Loose Connections

[Mike Monett]

Here is a very basic but essential suggestion. I have been having
problems with a Speedstream 5360 modem from eBay. It was working
fine, then it locked up. The WAN led on the router went out,
indicating it had lost connection to the modem.

TekSavvy support said Bell Canada had lost three area codes in
Ontario. All the DSL modems gave timeout errors and nobody could
connect. My area code was among them. After the Bell problem was
fixed, I assumed that was the cause and continued on my merry way.

But the same wall plug also powered the UPS for my entire computer.

The UPS had a habit of sounding an alert from time to time. I
figured it was a power problem upstream since I live in a small town
far from the nuclear plants in Pickering.

Yesterday, I happened to have the main electrical panel removed to
add a new circuit. While moving the wires, the UPS gave another
alert.

You guessed it. The neutral (white) wire went under a clamp screw
that had never been tightened. The wire was just resting against the
metal.

Checking the other connections, I found several more that were never
properly tightened. Now the lights in the bathroom seem much
brighter, and they don't flicker any more.

So I had fooled myself into believing the random system problems
were somewhere else and out of my control. My guess is the
Speedstream modem got a bad glitch and hung. It is working fine now.
The Bell problem was just a coincidence, and had nothing to do with
the real problem. All this was very poor troubleshooting technique
on my part.

If you have nothing to do some day, and are familiar with working on
AC line voltages, you might check the connections in your panel and
see if any are loose. You might find some that were never tightened
properly and have worked loose over the years. The same applies to
wire nuts and the screw connections in wall sockets.

It only takes a couple of minutes to check, but it could save many
hours tracking down random system problems.

Best Regards,

Mike Monett

 

[Mike Monett]

Fixed in a firmware update. See:
<http://dsl.snet.net/support/dsl/modems/5360/>

I already have the update but was reluctant to apply it since it was not
clear if it was needed. Do you have any information on what problems it is
supposed to fix?

The 5360 is not a very good DSL modem. There's no status information
available so troubleshooting line related problems is difficult.
Connection issue require numbers, such as the line levels, signal to
noise ratios, etc. I suggest you replace it with a 4100, 5100B, or
later DSL modem that provides web based diagnostics.

This is a cheap backup modem from eBay in case the main one dies. TekSavvy
can retrieve all the line information if needed, and they say all the data
is exceptionally good.

The switching office is less than two blocks away, so the SNR is excellent.
I checked the speeds using different online DSL speed checks, and they all
agree it is running about as fast as the connection allows. My neighbour
has some more expensive modems and he gets the same results.

Incidentally, please lose the right justified formatting and indented
margins. It looks nice with the proper proportional fonts, but looks
awful with monospaced font. Most usenet readers use monospaced.

Just for you, I'll make this post in XNews. But it won't have any spell
check, urls, or references from other sources.

Also, I have a problem with your links that are surrounded with "<>". This
makes it difficult to copy and paste the url from an editor to the browser
window.

Other than that, I really do appreciate your posts. They are on-topic, full
of very good information, and you usually include references for follow-up
if needed. I think you set a very good example for the others. Thanks.

Best Regards,

Mike Monett

 

[JW]

True. However, it prevents Forte Agent from word wrapping the URL's
and also blocks the URL into a clickable link if there are some odd
characters, such as parenthesis) in the URL.

Strange, I've never had that problem with Agent, and I have it set to wrap
at 74 characters. Let's see:
http://cgi.ebay.com/HP-Agilent-5454...04247QQcmdZViewItemQQ_trksidZp1742.m153.l1262

Looks ok to me. I'll check again when this message shows up.

 

[JW]

Strange, I've never had that problem with Agent, and I have it set to wrap
at 74 characters. Let's see:
http://cgi.ebay.com/HP-Agilent-5454...04247QQcmdZViewItemQQ_trksidZp1742.m153.l1262

Looks ok to me. I'll check again when this message shows up.

Still looks OK.

