FORWARD CONVERTERS HISTORY OF THE FORWARD CONVERTER by Rudy Severns For the past 50 years, the forward converter has dominated the market for commercial power supplies in excess of 50 W. Despite this long history, there is still an ongoing flow of articles claiming “novel” and “new” variations of this circuit. Many of these prove to be reincarnations of old ideas. We'll explore some of the milestones in the history of the forward converter in this article. The Basic Forward Converter Circuit Figure 1 shows a classic forward converter circuit. There are many other variations that share the same basic energy transfer mechanism: power is delivered to the output while the primary switch is conducting. The forward converter is single-ended. The transformer is driven in one direction with a single switch. Since the transformer windings must have zero volt-seconds on average, core reset is required whereby a reset volt- 20 Switching Power Magazine July 2000 age is applied to the transformer. The peak voltage that appears across the switch in the off state, in many variations, is the sum of the line voltage and the reset voltage. For most applications, the pulse width modulation (PWM) of the primary side switch varies with input line. The switch on-time regulates the output. For a constant output voltage, the volt-seconds applied to the primary will be constant but reset time varies, being relatively long at high line and short at low line. The switch voltage is minimized when the switch off-time is long and fully used for reset There is, however, a competing requirement. The rms current in the switch increases with shorter duty cycle. It is desirable to maximize on time, but this increases the reset voltage so a tradeoff is needed between switch current and voltage. All of this was recognized early on, and many new improvements have focused on the best way to reset the core. The circuit of Figure 1 uses a tertiary winding on the transformer for core reset. At low line, a well-designed converter is close to optimum. Almost half the time is used to deliver power, and the remainder is used to reset the core. However, at high line, the power pulse and reset time are short. The total off-time available for reset is not fully used, and the peak switch voltage is higher than necessary. FORWARD CONVERTERS Core reset, switch voltage and switch current are not the only issues in the design of a forward converter. The stress on the switch during switching transitions and the accompanying loss is also important, especially as switching frequencies increase. This has been the subject of many circuit improvements over the years. Historical Developments Single-ended converters have been around for a long time. Heinrich Hertz (1857-1894) demonstrated the existence of electromagnetic waves with a simple flyback circuit. The Ford Model T ignition system also used a flyback. Buck and boost converters appeared in the early 1920's, but a clearly identifiable forward converter didn't appear until 1956. Figure 2 is a circuit introduced by Paynter [1] in March 1956. It seems like another push-pull oscillator of that period, but a closer look shows one transistor (T1) passes power and the other (T2) is clearly labeled “reset”. Power is coupled to the output during the interval when T1 is conducting. The circuit is self-oscillating with the conduction of T1 terminating when the core saturates. T2 then conducts, resetting the core to saturation in the other direction. The duty cycle is controlled by the transformer turns ratio (W1/W2). In 1963, Dudley [2] continued using Paynter's circuit, however, he used the output voltage for reset as shown in Figure 3. Duty cycles in the range of 0.6-0.8 were possible with this circuit when operating from low-voltage power sources. PWM is not used, and the output filter is simply a capacitor. The forward converter received little further attention until the mid 70s. In 1975, there was a sudden burst of activity, mainly from Europe. Definitive discussion of the forward converter in modern form first appears in a March 1975 Philips applications note [3]. It is clear at this point that the circuit was well understood, and a two-transistor version of the forward converter is introduced as shown in Figure 4. There is a reference to an earlier application note [4] dated 1973. In addition to the two-transistor version where reset is done with two diodes, the applications note shows the use of a tertiary winding and a diode or an R-C-diode network for reset. Rudy Severns is a consultant in the design of power electronics, magnetic components and power conversion equipment. His 40 years of experience include both commercial and military designs with TRW, Lockheed, Hughes Aircraft, Magnavox, Intersil, Siliconix and the University