Share primary-side regulated flyback technology for LED lighting applications

This article describes a high power factor flyback converter for LED lighting that achieves all of these features and can be dimmed using a standard TRIAC-based dimmer .

I. Flyback Basics

For isolated power supplies up to about 100W , the flyback topology has been widely accepted because of its relative simplicity, low component count , cost-effectiveness and reasonable performance. Its basic operating principle is simple and easy to explain with the help of Fairchild Semiconductor Application Note AN-4137. When MOSFET Q1 turns on, the current in the primary of transformer T1 ramps up linearly, setting up a magnetic field that stores energy, and the polarity of the transformer winding is such that the secondary-side rectifier DRect is off during this period. Once the MOSFET turns off, the polarity of all voltages across the transformer reverses according to Lenz's law. DRect now begins to conduct and the energy stored in T1 is transferred to capacitor CFilt. The duty cycle of the PWM controller and the transformer turns ratio together determine the output voltage, which is stable with the help of the isolated feedback network . The networks DCL, CCL and RCL clamp the voltage surge because of the imperfect coupling between the primary and secondary, i.e. the presence of the leakage inductance . This is important to reduce the voltage stress on Q1, but is also a source of power loss because energy is dissipated in RCL.

Figure 1. Simplified schematic of a SMPS based on flyback operation.

Typically, switching power supplies can operate in two different modes: discontinuous conduction mode (DCM), where the MOSFET turns on only after the current in the diode DRec drops to zero; and continuous conduction mode (CCM), where it turns on while there is still current flowing through DRect. Sometimes a third mode is mentioned: transition or boundary conduction mode (BCM), where the MOSFET always turns on immediately after the diode current reaches zero. As the name suggests, this mode is between DCM and CCM.

II. Quasi-resonant operation

The flyback converter is so far a so-called hard- switching converter. This means that the MOSFET is turned off when the drain current is high and turned on when the drain voltage is high. Since in each switching cycle the falling/rising current and the rising/falling voltage overlap, their result is not negligible and each transition has considerable power losses called switching losses. In a DCM flyback, no current flows when the MOSFET is turned on, but the intrinsic capacitance CDS of the MOSFET must be discharged and the energy stored in this capacitance must be dissipated. If you remember that the stored energy is 0.5xCDSxVDS2, it is obvious that it is advantageous to switch on the MOSFET with the lowest possible VDS.

In a hard-switched flyback operating in DCM, it can be noticed that the drain voltage oscillates after the energy is fully transferred to the secondary and the transformer is demagnetized. This oscillation is caused by the transformer primary-side inductance Lp and the drain-source capacitance CDS of the MOSFET. The quasi-resonant topology monitors the drain waveform and detects the minimum of this oscillation to turn on the MOSFET. Using this method, the switching losses are reduced and can be further reduced by increasing the VDS at turn-off, at the expense of the increased cost of the MOSFET due to the increased VDS.

Without going into more detail, it can be said that conventional QR switching has the disadvantage that the switching frequency increases when the load decreases, because the switching is synchronized with the transformer demagnetization. The latter happens faster the lower the (load) current level. Even if the switching losses themselves are reduced with QR switching, the high-frequency operation at low load levels will disrupt the loss balance under these conditions.

Figure 2: Quasi-resonant switching

Therefore, advanced QR controllers use improved mechanisms to detect the minimum drain voltage. For example, the FAN6300A has a certain minimum time of 8μs, during which the synchronization circuit is disabled. Only after this time has passed, the next drain voltage minimum is detected. The result is that the nth minimum of the drain voltage oscillation is detected instead of the first minimum. If this minimum stop time is increased under reduced feedback level and thus reduced load conditions, it is even possible to reduce the switching frequency and the load current, resulting in excellent low-load current efficiency.

III. PRIMARY SIDE REGULATION (PSR)

LEDs should be driven with a constant current , since they are relatively constant and temperature and production parameters determine the on-state voltage . This is usually achieved by some circuit, as shown in the simplified schematic of Figure 1, where the output current is sampled and amplified to drive an optically isolated feedback network. The standard way to implement this circuit is to use an operational amplifier that requires an additional stable operating voltage, which significantly complicates the secondary side design. Despite this, looking at the performance of an optocoupler in a typical ballast application, the life of this device will be shortened at elevated temperatures.

One mechanism is to ignore the complex secondary-side circuitry and extend the lifespan, since no opto- coupling is required in the so-called primary-side regulation. The latter exploits the fact that the ratio of the two different flyback output voltages is mainly determined by the winding ratio of their respective transformer coils. If one of the outputs, that is to say the one that generates Vcc for the PWM controller, is stable, then the remaining output will also be relatively stable.

If output current regulation is involved, the situation becomes a little more complicated. Basic calculations show that the MOSFET on-time should vary with the square root of the load voltage, which is not easy to achieve. If the load voltage variation is limited to a smaller range, in fact, for LEDs, the linear approximation of the square root is acceptable.

IV. Dimming

Until now, the industry has used a number of different electronic dimmers to test ballasts. So-called 'Tronic ' or 'Phase Cut' dimmers, used with electronic transformers, work well with halogen lamps because the switching element in these dimmers is not a TRIAC and does not rely on a certain holding current.

Many standard TRIAC-based phase-cut dimmers also work well, but here the situation is more complicated. Because the TRIAC requires a certain holding current, which is related to the minimum controllable power, those dimmers have a lower minimum power, say 20W, and the low-power dimmers are better suited than dimmers with higher values. This is really no different than incandescent lamps using TRIAC-based dimmers. But because a 20W LED might replace a 75W incandescent, a dimmer with a built-in 50W minimum load rating might fail.

A second problem that can occur with some dimmers is the ringing of the input filter together with C102, which can cause false disconnection and re-triggering of the TRIAC. In this case, a damping network consisting of a resistor of about 470/2W in series with a 100nF/400V film capacitor can help. This network should only be included if necessary, as it dissipates some power and reduces efficiency.