Re: ST516 - Strange discovery in TekSavvy

Re: ST516 - Strange discovery

Sun, 11 Jan 2009 04:16:38 EDT

Bicephale : That's most probably not a SpeedTouch or 'DMT'
issue, i recall posting 'DMT' captures when i
1st got my hands on that FirmWare release...

Re: ST516 - Strange discovery

Sun, 11 Jan 2009 01:50:15 EDT

soniko : DELETED

Re: ST516 - Strange discovery

Sun, 11 Jan 2009 00:40:06 EDT

Bicephale : Well, 'user.tpl' most probably uses different
defaults which end up in your final 'user.ini'.

Re: ST516 - Strange discovery

Sat, 10 Jan 2009 22:31:50 EDT

maxotin : Anybody knows how to set up VPI/VCI of 0.35 for the 7.4.4.7 firmware? It only has standard, k-net and westend options and though I got syncronized, I won't able to get internet connection probably because of this :-(

Re: ST516 - Strange discovery

Wed, 26 Nov 2008 02:38:53 EDT

Bicephale : Hi,

I've found an answer to your initial question
about "Remote Tx Power", finally:

I see you also put "transmit power" to
maximum, but its only usefull with ADSL2+.

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Fri, 31 Oct 2008 15:52:07 EDT

Bicephale : Hi again TerTech,

Using an ST516 with FirmWare v6.2.29.2-GE (ZZQ0GE6.2T2)
i've been unable to access 'DMT's "Diagnosis" feature: it was
greyed-out. I found that 'STMT', 'OrbMT' and 'RouterStats'/
'RouterStats-Lite' were all useless relatively to error rates...

:huh:

...which shall make 'DSL Stat Scope' even more necessary!!!

:)

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Fri, 31 Oct 2008 14:19:29 EDT

Bicephale : Hi Ian,

I don't want to hijack TerTech's thread but i
must admit i was hoping you'd happen to be a
Green Acres phone-circuit topology candidate:

:p

So, you're certain Bell's drop-line gets straight
to your MoDem (thru centralized filtering) and
there's just no chance to collect noise from an
induction loop nearby?... I'm puzzled, it would
mean the twisted pairs are not as fine as we'd
like them to be!!! No street lights, whatever?

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Fri, 31 Oct 2008 13:58:25 EDT

Bicephale : Hi TerTech,

There must be something wrong with your setup-
file so erase it! I know it worked once, obviously:

From the ground up!, Bicephale, 2008-Jul-28

Oups! Oh, the "Diagnosis" mode, you mean... Well, i do
have that FirmWare installed on my ST516 so i'll check
and let you know what happens! I wish it does work...

:p

About 'DSL Stat Scope', it's safe to discard negative
values (i'd favour to flag them with -1 or so to keep
track) but i'm not sure why the previous counter value
must be cleared as well, unless you don't want to see
error rate bursts... Maybe this should call for one more
option: "Clip MoDem-Reset-time Error-Rate bursts"!...

;-)

A resolution of five seconds would be amazing already,
just make sure the SpeedTouch 5x6 remains responsive
if you open a 2nd TelNet CLI session via 'DMT', 'OrbMT'
or 'STMT': two simultaneous sessions are required when
the user tweaks, while one of the utilties is recording.

Having well "aligned" log-files (compatible time-slices)
would be good enough, you would find yourself a whole
new niche if 'DSL Stats Scope' were able to average a
set of records all by its own! Now, if the program could
launch under Linux 'WinE' (or JAVA?) from a "LiveCD" i'd
find it even more useful because anyone with a legacy
PC would get the means to collect detailed information
without much effort: just "Install & Forget", i might say.

:D

This afternoon announces really motivating times, even
better if you can deal with Thomson's HTTPS server...

Have a nice day!

:)

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Fri, 31 Oct 2008 13:02:31 EDT

tertech :

said by Bicephale :

Hi TerTech,

I bet you have found a way to use 'DMT' with FirmWare
v6.2.29.2 installed by now so i simply won't insist... It's
a real pleasure to read about your new diagnostic tool,
please allow me to express a suggestion while it's still in
the creation phase! You may find suitable to add a time
offset option so that long-term records can be averaged
easily, this might even be the core of some main feature:
the ability to reveal daily patterns through averaging...

:D

No, actually I haven't found how to make DMT diagnosis mode work :( I checked the 'allow untested firmware version' box, but the diagnosis option is still unavailable. That's what started me on DSL_Stat_Scope, so there's no turning back now. ;)

Thank you very much for the ideas for my new utility. Yes, I did run into negative numbers when calculating the difference between samples as you warned me. I just treat the previous count as if it were 0.

I capture the raw data from the modem stats (Telnet CLI) at a selectable rate from as fast as 1 per second. I hold at least 86400 samples so that an entire day can be captured at this fast rate. If the selected sample time is slowed to 1 per 10 seconds it could store a whole week+. Presently the 'scope' screen just shows the raw data for the selected parameters scaled to any min-max range. I think it's a great idea to build in some analysis tools that can be used to display such things as time-of-day or week-day averaging.

I'll also look into the HTML interface. CLI just was the fastest option for me to code at the time.

Thanks again.

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Fri, 31 Oct 2008 12:01:17 EDT

Ian D Allen :

said by Bicephale :

3) Almost insignificant noise at night, past midnight/before mornings ... Number three makes me wonder what Bell's drop was really capable of.

Exactly. I get great speeds, very low error rates, from midnight to 6am daily, on both lines (3Mbit and 5Mbit). (Reference: »Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R) above from October 27.)

Clearly the line itself is capable of a solid 5Mbit, if it weren't for the interference during the daylight hours. Almost every morning, between 6am and 7am, something clicks on and my line is awful until it clicks off after 11pm. Sometimes, I also get a little oasis of goodness in the early afternoon.

Yesterday was a great day. From midnight to midnight Thursday, the noise only knocked out the line between 08h00-09h00 and 15h30-21h30. Today is shaping up to be a great day - the usual 7am line catastrophe didn't happen, though errors were up a tiny bit. (Maybe the guy causing the daily interference took the day off.)

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Fri, 31 Oct 2008 07:57:05 EDT

Bicephale : Hi again TerTech,

Euh... About "daily patterns", lets not forget this was over five days:

What does this all mean?, Bicephale, 2008-Oct-22

I can see three daily patterns:

1) Noise dominated by FEC errors during late mornings and afternoons
2) Noise dominated by CRC/HEC errors during evenings/early mornings
3) Almost insignificant noise at night, past midnight/before mornings

Number three makes me wonder what Bell's drop was really capable of.

:p

I wouldn't even bet all of your Attenuation level results from length!...

Just can't wait to hear what the brand new MoDem will happen to tell!

;)

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Fri, 31 Oct 2008 07:22:59 EDT

Bicephale : Hi TerTech,

I bet you have found a way to use 'DMT' with FirmWare
v6.2.29.2 installed by now so i simply won't insist... It's
a real pleasure to read about your new diagnostic tool,
please allow me to express a suggestion while it's still in
the creation phase! You may find suitable to add a time
offset option so that long-term records can be averaged
easily, this might even be the core of some main feature:
the ability to reveal daily patterns through averaging...

:D

Lets suppose one 24 h set starts at 0:00:07, the next at
0:00:04, the 3rd at 23:59:56 and so on, which is exactly
what happened when i tried to analyze your samples last
week. The resolution being fifteen seconds in this case, i
am in favour of aligning the columns on 0:00:00 hours to
ensure the same time slices of successive days coincide.

This way, one can be sure to avoid some potential issues
before he starts working on his numbers in the future and
yet there are two more which i find most noticeable about
log files such as those from 'DMT': 1st, the error counters
are no error rates (a SpreadSheet must be used to obtain
them later); 2nd, negative values pop out after a counter
reset due to the MoDem rebooting or whatever and that's
a nuisance because the large negative values which these
events can generate are aberrations capable of obscuring
the real part of a curve by steering its scale off-range...

Considering that your application draws curves anyway, i
can infer that it already calculates the error rate between
two samples so you must have seen negative excursions.

My solution was to clip those values at -1 in order to keep
track of the related events while everything else remains
relatively unaffected. A few days later i realized that -0.1
or even -0.00000001 could have worked better, actually...

:p

Ideally, a 24 h record with 15 seconds resolution should
start no later than at 23:59:45 if there is to be any error
rate record for the next time slice (at midnight). Imagine
a SpeadSheet with a cell on the 23:59:45 line and another
cell on the 0:00:00 line, the 1st value must be calculated
at 0:00:00 and then the 23:59:45 line can be dropped. If
long-term records are fully automated via a set of timers
it means that two log files will overlap for one time slice,
which implies that the file names should be different and
include both the date and time of creation, quite likely.

One additional comment, if i may: the TelNet CLI is OKay
for those who experience no issue with a separate router
but i like how 'RouterStats' works with the HTTP internal
server instead as it opens the door to a more convenient
method: HTTPS would allow any individual with very little
knowledge of his router to obtain statistics from the WAN
side - there is a "Remote Assistance" ST option, after all!

Such a design would give access to a totally new niche.

;-)

Those are my thoughts about 3rd-party tools, please take
them for what they are. I've let myself dream out loud but
i'm well aware that there's a real life part in the balance...

You can be certain whatever you come up with i'll be glad
to give it a try if ready. Don't hurry up though, i'm patient.

:)

Sorry to hear that the issue of your underground wiring is
so confusing you can't determine who's responsible for it. I
will just consider this can't be fixed even if you're willing to
take action: Bell digging under your asphalt, to repair their
own mistake?! No way, it wouldn't happen in your lifetime!

:o

Good luck with trying FirmWares, i can vaguely recall that
my ST546v6 stopped bleeding SNR Margin with v7.4.4.7 or
so but this isn't really relevant since our profiles are hardly
comparable... No tweaks past v6.1.9.6, in any case, and i
rather recommend v6.1.0.5 to spare one's mental sanity!!!

;-)

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Thu, 30 Oct 2008 11:38:01 EDT

tertech :

said by Bicephale :

Hi TerTech,

First of all, i'm not familiar with 'DSL Stat Scope' so it
would be nice to know where to find it exactly: as you
can imagine, i'm always eager to evaluate new toys!...

:D

DSL Stat Scope is a 'work in progress'. When I got the new ST516 modem with firmware 6.2.29.2.0 it would not run with DMT's diagnostic mode. RouterStats-Lite only shows speed and SNRM. You indicated that these numbers are only part of the story without the CRC FEC and HEC error numbers.
On the rainy week-end I got ambitious and wrote this program to read the modem stats via a Telnet session and store them in a file of similar format to the DMT diagnostic CSV format. Once I had that, I figured why not add a real-time graphic display of the stats. It's not ready to distribute to the world, but once I've worked out some bugs, you are welcome to try it if you like. It only works on the ST516 so far.

I can appreciate how the humour notes are connected,
the tiny bug near your NID or that pacifying beaver on
your aerial drop-cable are simple variations on a theme
which were funny - although not accurate. You're quite
right to point out that this is an inferior DSL service, of
course; there's no way for us to deny this fact but i'm
afraid you happen to have more control over the issue
than most of us usually do: if your underground cabling
turns out to be a two hundred feet long induction loop,
as i suspect it is, then Bell is *not* responsible for that
3rd-party design fault. The good news is you're totally
entitled to see that it is fixed, the bad news is there's
no easy way to deal with that since it's carved in stone,
practically speaking! I'll bet Bell just failed to teach you
sound underground installation methods but did you call
them for advice? Would you have paid the extra cost of
low-noise environmentally-proof wiring should Bell have
told you that the benefits would be marginal, say 3008/
640 Kbps Interleaved instead? Since you disregard A.M.
radio interference as potential trouble then i guess you
would have done nothing different, would you? Months
later, though, i believe that you might have discovered
that those of us who observe related notches into their
'DMT' spectral response curves rarely have single-tone
4.3125 KHz voids: typical trenches are many tones wide
and their shapes conform to continuous variations... If
we can judge by the multiple 'DMT' captures posted here
so far then it's obvious the single-tone notches must be
theoretical at best because i don't see much around.

;)

All seriouslessness aside I truly don't know who laid that wire in the trench and if it's even in the same trench as the power wires. I hired a contractor to build my house; he sub-contracted tradesman to do the work. It might have even been Bell who ran that wire from their box beside the road to by demarc. When I showed up on the construction site, the trench was already filled in. No it's not carved in stone, just buried under asphalt. I also thought about running a wire on top of the ground to replace the buried one. I don't know if I can do this without Bell sending a technician. If I accidentally cut my neighbor's phone line I might be in big trouble. :uhh:

To make it worst, it's not only one local A.M. station
but a plethora of sources from diverse locations which
you will have to account for: A.M. stations don't need
to be audible by the human ear to interfere with a DSL
MoDem, in my opinion. Now, top that with appliances in
the neighbourhood which are generating electric arcs,
etc., and finally, keep in mind that mother nature also
spits out a wealth of noise in the clouds right above...

To me it looks like noise is everywhere around us so all
it takes to collect it is a long piece of copper cabling!

:p

Exactly what I'm thinking. Random noise from motors, arc-welders, etc is undisciplined (spread spectrum) unlike radio transmission which are well structured narrow band. From the point of view of the radio station it's not 'noise'; too bad you're trying to use their alloted frequency for DSL. In any case, my modem only uses tones up to #128 (552kHz); that's just at the edge of the AM radio band. Even the powerful CFRA 580 should not bother my DSL.

Which brings me to my next question... The transformer
being so close to your underground setup it leaves very
little wiring exposed to the airborne noise, i presume, so
i wonder: how much aerial electrical wiring is there that
might behave like an antenna, actually? If there's so few
and the new device works so well then perhaps a simple
explanation would be that the noise came from your old
DSL unit itself! Would you say it was just failing and/or it became over-sensitive to noise somehow and it wasn't
resetting because of excessive noise, after all?... Euh...

