Slightly OT: Disconnecting Ethernet

I have an application in which I need to disconnect an ethernet channel most of the time unless there is a special set of circumstances. I have a processor to control the disconnection, and access to the ethernet (CAT5) cable and its conductors.

The application is in a pressure environment and that precludes me from using the obvious relays on the 4 cores of the cable. I have no software access to the routers on either end of the cable, so I can't simply program them to turn off the channel. So, I'm looking for some way of disabling the channel by only accessing the transmitted signals.

To put some context: We have a subsea application in which two separate ethernet circuits are used to transmit data between the surface and the equipment subsea. In normal use, one circuit is used for bidirectional control and monitoring of the subsea systems, and the other is a dedicated, one-way high-speed data link used for monitoring. In order to minimise the jitter and latencies on that link there will be only one data channel open and transmission will be as fast as possible with no other data on the link. Normally the two networks are totally unconnected at both the topside and bottom ends.

However, it's all too possible that one of the ethernet connections becomes broken. In this case we want to be able to bridge the two independant networks onto the remaining wire, albeit with some degradation in the link speeds. On a surface system, this would mean that we simply unplug the bottom end ethernet cable from its router and plug in into the other network, and do the same thing at the top end, so that both networks share the same bit of wire for the uplink. Everything should then work as before.

However, we can't find any volunteers to replug the networks in the subsea end of the link, so we need to make the switch automatically. That'd be easy enough in a 1-bar environment, because we'd just use changeover relays in the cable. However the whole system's running in an oil- filled box at whatever the pressure is at depth, which could be up to 600MSW (or 60 atm). So, I need to use solid state stuff to do the switching. I don't think that I can just stick semiconductor switches in the circuit either because they have quite a large on- resistance (although I've not tried it).

I was wondering if anyone here had any ideas?

Trev

--
Trevor Barton
Isotek Electronics Ltd, 9 Clayton Wood Bank, Leeds, LS16 6QZ, UK.
 Click to see the full signature
Reply to
Trevor Barton
Loading thread data ...

bidirectional

dedicated,

Probably you could use semiconductor switches - many FETs have on-resistences well below 1 Ohm.

The signal on a twisted-pair Ethernet is a little less than 1 V on 100 to

150 Ohm impedance. The network is transformer-isolated from the endpoint electronics. IMHO, the greatest problem may be isolating the control signals of the FET switches. If there are no significant other voltages to block, the rest should be pretty straightforward.

HTH

Tauno Voipio tauno voipio @ iki fi

Reply to
Tauno Voipio

I think Fairchild or ADI may have had an app note or a flier on using analox MUXes for routing Ethernet signals in things like laptops docks/mini-docs.

Their idea was something along the lines of a user might want to have an Ethernet port on the laptop itself, but also have the dock wired to Ethernet so when they reach "home base" they could dock the laptop and not have to mess with hooking up connections.

The point however is that they were using analog muxes with BW upto

350MHz for switching Ethernet signals...you could probably do the same to connect/disconnect your device.

-- Jay.

Reply to
Jay

I think that you could do it with FET switches, which will easily go sub-ohm for about 30 cents. These are designed for switching power, but they should work here.

But, also, I haven't thought it though, but it just seems like you could do this with an Ethernet switch chip in the undersea unit. Broadcom sells these, for maybe $10. Open op a commercial switch/hub and you'll see it in there. It could possibly be configured to do what you want as well.

Reply to
Ian McBride

"Ian McBride" schreef in bericht news:sQwib.44078$ snipped-for-privacy@newsread1.news.atl.earthlink.net...

subsea

sub-ohm

should

Yes, but these powermosfets all have internal (reverse) diodes. The fets that don't often have Rds-on >25 ohm.

[snip]
--
Thanks,
Frank Bemelman
 Click to see the full signature
Reply to
Frank Bemelman

You can use 2 FET's back-to-back. This will conduct in both directions if driven high enough and not conduct at all when switched off.

For the OP: If the oil is totally non-conductive, it might be possible to immerse non-sealed relays in it completely. And how presure resistant are reed relays? The glass tubes are soo small that they might withstand a lot.

--
Stef
Reply to
Stef

Fantastic, the url is

and it's pretty much what I want! Thanks very much.

Trev

--
Trevor Barton
Isotek Electronics Ltd, 9 Clayton Wood Bank, Leeds, LS16 6QZ, UK.
 Click to see the full signature
Reply to
Trevor Barton

Yes, you can use relays and they work for a while, but they are not very reliable long-term. We have done it in the past with normal plastic encapsulated relays by drilling small holes in the plastic case or removing it altogether, but they soon become unreliable.

I think that there are a couple of problems. It'd difficult to ensure that the oil is always clean in the environment they are used in, the equipment can be pulled apart on deck in a force 9 gale for diagnostic work and reassembled in less than ideal conditions. Also, I suspect on low power relays the oil ends up gunging up the workings and probably attacks the plastic parts at least at the interface level which makes tham sticky, and they don't have a sufficiently large pull-in force to overcome that. They work to start with but after a few weeks or a year or so they just stop working.

