Smart Rectification Benefits Half-Bridge Converters

Mar 1, 2006 12:00 PM
By Jess Brown, Power IC Market Development Manager, Vishay Siliconix, Bracknell, Berks, United Kingd

MOSFET gate-drive signals derived from predictable effects or real-time feedback can enhance the synchronous rectification of half-bridge converters in power-supply applications.

For the PDF version of this article, click here.

The half-bridge converter has been in existence for several years, but these days, it is enjoying renewed interest due to the development of eighth-brick and sixteenth-brick dc-dc converters. In the past, the half-bridge topology was considered for half and quarter bricks, especially at relatively low power levels. As increasing power density in these half and quarter bricks leads to higher power levels (Table 1), it becomes more and more difficult to implement the half-bridge topology efficiently in these converters (Table 2).

However, the emergence of eighth and sixteenth bricks has resulted in a new demand for the half-bridge converter, since the power levels in these packages are more in tune with those at which the half-bridge topology excels. As a result, the half-bridge converter is emerging as the topology of choice for eighth and sixteenth bricks.

Along with a reduction in physical size and an increase in power density, there is also an increasing number of features being added to the dc-to-dc converters that are required from a system level. These features include prebias startup, tracking, controlled startup and controlled shutdown capabilities.

The half-bridge format also has inherent difficulties from an applications point of view that need to be addressed to meet the efficiency and power-density requirements for these converters. For example, there are intrinsic problems when driving synchronous MOSFETs in half-bridge converters due to the difficulty of producing the correct drive signals during dwell time and also predicting when the primary MOSFETs turn on and off.

Half-Bridge DC-DC Converter

The half-bridge dc-to-dc converter configuration consists of two large, equal capacitors connected in series across the dc input, providing a constant potential of one-half V_IN at their junction (Fig. 1).^[1] The MOSFET switches SW1 and SW2 are turned on alternately and are subjected to a voltage stress equal to that of the input voltage, rather than twice the input voltage that is usually the case in push-pull and forward converters.

Also, due to the capacitors providing a mid-voltage point, the transformer sees a positive and negative voltage during switching. This results in twice the desired peak flux value of the core, because the transformer core is operated in the first and third quadrant of the B-H loop and experiences twice the flux excursion of a similar forward converter core.

This is an advantage of the half-bridge topology over that of double-ended forward topologies, in that the half-bridge primary transformer winding has half the turns for the same input voltage and power. This is because the forward converter transformer must sustain the full supply voltage, compared to half that voltage for the half-bridge transformer.

Another benefit of half-bridge converters is the lower winding costs and proximity effect losses. Proximity effect losses occur when eddy currents are induced in one winding layer by currents in adjacent layers. These losses increase significantly with the number of layers. Because the half-bridge converter has fewer turns than the forward topology, it probably will have fewer layers and, thus, lower proximity effect losses.

Another significant advantage of the half-bridge over the double-ended forward converter is that the half-bridge secondary produces a full-wave output rather than a half-wave output. Thus, the square-wave frequency in the half-bridge converter is twice that of the forward converter, and the associated output inductor and capacitor can be smaller.

With half-bridge converters, the synchronous MOSFETs are normally on, and are only turned off when one of the primary MOSFETs is turned on. However, there is no inherent way of providing the drive signals for the designer to control the secondary synchronous MOSFETs during dwell time. There is no information readily available that predicts when the primary MOSFETs are about to turn on and enable the secondary-side MOSFETs to be turned off, thus ensuring neither are on at the same time to prevent shoot-through. And, since the secondary-side MOSFETs are normally on, application circuits must be able to handle startup and shutdown. If not managed correctly during startup and shutdown, the nature of synchronous MOSFETs being inherently on can cause undesirable conditions such as ringing, negative voltage levels and disturbances on the output.

