Gigabit Ethernet switch

[HertZ]

Hi everyone,

I'm designing a project for someone who want to be able to connect/disconnect an ethernet line using a dry contact entry as a command signal.
It's my first time with networking application, so even the function needed seems to be simple i've got a lot of questions.

The schematic of the design that i've come to is attached. Because i've no experience i've planned to use IC's that are designed to deal with ethernet applications. I plan to use RJ45 connectors with integrated magnetics (https://www.cuidevices.com/product/resource/crj001-ml1-th.pdf) and for switch to close or open communication line i've thinked of using a 1:2 ethernet mux (https://www.ti.com/product/TS3L501E)and using just one output and let the other free. (Does the second output signals as to be free or connected to the ground directly or via a 50 ohm trace ?).

Do you think it can work ?

Thank you for your support,
Maxime

[ogden]

Take manageable switch supporting rs232 terminal and arduino or any other tiny uC board you are familiar with. Solution could be ready during day or so.

[Terry Bites]

Do you mean no volt switching?
get an Embedded Ethernet Switch PCB.
and assert its Enable, Reset or Powerdown pin via an optocoupler.

[NiHaoMike]

Use diode switches? Small signal Schottkys would probably be the best choice. Easiest would be to use 8 dual series diodes (very common for ESD protection), midpoint to each line, cathodes commoned to each other and anodes commoned to each other. Then bias the diodes on to shunt the signal and bias them off to let the signal through.

[ajb]

Sure, you could use a mux and some additional magjacks, or an embedded ethernet switch and turn it on/off, or you could probably just use relays.  Make sure they're good low-signal types, and take care with the routing in and out of the relays, and the impact on signal integrity is unlikely to be more significant than any of the other options except the full on switch.  As a bonus you can choose your failure behavior (failing closed or open) if that matters to you.

[AaronD]

Use diode switches? Small signal Schottkys would probably be the best choice. Easiest would be to use 8 dual series diodes (very common for ESD protection), midpoint to each line, cathodes commoned to each other and anodes commoned to each other. Then bias the diodes on to shunt the signal and bias them off to let the signal through.

You mean roughly like this?:
https://sound-au.com/articles/vca-techniques.html#s9

That entire page is about analog VCA techniques (voltage-controlled-amplifier), and Figures 10 and 11 show two ways to do it with diodes.  Is that the connection and principle that you had in mind?

[AaronD]

...connect/disconnect an ethernet line using a dry contact entry as a command signal.

There are reports of early online gamers building multi-pole switch boxes for their ethernet cables.  Entirely mechanical, with an 8-gang (or 4-gang if they knew the spec) dry switch.  The idea was that they would physically disconnect themselves from the network, quickly get behind their (frozen) opponent, and re-connect.  The server would assume that it was a random glitch and instantly correct their position, hence the phrase "teleports behind you!"

10/100 was a lot more forgiving for that sort of thing than gigabit is, but maybe you can still do the same thing with transmission gates?  Often sold as "analog switches" with a digital control input for each, or sometimes ganged inside the chip.  It's not hard at all, to drive that digital input from a dry contact.  Match voltages, debounce, and you're set!

[HertZ]

Take manageable switch supporting rs232 terminal and arduino or any other tiny uC board you are familiar with. Solution could be ready during day or so.

I'm don't think i've understand your solution, why a microcontroller ?

[HertZ]

Do you mean no volt switching?
get an Embedded Ethernet Switch PCB.
and assert its Enable, Reset or Powerdown pin via an optocoupler.

Maybe it's me but i've not found any module already done for that job. It's not a an ethernet switch/router but more an alternative to a physical switch to connect/disconnect cables.

[HertZ]

Use diode switches? Small signal Schottkys would probably be the best choice. Easiest would be to use 8 dual series diodes (very common for ESD protection), midpoint to each line, cathodes commoned to each other and anodes commoned to each other. Then bias the diodes on to shunt the signal and bias them off to let the signal through.

Interesting idea, thank you i will study that.

[HertZ]

Sure, you could use a mux and some additional magjacks, or an embedded ethernet switch and turn it on/off, or you could probably just use relays.  Make sure they're good low-signal types, and take care with the routing in and out of the relays, and the impact on signal integrity is unlikely to be more significant than any of the other options except the full on switch.  As a bonus you can choose your failure behavior (failing closed or open) if that matters to you.

Thank you for your opinion. I was afraid about the power consumption using relays 

[HertZ]

...connect/disconnect an ethernet line using a dry contact entry as a command signal.

There are reports of early online gamers building multi-pole switch boxes for their ethernet cables.  Entirely mechanical, with an 8-gang (or 4-gang if they knew the spec) dry switch.  The idea was that they would physically disconnect themselves from the network, quickly get behind ...  input from a dry contact.  Match voltages, debounce, and you're set!

Thanks for your response, do you think of this https://www.renesas.com/us/en/document/dst/qs4a105-datasheet?language=en&r=13810 will it be suitable ?

[ajb]

Sure, you could use a mux and some additional magjacks, or an embedded ethernet switch and turn it on/off, or you could probably just use relays.  Make sure they're good low-signal types, and take care with the routing in and out of the relays, and the impact on signal integrity is unlikely to be more significant than any of the other options except the full on switch.  As a bonus you can choose your failure behavior (failing closed or open) if that matters to you.

