DComTalk.com

I need to put a piece of hardware between an ethernet hub/router/switch
port and a PC that can kill the connection (10-BaseT or 100-BaseT) on
demand. I don't want to pass the signals through a relay's contacts if
I can help it, because of losses, but can I safely short the pairs to
kill the link? Ie a mechanical relay contact that shorts orange to
orange-white and another contact that shorts green to green-white. I
could use a mechanical relay but an opto-isolator might be neater. I
will know the hub/router/switch port (since I will be providing it) but
the PC could be anything. Do I need a dead short, or would a simple
resistance be OK, like 1/10th the line impedance (4.7ohms perhaps)? Is
there a way to know that the PC is plugged in from the physical line?
If I measured the current flowing in the short could I detect the PC
with one pair (and my hub/router/switch port with the other pair
presumably) and which pair would it be (orange or green)?
Any ideas or points to the information I need would be gratefully
received.
Have I placed this message in the correct group? Is there a better
forum for this message?

Howard wrote:

I need to put a piece of hardware between an ethernet hub/router/switch
port and a PC that can kill the connection (10-BaseT or 100-BaseT) on
demand. I don't want to pass the signals through a relay's contacts if
I can help it, because of losses, but can I safely short the pairs to
kill the link? Ie a mechanical relay contact that shorts orange to
orange-white and another contact that shorts green to green-white. I
could use a mechanical relay but an opto-isolator might be neater. I
will know the hub/router/switch port (since I will be providing it) but
the PC could be anything. Do I need a dead short, or would a simple
resistance be OK, like 1/10th the line impedance (4.7ohms perhaps)? Is
there a way to know that the PC is plugged in from the physical line?
If I measured the current flowing in the short could I detect the PC
with one pair (and my hub/router/switch port with the other pair
presumably) and which pair would it be (orange or green)?

Why not just get a cheap hub or switch to place on that connectiona and just
kill the power to it? With the prices these days, it would likely be
cheaper than trying to build something and more reliable too.

"Howard" <h.shaw@dmkent.co.uk> wrote:

to detecting the PC, there must be some current flow when the pair is
shorted, it will be small, but there must be some.

It's a transformer coupled 2-3 volt peak-to-peak signal with not a lot
of drive capability. In fact, measuring it is probably really bad for
your signal integrity. Now that I think about it, a shorted line will
only see link pulses, which don't happen that often...

Is this for coax, 10BaseT, 100BaseT, or Gigabit? Each of them has
separate signalling, negotiation, and duplex...

What else does this product (need to) do, maybe there's a way to do
what you want...

I haven't used newsgroups since I was at University back in 1980, so
its good to see that people are still as quick to respond as they were
when the internet was young. Thank you all.
Thanks for the suggestion, but cheap hubs are not that cheap. I have a
budget of around £1 per unit to add this ethernet killing opton to an
existing product (that does lots of other things) and I can get a
simple two pole relay or a couple of optoisolators for around 25p. As
to detecting the PC, there must be some current flow when the pair is
shorted, it will be small, but there must be some. I can't take the
hub apart and connect in to the back of the leds because of keeping
approvals. The hub is too far away to fix an optodetector to the
outside to "see" the led. I am sure that shorting with a relay or
semiconductor device would kill the connection, but there is no point
in me going ahead if I can't get approval later. I'm guessing at the
moment because I still can't find the actual physical layer
specification, but I will keep looking. If the data was sent on top of
a continuous carrier then I could detect that, but if the pair floats
when there are no packets flowing then I might be stuffed.

"Howard" <h.shaw@dmkent.co.uk> wrote:

I need to put a piece of hardware between an ethernet hub/router/switch
port and a PC that can kill the connection (10-BaseT or 100-BaseT) on
demand.

Is there an echo in here?
http://dalco.com/ProductsList2.aspx?Cat ... lection=30 is
manual, and as others have pointed out a cheap switch will stop
forwarding packets when it's power is removed.

There isn't any current flowing in the cable to be able to detect PC
connection, but there's usually an LED on a switch...

Howard wrote:

As
to detecting the PC, there must be some current flow when the pair is
shorted, it will be small, but there must be some.

There is no current. The cable is transformer connected to the NIC, so
there's no DC at all. There's only the transmitted signal. I doubt you
could create a reliable, non-interfering detector, for around £1.

Howard wrote:

I need to put a piece of hardware between an ethernet hub/router/switch
port and a PC that can kill the connection (10-BaseT or 100-BaseT) on
demand. I don't want to pass the signals through a relay's contacts if
I can help it, because of losses, but can I safely short the pairs to
kill the link?