I still have the old modem. At some time, I'll try it again with the same firmware as the new modem. That should tell if it's the modem hardware. I don't know the cause of it re-syncing, but it did it often and there were those periodic total chaos times when it wouldn't hold a connection for more than a few seconds unless I tweaked.

As i argued before, "Que Sera, Sera"!!! Lets hope this is
the end of your misery and that snow absorbs the rest!

Amen brother!

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Thu, 30 Oct 2008 10:38:39 EDT

Bicephale : This thought crossed my mind. The Europeans
got their giant particle collider so why wouldn't
Canada have its own underground project?! I'll
bet TekSavvy may even want to provide funds
for research on TerTech's 200' noise collector!

:p

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Thu, 30 Oct 2008 08:57:19 EDT

Ian D Allen : If it's only 200 feet, have him buy a 250-foot spool of CAT-5 and run a temporary wire above ground from the road to your home. If that fixes the problem, you know it's your underground and you have to find a way to run a second wire. If it doesn't fix the problem, you can look elsewhere.

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Thu, 30 Oct 2008 07:43:02 EDT

Bicephale : Hi Ian,

Was your home empty on October 24 between
12 h and 18 h, which was a Sunday afternoon?

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Thu, 30 Oct 2008 06:40:03 EDT

Bicephale : Hi TerTech,

First of all, i'm not familiar with 'DSL Stat Scope' so it
would be nice to know where to find it exactly: as you
can imagine, i'm always eager to evaluate new toys!...

:D

I can appreciate how the humour notes are connected,
the tiny bug near your NID or that pacifying beaver on
your aerial drop-cable are simple variations on a theme
which were funny - although not accurate. You're quite
right to point out that this is an inferior DSL service, of
course; there's no way for us to deny this fact but i'm
afraid you happen to have more control over the issue
than most of us usually do: if your underground cabling
turns out to be a two hundred feet long induction loop,
as i suspect it is, then Bell is *not* responsible for that
3rd-party design fault. The good news is you're totally
entitled to see that it is fixed, the bad news is there's
no easy way to deal with that since it's carved in stone,
practically speaking! I'll bet Bell just failed to teach you
sound underground installation methods but did you call
them for advice? Would you have paid the extra cost of
low-noise environmentally-proof wiring should Bell have
told you that the benefits would be marginal, say 3008/
640 Kbps Interleaved instead? Since you disregard A.M.
radio interference as potential trouble then i guess you
would have done nothing different, would you? Months
later, though, i believe that you might have discovered
that those of us who observe related notches into their
'DMT' spectral response curves rarely have single-tone
4.3125 KHz voids: typical trenches are many tones wide
and their shapes conform to continuous variations... If
we can judge by the multiple 'DMT' captures posted here
so far then it's obvious the single-tone notches must be
theoretical at best because i don't see much around.

;)

To make it worst, it's not only one local A.M. station
but a plethora of sources from diverse locations which
you will have to account for: A.M. stations don't need
to be audible by the human ear to interfere with a DSL
MoDem, in my opinion. Now, top that with appliances in
the neighbourhood which are generating electric arcs,
etc., and finally, keep in mind that mother nature also
spits out a wealth of noise in the clouds right above...

To me it looks like noise is everywhere around us so all
it takes to collect it is a long piece of copper cabling!

:p

Which brings me to my next question... The transformer
being so close to your underground setup it leaves very
little wiring exposed to the airborne noise, i presume, so
i wonder: how much aerial electrical wiring is there that
might behave like an antenna, actually? If there's so few
and the new device works so well then perhaps a simple
explanation would be that the noise came from your old
DSL unit itself! Would you say it was just failing and/or it
became over-sensitive to noise somehow and it wasn't
resetting because of excessive noise, after all?... Euh...

As i argued before, "Que Sera, Sera"!!! Lets hope this is
the end of your misery and that snow absorbs the rest!

By the way, which FirmWare version is it again, please?

:)

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Wed, 29 Oct 2008 16:10:02 EDT

tertech :

said by Bicephale :

Hi TerTech,

I wasn't sure what to reply so i needed to think it
over... My impression is that we didn't understand
each other at some point, the beaver joke might be
funny but the underlying assumption is wrong in the
present context: correct me if you must but i think
Bell just can't be responsible for that major design
fault relatively to your underground wiring!!! You're
the owner of this cable, if i can trust what i could
gather from your posts, at best those beavers must
be replaced to point at Hydro-Ontario instead!... I
appologize to you if my remark sounds bad, i did see
the potential for a joke but i'm not even convinced
this noise is from Bell at all, quite on the contrary.

If Bell is to blame it's for not letting you have some
closer DSL feed.

:uhh:

Excuse my goofy humour. Since you saw the tiny little bug on the photo of my NID, I was joking that I found a tazer wielding beaver on your drawing. I don't have much clip-art; I had to choose from a beaver, a horse or a dog. The beaver looked the most threatening of the three.

W.R.T. Both Bell and Rogers - they have sorely neglected updating their infrastructure in the south suburbs of Ottawa. For years, I was on an old CO that was barely sufficient when the area was all farms. Rogers was analog only, and was several years to bring Digital TV & Internet; ExpressVU and StarChoice took away a lot of their TV business for the new homes. Rogers finally provided internet but not all the digital TV features. Eventually, Bell built additional remotes and offered DSL. So, yeah, I blame them for taking my money and offering me an inferior service.

It's not clear to me how i should explain my concept
but the noise doesn't have to come from any specific
location: this nuisance may very well not come from
Bell nor even from your home, i've suggested the idea
that this is a mix of radio interference picked up by
Hydro-Ontario's poles and it's being injected into the
DSL signal through that underground "noise collector"
of yours... Perhaps spectral analysis would tell if the
noise present on those lines is related so i hope that
trouble-shooting tool is still an option in the future...

My opinion is that you'll find your domestic appliances
are no noise source but they're a vector through which
the airborn noise is allowed to reach an induction loop.

As a result, daily patterns originating in the ionosphere
(and only God knows what else) are likely to combine
in an apparent complex, hard-to-predict manner. Once
again, at an off-scale Attenuation range such as yours,
i would expect the picture to get blurred, so to speak.

Yeah, I follow you. I discounted AM radio interference because each station is contained in a narrow-band specific frequency; they should not radiate out-of-band. The Discrete Multi-Tone modulation of ADSL should easily avoid such a narrow-band interference. I suspect a spread- spectrum noise; but I have no theory about the source. Normally, spread-spectrum signals are lower in amplitude than narrow-band; again the DMT modulation with interleave and Reed Solomon FEC should be able to handle that too.

In the meantime, i guess Hydro-Ontario transformers
attenuate radio signals due to their intrinsic inductance
so i'm curious: how far are you from their transformer?

Thanks for the input!

:p

The transformer is directly across the road from the end of my driveway, so I'm about 230ft from it to my panel.

Just to note... My new modem has been behaving very well since I installed it last Thursday. The screen-shot shows my d/l speed at 2496kbps(yellow) and SNRM averaging at 7.0dB (white). The FEC error counts are 0, and the CRC and HEC errors occur at a rate of less than 3 hits per second. (red,maroon,purple).

Today's snow on the wires has not had any adverse effect. Maybe snow is an RF insulator ? j/k ;)

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Wed, 29 Oct 2008 11:50:27 EDT

Bicephale : Hi,

580 KHz sounds quite right and the one question
i have in mind is where does it penetrate the DSL
signal path exactly?... If Bell's aerial twisted pairs
do not pick up significant radio interference then
it's from wiring between the pole and the MoDem;
if it comes through an induction loop then there
is a well defined location to focus attention on...

Perhaps it's not possible to break some types of
induction loops, i wonder as this is a new theory
to me but i believe that's just what i done before
i moved from my former place: 1" galvanized iron
tubes reduced the noise enough for me to get a
FastPath profile and i suspect i could have used
shielding paint just the same... Shielding sleeves
which can be grounded will also provide excellent
isolation, as i recall. The offending noise must find
an entry point somewhere, find that point and it
should become clear what cure there is. A person
who reads this site tells me that it would help to
do some pass-band filtering close to the MoDem:
this will limit input saturation due to strong local
A.M. stations, if i understood him correctly. This
makes a lot of sense to me, i'm not certain which
range of frequencies will work best though.

:o

When you write about canned packages i imagine
a LiveCD (like Davix, to which i refered in another
thread). In your case, i guess it's a normal Linux
installation instead. I'm afraid i'm no more at ease
with fixed Linux than i am with databases, sorry...

:huh:

-=*=-

I was told to visit this site very recently, maybe
this is some good reading for our audience too:

»www.arrl.org/FandES/field/regula···i-legal/

Now, here is a must-read reference document:

»www.audiosystemsgroup.com/RFI-Ham.pdf

I advise anyone to pay special attention to page 23!

:D

Here are some more links of my own about shielding:

»www.magneticshield.com/products/cable.html
»www.spira-emi.com/products/groove.htm
»www.lessemf.com/paint.html
»www.safelivingtechnologies.ca/RF···SF54.htm
»www.csishield.com/mp_8.htm
»www.swift-textile.com/products.htm
»eastcoastshielding.com/s150.html

The part about "Inter-8 Weave Cable" is stimulating!

;)

That's the filter used by my European correspondant:

»www.k-comfilters.com/p_rf1c.asp

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Wed, 29 Oct 2008 10:09:42 EDT

Ian D Allen : I've been custom-massaging the extended "adsl info" output, and just started looking at the bit-per-tone data from the debug commands. I extract one sample per minute from each modem and store it all. One line has been going for 22 months, the other since last week. (Disk is cheap.)

When I'm curious about something, I write a shell/Perl script to extract one or more columns and rows of the data, perhaps average it a bit, and produce a file that I can plot using the Linux "gnuplot" command.

A canned package such as RRDTool will collect and compress large volumes of data, averaging older stuff together. It's an efficient way to manage a lot of data; but, it loses the detail of older event information in the averages.

The awkward part now is that I have this huge data file that's too large to to process as one chunk and equally awkward to search to extract just part of the data. I really need an ISAM file!

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Wed, 29 Oct 2008 09:33:26 EDT

Ian D Allen : Hey - CFRA blasts out a strong signal at 580Khz on the AM dial here in Ottawa. Is that what's creating the big interference notch on both my DSL lines between 552 and 621 Khz?

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Wed, 29 Oct 2008 09:20:37 EDT

Bicephale : Salutations Ian,

I'd like to read more, especially on your setup!

:)

Perhaps you can bring hope into this thread:

»Informative scripts to run with RRDTOOL.

:D

Your demonstration on averaged daily records
was wonderful, i believe we can even observe
weekdays vs weekend differences! It would be
interresting to have curves reflecting domestic
current consumption (and hence low resistivity
path) history for later comparisons. Would you,
by luck, have access to an RS-232 compatible
digital meter?... Maybe your electrical circuit is
more involved than you think.

;)

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Wed, 29 Oct 2008 05:47:39 EDT

Ian D Allen : Above are the current bit profiles for my two lines. The wet DSL line on top is the one for which I have minute-by-minute "adsl info" log data going back about 22 months. The dry DSL line below it is the new one from NCF/Teksavvy put in last week. The green error bars show the bit range for each tone. The black line is the average. You'll note that both lines have a severe interference notch betwen 552 and 621KHz.

The current storm in Ottawa is not helping anything! I have severe static and AC hum on my third phone line which has caused my third DSL connection (Bell legacy 2.2Mbit) to drop all night. That line has even lost dial tone a few times. I put in a call to Repair Service for that. Good luck finding it.
--
| Centretown Ottawa (K1R 6S9), Ontario, Canada
| Home Page: »idallen.com/ Contact Improv: »contactimprov.ca/
| College professor (Open Source / Linux) via: »teaching.idallen.com/
| Defend digital freedom: »eff.org/

wet DSL line 3Mbit profile
dry DSL line 5Mbit profile

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Wed, 29 Oct 2008 04:46:11 EDT

Bicephale : Hi,

I'm speachless: your 2nd name ought to be patience!

:D

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Wed, 29 Oct 2008 04:44:46 EDT

Bicephale : Hi TerTech,

I wasn't sure what to reply so i needed to think it
over... My impression is that we didn't understand
each other at some point, the beaver joke might be
funny but the underlying assumption is wrong in the
present context: correct me if you must but i think
Bell just can't be responsible for that major design
fault relatively to your underground wiring!!! You're
the owner of this cable, if i can trust what i could
gather from your posts, at best those beavers must
be replaced to point at Hydro-Ontario instead!... I
appologize to you if my remark sounds bad, i did see
the potential for a joke but i'm not even convinced
this noise is from Bell at all, quite on the contrary.

If Bell is to blame it's for not letting you have some
closer DSL feed.

:uhh:

Well, there's impressive feed-back from Ian D Allen
who may be in a better position to help you but i'll
insist on some other detail which appears to escape
your attention, nonetheless: the records did reveal
a daily pattern, only it simply occurs at night instead.

It's not clear to me how i should explain my concept
but the noise doesn't have to come from any specific
location: this nuisance may very well not come from
Bell nor even from your home, i've suggested the idea
that this is a mix of radio interference picked up by
Hydro-Ontario's poles and it's being injected into the
DSL signal through that underground "noise collector"
of yours... Perhaps spectral analysis would tell if the
noise present on those lines is related so i hope that
trouble-shooting tool is still an option in the future...

My opinion is that you'll find your domestic appliances
are no noise source but they're a vector through which
the airborn noise is allowed to reach an induction loop.

As a result, daily patterns originating in the ionosphere
(and only God knows what else) are likely to combine
in an apparent complex, hard-to-predict manner. Once
again, at an off-scale Attenuation range such as yours,
i would expect the picture to get blurred, so to speak.