 

[Mike Monett]

[...] >> I already have the update but was reluctant to apply it since
it >> was not clear if it was needed. Do you have any information on

No real data but I have a guess as to what it fixes. The obvious one
is that 10/100 full/half duplex detection is broken. I had some
random short term (< 5 sec) disconnects that went away with the
update. Some of my customers have experienced occasional loss of
sync problems with long lines (12K to 15K feet) that was also
reduced. I tested a 5360 at home with the help of my ISP. They
obtained stats from the DSLAM before and after the update. They were
greatly improved in terms of staying connected after the update. My
guess(tm) is that there were some parameter tweaks to more closely
resemble what's found in the real world. I also did some bench
testing to see if various modems can handle 6Mbits/sec.

Most could but the 5360 was a bit flaky. It would sometime retrain
the line for no obvious reason. I don't recall if that was with old
or new firmware, probably old as I did it long ago.

The modem seems to be running fine after fixing the loose connection.
I'll keep monitoring it, but if it ain't broke, don't fix it

[...]

That's not what I found with a before/after test. Got any numbers?

I'm not sure what you mean by a "before/after test". Yes, I have the
Teksavvy numbers around someplace, but I don't know where I put them.
They did the test several times while I was fixing TCP problems on
Win98. Each time they remarked how good the numbers were.

In the meantime, here's some runs I just took using the test on
http://www.speedtest.net/index.php

Chicago : 4316
Dallas : 3740
Hamilton : 4227
LA #1 : 3560
LA #2 : 3717
LA #3 : 3540
London : 3273
Toronto : 4229
Vancouver : 4090
Denver : 1953

These were taken while listening to streaming classical music from DR
Klassisk, in Copenhagen. That eats 128k:

http://www.dr.dk/netradio/Metafiler/ASX/DR_Klassisk_128.asx

NetStat shows the instantaneous rate can vary quite a bit:

http://www.analogx.com/contents/download/network/nsl.htm

but the results are very similar to what my neighbor gets down the
hall.

Overall, I'm quite satisfied with the performance. TekSavvy advertises
5 MB for their DSL, and I'm getting pretty close to what I pay for:

http://www.teksavvy.com/en/resdsl.asp?ID=7&mID=1

Incidentally, I have an Acterna/JDSU HST-3000a tester. Some results
from typical lines: <http://dsl.11junk.com> The various graphs and
tables from "JV" are before and after results found after re-routing
the inside wiring and replacing the microfilter mess with a splitter
at the MPOE. It sorta worked both before and after, but the numbers
improved dramatically after rewiring.

That's roughly what I've been finding as I go around testing
everyones DSL lines. Lots of problems with inside wiring and
microfilters.

The graphs are too small to tell what is happening. I couldn't get
much information on the tests it makes, but the spec looks
interesting.

What causes the problems and why does rerouting fix them?

As for as the wiring here, it enters the building from underground,
then goes directly to the switch panel. From there, I believe it goes
through pipes to the different apartments. The building is constructed
with 10.5 inch thick prestressed concrete beams 4 feet wide. Any holes
cost a fortune to drill, but I think the power lines are routed
through individual aluminum pipes to each apartment. This would
provide good shielding from spikes on the power lines. I don't know
how the telephone lines are routed, but it is very unlikely that I
could do anything about it if there were a problem. But they seem to
be fine.

Hmmmm. Not good. Any chance you're suffering from a "disturber"
signal in the same bundle?

Yes. I did have problems before while I was on dialup, and they gave
me to a new pair. Then DSL came to town and I finally switched over
after everyone else in the apartment started getting reliable results.

Incidentally, the Bell installer did measure the performance of the
line withan instrument similar to yours, and he said the numbers were
very good.

I had interference problems from a mess of T1's that were sharing
the same bundle for about a mile. However, it showed up in the error
rate and a very erratic looking channel response graph (that I
forgot to save).

Your story about the loose neutral wire is interesting. We had an
earthquake in 1989 that jarred loose all kinds of underground
wiring.

