Startup circuit in Flyback SMPS

[Red_Micro]

Hello,

I'm studying this App Note because I need to build a similar SMPS, which requires that wide input voltage. You can directly see the schematic in this link: http://www.ti.com/lit/df/tidrlf7/tidrlf7.pdf

My question is about the startup network. For this, it uses an active network with the transistor Q3 (STP3N150). I'm looking for a cheaper alternative to this approach to avoid the pricey STP3N150. I thought of using another controller like the UCC28711 which has an HV pin. For example, this App Note uses a passive approach based on resistors and diode for the startup, but I don't think this would be as quick as the first one.

Has anyone had experience with this kind of power supplies?

[Weston]

In that app note STP3N150 is used as a source follower.  ~ Any N-channel MOSFET of suitable voltage rating and power dissipation (should be low as its only for startup) should work. That design has a maximum DC bus of 1260V so they use a 1.5kV mosfet which is somewhat expensive. If you need that same input voltage range digikey has some cheaper 1.5kV MOSFETs. If your maximum input voltage is less you can use a lower voltage MOSFET which would be a more common (and cheaper) part.

A startup circuit has to balance power dissipation and time to startup. Using a MOSFET for the startup network allows you to stop power being dissipated in the startup network once the converter has started, allowing you to source more current for startup. It seems that the UCC28711 has that functionality internally but the switch is only rated for 700V so they use that additional network to limit the voltage it sees.

[dmendesf]

Take a look at the BM2SCQ12xT family. It's outrageously expensive but has a 1700V SiC integrated switch. Maybe it can substitute enough parts to became feasible?

[Red_Micro]

In that app note STP3N150 is used as a source follower.  ~ Any N-channel MOSFET of suitable voltage rating and power dissipation (should be low as its only for startup) should work. That design has a maximum DC bus of 1260V so they use a 1.5kV mosfet which is somewhat expensive. If you need that same input voltage range digikey has some cheaper 1.5kV MOSFETs. If your maximum input voltage is less you can use a lower voltage MOSFET which would be a more common (and cheaper) part.

A startup circuit has to balance power dissipation and time to startup. Using a MOSFET for the startup network allows you to stop power being dissipated in the startup network once the converter has started, allowing you to source more current for startup. It seems that the UCC28711 has that functionality internally but the switch is only rated for 700V so they use that additional network to limit the voltage it sees.

The cheapest 1.5 kV MOSFET found in Digi-key is NDFPD1N150CG. Its drawback is the current. I have to check if 100 mA is OK. As you say, the UCC28711 has that functionality internally through the HV pin, and the second app note uses a clamp diode as the internal FET is rated for 700V. This approach seems OK and a cheaper option, however, I have to see if I can achieve the quick startup as the first app note for that wide range input voltage.

Take a look at the BM2SCQ12xT family. It's outrageously expensive but has a 1700V SiC integrated switch. Maybe it can substitute enough parts to became feasible?

I had looked into it. Pretty expensive. I was looking into some reference designs using this part, but couldn't find it. It's a relatively new part in the market.

[xavier60]

SMPS controller ICs like the UC3842 need a small startup bias current into the Vcc pin supplied by high value resistance either from the HVDC rail or one of the mains phases.
The LM5020 shouldn't mind being started the same way. D11 and the UVLO resistors will still be needed.
Resistors used at high voltage tend to have a limited life span. I use 3500V rated resistors.
 

[Weston]

The current rating of the MOSFET should not be an issue, you can size the resistor string in series with it to limit the current and the current is only seen momentarily.

You can also use the approach of just a resistor string and a low voltage zener diode (~20V or whatever these chips need) to directly derive the supply voltage for startup. It will just always consume power, which may or may not be an issue depending on your application.

You determine the maximum resistor value based on the expected leakage of the controller IC / anything else before startup, minimum size would probably be based on how much power you are willing to dissipate. Based on the resistor value and the capacitance for the controller you can determine what your startup time is. Do you have a fast startup requirement or something? Doing some back of the envelope math it looks like your startup time would be on the order of a second with reasonable component values for just a resistor string + zenier diode.

[Red_Micro]

You determine the maximum resistor value based on the expected leakage of the controller IC / anything else before startup, minimum size would probably be based on how much power you are willing to dissipate. Based on the resistor value and the capacitance for the controller you can determine what your startup time is. Do you have a fast startup requirement or something? Doing some back of the envelope math it looks like your startup time would be on the order of a second with reasonable component values for just a resistor string + zenier diode.

Yes, I have a fast startup requirement. One second is way too long for my application. FYI, the first design uses the LM5020 controller, but for my application I will use the UCC28711 that has the startup switch internally. I just need to check if I can achieve the 70 ms startup as the first app note.

[Red_Micro]

A startup circuit has to balance power dissipation and time to startup. Using a MOSFET for the startup network allows you to stop power being dissipated in the startup network once the converter has started, allowing you to source more current for startup. It seems that the UCC28711 has that functionality internally but the switch is only rated for 700V so they use that additional network to limit the voltage it sees.

As the UCC28711 has an internal 700V switch, do you think it could be good to add an external series MOSFET, like a cascode setup? That way I could use the same approach but this time Q3 can have a lower Vdss rating, hence it will be cheaper.