Hi Paul..
I think Walter has nailed it....
My layman's understanding of your line stats graphs is that they suggest that something is physically wrong with your loop.
I just made a blog post about the four metrics in the graphs (BITS, SNR, QLN, and HLOG).[1]
Perhaps most interesting in your case is your chart of Hlog(f). I believe it illustrates the problem.
Hlog(f) is a logarithmic measure of the channel insertion loss. In other words, it measures loop attenuation at individual sub-carrier frequencies across the band-plan spectrum.
The plot of Hlog(f) needs to be read upside-down. It is actually inverted, insofar as the lowest point of the red zone charts the attenuation reading for a given frequency. (An attenuation measurement of -96dB means the data is not available or is out of range).
We can identify the following characteristics of the attenuation on your loop from Hlog(f).
The lowest utilised sub-carrier in the lower D1 downstream band is DMT ~#36 which is transmitted at ~150kHz. (Remember vDSL Profile 8c uses 4.3125kHz channel spacing, so 36 * 4.3125 = 150kHz.)
In your 'good' Hlog graph, a channel insertion loss of -6dB was measured at 150kHz. In the 'bad' Hlog graph, a channel insertion loss of -12dB was measured at the same frequency.
The highest utilised sub-carrier in the lower D1 downstream band is ~#690. This tone is transmitted at ~3.0MHz. In the 'good' graph, a channel insertion loss of -17dB was measured at that frequency. However, in the 'bad' graph, attenuation at 3.0MHz has jumped to -45dB.
In the second downstream band (D2), the lowest tone is ~#1195. This tone is transmitted at ~4.9MHz. The loop attenuation measured at the frequency in the 'good' graph was ~-25dB. However, in the bad graph, the measurement was ~-65dB.
The highest utilised tone in D2 is DMT #1625. It is transmitted at ~7.0MHz. The 'good' graph shows loop attenuation of ~-27dB at that frequency, whereas the 'bad' graph shows a loss of ~ -75dB.
So in the period of time between the data collection for those two graphs, your loop mysteriously developed a high attenuation profile across the whole bandplan spectrum.
The attenuation measurements by themselves are not that useful. The sensitivity of the ADSL transceiver units will actually determine whether the signal at a given frequency can be recovered.
What's more important for diagnostic purposes is the slope of the Hlog graph which takes the form of an exponential decay function. Characterising the Hlog graph is not straightforward. Since the ordinate scale shows negative decibels you could call it an exponential growth function as attenuation grows as the frequency gets higher.
It would be useful to use non-linear regression (curve fitting) to discover the parameters of the attenuation function for a given loop. That's what I am trying to develop as part of the port of the dmt tool for use with vDSL connections. I see that Huawei Corp has patented a couple of ideas in that sphere. If anyone would like to collaborate on an open source project to do this, please contact.
Once the parameters of the attenuation function are identified, it could be superimposed upon the 'classic' attenuation functions for the typical loop, enabling diagnosis by comparison.
For example, the parameters of the attenuation function for a (faultless) single segment twisted pair constructed to BT cable specification CW1128 - class 1 solid annealed 0.5mm plain copper conductors with cellular polyethylene (CEL-PE) insulator - will be well known to the egg-heads at Martlesham Heath.
They will also know the (higher) attenuation profile of the aluminum twisted pairs that were used as a substitute when copper prices soared at the hands of the City gangstas in the 1970s.
Here are some plots of the attenuation functions for twisted pairs using various wire gauges (AWG24 has 0.5mm conductors). The plots were taken from [2] and [3] and are only meant for illustrative purposes. The units of measure and the graph scales are all over the place. To develop anything out of it, we need the parameters of the functions to plot those graphs. Then they could be used for various comparisons.
According to a European patent filed by Huawei in Sweden, the attenuation profile of a loop can be used to accurately estimate its length and to gauge its physical condition.
Enough rambling!
Cheers,
A
[1]
http://huaweihg612hacking.wordpress.com/2011/10/01/measuring-line-characteristics-on-the-huawei/[2]
http://www.cc.gatech.edu/classes/AY2005/cs4251_fall/class5.ppt[3]
http://www.serialphy.com/attenuation-chart-24-awg-wire.html