That should work fine. Most likely you only need to short one
direction, depending on the situation. To minimize the effect on the
line, you might use a double pole relay with two poles shorted together,
and the ethernet lines on the contact side.

Ie a mechanical relay contact that shorts orange to
orange-white and another contact that shorts green to green-white. I
could use a mechanical relay but an opto-isolator might be neater. I
will know the hub/router/switch port (since I will be providing it) but
the PC could be anything. Do I need a dead short, or would a simple
resistance be OK, like 1/10th the line impedance (4.7ohms perhaps)?

1/10th might not do it. If you figure the attenuation on a 100m cable,
your signal must be much smaller than that to qualify as shorted.
1/100th I would believe, though, without actually looking at the numbers.

Is
there a way to know that the PC is plugged in from the physical line?
If I measured the current flowing in the short could I detect the PC
with one pair (and my hub/router/switch port with the other pair
presumably) and which pair would it be (orange or green)?
Any ideas or points to the information I need would be gratefully
received.

Have I placed this message in the correct group? Is there a better
forum for this message?

Unlike about half the posts here, which belong in a TCP/IP newsgroup,
this is the correct group.

Even more interesting, this question came up a month or two ago.

-- glen

glen herrmannsfeldt wrote:

Even more interesting, this question came up a month or two ago.

-- glen

I have looked back through all the topics and found "Ethernet Breaker"
which seems to be what I need. Now I know that there is a transformer
at each end then I think I am OK just shorting the TX and RX pairs.
Mechanically I think I'll start with a standard cat5 cable, remove an
inch or so of the outer sheath in the middle, untwist a bit of the
orange and green pairs and use a 4-way 0.1" IDT connector to give me a
tap that can be plugged into a bit of PCB with the relay on it. It
would be nice to run 100BaseT through this, but it only for a PC to
connect to the internet via broadband at 512k so I can actually get
away with 10BaseT. I will report back here when I get something
working (or give up) in a week or two.
Thanks to you all.

Howard wrote:

glen herrmannsfeldt wrote:

Even more interesting, this question came up a month or two ago.

I have looked back through all the topics and found "Ethernet Breaker"
which seems to be what I need. Now I know that there is a transformer
at each end then I think I am OK just shorting the TX and RX pairs.
Mechanically I think I'll start with a standard cat5 cable, remove an
inch or so of the outer sheath in the middle, untwist a bit of the
orange and green pairs and use a 4-way 0.1" IDT connector to give me a
tap that can be plugged into a bit of PCB with the relay on it. It
would be nice to run 100BaseT through this, but it only for a PC to
connect to the internet via broadband at 512k so I can actually get
away with 10BaseT. I will report back here when I get something
working (or give up) in a week or two.

For 10baseT that should be fine. For 100baseTX you will need to be
a little careful with the wiring, such as trying to maintain 100 ohm
impedance through the PC board. It should then be done with two
connectors, such that the signal goes into the PC board, past the relay,
and then out again, maintaining 100 ohms as much as possible.

For 10baseT you don't need to be quite that good, but try to minimize
the length of the PC traces from the connector to the relay pins.

-- glen

Howard wrote:

I will report back here when I get something
working (or give up) in a week or two.

It works! I have tapped into a 5m cat5 cable with a 0.1" IDT connector
onto a bit of cheap Vero bread board with a cheap double pole relay
close to the connector. I didn't bother with using two poles together
to keep the data off the moving contact. I can sucessfully run
100BaseT over the link and when the relay contacts close the link leds
at both ends of the cable go out, and as soon as the contacts open
again the link leds come back on. I have tried to determine if the
relay has any effect on the link, such as an increased error rate but I
can't see any difference in the time taken to throw big files around
the network. I have tried a demo version of "Colasoft Capsa 5.0
Enterprise" and it doesn't show any hardware errors on the network with
or without the relay in place. Does anyone know of a suite of test
software specifically for this sort of hardware testing? I have sent
the prototype to my customer, so we shall see what he thinks. Thanks
to you all.

"Howard" <h.shaw@dmkent.co.uk> wrote:

It works! I have tapped into a 5m cat5 cable with a 0.1" IDT connector

Hooray!

it doesn't show any hardware errors on the network with
or without the relay in place. Does anyone know of a suite of test
software specifically for this sort of hardware testing?

Umm, you _REALLY_ need to certify this device to CAT5 so that it'll
work on any network, not just the network you set up in the lab. Bone
up on the Cat5 specs and rent a certification tool!