The trend of isolating each and every interface on the planet creates an enormous demand for isolated power supplies. Of the many types of isolated DC/DC converters available on the market, the initial favorites were expensive converter modules housing miniature circuit boards plus isolation transformer in a chunky metal box. A more modern design is integrated fly-back converter ICs with isolated feedback. However, according to customer feedback, they are “as noisy as hell!” These, too, require external transformers to boost the input-voltage to the desired output level.
In addition to the extensive noise level they produce, their isolation voltage is limited by the weakest link in the chain, namely the integrated converter circuit itself. Hence, what’s left is the discrete design of a miniature push-pull converter using a tiny transformer driver in combination with optimized, matched isolation transformers that ensure design simplicity, low-noise, small form-factor, and low cost. This article explains the functional principle of such converters and suggests two application examples.
A push-pull converter requires a transformer driver (Figure 1) driving a center-tapped transformer. This driver includes an oscillator that feeds a gate-drive circuit comprising a frequency divider and a break-before-make (BBM) logic. The BBM logic then provides two complementary output signals, which alternately turn two output transistors on and off.
Figure 1. Transformer driver block diagram and output timing with break-before-make action.
The output frequency of the oscillator is divided down by an asynchronous divider that provides two complementary output signals, S and /S, with a 50 percent duty cycle. The following break-before-make logic inserts a dead time between the high-pulses of the two signals. The resulting output signals, G1
, present the gate-drive signals for the output transistors Q1
. As shown in Figure 1 (right), before either one of the gates can assume logic high, there must be a short time period during which both signals are low and both transistors are high-impedance. This short period, known as break-before-make time, is required to avoid shorting out both ends of the transformer primary winding.
Push-pull converters require transformers with center-taps to transfer power from the primary to the secondary. The switches Q1 and Q2 in Figure 2 represent the output transistors of the transformer driver in Figure 1.
Figure 2. Switching cycles of a push-pull converter
conducts, the supply, or input voltage, VIN
causes a current to flow through the lower half of the transformer primary. This current flow generates a negative potential at the lower primary end with regards to the VIN
potential at the center-tap.
At the same time the voltage potential at the upper primary end is positive with regards to the center-tap. This maintains the previously established current flow through Q2
, which now has turned high-impedance. Both, the voltage across the upper primary and the voltage across the lower primary are equal in magnitude to VIN
. Because these two voltage sources appear in series, they cause a voltage potential of 2·VIN
at the open end of the primary winding with regards to ground.
Per dot convention the voltage polarities at the primary also occur at the secondary. The positive potential at the upper end of the secondary winding, therefore, forward biases the diode CR1
The secondary current, starting from the upper secondary end, flows through CR1
, charges capacitor C, and returns through the load impedance RL
back to the center-tap.When Q2
goes high-impedance and the voltage polarities at the primary and secondary reverse. Now the lower end of the primary presents the open end with a 2·VIN
potential against ground. In this case CR2
is forward biased while CR1
is reverse-biased. Current flows from the lower secondary end through CR2
charging the capacitor, and returning through the load to the center-tap.Core magnetization
Figure 3 shows the ideal magnetizing curve for a push-pull converter with B as the magnetic flux density and H as the magnetic field strength. When Q1
conducts, the magnetic flux is pushed from A to A’. When Q2
conducts, the flux is pulled back from A’ to A. The difference in flux, and thus in flux density, is proportional to the product of the primary voltage, VP
, and the time, tON
, it is applied to the primary: B ˜ VP
Figure 3. Core magnetization and self-regulation through positive temperature coefficient of RDS-ON.
This volt-seconds (V-t) product is important as it determines the core magnetization during each switching cycle. If the V-t products of both phases are not identical, an imbalance in flux density swing results with an offset from the origin of the B-H curve. If balance is not restored, the offset increases with each following cycle and the transformer slowly creeps toward the saturation region.
Fortunately the output transistors of modern transformer drivers utilize MOSFET technology, rather than bipolar junction transistors (BJTs). The on-resistance of a MOSFET possesses a positive temperature coefficient, which has a self-correcting effect on any occurring V-t imbalance. In the case of a slightly longer on time, the prolonged current flow through a FET gradually heats the transistor, which leads to an increase in on-resistance, RDS-on
. This higher resistance causes the drain-source voltage, VDS
, to rise. Because the voltage at the primary is the difference between the constant input voltage and the voltage drop across the MOSFET, VP
, the primary voltage, VP
, is gradually reduced and V-t balance restored.