Solving the valley frequency hopping problem of quasi-square wave resonant power supply

 Quasi-square wave resonant converters, also known as quasi-resonant (QR) converters, enable flyback switch mode power supply (SMPS) designs with lower signal electromagnetic interference (EMI) and higher full load efficiency. However, since the switching frequency increases when the load decreases, the frequency drift must be limited to avoid additional switching losses. Conventional quasi-resonant controllers use frequency clamping techniques to limit the frequency drift. When the system switching frequency reaches the frequency clamping limit, valley frequency jumping occurs: the controller jumps back and forth between the two possible valley frequency selections, causing transformer instability and noise. A new technique to overcome this problem is to change the valley frequency as the load decreases, thereby gradually reducing the switching frequency. Once the controller selects a valley, it remains locked to this valley frequency until the output power changes significantly: this is the valley locking technology recently introduced by ON Semiconductor.

　　In addition to briefly introducing the quasi-resonant power supply, this article will further explain the valley frequency hopping problem, introduce the valley locking technology to solve this problem, and share practical application cases where experimental results support theoretical research.

　　Introduction to Quasi-Square Wave Signal

　　Quasi-square wave resonant power supply is also commonly called quasi-resonant power supply, which is widely used in laptop adapters or TV power supplies. The main feature of this architecture is zero voltage switching (ZVS) operation, which can reduce switching losses and help weaken electromagnetic interference (EMI) signals. After the transformer is demagnetized, the MOSFET is turned on when the voltage is at the lowest value of the free oscillation (i.e. "valley switching") caused by the resonance of the inductor capacitor (LC) network at the drain node of the MOSFET, thereby achieving ZVS operation. This network actually consists of the primary inductor Lp and the parasitic capacitor Clump at the drain node.

Figure 1: MOSFET conduction at valley bottom

　　The switching frequency of a quasi-resonant power supply is inherently highly variable, depending on load conditions. Unfortunately, the switching frequency increases as the load decreases, resulting in poor light-load efficiency because the switching loss budget is increased. To improve light-load efficiency, a method must be found to clamp the switching frequency even lower.

　　Conventional Quasi-Resonant Converter

　　Traditional quasi-resonant controllers contain an internal timer to prevent the free-running frequency from exceeding an upper limit. The frequency limit is usually fixed at 125 kHz to keep the frequency below the 150 kHz starting point of the CISPR-22 EMI specification. The figure below is a simplified diagram of the internal architecture of a quasi-resonant controller with an 8 µs timer to clamp the switching frequency.

Figure 2: Circuit diagram of a conventional quasi-resonant controller.

　　To turn on the MOSFET, not only must the valley be detected by the zero-crossing detection (ZCD) comparator, but the 8 us timer must also have expired (Figure 2). If the valley occurs within the 8 µs time window, the MOSFET is not allowed to turn on. Therefore, the off time of the power MOSFET can only be changed by different steps within one free oscillation period.

　　At low line voltage and high output load, the transformer demagnetization time is long and can exceed 8 ?s: the controller will switch on the MOSFET in the first valley. However, as the power demand decreases, the demagnetization time decreases and when it decreases to less than 8 ?s, the frequency is clamped. In this case, the transformer core will be indicated to reset before the 8 ?s timer expires (indicating that the secondary current has reached zero and the internal magnetic field has returned to zero). The MOSFET will not restart immediately, the 8 ?s time window will keep the MOSFET in the blocking state and some valleys will be ignored. If the output power level is such that the off time required for cycle-by-cycle energy balance falls between two adjacent valleys, the power supply will operate with switching cycles of unequal size: this is called valley skipping. Longer switching cycles are compensated by shorter switching cycles and vice versa. In Figure 3, 2 or 3 cycles of first valley switching are followed by 1 cycle of second valley switching. The valley skipping phenomenon causes large changes in the switching frequency, which are compensated by large peak current jumps. The current jumps cause audible noise in the transformer.

Figure 3: Valley hopping: The controller frequency jumps back and forth between two adjacent valleys

　　Clamping the switching frequency alone can solve the instability problem under light output load conditions, but it will not improve the energy efficiency at that specific operating point. Therefore, in traditional quasi-resonant converters, frequency clamping involves either cycle skipping circuits or frequency foldback circuits.

　　Frequency Reverse

　　The frequency foldback circuit is usually a voltage-controlled oscillator (VCO) that reduces the switching frequency when the frequency is clamped (Figure 4). By reducing the operating frequency, switching losses are also reduced, and light-load efficiency is improved accordingly. However, during the frequency foldback mode, the MOSFET turn-on event is still synchronized with the valley detection: valley jumping occurs when the controller frequency jumps back and forth between two adjacent valleys, which also causes audible noise in the quasi-resonant power supply.

Figure 4: Quasi-resonant mode with frequency foldback

　　Another constraint imposed by this technique is the choice of minimum frequency at full load and low input voltage. In practice, frequency clamping requires the choice of a low minimum frequency, and this value must be above the audible frequency range (usually around 30 kHz). Due to this low minimum frequency, the primary inductance value is increased to provide the necessary output power, and the transformer size is increased accordingly.

