicely).
The allowable line profile areas were computed by assuming that a line capacity of 80% is required for properly working service. Line capacity = downstream profile rate / max line rate. For example, for a properly working 32/5 profile, the downstream profile rate is 32200 Kbps. FOr a line capacity of 80%, that means that Max Line Rate = Downstream Profile Rate / 80% = 32200 / .8 = 40250 Kbps, which I rounded to 40000. The other rates separating the profiles were computed similarly. The exception is the border for No Service, where I upped the allowed line capacity to 85% to allow just a little less max line rate.
The people whom I believe have a bridge tap on their line are very interesting. The bridge tap reduces the max line rate that the line could theoretically carry by nearly 30%. In the particular case of the person with an 18.5 dBm noise margin and 29000 Kbps max rate, having the line conditioned could conceivably get him a max line rate of 41000 Kbps, which would boost his allowable profile from 19/2 to 32/5 ! He could conceivable go from 1HD/3SD to 3HD/1SD just by having the bridge tap removed.
The person with the 10.0 dBm noise margin and a 19200 Kbps max rate is certainly experiencing poor service. The line capacity in his case is at 100%, with no margin for errors.
In short, this chart can tell you at a glance whether your service is operating in agreement with everyone else's service. It can immediately identify what profile you should be on or be able to get. It can tell you if your line needs conditioning to have a bridge tap removed. Based on my assumptions, it can also tell you your wire gauge, although that parameter is questionable.
When people post their line stats, this chart can be very useful in quickly determining if they are experiencing line problems or if their service is in agreement with normally working service.
Hey SomeJoe, many MANY thanks!
Now, let me play the part of the forum dummy (a role I'm good at) and ask some dumb questions. Looking at the chart, I see four red squares in the yellow area. These four squares represent four Uverse users in the 19/2 profile, correct? The lowest square is somewhere around 9 -10 dBm and the other 3 are around 12 dBm. If the two users with the highest squares could clean up their stats somehow (and get an additional 4-5 dBm improvement) they could possibly get into the 25/2 profile (green area), correct again? Now, if my first two assumptions are correct, and knowing my own stats, I believe I'm one of the two red squares just above the line into the blue 32/5 profile range. That is probably not good enough to have reliable trouble free 32/5 but I should have excellent 25/2 service (?).
The five folks with green diamonds need to call Tech Support.
You folks with blue diamonds...... may the fleas from your neighbor's dog hop the fence and infest your yard! 
Yes, the four red squares in the yellow area are users with properly working service on the 19/2 profile. And yes, your square is the 17 dBm/41000 Kbps. As long as your uncorrected blocks are zero or very low, it should be enough for 32/5 in my opinion. Whether AT&T believes that or not, I don't know, so I don't know if they will push the 32/5 profile to you.
But unfortunately, there is not much that can be done to improve the noise margin if the service is working properly to begin with. The noise margin is dependent on attenuation (which depends on distance and wire gauge), and the background noise and intererence present on the line.
Sometimes for some people, adding a home run from the NID to the RG can improve the stats because the home run picks up less noise than the normal house telephone wiring. This can improve the noise margin and raise the max line rate. Other items can also add noise, like improper grounding or a bad VDSL balun. But most of the differences on the graph in max line rate for a constant noise margin reflect differences in the frequencies that the line can carry. The bigger wire (22 gauge) can carry more high frequencies, thus the max line rate can be higher. Bridge taps cause reflections and interference down the line that block the high frequencies, thus lowering the max line rate.
Gamers constantly curse the people connected to the gaming server who have the lowest latency, and there's a name for them -- the Low Ping B*stards (LPBs). I believe you have insinuated that I've now created a new category of people, the Blue Line B*stards (BLBs).
But seriously, the guys in the gray who are capped at 64 Mbps are the true snobs.
Great work and deductive reasoning SomeJoe. I can see where I'm on the graph at a 54778 max line rate and was just upgraded to the 32/5 profile. It should be noted that I worked closely with the prem techs and helped with the install. This is how I found out I'm 1,325 from the VRAD and the tech showed it to me on his meter. They also said it's great that I have the heavy duty 22 gauge wire. One of the tech's also went down and removed a bridge tap for the strongest signal. So in my case you are exactly right about having the 22 gauge wire, no bridge tap, and this is supported by the prem tech's findings.
Thanks for all your hard work on this project.
