@O.P,
*NOT* a good idea - as described you may run into power sequencing problems and possibly damage your board.
What you need to do is make the presence of power from the ATX PSU switch on the substitute 5V supply output
So, in order to rephrase this so I know if I understand, I want the substitute 5v supply to have a constant input upon applying power to the main DC input, and wire up a circuit to switch its output — rather than switching its
12V input by connecting that to the output from the ATX PicoPSU? This would solve something else I was slightly concerned about — the PicoPSU’s manual suggests not drawing more than 7A max, 10A peak from the 12V output, although it states that forced air cooling is not required for a sustained max load on
that rail (presumably because it’s only regulated by switching the input) so I was figuring I could cheat a little there. The spec for 5V on the stock power supply I’m replacing is 13.2A, which is why I chose 15A (15A 12V to 5V DC-DC converters are readily available, and on the off-chance that 10A isn’t actually enough 15A would be).
Is the power sequencing problem you mention something like the Pico turning on PS_OK, which then would connect to my AND gate, and then as soon as the 5V substitute supply is high enough to trigger the other input of the AND gate (regardless of what the actual voltage is) its output would be HIGH thus signaling the motherboard to boot (whether or not 5V actually is stable)?
I don’t think I mentioned clearly enough how much of a newbie I am to this. I’ve never taken an electronics course, and I’ve learned everything I know between reading and experimenting. (Please don’t try to discourage me from this project, the entire reason I’m wanting to pursue this myself is to learn more about proper power supply logic; knowing what I know about the type of motherboard I’m powering with this it’s very likely I could get away with cheating a
lot on the spec, but I want to remain (at least mostly) ATX compliant for the sake of learning how this works).
this would be easy to do with a 20A logic level MOSFET, gate to +12V via a 100R resistor, drain to substitute +5V, source to the 5V load - and intercept the active low ATX PS_ON# signal and OR it with the output of a PSU supervisor IC* monitoring the substitute 5V supply output before the MOSFET. If the new 5V rail is out of spec its disconnected from the load and the ATX PSU is held in shutdown. The OR gate should be TTL and powered from the ATX 5V standby supply.
I’m hoping to do this without a dedicated PSU supervisor IC. If I can make something myself out of discrete parts that functions as one that’d be better. However, I’m also open to either buying one new (if they’re not that expensive) or stealing one from an old PSU.
That said, again to reiterate what you’re saying to ensure I understand: the idea is that the MOSFET becomes the switched output of the 5V supply (and as mentioned above it’s powered by the direct, “always on” 12V input instead of being switched on via 12V output from the PicoPSU) and then… I’m a little lost at the OR gate — is that simply because PS_ON# is active-low, as opposed to the PWR_OK signal from the ATX PicoPSU that is held high when the PicoPSU is ready?
So really, I wouldn’t be influencing the PWR_OK line at all; I’d just be leaving that to the ATX supply. Rather I’d be switching the output of my substitute 5V rail (which is otherwise “always on” before the MOSFET, like the 5Vsb line) and if the PSU monitoring chip says it’s NG then the OR gate would raise PS_ON# high, turning the PicoPSU off.
Why would I be better off raising PS_ON# high in the event of bad 5V than de-asserting PWR_OK?
You can get ATX PSU cable extenders - better to hack one of those rather than the original PSU wiring harness.
I actually have no choice but to hack an ATX extender because the PicoPSU is just two double-sided PCBs soldered directly to the Molex Mini-Fit Jr. plug, anyway. Not to mention I have pin extraction tools to leave cutting to a minimum. I actually have an ATX extender on my desk right now that I’m intending on using for this. It’s 24-pin, and the PicoPSU is only 20-pin (and in fact this extender is a 24-to-20-pin adapter), but thanks to the genius keying design I just have to be aware of the extra 4 pins (and which four they are!)
* Or a reset controller. However whatever you use needs an active low output when the rail its monitoring is in spec. If all you can find is active high, use a 74LS00 and De Morgan's laws to implement the OR function with the input from the PSU supervisor inverted.
What is a reset controller?
What are De Morgan’s laws?I ended up bumping into this while still writing this: De Morgan’s laws are logical laws that (to oversimplify for this) allow one to implement the OR function via AND gates and inverters, and vice versa I’m going to look this up online as soon as I hit post, but I figure I’ll ask here, too.
Something like one side of a dual 4-input schmitt cmos nand gate would do. Followed by the other nand wired as an invertor.
You then need to work out what potential divider is needed on each input to take the gate past its threshold if the supply is adequate. The schmitt input is needed to give a clean output change. Basically if any i/p is below its threshold the first gate o/p is high, so the second gate o/p is low and the computer doesn't start.
Remember to tie all unused cmos inputs to +Vdd, or to another input.
One chip, and two resistors per supply rail. You could make something more precise and repeatable using a TL084 quad opamp and a reference zener, but the arrangement I've outlined should do for a one-off.
So this would be essentially making my own rudimentary PSU monitor by biasing (I’m not sure if this is the right word for this) the transistor inputs such that an under voltage would not turn on PWR_OK? Is there a way to use another gate/inverter/something to also account for an overvoltage? Say, biasing a gate such that it requires 5.25V to go HIGH, inverting that, and then using it
to pull PWR_OK back low? as input to a gate with the undervoltage gate as the other input?
EDIT: something didn’t feel right, I want to use transistor logic and not have resistors pulling things up or down here if I understand it correctly, or at least that’s the way I intend to implement it. I think. One gate goes high if the voltage is above 5.25V (overvoltage, and low if not), another gate goes low if the voltage is below 4.75V (the undervoltage limit, and high if not), both are inputs to an XOR gate (or whatever, De Morgan’s laws are such that I can invert an input and use an AND gate instead of an XOR, and similar, right?) and that’s the final PWR_GOOD output?I don’t know what Schmitt means. I know I’ve heard the term used in this sort of context before, but I’m not quite sure. Schmitt is referring to a Schmitt trigger? If I’m understanding Wikipedia’s article on it correctly, I’d have the Schmitt set to “trigger” high at 5 volts with a .25v hysteresis (how do I do this?), and it would be the input to my XOR that goes low if the voltage falls to low?
…am I making sense here? I apologize sincerely for how much hand-holding I might need here… This is really my first attempt at understanding semiconductors.