This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

SN3257-Q1: Can SN3257-Q1 demux the QSGMII into 4 Ethernet PHYs?

Part Number: SN3257-Q1
Other Parts Discussed in Thread: TS3L110, TMUX136

E2E, 

There is a quad SGMII port on the TDA and we are looking for a way to split the QSGMII signals into individual SGMII signals. I’ve spoken with the Ethernet team and it does not sound like we have a single device that would accomplish this, we would need to demux the QSGMII into 4 Ethernet PHYs.

 

I found 4-channel 2:1 TS3L110 online, and saw there was an automotive 2:1 4-channel device called SN3257-Q1.  Would something like this work for this application? 

Thank you,

Adam

  • Hey Adam, 

    What's the application here? Is being automotive qualified a preferred feature?

    It sounds like you'll be needing a bandwidth of 625MHz. This is the clock frequency of a single SGMII clock signal. It's a DDR protocol so the TX and RX data will also be maxed out at 625MHz. The TS3L110's bandwidth wouldn't be enough here, so I wouldn't recommend it.

    What's the setup here? SN3257-Q1 may work but I don't believe it would provide us with the correct configuration. My understanding is that the SGMII has a differential pair for TX clock and data and RX clock and data for a total of 4 pairs or 8 signal paths. I believe the clock itself may not need to be muxed but I could be mistaken. If the clock is being muxed, it sounds like you'll need 8-channels of a 4:1 to mux to split the quad into individual signal SGMII then. This could only be accomplished with multiple devices however. We won't have a 4:1 that can support the bandwidth requirements so the best solution here would be to short the outputs of multiple 2:1 devices together. 

    Unfortunately, of the 2:1 devices that can support the bandwidth requirement, none of them allow for individually disabling one of the channels in the device. So if we short the outputs, we would be shorting two signals together. So our solution here would be to use 1 channel of a 2:1 mux and leave the other channel unused.
    For example, here would be the RX differential pair data line :



    If you wanted the +RX_data1 path connected, you could select it in mux#1 and then disable mux#2. This logic would be used to select all the different paths. 
    Since we're shorting the outputs together, the off-capacitance of the unused channel would negatively affect our bandwidth, so I selected the TMUX136 as the mux as it will provide us with enough headroom for this solution to work. Bare in mind, that this would be the setup for every signal. So RX+, RX-, TX+ and TX- for a total of 16 muxes needed.

    I don't believe you'll need to do this for the clock, however, since you can short the clocks all together and allow them to share the clock. I could be mistaken though. If we need to mux the clocks too, we'd need double the count for the differential TX and RX clocks.

    Let me know if this is clear or if you have any other thoughts or questions on this.


    Thanks!
    Rami