This whole situation might be a severe test-bench for
FirmWare/HardWare evaluation and your observations
may teach us a few lessons, hopefully. I wouldn't put
too much faith in easy solutions but "Que Sera, Sera"!

In the meantime, i guess Hydro-Ontario transformers
attenuate radio signals due to their intrinsic inductance
so i'm curious: how far are you from their transformer?

Thanks for the input!

:p

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Mon, 27 Oct 2008 09:30:24 EDT

Ian D Allen :

said by tertech :
What kind of attenuation does your modem show?

For both lines: 48-49

I have not seen a consistent repeatable time-of-day pattern yet.

I just took the last 20 weeks and 93 weeks of modem data and averaged them onto one week each. The results show a clear pattern of bad signal quality from about 6am to midnight daily. Midnight to 6am is pretty good, every day of the week. The attached graphs show the "15 minute error seconds" and the number of times the line reset, for the averaged week.

Average errors/resets for previous 20 weeks
Average errors/resets for previous 93 weeks

Re: Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Sun, 26 Oct 2008 12:47:10 EDT

tertech :

said by Ian D Allen :

I've been having problems with my 3Web ADSL service dropping sync at
certain times of the day for almost two years. Last week I got NCF
(Teksavvy) DSL on a dry line to see if anything would improve. I also
borrowed a second Speedtouch 516v6 modem, which had newer software.

The newer modem and software makes a big difference in a better error rate
and reduced line resets; but, both lines do show sudden-onset "storms"
of unusability at the same times of the day. Nights are usually good,
with full speeds possible on both lines. (The 3Web line has had its
profile reduced from 5Mbit to 4Mbit and then 3Mbit in an effort to
eliminate the line resets. Hasn't worked.)

For example: This weekend (Oct 25-26) the oasis of good night-time signal
quality was destroyed by a noise storm starting around 6:30am Saturday.
The storm continued on both DSL lines all day until 21h30 Saturday and
then suddenly stopped. Signal was great all Sat-Sun night, with the
storm returning suddenly at 6:15am Sunday. It's still stormy now.

I checked my logs. I don't have any 'storms' in the time slots that you mention. With Saturday's rain, I did get some lower SNRM numbers with a few re-syncs through the day that recovered instantly, but I've stayed at my profiled max speed of 2496kbps throughout the weekend so far.

You've been putting up with this for two years and haven't found anything definitive yet? That's discouraging :(

What kind of attenuation does your modem show? We may be looking at a similar problem due to long line lengths, but I have not seen a consistent repeatable time-of-day pattern yet.

edit: BTW - I'm near Manotick; I doubt the exact same noise source could affect us both.

Ottawa Centretown line noise stats - Preston/Somerset (K1R)

Sun, 26 Oct 2008 11:53:32 EDT

Ian D Allen :

That firmware has been holding my sync well !! 6.5-7db SNR.

I've been having problems with my 3Web ADSL service dropping sync at
certain times of the day for almost two years. Last week I got NCF
(Teksavvy) DSL on a dry line to see if anything would improve. I also
borrowed a second Speedtouch 516v6 modem, which had newer software.

The newer modem and software makes a big difference in a better error rate
and reduced line resets; but, both lines do show sudden-onset "storms"
of unusability at the same times of the day. Nights are usually good,
with full speeds possible on both lines. (The 3Web line has had its
profile reduced from 5Mbit to 4Mbit and then 3Mbit in an effort to
eliminate the line resets. Hasn't worked.)

For example: This weekend (Oct 25-26) the oasis of good night-time signal
quality was destroyed by a noise storm starting around 6:30am Saturday.
The storm continued on both DSL lines all day until 21h30 Saturday and
then suddenly stopped. Signal was great all Sat-Sun night, with the
storm returning suddenly at 6:15am Sunday. It's still stormy now.

Here are the abysmal stats during the storm:

Location: Ottawa K1R - Centretown - Preston/Somerset
 
Speed test:  »speedtest.cogeco.net/
 
NCF (Teksavvy) Dry DSL 5 Mbit:
 
    Modem in use: Speedtouch 516v6 firmware 6.2.16.3
 
    5 line resets in the past 24 hours ending 9am Sun Oct 26.
    0.57% ping failures
 
    Graph:  »idallen.org/tmp/teksavvy.png
 
    Speed Test #59926348 by dslreports.com
    Run: 2008-10-26 08:51:11 EST
    Download: 193 (Kbps)
    Upload: 652 (Kbps)
    In kilobytes per second: 23.6 down 79.6 up
    Tested by server: 7 java
    User: 2 @ dslreports.com
    User's DNS: teksavvy.com
    Compared to the average of 121 tests from teksavvy.com:
    * download is 95% worse, upload is 8% better
 
    Speed Test #59927232 by dslreports.com
    Run: 2008-10-26 09:11:25 EST
    Download: 185 (Kbps)
    Upload: 630 (Kbps)
    In kilobytes per second: 22.6 down 77 up
    Tested by server: 7 java
    User: 2 @ dslreports.com
    User's DNS: teksavvy.com
    Compared to the average of 122 tests from teksavvy.com:
    * download is 95% worse, upload is 4% better
 
Existing 3Web (formerly iStop) DSL 3 Mbit profile:
 
    Modem in use: Speedtouch 516v6 firmware 6.1.0.5
 
    77 line resets in the past 24 hours ending 9am Sun Oct 26.
    4.87% ping failures
 
    Graph:  »idallen.org/tmp/3web.png
 
    Speed Test #59927108 by dslreports.com
    Run: 2008-10-26 09:08:37 EST
    Download: 2107 (Kbps)
    Upload: 667 (Kbps)
    In kilobytes per second: 257.2 down 81.4 up
    Tested by server: 7 java
    User: 2 @ dslreports.com
    User's DNS: istop.com
 
    Speed Test #59927162 by dslreports.com
    Run: 2008-10-26 09:09:50 EST
    Download: 1683 (Kbps)
    Upload: 671 (Kbps)
    In kilobytes per second: 205.5 down 81.9 up
    Tested by server: 7 java
    User: 2 @ dslreports.com
    User's DNS: istop.com
 
Legacy Sympatico DSL 2.2 Mbit static IP:
 
    Modem in use: Westell A90
    0.009% ping failures
 
    Speed Test #59926504 by dslreports.com
    Run: 2008-10-26 08:54:19 EST
    Download: 1779 (Kbps)
    Upload: 987 (Kbps)
    In kilobytes per second: 217.1 down 120.4 up
    Tested by server: 7 java
    User: 2 @ dslreports.com
    User's DNS: bellglobal.com
 
    Speed Test #59926860 by dslreports.com
    Run: 2008-10-26 09:03:00 EST
    Download: 1750 (Kbps)
    Upload: 996 (Kbps)
    In kilobytes per second: 213.7 down 121.6 up
    Tested by server: 7 java
    User: 2 @ dslreports.com
    User's DNS: bellglobal.com

The graphs show that both the non-legacy DSL lines are affected by
the same noise events. (It doesn't seem to bother the Westell line.)

The noise doesn't originate inside this house, since I was the only
one here and I wasn't doing anything different at 7am or 9pm. The dry
DSL line goes from the new NID via Cat 3 and Cat 5 up to the modem;
it doesn't use any house wiring, and it's even a separate wire from the
pole going into the house in a different corner. The noise has to be
"out there" somewhere.

If enough of us Centretown people complain, can we get Bell to find the
noise source and get it away from the lines?

I have minute-by-minute "adsl info" stats saved on the 3web line since
January 1997. The graphs are pretty. »ian.idallen.ca/dsl/

I am saving NCF/Teksavvy stats since October 22 2008.

--
| Centretown Ottawa (K1R 6S9), Ontario, Canada
| Home Page: »idallen.com/ Contact Improv: »contactimprov.ca/
| College professor (Open Source / Linux) via: »teaching.idallen.com/
| Defend digital freedom: »eff.org/

3Web DSL line quality - past 30 hours
NCF/Teksavvy DSL line quality - past 30 hours

Re: ST516 - Strange discovery

Sat, 25 Oct 2008 17:58:34 EDT

Shikomu : Its got 18 VAC output.

Re: ST516 - Strange discovery

Sat, 25 Oct 2008 16:37:33 EDT

tertech :

said by Shikomu :

That firmware has been holding my sync well !! 6.5-7db SNR.
Its the one that seems to hold during noise bursts. Albiet the errored seconds occur then.

6 days without re-sync would be great! I just bought this new ST516 on Thursday; it has resynced just a few times since then. It came with 6.2.29.2; since it worked best for you, I figured I'd try it out before loading any v7 into it.

Does yours have an AC or DC power supply?

Re: What does this all mean?

Sat, 25 Oct 2008 16:33:07 EDT

tertech :

said by Bicephale :

Hi TerTech,

Here is what i imagined your situation is like... I
understand you can't install chokes on the main
power lines even if this should turn to be a valid
cure. Take note that the MoDem's power supply
is not a part of the noise induction loop in this
illustration, by the way...

:p

When I zoomed in to the power lines of your drawing, I found my problem!

Re: What does this all mean?

Sat, 25 Oct 2008 15:49:32 EDT

Bicephale : <Inner Voice Says>

Bicephale thinks it would be great to play
"Lucy in the Sky" followed by "Major Tom"!

Inner Voice Says>

:D

Re: What does this all mean?

Sat, 25 Oct 2008 15:42:17 EDT

Bicephale : Hi TerTech,

Here is what i imagined your situation is like... I
understand you can't install chokes on the main
power lines even if this should turn to be a valid
cure. Take note that the MoDem's power supply
is not a part of the noise induction loop in this
illustration, by the way...

:p

Re: ST516 - Strange discovery

Sat, 25 Oct 2008 15:18:40 EDT

Shikomu : That firmware has been holding my sync well !! 6.5-7db SNR.
Its the one that seems to hold during noise bursts. Albiet the errored seconds occur then.

Re: What does this all mean?

Sat, 25 Oct 2008 12:31:22 EDT

tertech : Bicephale,

Thanks for all your effort analyzing my data logs.

quote:
So, you explained to
me that the line had been relatively stable (read
"quiet") for a while and then something took place
at lunch time (past a noise burst) which caused the
DownStream SNR Margin level to improve. Well, for
me it now seems reasonable to suggest that this was
most likely related to something i like to call the
"cooking hour syndrome": a person in the kitchen
turns on a power appliance (even many), it completes
the noise induction circuit i refered to previously
with even more noise coming from the initial spark
and then your DSL device renegociates a better SNR
margin... How about that?

I believe that there is a randomly timed, infrequent electrical 'event' that causes the noise on my lines to increase, or perhaps the signal to decrease. In either case it leads to my modem losing sync and not being able to re-sync ( without tweaking ) until the event ends. I am 99.99998% sure that this event is not occurring within my house. In addition, the modem is behind an APC UPS with built-in surge suppression; it also seems to block high frequencies from getting through from the power lines to anything plugged into it. The old ST516 had an 18VAC power supply; probably just a transformer. The new one has a 22VDC switching power supply. I don't know if that should make any difference, but it's interesting that SGT changed their power supply from AC to DC.

quote:
n any case, i'll bet there's not only one type of
noise involved. With such a faint signal, i think
you have a nearly ideal test-bench with that noise
collector of yours! One can hardly imagine a more
severe test for Thomson's SpeedTouch and yours was
not so bad, after all, considering it's tweakable.

I'm glad that I can be the poster-child for E.N.S ( Electrical Noise Syndrome ). Maybe we can have a bike-a-thon to find a cure? :D

quote:
I don't know about you but the previous curves are
suggesting that your DSL signal is a lot more quiet
at night. May i ask if you ever tried to evaluate
the maximum attainable rate during that period? I
mean between 1 h and 6 h when you're all asleep...

I haven't yet stayed awake to the wee hours to see what kind of SNRM I would get. The 2496kbps is my max profile speed, so as long as it stays there, and the CRC, FEC, and HEC numbers stay very low, what's the difference if I get a slightly better SNRM?

quote:
Also, i regret you couldn't use a spectroscope in
order to compare the noise present both on the DSL
and power lines. If the envelopes had matched then
it would have been an strong argument in favour of
decoupling your electrical lines in the 0 - 1 MHz
range (where the DSL signal lives). Perhaps large
ferrite chokes would be sufficient to cut the noise
induction loop (providing this theory is correct),
power suppression coils and capacitors may be part
of a possible fix but it's easier to protect against
noise than to try to find a cure for it, evidently.

My wires are underground, and only pop up briefly to the meter; then they go back down under the garage and come through the basement wall right behind the main panel. The only place I see them is below the meter. If I try to put anything outside around the PVC pipes leading to the meter, I'm sure that Ontario Hydro would think I was trying to steal power with some hair-brained scheme, and cut my power off.

The new modem has been holding pretty well stable with 2496kbps and around 8dB to 9dB SNRM since Thurs. evening. I haven't seen any electrical mayhem yet. The old modem was getting only 3 to 4 dB when it made a 2496 connection.

Today it's raining; that has been the bane of my connection from the start. Now it doesn't seem to be having any adverse effects at all.

Only time will tell :uhh:

Re: What does this all mean?

Fri, 24 Oct 2008 07:17:06 EDT

Bicephale : Euh... "Protect against noise"...

To not let it in, in the 1st place, i mean!

;)

Re: What does this all mean?

