Re: Company network slowdown

On Sat, 17 Sep 2005 20:17:29 -0500, ibuprofin@xxxxxxxxxxxxxxxxxxxxxx
(Moe Trin) wrote:

>>Nope. I wasn't looking at the ethernet waveforms although those were
>>interesting to look at. I was sending a pulse down the line and
>>looking for discontinuities.
>
>The rational is that the DIX specification isn't really covering the
>higher frequencies because what happens up "there" doesn't have the
>effect as the 10/20 MHz frequencies.

Well, that's correct. The 802.3 MPE (Manchester Phase Encoding)
waveform concentrates most of it's power around 10MHz. The
10Mbits/sec data rate only requires about 30Mhz or bandwidth to
operate. I don't see any reason to characterize the cable beyond its
highest operating frequency.

>Some of what you see on a 5 nano
>rise/fall time (200 MHz = about 2 foot resolution) isn't there as far
>as the Ethernet signal is concerned. None the less, having the sharp
>rise/fall really does help in multiple mismatch conditions. A long
>time ago, I had something similar (a NE555 driving half of a 74S74
>driving an LH0002) when I was looking at a number of coax runs in a
>data acquisition trailer, and it certainly was useful in locating
>the fault, but we used VSWR meters to go/no-go the lines.

Mine was a bit fancier. 555 driving some ECL gates to a fast
switching xsistor with a clamp diode to prevent saturation (which
would trash the risetime).

This one seems to be too crude:
http://www.tkk.fi/Misc/Electronics/circuits/tdr.html
but does explain the principles involved.

>Major problems, and breaks in the cable itself.

I never had to deal with breaks in the cable. However, damage to the
cable was a major cause of extended troubleshooting exercises. In one
warehouse, the yellow cable was snaked through the overhead. No
problem getting to the cable except that it was 20 ft off the floor
which was about 3 ft too high for my tallest ladder.

>>kinks that cause the dielectric to migrate
>
>That would have to be a pretty wicked bend - certainly down below the
>MIL-C-17 bend radius requirements. For the orange stuff, we never had
>bends shorter than about 2 foot radius.

The yellow coax was so stiff that it wouldn't bend with less than a
1ft radius anyway. However, that didn't stop anyone from trying to
bend it across a sharp corner. The problem was always the same.
Someone's workstation was a bit too far from the probe transceiver.
DB15 extensions were impossible to find. The culprit certainly was
not going to re-arrange their office layout for the convenience of the
cabling. So, they give a good hard tug on the cable and try to brute
force into giving them a few extra inches of cable length. If the
cable bend radius started out at 1ft, it was now about 1 inch. Some
of the buildings had metal studs in the walls and were capable of
cutting the cable in half. I recall one 300ft run that had 3 or 4
cable splices installed. Cheapernet installs had exactly the same
problem except that the cable was even more fragile. The problem
didn't go away until we went to 10baseT.

>>The transceiver loading is fairly light. A few PF at most.
>
>Two puff max - shunt resistance over 100K

As I vaguely recall, 4PF maximum. Very lightly loaded considering
that the coax cable was about 25PF/ft.

>Terminators speced as 49.9 Ohms +/- 1 percent, at 0 to 20 MHz with the
>phase angle of the impedance not to exceed 5 degrees, which is relatively
>good.

Major overkill. I don't think any of my junk VSWR test equipment or
directional couplers are accurate enough to measure that at low
VSWR's. I would have to use a bridge to get the accuracy over the
frequency range.

The reason for the tight specs had nothing to do with VSWR. 10base5
and 10base2 both use the DC levels on the coax for collision
avoidance. The transceiver has a current source and uses the two
terminators as a load to get the exact voltage required. Variations
in typical production and installation, plus coax copper losses,
caused the tolerance accumulation to potentially cause this voltage to
drift out of spec. Rather than transfer the cost of a high tolerance
current source to the transceivers, it was cheaper and easier to
demand that the terminators were close to perfect. Eventually,
everyone figured out how to make cheap precision current sources, so
the terminator tolerances didn't need to be that critical. However,
once written, such specs tend to be cast in stone.

The 10base2 Cheapernet terminator specs are something like 51 ohms,
+/- 5 percent. 10% will work. Unfortunately most of them were built
with carbon composition resistors, which a slightly hygroscopic and
will therefore tend to drift over time. I recently tested my pile of
50, 75, and 93 ohm terminators and found many of them were way out of
spec. A quick bake in the oven solved that problem.

>The coax is allowed to have sinusoidal +/- 3 Ohm ripple at spacing
>of two meters or less (on top of the 50 +/- 2 Ohms of the cable), and
>that's probably dominant, I've seen people hand select MIL-R-11 carbon
>composite resistors for "50" ohms, and then wonder why the VSWR is so
>gross at higher frequencies. They might as well have used a wire wound.

Sigh. You must work in a research or government environment. Nobody
else I know could afford or has any interest in such details. Carbon
comp resistors are terrible at higher frequencies but are probably
just fine up to 30MHz. I vaguely recall tearing apart a 10base5
terminator and finding a single 51 ohm carbon comp resistor in
parallel with something to get it down to 49 ohms. Resistor lead
length inductance is probably the real killer.

>(...) He whips out his Simpson 260, center-center, shield-shield,
>nothing (open) from center to shield. "Yup - it's OK." We had a nice
>half hour lesson as I showed him how to use a slotted line and VSWR
>meter, and how to measure insertion loss with a shorting plug, and how
>it differed from the results using an open.