In one case, the neutral wire for a large electronics manufactory
was lost at the xformer. Of course, the phases were not balanced and
resulting in brownouts or burnouts depending on location. I've had
a similar problem at my house, where the combination of an
unbalanced load and an inadequate high resistance neutral wire
resulted in some odd effects. The only real clue was that there was
a loud 60Hz buzz when a high current drain appliance would kick in
coming from the copper water pipes under the house. Since the
neutral wasn't carrying the current, the water pipes were trying to
do its job, and buzzing. I measured about 3A AC current around the
water pipes with a clamp on ammeter. I'm sure it was more as there
were multiple paths to ground.

Congrats on analyzing and fixing the problem.

Thanks.

The whole point was to indicate we tend to take things for granted.

Sometimes it pays to look under the rug and find out what uglies are
lurking there

Thanks. I appreciate it. Content before form is good.

I wonder if you haev any problems reading newspapers? Or IEEE
articles? Maybe datasheets or app notes? Pretty much anything in pdf
files is right justified. There's a good reason for this. It greatly
speeds reading text since you don't have to track the edge of
individual lines. Your eyes can take in one or more lines at once,
without wasting time fooling with the right margin. That's why
newspapers and professional articles have been right-justified for
centuries.

I also find it makes my notes much easier to read. But I wrote my own
editor so I get the advantages of being able to make it do whatever I
need.

True. However, it prevents Forte Agent from word wrapping the URL's
and also blocks the URL into a clickable link if there are some odd
characters, such as parenthesis) in the URL. I've experimented with
various ways to insure that the URL is mutilated by the news
software and the <xxx> method seems to be the least disgusting.

The clients won't wrap short urls. Why not use TinyUrl or some such?

Thanks. Oh-oh, you've been reading my stuff. That usually causes
brain damage. If you dig more carefully, you'll also find that
they're also mixture of guesswork, satire, humor, and anecdotes.

Try not to take anything I write seriously.

No problem

Best Regards,

Mike Monett

 

[Mike Monett]

Yep. Now try that with a URL that has:

1. Spaces in the URL such as:
http://802.11junk.com/jeffl/antennas/Sinclair SRL441- 2P/slides/SRL441 back.html

2. () in the URL
http://img.villagephotos.com/p/2003-8/361947/DXNHA-IMG_7461(Medium).JPG

3. There were also some oddities with other special symbols in the
URL, but I can't seem to find any examples.

4. Try indenting a URL with a few spaces such as:

http://cgi.ebay.com/HP-Agilent-54540A-500-MHz-2-GS-s-4-Ch-Oscilloscope_
W0QQitemZ200254042495QQihZ010QQcategoryZ104247QQcmdZViewItemQQ_trksidZp
1742.m153.l1262 As I recall, sometimes it works but sometimes it
wraps. I'm not sure what controls it.

Jeff, those examples all belong in TinyUrl!

When you come across a link that is too long, or so badly done that you
cannot predict what will happen in different clients or browsers, you owe
it to your readers to go out of your way and fix those kinds of problems.

Many people might be very interested in what you have to say, but maybe
they are not skilled enough in the ways of the web to figure out how to fix
those kinds of problems. But you can, and save everyone a big headache
from wasting time on those cases. That's exactly what TinyUrl and all the
other sites are for.

I find email clients start wrapping at column 70 or so. If a reply is
quoted, it adds two or three characters to the beginning. This often makes
the line long enough to wrap, and urls are probably the most likely to get
garbled. But a TinyUrl link is very short, so it will probably survive in
the longest thread.

Now you can guarantee your links will survive, and you no longer need the
"<>" separators. That will save you time as well as your readers. So be
kind to your readers, and use TinyUrl for any link longer than 50
characters or so.

Golly, I think I just persuaded myself to do the same thing

Best Regards,

Mike Monett

 

[Mike Monett]

[...]

Try:

<http://speedtest.teksavvy.com/>

Yes, I know. Teksavvy suggested the same one. Unfortunately, it
requires Java. I run Win98 with 2 gig partitions for DOS. I don't have
Java installed, and there's not enough room on the hard disk.

