Design Ideas

February 3, 1997

Saturable bead improves reverse recovery

Giampaolo Carli, Computer Products, Boston, MA

Power-supply designers struggle to reduce the ill effects of the reverse-recovery charge in switching diodes, especially in high-voltage circuits. The recovery mechanism causes unwanted ringing and high-voltage spikes in the circuit. These spikes, in turn, generate high EMI emission, which can pollute the line and produce a noisy environment for the control circuitry. Active snubbers can alleviate these problems but are complex and inappropriate for low-end designs; however, passive RC snubbers are lossy and bulky.

The circuit in Figure 1 uses a saturable inductor bead to control the switching diode's reverse-recovery time and, therefore, to reduce EMI. In addition to lowering EMI, this circuit allows the use of slower diodes, which may be cheaper and more available, if a slight reduction in efficiency is acceptable. For example, using a MUR860 rather than the MURH860 results in an efficiency reduction of less than 0.5%. The cost difference, however, including the addition of the 1/2W resistor, signal-switching FET, and bead, is relatively large, with savings of as much as 50%.

This circuit is also less sensitive to specific diode device characteristics. These types of switching diodes are very specialized, and their turn-off profile changes quite markedly from manufacturer to manufacturer, among production batches from the same manufacturer, and with temperature. These variations may cause concerns with the predictability of performance on production units when exposed to different environments. This circuit eliminates these worries. For example, an HFA08TB60 diode in place of a MURH860 results in a marked change in the turn-off waveform without the bead. With this snubber in, the waveforms become similar. The only difference in performance with temperature is a prolonged reverse recovery time and consequent loss of efficiency.

Although you can apply the proposed circuit to many different topologies, the boost-converter topology in Figure 1 demonstrates how the bead works. The saturable inductor bead connects to the source of switching FET Q₁, which in turn couples to the main inductor (L₁) current. Before Q₁ turns on, the main inductor current flows through the saturable bead and through the output diode into the load. Because the bead's saturation current is small, the bead is deeply into saturation at this point.

As Q₁ turns on, L₁'s current redirects from the load into Q₁'s opening channel as well as the other winding of the bead. When the increasing current in Q₁ becomes equal to the current in L₁, and when diode D is ready to recover, the total NI product in the bead approaches zero, and the bead itself quickly drifts out of its saturated state. The bead then becomes a high impedance and reacts to the increasing Q₁ current by asserting a positive voltage, V_B. This voltage detracts from the gate-drive voltage and considerably slows turn-on. The di/dt rate decreases from several hundred amps per microsecond (as determined by circuit stray and parasitic inductances) to a controlled value equal to V_B/L_M, where L_M is the unsaturated inductance of the bead.

You can estimate the value of V_B as V_T+(I_L1/K), where I_L1 is the current through L₁, and V_T and K are Q₁'s threshold voltage and transconductance, respectively. For high-permeability ferrite or amorphous cores, the resulting di/dt rates range from 15 to 2A/m sec.

Source-resistor R provides an offset to the current balance between the three circuit branches. Without this resistor, NI cancellation in the bead would occur before the current in the diode becomes zero because of the nonzero value of the bead's saturation current. Under this condition, the snubbing circuit would become active when the diode is not recovering. Furthermore, to recover reasonably fast, silicon diodes need some sizable negative current to sweep out stored charges. By adding R in the circuit, the FET branch needs extra current equal to V_B/R before the snubbing circuit becomes active, thereby ensuring some controlled reverse current in the diode.

The higher this controlled reverse current, the faster the recovery time, and, in most cases, the lower the power dissipation in the FET. The value of R is on the order of 5ohms, and you can modify the value to achieve better performance in specific areas.

Because this snubber circuit tends to slow FET turn-off and -on, an undesired decrease in conversion efficiency results. However, a small switching FET in the gate circuit removes this problem. As the main FET, Q₁, turns off, V_B immediately becomes negative because of the decreasing current, rapidly turning Q₂ on and Q₁ off with minimal switching loss. During this sequence, V_B is equal to –(I_L1´R), and this voltage persists until it drives the bead back into saturation, ready for the next turn-on transition. Note that V_B may now exceed the maximum allowed gate voltage for Q₁ and Q₂, so an additional diode and resistor may be necessary across R and the bead.

Figure 1's configuration and component values implement a typical 350W boost converter regulating at 390V dc with an input of 100V ac. Switching frequency is 75 kHz. This design reduces the recovery current at 25° from 12 to 3.5A. Consequently, the circuit also reduces the dv/dt at the boost diode's anode with the direct effect of lower EMI. A similar circuit was run that compared the EMI performance of a SGS STTA12 diode, which already features soft turn-off for EMI reduction, with and without the proposed snubber. The emitted noise is approximately 6 dB lower everywhere and as much as 15 dB lower at specific frequencies. This improvement may represent some savings in the implementation of the EMI input filter. The circuit also does not adversely affect conversion efficiency, which remains at 95%.

The snubber acts on the driving-circuit side of the switching device rather than the power-handling side, which is the case for most conventional snubbers. This configuration reduces the size and cost of the snubber magnetic component by orders of magnitude, because of the greatly reduced operating volt-per-second requirements. Also, unlike most conventional magnetic snubbers, no external components exist at the drain-anode connection, which allows the otherwise-impossible implementation of tight layouts. Low parasitic and stray-inductance layouts are desirable because they reduce voltage stress on the switching devices.

Implementing this circuit, which takes up virtually no pc-board real estate, is easy. The 1/2W resistor is the largest of the three added components. The layout is important, however, because the bead sits in the switching FET driving circuit, which has high gain and bandwidth. High-frequency oscillations may break out during the FET's turn-off transitions. You can usually avoid oscillations by using a better layout or by adding a small noise-suppression bead directly in series with the gate of Q₁. (DI #1981)

FIGURE 1

A saturable inductor bead controls the switching diode's reverse-recovery time, which reduces EMI and provides other benefits.