0
\$\begingroup\$

Daisy Chaining

MagJack Diagram

EDIT: The Microchip Ethernet controller only supports 10/100 Mbs

I have verified on my bench that two Magjacks with magnetics built in, and hard wired together, works on the bench using just the communication lines as a bypass to simulate a POWER LOSS from a device with dual ports.

Currently I am using only ONE port. But I want to add dual ports and a switch for daisy chaining and connecting the embedded switch in strap mode so won't have to be managed by the MCU, and KISS concept of not having to implement the STP or RSTP spanning tree protocol.

That way if a daisy chain of devices is used on a network, and one device loses power, that the rest of the networked devices down the line will still have communications.

So if using a "mechanical signal relay" with normally closed NC contacts is used, and there is power, the contacts/switches will be open allowing normal communication. When power is lost, the relay defaults back to its default state of Normally closed NC and hard bypassing the communication lines to bypass the device with lost power so communications can still be maintained to the following devices in the daisy chain.

Pin Connections:

Mag 1 Mag 2
TD+ TD+
TD- TD-
RD+ RD+
RD- RD-

Using simple jumpers, I hard connected the MagJacks pins together. I plugged an ethernet cable into a basic consumer Netgear 5 port switch on my network. Plugged the other end into one side of the "MagJacks hard wired input", and then plugged another ethernet cable into the other end of the "MagJacks hard wired other input" to my device with embedded TCP/IP Server that uses both TCP/IP and UDP.

Communication to Web server works like a charm with the magjacks hardwired.

I know there will be cable length limits, but I'm no networking engineer.

Does anyone have any input as to any pros and cons on network cable losses etc using the MagJacks with the built-in magnetics? Or should the magnetics be separate from the ports? With board space, the built-in magnetics help. But I'm not sure if I should use a separate RJ45 connector and separate magnetics/transformer with a true hardware bypass.

There will also probably be a limit of devices that can be in the chain also.

Thanks for any pros and cons feedback anything you could think of. The concept works, but again I am no networking IT expert engineer and any feed back on pros and cons any why's would be greatly appreciated.

\$\endgroup\$
3
  • 1
    \$\begingroup\$ Ethernet signals (especially gigabit) are high frequencies, so you might want to get an "RF relay" with good bandwidth and low loop area. \$\endgroup\$ – user253751 Feb 3 at 22:52
  • 1
    \$\begingroup\$ I remember someone using this concept to implement "movie-style hacking" - he was able to tap into an Ethernet security camera cable and replace the camera feed with a loop, without breaking the connection. All just for a demo, of course :) \$\endgroup\$ – user253751 Feb 3 at 22:53
  • \$\begingroup\$ The Microchip Ethernet controller only supports 10/100 Mbs \$\endgroup\$ – blackedpi Feb 3 at 22:57
1
\$\begingroup\$

This is a known technique - see for example https://www.win-ent.com/1U-Rackmount-Platforms/pl-80310.

My understanding is that it is done right at the RJ45 connectors, rather than at the PHY output. To save cost on a 1Gbps port you may choose to only bypass the two pairs used for 10/100mbps, leaving the connection to reestablish at the lower speed.

\$\endgroup\$
4
  • \$\begingroup\$ you mean like the top picture in this link ? link \$\endgroup\$ – blackedpi Feb 3 at 23:03
  • \$\begingroup\$ I found that picture in a masters thesis I found online. Some one else mentioned it wouldn't work. But will try and remove some jumpers from the MagJack and see what happens. Just standard 10/100Mbs speed though. link \$\endgroup\$ – blackedpi Feb 3 at 23:07
  • 1
    \$\begingroup\$ OK, now I see that you are using RJ45 ports with integrated magnetics. In the case of a separate RJ45 connector and PCB mount magnetics, it would typically be done between the RJ45 connector and the magnetics. I haven't seen it done on the output of the PHY, but in principle it could work. \$\endgroup\$ – elchambro Feb 3 at 23:10
  • \$\begingroup\$ It does work with the integrated magnetics. But i've only tested it with one device. So the ethernet cable is plugged into the hard wired MagJacks (simulated device 1) from a netgear consumer switch and then to my device (device 2) similar to the top picture as last device 2. All works. But again just one device. I agree that separating the magnetics from the RJ45 connector would be ideal, and bypass just the RJ45 connectors without magnetics in the loop would be better, but Trying to limit PCB space. \$\endgroup\$ – blackedpi Feb 3 at 23:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.