Fri, 24 Oct 2008 07:07:03 EDT

Bicephale : Hi TerTech,

My appologies for the delay. So, you explained to
me that the line had been relatively stable (read
"quiet") for a while and then something took place
at lunch time (past a noise burst) which caused the
DownStream SNR Margin level to improve. Well, for
me it now seems reasonable to suggest that this was
most likely related to something i like to call the
"cooking hour syndrome": a person in the kitchen
turns on a power appliance (even many), it completes
the noise induction circuit i refered to previously
with even more noise coming from the initial spark
and then your DSL device renegociates a better SNR
margin... How about that? The next cooking period
might have occured when the noise picked up by the
electrical lines was more severe so the MoDem could
not cope with it as well as before anymore, hence
the SNR Margin drop and then continuous retrains a
bit later - which lasted for a couple hours. Oh,
the SNR margin was great when you decided to tweak
but we can see how it affected the DownStream rate:
it seems even 20 dB of DownStream SNR Margin is no
garantee for reliable connectivity! Knowing what
you admited lately, i'm not really surprized. The
Attenuation level is out-of-range, a professional
electric contractor managed to couple your signal
with airbone noise over a distance of two hundred
feet via a parallel underground run, euh... Maybe
it has to do with using 120 vs 240 Volts: one is
assymetrical while the other is balanced, i wonder.

:o

In any case, i'll bet there's not only one type of
noise involved. With such a faint signal, i think
you have a nearly ideal test-bench with that noise
collector of yours! One can hardly imagine a more
severe test for Thomson's SpeedTouch and yours was
not so bad, after all, considering it's tweakable.

:p

I don't know about you but the previous curves are
suggesting that your DSL signal is a lot more quiet
at night. May i ask if you ever tried to evaluate
the maximum attainable rate during that period? I
mean between 1 h and 6 h when you're all asleep...

Also, i regret you couldn't use a spectroscope in
order to compare the noise present both on the DSL
and power lines. If the envelopes had matched then
it would have been an strong argument in favour of
decoupling your electrical lines in the 0 - 1 MHz
range (where the DSL signal lives). Perhaps large
ferrite chokes would be sufficient to cut the noise
induction loop (providing this theory is correct),
power suppression coils and capacitors may be part
of a possible fix but it's easier to protect against
noise than to try to find a cure for it, evidently.

:uhh:

TerTech's DMTDiag (2008-Oct-19)

Re: What does this all mean?

Wed, 22 Oct 2008 21:36:21 EDT

Bicephale : Hi again TerTech,

I haven't looked at the other records yet but if
there's a daily pattern which can be associated
to your own home then maybe huge ferrite clamps
installed around the main power lines would break
the induction loop - or at least affect it enough
to provide sufficient improvement. Do you people
go to bed quite early? Something like 20h? Right
now i'd say we've made progress because "noise"
was only suspected to be a problem before and it
wasn't clear if the MoDem was working correctly.

Re: What does this all mean?

Wed, 22 Oct 2008 21:24:25 EDT

tertech : Hello Bicephale,

I'm happy you're enjoying my data by making it into psychedelic art. :D

Yes I agree with your earlier hypothesis. Noise is the ugly monster lurking in my wires - as it does in all data transmission medium. The telephone wires buried alongside the power lines for 200ft could very well be the induction point for the noise transfer. I thought that the phone cable was shielded. I peeled away some insulation to get at the shield with the intent of grounding it, but there was no shield! I guess the original installer had no idea that this would someday lead to ADSL problems as well as to hearing CFRA 580 AM on my phones. Too late to do anything about it now. :(

The thing is though... ADSL was designed to operate in noisy environments. For the most part, noise is actually other transmissions that are being induced onto the wires. These transmissions by nature are relatively narrow band and are constantly present although they may vary in amplitude from time to time. The tone selections made by the ADSL should avoid those bands with too much interference from these other signals. Most of the time it seems that my modem is doing fine separating the signal from the noise; even when the noise margin is barely above the minimum SNR. Given my off-scale attenuation it may not be the most optimal connection, but at least it's stable.

What I still haven't figured out yet is what causes my modem periodically to suddenly lose connection and fail repeatedly to re-connect. It doesn't seem to happen at any particular time that I can associate with any electrical event occurring in my house. It can last a few minutes or a few hours. I imagine it must be a spread spectrum noise of sufficient amplitude that it prevents the ADSL from finding enough usable tone slots to create a stable connection. The spectrum analyzer ( had it worked ) would have verified or refuted that. :huh:

One other possible problem could be the modem itself. Either the physical hardware or the firmware load might not be doing its job. To eliminate this possibility I bought another ST516 last night. It came with v6.2.29.2. Since Miro had success with that firmware, I thought I would try it first. So far for 1 day it's hanging on with the max speed 2496 and the SNRM is bouncing around 9dB or 10dB and only a handful of single event disconnect/re-connects.. I'll just have to wait for a week or so to see if it goes into crazy mode. This firmware does not support the modemoptions or diagnostic mode of DMT so I am only using RouterStats-Lite to log the connection speed and noise margins.

Re: What does this all mean?

Wed, 22 Oct 2008 20:34:40 EDT

Bicephale : Hi TerTech,

Here's today's post...

Dark colours: October 16, 2008 Light colours: October 17, 2008

I'll comment when i'm done painting tomorrow!

;)

1) Approximative timeframe added...
2) Complementary graphic added...

Re: What does this all mean?

Tue, 21 Oct 2008 23:22:08 EDT

Bicephale : Hi TerTech,

I've been busy working on a Linux LiveCD lately,
so that i can use 'GNUmeric' in a suitable way...

Here's what the 1st 'DMTDiag' log-file looks like:
just as i expected, the SNR Margin curves don't
convey much information while the "Noise" ones,
on the other hand, certainly provide enough clue
for me to suspect that you'll find no Mr. Hankey
in that MoDem. Oh, i'm afraid there sure is some
Hankey stuff involved but it must result from the
way your phone-line was laid out, if you ask me.

My hypothesis is simple. Electrical power lines as
those we've been used to see on top of poles for
decades behave like antennas picking up all sorts
of nuisances in the DSL frequency range. It's not
the 60 Hz signal itself which causes our MoDems
to go berzerk, it's the coupling between electrical
power lines and Bell's telephone lines which is at
fault for so many DSL customers: if a phone cable
has been tightly attached to a PVC power conduit
over long distances then it happens to be coupled
to it and low resistance power appliances such as
ovens, heating plates, etc., provide a return path
to close the loop which in turn induces that noise
into the DSL signal. Well, there are other sources
of noise, like CrossTalk, but i doubt it's what you
have here... With out-of-range Attenuation levels
such as yours, i guess a few feet of this parasitic
coupling would suffice to ruin the day.

Re: What does this all mean?

Mon, 20 Oct 2008 22:29:43 EDT

Bicephale : Hi TerTech,

Great! I assure you this makes my life simpler;
i will enjoy and maybe a few more readers too.

:)

Re: What does this all mean?

Mon, 20 Oct 2008 22:01:17 EDT

tertech :

said by Bicephale :

Hi TerTech,

It would really ease my reading if those .XLS files
were available: i need to see the overall picture!

O.K. I just don't want you to think you have any obligation to spend your time analyzing this data. If it's for your curiosity, enjoy; there's about a week's worth of capture. :D

DMT_DIAGS.ZIP 2,121,159 bytes

Re: What does this all mean?

Mon, 20 Oct 2008 20:06:59 EDT

Bicephale : Hi TerTech,

It would really ease my reading if those .XLS files
were available: i need to see the overall picture!

Re: What does this all mean?

Mon, 20 Oct 2008 19:31:49 EDT

tertech :

said by Bicephale :

Hi TerTech,

I wrote about this before, in the very 1st reply
which i published in this thread, actually... Your
SNR Margins are only part of the preliminaries, i
suggested that one must optimize those against
other factors such as Error rates/RCOs so that,
in the end, a real improvement is observed thru
"browser" speeds testing. You were provided an
example where connection speeds from one test
were slower than those of another and yet this
gave some better "browser" speeds, most likely
because of the lesser noise.

Hi Bicephale,

What I understood from your previous post was that it was possible to have a lower connection speed and still a better end-to-end throughput because of the error retries. Using a speed test would therefore be a better benchmark than simply looking at the connection speed. I understand this and agree it's true. What I was trying to show in my exhibit A-B-C-D evidence was that I was seeing lower FEC and CRC errors with a high speed (2496) low SNRM (3.5dB) connection than I was with the very slow speed (832) very high SNRM (20dB) connection. Again, I'm not trying to optimize my line; I just want it to stay alive, or at the very least only disconnect infrequently.

As I understand it... we generate a CRC on the payload data, then apply a Forward Error Correction algorithm, and finally interleave the data to spread a short burst 'hit' over a wider area and give the FEC a better chance at correcting it. On the receiver end, we follow the process in reverse. The final determination is CRC. If the calculated CRC matches the transmitted CRC, the packet was successful; if not, it's rejected. The whole reason for all of this is to overcome the effect of noise which is always present on any transmission medium. The noise can garble a portion of the transmission. The less noise, the more likely to get a successful transmission. The more noise, the less likely a successful transmission. I would then expect to see a higher FEC and/or CRC errors on a line where the SNRM was low.

I'll gladly comment
the data which you captured in graphic form but
don't expect me to copy the values cell by cell!

:p

Please, include your genuine .XLS files as a .ZIP
archive in order to allow me to proceed with the
rest of your post.

Very kind of you to offer, but I can't expect anybody to take their time to analyze hours and hours of my data. You had told me to use the DMT diagnostic mode to see if the error counts would rise just before the apparent line instability. The snaps I posted are around those events and I don't see any significant increase in FEC or CRC errors. The big surprise was the high CRC & FEC counts at the 832kBps 20dB SNRM.

Everything I've read about ADSL tells me that what I'm seeing is impossible, yet there you have it. Most of the time, I have a solid stable connection at 2496 with low FEC/CRC errors with an SNRM of half the minimum specification. At other times for a relatively short period of time I have a very poor connection that hardly makes ADSL practical. As stated before... like a switch was thrown.

At this point, there's no benefit to continue this thread; I'm sure everybody's sick of my whining. If I discover that my neighbor's hobby is re-animating corpses, and his Jacob's ladder arcing electricity is causing the interference, I'll let you know. :uhh:

Re: What does this all mean?

Mon, 20 Oct 2008 04:07:57 EDT

Bicephale : Hi TerTech,

I wrote about this before, in the very 1st reply
which i published in this thread, actually... Your
SNR Margins are only part of the preliminaries, i
suggested that one must optimize those against
other factors such as Error rates/RCOs so that,
in the end, a real improvement is observed thru
"browser" speeds testing. You were provided an
example where connection speeds from one test
were slower than those of another and yet this
gave some better "browser" speeds, most likely
because of the lesser noise. I'll gladly comment
the data which you captured in graphic form but
don't expect me to copy the values cell by cell!

:p

Please, include your genuine .XLS files as a .ZIP
archive in order to allow me to proceed with the
rest of your post.

What does this all mean?

Sun, 19 Oct 2008 21:48:03 EDT

tertech : After a few days of a relatively stable connection - only dropping the line and resyncing a few times a day, this evening, my modem went into spasms.

I looked at the output of the DMT diagnostic that I had started at around 11:30 this morning. For about a half hour it was synced at 2496 with an SNRM hovering around 3 or 3.5dB; the rxFEC was 0. Just after 12:00 it hiccuped a few times, but finally resynced at 2464 with an SNRM of 7. Strangely, the rxFEC count started to climb. Exhibit-A

At 17:25 it dropped connection and resynced. This time at 2496 with an SNRM back down to 4dB. The rxFEC stayed at 0, as it had done back at 11:30. Exhibit-B

At 18:08 it dropped again; this time it re-synced at 2464 with SNRM of 5.5 to 6dB. The rxFEC shot up rapidly. It couldn't hold the connection for very long. Exhibit-C

From then on it was up and down until I finally laid my hand upon it at 20:07 and adjusted the target SNRM. It synced-up at 832/576 with an SNRM around 20dB. The rxFEX shot up rapidly, but the connection has held since then. Exhibit-D

My belief system has been shattered :( I had always believed that the higher the SNRM, the less errors would occur. This data clearly shows there is a higher incidence of CRC errors as well as FEC detected errors at 20dB than there is at 3dB SNRM.

Bicephale, If you can explain this to me, you're a genius! I am at a total loss. :huh: It hasn't rained today, and I'm not aware of any electronic equipment being turned on at this time. There are no street lights nearby; no lights with or without dimmers were turned on until after 18:00.

Exhibit A
Exhibit B
Exhibit C
Exhibit D
RouterStats

Re: Time to recap.

Fri, 17 Oct 2008 01:43:26 EDT

Bicephale : Hi TerTech,

'RouterStats-Lite' only monitors Connection Speeds
and SNR Margins but one can reboot his MoDem all
he wants and yet the recording won't be interrupted
so, considering that rebooting is triggered long after
a glitch, if it was manifesting itself as an SNR Margin
event then 'RouterStats-Lite' should have caught it
(provided that resolution was good enough, that is).

Here's a long-term (24 h) capture of mine. This one
was made up of 1440 samples (1 minute resolution),
we can see an abrupt change in the SNR Margin and
its effect on the Connection Speed - which strongly
suggests a MoDem reboot/resynchronization phase:

'RouterStats-Lite v3.5' + Thomson SpeedTouch 5x6v6, Bicephale, 2008-Sep-12

This capture tells us there was no resynchronization
possible after the 2nd event; i also remember being
able to verify that the pseudo-stable plateau near 4
Mbps just didn't provide real connectivity, actually.

:p

I wrote many times before that i don't trust the SNR
Margin curves alone to trouble-shoot... The numbers
below came from 'OpenOffice v2.4.1 for Windows XP'
(i don't recommend it, by the way) after reading the
log-file produced by 'DMT' while in "Diagnosis" mode:

From the ground up!, Bicephale, 2008-Oct-3

Notice the yellow line filled with zeros? This line was
generated during a MoDem reboot event and we can
see the bursts taking place right before. So, it's not
gone: the information remains available nonetheless!

