I'm building a manual switch to cut off a 10G ethernet connection. At the moment, I'm leaning toward using a Grayhill 71B30-04-2-03N (4 decks, 2 poles per deck). It's just a stab in the dark, but I thought I'd try running the twisted pairs through separate decks for some measure of isolation.

I can't find many examples of other available manual switches that are spec'd for 10G.

Any clues whether this might work? Or suggestions what might work better?

Thanks!

 

I think the frequencies involved will squirt rigth trhough that kind of switch. There must be extremely low capacitance between the contacts to act as an open switch.

You need a switch with a very long throw like was used in old video games to switch between the game and the antenna.

There are probably switches available but I expect them to be pricey. I don't have time to check right now.

 

Ask yourself why this product is not available at your local Radio Shack store.

It's not going to work well.

 

Agreed that the conventional manual ethernet switches won't work well. I bought one such, which was functional up to maybe 10mbps, barely 100mbps, but there was no shielding & certainly no pair twisting involved...

@RichardO I'm surprised that there might be signal bleeding across an open contact like that... I guess I've only considered problems maintaining a connection, not breaking one. I hate to second guess, but for such a low voltage is it really possible for the signal to reliably jump even a 1mm pcb gap (i.e. to maintain a valid ethernet connection)?

@Sensacell thx for your reply, I understand there are challenges. Do you think there's no creative way to overcome them?

How about pair isolation? eg. four separate 2 pole switches with shielding inbetween (synchronized with manual actuation)? Would that not satisfy shielding requirements?

thanks!

 

@RichardO I'm surprised that there might be signal bleeding across an open contact like that... I guess I've only considered problems maintaining a connection, not breaking one. I hate to second guess, but for such a low voltage is it really possible for the signal to reliably jump even a 1mm pcb gap (i.e. to maintain a valid ethernet connection)?[/QUOTE]
At 100 Mb/s only a few pf of capacitance is enough to let the signal through. The switch you are looking at is likely to have more than that between the contacts when they are open.

Here is an example of what a typical rotary switch looks like inside:

 

At 100 Mb/s only a few pf of capacitance is enough to let the signal through. The switch you are looking at is likely to have more than that between the contacts when they are open.

Here is an example of what a typical rotary switch looks like inside:

View attachment 138147

Thanks so much for the info... if you were selecting a manual switch for 10G, is there a type of 2 pole switch that comes to mind that might better fit the bill?

 

@RichardO Or to ask the question the other way, is there a reference where I can look up the capacitance of switches?

My mental framework has a standard Cat6 jack/plug in mind--where it doesn't take much separation of either to sever an ethernet connection. This makes me think that finding a suitable switch can't be too hard... (?)

 

@RichardO Or to ask the question the other way, is there a reference where I can look up the capacitance of switches?

My mental framework has a standard Cat6 jack/plug in mind--where it doesn't take much separation of either to sever an ethernet connection. This makes me think that finding a suitable switch can't be too hard... (?)

Look at specifications for RF (Radio Frequency) switches and relays. Normal switches do not have capacitance spec's.

At Ethernet frequencies, everything is hard.

Another difficulty that has not been mentioned is proper signal termination. That is a subject all by itself.

 

Any switch that can do this at 10G will be terribly expensive. But I see another much cheaper option - a mechanical rig with a servo that physically unplugs the connector from the switch and moves the plug to whatever isolation distance you desire. It is ultimately just a few 3D printed parts and one servo, it is cheap and simple, and it will maintain all the required electrical specs that you need for a proper 10G ethernet.

That is, if you want it electrically operated. If purely manual, why not just pull the plug by yourself?

 

Thanks so very much for all the replies... It's been illuminating for sure.

Looks like I'll probably go the route of using a cat6 jack/plug & rig up a sliding mechanism to do the job. Would I be right thinking that sliding the ethernet plug out about 1/2" would sufficiently isolate the terminals? This would give probably >5mm separation distance...

thanks!

 

... Would I be right thinking that sliding the ethernet plug out about 1/2" would sufficiently isolate the terminals? This would give probably >5mm separation distance...

That depends on what that separation is for. If you just want to disconnect data, then even a millimeter should be enough. If you are going for some overvoltage protection, then that depens on what kind of voltage are we talking about.

 

@RichardO - I'll probably just use an ethernet jack/plug combo to create the switch, but that leads to a second question: would it work to wire an B/C switch from a common A, by wiring two ethernet jacks (B & C) in parallel & initiating the circuit to A from either one?

I guess I'm just wondering if creating a dangling open jack (the one in parallel) on a Cat 6a circuit is going to cause problems?

thanks!

 

I guess it would work on 100mbit, it might work on 1G if you are careful and keep the wires short, but I dont think it is possible on a 10G link.

 

thanks @kubeek ! I'm stuck in a mental quandry. I look at the circuit having a parallel dangling open jack/plug & it seems unlikely to work well. Then I think of a closed jack/plug and I wonder where's the big difference (!)

Any guesses if using a dangling open plug instead of a dangling open jack would improve the design (more insulation in the plug--jacks seem kind of open when they don't have plugs in them)?

 

I doubt plug or jack makes any difference. The problem is in the T connection of the wires and resulting impedance mismatch and unterminated end reflections.

Why dont you describe why exactly do you need this arrangement? This is just a lot of guesswork and not enough information.

Could you use some software solution for the changeover, like a managed switch or router and have just a single disconnect? Where do you even have a 10G metallic link that needs such a weird treatment? Why dont you use fiber optic instead?

Please try answering all my questions, i believe it would be a tremendous help for finding a proper solution for your actual problem, and not band-aiding what you see as the only solution.

 

Why dont you describe why exactly do you need this arrangement?

I have an isolated 10G NAS in my shed that needs to be disconnected when not in use (I'm horrible with crawling under my desk to connect/disconnect an ethernet jack--a recipe that mean's I'll eventually forget when it's connected). Being 10G, I'd prefer not to budget for a switch, and for my uses it's not necessary. The NAS is connected directly to my win10 computer via a second 10G ethernet card in the win10 machine. The consequence is that the NAS in my shed doesn't have a gateway to the internet, and so no easy way to update software. I could probably configure software connection sharing in the win10 machine somehow, but I like simple solutions. I would rather just switch the 10G non gateway ethernet cord at my desk for the household 1G network cord. This would provide network & internet access, so I can update the software on the NAS.

Could you use some software solution for the changeover, like a managed switch or router and have just a single disconnect?

Not really, at least not without a switch (10G switches are not in my budget).

Where do you even have a 10G metallic link that needs such a weird treatment?

It's the link I have to my NAS in my shed. I like quick file backup solutions, so 10G enables me to compare/sort/merge files on 5 mirrored HDDS concurrently & backup time is reduced to <10 mins (I have >6TB of data to account for)... I delay backups all too easy, so keeping them fast means it's never burdensome and it just tends to get "done." Using other solutions, I can delay weeks on end & jeopardize files.

Why dont you use fiber optic instead?

Haha, it's purely price point for me. I can set up 10G ethernet for <$250. When I did the fiber optic price comparison, it looked prohibitive for a 75 foot connection... plus, I didn't want to convert my home network to fiber optic, so this seemed more simple...

 

I could probably configure software connection sharing in the win10 machine somehow, but I like simple solutions.

You can setup a bridge between the 1G and 10G connections in windows in about five clicks. Then the NAS can access your lan and you need to pull just one wire to do the whole process.