Thank you for your opinion. I was afraid about the power consumption using relays 

Latching relays can be a good option to reduce power consumption (as long as you're not in a high vibration/shock environment).  They don't even need to be that complicated to drive, a push-pull output with a capacitor in series with the coil can provide the necessary pulse drive from a simple logic signal.  The only downside to that method is needing a bit of time between state changes to ensure the capacitor voltage stabilizes, otherwise there may not be enough energy to fully switch the relay, but that's pretty simple with an RC delay or a little 6-pin MCU if you prefer.

I used this method to provide fail-to-wire functionality for an (100Mbit) Ethernet switch with two external ports. Since the equipment is designed to be daisy-chained via the two external ports the switch will bypass the switch ASIC and directly connect the two external ports in case of power failure, so that the rest of the equipment can still communicate even if one unit in the daisy chain loses power (subject to length limits of course).

[ogden]

Take manageable switch supporting rs232 terminal and arduino or any other tiny uC board you are familiar with. Solution could be ready during day or so.

I'm don't think i've understand your solution, why a microcontroller ?

Manageable Ethernet switches can be commanded through rs232 terminal commands to "connect/disconnect an ethernet line". But you need "dry contact entry as a command signal". So microcontroller will input dry contact signals and translate to manageable Ethernet switch commands. That's it. Keep It Stupid Simple. No overengineering as you and many are so eager for.

[AaronD]

do you think of this https://www.renesas.com/us/en/document/dst/qs4a105-datasheet?language=en&r=13810 will it be suitable ?

Possibly.  Look at the specs, compare it to GbE, and see if "ON" is sufficient to guarantee that GbE works and "OFF" to guarantee that it doesn't.  Don't just look at a single gate, but also at the interactions between gates.  And if "OFF" doesn't work for GbE, will the system fall back on something else that still does?  There are also 8! (8 factorial, = 40,320) different ways to hook it up.  Maybe some of those work and some don't because of the interactions?  Lots of things to think about, if it doesn't explicitly say for itself that it's suitable for something.

One of the applications is "high-speed networking", but "high-speed" is meaningless without a number with units, or a specific standard that itself has numbers with units.  That phrase *might* refer to GbE, or it might refer to 10 base-T because that's faster than a 56k modem.  Read and understand the specs to be sure.

Or if it's cheap enough, just buy it and try it.

So microcontroller will input dry contact signals and translate to manageable Ethernet switch commands. That's it. Keep It Stupid Simple. No overengineering...

Except that for someone that only knows hardware, a microcontroller that isn't already programmed by someone else can represent an insurmountable barrier.  It's trivial for those of us who are already comfortable with it, so it's easy to forget that.

If I had a managed switch with RS232 input, then I'd probably go the same route.  Get a cheap 8-pin 8-bit MCU, and program it to spit out a canned phrase on the UART in response to a software-debounced button.  But the programming part turns a lot of people off.  Far more than it should, but it does nonetheless.

(If you include all of the circuit complexity on the silicon die of even the smallest and cheapest MCU, then it becomes a ridiculous example of Rube-Goldberg-ism for something like this.  But we normally don't look at that.    And I'd still use it anyway.)

[NiHaoMike]

You mean roughly like this?:
https://sound-au.com/articles/vca-techniques.html#s9

That entire page is about analog VCA techniques (voltage-controlled-amplifier), and Figures 10 and 11 show two ways to do it with diodes.  Is that the connection and principle that you had in mind?

It would be something like the second one.

[HertZ]

Latching relays can be a good option to reduce power consumption ... so that the rest of the equipment can still communicate even if one unit in the daisy chain loses power (subject to length limits of course).

Nice design , it can do the job indeed , thank you !

[HertZ]

Take manageable switch supporting rs232 terminal and arduino or any other tiny uC board you are familiar with. Solution could be ready during day or so.

I'm don't think i've understand your solution, why a microcontroller ?

Manageable Ethernet switches can be commanded through rs232 terminal commands to "connect/disconnect an ethernet line". But you need "dry contact entry as a command signal". So microcontroller will input dry contact signals and translate to manageable Ethernet switch commands. That's it. Keep It Stupid Simple. No overengineering as you and many are so eager for.

Oh yes, now I see your idea, thank you !

[ledtester]

A variant of the managed switch idea...

Use a powered switch and turn the power to the switch on and off as desired.

You can keep everything low voltage by interrupting the power at the DC input jack of the switch (like with a relay).

[HertZ]

s to think about, if it doesn't explicitly say for itself that it's suitable for something.

One of the applications is "high-speed networking", but ...
silicon die of even the smallest and cheapest MCU, then it becomes a ridiculous example of Rube-Goldberg-ism for something like this.  But we normally don't look at that.    And I'd still use it anyway.)

Ok thanks for your advices ! the two solutions are ok for me.

[HertZ]

A variant of the managed switch idea...

Use a powered switch and turn the power to the switch on and off as desired.

You can keep everything low voltage by interrupting the power at the DC input jack of the switch (like with a relay).

Why not , but power down and power up can take some seconds 

[ledtester]

Why not , but power down and power up can take some seconds 

That's true, but the original question didn't specify any timing constraints, so I thought I'd just throw it out there.

[HertZ]

Why not , but power down and power up can take some seconds 

That's true, but the original question didn't specify any timing constraints, so I thought I'd just throw it out there.

Yes completely , thanks