Kitz Forum
Broadband Related => Router Monitoring Software => Topic started by: Bald_Eagle1 on July 19, 2014, 05:24:35 PM
-
Finally, a comment about the size of the y-axis scale for the QLN and HLog plots. I have never seen a QLN plot where the curve is below -150 dBm/Hz. So perhaps using the range -80 to -160 dBm/Hz would make better use of the space? Similarly I have never seen a HLog plot where the curve is below -100 dB. So, again, using a range of 0 to -120 dB would make better use of the space? Perhaps Bald_Eagle1, having viewed many QLN and HLog plots from different circuits over the last few years, has an opinion upon the y-axis scales?
Here's a snippet of recently obtained QLN data from a G.DMT connection:-
Tone number QLN
0 -88.0000
1 -146.0000
2 -147.0000
3 -148.0000
4 -150.0000
5 -149.0000
6 -150.0000
7 -150.5000
8 -150.5000
9 -150.0000
10 -150.0000
11 -150.5000
12 -149.0000
13 -150.0000
14 -150.0000
15 -148.0000
16 -147.0000
17 -146.0000
18 -144.0000
So, with a little spare, perhaps -160dBm/Hz should be the value for bottom of the Y axis.
I have attached 3 versions of the same data.
One uses 0 to -160 one uses -80 to -160 & the other uses -110 to -160
My preference is 0 to -160 as some users might panic on seeing larger spikes as shown in the -80 to -160 & -110 to -160 graphs & it might prompt them into asking many (possibly) unnecessary questions? ??? ;)
Whichever values are used, if wishing to compare stats from different connections, in my opinion it is wise to have fixed rather than auto-scaled axes.
-
Making use of DSLstats, this is what I see for QLN and HLog plots --
-
Similarly I have never seen a HLog plot where the curve is below -100 dB. So, again, using a range of 0 to -120 dB would make better use of the space? Perhaps Bald_Eagle1, having viewed many QLN and HLog plots from different circuits over the last few years, has an opinion upon the y-axis scales?
Neither had I until I recently obtained the stats from the G.DMT connection mentioned above.
It prompted me to adjust the Hlog axis values for ADSL connections.
I have attached an example I replotted from some ADSL2+ testing data you sent me from your own connection.
-
Ah, I vaguely remember!
The violent swings present in the "mush" from tone 330 upwards should really be discounted. I put it down to the CPE transceiver attempting to "find something that isn't there".
Perhaps we should make a study of Kitizens' QLN and HLog plots in the region of tones 0 to 10. That might indicate what could be regarded as a sensible "lower limit"? :-\
-
I'm certainly interested in optimising the scales of the QLN and HLog graphs. It seems to me that the best thing for all concerned would be that HG612-stats and DSLstats use the same scales, and for the same reason I agree that a fixed scale is better than autoscaling.
I think my preference for the QLN graph would be a range of -60 to -160, because it would seem to cover all the likely values and doesn't leave a lot of empty space (which starting at zero would).
For HLog, 0 to -100 would cover most cases, but there would be the few that go below -100. Perhaps in this case a degree of autoscaling might be appropriate, rather than leaving a lot of empty space on most connections.
For HLog, 0 to -120 would seem to cover all likely cases, without too much empty space.
-
I've split this into a separate topic, as it's relevant to any monitoring program.
-
I've split this into a separate topic, as it's relevant to any monitoring program.
A good move! :thumbs:
-
For reference, here are my QLN and HLog graphs using the scaling I suggested above.
-
I have never seen a QLN plot where the curve is below -150 dBm/Hz. So perhaps using the range -80 to -160 dBm/Hz would make better use of the space? Similarly I have never seen a HLog plot where the curve is below -100 dB.
My line does.
My QLN already goes as low as -160 dBm/Hz in several regions so perhaps just a bit more is needed to say -170 please to make the graph look like its not falling off the bottom of the page, and also allow for the fact that there are many lines shorter than mine.