I don't know about reed relays. We use light bulbs in pressurised gas up to about 18 bar subsea, the small 12V halogen bulbs and they are OK it seems but they are replaced regularly (after only a few

10s of hours use). I *guess* reed relays are probably OK, especially if they have secondary encapsulation (eg epoxy or something) but it's be only a guess and I'd shy away from it if at all possible!

Anyway, I'm looking at the Fairchild device now that seems to be made for just my purpose!

Cheers, Trev.

--
Trevor Barton
Isotek Electronics Ltd, 9 Clayton Wood Bank, Leeds, LS16 6QZ, UK.
 Click to see the full signature
Reply to
Trevor Barton

... snip ...

Why can't you drive both links through a virtual driver, and do the switching entirely in software?

--
Chuck F (cbfalconer@yahoo.com) (cbfalconer@worldnet.att.net)
   Available for consulting/temporary embedded and systems.
 Click to see the full signature
Reply to
CBFalconer

I have been pondering (though not very seriously) a slightly similair problem for a while now. Seeing as the topic is on the table I thought I may as well ask and see what gets suggested :

I want to be able to have two CAT5 cables that are normally independent of each other coming into my device. E.g. in a VOIP telephone with a two (three) port ethernet switch. However, when power is lost on the device, I want the two cables to be bridged, and still support high speed comms :

cable from hub device with 2-port ethernet switch PC

but when power to device goes down, or other fault :

cable from hub [low resistance through-connection requiring no power] PC

obviously one would have to take into account resultant cable lengths, particulalry given that the analogue switch probably has worse impedance problems than a normal twisted cable .

The "requiring no power" (i.e normally closed) part is the biggest problem to me....

Any suggestions ?

Reply to
Simon Berry

'Cause although I have a processor, I'd need 4 ethernet ports (two to connect to the bottom end units, and two to connect to each of the uplinks). That's problematic.

Trev

--
Trevor Barton
Isotek Electronics Ltd, 9 Clayton Wood Bank, Leeds, LS16 6QZ, UK.
 Click to see the full signature
Reply to
Trevor Barton

It shouldn't be. A simple mechanical relay with "opener" switches should do that for you. I.e. a relay switch that is closed by default, and opened by applying current to the magnet. May not be the most reliable or long-term stable solution in the world, and it's everyone's guess what happens if you try to transmit 1000baseT (gigabit ethernet) over it --- but simple it definitely is.

--
Hans-Bernhard Broeker (broeker@physik.rwth-aachen.de)
Even if all the snow were burnt, ashes would remain.
Reply to
Hans-Bernhard Broeker

A NC relay? If you've no special environmental concerns a normally closed relay connected to your power supply will do that, or to a digital output of you want to more intelligently turn it on (eg after your device has booted). I doubt the effect on the circuit impedances would be more notifable than an RJ45 connector/wall connector/crap homemade wiring behind the trunking.

Trev

--
Trevor Barton
Isotek Electronics Ltd, 9 Clayton Wood Bank, Leeds, LS16 6QZ, UK.
 Click to see the full signature
Reply to
Trevor Barton

And just to follow up my post there are "lots" of analogue switches that threaten to be able to do ethernet switching in their app notes. They all suggest the switch side of the magnetics, which is probably OK for me, but I'm sure that with a bit of though it'd be easy enough to do on the isolated side of the magnetics with appropriate isolation.

Analog ADG774 for example.

Duh, I don't know why I never though of analogue switches myself, it's bleedin' obvious when someone points it out to you! Thanks everyone for the input!

Cheers, Trev

--
Isotek Electronics Ltd, 9 Clayton Wood Bank, Leeds, LS16 6QZ, UK.
Tel: +44 (113) 275 1339, Fax +44 (113) 224 9827
 Click to see the full signature
Reply to
Trevor Barton

Rather than switching the Ethernet, why not switch the MII interfaces to the PHYs? These are digital signals and can be easily switched with normal buffers, transceivers, or quickswitch parts.

If you were planning to use integrated MAC/PHY parts, you will have to buy separate PHY chips, but they aren't too expensive.

Much better than using relays, FETs, etc. on the analog side, IMHO.

Is there some reason why *software* can't control which Ethernet is used?

Reply to
Eric Smith

Yes, it's less to do with the cost than the board acerage. I need 8 channels on one side and probably the same number of the other. Although, of course, the magnetics are fairly large.

Not any electronic reason, but a practical one; I can save hours and hours of software development time by removing the need for someone to learn how to drive a switch through its serial interface by reducing the control to a single digital output. By the time someone has trawled through and understood the datasheet, written the code to drive whatever the serial interface is, debugged and tested it, we'll have used about a week of software effort. Now, use the frig factor someone posted on here a while ago (add one and increase to the next time unit) that maps to about two months of software development. If I wanted to do anything more complicated, I agree, the software route is the correct way to go, but to just switch an input on or off it's overkill.

Trev

--
Isotek Electronics Ltd, 9 Clayton Wood Bank, Leeds, LS16 6QZ, UK.
Tel: +44 (113) 275 1339, Fax +44 (113) 224 9827
 Click to see the full signature
Reply to
Trevor Barton

You're very welcome!

Please let us know how it turns out.

Best regards.

-- Jay

Reply to
Jay

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.