Two SBC's sharing one ethenet port?

[Theguywhobea]

Trying to figure out if something is plausible at work. I have one SBC (lets consider it to be a Nano Pi Neo) which has it's own ethernet hardware and RJ45 port. Then I have an electrical meter, that also has it's own hardware and RJ45 port. If I de solder the RJ45 port on the Nano Pi Neo, and then solder wire from where the RJ45 port was, to the pads of the electrical meters RJ45 port, will that let the Nano Pi Neo and the electrical meter talk with each other? Or does it not work like that? And if they could still talk to each other, If I plugged the electrical meter's shared port into the LAN side of a switch with network access, would I be able to access both the meter and the Nano Pi Neo still as if they were seperate devices?

 

It seems like it would work in my head, just want to be sure before I go and ruin anything.

[WereCatf]

7 minutes ago, Theguywhobea said:

If I de solder the RJ45 port on the Nano Pi Neo, and then solder wire from where the RJ45 port was, to the pads of the electrical meters RJ45 port, will that let the Nano Pi Neo and the electrical meter talk with each other?

Depends on whether the magnetics are in the RJ45-connector or as separate components on the board. In all likelihood, they are in the RJ45-connector, in which case the answer would be no.

 

8 minutes ago, Theguywhobea said:

If I plugged the electrical meter's shared port into the LAN side of a switch with network access, would I be able to access both the meter and the Nano Pi Neo still as if they were seperate devices?

No, that's not how ethernet works.

[Windows7ge]

My initial advice that you probably expect people to say would be to just buy a network switch. 4 port 1Gbit switches come pretty cheap today.

 

Attacking your question head-on. You're asking if you can connect the two devices directly without the assistance of a switch/router(DHCP server), correct? Assuming the meter device follows standard networking protocols assign it a static IP. Assign the Pi a static IP in the same subnet. Connect the two with a standard network cable (without cutting or desoldering anything.

 

They should be able to talk to one another.

[Theguywhobea]

4 minutes ago, Windows7ge said:

My initial advice that you probably expect people to say would be to just buy a network switch. 4 port 1Gbit switches come pretty cheap today.

 

Attacking your question head-on. You're asking if you can connect the two devices directly without the assistance of a switch/router(DHCP server), correct? Assuming the meter device follows standard networking protocols assign it a static IP. Assign the Pi a static IP in the same subnet. Connect the two with a standard network cable (without cutting or desoldering anything.

 

They should be able to talk to one another.

Right, so I have made like, a 3inch log patch cable. I know they talk to each other fine just like that as I have them both at static IP's. I just want to know if it would still work if I soldered the two ports together instead of using patch cable.

 

The whole purpose of this is to try and fit this SBC inside the case of the electrical meter without having wires running in and out of it. It's just a prototype so it doesn't have to be perfect, I'm just trying to figure out how feasible some aspects are.

[Windows7ge]

1 minute ago, Theguywhobea said:

Right, so I have made like, a 3inch log patch cable. I know they talk to each other fine just like that as I have them both at static IP's. I just want to know if it would still work if I soldered the two ports together instead of using patch cable.

 

The whole purpose of this is to try and fit this SBC inside the case of the electrical meter without having wires running in and out of it. It's just a prototype so it doesn't have to be perfect, I'm just trying to figure out how feasible some aspects are.

Oh, all I can say for sure is it should.

 

The gold coating on the pins of the RJ-45 connectors are to help connectivity with the modular design. If you directly solder the copper wires to the holes on the PCB this should result in the same conductivity (if not better).

 

You will want to ensure the wires maintain the same twists (Twisted Pairs). Although the shorter the wire the less interference can be experienced causing packet loss. So you may be fine with no twists (if the wires are incredibly short). These are things you'll have to test or research.

 

You'll also have to find the pin-out of the PCB or reverse engineer the pin-out so as to know which holes go to which wires.

 

But other than those things. It should work, yes. I'd actually be surprised if it didn't.

[mariushm]

If both devices have ethernet and you're fine with 100 mbps speed, you can make a crossover cable to create a direct connection between them.

The cable needs to have T-568A arrangement on one end, and T-568B on the other end.

For 100 mbps, only 4 wires are needed (the two green wires and the two orange wires), the other 4 are not used at 100mbps and can even miss, if you want a thinner cable.

If the distance is very short, like let's say less than 1m you would probably be fine even with two twisted pairs of wires between two ethernet jacks, without the sleeve (if you really need thin)

With this crossover cable, you'd have to configure the IP address and subnet mask of each device manually, or you must configure a DNS server on the PI and give the other device an IP address automatically.

If you do it manually, you can simply set one to 192.168.0.1 and the other to 192.168.0.2 and set subnet mask to 255.255.255.0 on both.

 

If both devices support 1 gbps, technically all gigabit ethernet cards are supposed to auto detect pairs, so you could use a regular patch cable, like the one between your computer and router/switch/cable modem. The ethernet chips should - in theory - pick the wires and talk to each other without your intervention, and you'd get 1 gbps link.

