0
\$\begingroup\$

In differential signal circuits (specifically, RS-485), there is usually an isolation between the signal processing / generation stage (eg, microprocessor) and the line driver chip (something like a MAX485 or SN75176).

This is usually accomplished with an isolated DC-DC power supply, or a galvanically isolated AC-DC transformer / power supply.

What I don't understand, is why isolating the power between the microprocessor and driver chip is neccesary, if instead the entire circuit is powered by a galvanically isolated power supply.

Does this not achieve the same level of isolation, if there are no other external connections sharing this power supply?

(My understanding is that the isolation is to prevent ground-loops / large voltage potential between the transmitter / receiver - and the isolated power allows the receiver to be grounded (here I'm assuming the trans/rec are connected with d+, d- and ground) at the same level as the transmitter. In my case, the galvanically isolated power supply would also be connected to the transmitter's ground along with the data wires)

\$\endgroup\$
  • 1
    \$\begingroup\$ What examples do you have of isolated supplies? Granted that it is best is all connection to a RS-485 node have isolated supplies, as RS-485 often requires a separate signal ground wire the goes to a ground terminal at each port in a hopscotch fashion. This is to maintain some control of DC drift, else data voltages would drift too much and errors would occur. \$\endgroup\$ – Sparky256 Oct 20 '18 at 23:11
2
\$\begingroup\$

I would not say that "usually" there is galvanic isolation of the RS-485 bus. Only in severe applications that can bear the cost or where the lines are very long and common mode voltages are difficult to control.

As far as whether isolation of just the power supply is equivalent- you've mentioned the main reason why it isn't - usually there is something else connected to the circuit on top of the power supply and RS-485 connections. Even if there isn't, noise coupled in on the RS-485 lines would drive the circuit relative to the (presumably grounded) housing/chassis so you could get issues from capacitive coupling of that noise to anywhere in your circuit. Of course you could add a shield, connected to the RS-485 ground so the entire circuit was encased in a Faraday cage at the common-mode voltage of the communication bus but there is still coupling through the power supply input-output capacitance, which can often be in the nF or tens of nF range for switching circuits (more like pF for typical RS-485 isolators, so 1,000-10,000x less).

Which is not to say that there are not possible situations where it might be appropriate, but it would be a fraction of a fraction of RS-485 applications.

\$\endgroup\$
1
\$\begingroup\$

The problem occurs when your device has other ports

If you were talking about a device that only connected to the RS-485 or CAN bus, you'd be correct; but then again, do you really need signal isolation to begin with in that case?

However, most devices have other inputs and outputs as well, and these also require a stable reference to work from unless they, too, are isolated. Given that industrial devices that take a DC input often do not control what they share their power supply with, and RS-485 requires a signal ground to control common-mode differences (unlike Ethernet with its fancy-pants transformer isolation), it's simply easier to isolate the RS-485 port than to try to worry about what all combinations of things can be plugged into your device.

That being said, there are devices that do not isolate their RS-485 interface (many BACnet devices I have seen are this way) -- they usually place special constraints on power supply sharing, though (like requiring their own transformer or DC supply, not to be shared with other devices on the network).

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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