0
\$\begingroup\$

At a previous employer, we had an issue where some of our power supplies would brown-out when turned off, and this destroyed some hardware. By 'brown-out' I mean they acted like there was a capacitor connected directly across the output terminals (probably because there was), and the output voltage would drop gradually from the set-point to zero over the course of a few seconds. Under certain conditions, this relatively slow drop in voltage could result in unexpected conduction paths that permanently broke things.

It seems to me that an ideal power supply should open the circuit when output is disabled, e.g. via relay-like action. Thus, it would be like pulling a plug from the wall. Is that desired behavior, or am I missing something?

If that is desired behavior, how can I achieve that without resorting to a solid state relay? I can't imagine the contact bounce on a mechanical relay would be good. If I connected a thyristor in series with the load, would that eliminate the bounce when the relay is opened? When enabling output, I'm thinking close the relay, then trigger the thyristor. When disabling output, I'm thinking the thyristor will choke the current as soon as it first drops, and then it will stay closed during subsequent bouncing of the relay.

If I recall correctly, the issue was with peripheral hardware remaining at least partially functional after the main hardware shut down. The main hardware involved an FPGA, and the peripheral continued to apply a voltage to one of the FPGA's I/O pins. And it so happens that when an FPGA is unpowered, applying a voltage to an I/O pin can fry the chip by creating a short between the power rail for the pin's I/O bank and ground. So the FPGA becomes a solid state heater.

\$\endgroup\$
4
  • \$\begingroup\$ "Brown out" suggests a temporary dip in supply voltage so this isn't the right term for the voltage decay you are describing. It sounds as though you have two power supplies but your question doesn't state this. \$\endgroup\$
    – Transistor
    Mar 3, 2020 at 0:29
  • 1
    \$\begingroup\$ Sorry, that just seems like incompetent design. You ought to be able to turn equipment off, and smoothing capacitors on DC supplies are the most reasonable thing on earth. \$\endgroup\$ Mar 3, 2020 at 1:02
  • \$\begingroup\$ Usually the complaint about some power supplies is that they generate a transient voltage above the set voltage when turned off. That could be expected to cause problems with some equipment. A supply that gradually returns to zero with no overshoot when turned off is normally considered desirable behavior. \$\endgroup\$
    – Barry
    Mar 3, 2020 at 1:47
  • \$\begingroup\$ Simple answer is you are responsible for switching off CPU or DC OK detect not the power supply having a relay. \$\endgroup\$ Apr 16, 2020 at 21:03

1 Answer 1

1
\$\begingroup\$

Power down (and up) sequencing can be a problem in some circuits. For example an MCU or FPGA might have 5 V tolerant pins, but only when the chip is powered, or a circuit might get powered through a device's input protection diode which can only handle a few milliamps. Some chips have multiple supply rails that have to go up and down in a particular sequence to avoid damage (eg. the Intel 8080 CPU).

A power supply can be designed in such a way that some voltages are guaranteed to come up and go down before/after others, or all at the 'same' time. This might be worthwhile in a dedicated supply, but could make it more complicated and less flexible (= more expensive) than a 'generic' supply.

Most supplies are designed to hold their outputs up as long as practicable in a 'brownout', and the time they hold up for depends on the current drawn. It is then up to the system designer to make sure the devices being powered can handle it, or add circuitry to produce the required sequencing.

You could 'open' the supply using a power MOSFET controlled by a voltage comparator that turns it off when a brownout is detected. However this still might not be sufficient if there is a lot of 'bulk' capacitance further on in the circuit. The power on sequence may also need to be considered, which could make the control circuit quite complex.

Other solutions include powering everything from the same voltage, adding more bulk capacitance to parts of the circuit that have to stay powered up longer, disabling regulators when brownout is detected, protecting inputs with series resistors to limit current (possibly combined with external protection diodes) optocouplers, level shifters etc. In some circuits a relay may be the best option!

\$\endgroup\$
1
  • \$\begingroup\$ Well, it definitely sounds like the system designers didn’t do their job with respect to power sequencing. Which is not surprising, and part of why I no longer work there. \$\endgroup\$ Mar 3, 2020 at 7:13

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.