1 edit | Yet Another Modem Statistics tool for the ST516.
I recently started a thread to discuss the bizarre behaviour of my ST516 modem. As the discussion progressed, I was convinced that I would need to do some long-term analysis of my modem's connection metrics, especially the error rates.
»ST516 - Strange discovery
The popular DMT tool has a diagnostic option to output a comma-separated-variable (CSV) file containing all the modem metrics. This can be fed into several programs such as MS-Excel or Matlab to create summary reports and graphs. The only problem is that the DMT diagnostic feature is only supported on some versions of the ST516 firmware. Once I upgraded to newer firmware loads, I had no way to capture the detailed results to evaluate the new firmware's performance. Out of necessity I created my own simple program to just grab the data from the ST516 and store it in a .CSV format similar to that created by DMT. After that, I got a little carried away and added some real-time graphing capabilities to the program. I call the program DSL_StatScope. I was tempted to call it 'YAMS' but then you would think me an old UNIX hack 
It has been suggested to me that other Teksavvy customers with highly attenuated, noise infested DSL connections might make good use of this program to measure the effect of new firmware, or upgraded wiring, or whatever. So here it is; just unzip the file into a directory and you'll find a readme.txt file. It's pretty simple to use. |
|
 CanerisErikCanerisPremium,VIP
join:2007-10-03
Toronto, ON
kudos:2 | Great work! I look forward to playing with it when we get more 516s next week. |
|
| reply to tertech
Hi TerTech,
Excellent! I've put my ST516v6 (FW v6.2.29.2)
on-line just a few minutes ago and it happens
i needed an alternative to the "Diagnosis" item
of 'DMT' (which is a greyed out option if using
this FirmWare). I wished to compare my MoDems
for a very long time, now i'm sure Santa exists!
My plan is to start an evaluation by collecting a
high-resolution (15 sec.) long-term sample under
"Linear" mode and with the "Follow current time"
feature enabled (i presume the later option has
to do with time-slice round-up synchronisation).
I can't wait to see the curve averaging function
in action since it's specific to 'DSL_StatScope'
and no other 3rd-party tool did that until now
but i'm patient and these tests take time. You
can bet i'll be back to share my observations, it
brings us to mid-December or so but i'll be there!
 |
|
|
Addendum:
I'm so impressed looking at this very unique utility
showing evidence of the "Cooking Hour Syndrome"
in real time i've now made it a channel on my TV!
 |
|
| reply to Bicephale
said by Bicephale:Hi TerTech, Excellent! I've put my ST516v6 (FW v6.2.29.2) on-line just a few minutes ago and it happens i needed an alternative to the "Diagnosis" item of 'DMT' (which is a greyed out option if using this FirmWare). I wished to compare my MoDems for a very long time, now i'm sure Santa exists! My plan is to start an evaluation by collecting a high-resolution (15 sec.) long-term sample under "Linear" mode and with the "Follow current time" feature enabled (i presume the later option has to do with time-slice round-up synchronisation). I can't wait to see the curve averaging function in action since it's specific to 'DSL_StatScope' and no other 3rd-party tool did that until nowbut i'm patient and these tests take time. You can bet i'll be back to share my observations, it brings us to mid-December or so but i'll be there! Ho Ho Ho Merry Christmas Well, I hope you can put this program to good use. I did a lot of testing myself, but I am interested to hear any problems you may encounter with your long term captures. It's only version 1.0, so I don't expect it to be perfect.
Just to clarify... the "follow current time" option simply automatically scrolls the graphical display so that the most recent capture is always visible, that's all.
Enjoy |
|
|
|
1 edit | 
DSL_StatScope v1.0 - CRC vs HEC & FEC |
My test only starts at midnight so i can still
play around with 'DSL_StatScope v1.0' a bit,
after that i'll leave it alone doing it's job.
Here are my preliminary observations, something
may very well have slept through... Well, you've
brought the SpeedTouch owners a unique piece
of work and i can only think of cosmetic aspects.
•The CRC dots being critical they should have
priority over the HEC and FEC dots (in this
order) which are presently masking it. See my
animated .GIF from above. This was captured
using the [Load] option (playback) after the
log file was copied to my everyday machine.
•When trying to play back a tiny log file with
only a few time-slice records nothing happened
and there was no error message to explain why
but maybe i was supposed to restart 'DSS' 1st;
i even thought it was about 8.3 DOS filenames.
•You can make the window's width two characters
less using "DS" (DownStream) instead of "D/L"
(DownLoad) and "US" (UpStream) instead of "U/L"
(UpLoad).
•This should have been called v1.25 or something:
some persons won't even try it when it's a first
release. I say the loss is theirs, though!
