1
\$\begingroup\$

I have a scenario, where device A provides 24V, device B converts 24V to 3.3V and feeds it back to device A. Both devices communicate with TTL (@3.3V), so we have 2 signal lines as well. What circuit is required so that I can safely plug device B in or out while device a is powered and so that I'm not limited to connector types that are hot-plug-safe. Also there is no requirement to limit inrush currents, because all inputs have low capacitances and the 24V rail may drop. I know there are ICs for hot swap circuits, but they are quite overpowered and I prefer a circuit consisting only of passive components.

Does something like the following circuits make any sense (using diodes like for esd protection)? From my understanding, the act of connecting two devices is simmilar to an electrostatic discharge in some way.

Version 1: version 1

Version 2: version 2

Edit: as desired, the DCDC-circuit: DCDC circuit

Block diagram of TPS54240 : enter image description here

\$\endgroup\$
1
  • \$\begingroup\$ For ESD protection you would ideally need transient diodes (bipolar) not standard diodes, you wouldn't know what polarity an ESD event would be. \$\endgroup\$
    – Nedd
    Nov 18 '19 at 3:49
3
+50
\$\begingroup\$

To analyze what happens when you connect two powered circuit, without any special connector, try to imagine that you connect each wire individually in every possible order. Usually the worst case scenario is when you connect GND last.

For example: assume that you connect the 24V first and then the 3.3V and try to imagine which path the currect would flow. What would happen inside the DCDC when it does not have a GND connection? Will it isolate the 24V from the 3.3V, or will it allow some current to flow through? What will happen to this current in device A? If it increases the 3.3V, will device A tolerate it?

The diodes will not do anything. If the corresponding wire has already been connected, it is reverse biased, so it is closed. And if the corresponding wire has not been connected yet, then the diode is not connected either, so it does nothing.

Please copy the schematics of the DCDC here, and I will try to help more.

\$\endgroup\$
6
  • \$\begingroup\$ thanks for your answer! The DcDc converter is only one part of the actual circuit and it's mainly meant to be an example for a critical scenario. I think, providing further schematics would blow this question, but your hint on how to analyse such a problem cleared my mind. Can you imagine an alternative circuit of passive components that would solve my problem? Actually, now I doubt that what I want is possible at all... \$\endgroup\$
    – Sim Son
    Nov 23 '19 at 16:14
  • \$\begingroup\$ I added the schematic for the DCDC converter \$\endgroup\$
    – Sim Son
    Nov 23 '19 at 16:21
  • \$\begingroup\$ It is hard to tell what will happen to this IC without any GND. Try to connect 24V and 3.3V directly and GND through a 10kΩ resistor. This is about the worst thing that could happen during an unfortunate hot plug in. I guess EN will be low enough to enable the DC/DC, GND on the B side will be about 1V, SENSE will be 2.25V, and 3.3V will be 4.3V, enough fry circuit A. \$\endgroup\$ Nov 25 '19 at 5:24
  • \$\begingroup\$ Use GND of side A for the measurements, and replace your load on side A with a resistor. \$\endgroup\$ Nov 25 '19 at 5:31
  • \$\begingroup\$ Yes, I think having 24V and 3V3 connected is the worst case. If I do so, the whole device B is on 24V with respect to GNDa. While this is no problem for device b itself, device a does get 24V on the 3V3 line and the signal lines if they are connected next. I also measured the resistance between 24V and 3V3 on device b (ground not connected) to be ~200kOhm. This high resistance should make it possible to protect device a with z-diodes at every input line, right? \$\endgroup\$
    – Sim Son
    Nov 25 '19 at 14:07
4
\$\begingroup\$

The single most-important feature that your connector must have is that the ground pin makes connection first and breaks last.

Sketch up your circuit and connectors and observe what happens if both of your power supply connections are made but the ground connection is open. There is a very good chance that both your low-voltage rail AND the signal connections have 24V presented to those nodes.

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

The most common necessary part and that is the hot-insertion type connector which connects ground first. At the same time you must provide ESD transient protection. The requirements to avoid SCR-latchup are that the signal lines must never go outside the power rails by more than 0.2V.

I doubt there will be any diode solution for this as transient currents can easily exceed 0.2V unless you shunt every I/O port until DC is OK. That would require a DC OK signal and some handshaking to activate Reed Relays or solid-state switches to go from "protected disconnect" to connected mode.

I would suggest you add 10k resistors in series to limit the current so that the diode protection you have on CMOS inputs can work better with power rails not connected.

\$\endgroup\$
2
  • \$\begingroup\$ I understand why my diode approach doesn't make any sense. But right now I'm positiv that using z-diodes (Vz=3.4V) and weak pulldown resistors to ground on each pin can avoid illegal voltages in any order of contact. Though, the point you mention seems to be relevant. Unfortunately, the signal lines are open-drain, so I can't use series resistors that high... Do you think, external low-drop diodes, used like for esd-protection, can solve this? \$\endgroup\$
    – Sim Son
    Nov 30 '19 at 1:48
  • \$\begingroup\$ ...or maybe additional pullups on the signal lines to kind of bypass the remaining voltage drop of the esd diodes? \$\endgroup\$
    – Sim Son
    Nov 30 '19 at 1:52

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.