Want to physically disconnect Ethernet on/off - design recommendations needed

[jsquaredz]

I am looking to build something I can plug ethernet into one side, and out of the other side, and be able to physically isolate it with a remote signal.

Only need help on the piece that is isolating disconnecting the ethernet.

Was thinking of using a few of these. http://www.semicon.toshiba.co.jp/info/docget.jsp?type=datasheet&lang=en&pid=TC7S66F
Any reason it wouldn't work?

Open to ideas.  How is this done in a managed switch? Is there a better IC for this job?

[c4757p]

So by 'isolate', you mean 'disconnect', right? Not like high voltage isolation?

[jsquaredz]

So by 'isolate', you mean 'disconnect', right? Not like high voltage isolation?

Correct.

[free_electron]

dual pole relay ? simply cut the rx lines

[Richard Crowley]

Nowhere near enough details to give a definitive response, but I would have very serious doubts whether two (or four, as needed) of those little things would adequately conduct even 10BASE-T, much less any higher speed, and certainly not POE.  In particular there are issues of current capacity, and perhaps even more important are impedance and frequency response to maintain a network connection.

Since we know nothing about the motivation here, we must ask why is such an unusual method required when there are more conventional methods of logically controlling an ethernet connection.

Without any more details than you have offered, I would have to say it is extremely doubtful that will work reliably.

[dom0]

Keystone panel on one side, piece of patch cable on the other with the latching tab removed. A servo / stepper motor inserts or removes the plug from the keystone panel when commanded to do so.

[jsquaredz]

Keystone panel on one side, piece of patch cable on the other with the latching tab removed. A servo / stepper motor inserts or removes the plug from the keystone panel when commanded to do so.

LOL.   

[Muttley Snickers]

You might try a separate switch or hub on that branch that you want to disconnect and use your rf relay to cut power to that switch or hub.

Muttley

[jsquaredz]

Nowhere near enough details to give a definitive response, but I would have very serious doubts whether two (or four, as needed) of those little things would adequately conduct even 10BASE-T, much less any higher speed, and certainly not POE.  In particular there are issues of current capacity, and perhaps even more important are impedance and frequency response to maintain a network connection.

Since we know nothing about the motivation here, we must ask why is such an unusual method required when there are more conventional methods of logically controlling an ethernet connection.

Without any more details than you have offered, I would have to say it is extremely doubtful that will work reliably.

Just looking for a way to electrically disconnect an ethernet (100base T) from an end user that can be controlled with a signal.  I am looking at other solutions to my problem in parallel, but I wanted to see if this option was feasible. 

[jsquaredz]

You might try a separate switch or hub on that branch that you want
to disconnect and use your rf relay to cut power to that switch or hub.

Muttley

Yes, this is one of the options I am exploring, but ideally I would just have a single board that would have an Ethernet in and Out that I can electrically disconnect with a singal from my MCU.

[Richard Crowley]

Just looking for a way to electrically disconnect an ethernet (100base T) from an end user that can be controlled with a signal.  I am looking at other solutions to my problem in parallel, but I wanted to see if this option was feasible.

100BASE-T is a quite sophisticated and precise design standard. It seems very doubtful that those little CMOS switches are up to the task of adequately passing the signal while maintaining the series resistance, impedance, and frequency response necessary to support 100BASE-T.  You would be more likely to find success with a relay.  Although any mechanical methods seem very crude in such a sophisticated design as a LAN connection.

There may be even a dozen ways to "disconnect" a network connection LOGICALLY without any physical components.  But since we don't see the big picture here we have no way of knowing which might be practical?

[Muttley Snickers]

Sorry, but your first post said remote signal not remote command, silly me...

[PeterFW]

Open to ideas.  How is this done in a managed switch? Is there a better IC for this job?

With a relay, or multiple relays depending on the contacts.
I horribly botched together cables and had not problems on 50 meters worth of unshieldet cable that i would not expect the relay to produce much of a problem.

You only need four conductors annyway for 100Mbit, the other two pairs are not used.

[jsquaredz]

Open to ideas.  How is this done in a managed switch? Is there a better IC for this job?