858 -124.8750
859 -127.3750
860 -160.0000
861 -160.0000
862 -160.0000
863 -160.0000
864 -160.0000
865 -160.0000
866 -160.0000
867 -160.0000
868 -160.0000
869 -160.0000
870 -160.0000
871 -160.0000
872 -160.0000
873 -160.0000
874 -160.0000
875 -160.0000
876 -160.0000
877 -160.0000
878 -160.0000
879 -160.0000
880 -160.0000
881 -160.0000
882 -160.0000
883 -160.0000
884 -160.0000
885 -160.0000
886 -160.0000
887 -160.0000
888 -131.5000
889 -131.5000
890 -131.5000
891 -131.5000
1191 -132.0000
1192 -160.0000
1193 -160.0000
1194 -160.0000
1195 -160.0000
1196 -160.0000
1197 -160.0000
1198 -160.0000
1199 -160.0000
1200 -160.0000
1201 -160.0000
1202 -160.0000
1203 -160.0000
1204 -160.0000
1205 -160.0000
1206 -160.0000
1207 -160.0000
1208 -160.0000
1209 -160.0000
1210 -160.0000
1211 -160.0000
1212 -160.0000
1213 -160.0000
1214 -160.0000
1215 -160.0000
1216 -128.9375
My Hlog in the corresponding frequencies reaches -96
850 -17.1875
851 -17.1875
852 -17.1875
853 -17.1875
854 -17.1875
855 -17.1875
856 -17.1875
857 -17.1875
858 -96.0000
859 -96.0000
860 -96.0000
861 -96.0000
862 -96.0000
863 -96.0000
864 -96.0000
865 -96.0000
866 -96.0000
867 -96.0000
868 -96.0000
869 -96.0000
870 -96.0000
871 -96.0000
872 -96.0000
873 -96.0000
874 -96.0000
875 -96.0000
876 -96.0000
877 -96.0000
878 -96.0000
879 -96.0000
880 -96.0000
881 -96.0000
882 -96.2500
883 -96.2500
884 -96.2500
885 -96.2500
1192 -19.5625
1193 -19.5625
1194 -96.0000
1195 -96.0000
1196 -96.0000
1197 -96.0000
1198 -96.0000
1199 -96.0000
1200 -96.0000
1201 -96.0000
1202 -96.0000
1203 -96.0000
1204 -96.0000
1205 -96.0000
1206 -96.0000
1207 -96.0000
1208 -96.0000
1209 -96.0000
1210 -96.0000
1211 -96.0000
1212 -96.0000
1213 -96.0000
1214 -96.0000
1215 -96.0000
1216 -96.0000
1217 -96.0000
1218 -96.2500
1219 -96.2500
1220 -96.2500
1221 -96.2500
1222 -96.2500
1223 -96.2500
1224 -21.6875
1225 -21.6875
1226 -21.6875
1227 -21.6875
1228 -21.6875
1229 -21.6875
1230 -21.6875
1231 -21.6875
-
I have never seen a QLN plot where the curve is below -150 dBm/Hz. So perhaps using the range -80 to -160 dBm/Hz would make better use of the space? Similarly I have never seen a HLog plot where the curve is below -100 dB.
My line does.
My QLN already goes as low as -160 dBm/Hz in several regions so perhaps just a bit more is needed to say -170 please to make the graph look like its not falling off the bottom of the page, and also allow for the fact that there are many lines shorter than mine.
I would be interested to know, for tones 0 - 10, if you see a horizontal line and the level it indicates. (If not at -160 dBm/Hz, then my postulation will be false.)
-
Taking note of Kitz' comments, I propose the following:
QLN: -60 to -170 dBm/Hz
HLog: 0 to -120 dB (as before)
-
I suspect that kitz is reporting values in unused bands on vdsl. They do go to the values she notes but it probably would not matter is the graphs did not get that far. See examples below.
-
I prefer to show the data from usable & unusable tones.
If nothing else, it partly provides an explanation as to why those tones can't be used (Attenuation too high etc.)
