This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Obtaining stable PFC operation from 0% to 100% load

Happy Holidays to all,

I am working on several continuous-duty designs that vary from about 1.5 kW to 5 kW. Universal AC input voltage is NOT wanted in order to improve overall power efficiency. The output of the PFC (about 200 VDC for 120 VAC input and about 400 VDC for 208/240 AC input) drives a DC-DC converter. The load on the output of the DC-DC converter can ramp from 0% up to 100% and back, or anywhere in between.  

Most of the PFC controllers on the market are designed for relatively large input voltage variation and relatively small output power variation at lower voltage. But for some applications, such as audio and industrial control, the situation is just the opposite: the input voltage variation is relatively small and the output power variation is relatively large at higher voltage. The average power may be10% of full load (or less), but the capability to stably and efficiently supply anywhere from 0% to 100% indefinitely must be there. 

I have designed linear power supplies for decades, but I am just now getting into SMPS. I have studied the textbooks and the literature from many manufacturers and I have come to several conclusions about wide load range, high power PFC controllers. Below I list some of those conclusions. I would appreciate your comments, especially if you think that I am wrong about something.

1. Regardless of manufacturer, if the data sheet doesn't specifically address operation at low load, the controller won't work well at low load. Almost any controller works well at >50% load and many work O.K. down to 20%. That is why almost all of the power factor (PF) and efficiency curves stop at about 20% of full load.

2. According to the regulations, the PF and harmonics are to be measured at the Point of Common Connection to the public power supply (i.e., at the plug for the devices that I am designing). This is an important point because the EMI filter on the input can change the voltage and current waveshapes, especially at low load. At least in principle, the EMI filter should be included in the PFC loop.

3. For the PFC controller to match the input current waveshape to the input voltage waveshape accurately, both waveshapes must be inputs into the PFC control function. Sampling partial currents at various points in the circuit and then using a model to compute the actual waveshapes invariably leads to distortion. The distortion is always worse at low load, for a variety of reasons (lower S/N ratio, change from CCM to DCM, oscillations, model errors, etc.).

4. For reliable PFC operation at <20% of full load (and certainly at <10% of full load), only Average Current Control Mode works well. For some topologies, measurement of the input voltage and (especially) the input current is not straightforward, so many approximate models have been devised, but they all fail at low load. The use of a Hall-Effect AC/DC current transformer (such as the LEM LAX100-NP) allows accurate measurement of the total current at any point in the circuit. The cost (~$20 quantity one) is certainly reasonable for multi-kilowatt systems.

5. Resonant topologies may be useful, but I need to learn more about them before I can decide. To maintain resonance, there must be a minimum circulating current and whether that can be maintained at 0% load remains to be seen. Even if it can be maintained, it is not clear how the efficiency vs. load will compare with a more conventional topology.

Thanks for your comments,   Larry

  • Re your comments:

     Regardless of manufacturer, if the data sheet doesn't specifically address operation at low load, the controller won't work well at low load. Almost any controller works well at >50% load and many work O.K. down to 20%. That is why almost all of the power factor (PF) and efficiency curves stop at about 20% of full load.

    True Powwer Factor requiements must be met at maximum load. There are no requirements for other than maximum load but it is desirable to have them work to as light a load as possible

     According to the regulations, the PF and harmonics are to be measured at the Point of Common Connection to the public power supply (i.e., at the plug for the devices that I am designing). This is an important point because the EMI filter on the input can change the voltage and current waveshapes, especially at low load. At least in principle, the EMI filter should be included in the PFC loop.

    True.

    For the PFC controller to match the input current waveshape to the input voltage waveshape accurately, both waveshapes must be inputs into the PFC control function. Sampling partial currents at various points in the circuit and then using a model to compute the actual waveshapes invariably leads to distortion. The distortion is always worse at low load, for a variety of reasons (lower S/N ratio, change from CCM to DCM, oscillations, model errors, etc.).

    There are new control ICs that do not requie the input voltage monitoring.

     For reliable PFC operation at <20% of full load (and certainly at <10% of full load), only Average Current Control Mode works well. For some topologies, measurement of the input voltage and (especially) the input current is not straightforward, so many approximate models have been devised, but they all fail at low load. The use of a Hall-Effect AC/DC current transformer (such as the LEM LAX100-NP) allows accurate measurement of the total current at any point in the circuit. The cost (~$20 quantity one) is certainly reasonable for multi-kilowatt systems.

    No Comment.

    Resonant topologies may be useful, but I need to learn more about them before I can decide. To maintain resonance, there must be a minimum circulating current and whether that can be maintained at 0% load remains to be seen. Even if it can be maintained, it is not clear how the efficiency vs. load will compare with a more conventional topology.

    No Comment.