With a relay, or multiple relays depending on the contacts.
I horribly botched together cables and had not problems on 50 meters worth of unshieldet cable that i would not expect the relay to produce much of a problem.

You only need four conductors annyway for 100Mbit, the other two pairs are not used.

Any recommendations on a relay(s) for this?  Needs to be pcb mounted.

[SeanB]

4 reed relays, either as 4 separate relays or as 4 reeds with a common coil wound around them all. Good isolation, and low capacitance and will not interfere much with the signal when on. Control will have to be either a separate port with some sort of network switch ( arduino with ethernet shield or a rasberry Pi using one of the GPIO lines) to switch on command.

[Richard Crowley]

Any recommendations on a relay(s) for this?  Needs to be pcb mounted.

You will need TWO of these (or something like them)....

4.5VDC DPDT SURFACE-MOUNT RELAY  --  $1.25 each   
http://www.allelectronics.com/make-a-store/item/rly-561/4.5vdc-dpdt-surface-mount-relay/1.html


They are cheap enough that you aren't wasting much $$ if they don't work.

[jsquaredz]

Any recommendations on a relay(s) for this?  Needs to be pcb mounted.

You will need TWO of these (or something like them)....

4.5VDC DPDT SURFACE-MOUNT RELAY  --  $1.25 each   
http://www.allelectronics.com/make-a-store/item/rly-561/4.5vdc-dpdt-surface-mount-relay/1.html


They are cheap enough that you aren't wasting much $$ if they don't work.

Looks good.  What contacts should I be breaking?  Just the TX for each pair?

[PeterFW]

4 reed relays, either as 4 separate relays or as 4 reeds with a common coil wound around them all.

Yes, a reed relay, not a "normal" one, would be my choice as well.
They are sealed, compact and designed to run of 3-5V depending on the model.
You get them in SMD or THT, along with a transistor, diode and resistor that is everything you need.

Something like this:
http://static.rapidonline.com/catalogueimages/Module/M029535P01WL.jpg

You can get them with very high specs as well in tiny, tiny packages, it all depends on were you buy them and how much you want to spend.

[PeterFW]

Looks good.  What contacts should I be breaking?  Just the TX for each pair?

That is not a sealed relay, at least it does not look like one.
See my other reply.

[Richard Crowley]

None of the other components (connectors, transformers, etc.) in the 100BASE-T circuit path are "sealed".
It is not clear what is the motivation for a "sealed" relay?

The TX is one pair (pin 1 & 2), and the RX is another pair (pin 3 & 6).  They are differential/balanced signals.
To make a complete break, you would interrupt BOTH pair.
Although it would disrupt communication to break just one of the pair of wires for TX and for RX

[DanielS]

If the intent is to disable a port for administrative reasons rather than electrical ones, I would simply use a managed switch and enable/disable the port through SNMP or the web-UI.

[Richard Crowley]

If the intent is to disable a port for administrative reasons rather than electrical ones, I would simply use a managed switch and enable/disable the port through SNMP or the web-UI.

But that is exactly what managed hubs are made to do. You can enable/disable ports over the network. No fooling around with extra hardware.

[PeterFW]

None of the other components (connectors, transformers, etc.) in the 100BASE-T circuit path are "sealed".

It is not clear what is the motivation for a "sealed" relay?

Because the signal does not carry that much power, not that much wetting current.

He does not say were it will be used (hot, cold, dry, wet) how often/quick it will be switched.
The additional cost is neglectable in this case since you get a contact surface that will stay much longer clean then a standard relay for just a bit more money and board space.

Or in short: Way longer lifetime and less problems with contact wear/corosion.

[agehall]

But that is exactly what managed hubs are made to do. You can enable/disable ports over the network. No fooling around with extra hardware.

Not to be picky, but I don't think you can find a managed hub these days. Not even sure if there ever were managed hubs. I have not seen a hub in years...

[Richard Crowley]

Not to be picky, but I don't think you can find a managed hub these days. Not even sure if there ever were managed hubs. I have not seen a hub in years...

Perhaps we are now using managed switches. Seems like a semantic distinction from this distance.
But when I searched for managed hub on Ebay, I got a long list of available gadgets.
Most of them cost far less than @DanielS will spend to cobble together his kludge.