•Perhaps i didn't find it but there should be a hidden
"easter egg" with Santa's face in it somewhere!...
In any case, thank you for this precious opportunity!
Note:
I was hungry late last night so i started cooking
something to eat (the CRC burst did confirm it).
Now, imagine if i conducted such a test with help
from the powerful averaging feature for a week!
|
|
| reply to tertech
Oh, i almost forgot...
One question about reading aberrations: 'DMT'
in "Diagnosis" mode often generates very large
numbers during transitions which would totally
offset a curve's range. I've been working with
'GNUmeric' SpreadSheets lately and my solution
to that problem was to "clip" those events to a
negligible negative fraction if negative or what
appears to be the curve's normal peek, my goal
being to conserve as much details as possible.
I'm wondering how you dealt with that relatively
to the log files (the display itself already has two
customisable fields for the upper/lower limits).
|
|
| reply to Bicephale
said by Bicephale:Hi TerTech, •The CRC dots being critical they should have priority over the HEC and FEC dots (in this order) which are presently masking it. See my animated .GIF from above. This was captured using the [ Load] option (playback) after the log file was copied to my everyday machine. O.K makes sense. I usually just enable one error stat at a time because it is too crowded to see everything.
•When trying to play back a tiny log file with
only a few time-slice records nothing happened
and there was no error message to explain why
but maybe i was supposed to restart 'DSS' 1st;
i even thought it was about 8.3 DOS filenames.
The intent was that either you were doing a Telnet capture or playing back a file. I prevent a file being loaded when a Telnet session is active; I thought I made it allow a file load after a Telnet session was stopped. It looks like it doesn't allow a file load if a Telnet session had been run, even if it was stopped. You will have to restart the program to load a file if you had a Telnet session.
•You can make the window's width two characters
less using "DS" (DownStream) instead of "D/L"
(DownLoad) and "US" (UpStream) instead of "U/L"
(UpLoad).filenames.
It's UpLink and DownLink ( I'm a satellite guy ). You're right though about giving more room to the U/L SNRM and D/L SNRM labels.
•This should have been called v1.25 or something:
some persons won't even try it when it's a first
release. I say the loss is theirs, though!
It's the first release. If I counted every version that I didn't release it would be like v25. The first number is a major release and will increment with new features or some incompatibility with previous releases. The second number is a minor release with just cosmetic changes that is still compatible with previous versions of the same major release number. For example if the log file format changed that would need a new major release number.
•Perhaps i didn't find it but there should be a hidden
"easter egg" with Santa's face in it somewhere!...
Keep looking. |
|
| reply to Bicephale
said by Bicephale:Oh, i almost forgot... One question about reading aberrations: 'DMT' in "Diagnosis" mode often generates very large numbers during transitions which would totally offset a curve's range. I've been working with 'GNUmeric' SpreadSheets lately and my solution to that problem was to "clip" those events to a negligible negative fraction if negative or what appears to be the curve's normal peek, my goal being to conserve as much details as possible. I'm wondering how you dealt with that relatively to the log files (the display itself already has two customisable fields for the upper/lower limits). I'm not sure what situation you're referring to. IM me with details please. |
|
| Hi TerTech,
I'll make contact shortly. In the meantime, this is
a compatibility chart i built after wondering if the
log file format may allow me to playback logs from
other brands/models using 'DMT' Diagnosis mode:
.-----------------+-------------------------+-----------------------+------------------------.
| 'DSS v1.0' | 'DMT v7.35' | 'DMT v9.01' | 'DMT v2.20' |
| Thomson | Thomson | Siemens | GNet |
| ST5x6 | ST5x6 | SS4200 | BB0060B/GBB2060-Xi |
+-----------------+-------------------------+-----------------------+------------------------+
| | # DMT Diagnosis... | # DMT Diagnosis... | # DMT Diagnosis... |
| | # creation date... | # creation date... | # creation date... |
| | Tones=255 Separator=, | Tones=255 Separator=, | Tones=255 Separator=, |
+-----------------+-------------------------+-----------------------+------------------------+
| Time,Up Rate... | Time,USrate,DSrate... | Time,USrate,DSrate... | Time,USrate,DSrate... |
| Up Rate | USrate | USrate | USrate |
| Up SNRM | USsnr | | USsnr |
| Dn Rate | DSrate | DSrate | DSrate |
| Dn SNRM | DSsnr | DSsnr | DSsnr |
| Rx FEC | rxFEC | | rxFEC |
| Rx CRC | rxCRC | rxCRC | rxCRC |
| Rx HEC | rxHEC | rxHEC | |
| Tx FEC | txFEC | | txFEC |
| Tx CRC | txCRC | txCRC | txCRC |
| Tx HEC | txHEC | txHEC | |
+-----------------+-------------------------+-----------------------+------------------------+
| Rx FEC/s | | | |
| Rx CRC/s | | | |
| Rx HEC/s | | | |
| Tx FEC/s | | | |
| Tx CRC/s | | | |
| Tx HEC/s | | | |
| Index | | | |
| Con | | | |
`-----------------+-------------------------+-----------------------+------------------------'
I haven't tried anything yet and i'm not sure what
to do with the last section (N.B.: i rearanged the
sequence to render this table easier to read)... If
'DSL_StatScope v1.0' accepts log files missing the
"/s" data then long-term records averaging may be
available even for devices not directly supported!