I still have some slotted coax and waveguide lines floating around. I
use them for skool demonstrations. These days, I use a network
analyzer. It's nice to have everything displayed on a single Smith
chart.

I'm not going to try and justify my use of crude tests. I was looking
for continuity, not compliance with specifications. I don't have a
test lab available at the customers. I also can't do much if I find
something wrong with the terminator except replace it with another
one. Also, knowing exactly why the specs are so tight on the
terminator was a big help in knowing what I could get away with. I
did find a few failures with the ohms-guesser method. Shorts were
common. Defective transceivers were a problem. 117VAC on the coax
shield (long story here) was found with a volts-guesser.

Incidentally, I still have about 4 Simpson 260 voltmeters in various
levels of functionality. However, these days, I use a DVM.

>Oh, so you don't like my Exabyte 8205s and 8505s? ;-)

I have an 8205 somewhere. I never got into the 8mm drives. Too
expensive at the time and they had already developed a rather bad
reputation. I didn't see it as much of an improvement of DC-600 size
drives except for capacity. Instead, I went directly to 4mm drives
and jukeboxes. Big mistake as it took a few years to demonstrate
their shortcomings. I later went to AIT which seems to have fixed
most of my complaints.

>I dunno - we've been using it for years, and it hasn't killed us yet.
>(I know, I know - it will).

Have you ever had to do a massive restore from tape (while under
pressure)? I have and I can assure you that reliability is not one of
the better features of tape. I have a small collection of recovery
tools that I use in case I get the all too common read errors. I've
also had to use a tape recovery service to deal with tape errors,
where the tape head had worn enough to be unable to read an old tape
or where a new replacement drive would not read an old tape. For a
time, one customer would put the old DDS-2 drive in the safe along
with the tapes just to be sure they had the hardware to read the
tapes. To add a challenge, HP was constantly screwing around with the
tape format so that a random version of their firmware would not
necessarily read tapes made with a different version. Sony was doing
a heroic job of trying to stay HP compatible but eventually gave up.

This brings back nightmares of baby sitting tape restores that took
all night and had to be watched constantly. I'm glad those days are
over.

>Actually, I shouldn't complain - my "RF" tool box has more adapters
>than you can shake a stick at, though I am missing my WR-112 to clip
>leads. ;-)

Oh, be serious. I have about 20 lbs of adapters and connectors. I
would visit the local hamfests and retired hams and buy up all the
adapters I could find. It didn't matter what type or flavor. Best
investment I ever made. Every Field Day, about a dozen adapters
evaporate, but I have spares. In this case, BNC to F adapters are
very common in CATV work to interface to the test equipment. I carry
a pile of them. There's almost no loss but admittedly, they are a
rather bad 75 ohm to 50 ohm match. At 10Mhz, it's tolerable.

Incidentally, a demonstration I like to give at radio club meetings is
grabbing one of my adapter boxes and stringing as many adapters in
series as possible. I usually have a string about 6ft long. I attach
a VSWR meter at both ends, dummy load, and 450Mhz transmitter. I then
ask the assembled horde what they would predict for the loss.
Conventional wisdom says that adapters are evil abomination and should
be avoided at all cost. It turns out that the adapter chain has about
the same loss as an equivalent length of RG-8/u. So much for the
lossy adapter theory.

>When I moved to this house, I initially set up using 10Base2, because
>it was quick. The following winter, I spent one weekend in the attic
>pulling CAT 5 to each room (why not), including eight drops in the
>den. Only when I was finished did I think it might have been a good
>idea to replace the US$2/mile cable that the klowns had for the phones.

I use CAT5 for everything including video. About 5-7 cents per ft
which is cheaper than coax. I did a remodel in about 1995 and was
able to run conduit to various places. The size varies from 1/2" to
1" schedule 40. If I need to run something, I just add it to the
tangled mess. Of course, I ran the conduit from where I needed it
least, to where I thought it might be useful. That resulted in more
cables under the desk and along the floor. I also have about 500ft of
fiber in the pipes, which has yet to be useful. If I had to do it
today, I would use the blue flex plastic wiring conduit instead of
schedule 40.


--
Jeff Liebermann jeffl@xxxxxxxxxxxxxxxxxxxxxx
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558
.

Relevant Pages

  • Re: cutting psu wires
    ... I don't like black electricians tape because over time the glue comes off on ... What you should not do is use elastic rubber bands to hold the wires back! ... don't use tie wraps on flat IDE and floppy cables as you can ... one wrap, it is now a flat oval adjusted to the size of the IDE cable. ...
    (comp.os.linux.hardware)
  • Re: Digital TV question
    ... there are no cables!!!! ... I think the vcr will be used for many more years yet. ... I would strongly suggest looking at Tivo. ... The only thing it can't do is record the material to a tape or DVD. ...
    (alt.home.repair)
  • Re: Company network slowdown
    ... All of our overhead drops go to wall plates. ... Cables hanging from the overhead were ... came down from the TV tower because the automagic VSWR meter relay had ... My string of adapters showed almost no VSWR at 450Mhz. ...
    (alt.internet.wireless)
  • Re: old 68pin cables for new U320 cards and HDDs ? ok ?
    ... >>> card and old scsi 7.5K RPM drives. ... >>> Does anybody know is it possible to use these cables with new U320 ... >Nowhere does it say that you always have to use the last connector. ... >A terminator terminates the cable, ...
    (comp.periphs.scsi)
  • Re: FC adapters?
    ... I must have adapters for H80s and p650s. ... > redundant and the two 9216 switches are not cross-connected. ... they are compatible with 5716 HBA. ... I always saw yellow cables. ...
    (comp.unix.aix)