I'm trying to design a quick startup network for the controller UCC28704. See picture attached. The datasheet suggests using a depletion MOSFET, but in my case it is hard to find them with high voltage for this application, so I'm using a enhancement MOSFET. The circuit attached is from a reference design, and I'm trying to calculate the time it takes to charge C7 and start the controller at the minimum and maximum input voltage levels. I mean, how much it takes when input voltage is 100VAC, and when it is 600VAC? The NTC pin just supplies a current when the controller starts so it turns off the MOSFET. I want to have startup time about ~10 ms, and the circuit with the FET is meant to substitute the commonly used lossy RC network.

Do you know other startup approaches?

 

Another way could be like attached below where I can have a known current

 

At digikey.com there are two pages of N-mosfets, 1000V, 100mA to 500mA. There are some 1kV, 100mA 100 ohm parts.

I was thinking of getting Q4 turn on voltage with out adding a string of resistors. The voltage at R26 is 24 to 140 volts. Remove C17.
One option is to turn Q4 into a constant current source. When there is 0.65 volts across the 20 ohm resistor the NPN transistor pulls off on Q4 limiting the current. Now C22 charges at 100mA. Simple math.

 

At digikey.com there are two pages of N-mosfets, 1000V, 100mA to 500mA. There are some 1kV, 100mA 100 ohm parts.

I was thinking of getting Q4 turn on voltage with out adding a string of resistors. The voltage at R26 is 24 to 140 volts. Remove C17.
One option is to turn Q4 into a constant current source. When there is 0.65 volts across the 20 ohm resistor the NPN transistor pulls off on Q4 limiting the current. Now C22 charges at 100mA. Simple math.

Good idea. I have some thoughts:

1. I think C17 is recommended by the datasheet because the controller enters a "wait state" when it can turn Q4 on unintentionally. Correct me if I'm still wrong.
2. So, I have fixed 100 mA when min input voltage (100VAC) and max voltage (600VAC), right?
3. D14 is a 24V zener, and D15 is 15V zener. Are both necessary?
4. I think this FET can do the job: https://www.digikey.com/product-detail/en/ixys/IXTY02N120P/IXTY02N120P-ND/2354490

 

1. I think C17 could be 0.1uF.
2. Yes 100mA at any voltage above 40 volts and under 1020 volts. (DC on the filer caps)
3. That hurts my head. Not certain. Need to think more.
4. Looks good to me. Worst case 850V x 100mA = 85 watts for a very short time. What about 10mA for a longer time? Or 1mA for even longer. 0.85 watts is much better. How long does it take to charge 22uF from 0V to ?? startup voltage with 1mA? Many of my power supplies take 0.1 second to get started. Will 30mA charge you cap in time?

 

1. I think C17 could be 0.1uF.
2. Yes 100mA at any voltage above 40 volts and under 1020 volts. (DC on the filer caps)
3. That hurts my head. Not certain. Need to think more.
4. Looks good to me. Worst case 850V x 100mA = 85 watts for a very short time. What about 10mA for a longer time? Or 1mA for even longer. 0.85 watts is much better. How long does it take to charge 22uF from 0V to ?? startup voltage with 1mA? Many of my power supplies take 0.1 second to get started. Will 30mA charge you cap in time?

3. I think I can leave just a Zener as shown in my post #2.
4. Actually, 22 uF was in the reference design. In my case, I would need to recalculate the capacitor depending on the controller. But for most of them the UVLO is 21V. So if I choose a startup current of 30 mA, I need to calculate how long it takes to from 0V to 21V. Let's say C= 1uF, then t = (C*V)/i = 700uS. If C = 22 uF, then t = 15.4 ms, which is OK.

 

I see the need for C17 right on the PFC. I just don't see the need to delay the Gate voltage. I would pick some thing small. Data sheet shows 1nF.