Primary-side regulation (PSR) is an observer-based approach to estimate the output voltage of a flyback converter using the reflected voltage on a primary-referenced winding. Previous Analog Design Journal articles^{[1, 2]} have described the operation and switching behavior of an auxless PSR flyback DC/DC converter for automotive and industrial applications, where appropriately timed sensing of the primary switch voltage at its resonant knee position provides a suitable proxy for the output voltage. This magnetically-sensed regulation technique supports a sub-1% output-voltage accuracy across load, line and temperature ranges.^[1] Moreover, it avoids the secondary-side optocoupler and error amplifier normally used for isolated feedback, thereby eliminating a component crossing the isolation barrier and enabling a cost-effective, high-reliability design with low component count.

Within this context, this article reviews the modes of operation and salient characteristics of a PSR flyback DC/DC converter and specifically examines output capacitor selection to meet system specifications for output voltage ripple and small-signal stability.

Figure 1 shows the schematic of a PSR flyback converter^[3] with integrated primary switch and loop-compensation components. The converter supports magnetically-sensed regulation of the output voltage through the transformer’s primary winding. Equation 1 gives the output voltage setpoint:

where N_PS is the primary-to-secondary transformer turns ratio, V_REF is the internal bandgap reference voltage, and V_D is the flyback diode drop (at close to zero current).

Using a variable switching-frequency-control law with peak current-mode control, the converter operates in boundary (BCM) or discontinuous (DCM) conduction mode depending on the load current, as depicted in Figure 2.^[4] Modulating the switching frequency and peak primary current amplitude helps maintain high efficiency across wide operating ranges of load and line.

More specifically, the converter operates in BCM at heavy loads, and the primary switch turns on with a resonant-half-period delay after switch voltage-knee detection (core reset) to achieve a quasi-resonant switching transition. As shown in Figure 2, the switching frequency in BCM increases as the load current decreases. To prevent high-frequency operation at medium load, the mode changes from BCM to DCM such that the switching frequency remains constant at its maximum value (350 kHz in this example). Equation 2 gives the critical output current at the DCM-BCM boundary.

Operation at light load changes to frequency-foldback mode (FFM), which is effectively DCM with a variable switching frequency and constant peak current. Because magnetic regulation relies on sensing the output voltage during switching cycles, it is necessary to maintain a certain minimum switching frequency at no load to continue sensing the output voltage (12 kHz in this example).

As shown in Figure 2, BCM has a lower switching frequency and a higher peak current than DCM. As such, BCM dictates the output capacitor sizing for a given ripple voltage specification. Figure 3 shows the secondary-side waveforms in BCM.