However, I wouldn't just hope it's the case, unless you really need the transfer speed, I'd rather force something like this to 100mbps and use proper crossover cable.

 

[Theguywhobea]

3 minutes ago, mariushm said:

If both devices have ethernet and you're fine with 100 mbps speed, you can make a crossover cable to create a direct connection between them.

The cable needs to have T-568A arrangement on one end, and T-568B on the other end.

For 100 mbps, only 4 wires are needed (the two green wires and the two orange wires), the other 4 are not used at 100mbps and can even miss, if you want a thinner cable.

If the distance is very short, like let's say less than 1m you would probably be fine even with two twisted pairs of wires between two ethernet jacks, without the sleeve (if you really need thin)

With this crossover cable, you'd have to configure the IP address and subnet mask of each device manually, or you must configure a DNS server on the PI and give the other device an IP address automatically.

If you do it manually, you can simply set one to 192.168.0.1 and the other to 192.168.0.2 and set subnet mask to 255.255.255.0 on both.

 

If both devices support 1 gbps, technically all gigabit ethernet cards are supposed to auto detect pairs, so you could use a regular patch cable, like the one between your computer and router/switch/cable modem. The ethernet chips should - in theory - pick the wires and talk to each other without your intervention, and you'd get 1 gbps link.

However, I wouldn't just hope it's the case, unless you really need the transfer speed, I'd rather force something like this to 100mbps and use proper crossover cable.

 

Yeah I only need very small amounts of data to transfer back and forth so 100Mbps is fine, so I'll probably do that. I would expect the length of cable I would be using would be between 3 and 5 inches, since I would ideally have these two as close a possible, likely with their PCB's back to back, separated with a thin piece of plastic to insulate them.

[Theguywhobea]

8 minutes ago, Windows7ge said:

Oh, all I can say for sure is it should.

 

The gold coating on the pins of the RJ-45 connectors are to help connectivity with the modular design. If you directly solder the copper wires to the holes on the PCB this should result in the same conductivity (if not better).

 

You will want to ensure the wires maintain the same twists (Twisted Pairs). Although the shorter the wire the less interference can be experienced causing packet loss. So you may be fine with no twists (if the wires are incredibly short). These are things you'll have to test or research.

 

You'll also have to find the pin-out of the PCB or reverse engineer the pin-out so as to know which holes go to which wires.

 

But other than those things. It should work, yes. I'd actually be surprised if it didn't.

Alright thanks for the help, Finding out the pinout of each side is the hard part it seems. I thought I could ohm it out with a meter, however I don't get conductivity between the pads on the PCB and a CAT5 cable with the bare wires exposed. I'll have to see if anyone has pinouts for these on the actual PCB or just trial and error a little bit.

[Windows7ge]

3 minutes ago, Theguywhobea said:

Alright thanks for the help, Finding out the pinout of each side is the hard part it seems. I thought I could ohm it out with a meter, however I don't get conductivity between the pads on the PCB and a CAT5 cable with the bare wires exposed. I'll have to see if anyone has pinouts for these on the actual PCB or just trial and error a little bit.

You can try to find a diagram or:

1. Remove the connector on the PCB (without destroying it).
2. Terminate one end of a Ethernet cable with T568-B
3. Strip the other end. Expose the copper on the 8 wires.
4. Continuity test between the pins on the PCB connector and the bare wires on the cable.

[Theguywhobea]

1 hour ago, Windows7ge said:

You can try to find a diagram or:

1. Remove the connector on the PCB (without destroying it).
2. Terminate one end of a Ethernet cable with T568-B
3. Strip the other end. Expose the copper on the 8 wires.
4. Continuity test between the pins on the PCB connector and the bare wires on the cable.

I did 2-4 with the RJ45 port still connected to the PCB. But on both my SBC and the electrical meter I had no continuity between the pins on the PCB and any of the bare CAT5 cable wires. I tested all the wires to all of the pins, none of them rang out.

[Windows7ge]

47 minutes ago, Theguywhobea said:

I did 2-4 with the RJ45 port still connected to the PCB. But on both my SBC and the electrical meter I had no continuity between the pins on the PCB and any of the bare CAT5 cable wires. I tested all the wires to all of the pins, none of them rang out.

I would wonder if the wire has a think non-conductive coating on it. I'd try to tin the copper wire with solder and try again. The continuity has to be there or else the network wouldn't work at all.

[mariushm]

No, it's most likely because the isolation transformer is inside the connector, the jack, whatever you want to call it.

It's a miniature transformer inside, saving room on the circuit board ... here's examples: mile long digikey link

 

See https://www.molex.com/pdm_docs/sd/936273608_sd.pdf for an example of such a jack

You can sort of figure the individual pairs by measuring the resistance between two pins at a time but for most jacks with two rows of pins, staggered, you should have them like in the datasheet above: