Kitz Forum
Broadband Related => ADSL Issues => Topic started by: konrado5 on February 24, 2016, 12:18:10 AM
-
It seems I've finally discovered why my Hlog has a lot of undulations and lazy roll. It is related to the PSD masks or something of this kind.
The first attached Hlog is my Hlog. The second attached hlog is hlog of other user from Poland. This Hlog does not contain strange undulations. The third Hlog is Hlog of one other Polish user. This is somewhat undulated Hlog. I've highlighted similar patterns of Hlog by green.
On the second Hlog there is noticable risement of attenuation at the place of my lazy roll. On the third Hlog there are noticable similar undulations at the same place. It implicates that my Hlog measurement is the most sensitive to the PSD masks or something else of this kind. Perhaps my DSLAM reports very inaccurately output power at the attenuation measurement moment. It is important because attenuation is difference between output power and input power.
Best regards
konrado5
-
What do you think about my explanation?
-
What do you think about my explanation?
Not sure.
-
Why not sure?
-
burakkucat: what do you think about it? Especially with you I've talking about this Hlog.
Best regards
konrado5
-
Three different circuits, in Poland.
- Do you know if they are all with the same ISP/CP?
- Do you know if they are all terminated on the same make and model of DSLAM?
- Do you know if they all have the same PSD mask applied?
I am trying to think why there should be evidence of the PSD mask shown in the corresponding Hlog plot. At the moment I am having trouble convincing myself that it could be possible.
I am happy to agree that the attenuation may be expressed in terms of the received and transmitted power.
I am prepared to accept that certain equipment may be incorrectly reporting certain values.
At the moment, all I can say is that you have prepared a plausible theory but not, unfortunately, a proof.
-
Do you know if they are all with the same ISP/CP?
These users have other ISP.
Do you know if they are all terminated on the same make and model of DSLAM
They also have BDCM DSLAM but not the same model.
I am trying to think why there should be evidence of the PSD mask shown in the corresponding Hlog plot. At the moment I am having trouble convincing myself that it could be possible.
Is there other way to explain coincidence of patterns on Hlogs?
Furthermore, G992-3 mentions some accurracy issues with Hlog for ADSL2.
The accuracy requirements for the downstream HLOGps (HLOGps_ds) shall apply only to the
following subcarriers (with the corresponding frequency ranges being a part of the passband), and
only if not within the downstream BLACKOUTset (see clause 8.13.2.4):
•
Annexes A and I:
Subcarriers 46 to 208.
•
Annex L:
Subcarriers 46 to 104.
•
Annexes B, J and M: Subcarriers 92 to 208.
Moerover, my lazy roll correspond to the end of AnnexM EU-56 upstream band.
Do you know if they all have the same PSD mask applied?
It is possible. Notice: all of them have HAM band mask (1.8-2.0 Mhz).
Best regards
konrado5
-
I'm not at the pc, so haven't looked at the graphs... but iirc and going from memory which ones they are, those HAM tones which are blocked are specific to the msan make and model.
-
Meant to also add
Again I'm going entirely off memory, but when I was looking deeply into konrados stats last year, there was something which transmitted on those frequencies in Poland, but can't remember now what it was. I suspect the differing amount of undulations in the various graphs could depend on how close to a transmitter they are?
-
By chance I've got evidence that my Hlog is affected by measurement errors. I had reconnection and my DSLAM by mistake connected with target SNR margin 15 dB for donwstream and 17 db for upstream and other power output profile. By mistake because as I reconnected again I got again usual 6 dB SNR margin.
Bearer: 0, Upstream rate = 381 Kbps, Downstream rate = 10647 Kbps
Link Power State: L0
Mode: ADSL2+
TPS-TC: ATM Mode
Trellis: U:ON /D:ON
Line Status: No Defect
Training Status: Showtime
Down Up
SNR (dB): 15.0 17.2
Attn(dB): 23.0 14.1
Pwr(dBm): 18.9 12.4
Usually I have following stats.
Bearer: 0, Upstream rate = 1221 Kbps, Downstream rate = 15725 Kbps
Link Power State: L0
Mode: ADSL2+
TPS-TC: ATM Mode
Trellis: U:ON /D:ON
Line Status: No Defect
Training Status: Showtime
Down Up
SNR (dB): 6.2 6.7
Attn(dB): 23.0 15.2
Pwr(dBm): 18.6 12.1
On mistaken connection I've got slightly better Hlog (less undulations) and utterly other QLN. QLN on 15 dB target margin seems to indicate bridge taps.