:D

I found that 'OpenOffice Calc for Linux' or 'GNUmeric'
worked better for reading large 'DMT' Diagnosis files;
this is another example where the Connection Speed
dropped to zero and yet recording simply continued:

From the ground up!, Bicephale, 2008-Oct-7

'DMT' in Diagnosis mode collects practically all of the
parameters you can obtain from your MoDem, it does
keep track of major events causing a MoDem reboot
so i'm not sure what more you expect of a 3rd-party
trouble-shooting tool, exactly! Perhaps the MoDem's
own logs? Well, i wouldn't be surprized if this should
happen to be a supported option waiting to be found.

;)

Re: Time to recap.

Thu, 16 Oct 2008 14:39:20 EDT

tertech : Hi Bicephale,

I have been running both DMT in diagnosis mode as well as RouterStats-Lite.

I haven't seen a 'daily' pattern yet; I've only caught it happening twice - both times around the 16:00 - 17:00 time frame. The past day or so the stats have been corrupted by my attempt at connecting the spec analyzer but I can ignore those times.

If the connection gets really bad and the modem keeps dropping, the stats from DMT and RouterStats goes to 0. It would be nice to see what's really going on the line when the modem can't even get a sync. Once the modem does sync and gives stats, the worst has already passed.

Re: Time to recap.

Thu, 16 Oct 2008 14:13:49 EDT

Bicephale : Hi TerTech,

A while ago i suggested you to collect some long-term
noise curves using 'DMT' in "Diagnosis" mode as i know
all too well there's just no clue in the SNR Margin ones
at times. Are you certain this calls for spectral analysis
using costly test equipment?... I, for one, would pretty
much appreciate it if you could show us a "voiceprint"
of the nuisance but it seems excessive and premature
for those who didn't see the evidence which you appear
to have accumulated so far... If i understand correctly,
there is a daily pattern so is it not time for Error rate
curves?

:p

Time to recap.

Thu, 16 Oct 2008 13:21:36 EDT

tertech : Before I go drifting off into 'tangent land' with debugging test equipment, let me just recap the reasons I started this discussion...

When my service was first installed I had connection problems with frequent line drops. I phoned TSI who opened Bell tickets. Various Bell Techs showed up at my house and tried swapping my wire pairs, etc. Finally someone showed up with a proper DSL analyzer and said my d/l profile needed to be lowered to 2496.

That worked O.K. for a while, but changes in weather, like rain would cause frequent disconnects. I read here about the DMT tool that allowed me to view the line stats. It showed that my SNRM was marginal and probably the cause of the disconnects. I also found that DMT allowed me to set a target SNRM which the modem would negotiate. Now I could overcome the disconnects by simply setting a higher (10dB) SNRM target.

For about a year I used this process when my line went wonky on rainy days to push up the SNRM and lose some d/l speed. As long as I got over 1mBps, I was still happy about switching to DSL from my 1mBps Rogers plan.

About 6 months ago, it started to get worse and worse. I found that I had to set an SNRM over 15dB just to get an 800kBps d/l speed. Now the thought of going back to Rogers kept creeping into my brain - but I resisted the dark-side.

When I read here that TSI was reviewing their customer lines, I thought they would surely find my flaky connection. While waiting for them to contact me, I wanted to collect some stats to help them, and also to try some stuff on my own.

By my reasoning, a low SNRM either means that the noise level has gone up, or the signal level has gone down. I played with the 'remote tx pwr' setting of DMT and did a re-sync. Magically, my speed jumped up to 2496kBps, and the SNRM went as low as 3dB. I had never seen a stable connection that lasted more than a few seconds with such a low SNRM; this went for several hours. "Eureka" I though; I fixed it!

Later that week, the modem went right back into its bad behaviour with no apparent reason. Later on it recovered and went back to a stable 2496 connection. This occurred again a day or so later; both times between 16:00 and 17:00. That's not enough samples to make a conclusion, so I started collecting stats every day.

I also discovered that it didn't matter what I did with the 'remote tx pwr' setting on DMT. It made no difference. It was just coincidence that my first re-sync after changing that setting resulted in a good connection.

So, here I am now. I'm still gathering daily statistics, but I would like to look at the DSL line while the modem is exhibiting the bad behaviour. Since I can't really see what the DSL line stats are while the modem is disconnected attempting to re-sync, I thought of using the spectrum analyzer. Unfortunately the only SA I managed to get my hands on is broken. I have used SAs before to view L-Band satellite signals; I know how they should behave - this one isn't.

So, I'll keep collecting daily stats for the time being until I detect a repeatable pattern, and/or another SA becomes available. We have all kinds of couplers and isolators that I can use to match impedance. These are very expensive pieces of equipment; they should not induce 60hz noise into the circuit you're testing.

Re: Update.

Thu, 16 Oct 2008 08:57:32 EDT

minoe : A balanced input adapter or a simple isolation transformer on the input of the spectrum analyser will solve the problem.
The circuit is a simpler version of the one shown above.

Also, if you put an A/C line isolation transformer on the test equipment, that might fix the trouble.

Re: Update.

Thu, 16 Oct 2008 00:51:13 EDT

Bicephale : Hi TerTech,

Since you raised the matter of tempering with
Bell's equipment i guess galvanic isolation may
be called for so this circuit should do the job.

You only want the frequency spectrum curve,
i see no need to get actual amplitude levels...

Non-Invasive DSL Monitoring

Re: Update.

Wed, 15 Oct 2008 23:06:14 EDT

tertech :

said by minoe :

The ground connection on the test equipment is most likely to blame for the 60 cycle buzz.

If your scope has a the option of a differential input (using channel A and B without the ground) you will probably be able to scope it without the buzz.

I think you're right. It unfortunately has just a single input; it's a spectrum analyzer. I don't think any of the spectrum analyzers at work have a differential input if there even is such a thing. Among the accessories was a filter that blocks DC and attenuates low frequencies. It didn't help. The fact that a shorted input still shows -30dBm signals at various frequencies tells me something's gone haywire inside it. I'll have to wait until one of the calibrated working units is available, and I might be able to borrow it overnight sometime.

Re: Update.

Wed, 15 Oct 2008 22:01:15 EDT

minoe : The ground connection on the test equipment is most likely to blame for the 60 cycle buzz.

If your scope has a the option of a differential input (using channel A and B without the ground) you will probably be able to scope it without the buzz.

Update.

Wed, 15 Oct 2008 20:52:56 EDT

tertech : Bicephale,

The splitter in your picture looks like a Wilcom PS11. It's meant to go right inside the NID on the Telco side of the demarc. We know better than to tamper with the Telco side of the demarc now don't we? :uhh:

»www.wilcominc.com/product_pdfs/PS11.pdf

My splitter is designed for the customer side of the demarc - outside near the NID...

»www.suttleonline.com/pdfs/649A1.pdf

As far as getting the spec analyzer connected, the impedance does not seem to be as big an issue as I thought. An RF engineer at work suggested to just put a 50 Ohm resistor in series with the probe. If I'm not interested in making precise dBm measurements, it won't matter that I'll be attenuating the signal slightly. The bigger issue seems to be that the instrument is injecting a loud 60 cycle buzz into the phone circuit. Even though it's connected on the DSL side of the splitter it makes it impossible to use the phone. Reversing tip-ring makes no difference. I even tried putting an isolation filter LF blocker on the probe. No help. I suspect that this spec analyzer is toasted; even with the input shorted, I see phantom signals here and there across the whole base-band spectrum at around -30dBm. The power supply might be bleeding through to the amplifier and out the input connector. No wonder nobody cared if I took it home :huh:

I'll have to re-think my plan. This weekend I will look in the NID to see if my underground wire's shield is grounded.

Re: Noise, that monster...

Wed, 15 Oct 2008 01:27:58 EDT

Bicephale : Euh...

On 2nd thoughts, if i were to use a simple
resistor i guess i'd want to insert it on the
"Tip" side in order to use my Tip connector
as a zero Volts reference. Hummm... But it
wouldn't matter using a series transformer
so i'd perfer this later solution a lot better.

Re: Noise, that monster...

Wed, 15 Oct 2008 01:10:51 EDT

Bicephale : Hi TerTech,

It seems i missed the fun! Sorry, i wanted to be
outside for the full-moon/election day conjunction!

;)

So, when you write POTS Splitter you really mean it:

Phone Jack Wiring info?, Bicephale, 2007-Oct-23

I guess yours has D.C. Blocking as shown above...

I'm curious, does it look anything similar to that one:

Phone Jack Wiring info?, Bicephale, 2007-Oct-24

:)

About the interfacing problem, perhaps instead of
trying to connect in parallel it would be OKay to
insert some low value resistor in series with the
"Ring" side and to measure the signal across that
resistor's leads: it would be minute but shunt it
all you want this shouldn't disconnect the MoDem,
on the contrary: it would make the drop lesser...

Even better, if you've got a large ferrite toroïdal
core handy, i'd suggest you just pass the MoDem's
Ring line through it while you feed your measuring
instrument via a "secondary" made up of insulated
wire coupled by a capacitor. Two clamping diodes
connected across the secondary will limit voltage
to a value below one Volt while the capacitor is
taking care of this D.C. path... Do i make sense?

:D

Re: Noise, that monster...

Tue, 14 Oct 2008 22:13:41 EDT

tertech :

said by pnjunction :

Not surprising. If you're looking with a spectrum analyzer, it's going to have 50 or 75 ohm inputs, while phone lines are high-impedance (kilo-ohms) AFAIK. So you're effectively shorting it. (edit: looks like the impedance of the lines is supposed to be 100 ohms, still the modem won't be happy with 50 ohms to ground on there)

An oscilloscope (set to high-impedance) or probe shouldn't have the short problem and they usually have only 5-15+ pF of input capacitance (compared to many nF for the phone line). However, you're not going to anything meaningful of a DMT signal in the time-domain.

Really all you might be able to do is check for interference sources on the line (with the scope or SA) while the DSL isn't connected.

Thanks for verifying my suspicions. For sure the SA is 50 Ohms. I had no clue of the impedance of a phone line after a DSL splitter; I hoped that I might get lucky :uhh:. I guess I'll have to dig up the old books and see how to breadboard an op_amp in a non_inverting impedance matching configuration with 1:1 gain. In the meantime, yeah a scope is useless to view the DSL spectrum, I will use the SA to view the open DSL line with the modem disconnected.

Re: Noise, that monster...

Tue, 14 Oct 2008 21:46:51 EDT

pnjunction :

said by tertech :

Unfortunately, as soon as I connect anything across the DSL modem's input, it instantly drops connection and will not re-sync until I disconnect the equipment. Must be an impedance mismatch. :huh: For the time being, I can still examine the line for stray noise without the modem connected.

Not surprising. If you're looking with a spectrum analyzer, it's going to have 50 or 75 ohm inputs, while phone lines are high-impedance (kilo-ohms) AFAIK. So you're effectively shorting it. (edit: looks like the impedance of the lines is supposed to be 100 ohms, still the modem won't be happy with 50 ohms to ground on there)

An oscilloscope (set to high-impedance) or probe shouldn't have the short problem and they usually have only 5-15+ pF of input capacitance (compared to many nF for the phone line). However, you're not going to anything meaningful of a DMT signal in the time-domain.

Really all you might be able to do is check for interference sources on the line (with the scope or SA) while the DSL isn't connected.

Re: Noise, that monster...

Tue, 14 Oct 2008 21:20:56 EDT

tertech :

said by pnjunction :

Be careful hooking up a spectrum analyzer (especially an expensive one that you are only borrowing!) to your phone line as phone lines carry DC voltage (40-something volts) and many spectrum analyzers DO NOT tolerate DC voltages on the input.

You might be aware of this and plan to use AC-coupling (or your model can tolerate DC, although many only go up to 25V or so), but just in case...

Would I know when the phone was ringing by the puff of smoke? :D

Thanks for the warning. There's a yellow label on the unit... "DC or static discharge may damage input". I took heed.

I'm connected after my DSL splitter, so I imagined that any DC component is already removed. Just to be safe, I tested the line first with a voltmeter; there's no DC. Then I hooked up an oscilloscope just to see the AC levels.

Unfortunately, as soon as I connect anything across the DSL modem's input, it instantly drops connection and will not re-sync until I disconnect the equipment. Must be an impedance mismatch. :huh: For the time being, I can still examine the line for stray noise without the modem connected.

Re: Noise, that monster...

Tue, 14 Oct 2008 20:30:46 EDT

pnjunction : Be careful hooking up a spectrum analyzer (especially an expensive one that you are only borrowing!) to your phone line as phone lines carry DC voltage (40-something volts) and many spectrum analyzers DO NOT tolerate DC voltages on the input.

You might be aware of this and plan to use AC-coupling (or your model can tolerate DC, although many only go up to 25V or so), but just in case...

Re: Noise, that monster...

Tue, 14 Oct 2008 14:26:01 EDT

Bicephale : Well, being shielded & underground would
be an advantage if the power and phone
lines were not traveling parallel to each
other, at close range and over very long
distances (two hundred feet definitively
falls into this category). A grounded shield
will stop electrostatic pickup but i'm not
confident it can protect against magnetic
induction. I'm sorry but the configuration
which your post suggests is more likely to
reinforce the coupling, actually. Lets just
cross fingers in hope the shield is floating
because if it is then there's a cure for it!

:p

You sound like you're observing predictible
events taking place at 16h30 each day but
is that work days, week-end days or both?

You may have a neighbour getting home at
4h30 or it could be someone's timed heater
system, etc. My attention is captured by
this kind of suspense so keep us posted!...

:o

Re: Noise, that monster...

Tue, 14 Oct 2008 12:06:05 EDT

tertech :

said by Bicephale :

Hi TerTech,

C'mmon, spit it out! Tell me how many feet...

