simonc wrote:
> I've changed the network cable, but with no improvement. On changing the
> modem I was researching the cost of a replacement when it finally dawned on
> me that as there are two other computers connected directly to the same
> router and none of those show the same problem it must be a problem with the
> machine I'm using. Maybe there's a dodgy RJ45 connector.
>
> Does anyone know of a software issue that might cause this or is it much
> more likely to be a hardware problem?
I don't know what the root cause of that problem is, but I have an
experiment you can try.
Ethernet consists of two blocks of logic. MAC (higher level) and
PHY (lower level). The PHY block usually contains some autonomous
logic, responsible for auto-negotiating the cable interface rate.
It can carry out that negotiation, even before the OS boots.
On a gigabit Ethernet chip interface, that logic has the choice of
running the interface at 10, 100, 1000, and selecting half or
full duplex.
Both ends of the link, play a part in the negotiation. So they
respond to what each other are doing. And there is a protocol
defined (state machines), to control what they do.
Now, if you go to Device Manager (Start : Run : devmgmt.msc), then
look for "Network Adapters", there should be an entry for your LAN chip.
For example, I have a "Marvell Yukon" entry under Network Adapters.
Now, right click the entry and select "Properties" from the menu.
Look for a tab named "Advanced". It contains a column of property
names and a menu with values. On my chip, the entry I want is
"Speed & Duplex". Normally, mine is set to "Auto Negotiate", which
allows the PHY to pick a speed it likes. To stop the PHY from
going nuts, you can attempt to fix the negotiated rate. For
example, mine offers "100 Mbps full duplex". I could choose that
setting, then see what happens to the link state, after it has
been changed. I know that value is appropriate, because the
device at the other end of the link, happens to support it.
If it stops oscillating, then you're done for now.
*******
There are Ethernet chips, that have built-in cable test capability.
The chip I named in my example, is a chip that supports VCT or
Virtual Cable Tester. It uses TDR or Time Domain Reflectometry, to
evaluate the cable. On a gigabit Ethernet interface, there are eight
wires, arranged in four pairs. The pairs operate in differential mode,
with one signal wiggling downward, when the other signal wiggles upward.
Those signals expect to work into a fixed load impedance. If the
device at the other end of the cable, present the correct impedance
(matching the cable), there are no reflections. If the connector
pins are making bad contact or the wire has been crushed, the impedance
can be a value either lower than or higher than the desired value.
With TDR, the Ethernet chip sends a pulse down the line, It looks
for either no reflection, a downward pulse coming back, or an upward
pulse coming back. The software (even a BIOS screen) driving the chip,
looks at the arrival time, the polarity of the pulse, the pulse amplitude,
and can tell you there is a break in the cable XX feet away, or can tell
you the cable is shorted or is open somewhere. The Ethernet chip can
measure the returning pulse, to the nearest +/- 1 nanosecond.
So if one end of the link, has that test capability, then you have an
additional test you can run.
I actually had trouble getting the Ethernet interface on my current
motherboard to work at first. And the cable tester was indicating
a short on a couple pairs. But that turned out to be perfectly normal.
My modem/router uses four wire interfaces (only four of eight wires
are used for signals), and the modem/router designer decided to short
the unused pairs together. So the diagnostic was a red herring, and
eventually, after enough fiddling with settings in WinXP, the thing
started to work.
I've also used VCT on another motherboard, where the RJ45 connector was
dirty from the factory. At the factory, the motherboard PCB is washed
after soldering is completed. The wash water contains solder flux or paste,
which is water soluble on purpose. If some of that dirt gets in a brand
new connector, when the customer plugs a cable into it, the dirt prevents
a good electrical contact (even though the pins are gold plated). I
inserted and removed the connector five times in a row, to allow the
wiping action to clean the contacts, and that was enough to get it
to work. I could monitor the progress, by using VCT and checking to
see if all pairs read "Normal". So for that previous computer, the
diagnostic worked perfectly, to tell me one pin was dirty. And it
could also tell me, when the connector was making contact.
I expect there is a patent on that, so don't expect every Ethernet
chip to have that feature. As far as I know, it is not part of the
IEEE 802 spec.
So that is an example of a diagnostic capability for Ethernet. Good
for checking cables or connectors.
Paul
>
> "Unknown" wrote:
>
>> If possible you should try another modem.
>
>> "simonc" wrote in message
>> news
57BE29E-BC20-4788-9805-85CF03C59AD6@microsoft.com...
>>> I've rebooted the ADSL modem, unplugged and replugged all the cables but
>>> still every few seconds the network icon in the system tray changes to one
>>> with a red cross, and the message "A Network cable is unplugged" comes up,
>>> followed a fraction of a second later by a "Local Area Connection is now
>>> connected" message.
>>>
>>> Can anyone suggest what might be causing this and where to look for a fix?
>>>
>>> Grateful for advice.
>>>
>>> I'm using XP Professional SP3 (32 bit)
>>
>> .
>>
