0
\$\begingroup\$

REMOVE THIS WHEN YOU ANSWER THE QUESTIONS IT ASKS

YOUR DESCRIPTION IS UNCLEAR. PEOPLE MAY NOT UNDERSTAND THE QUESTION PROPERLY AND SO WASTE TIME ANSWERING IT. You talk about transcievers and devices. My understanding (which may be wrong) is that you have maybe 500 to 600 'devices'. Please provide a diagram and description that makes it clear the number of each component/module/subsystem and their relative distance from their contrioller/provcessor / ... . \ What is the nature of the interface between the transceiver and the devices?


I am routing a UART connection between multiple devices, the baud rate is 460800. I have a microcontroller which sends to and reads from 20 transceivers using UART. The TX pin from the microcontroller is directly connected to the RX pins of all transceivers. And the TX pin of each transceiver is connected to the RX pin of the microcontroller via a diode. Each transceiver controls different sets of devices at different locations. Each set of slave devices comprises 25 to 30 devices and is about 200 meters away from my device to the last device of the set

Simplified schematic: enter image description here

I have the transceivers layout as a 4x5 matrix, what is the best way to route the UART lines to minimize EMI? Is it a) or b) or some other way?

a) enter image description here

b) enter image description here

\$\endgroup\$
22
  • 2
    \$\begingroup\$ You must first be sure that the idea of using diodes, to combine 20 signals together, at almost half a megabit per second, even has a chance to work. Without schematics to prove the design will work (I guess it doesn't) there is no need to worry about best routing for EMI or EMC. \$\endgroup\$
    – Justme
    Nov 14, 2023 at 20:47
  • 1
    \$\begingroup\$ Assuming these are all high impedance inputs and the source is very low impedance, I would make a controlled impedance line in the shape of A) and then terminate to ground after the last so that there are no reflections. The devil is in the details though (length of line, source/load impedances, high/low logic levels, etc). \$\endgroup\$ Nov 14, 2023 at 20:52
  • 1
    \$\begingroup\$ @HV16 My simplified question is, why do you think that will work at 0.5 megabits? How much wiring capacitance is compared to pull-up resistor value? Or when multiple interfaces receive data simultaneously? \$\endgroup\$
    – Justme
    Nov 14, 2023 at 21:20
  • 1
    \$\begingroup\$ You are using RS485 chips so why did you not use an RS485 to wire all the slaves together as per what folk would consider absolutely normal for this? Please explain why you are using this strange and flawed method. There must be a good reason for what looks such a complicated method but, for the life of me I can't see it. \$\endgroup\$
    – Andy aka
    Nov 14, 2023 at 21:26
  • 2
    \$\begingroup\$ I have used both topologies. While A is preferred, the driver has to be strong enough to drive the properly terminated line. \$\endgroup\$
    – SteveSh
    Nov 14, 2023 at 21:40

2 Answers 2

3
\$\begingroup\$

From comments: -

You are using RS485 chips so why did you not use an RS485 to wire all the slaves together as per what folk would consider absolutely normal for this? Please explain why you are using this strange and flawed method. There must be a good reason for what looks such a complicated method but, for the life of me I can't see it. – Andy aka

I am not sure I understand your point on using an RS485 to wire them all together. Could you please elaborate?

This is the normal way of wiring multiple devices using RS485. It's clearly far less complex and less vulnerable but, maybe you can explain why this won't work: -

enter image description here

\$\endgroup\$
8
  • \$\begingroup\$ Each of my RS485 controls a set of devices at different locations. Imagine my device is at the center of a room, devices A, B, C are in one corner, devices D, E are in the opposite corner. If they are on the same bus, the bus will be awfully long. Using a second transceiver on my device to talk to D, E shortens the bus(es) \$\endgroup\$
    – HV16
    Nov 14, 2023 at 21:54
  • 2
    \$\begingroup\$ @HV16 How big is the room, i.e. how long a bus would be "awfully long"? RS-485 is made for "awfully long" buses. \$\endgroup\$
    – Justme
    Nov 14, 2023 at 22:05
  • \$\begingroup\$ @Justme Each set of slave devices comprises 25 to 30 devices and is about 200 meters away from my device to the last device of the set. My device is at the center. If one bus is used, the length is 200m(center to corner 1)+400m(corner 1 to corner 2)=600 meters for two sets of devices. I am afraid the capacitance will affect the speed of my bus, and a repeater(s) will be needed. It will be a mess with 20 sets. \$\endgroup\$
    – HV16
    Nov 14, 2023 at 22:24
  • 1
    \$\begingroup\$ @HV16 Can you edit the question to include the information that Each set of slave devices comprises 25 to 30 devices and is about 200 meters away from my device to the last device of the set, since is a constraint on the possible answers. \$\endgroup\$ Nov 14, 2023 at 22:45
  • 1
    \$\begingroup\$ @ChesterGillon this information has been added to the question. \$\endgroup\$
    – HV16
    Nov 14, 2023 at 22:48
1
\$\begingroup\$

The 'best way' to do anything for EMC is very nuanced, and the answer will always be some form of "it depends". It's incredibly dependent on routing, and other board-specific constraints, and ultimately can't be answered concisely, and can't really be answered without writing a book's worth of "but if A then B"'s caveats. I think you'd be better off having a go at the routing, and then asking for a review. It's also worth noting that this bus is only one part of your design, and there may be others that are more of an EMC problem.

That being said, in my opinion, B. Personally, I would go for a more star topology than a snake. You don't need to terminate the line, so you can go for the star, and a long snaking trace that repeatedly folds back on itself as it travels the board is far more of a risk IMO. A snake also stretches the line out to its maximum length giving it more chance of radiating. Again, actual constraints when you lay this out might take prority, so come back when you've made a first pass.

\$\endgroup\$
4
  • \$\begingroup\$ For the star topology, can the layout I made for B be improved in any way? \$\endgroup\$
    – HV16
    Nov 14, 2023 at 23:31
  • 1
    \$\begingroup\$ No idea, you have to lay it out first. Maybe it'll make sense in your design to have three columns of 10,10 and 5? 4x5 seems like a smart place to start though. \$\endgroup\$
    – LordTeddy
    Nov 14, 2023 at 23:33
  • \$\begingroup\$ When I've done this, I used a simulation tool and just played with the various routing possibilities and source termination resistor until I got a response I was happy with (monotonic edges, slight amount of overshoot, and edge rates fast enough to satisfy the requirements of the receiver). \$\endgroup\$
    – SteveSh
    Nov 15, 2023 at 1:06
  • 1
    \$\begingroup\$ And this is why you need to model/simulate these options. \$\endgroup\$
    – SteveSh
    Nov 15, 2023 at 1:11

Your Answer

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

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