[SI-LIST] Ethernet switch problems: detailed questions
- From: "Jan Vercammen" <jan.vercammen.jv1@xxxxxxxxxxxxxxxx>
- To: si-list@xxxxxxxxxxxxx
- Date: Thu, 12 Sep 2002 16:28:34 +0200
Hello SI-list,
about 5 weeks ago a posted questions concerning Ethernet switch problems.
The old mail is included below
as a reference. First of all, many thanks for your helpful comments.
I have more details on the switch problem:
-1- the switch is not physically damaged, however the swich goes into
hangup and needs a reset (power off/on)
-2- disconnecting ports (if necessary all of them) does not help
-2- the problem is repeatable but we do not have an explanation why the
switch freezes
Here are the details of the switch configuration: it is a small switch fed
by an AC/DC adaptor (220V). There are two
10Mbit ports connected to printers, two 100Mbit ports (connected to input
devices that generate images for the
printers) and one 100Mbit port connected to a printer buffer. All cabling
is unshielded twisted pair. Maximum length
of the cables is 3-5m. In short: a small test setup for a network in an
office environment. There is lot's of traffic
between the printers, input devices and buffer device. The hardware or
software has not changed on any device,
except for the buffer (see belwo).
We can reproduce the hangup, here are the details. The culprit device is
the buffer device. It is the size of a PC
with a mother board, hard disk, SDRAM modules, power supply and keypad. The
processor platform is PowerPC
and the CPU type is selectable: 603 or 740. The problem is apparently
caused by the voltage regulator module (VRM)
which generates 2.6V for the core voltage of the MPC740. The VRM uses the
5V as input and as a by-product
contaminates the 5V with a 470kHz ripple.
The 5V is also used for the Ethernet physical National DP83840 and used
specifically for the analog power supply
part of the physical. Here lies the problem: the 5V is filtered by means of
an RC filter, using a ferrite bead, a small resistor
and 100nF+1nF decoupling. This design dates from 1996 and was found to work
OK with a MPC603 which uses
the 3.3V for its core supply and does not have a VRM.
Since we switched to MPC740 and the VRM we had a problem with this (a
single type of) Ethernet switch. We can correct
the situation as follows: replace the RC filter decoupling 100nF+1nF with a
single 10uF ceramic capacitor. We have tested
both versions: 100nF+1nF and 10uF several times and the 10uF is a big
improvement:
test with 100nF+1nF test with 10uF
hangup after 24 to 36 hours no hangup after one week of
testing!!
TX and RX error packets about 7-10% TX and RX error packets 0%
So the problem is caused by a bad filtering of the analog supply (our fault
and not of the componenets) of the Ethernet
physical. This shows by a 7 to 10% error packets on transmit and receive.
The switch will freeze after about 1 day to 1.5
days (tested 3 times). The 10uF capacitor produces errorless TX and RX and
no hangup, even after a week of testing!
I measured, with a spectrum analyser, the analog power supply on the
physical: 100nF+1nF gives about 50mV
of noise at 470kHz, with a 10uF cereamic this is only 0.5mV. So the 10uF
ceramic does a better job on
providing a clean analog supply.
Although we seem to have solved the problem a nagging question remains: why
does this bad filtering freezes
an (external) switch? Any ideas what is going on?? So our machine when it
provides a badly filtered analog
supply to the Ethernet physical can freeze an external Ethernet switch.
Amazing!!
Kind regards,
Jan Vercammen
Agfa-Gevaert
Belgium
****************** old mail below *******************
I have an urgent question concerning Ethernet twisted pair. We have
received from the field three reports
that one of our products seems to have the capability to physically damage
the port of a switch to which it
is connected.
To be more specific: our device uses a Bob Smith termination. The RJ-45
unused pins are connected with
75ohm to a common-rail, which in turn is capcitively coupled (using 1nF) to
the chassis (=housing of device),
which is also connected to the logic ground of the PCB.
The center taps of both transmit and receive are also connected with 50ohm
to the same common rail.
The other port is a switching hub and I do not know if shielded or
unshielded cable was used.
What seems to happen is that after a while the communication hangs and the
port at the switch does not operate
anymore, not with our device or any other.
For now I cannot confirm that resetting the switch solves the problem and
it looks that there is hard damage.
I have heard of cable sheath discharges (the tubing of the cable charges
and next discharges to the environment,
damaging hardware). However, for now I would like to disregard such exotic
scenario.
So my questions are:
-1- as anyone encountered similar switch burnouts (whether hard or soft)?
-2- if so, what is the possible explanation and fix?
Kind regards,
Jan Vercammen
Agfa-Gevaert, Belgium
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