-
QLN attached.
-
burakkucat: what do you think about it?
-
To me, it still looks like a mystery.
At some time, we have all experienced an anomalous resynchronisation event. You have captured the data from one such event. I am not sure what conclusion can be deduced from that data.
I will agree that the shape of the dips, present in the QLN plot (from the anomalous resynchronisation event), look similar to the dip(s) that would be seen in an Hlog plot from a circuit which has the defect of a bridging tap. However it is wrong to make any conclusion that a bridging tap is present in the circuit from what can be seen in the QLN plot.
A QLN plot would give indications of possible RF ingress, the state of the overall AC balance of the pair, cross-talk from adjacent circuits, etc.
An Hlog plot would give indications of possible physical defects in the circuit, the presence of bridging tap(s), of oxidised high resistance or semi-conducting joints, the overall attenuation loss of the conductors, etc.
-
Today I had second anomalous resynchronization event with SNR margin 15 dB. I had hlog15dB again . Do you see that this Hlog have more regular shape. It implicates that usual hlog is strongly affected by measurement errors related to power output configuration.
-
Ah, now I understand what you were trying to explain! Thank you.
Yes, I agree that the Hlog plot (obtained with a 15 dB SNRM) does look the better of the two. There is not a great deal of difference between the two plots but a small difference can be seen.
I wonder if a "thought experiment" can be performed? Let us assume that you have the ability to remotely configure the DSLAM to adjust the target SNRM that is applied to your circuit. How would we expect to see the Hlog plot vary for each 3 dB increase in target SNRM? I am not sure . . .
Having just typed the above paragraph, I now wonder if the target SNRM is even applied to the circuit before the DSLAM and modem have measured & exchanged the data required to calculate the transfer function. (The transfer function is what is being displayed in an Hlog plot, in logarithmic form.) The question is what occurs first. My feeling is that the transfer function is first determined and then, once the loop-loss of the circuit is know, the power output of the transceivers (at each end of the loop) are adjusted. Finally the target SNRM is applied. But I may be wrong . . .
Perhaps you could examine the ITU-T Recommendations Document G.992.3 [1] and also the Broadband Forum Technical Report TR-138 [2] to see if those documents clarify the process?
[1] https://www.itu.int/rec/T-REC-G.992.3/en
[2] https://www.broadband-forum.org/technical/trlist.php
-
Perhaps you could examine the ITU-T Recommendations Document G.992.3 [1] and also the Broadband Forum Technical Report TR-138 [2] to see if those documents clarify the process?
The HLOGps accuracy requirements shall apply only to those subcarriers with an SNR (as defined
in clause 8.12.3.3) ≥ 12 dB, where the SNR is the SNR value measured during initialization.
The accuracy requirements for the downstream HLOGps (HLOGps_ds) shall apply only to the
following subcarriers (with the corresponding frequency ranges being a part of the passband), and
only if not within the downstream BLACKOUTset (see clause 8.13.2.4):
•
Annexes A and I:
Subcarriers 46 to 208.
•
Annex L:
Subcarriers 46 to 104.
•
Annexes B, J and M: Subcarriers 92 to 208.
It implicates that SNR is measured before Hlog. Furthermore, I have additional evidence, if I have connection with slightly higher power output for tones belonging to HAM band mask I get sligthly less jigged Hlog for this band.
Furterhmore, there is one issue which I have explained unclearly before. Do you see that other circuit Hlog (normal hlog.png) have also sligthly higher attenuation at the place of my lazy roll (highlighted by greem, tone 55)? I see similar pattern at this place (however on normal hlog hardly noticable). Furthermore, these tones are ending tones of upstream Annex M band. I think it is Annex M PSD mask. I think overall shape of my Hlog is caused by measurement errors related to PSD masks or something similar. On my DSLAM there are significantly more errors.
-
The HLOGps accuracy requirements shall apply only to those subcarriers with an SNR (as defined
in clause 8.12.3.3) ≥ 12 dB, where the SNR is the SNR value measured during initialization.
The accuracy requirements for the downstream HLOGps (HLOGps_ds) shall apply only to the
following subcarriers (with the corresponding frequency ranges being a part of the passband), and
only if not within the downstream BLACKOUTset (see clause 8.13.2.4):
•
Annexes A and I:
Subcarriers 46 to 208.
