I am designing a hobbyist's multi-fixed voltage channels using ATX PSU breakout. Of course the fixed voltages are 3.3V, 5V, 12V, and if possible, -12V too which are supported by ATX.

Now I want to implement overcurrent protection for each channels.

For TLDR, questions are:

  1. How fast should the overcurrent protection turn off the switch to protect connected devices? Of course, it will depend on the connected devices but I need a typical value because I am designing a PSU to test various my projects.
  2. What is recommended way to implement this?
  • This protection should be user-adjustable by MCU to protect connected devices not PSU itself. PSU has some fuses to protect itself and also ATX PSU contains OCP already.
  • Maximum current rating should be at least 5A, and preferably, 10A.
  • This is a project of hobbyist, so maybe lacks some electronical details. If you need any further information or wrong description, please let me know.

At first, I thought combination of INA301 which supports overcurrent alert digital output and a PFET for high-side switch. Since my PSU has multiple voltage channels, I adapted high-side switch instead of low-side to avoid ground shift.

And then, I noticed that the alert output of INA301 is active low, so I have to invert the signal to turn off P-MOSFET on alert and added a digital inverter.

And again, I noticed that the current from digital output is too low to saturate the P-MOSFET fast enough.

Here, fast enough time means us order because response time of INA301 is 1us max. For instance, to fully charge the P-MOSFET with 50nC of Qg in 1us, I will need 50mA at least.

By the way, I have no idea how fast the switch should be turned off to protect loads.

Anyway, now I need a MOSFET driver to control MOSFET because digital output current is too low. For instance, in the case of INA301, the current from ALERT is uA order only and I will need

Now I started to think it became too complex because it is already combination of 4 components: INA301, P-MOSFET, digital inverter, MOSFET driver.

I have heard that there's a thing such as high-side switch IC, and searching products. However, before further digging, I think maybe I am not on the right track to implement adjustable OCP.

Also, I need to know a typical time scale for OCP to filter out not adequate products but I have no clue.

So, rephrasing the question,

  1. How fast should the overcurrent protection turn off the switch to protect connected devices?
  2. What is recommended way to implement this?

Any advice would be appreciated. Thank you.

  • \$\begingroup\$ I guess my first question to you is "Are you familiar with current foldback as a technique?" And if so, is it out of the question for some reason? (Yes, I just skimmed your TLDR.) \$\endgroup\$ Commented Aug 14, 2023 at 14:20
  • \$\begingroup\$ @periblepsis I didn't know that term but from a brief search, it seems a techinque to drop voltage to fall current draw which I can see in CC mode of battery charger. Well, it would be nice but I thought it is more difficult or complex thing and I thought just turn off the the channel would be more easy because my power source is fixed voltage from ATX PSU. Or is it more simpler way to protect devices from overcurrent? \$\endgroup\$
    – slyx
    Commented Aug 14, 2023 at 15:51
  • \$\begingroup\$ I've used it and liked it for linear systems (not switchers.) So in modern usage, perhaps more study is required for switchers. I assume they can handle foldback,,designed correctly. But I don't know the complexities. But it works very well in linear systems. I can provide examples there. Just not with switchers. \$\endgroup\$ Commented Aug 14, 2023 at 16:02
  • \$\begingroup\$ @periblepsis That would be very helpful. Alternative suggestion also acceptable. btw, I found an implementation with OP-AMP which I didn't look the detail yet previously here: electronics.stackexchange.com/questions/159922/… I didn't try it because it has MOSFET at low-side which is not preferable for my system since I don't want different ground shift of each voltage channel. \$\endgroup\$
    – slyx
    Commented Aug 14, 2023 at 16:07
  • 1
    \$\begingroup\$ @periblepsis I did and still am doing. Thank you for keyword. \$\endgroup\$
    – slyx
    Commented Aug 14, 2023 at 16:15

1 Answer 1


And then, I noticed that the alert output of INA301 is active low, so I have to invert the signal to turn off P-MOSFET on alert and added a digital inverter.

Normally, the \ALERT output can be used like following:


simulate this circuit – Schematic created using CircuitLab

VP is your respective output voltage (12V, 5V or 3.3V) and VD is the supply voltage you plan to use for low voltage control stuff (e.g. INA301), probably 5V.

The M1 is normally on with the help of Q1 i.e. if \ALERT is floating (no error) then Q1 will be on through R1, making Vgs of M1 negative and turn it on. One thing to note here: The threshold Vgs (or the absolute value) of M1 should be low enough than VP or else it can't be turned on fully. For example, if VP is 3.3V and threshold Vgs of M1 is -4V then M1 can't be normally on because its Vgs will never be higher than -3.3V.

You may want the INA301's \ALERT output to be set for latched protection (RESET = HI) because the IC may try to "regulate" or "limit" the output current by adjusting the M1's on-state resistance by giving or releasing the \ALERT output. That will result in output voltage drop. The disadvantage of latched protection is that either the system needs a power cycle or the RESET input should be set LOW when the overcurrent situation is cleared.

Now, as for your main questions...

How fast should the overcurrent protection turn off the switch to protect connected devices?

You normally don't want it to be slower than the PSU's internal protection (because it'd be meaningless) which kicks in within milliseconds, but I wouldn't worry about the speed of protection because it's not a thing you can keep under control completely. The reaction time of the amplifier-comparator and the rise (or fall) time of \ALERT output, the input capacitance of the high-side switch (M1 above) etc. all will affect the overall time. Also, hardware (analog) protection is generally faster than software (MCU-controlled) protection because the ADC constantly (or periodically) measures the voltage, generates a flag (or an interrupt) which is constantly (or periodically) checked by the software, then it makes a comparison and decides to run a specific code block that turns the series switch on or off. It can still be fast but not faster than a hardware solution.

What is recommended way to implement this?

There's no single way or recommended way of doing this. This can be a hardware- or software-controlled solution. The circuit above might be useful but since the limit is set with a resistor you may not be able to adjust the limit by an MCU. So you may want to have a look at different chips that accepts voltage inputs as reference. In my adjustable power supply design (a few years ago) I used high-side current sensing, and a high-side series switch for output on/off functionality (for both normal operation and OVP situation), not for overcurrent because the PSU was capable of limiting the current and decreasing the voltage accordingly.

  • \$\begingroup\$ Thank you for answer. INA301 supports voltage control instead of resistor and latch mode is actually what I wanted, which is why I choose INA301 to sense overcurrent situation. Of course, its fast respond time advertised as 1us is also considered. \$\endgroup\$
    – slyx
    Commented Aug 16, 2023 at 8:43
  • \$\begingroup\$ What I concerned was, the alert pin's current level. If I have read the datasheet correctly, the typical current of digital output from INA301 is only 0.1uA and maximum 1uA which is too small to charge the capacitance of MOSFET. My initial designed adapted P-MOSFET whose Qg is about 50nC which means the delay of typical 0.5s and 50ms at best. I am not familiar with BJT but isn't this be a problem with BJT? \$\endgroup\$
    – slyx
    Commented Aug 16, 2023 at 8:43

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.