:uhh:

How bad it is, actually?... I bet Bell strapped
their flat non-twisted phone cable to that PVC
tubing, with main power electric cables inside,
on a distance starting at the roof and ending
at the NID, right? So, how many feet is that?

Two, four, ten, fifteen? Twenty feet maybe or
is it even worst because there's many stages!

:o

About 200 feet! No, not a huge pole, it's underground. The power transformer is directly across the street from my 200' long driveway; the phone/cable boxes are right next to my driveway on my side of the road. I imagine the builder dug one trench and ran all the wires in the same hole. I don't know if the building code requires the power wires to be separated from other services. At least I know the Natural Gas is in a different trench. :uhh:

The phone wire is the thick shielded grease-filled cable with 3 pairs. It was installed by my building contractor's guy, not Bell. I'll have to check when I get home if they are twisted, and also if the shield is grounded at my NID and/or at Bell's box at the end of my driveway ( it's not locked :D ).

Of course, you're aware that aerial electrical
power lines collect all sorts or electromagnetic
radiation like there is no tomorrow... Right? Is
it possible this physical configuration turns out
to couple the power and phone lines together,
injecting airborn noise into a DSL signal which
may very well happen to be relatively cleaner
otherwise? You considered the "Green Acres"
phone-circuit topology to cure this issue?...

:p

More bugs up there but potentially better signal!

;)

That probably explains why we hear CFRA 580 AM on our phones. That's a constantly present frequency, and should be avoided by the ADSL. I still want an explanation of what happens around 16:30 when my modem says 'hi-de-ho' and jumps right back in the crapper ( like Mr Hankey on South Park ). I have secured two of the junkey old spectrum analyzers at work; one might actually be functional. I can take them home to look at the phone line. If they don't work, I can borrow one of the good expensive models, but I have to take it after 5pm and return it by 9:00am each morning. :huh:

I can't wait to see what this line looks like!

Re: Noise, that monster...

Tue, 14 Oct 2008 11:34:11 EDT

Bicephale : Hi TerTech,

C'mmon, spit it out! Tell me how many feet...

:uhh:

How bad it is, actually?... I bet Bell strapped
their flat non-twisted phone cable to that PVC
tubing, with main power electric cables inside,
on a distance starting at the roof and ending
at the NID, right? So, how many feet is that?

Two, four, ten, fifteen? Twenty feet maybe or
is it even worst because there's many stages!

:o

Of course, you're aware that aerial electrical
power lines collect all sorts or electromagnetic
radiation like there is no tomorrow... Right? Is
it possible this physical configuration turns out
to couple the power and phone lines together,
injecting airborn noise into a DSL signal which
may very well happen to be relatively cleaner
otherwise? You considered the "Green Acres"
phone-circuit topology to cure this issue?...

:p

More bugs up there but potentially better signal!

;)

Re: Noise, that monster...

Tue, 14 Oct 2008 10:03:06 EDT

tertech :

said by Bicephale :

Oh my! Oh my! Don't tell me: i've got
a tip what may be next on the menu!

...and it's not about the little bug on
the upper-left corner of Bell's NID...

hahahahaha you found my 'bug'. I noticed that little guy when I was downscaling the photos from the high res. You have a good eye!

Re: Noise, that monster...

Tue, 14 Oct 2008 02:45:52 EDT

Bicephale : Oh my! Oh my! Don't tell me: i've got
a tip what may be next on the menu!

wiring from NID, TerTech, 2008-Oct-13

...and it's not about the little bug on
the upper-left corner of Bell's NID...

Re: Noise, that monster...

Sat, 11 Oct 2008 15:42:57 EDT

Bicephale : Hi TerTech,

I must confess i have no expertise as a matrimonial
advisor; professionnals are in a better position to
help in this matter and they will save you money in
the long run!!! It seems possible to leave your ST
alone while it does its things automatically with a
bit of help from 'DMT' or even 'OrbMT', though. In
'DMT' go to the "automatic DSL-Monitoring" section,
its "DMT Automat" tool is accessible by clicking on
the [Configuration] button; under 'OrbMT' you have
the "Actions" section on the "Settings" tab, if all
it takes is a way to resync/reboot a SpeedTouch 5x6
then i bet you can avoid a costly divorce using one
of these programs. Resolution is as good as 1 min.

:D

Most unfortunately, i can't really assist much with
cutting your reading time. I exposed the problems/
solutions as it occured using graphic illustrations
when possible. This was all i could do to save the
reader from going through all of the frustration...

I'm sure you will be able to collect some long-term
(24 h) SNR Margin curves using 'RouterStats-Light',
no matter what happens to your connection. But, if
you suspect that the problem is noise, then it's not
wise to try to monitor it by watching side-effects:
just use 'DMT' in "Diagnosis" mode as i suggested!

:p

A spectrum analyser would probably have proven that
my downstairs neighbour's dimmer is radiating 60 Hz
nuisible harmonics all over the DSL spectrum, which
is why i could use multiple A.M. radio frequencies.

Lets hope the noise source is similar in your case:
auto-transformer dimmers are the perfect substitute
for noisy ones, for example, but it's another thing
to deal with street lamps if that's your problem...

It all depends what type of noise source you got to
work with and where it's located. Sorry but even a
"clean" DSL signal would be ruined by a few feet of
telephone wire strapped along a power line enclosed
in PVC tubing. This is what happened to me while i
lived in Louiseville and i'm pretty sure i'm facing
the very same issue again. I've demonstrated how a
para-magnetic/electrostatic shield could reduce the
level of noise before, the graphic from my previous
post should speak for itself... Pray it's not more
difficult than this: there's hope if you're lucky!

;)

Re: Noise, that monster...

Sat, 11 Oct 2008 10:33:27 EDT

tertech : Bicephale,

Thanks again, you've been very helpful. I'll try to re-read all those posts that you linked, but I think the wife is getting a little annoyed I am spending all day at the computer :uhh:

Yes, it's like a switch being thrown. It is obvious when the noise turns on, but not so obvious when it turns off. Once I set the modem to a target SNRM of 15dB or more, it will keep a stable connection in spite of the noise. When the noise does turn off, the modem will not automatically re-sync. I can't stay awake 24 hours a day clicking the re-sync button, so I can only see like this morning at 08:00 when I did a manual re-sync that I'm back to a 3.5dB SNRM with a stable 2.496mBps connection.

The ability of 'Mr Hankey' to stay locked-in with such a poor noise margin is amazing; which makes it all the more frustrating when it goes into the horrible performance mode.

My plan now....

Just leave the modem set to automatic SNRM; no monkeying with the DMT sliders. Keep RouterStats and DMT diagnostics running ( I set up another machine for this ). This will probably mean that during this test I can't use the internet when the connections drops every few seconds. :huh:

On Tuesday, I'll borrow a spectrum analyzer from work and hook it up somehow to my phone line without creating a new problem. I'll take measurements scanning the band from 20kHz to 1000kHz with the scope set to mark/hold peak signal level detected across this band. I will take measurements when the modem is in both the stable and unstable modes. If I see a section of the spectrum where there is a large signal during the 'bad time' that is not present during the 'good time', I will at least have a better idea of the type of interference I'm dealing with. I don't suspect that I will find a narrow offending frequency like a radio station; I expect to see a spread spectrum of noise. If it were a very narrow noise band, wouldn't the ADSL simply not use those affected tones?

Finally, after all this effort, if I find a noise that comes on my line at a certain time of day, what could I do about it? I don't believe that this noise is originating from within my house, or even that it is entering my wires inside my house; I think it's occurring on the wires between me and the DSLAM. The neighbors I talked to say they all have Rogers cable internet, so I don't know anybody else with DSL in my neighborhood to ask if they see similar problems. I have already installed an outdoor Voice/DSL splitter next to the NID. I have new CAT-5 outdoor-grade cable from there to the closest point inside the house where the modem is situated. Will my shielding a few feet of cable make a difference in a 5+ Km wire run?

Thanks for listening to me rant and complain, now I better go out and rake some leaves if I don't want a divorce :(

Noise, that monster...

Fri, 10 Oct 2008 23:47:02 EDT

Bicephale : See what i mean?

This nuisance is noise!

Oups! It was "The Beast", actually...

Well, you get the idea!

:o

VelCom - The Monster Walks Away! (Sep. 2008)

Re: If I were paranoid......

Fri, 10 Oct 2008 23:07:36 EDT

Bicephale : Hi TerTech,

Those graphs feel very familiar to me: it's
like flipping a switch, wouldn't you say?...

My Siemens SpeedStream 4200 is On-Line
since the begining of October, i had weird
noise curves during the 1st couple days (it
was even resetting by itself on occasions),
then my daily CRC Error counts (recorded at
0 h) seemed to improve quite significantly:

October 6: 1439 CRC errors in 86095 seconds October 7: 2171 CRC errors in 86005 seconds October 8: 4657 CRC errors in 86013 seconds October 9: 1449 CRC errors in 86011 seconds

Despite all my efforts to null out marginal
events i still measure a 224 % noise raise
if i compare the best and worst days - and
no, i'm not paranoïd, well not yet! My own
attenuation is 45 dB while yours is off-scale
so if strong variations can hit me then lets
imagine the issues which you're exposed to!

:p

Now that you can see it, you may want to
hear it as well. If there's a daily pattern it's
possible to catch the transition in audible
form over a simple A.M. radio: i've done it.

It could provide you the means to pin-point
the noise source or at least it may indicate
in which direction to move the MoDem. Once
the problem is known it's easier to deal with
it, shielding worked for me when i was facing
strong local noise, before i went to Limoilou:

Ref.:

The customer's own wiring, Bicephale, 2007-Jun-13
Tweaks, Bicephale, 2007-Aug-27
A wire is a wire is a wire, Bicephale, 2007-Sep-6

It was about telephone wire cleaning, this is
more closely related to what you just shown:

From the ground up!, Bicephale, 2008-Jul-9

Perhaps Mr Hankey isn't designating a MoDem
at all, i bet it's like what i called "the monster":

From the ground up!, Bicephale, 2008-Sep-21

;)

If I were paranoid......

Fri, 10 Oct 2008 21:35:07 EDT

tertech : Today at just about 4:30pm, my modem suddenly contracted the frequent disconnection sickness (FDS). It had been running smoothly at 2.496mBps/3dB for a few hours prior.

Hmmmmmm what happens around 4:30 every day? Tea-time break? A new heinous throttling method? No, No I mustn't think that. That's just crazy talk! Like Bicephale says... don't base any conclusions on a single occurrence. ;)

To keep the connection alive, I had to intervene and adjust the SNRM target up until I finally got a stable connection at around 1mBps/13dB.

So it's back into the crapper once more until it magically pops back out. New nickname for my modem.... Mr Hankey. :D

Re: Arrrghhhhhhh!

Fri, 10 Oct 2008 13:39:21 EDT

Bicephale : Hi TerTech,

I see. I often advise against only using
static numbers for the reason that one
might risk to fall on some glitch. Multiple
measurements, at least, should help to
reject marginal data but i prefer curves.

About FirmWare v6.2, no MoDemOption
support there! The very last version to
work with the 'DMT' tweak feature was
v6.1.9.6 which is Thomson's 2nd 'DMT'-
hostile experimental release, apparently.

'DMT' tweaking stops at v6.1.0.5 for all
practical purposes.

Re: Arrrghhhhhhh!

Fri, 10 Oct 2008 12:25:21 EDT

tertech :

said by Shikomu :

Yes as a matter of fact I had the SNR up at around 9.5db with the sliders speed was lower than 2496 and yet randomly it would resync. I tried many positions of the slider TX full power min power you name it. Apply save and reboot at least 40 times.

Ditto for me. I can set the Remote Tx Pwr slider to default, min, max, or anything in between. I apply/resync and save settings after each change. I see that the REGS 0D & 0E contain new values. None of the settings appear to make any difference in any of the numbers I get after re-sync. It was just a fluke that it appeared to do something the first time I tried it last week.

Miro: version 6.2.29.2 doesn't have the ability to adjust the SNRM with modemoptions? I'll add it to my list of versions to try over the long weekend when I'm not stuffing myself with turkey. :D

Have a great Thanksgiving week-end.

Re: Arrrghhhhhhh!

Fri, 10 Oct 2008 07:18:53 EDT

Shikomu : Yes as a matter of fact I had the SNR up at around 9.5db with the sliders speed was lower than 2496 and yet randomly it would resync. I tried many positions of the slider TX full power min power you name it. Apply save and reboot at least 40 times.

Re: Arrrghhhhhhh!

Thu, 09 Oct 2008 23:54:28 EDT

Bicephale : Thanks TerTech for the fancy interlude,
Thanks Miro for your late contribution!!!

Euh... You did apply, save and reboot
between each move of the slider, right?

Re: Arrrghhhhhhh!

Thu, 09 Oct 2008 22:59:29 EDT

Shikomu : Well after days of testing every 7.x firmware and 6.1.0.5 I found the one that is most stable for my stats which are very close to yours. 6.2.29.2. It holds the SNR between 6-8db but does not resync like the others. They all resync at least 1 time every 24 Hours yet this firmware is over 3 days. It wins for me.

Arrrghhhhhhh!

Thu, 09 Oct 2008 20:47:51 EDT

tertech : Since I have been monitoring for the past two days with RouterStats-Lite and DMT Diagnostic, my modem has been behaving perfectly. It's back up in the 2.3/2.4 mBps speeds with SNRM in the single digits. The SNRM may vary .5dB up or down periodically, but for the most part the RouterStats graph is a straight line. Even though it rained overnight, a certain connection killer for me in the past, the modem held rock-steady.