For VDSL2 connections, my Hlog graphs currently display everything between 0 & -95.999 (I assumed -96.000 meant 'invalid' or N/A data).
For G.DMT & ADSL2, the parameters currently used are 0 & -140.999
For ADSL2+, 0 & -120.000
For QLN, the parameters are currently 0 & -159.999 for all modes.
As per the attached Hlog graph, attenuation in the D3 band was too poor for my connection to use any of it, but at least I can see how poor it is.
It has deteriorated since then & I have nothing at all better than -96.000 in the U2 & D3 bands (see the other graph).
I intentionally chose not to show a flat line for 'invalid' or N/A data, thinking it 'could' be misleading.
EDIT:
Just for 'fun', I replotted the 2nd Hlog graph (attached) with parameters 0 to -100.
It has alerted me to the fact that since the October firmware update, -96.250 now seems to be the limit rather than -96.000 for VDSL2 connections.
Does anyone have a preference for what is actually plotted & if so, please explain why?
-
b*cat, having been the one to instigate this round of discussions prefers to stand back and let others express opinions.
It would be nice to have both DSLstats and modem_stats using the same y-axis range for QLN and HLog plots.
My experience is solely with DSLstats and an ADSL2 mode service -- and there is a lot of wasted vertical space in the HLog plot.
-
Thank you, yes I'd be happy with -170.
Yes bcat you are correct -160 is my base line for the starting tones and also any unused tones such as stop bands. I'm not at the pc atm to check but iirc it was tones 1-7 that were at -160 too.
Like bald eagle, if possible I would like to be able to see data both from usable and unusable tones, otherwise if the graph stopped at -160' it would just look like the bottom of the graph had been chopped off, aside from the fact I'd never know if it did get any lower.
Finally it would be confirmation of my base noise floor measurement, which when it comes to QLN can be a very important figure in its own right.
@BE I'm not using the hg612, and I can't double check right now, but I'm 99% certain that I also see -96.250
-
Billion 7800DXL on TalkTalk ADSL2+:
adsl info --Hlog
adsl: ADSL driver and PHY status
Status: Showtime
Last Retrain Reason: 8000
Last initialization procedure status: 0
Max: Upstream rate = 1024 Kbps, Downstream rate = 8408 Kbps
Bearer: 0, Upstream rate = 1020 Kbps, Downstream rate = 7393 Kbps
Tone number Hlog
0 -96.2500
1 -96.2500
2 -96.2500
3 -96.2500
4 -96.2500
5 -96.2500
6 -96.2500
7 -21.0000
8 -20.1875
9 -20.0000
10 -20.5625
Values for tones beyond the used range bottom out at -96.3125
-
Taking note of Kitz' comments, I propose the following:
QLN: -60 to -170 dBm/Hz
HLog: 0 to -120 dB (as before)
Taking recently obtained data from a G.DMT connection into account, perhaps we should consider these parameters for QLN & Hlog:-
QLN: -60 to -170 dBm/Hz (all modes)
HLog: 0 to -140 dB (ADSL connections) 0 to -100 (VDSL2 connections)
While we are at it, looking at the attached montage, maybe we should consider 70 to -20 for SNR (all modes).
-
Seems fine by me :)
-
Yes, I can agree to that. From the point of view of DSLstats, I can certainly change the scaling of the SNR per tone graph as a separate graph, but the combined bitloading/SNR per tone graph is a bit more problematical and I don't think I'll be changing it at this stage.
-
but the combined bitloading/SNR per tone graph is a bit more problematical and I don't think I'll be changing it at this stage.
Ah, yes.
I had a very quick look at that graph last night & also decided to 'park' it for the time being :)
-
We seem to be reaching an agreement. Does anyone else have any comments?
-
We seem to be reaching an agreement. Does anyone else have any comments?
I'm quite happy to leave things to those who know their own code. :)
Once changes have been made, I'll be ready to test and examine the result(s).