EEVblog Electronics Community Forum
Electronics => Beginners => Topic started by: rakeshm55 on March 26, 2018, 12:20:43 pm
-
On of our product requirement is to have one external RJ45 connector (RJ45+ Magnetics) and this should communicate with two different Ethernet modules (PCBs) ie two different PHYs... At a time only one module will communicate with the port and the other remain idle...
One module act as the Master . Default access to RJ45 is for Master. Once the master is through with the communication it can free-up the port and inform the other module to start communication . There is a provision to inform the other module via another interface(UART).
is this a reliable implementation?? Is this switching possible.? Any chipset for doing this?? What could be the pitfalls given that Ethernet is a complex protocol??
-
You will need a switch. Lucky for you there are IC's capable of skipping the PHY and directly talking RMII for a few ports.
But, ethernet switches are high speed and specific parts. You might be better of just buying a switch if the enclosure accepts it.
KSZ8775CLX
-
I hav a battery powered project so using an ethernet switch may a tough choice to consider.... I came to know of TS3L110 ...10/100 ethernet analog switch...will it be solution??
-
I hav a battery powered project so using an ethernet switch may a tough choice to consider.... I came to know of TS3L110 ...10/100 ethernet analog switch...will it be solution??
Analog switches can be used for Ethernet indeed, but you have to take care about ESD and surge protection. On safe side I would pick purpose-built mux like MAX4927.
-
I hav a battery powered project so using an ethernet switch may a tough choice to consider.... I came to know of TS3L110 ...10/100 ethernet analog switch...will it be solution??
That TS3L110 is just a "dumb" 4PDT switch. It is not a "switch" in Ethernet terms. Unless you want to write the code to perform Ethernet packet switching.
I had this same question a couple years ago and someone suggested a chip from NXP that had a phy port (for your local gadget) and two 10-base T ports for proper RJ45 connections to the outside world. Perfect solution for what I was looking for. But now I can't remember the name of the chip?
-
There are a lot of ethernet switch chips, that have multiple PHY-less RMII/MII/RGMII interfaces and built-in PHYs as well. AFAIK the PHYs are power hungry when link is up; multiple (R)MII interfaces without PHYs are not that bad. If the modules are your design and you can wire up them to the switch directly without PHYs, it would be a viable solution.
-
That TS3L110 is just a "dumb" 4PDT switch. It is not a "switch" in Ethernet terms. Unless you want to write the code to perform Ethernet packet switching.
Right. It is ok to switch ethernet analog signals, which is more or less like reconnecting cable. That's what mentioned MAX chip is used for, mostly for laptops to switch between docking station and on-board port. On the other hand using PHY-less simple switch IC would be much more desirable because it is "always on" solution that does not have PHY sync/negotiation and MAC table reset delays.
[edit] OP shall note that using analog mux between ports can have MAC table timeout penalty up-to 300 seconds unless you properly reset both interfaces in question. Established TCP/IP connections can timeout in result. That's why proper Ethernet switch is better solution.
-
1. One Ethernet PHY is connected to a Secure device which is a non OS and device may use low layer ethernet communication
2. Second Ethernet PHY connected to a non Secure device with OS. Will be using TCP/IP
Earlier it was intended to have two RJ45 connectors for each of them. Later decided to combine becoz both of them are not simultaneously intended to communicate. So using an analog switch is as equivalent as, plug OFF and re-plug to a different port...
-
Yes this could work however you will have to take into account the rest of the network , eg DHCP server renegotiations etc., also depending on the PHY could take some time. So a time penalty of a few minutes before the second link is up and running again after the "hard" switch should be considered in the design.