I'm starting to believe this is an example of the Heisenberg Principle where the act of just measuring something affects that measurement. :uhh:

At this point, I throw up my hands. I don't really know if my adjusting the Remote Tx Power setting actually had any effect. Now I can move the sliders up and down and it doesn't make the slightest difference in the connection statistics. Maybe it was just a coincidence that on the same day I first made those adjustments, there were repairs done in the Remote CO. I'm ready to believe in the DSL Fairy; it's just as plausible as any other explanation I can think of why the characteristics of my line have suddenly changed so dramatically.

Re: Too good to last!

Tue, 07 Oct 2008 14:59:37 EDT

Bicephale : Hi TerTech,

Now that i think of it, didn't you start this
thread after finding that you had control
over the remote power level? I still had it
in mind when i refered you to a section of
this very old post (2007-Jul-24), after all:

5056/800 Kbps
       DownStream Power: 19.59 dBm
         UpStream Power: 10.95 dBm
DownStream Attenuation: 32.0 dB
   UpStream Attenuation: 30.0 dB
DownStream SNR Margin: 18.0 dB
    UpStream SNR Margin:   7.0 dB

5056/288 Kbps
       DownStream Power: 19.44 dBm
         UpStream Power:   6.36 dBm
DownStream Attenuation: 32.0 dB
   UpStream Attenuation: 30.0 dB
DownStream SNR Margin: 20.0 dB
    UpStream SNR Margin:   6.0 dB

Note how the UpStream Speed and Power
level appeared to be linked in this example.

There was a 2 dB gain on the DownStream
SNR Margin side, do you see it?

:)

Won't this help to improve the DownStream
parameters if you lower the UpStream Power
level?... It's hard to tell where your Bottle-
Neck really is with that UpStream SNR Margin
stuck at 6.0 dB exactly; 'RouterStats-Lite'
is nice for gathering SNR Margin information
but you won't get a clue about the UpStream
signal with such a steady SNR Margin anyway
as i'll bet it's because of a bug in the Texas
Instruments (MoDem peer) FirmWare, actually.

You'd get long-term noise curves using 'DMT'
in its "Diagnosis" mode, it's only a matter of
choosing options which won't generate a file
that is so large no SpreadSheet can read it
(unselect the "Bitalloc" item, for example):

STMT crashing, Bicephale, 2008-Sep-17

In your case, e.g. for an Interleaved profile,
the "txFEC"/"rxFEC" items are needed, though.

;)

Most unfortunately, 'STMT' needs compatible
FirmWare which won't support MoDemOption
features while displaying noise curves; 'OrbMT'
will always ignore error rates (by design) so
this tool isn't an option neither... As for the
'RouterStats-Lite' log file, it just wasn't being
created the last time i tried - leaving 'DMT' in
"Diagnosis" mode when it comes to long-term
(24 h) noise curves. The later application goes
to "Diagnosis" mode in an exclusive fashion and
there can't be two instances of 'DMT' running
on the same PC so you'll have to run 'DMT' on
a second computer if you must retain tweaking
abilities while a long-term record is being made.

:p

Correction:

Power levels must be expressed in "dBm" units...

Re: Too good to last!

Tue, 07 Oct 2008 12:59:29 EDT

tertech :

said by Bicephale :

Hi TerTech,

Sorry about your misfortune. Maybe it's about time for
long-term curves in order to get a better idea what's
going on! You might also want to consider asking for a
profile with some lower UpStream speed, something like
2496/388 Kbps Interleaved if this is possible at all: that
would possibly help on the DownStream side somewhat.

Thanks,

I just started RouterStats-Lite[v3.5]; I selected logging and capture graphs. Once I have some data, I'll contact Teksavvy and suggest the lower upstream profile. Problem is if that doesn't help; then I'm stuck with it. I may still consider the 512/512 service - especially if CRTC rules for Bell, and Conservatives get elected :uhh:

Re: Too good to last!

Tue, 07 Oct 2008 12:12:39 EDT

Bicephale : Hi TerTech,

Sorry about your misfortune. Maybe it's about time for
long-term curves in order to get a better idea what's
going on! You might also want to consider asking for a
profile with some lower UpStream speed, something like
2496/388 Kbps Interleaved if this is possible at all: that
would possibly help on the DownStream side somewhat.

Too good to last!

Tue, 07 Oct 2008 08:34:39 EDT

tertech : Well, Last night before my eyes I watched my connection circle the bowl then head right down the ol' crapper.

This AM, I'm right back to where I was for the last several months; that is - needing high SNRM numbers just to get a stable 1Mbps connection.

Oh well, it was nice while it lasted. :huh:

Re: Curiouser and Curiouser

Mon, 06 Oct 2008 18:07:26 EDT

Shikomu : Ive been doing tests also and would get some odd resyncs but this week Teksavvy is doing upgrades so I will wait until they are done so if mine resyncs its not from upgrades.

Re: Curiouser and Curiouser

Mon, 06 Oct 2008 13:33:53 EDT

Bicephale : Hi TerTech,

As you wish, those are your shoes.

:p

I'll sure read you with curiousity when
you catch an invisible hand in the act!

Curiouser and Curiouser

Mon, 06 Oct 2008 12:56:52 EDT

tertech : This morning at 9:00 I noticed that a single re-sync had occurred overnight at an hour or so earlier. The really amazing thing was that the SNRM was as low as 3.5dB, and it held over an hour! The SNRM gradually rose to as high as 7.5dB over the next hour and then it re-synced to a slightly lower speed.

06.10.2008 08:59:53 - C1L1D1T00 | BitrateDN: 2496 kbit/s | BitrateUP: 576 kbit/s | SNRM: 3.5 dB
06.10.2008 09:09:54 - C1L1D1T01 | BitrateDN: 2496 kbit/s | BitrateUP: 576 kbit/s | SNRM: 4.0 dB
06.10.2008 09:19:54 - C1L1D1T01 | BitrateDN: 2496 kbit/s | BitrateUP: 576 kbit/s | SNRM: 7.5 dB
06.10.2008 09:29:54 - C1L1D1T01 | BitrateDN: 2496 kbit/s | BitrateUP: 576 kbit/s | SNRM: 7.5 dB
06.10.2008 09:39:54 - C1L1D1T01 | BitrateDN: 2496 kbit/s | BitrateUP: 576 kbit/s | SNRM: 7.0 dB
06.10.2008 09:49:54 - C1L1D1T01 | BitrateDN: 2496 kbit/s | BitrateUP: 576 kbit/s | SNRM: 7.0 dB
06.10.2008 09:59:54 - C1L1D1T01 | BitrateDN: 2464 kbit/s | BitrateUP: 544 kbit/s | SNRM: 8.0 dB
06.10.2008 10:09:54 - C1L1D1T01 | BitrateDN: 2464 kbit/s | BitrateUP: 544 kbit/s | SNRM: 8.0 dB
06.10.2008 10:19:54 - C1L1D1T01 | BitrateDN: 2464 kbit/s | BitrateUP: 544 kbit/s | SNRM: 8.0 dB
06.10.2008 10:29:54 - C1L1D1T01 | BitrateDN: 2464 kbit/s | BitrateUP: 544 kbit/s | SNRM: 7.5 dB

Now, I am totally baffled as to what is going on. I'm beginning to doubt that my tweaks had anything to do with it; maybe it was just coincidence that someone fixed the DSLAM at around the same time I decided to try the remote tx pwr setting. :uhh:

I appreciate all the suggestions expressed here. All I want out of my modem is a stable connection that doesn't drop the line every few minutes. I was even happy to lower the speed to around 1mBps if it meant I got a stable link. Dropping down to 700kBps was a bit too much sacrifice. One or two re-syncs a day doesn't bother me, and having a speed over 2mBps is just sweet icing on the cake. If things stay this way; I'll try my old 2Wire modem and also load the v7 firmware into the ST516.

I'll post back if anything else bizarre happens. Thanks again, all :D

Re: ST516 - Strange discovery

Sun, 05 Oct 2008 23:38:23 EDT

Bicephale : Hi,

What about noise? This capture doesn't tell when
the 10th reset took place nor what your counter
values were right after that last reset. As for the
mysterious remote transmission power level, i've
suggested that less UpStream power may lead to
some noise reduction; that's why i'd keep an eye
on the RCO and Error rate if i were in your shoes.

:)

Re: ST516 - Strange discovery

Sun, 05 Oct 2008 15:56:23 EDT

tertech :

said by Bicephale :

Hi TerTech,

I noticed that your UpStream SNR Margin doesn't move much:

Does it ever change to other values?

It does stay around the 6dB point; the bit-rate varies up and down though. I'm not as concerned with the upstream as I am with the downstream so I haven't paid as much attention to it.

My stats from Sep 21 that you posted really illustrate my point. At that time I had the Target Noise Slider up near the top and got an SNRM of 15.5dB but only a speed of 1312kbps. The tx pwr was 16.5dBm. Today I am at the limit of my profile (2496kbps) with an SNRM of ~8dB and the tx pwr is also 16.5dBm. It has been running 15.5 hours so far without re-sync. The only thing I did on Friday was to adjust the tx pwr setting in Reg 0D & 0E to "max" instead of "default".

This is why I titled my original post "strange". I can't explain it, unless someone at the Remote repaired my DSLAM at just around the time I started my ~~monkeying~~ empirical testing.

Re: ST516 - Strange discovery

Sun, 05 Oct 2008 11:59:05 EDT

Bicephale : Hi TerTech,

I noticed that your UpStream SNR Margin doesn't move much:

SNR drops at Night (This may help), TerTech, 2008-Sep-21

Does it ever change to other values?

Re: ST516 - Strange discovery

Sun, 05 Oct 2008 11:49:45 EDT

Bicephale : Hi JFMezei,

My reply of October 4 to TerTech (which refers to links
of September 27) lead to Bell portal captures where the
MoDem's peer didn't stick to some preset profile... The
ADSL standard is supposed to Fall-Back to a lower set
of speeds and then to recover as required but this just
won't happen when using a Thomson ST5x6, a Siemens
SpeedStream 4200 and i wouldn't even want to bet on
GNet's BB0060B and/or GBB2060-Xi for that. We're lucky
when a DSL unit provides some form of Spectral Shaping
and a lower UpStream rate would be most welcomed in
the present case, i believe.

Re: ST516 - Strange discovery

Sun, 05 Oct 2008 11:10:29 EDT

tertech : 10.5 hours at my max d/l profile speed of 2496 with no re-sync disconnects. :D :D :D

Now I'll leave it alone all day to test the effects of the diurnal variances such as cosmic rays, solar flares, moon gravity, whatever. The SNRM was 9.0dB last night when I finally went to sleep; this morning after sunrise it's down to 7.5dB but still hanging on to sync. I have the Tx Pwr set to MAX and it's showing 16.5dBm.

I dare not click that [DSL Resync] button for fear that my speed may come crashing down.

Re: ST516 - Strange discovery

Sun, 05 Oct 2008 06:04:59 EDT

jfmezei : Bicephale, I was not suggesting hacking to raise one's profile.

With MNP modems, they would automatically retrain down to a lower speed when conditions were bad, and then retrain back up when conditions imporved, and you could set a ceiling for negotiations.

For instance, when I was in australia, I would connect back to my system via long distance dial-up with an accoustic coupler on a public payphone. I found it was faster to limit my tiny modem to 2400 baud than to let it negotiate higher speeds and spend time retraining down and up again.

From Rottness Island On the indian Ocean near perth (about as far as one could get from Montreal on the planet), I was able to get a usable 14.4kbps connection over a payphone. But from other places, 2400 was barely usable.

Now, back to DSL: does the client modem really have the ability to tell the DSLAM to use a lower speed than is provisioned ?

Or is the signal still saying 5056000 bps but because of blocked frequencies, the actual throughput is less ? Or does it actually negotiate a lower speed ?

And can the modem do this dynamically as conditions change, (like MNP modems) or once synched, it remains at those settings no matter what ?

Re: ST516 - Strange discovery

Sun, 05 Oct 2008 01:26:39 EDT

Bicephale : Hi JFMezei,

With DialUp devices you could define the negociated
speed directly, with DSL MoDems you might find that
what is done actually refers to a form of Spectral
Shaping where the customer has some control over a
few parameters such as the Bits-per-Bin setting and
the Coding Gain which are available to GNet owners
only and maybe some others but not that's not to be
found in a Thomson ST5x6: i would have noticed!!!

Better Spectral Shaping control is allowed with the
3Com USR-910x products but i've never been able to
see one single capture which could seem conclusive.

:(

If we were able to raise our profile ourselves this
would be called hacking, actually, and it's not open
for discussion - which doesn't matter as it doesn't
work like that even if you try very hard. All we do is
trade-off (sacrifice) some speed for more stability
instead of having Bell involved in a gross way...

;)

Addendum:

"De-Luxe" Spectral-Shaping:

SpeedTouch 516 Tweaking :), Bicephale, 2008-May-8

GNet Tweaking:

Need help/tips about DSL connection speed, Bicephale, 2008-May-25

The Trade-Off:

Stats help from someone at Teksavvy, Bicephale, 2007-Jul-24

:D

Re: ST516 - Strange discovery

Sun, 05 Oct 2008 00:52:23 EDT

Bicephale : Hi TerTech,

Thanks for your motivating comment, as an ordinary
DSL customer i do little more than to simply use a
set of utilities which are at everyone's disposal.

Well, i guess it's true that some individuals might
risk to regress after a while when doing this but i
find "empirical testing" much more suitable myself!

:D

About the difference between copper vs empty space,
i'm thinking about parasitics right away but maybe
you just refer to the "monkeying" work required...

;)

It didn't occur to me i couldn't define MoDemOption
register #0Bh (and beyond) until you mentioned this
detail to me. Possibly, that's only a result of the
character set used by 'DMT' but there's a chance it
is why i couldn't set the Remote Tx Power level (?).

