1
\$\begingroup\$

I'm designing the PCB for a robot, where I plan to add some connectors for external RS232 tools (provided by the client).

Each connector will provide power (either 24 or 48V, at 5A max) and an RS232 connection.

Do I need to add a "signal" ground wire for the RS232, or can I share it with the power ground?

RS232 being rather robust to noise, I suspect that it will work just fine, but on the other hand, I have no control at all about what load will be connected (excepted that I will limit the current to 5A), so the current might vary quite brutally.

Cable length will not exceed 2m.

\$\endgroup\$
9
  • \$\begingroup\$ We can't know how you are going to use the two grounds. If you have two ground wires, you can use one for RS232 comms and one for returning motor current, but then how will you do the motor module as you likely need to connect a MCU to RS232 and the motor anyway, so it really depends how you are going to use them. \$\endgroup\$
    – Justme
    Commented Dec 15, 2023 at 9:45
  • 1
    \$\begingroup\$ I generally tend to keep signal grounds and power grounds separate because power ground carries the return current, therefore it can be "fluctuating" which sometimes can lead to communication errors. I'm not sure if this can be the case for a relatively slow and high-amplitude communication such as RS-232. \$\endgroup\$
    – Rohat Kılıç
    Commented Dec 15, 2023 at 10:00
  • 1
    \$\begingroup\$ Well... what's the power doing? Filtered at both ends, heavy switching, the works? Are there any other ground paths (chassis, other connectors) that might make a loop? Is everything more-or-less common-ground or is isolation needed? \$\endgroup\$
    – Tim Williams
    Commented Dec 15, 2023 at 10:22
  • 1
    \$\begingroup\$ @Sandro But if you do provide two ground wires that connect together at your connector, but can't control how clients design their equipment, it does not matter if there is one or two grounds. Clients could connect the two grounds at their connector too. Clients may forget to connect the big ground wire and all 5A flows through small ground wire. If the client sensor powers a MCU from the 5A positive wire, then it uses the 5A ground return wire for power, and if it still needs to communicate over RS232, it will already be connected without the small ground wire. \$\endgroup\$
    – Justme
    Commented Dec 15, 2023 at 12:25
  • 1
    \$\begingroup\$ @Sandro Not really. If you do that, you need a big 5A wire just for RS232 if you do not even use the supply. And it will prevent your submodules from having isolated comms interface from power. And if you plug in connector so that +48V connects first before the single ground wire, that's going to explode most RS-232 transceiver chips rated for 30V. \$\endgroup\$
    – Justme
    Commented Dec 15, 2023 at 12:43

1 Answer 1

Reset to default
1
\$\begingroup\$

I would always provide ground for the RS232 signal to minimize the path inductance and minimize interference (i.e. the minimum RS232 connector would normally have Tx, Rx & Gnd).

However, in your case, a key question will be whether the client-provided PSU and RS232 kit are isolated from each other. If they share a common ground (which your question implies), then potentially a significant fraction of the motor return current could flow in the RS232 "Gnd" wire (whether this is separate or combined). This can cause problems due to shifts in the reference level, especially if the motor current has lots of transients. It may also burn out the wire.

RS232 is pretty tolerant, so you might not have issues, but for industrial environments, my default approach is always to have galvanic isolation on control interfaces to avoid this kind of problem.

It comes from free with Ethernet (which is transformer isolated), and there are off-the-shelf isolator chips for RS232 (including charge pump supplies for Tx) that work very well.

Summary: Do provide a "return"/"gnd" wire for your RS232 interface, but galvanically isolate the RS232 interface from the rest of your circuit to avoid "entertaining" problems in the field.

\$\endgroup\$
3
  • 1
    \$\begingroup\$ OP: Welcome to the wonderful world of ground loops. Where avoidance is expensive, mitigating is tedious, ignoring is ignorance, and when customer equipment integration is involved: nothing is guaranteed. \$\endgroup\$
    – Chris Knudsen
    Commented Dec 15, 2023 at 15:06
  • 1
    \$\begingroup\$ Also, my fellow enginerds (and OP), let RS232 fade away already. Switch to Ethernet (or some industrial version of it), and save yourself all of these nasty ground loop issues. When matching (and beating) rs232 speeds, Ethernet is stupid easy and nearly free. \$\endgroup\$
    – Chris Knudsen
    Commented Dec 15, 2023 at 15:11
  • \$\begingroup\$ @ChrisKnudsen Indeed. As my answer probably gives away, my advice to use galvanic isolation comes from past bitter experience. My "top" problem was with -48V & +24V telco kit connected via a serial cable, where it turned out the incorrect use of a 48V-24V converter to power the 24V kit, meant its serial "ground" was actually at -24V relative to true ground. When the serial connection was made, we discovered that typical RS232 wiring isn't rated for 100A+. Very lucky the person making the connection avoided serious burns. \$\endgroup\$
    – colintd
    Commented Dec 15, 2023 at 16:02

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