After the FB above Vfbth, how did the link626P to make the MOSFET on

I'll refer to the DI-201 document for this explanation.


Our LinkSwitch family of controllers use a simple On/Off control technique coupled with primary side feedback. The idea behind the primary side feedback is fairly straight forward.


When the MOSFET in U1 turns OFF, D7 becomes forward biased and the output voltage on C7/C8 appears across the secondary winding (pins 9 and 6). The voltage across the secondary winding will also appear on the FB/Bias winding (scaled by the turns ratio squared). The controller waits 2.5 us after the MOSFET turns off and samples the voltage present on the FB pin and uses this to determine if the output is in regulation or not.


For the On/Off control, there is an internal state machine that changes states depending on information received from the FB pin. The highest state corresponds to max duty cycle and max current limit. As the secondary output load decreases, the state machine changes states, reducing the MOSFET duty cycle and current limit. At very light loads, the state machine begins cycle-skipping where the MOSFET is kept OFF during most of the switching frequency periods. In its absolute lowest state, the MOSFET will still turn on occasionally, but not off completely, so that the output voltage can be sampled via the FB pin. If the MOSFET was disabled completely, there would be no way for the controller to sense the output voltage.


I hope that answers your question.


-The Traveler

Thanks for your reply.
For the 'off control' it's clear. But I am still not clear about the 'on control'.
For example, at very light load, most cycle will not turn on the MOSFET, what's the factor for the state machine to decide to turn on MOSFET occasionally?

Regards
Thomas

I don't know all of the specific details...but think of it like this:


Suppose there are a total of 64 states in the state machine and each state corresponds to 64 successive clock cycles.


In state 64, the device switches the MOSFET ON and uses the max mosfet current limit and max duty cycle for every single clock event to deliver the most power to the output.


As less output power is required, the state machine starts racheting down, first with smaller and smaller duty cycles. Then as output power is decreased further, the state machine starts lowering the mosfet current limit.


Once the state machine is at the lowest practical mosfet current limit, it will start cycle skipping. First it might skip only one of the 64 switching cycles. As less output power is required, even more switching cycles get skipped until eventually it's only switching once out of every 64 clock cycles.


You can't have the state machine disable switching completely because the FB pin can only see the output voltage of the power supply after a MOSFET switching event. So even in it's lowest state, the MOSFET still switches ON every once in while to maintain output regulation at light loads.


As far as "what decides" when the MOSFET switches...it's hardwired into the state machine. It's not designed to stop switching completely unless there is some kind of fault condition on the output.


Bear in mind, this is a very simplified version of whats happening inside the controller. Much more and I'd have to get you to sign an NDA :-P

-The Traveler

I pick up this topic, because I want to decrease the lightest load power consumption for my design based on LNK626P. It's not a must requirment, I just want to try.
Any way, I had understand much more about this power supplys working.
I am very appreciated for your reply. Have a nice day!

Regards
Thomas

Thomas -


This is a very common design goal. I would suggest getting the majority of your design completed. Once you're happy with the performance of the supply, you'll want to experiment with different values of pre-load resistance on your output. The final choice will be a compromise between no-load input power and no-load output voltage. As you use higher values of pre-load resister, your no-load input power will go down but your no-load output voltage will tend to increase as a side effect.


When you've reached a good compromise on your pre-load resister, experiment with different values of your bypass pin resister (the resister that feeds the bypass pin from your bias winding). Larger values of resistance will decrease no-load input power but you don't want to make it too large or you'll turn on the internal current-shunt attached to the drain pin. When the drain pin current shunt turns on, your no-load input power will increase quite a bit.

If no-load input power is one of your primary design goals, consider our LinkSwitch-II family. They're designed to get down to < 30 mW of no-load input power. You can also check out our LinkZero products. These can typically be designed for less than 5 mW of no-load input power.

-The Traveler