　　Solving the valley frequency hopping problem

　　A new solution to avoid the valley frequency jumping problem is to change from a certain valley position to the next/previous valley position when the output load changes, and lock the controller frequency at the selected position. This is called the "valley lock" technique. Once the controller is selected to operate at a certain valley, it remains locked at this valley until the output power changes significantly. In practice, the output power change can be observed by monitoring the feedback voltage VFB. A counter is required to count the valleys. Valley lock is achieved by allowing the power supply to have two possible operating points under a specific output load. Therefore, when the output load value makes the off time required for cycle-by-cycle energy balance between two adjacent valleys, the peak current is allowed to increase enough to find a stable operating point at the next valley.

Figure 5: For each output load, there is a corresponding operating point between two adjacent valleys

　　Thanks to this technique, the valley frequency instability problem no longer exists and no audible noise can be heard in the transformer.

　　Another feature of this technique is that it provides a natural switching frequency limitation. In fact, each time the controller valley increases, the frequency is reduced in different steps, as shown in Figure 6. The reduction in switching frequency depends on the free oscillation period:

(1.1)

　　Where: -Lp is the primary inductance

　　-Clump includes all parasitic capacitances present at the drain of the power MOSFET (output capacitance COSS, transformer capacitance, etc.)

　　Figure 6 depicts the switching frequency variation of an adapter using a controller with valley lock, such as the NCP1380 from ON Semiconductor. At an input voltage of 115 V rms, the switching frequency drift is limited to between 65 kHz and 95 kHz without the use of any frequency clamping.

Figure 6: Controller switching frequency versus output power with valley lock

　　Another advantage of this technique is that it optimizes the efficiency over the entire load/input voltage range, especially at high input voltages. At high input voltages, there is no longer zero-voltage switching operation: the switching losses increase. Therefore, for example, it is more advantageous to operate in the second valley instead of the first valley, or in the third valley instead of the second valley, allowing the power supply to switch at a lower frequency. This is best depicted in Figure 7, which shows the efficiency change for output powers between 24 W and 34 W when the controller is operated in the third or fourth valley. It can be seen that turning on the MOSFET in the fourth valley provides 0.3% higher efficiency than turning on the MOSFET in the third valley. The switching frequency in the fourth valley is 15 kHz lower than in the third valley.

Figure 7: Energy efficiency differences between the third valley operation and the fourth valley operation in actual application cases

　　Applying valley locking technology in integrated circuits

　　The valley locking technique is implemented in the quasi-resonant controllers NCP1379 and NCP1380 manufactured by ON Semiconductor. In practice, a set of comparators is used to monitor the voltage at the feedback pin and feed the information to a counter. The hysteresis on each comparator locks the operating valley. Therefore, for a given output power, there are two possible operating points: stable operation is guaranteed and there is no valley jump. To further improve light load efficiency, a frequency flyback circuit based on a voltage-controlled oscillator reduces the switching frequency when the output power decreases. The figure below shows the circuit diagram of a 19 V, 60 W quasi-resonant adapter controlled by the NCP1380.

Figure 8: 60 W adapter circuit diagram using NCP1380

　　Thanks to the valley lock technique, this controller changes the valley (from the first valley to the fourth valley) when the load decreases without any instability issues. This helps extend the quasi-resonant operating range down to 20 W at 230 Vrms. The filter screenshot below shows the operating valley when the load decreases at 230 Vrms input voltage. No valley jumping is observed.

Figure 9: First valley at 60 W, 230 V rms Figure 10: Second valley at 45 W, 230 V rms

Figure 12: Fourth valley at 24 W, 230 V rms Figure 11: Third valley at 30 W, 230 V rms

　　Locking technology optimizes efficiency over the full line voltage/load range and improves overall efficiency:

　　At Vin = 115 V rms, the average efficiency measured is 87.9%

　　Average efficiency is 87.7% at Vin = 230 V rms, which is above the 87% limit specified in the Energy Star EPA 2.0 standard

　　When the output is lightly loaded, the efficiency is further improved by the frequency foldback circuit. At 0.7 W output power, the adapter consumes less than 1 W from the AC mains. The following table summarizes the efficiency at light load:

Table I: Light load energy efficiency

　　Frequency foldback technology also reduces the power consumed by the adapter in standby mode (meaning no output load is connected to the adapter) by reducing the switching frequency. At 230 Vrms, the power consumed by the adapter in standby mode from the AC mains (including the discharge resistor of the X2 capacitor) is 85 mW, which is a very good result for a controller without a high-voltage startup circuit.

Table II: No-load energy consumption

　　in conclusion

　　Conventional quasi-resonant controllers are susceptible to the so-called valley hopping problem, which generates switching cycles of different sizes and audible noise in the transformer. Valley hopping occurs when the off time required for cycle-by-cycle energy balancing falls between two adjacent valleys under certain line voltage/load conditions. To solve this problem, this article introduces the valley locking technique. This technique enables the power supply to select two possible stable operating points under given output load conditions, not only does the instability problem disappear, but also, in combination with a voltage-controlled oscillator, the energy efficiency values ​​in this application are significantly improved. Actual test results based on the NCP1380 controller confirm the effectiveness of this approach.