SomeJoe7777 wrote:
But seriously, the guys in the gray who are capped at 64 Mbps are the true snobs.
We're not snobs. We're just better than all the little people. 
oobleck wrote:
Was that me - 18.5 dB and 28,997 kbps? My line originally did have a bridge tap, but it was removed a year and a half ago. However, they switched me to a different pair 10 months ago. I wonder if the new pair might have a bridge tap on it. Presumably they would have checked that when making the switch, but who knows? Something to consider when I go HD.
The I&R's haven't been able to improve things much since then. Downgrading to 19 Mbps reduced the errors a lot, as did the switch to VDSL2.
You are indeed a choice test case, oobleck. Yes, that is your data point.
I would love to see I&R test your line and see if you indeed do have a bridge tap on it. I'm betting that you do, because your stats certainly don't line up with everyone else's.
Can you do me a favor and also run the Dohrenburg analysis test? Go to http://adslm.dohrenburg.net/tools/3800hg-stats.php and follow the instructions, which I'll also print here:
1. Open up a web browser and pull up the DSL diagnostics page from the RG's MDC using http://192.168.1.254/xslt?PAGE=J42&THISPAGE=A02_PO
2. Right-click on the page and select "View Source".
3. In the HTML source code window that pops up, click your mouse in it, then hit CTRL-A on your keyboard to select all the code. Then hit CTRL-C on your keyboard to copy all of it to the clipboard. Close the source code window.
4. On the Dohrenburg analysis page, click your mouse in the first box and hit CTRL-V on your keyboard to paste all the HTML source code. Then click the Process Form button.
5. The analysis page that appears will have a URL in green at the top of the page. Copy that URL and paste it in a post here.
The shape of the DMT distribution chart in the analysis page can also tell us if there is a bridge tap.
SomeJoe, thanks for all the time and effort spent on this project. It may (will!) take me a few readings to fully comprehend it all. Nonetheless, as a novice in the studies of UV, I find this fascinating. Thanks.
Any idea regarding gauge of wire as to when installed? 22 gauge more recent? Just curious. With a 59.9K max & 24 db noise margin in a sub division 20/25 years old, most likely 22 gauge?
As it turns out, distance from the VRAD isn't the most important thing regarding the type and quality of service that you get anyway. Instead, the wiring gauge, presence/absence of bridge taps, and the noise margin (signal-to-noise ratio) of the line are much more important.
I find this most interesting as we were all hung up on the distance thing.
aross99 wrote:
My noise margin moved up to 12dBm, and my max sync rate went up to 22533. I don't know if they physically did anything, but I will follow up tomorrow and find out why they didn't show.
That's good, improving the noise margin raised your max rate, but you've still got a bridge tap on the line. Your line capacity is no longer at 100% either, which will improve your errors as well.
Your Dohrenburg analysis is a classic bridge tap pattern. A bridge tap is a "T" connection in your line between the VRAD and you, where the end of the "T" wire isn't connected to anything. What this does is reflect the signal back to your line, which causes interference. The interference emphasizes some frequencies and attenuates others. What this does is give a sine wave pattern of noise on the line in the frequency domain. By looking at the DMT graph from the Dohrenburg analysis, you can see the sine wave noise show up in the bitloading of the different tones.
Here is your DMT graph from the Dohrenburg analysis page:
You can see the sine wave pattern that the bridge tap causes:
Continue to work with AT&T and the I&R group and see if then can remove the bridge tap. Between the removal of the bridge tap and the increase in the noise margin to 12.0 dB, I predict that you may be able to get a sync rate near 30-31 Mbps, which may be just barely enough to move you to the 25/2 profile, and therefore 2HD/2SD.
SomeJoe7777 wrote:You are indeed a choice test case, oobleck.
I would love to see I&R test your line and see if you indeed do have a bridge tap on it. I'm betting that you do, because your stats certainly don't line up with everyone else's.
Can you do me a favor and also run the Dohrenburg analysis test? Go to http://adslm.dohrenburg.net/tools/3800hg-stats.php and follow the instructions, which I'll also print here:
Here's the link:
http://adslm.dohrenburg.net/tools/db-link.php?link
I'll be interested to see what you find. When they changed my pair it didn't seem to make much difference one way or the other; I was still getting lots of uncorrected blocks. If they had changed me from a pair without a bridge tap to one with a tap I'd think it would have increased the error rate. But the error rate was pretty variable back then, so the effect might not be discernible.
--
oobleck
