0
\$\begingroup\$

I am designing a system based on this application note of TI. The overall design goal is to have multiple shift registers, connected via RS-422 and ethernet cables over fairly long distances(up to 5 meters between each node). Each node should convert the differential RS-422 signal to SPI for the shift register and after that back to differential the RS-422 for daisy-chaining nodes. Each node can have a single power supply or - if spatial possible - share a power bus with neighbouring nodes.

My plan is to have MOSI,MISO and CLK as differential signals in a twisted pair, since they will have the highest data rate(not specified yet). Using standard ethernet cables this setup will use 3 of 4 twisted pairs. This leaves me with one twisted pair.

For clean initial conditions I would like to have STROBE/LATCH and OUTPUT_ENABLE for clean operations as control signals. Is there any problem using the remaining twisted pair non-differential for the control signals?

\$\endgroup\$
7
  • \$\begingroup\$ You'll have 5m of the two signals tightly twisted together, so crosstalk will be severe unless the edge rate of the signals is very low. \$\endgroup\$
    – user1850479
    Commented Oct 6, 2021 at 2:48
  • 2
    \$\begingroup\$ What's providing the ground reference among the nodes? This would also be carrying the return current for any single-ended signals. \$\endgroup\$
    – Dave Tweed
    Commented Oct 6, 2021 at 3:23
  • \$\begingroup\$ Dave: Thats a good call! Thanks - haven't thought of that. In this scenario the ground reference would be provided by the different power supplies or the shared power bus, which would create a hell of a ground loop - not good. I guess I need to have one ground wire in the ethernet cable and do some circuit magic at each node for my control signals \$\endgroup\$
    – Dennis Ernst
    Commented Oct 6, 2021 at 3:29
  • 1
    \$\begingroup\$ SPI may not work effectively when slave responds to master due to accumulated delays. \$\endgroup\$
    – Andy aka
    Commented Oct 6, 2021 at 7:08
  • 2
    \$\begingroup\$ For 5m distance if everything is well done, you could push max. 25Mbit/s. Not to mention the complexity and cost of such circuit. It would be economically better to use a separate MCU for each node and some standard communication like CAN, RS485,... \$\endgroup\$
    – Marko Buršič
    Commented Oct 6, 2021 at 7:30

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.