•
Annex L:
Subcarriers 46 to 104.
•
Annexes B, J and M: Subcarriers 92 to 208.
It implicates that SNR is measured before Hlog.
Thank you for finding the details and clarifying the sequence of events.
Furthermore, I have additional evidence, if I have connection with slightly higher power output for tones belonging to HAM band mask I get sligthly less jigged Hlog for this band.
Ah, yes, the "Top Band" mask for the 1.8 - 2.0 MHz frequency band.
Furterhmore, there is one issue which I have explained unclearly before. Do you see that other circuit Hlog (normal hlog.png) have also sligthly higher attenuation at the place of my lazy roll (highlighted by greem, tone 55)? I see similar pattern at this place (however on normal hlog hardly noticable). Furthermore, these tones are ending tones of upstream Annex M band. I think it is Annex M PSD mask. I think overall shape of my Hlog is caused by measurement errors related to PSD masks or something similar. On my DSLAM there are significantly more errors.
I can see your reasoning and, thus, I am prepared to accept your proposal. However, overall, it is quite a small effect.
As we have discussed before, going back many, many months, you have a very stable and well behaved circuit. Far more stable than the average ADSL2+ circuit here in the UK. For example, I could never configure my circuit to operate with a target SNRM of just 1 dB.
-
However, overall, it is quite a small effect.
What do you have on your mind? On normal hlog there is very small effect on the tones 55-100 but on my Hlog there is large effect.
As we have discussed before, going back many, many months, you have a very stable and well behaved circuit. Far more stable than the average ADSL2+ circuit here in the UK. For example, I could never configure my circuit to operate with a target SNRM of just 1 dB.
This is additional argument that my Hlog measurement have a lot of measurement errors. :)
Best regards
konrado5
-
burakkucat: This is Hlog for circuit with only one conductor connected. Do you see that dip on tone 55 is very deep?
Best regards
konrado5
-
Yes, I see it. But a circuit with only one wire of the pair connected is defective. Why do you want me to look at it?
Perhaps I have misunderstood. Are making a collection of Hlog plots for various circuit defects?
-
Yes, I see it. But a circuit with only one wire of the pair connected is defective. Why do you want me to look at it?
I wanted to show that weaker signal received by CPE gives larger measurement error (larger difference between 55 tone and for example 85 tone).
Furthermore, you've missed previous post.
-
burakkucat: do you see that on hlog15db.png on some tones there is higher attenuation than on usualhlog.png? It indicates that measurements errors are both up and down. Do you see that lazy roll has smaller depth on hlog15db.png? It indicates that roll is one of measurement errors. Do you still think my circuit consists many wires of different gauge?
Best regards
konrado5
-
I am sorry but I do not know. :no:
-
I've found someone other Hlog with the same pattern at the place of my lazy roll.
-
There is another one Hlog with similar pattern however this circuit has bridge taps unless I know.
-
In the case of the last Hlog plot, I would certainly suspect the presence of a bridging tap.
With access to the raw data that produced the plot it should be fairly easy to confirm my suspicion.
-
OK, I attach raw data.
-
Thank you. I've downloaded a copy of the data but I am not going to be able to perform the calculations today.
Perhaps you remember a thread from a few months ago when WWWombat and I were performing some calculations to determine the length of a bridging tap? (And I performed the calculation incorrectly . . . :-[ ) We used a table from a JDSU Application Note, titled "Detecting Bridged Tap and Noise Interference in VDSL2 Access Networks using the JDSU SmartClassTM TPS". (I think I have given you a copy of that document, in the past.) It is Table 2 on Page 7 . . . if you would like to try and see if you get a consistent answer for the bridging tap's length.
-
Having downloaded a copy of the raw data (hlogj3w9k.txt) my first action was to re-plot the Hlog graph so that there was more detail on the scale of the X-axis. (A copy of the graph is attached, below.)
Taking note of the approximate location of each minimum (from the graph), I then turned to the tabular data and attempted to specify each minimum by sub-carrier (tone). It was quite clear for the first, not quite so obvious for the second & third (neither were required for the calculations) and somewhat difficult for the fourth. For the latter minimum, I eventually decided to take note of the two extreme sub-carriers and perform the calculations twice.