So I have to pass on that one:

Good enough. I never could get much more than 3500 out of the
Efficient 5360 DSL modems. Looks like some users are getting 4 to 5
Mbits/sec. You might want to try their test (without the background
music):

<http://www.dslreports.com/archive/teksavvy.com>

Looks like most people are getting around 4300KB, the same as me. I am
pleased you found that link. Now I can sleep nights, knowing I'm just
about average

[...]

The problems with inside wiring are the side effects of the
spaghetti wiring topology. Wires can be either bus or star topology.
Some are unterminated, which look like a capacitor. Some have shorts
or leaks to ground or conduit resulting line balance problems. I'm
also beginning to suspect that there are some differences in quality
between brand of microfilters. The lines tend to run next to power
lines, resulting in noise pickup. CFL lamps seem to be a problem.

Usually, these only cause problems when you're at the line length
limit (15,000ft), but even the best of short lines can be trashed by
interference or an imbalance.

The HST-3000b has a built in TDR (time domain reflectometer). The
pulse width is too wide to show anything less than a few feet (I
forgot the exact number) from the tester. However, when I use it to
test inside wiring, it shows every rotten connection, flaky
microfilter, bridged tap, and broken wire. I can't use it to reverse
engineer the topology or locate specific devices, but it's a great
quality control test to see whether the inside wiring might be an
issue. Again, I'm still learning to interpret the data.

Thanks for the explanation. I'm amazed the problems can be so bad.

[...]

The problems usually start at the MPOE (minimum point of entry) also
known as the NIU (network interface unit) or demarc (point of
demarcation). One side is the responsibility of the telco and is
usually in good shape. The other side is the responsibility of the
property owner, which is usually a mass of spaghetti wiring.

Yes, we use the term demarc here in Canada. But the wiring was done
very professionally, as was most everything else. The only thing is
they forgot to tighten some clamp screws in my breaker panel

It's odd that you would have a problem so close to the CO (central
office). My guess(tm) is that your line goes well beyond your
building from the CO. The way it works is that the telco does NOT
run a pair directly from the CO to your building. They run a mass of
100 pairs (or more) down the poles or through the underground
conduit to as far as 3 miles away. Your pair is just a tap off this
bus which is a bridged tap.

<http://en.wikipedia.org/wiki/Bridge_tap> Your line could also have
entered several other buildings further downstream. Lots of ways to
create problems.

Yes, I know how the lines are run. This particular run is very short.
It only goes from the main office to this building, then terminates in
a subdivision a short distance away. Midland is quite small, so
there's not many places you could go three miles and not end up way
out in the country side

But that's the second time I had problems with a pair. I understand
about half the lines have drops that were never disconnected.

The tester has a go/no-go check for installers. I don't use it
because it will pass a really marginal line. The installer usually
tests the line from the MPOE, not from the DSL modem, since they're
only really responsible for the telco part of the puzzle.

The problem is usually between the MPOE and the DSL modem, which may
not have been part of the test.

No, he had a much more comprehensive test. Recall I'm in Canada, and
way out in the boondocks. So the procedures are quite different here.

If you had a better DSL modem, with internal diagnostics, you could
have obtained the line stats yourself. Incidentally, look on eBay
for an Efficient 5260. It has a huge amount of built in stats and
diags.

Funny you should mention that. My main modem is a 5260.

Unfortunately, I was never able to get into the modem and do anything,
so I had to rely on TekSavvy to get the readings from their end. Like
most people, I thought the 5360 would be an improved version. It looks
very much the same, but actually has less capability than the previous
model. But they both work fine, given the very short run and excellent
SNR.

I used one for line testing before I bought the HST-3000b. See:

Also, I just received a new battery for my tester that includes a
serial port suitable for driving a printer. (Don't ask me why they
put the serial port in the battery). The installers used to carry a
printer and leave a printout of the line stats in the phone closet.

Locally, they don't do that any more, but I plan to resurrect the
practice.

Sounds like you do a lot of this kind of work.

Not me. I'm a classic paranoid and am ready and willing to blame
anyone and anything for my problems. What I find amazing is that
after doing this for so many years, I can almost guess what's wrong
from the general symptoms, usually without any troubleshooting or
even a complete description.