Prebias is defined as a voltage that is present at the output of the converter before the converter is switched on. This prebias can be present for several reasons. The converter may be hot-swapped or there could be a forward path between this output and another output. The converter could be used in a redundant power-supply system (N+1), a parallel system or a battery backup system, to mention a few, where output voltages are already present.

The issue with a prebias is that if any converter has synchronous rectification, it would be possible to discharge the output during startup. That is, the synchronous MOSFETs could be on during startup before the converter has started to supply power, thus providing a discharge path to ground through the output inductor. This results in negative current flow in the output inductor, oscillation problems and negative voltage spikes at the output. It is necessary for the output voltage to charge to its required regulation voltage and not to go through any discharge of the output if a prebias exists prior to startup.

Soft-start is a feature to prevent component stresses during startup of the converter. In addition, it limits the inrush current at startup and allows a monotonic voltage startup. This is generally applied to the primary-side controller of the converter and inherently deals with the secondary-side MOSFETs. However, this article looks at how the secondary-side MOSFETs can also benefit from a soft-start mechanism in half-bridge converters.

Soft-stop is a feature that prevents disturbances on the output and allows a monotonic shutdown of the voltage output. As with soft-start, this is an essential feature in half-bridge converters because leaving the secondary-side MOSFETs to turn off indeterminately can cause negative voltage transients and uncontrollable discharging of the output voltage.

Secondary-Side MOSFET Drivers

Achieving the requirements for prebias, overvoltage protection (OVP), soft-start and soft-stop of the secondary-side MOSFETs generally requires a separate application circuit for each function. However, Vishay Siliconix has introduced a single chip that can be configured to address all the issues described previously, as well as incorporating the error amplifier and precision voltage reference required on the secondary side of the half-bridge topology. The block diagram of this integrated circuit (IC) is shown in Fig. 2.

The signals required to drive the synchronous MOSFETs (Si7108DN) are obtained from the primary-side half-bridge controller (Si9122). By using the primary-side controller, this allows the drive signals to be present even during dwell time. In addition, because the primary controller knows when the primary MOSFETs (Si7810DN) will be turned on, it can provide a signal to turn off the secondary MOSFETs before the primary ones turn on.

The V_IN voltage can be derived from any of the usual methods, such as an extra winding on the power transformer or from the output inductor. However, since a pulse transformer is required to transmit the secondary MOSFET gate signals across the isolation barrier, it is possible to generate the V_IN from these gate signals (Fig. 3).

During startup of the converter and intelligent driver, the MOSFET drivers initially need to be disabled, since the gate-driver voltage could be at an indeterminable level, causing extra losses and even failures in the synchronous MOSFETs. Therefore, the SiP11203 MOSFET drivers are disabled until V_L is at 90% of its final value. However, if the output drivers were left floating until the main drivers were enabled, the high dV/dT rate during the transition of the current from the body diodes of second SW1 and second SW2 (Fig. 1) could result in spurious turn-on of the MOSFET.

Thus, before the main drivers are enabled (V_L < 90%), the Input A and Input B drive paths are reversed and inverted (Fig. 4), and there is a device on each output to pull the relevant gate-driver outputs low at the appropriate time. Once V_L is established, the small inverting driver is switched out of the path, and as a result, the driver circuit will effectively hold the driver outputs low until the V_L voltage level has reached 90% of its final value.

Once V_L has reached 90% of its final value, the external V_REF also is released and able to rise according to the value of the V_REF capacitor (Fig. 5). The rate of the rise of V_L is determined by the external V_L capacitor, which is driven by a 20-mA current from the preregulator. This allows the designer to control the power-up period of the SiP11203.

If the secondary-side MOSFETs are disabled, there will be a situation in which the body drain diodes are conducting current rather than the MOSFET channels. By switching in the MOSFET channel, the voltage across the device will be reduced considerably, resulting in a disturbance on the output of the converter. Hence, the SiP11203 has a soft-start feature that brings the synchronous MOSFET into the power path over several cycles. The soft-start method, as shown in Fig. 6, prevents disturbances of the output voltage. The soft-start period is set by an external resistor placed on R_PD.

The Si9122 allows the break-before-make (BBM) delay to be set with one resistor (BBM1 = BBM2 = BBM3 = BBM4). However, this lacks some flexibility because gate delays can be introduced across the isolation barrier. Therefore, the SiP11203 incorporates a feature that introduces a rising edge delay, once the device has finished the soft-start. For definition of the BBM delays, please see the Si9122 data sheet.

A separate pin (OVP input) is provided in the SiP11203 to detect overvoltage conditions, which can occur according to two separate modes. The first is an overvoltage during the startup of the device. The second is an overvoltage during normal operation. At startup, an overvoltage condition is defined as the OVP input being 20% greater than the final value of V_REF, which is approximately 1.4 V.

If an overvoltage event occurs during startup, the driver outputs are disabled until the external V_REF has reached 1.1 V, which is 90% of its final value of 1.225 V. As V_REF reaches 1.1 V, the output drivers are released to respond to the input pulses. However, if the overvoltage set point (V_REF + 20%) is reached during normal operating conditions, or after V_REF has reached 90% of its final value, the OVP comparator is latched and the output drivers are forced to the on condition. The external V_REF is then discharged to 20% of its normal value. The output drivers will stay on for the SiP11203 (and off for SiP11204) until V_REF is discharged to 245 mV and the voltage at the OVP input pin is below 1.1 V. After the fault condition has cleared, the output drivers are released and the device experiences a soft-start condition.

To prevent the synchronous MOSFETs from staying on when the input pulses from Input A and Input B cease, the SiP11203 has a function that discharges the gates of the synchronous MOSFETs before the bias supply to the IC disappears. The inputs are monitored and, when there is no activity on INA and INB after a certain time, the main drivers are disconnected and the driver outputs are discharged under power-down control (Fig. 7).

The pull-down current will be a fixed ratio of the current set by an external resistor such that the discharge time can be a fixed number of pulses at the normal operating frequency. Without this function, no activity on the inputs of INA and INB — due to the primary-side shutting down — would result in the synchronous MOSFETs staying on. Allowing the MOSFETs to remain on once the primary controller has been shut down can cause the output to be discharged, and negative spikes can occur when the synchronous MOSFETs finally turn off.

Since the SiP11203 has a separate overvoltage pin, it is possible to configure the device to prevent the MOSFET drivers from being enabled under a prebias condition (V_OUT >1.4 V) at startup. During startup, the OVP_IN signal could be connected directly to the output, allowing the SiP11203 to detect a prebias voltage greater than 1.4 V as an overvoltage condition, turning off the MOSFET drivers and preventing any discharge of the output. Once the V_REF on the SiP11203 is established, the OVP_IN would then need to be connected to the output via an external resistive divider, to allow for overvoltage protection as per normal operation.

References

1. Brown, Jess; Davies, Richard; Williams, Dilwyn; and Bernacchi, Jerry. “High-Efficiency Half-Bridge DC-to-DC Converters with Secondary Synchronous Rectification,” PCIM 2001, Nuremberg, Germany.

2. Si9122 data sheet, www.vishay.com/docs/71815/71815.pdf.

Table 1. Typical values of power levels for different brick sizes.

Brick Size	Power Level (W)
Full	< 700
Half	< 350
Quarter	< 200
Eighth	< 100
Sixteenth	< 50

Table 2. Typical values of power levels for various topologies.

Topology	Common Characteristics
	Power Level (W)	Maximum Duty Cycle
Flyback	5 to 50	< 0.5
Forward Resonant Reset	10 to 250	< 0.5
Forward Active Clamp Reset	10 to 250	< 0.5
Push-Pull	15 to 150	< 1.0
Half-Bridge	50 to 200	< 1.0
Full-Bridge	200 to 2000	< 1.0