Here is the table of the data used in the calculations --
53 -25.5000
54 -25.6250 Minimum 1
55 -25.5625
56 -25.5000
176 -34.6875
177 -34.7500
178 -34.8125
179 -34.8125 Minimum 2 (approximate)
180 -34.8125
181 -34.8125
182 -34.6875
298 -42.5625
299 -42.6875
300 -42.6250
301 -42.6250 Minimum 3 (approximate)
302 -42.6875
303 -42.6875
304 -42.5625
409 -50.2500
410 -50.0000
411 -50.3750 Minimum 4 ?
412 -50.2500
413 -50.3750
414 -50.4375
415 -50.4375 Minimum 4 ?
416 -50.3750
How can we be sure that we are looking at the effect of a bridging tap on that circuit's transfer function?
(A) Look at the shape of the lowest frequency minimum. Notice how it is similar to an exponential growth (or decay) curve that has been rotated in the plane and an image of itself reflected along the vertical axis. In this particular case it is not too clear but in other examples it is a distinct diagnostic "give away".
(B) Consider the delta between each successive minimum. Is it consistent?
The delta between minimum 1 & minimum 2 is (approximately) 125.
The delta between minimum 2 & minimum 3 is (approximately) 122.
The delta between minimum 3 & minimum 4 is (approximately) either 110 or 114.
The three deltas are of similar order of magnitude hence, coupled with the shape of minimum 1, we can conclude that we are looking the effect of a bridging tap.
The calculations will be performed using minima 1 and 4, noting that there are two intervening minima.
Now working with the relevant data from Table 2 of the JDSU Application Note ("Detecting Bridged Tap and Noise Interference in VDSL2 Access Networks using the JDSU SmartClassTM TPS"):
Using the minima at 54 & 411. Delta is 357.
According to Table 2, for two sub-minima . . .
350 198.8
400 173.9
Interpolation for 357 gives either . . .
350 / 357 = N / 198.8
So N = 350 x 198.8 / 357 = 194.90
Or . . .
357 / 400 = 173.9 / N
So N = 173.9 x 400 / 357 = 194.85
Using the minima at 54 & 415. Delta is 361.
According to Table 2, once again, for two sub-minima . . .
350 198.8
400 173.9
Interpolation for 361 gives either . . .
350 / 361 = N / 198.8
So N = 350 x 198.8 / 361 = 192.74
Or . . .
361 / 400 = 173.9 / N
So N = 173.9 x 400 / 361 = 192.69
The arithmetic mean of the four above values is --
(194.90 + 194.85 + 192.74 + 192.69) / 4 = 193.80
The delta of 194.90 & 193.80 is 1.10
The delta of 193.80 & 192.69 is 1.11
Accepting the larger of the two deltas, 1.1, (to one decimal place) as the error, I propose that the bridging tap responsible for the effect shown in the Hlog plot has a length of 193.8 +/- 1.1 metres.
We cannot deduce the location of the start of the bridging tap (relative to a known location in the circuit) by considering the Hlog plot. Other techniques (Time Domain Reflectometry, for example) would have to be deployed to obtain the required datum point.
-
As I know this circuit has three bridge taps.
What do you say about difference between my usual hlog and hlog with target SNR margin 15 dB (and higher power output)? Do you see that better graph (hlog15db) has also higher attenuation at some frequiences? It implicates there are also measurements errors causing lowering attenuation value.
-
I also attach raw data.
-
As I know this circuit has three bridge taps.
Well that is interesting but in my above analysis and calculation I saw no obvious sign of more than one tap. It just goes to show that we must consider all information about a circuit. If, for example, I saw that Hlog plot the first thing I would probably do is to make TDR measurements from both ends of the circuit. Look at the attached PDF document. In the example trace, the length of the tap is given by the difference between the minimum and the maximum whilst the location of the tap is given by the point where the plot starts to drop down to the minimum.
What do you say about difference between my usual hlog and hlog with target SNR margin 15 dB (and higher power output)? Do you see that better graph (hlog15db) has also higher attenuation at some frequiences? It implicates there are also measurements errors causing lowering attenuation value.
Please be aware that the transfer function is a negative value, so that when we examine a Hlog plot the zero point of the Y-axis is at the top. Thus greater attenuation is indicated by the curve being further down the Y-axis scale. Greater attenuation is lower down; lesser attenuation is higher up.
You have shown me similar plots before. I really do not know why your circuit behaves that way. It would be nice if you could plot the normal in red and the abnormal in green on the same graph. That would make it so much easier to compare them both.