This reminds me of a link i gave to KPaul anyway:

»Re: SpeedTouch 516 Tweaking :)

I wonder if you wouldn't benefit from exploring the
matter since this post dates back to mid-2005: i'd
bet the author owned similar HardWare to your own.

:)

In any case, having no resets sure is a good thing;
oh, and PLEASE keep posting as i'm finally learning
from the late new input! My situation was different
if we consider that i had lots of noise while you're
having lots of attenuation... This could be another
reason why i went nowhere trying to adjust my Remote
Tx Power level, i don't know, but i wouldn't be too
surprized if your "sweet spot" happens to gravitate
around an SNR Margin of 12 dB. Look at my capture:

'RouterStats-Lite v3.5' as a Trouble-Shooting Tool, Bicephale, 2008-Sep-14

The part marked "Disruption" was when my downstairs
neighbour walked into her kitchen and lighted up a
noisy dimmer (so it seems); you can see my ST546v6
trying to synchronize again but to no end. Then, i
used a stable tweak for a while and started looking
for a sweet spot before that noisy event ended (the
constant-slope sections). It's clear on my high-res
version of this capture that the MoDem reconnected
past 10.5 dB - but it was still resetting frequently.

:(

13 dB was conservative, 12 dB would have been edgy.

This other capture clearly illustrates how my ST546
once just failed to gracefully "Fall-Back" far enough:

'RouterStats-Lite v3.5' as a Trouble-Shooting Tool, Bicephale, 2008-Sep-12

It continuously targetted around 4 Mbps/9 dB instead!

:uhh:

I keep my fingers crossed wishing that you can have
a stable connection well below the 12 dB mark but i
am sceptical. By the way, i explicitely associated
my last capture to FirmWare v7.4.4.7 so i feel this
one may be fine only until it hits the 12 dB barrier in
presence of strong noise...

:p

Re: ST516 - Strange discovery

Sat, 04 Oct 2008 21:49:08 EDT

tertech :

said by Shikomu :

tertech:

Mine is 17DB yours is 14 DB but I set the PWR to Max or MIN and it had no affect so I'm not sure yet. I'm also tweaking. See your bottom graph (Loop Length) mine is a bit longer than yours. When you lower your SNR and increase your speed does your TX Power go up to 17 DB like mine ?

I've just gone to 'Max Power' too ( wasn't that a Simpson's character? :) ). I'll start taking tiny steps in lowering the SNRM then waiting several hours with no re-syncs before trying the next step. I just went to 2080kBps @ 12dB. I do notice the tx pwr has gone up to 16.0dBm; we'll see if and when I get to 2208kBps like you if the tx pwr goes to 17dBm.
What happens to you if you put the Remote Tx Power settings back to 'default'? I instantly go back to crappy connection.

This is all very encouraging. Over the last week or so Ive torrented a lot ( of Linux distros, ahemmm ). It was pathetic at 700kBps even with s/l MLPPP.

Re: ST516 - Strange discovery

Sat, 04 Oct 2008 21:25:48 EDT

Shikomu : tertech:

Our lines are very similiar. Ive been testing the 7.4.4.7 for the last week and its very good it pushes your SNR up almost 2 DB. What I didn't like is if there was some DSL noise out there at night and errored sec. started it would re-sync. It was annoying. It would resync every evening at least once. So I'm also playing with open firmware 6.1.0.5. I have noticed the TX power in DB goes lower or higher depending on the speed you are syncing when you change your SNR. Mine is 17DB yours is 14 DB but I set the PWR to Max or MIN and it had no affect so I'm not sure yet. I'm also tweaking. See your bottom graph (Loop Length) mine is a bit longer than yours. When you lower your SNR and increase your speed does your TX Power go up to 17 DB like mine ?

Re: ST516 - Strange discovery

Sat, 04 Oct 2008 21:05:51 EDT

tertech :

said by TilhasBB :

I would recommend you try different firmwares. Some of them are already tweaked nicely.

I would try the following ones.
7.4.4.7
5.4.0.14
Or the 6.2 UK Firmware

I tried v7.417 because it was supposed to be so much better on marginal lines. It did the same as the 2Wire; it connected at the max profile speed. If I remember correctly, the SNRM was up around 8 or 9dB. Not that it mattered very much, it kept disconnecting and re-syncing every few minutes. The DMT ability to adjust the SNRM is not in that version; I went back to 6.1. The modem came with v5.4; I upgraded it to the current 6.1 at some point. I really didn't see any big difference in stability from v5 to v6. I haven't tried the 6.2 UK firmware.

I'll finish my little experiment with the Remote Tx Pwr first. and then start trying the other firmwares you recommended. Thanks for the suggestion.

As of now... 6 hours at 1664kBps 12.5dB SNRM without disconnect. Time to push the SNRM a little lower :D

Re: ST516 - Strange discovery

Sat, 04 Oct 2008 19:42:45 EDT

TilhasBB : I would recommend you try different firmwares. Some of them are already tweaked nicely.

I would try the following ones.
7.4.4.7
5.4.0.14
Or the 6.2 UK Firmware

In that order :)

Bicephale has a nice post on where to find them
»Thomson SpeedTouch 5x6v6 FirmWare v6.2.H (v6.2.17.5)

Re: ST516 - Strange discovery

Sat, 04 Oct 2008 18:54:34 EDT

tertech :

said by jfmezei :

I have a question for which I never got an answer:

With dial-up modems, the connection speed was negotiated between the 2 sides. (and then the various MNP capabilities.

With ADSL, I was under the impression that the connection speed was dictated by the DSLAM and that end user's modem didn't negotiate the speed.

Many here have pointed to the speedtouch having some magical ability to lower the negotiated speed to get a better SNR, which doesn't seem to be present in any other modem.

Does the speedtouch truly tell the DSLAM to not go higher than X speed ?

Or is the actual ADSL speed still dictated by the DSLAM, and the speedtouch simply adjusts the throughput to a lower value, hoping that this will result in fewer errors (aka: time gaps between packets instead of having packets one after the other) ?

I can't say with any authority how an ADSL connection is negotiated. I believe the DSLAM profile determines the maximum speed. The other modems that I have tried ( a 2Wire and a SpeedStream ) seemed to connect at that max profile speed; even if it resulted in a 6dB SNRM. Apparently that 6dB figure is the minimum acceptable number for a connection. I don't know if the 2Wire or SpeedStream would negotiate a slower speed if that 6dB margin could not be obtained. I never saw it happen in the short time I used those modems. They just kept dropping sync over and over.

I don't know if the ST516 has any magic in it. I do know that the ability to set a target SNRM was removed in later firmware versions; somebody at SGT didn't think it was a good feature.

As I set the target SNRM lower and lower, I see an increase in the connection speed and an accompanying increase in the frequency range. I guess this means that the ST516 is actually changing the negotiated modulation frequency range rather than some trick of spacing out the data stream. We have a radio station in Ottawa CFRA that broadcasts at 580kHz: sometimes I can actually hear it on my telephone handset. That frequency stomps right on the middle of the ADSL band. If I do go higher than that I see a notch at 580kHz.

Re: ST516 - Strange discovery

Sat, 04 Oct 2008 17:53:51 EDT

jfmezei : I have a question for which I never got an answer:

With dial-up modems, the connection speed was negotiated between the 2 sides. (and then the various MNP capabilities.

With ADSL, I was under the impression that the connection speed was dictated by the DSLAM and that end user's modem didn't negotiate the speed.

Many here have pointed to the speedtouch having some magical ability to lower the negotiated speed to get a better SNR, which doesn't seem to be present in any other modem.

Does the speedtouch truly tell the DSLAM to not go higher than X speed ?

Or is the actual ADSL speed still dictated by the DSLAM, and the speedtouch simply adjusts the throughput to a lower value, hoping that this will result in fewer errors (aka: time gaps between packets instead of having packets one after the other) ?

Re: ST516 - Strange discovery

Sat, 04 Oct 2008 17:44:00 EDT

Scoop : Yes, the 516 is a good modem. You can get it elsewhere for less money.

Re: ST516 - Strange discovery

Sat, 04 Oct 2008 17:07:52 EDT

Admiral Mike : Quick question. I'm switching to TS when my Bell cancellation date nears, is the ST516 a good modem to get? I would get it through TS as well.

Re: ST516 - Strange discovery

Sat, 04 Oct 2008 13:28:36 EDT

tertech : Hi Bicephale,

I don't believe there's anyone here who has done more meticulous research into the modem behavior than you. Thank you for sharing all that fine work. :)

I say that I'm 'monkeying', but actually my current job is in satellite communication - so I understand the basic principles. What's the difference between 5Km of copper and 36000Km of space?

I'm using ST516 v6.1.0.5.0 and DMT v7.33 (2007.07.28). I notice on your screenshot of DMT, your modem options only go to Reg 0A; you see on mine that they go to REG 0F. The remote Tx Pwr setting appears to change Reg 0D & 0E.

So far, with the setting shown on my screenshots I've been running for 2 hours without a re-sync disconnect.

Re: ST516 - Strange discovery

Sat, 04 Oct 2008 12:41:18 EDT

Bicephale : Hi TerTech,

I published a .GIF animation showing how the
"Noisemargin" gliding cursor has worked for me:

From the ground up!, Bicephale, 2008-Aug-26

Such an experiment just wasn't possible when i
tried to move the cursor at my former location
(it only produced random results) but here, in
Quebec city, its behaviour changed radically:
i can now observe a smooth progression between
the SNRM vs RCO (Connection Speed) poles... It
would seem a better RCO obtained at the cost of
a lower Connection Speed could lead to a better
Final "browser" Speed, in addition to improved
noise/latency figures. I'd call this finding
the Sweet Spot, others may prefer "Monkeying":

From the ground up!, Bicephale, 2008-Sep-27

From the ground up!, Bicephale, 2008-Sep-27

The remote transmission power level setting did
not work for me but perhaps the conditions are
not right. In your case anyway, i'll bet it's
contributing to some noise reduction which may
translate into a lower DownStream RCO figure...

My advice is to keep an eye on the RCO and Error
rate measurements while you experiment further!

;)

Can you mention your exact FirmWare version please?

Re: ST516 - Strange discovery

Sat, 04 Oct 2008 11:13:18 EDT

tertech :

said by justsomeguy :

Instead of monkeying around like this I would suggest
opening a ticket in the teksavvy DIRECT area and asking
if there is a remote you can be put on or if they can
have a tech look at your line.

I wouldnt put up with what you are having to put up with and maybe I would go as far as looking into cable if nothing could be done.

Instead of 'monkeying' if I called it 'empirical testing' would that be better? :D

I've gone through a few tickets. A Bell tech comes to my house with a voltmeter and says the line is fine. When they did bring a DSL line tester, all that they could suggest was a lower profile. My suspicion is that there might be a different solution with fine tuning my modem.

I'm on the closest remote that's less than a Km from me in a straight line - but my wires go over 5Km to get there. :huh:

I used to have Roger's cable TV & Internet. I switched to StarChoice and Teksavvy because I was sick of being reamed. Going back to cable would be my last resort, right behind dial-up.

Re: ST516 - Strange discovery

Sat, 04 Oct 2008 11:03:07 EDT

tertech :

said by GearHead360 :

I have a ST516 & ust DMT. How are you able to adjust your SNRM? The option isn't there other than trying to resync.

Under [Special][Misc. Options/Settings] is a checkbox "unlock experimental area". If this is gray, then your firmware doesn't support this feature. At some point in the SGT release cycle they thought it was a bad idea to empower their customers and removed this feature. You could downgrade to an earlier release, but unless you're having line difficulties there's no reason for you to do this.

Re: ST516 - Strange discovery

Sat, 04 Oct 2008 10:52:53 EDT

justsomeguy : Instead of monkeying around like this I would suggest
opening a ticket in the teksavvy DIRECT area and asking
if there is a remote you can be put on or if they can
have a tech look at your line.

I wouldnt put up with what you are having to put up with and maybe I would go as far as looking into cable if nothing could be done.

Re: ST516 - Strange discovery

Sat, 04 Oct 2008 10:50:24 EDT

GearHead360 : I have a ST516 & ust DMT. How are you able to adjust your SNRM? The option isn't there other than trying to resync.

ST516 - Strange discovery

Sat, 04 Oct 2008 10:40:43 EDT

tertech : For about the last year or so, I have been using the DMT tool to constantly adjust the target SNRM of my connection. With my attenuation of 63.5dB, I'm on a d2496/u640 Interleaved profile; but I can rarely keep sync active for more than a few minutes if I set an SNRM below about 15dB. This leaves me with a dl speed of around 850Kbps. Lately it seems to have gotten even more finicky and frequently losing sync unless I set the SNRM for the max of around 20dB with a resulting dl speed in the 700's. Even then it might re-sync every few hours.

After reading the other post here about how Teksavyy was reviewing their customer's line stats to uncover poor connections, I decided to do some of my own tests and log the results so that I would be prepared with a lot of stats to help with determining if my line should be serviced or my profile lowered. I even considered the downgrade to 512/512.

Anyway, in desperation last night I applied an engineering process of which the technical term is called 'monkeying'. I found a setting in the 'moptions2' tab of DMT to set the remote Tx power. Having absolutely no idea what that means (aside from the obvious), I moved the sliders from 'default' to about 3/4 the way between min and max. There are 2 sliders; both say 'REMOTE TX PWR'. One affects REG 0D and the other REG 0E. With that 3/4 setting, I also set the SNRM to around 12dB and synced at about 1.2Mbps. This morning I checked the log and found not a single re-sync disconnect all night. :D

Could that have been my problem all along? Unless someone can tell me exactly what those settings should be, or at least point me at a process to determine the best settings, I will keep tweaking them in small increments and also set lower SNRM targets until I start to see re-sync disconnects; then I'll back off to the previous setting that worked.

It looks encouraging!