The 'DMT' log file format allows comments and some
definitions in the header, maybe it would be a good
idea in an eventual future to exploit these features.
|
|
| reply to tertech
DMT Diagnosis Aberrations |
About aberrations, finally it's simpler for
me just to post an illustration right here.
This graphic was generated using 'GNUmeric'
to process a log file generated by 'DMT' run
in "Diagnosis" mode. The negative values are
resulting from a disconnection (zero SNR), it
is meaningless information but those numbers
are so large everything else will vanish into
the baseline if it's not corrected manually...
 |
|
2 edits | reply to tertech
SNR, rates, errors have same vertical limit |
But I think you should make three separate graphs or at least give the option to view SNR, rates and errors in separate graphs. Having one graph with different vertical limits for each variable is unconventional and makes it harder to read.
If not, same type data like (rates, SNR, errors) should have same vertical limits by default. Might want to consider new default colours that are easier to follow too. Would also be nice to change the background colour.
Also, what does increment do?
--
Argue opinions using facts. Not facts using opinions. |
|
| reply to Bicephale
Oups! TerTech, i just realized these negative
values don't exist yet in the log files: they're
the result of calculations done later instead...
Sorry to bother!
|
|
| reply to RobinK
said by RobinK:Nice tool, tertech. I like it and the graphs it can produce. But I think you should make three separate graphs or at least give the option to view SNR, rates and errors in separate graphs. Having one graph with different vertical limits for each variable is unconventional and makes it harder to read. If not, same type data like (rates, SNR, errors) should have same vertical limits by default. Might want to consider new default colours that are easier to follow too. Would also be nice to change the background colour. Also, what does increment do? Thanks for your input. You are one of the fortunate ones for whom error rates are somewhat constant. I created this utility because my stats were all over the place. In one single sample I could have 10's of CRC errors and 100's of FEC errors. By making separate scaling factors, I can spread the values vertically on the graph rather than just have an off-scale straight line across the top or bottom.
It happens frequently on weekday mornings at ~07:00 that I get an FEC storm. This results in a lowering of SNRM. To see them on the same graph makes the relationship of the one to the other more obvious. Generally I turn off all the uplink values, and only enable the downlink SNRM and error stats. When I see a situation that caused the modem to disconnect & re-sync, I back up in time and view each error type with a scaling that makes the values spread vertically.
The 'increment' selection box chooses the value to change the upper and lower limit values when the up/down arrow beside each value is pressed. In my case for spreading the error values across the vertical it's easier to just press the up-arrow and have the value increase by 10 or 100 and see the graph react to each incremental change.
The colours are a bit of a nuisance; I can't make up my mind what I like the best; it's subjective to each individual's preference. One thing I'm thinking of doing is to save the user's graphical preferences in an 'ini' file or the system registry. At least you wouldn't have to make the same adjustments every time you start the program. |
|
| reply to tertech
Hi TerTech,
This wouldn't have been possible without your
help, i wish to thank you warmly and publicly!
%20.GIF)
From the ground up!, Bicephale, 2008-Dec-13
|
|
 HiVoltPremium
join:2000-12-28
Toronto, ON
kudos:12 | reply to tertech
Cool utility, too bad its only for the ST... Mine is dead, using a SpeedStream at the moment...
--
GO LEAFS GO! |
|
| Hi HiVolt,
Well, it's too soon for me to draw conclusions
but maybe you're better off with the Siemens:
From the ground up!, Bicephale, 2008-Dec-13
|
|
| reply to HiVolt
Euh...
I mean, could this just be a simple coïncidence?
%20.GIF)
From the ground up!, Bicephale, 2008-12-7
|
|
HiVoltPremium
join:2000-12-28
Toronto, ON
kudos:12 | reply to Bicephale
Its no big deal anyway, my line is solid. I dont really need to monitor line stats all the time.
--
GO LEAFS GO! |
|
1 edit | Great HiVolt! Otherwise i would have refered you to this post:
»Re: Yet Another Modem Statistics tool
I'd bet you do have 'DMT v9.01' working in "Diagnosis" mode... |
|