Very good. Now you can impress people with magic and charge
accordingly:

Best Regards,

Mike Monett

 

[JosephKK]

Here is a very basic but essential suggestion. I have been having
problems with a Speedstream 5360 modem from eBay. It was working
fine, then it locked up. The WAN led on the router went out,
indicating it had lost connection to the modem.

TekSavvy support said Bell Canada had lost three area codes in
Ontario. All the DSL modems gave timeout errors and nobody could
connect. My area code was among them. After the Bell problem was
fixed, I assumed that was the cause and continued on my merry way.

But the same wall plug also powered the UPS for my entire computer.

The UPS had a habit of sounding an alert from time to time. I
figured it was a power problem upstream since I live in a small town
far from the nuclear plants in Pickering.

Yesterday, I happened to have the main electrical panel removed to
add a new circuit. While moving the wires, the UPS gave another
alert.

You guessed it. The neutral (white) wire went under a clamp screw
that had never been tightened. The wire was just resting against the
metal.

Checking the other connections, I found several more that were never
properly tightened. Now the lights in the bathroom seem much
brighter, and they don't flicker any more.

So I had fooled myself into believing the random system problems
were somewhere else and out of my control. My guess is the
Speedstream modem got a bad glitch and hung. It is working fine now.
The Bell problem was just a coincidence, and had nothing to do with
the real problem. All this was very poor troubleshooting technique
on my part.

If you have nothing to do some day, and are familiar with working on
AC line voltages, you might check the connections in your panel and
see if any are loose. You might find some that were never tightened
properly and have worked loose over the years. The same applies to
wire nuts and the screw connections in wall sockets.

It only takes a couple of minutes to check, but it could save many
hours tracking down random system problems.

Best Regards,

Mike Monett

Thank you. I will remember.

 

[David L. Jones]

Here is a very basic but essential suggestion. I have been having
problems with a Speedstream 5360 modem from eBay. It was working
fine, then it locked up. The WAN led on the router went out,
indicating it had lost connection to the modem.

TekSavvy support said Bell Canada had lost three area codes in
Ontario. All the DSL modems gave timeout errors and nobody could
connect. My area code was among them. After the Bell problem was
fixed, I assumed that was the cause and continued on my merry way.

But the same wall plug also powered the UPS for my entire computer.

The UPS had a habit of sounding an alert from time to time. I
figured it was a power problem upstream since I live in a small town
far from the nuclear plants in Pickering.

Yesterday, I happened to have the main electrical panel removed to
add a new circuit. While moving the wires, the UPS gave another
alert.

You guessed it. The neutral (white) wire went under a clamp screw
that had never been tightened. The wire was just resting against the
metal.

Checking the other connections, I found several more that were never
properly tightened. Now the lights in the bathroom seem much
brighter, and they don't flicker any more.

So I had fooled myself into believing the random system problems
were somewhere else and out of my control. My guess is the
Speedstream modem got a bad glitch and hung. It is working fine now.
The Bell problem was just a coincidence, and had nothing to do with
the real problem. All this was very poor troubleshooting technique
on my part.

If you have nothing to do some day, and are familiar with working on
AC line voltages, you might check the connections in your panel and
see if any are loose. You might find some that were never tightened
properly and have worked loose over the years. The same applies to
wire nuts and the screw connections in wall sockets.

It only takes a couple of minutes to check, but it could save many
hours tracking down random system problems.

I had a problem like this with my dishwasher.
It starting failing at the same point in the cycle, intermittent at
first, then all the time.
Not wanting to bother troubleshooting myself I called a repair dude
who promptly replaced the main control module and all was good (he
left me the old module).
A few days later it fails again, same problem. I swapped back the old
module and problem remained.
Repair dude comes back and finally figures it out after a lot of
frigging around.
Yup, you guessed it - bad contact on the back of the mains wall plug,
allowing operation at low currents, but fails at higher currents (like
when the heater comes on I guess).
Socket must have always been like that since I moved into the place.
The tech playing around with the mains plug was enough to fix it for a
day or two.

Dave.