I'm designing a power supply that has an input of 100VAC to 300VAC. On the primary side on the main transformer, I have a 15V rail to power the control chip (a FAN6920) and a gate drive chip (a FAN7382). On this 15V rail, I have roughly 400uF of capacitance. Because of this amount of capacitance, the hiccup time of the supply is about 10 sec. That's the time it take for the 15V rail to decay down to about 5V, at which point the control chip will attempt to start back up. This is acceptable during an actual fault (e.g. an over voltage condition on the output or the output shorted), but this hiccup time is also present when turning the line off and then back on within that 10 seconds. The supply is used to power lights, so as you can imagine, no one wants a potential 10 second delay when they hit the light switch.
To that end, I'm trying to figure out a way to design a circuit that senses whether the AC line has gone away, and if so, to pull down the 15V rail quickly, while at the same time not impacting steady state operation.
simulate this circuit – Schematic created using CircuitLab
My first attempt at this is above. This idea is that C1 will charge to 10V when the line voltage is applied, holding M2 on. Holding M2 on will hold M1 off. At the same time, C2 is charging to 10V. When the line voltage is disconnected, C1 discharges much quicker than C2, allowing M2 to open up, and allowing C2 to change the gate of M1, turning it on and rapidly discharging the 15V rail through a 10 ohm resistor. However, in practice, C1 doesn't discharge nearly quick enough. It maintains enough voltage to keep M2 on for several seconds, so it doesn't really buy me anything (and also, the charge of C2 decays enough that it can't turn M1 on at that point, anyway). I can't change the time constant of C1 and R2 by much because the voltage across C1 will start drooping significantly when the line crosses 0V.
I am not tied to this topology by any means; this was just my first pass at a solution. And while I've tried a few other things, nothing has really worked, so any suggestions would be appreciated. Thanks.
Best Answer
The circuit you've shown (assuming there is 120VAC connected to an unseen bridge rectifier with the '-' connected to circuit common) 'looks' like it should work. Obviously those diodes cannot actually be 1N4148s since they're rated at only about 75V.
The time constant of R2/C1 is less than 10msec so there is no way it should take seconds to discharge (even if you allow for the gate capacitance on the MOSFET- presumably you're not actually using that part number, but something much smaller anyway). If there was enough leakage from the main supply capacitor through the main bridge rectifier that could explain it (so the problem would be 'off the page').
Perhaps if you split the (off the page) bleeder resistor for the input filter into two resistors, each to the common, rather than one resistor across the line..