I also attach raw data.
I have downloaded the data for both Hlog plots. At the moment I'm not sure what I can do with them.
As I have spent hours performing the calculations and then typing up what I had found, earlier this evening, I just do not have any more time to spare.
There are other things in my life that require my time . . . so I am stopping looking at your other Hlog plots.
[Edited to correct a minor error in the description of the JDSU TDR plot example.]
-
It would be nice if you could plot the normal in red and the abnormal in green on the same graph. That would make it so much easier to compare them both.
It would be nice if the script could plot two graphs in this way.
-
konrado
Please take note that b*cat has other things on right now and has already spent an enormous amount of time trying to help answer your questions.
I'm only butting in because I know that sometimes you may not notice when people are under pressure and you can miss a hint when someone is trying to be polite.
If he can help, then he will, but just right now he needs to take a breather to do some other things in real life.
Thank you for understanding. :)
-
kitz: now I try to proof my theory. I want to convince burakkucat.
-
Konrado, please take note of what Kitz said.
-
I have taken another quick look at this and I think I realise what I have been missing. I believe that you have concluded that the equipment is not reporting values in a linear fashion across the entire range. When there is a change in output power there is not an equal change in the representation of the transfer function, visualised by an Hlog plot.
As for the last two data sets that you have provided (usualhlog.txt and hlog15db.txt), I have not plotted them on one graph in different colours (I don't know how to do so without spending time reading the gnuplot documentation) but have, instead, just plotted the delta of the two data sets. (Attached below.)
By considering the two basic Hlog plots (also attached below), we must ignore the two outer limits -- where the sub-carriers (tones) are less than & equal to 30 and also those that are greater than & equal to 400. In other words, we just look at --
400 > sub-carrier > 30
I have not attempted to explain the result. However we can see that the delta is inversely proportional to the sub-carrier (i.e. frequency) and the rate of decay (with frequency) appears to follow an exponential curve. As the frequency rises, fine structure becomes apparent in the actual curve of the delta.
-
I have taken another quick look at this and I think I realise what I have been missing. I believe that you have concluded that the equipment is not reporting values in a linear fashion across the entire range. When there is a change in output power there is not an equal change in the representation of the transfer function, visualised by an Hlog plot.
1. I assume that unusual-hlog is more accurate representation of real Hlog than usualhlog.
2. I've noticed that on unsual Hlog there are lower value on some tones but there are also higher value on some tones.
3. I conclude that hlog measurement errors are both: up and down.
Did you notice it?
Similarly, shortening the line causes higher attenuation on some tones.
http://forum.kitz.co.uk/index.php?topic=14668.0
-
Having seen many variations on the overall shape of Hlog plots for the three versions of ADSL (G.992.1, G.992.3 and G.992.5) and for VDSL2 (G.993.2), I would have to say that out of the two examples (usual-hlog and unusual-hlog) the one I have labelled unusual-hlog appears to better fit the expected shape.
By looking at the delta plot, we can see that there is most definitely measurement errors in both directions of data transfer. As the input to the plotting utility, the usual- data was the first and the unusual- data was the second. Hence when the value of the unusual- is larger (more negative) that the usual- then the delta will be positive and when the value of the unusual- is smaller (more positive) than the usual- then the delta will be negative. Examining the delta plot allows us to clearly see the differences without needing to state which data set should be accepted as a correct representation.
My statements, above, align with the three conclusions that you have made. :)
As for your other observation that there is higher attenuation on some tones (sub-carriers) after the physical line length was shortened is somewhat more difficult to rationalise. However I offer the suggestion that the length by which the line was shortened was a very small percentage of the overall line length, hence any such comparative measurements would be of the order of the overall experimental error. The real test must surely be a comparison of the overall (the sigma, the sum of all sub-carriers) attenuation of the line at its original length with that overall attenuation when it had been shortened by a small percentage of its length. :-\
-
I'm glad we came to agreement regarding to these two Hlogs. I'm still not sure if I explained clearly enough that all undulations on my Hlog are caused by measurement errors. I send again other circuit Hlog. Do you see the same pattern at tones 50-100 on very low errors hlog.png? On the remaining tones my errors are not noticable on low errors hlog because there is most disturbing factor on tones 50-100(PSD mask or something similar). Do you still believe that my circuit consists of different gauges?
EDIT: I now see on low errors hlog slightly steeper slope on tones 100-200. It reflects my roll on these tones.