Of California. Since 1978 he has lectured extensively in the US, Asia and Europe on power conversion, magnetics design and power semiconductor topics. He holds a B.S. Engineering from the University of California at Los Angeles and is the author of two books and over 70 technical papers. He is a fellow of the IEEE. July 2000 Switching Power Magazine 21 FORWARD CONVERTERS The two-transistor forward converter is a modern-day standard for off-line power from 200 W to 1 kW. Many companies use it as the basis of their product lines. In June of 1975, La Duca and Massey [5] addressed the problem of optimizing the reset by using a voltage clamp. The amplitude was adjusted to provide the minimum necessary voltage during the reset interval as shown in Figure 5. This marks the beginning of efforts to optimize the forward converter by changing the reset voltage with input line. Reset using a current source and an auxiliary winding was introduced by Heinicke [6] in a 1975 patent, shown in Figure 6. The resetting current source is formed from the source voltage and a series combination of an inductor and resistor. Heinicke's circuit also shows the use of a forward converter as a non-isolated boost converter. Before this, the transformer core energy was discharged into a voltage source. Only a portion of the potential change in core flux density could be applied, forcing the use of a larger core with a small air gap. The air gap required more energy stored in the core during reset. The current source allows a larger flux swing and eliminates the need for an air gap. In principle, the core could be reset to negative saturation, and the entire range of differential flux density up to positive saturation could be used. A later variation on this idea was to incorporate a ferrite permanent magnet in the core structure to provide reset bias. In 1976 we see a continued flurry of forward converter developments. Kamata and Katou [7] introduced the secondary inductor current in combination with the regulated output voltage for reset as shown in Figure 7. The reset current goes down as the load current drops but some reset can still be accomplished using the voltage clamp on winding N4. Following the Massey and La Duca circuit, Peterson published the variation [8] shown in Figure 8. An error amplifier referenced to the output controls the reset voltage clamp. Vermolen also introduced a variable clamp reset [9]. In 1976, Lilienstein and Miller [10] combined tertiary winding current as shown in Figure 9. Their circuit first showed an interleaved forward converter. The transistor switches employed mag-amps controlling the base drive to provide PWM (Note: the bias winding for the magamp is omitted for clarity). Interleaved forwards are still very popular and useful today, undergoing another cycle of rediscovery. In an important development in 1981, Carsten [11] used an active clamp circuit with several variations, shown in Figure 10. This simple circuit provides optimum reset voltage as the duty cycle changes, recovery of most of the core energy, symmetrical core excitation, and low loss switching transitions. That's just about everything you could ask for in a forward converter reset circuit. Since that time this circuit has been widely adopted. The idea of a voltage clamp suggested by La Duca and Massey in 1975, has been revived many times over the years. Another clamp circuit was suggested by Kuwabara and Miyachika in 1987 [12]. Carsten's further contribution in 1992 used a snubber circuit to simultaneously provide low stress switching and core 22 Switching Power Magazine July 2000 reset. This idea appears in other literature. Figure 11 shows a variation, where a well-known energy recovery snubber circuit has been designed to provide core reset. A number of papers have appeared with similar ideas [14], frequently using the transformer leakage inductance in a resonant energy recovery scheme. The forward converter remains the subject of vigorous innovation and debate today. Although the basic energy transfer mechanism is the same as Paynter's original circuit, the details of implementation are still being explored. The only real change has been the increase in switching frequencies from about 2 kHz to 500 kHz or more. This is due mostly to device improvements, packaging advances, and reduced size requirements. The most recent versions of the forward converter have focused on the use of MOSFET synchronous rectifiers with various gate drive schemes and reset clamp arrangements. References [1] Paynter, D.A., AN UNSYMMETRICAL SQUARE-WAVE POWER OSCILLATOR, IRE transactions on Circuit Theory, March 1956, pp. 64-65 [2] Dudley, William, UNSYMMETRICAL LOW VOLTAGE CONVERTER, 17th Power Sources Conference proceedings, 1963, pp. 155-158 [3] van Velthooven, C., PROPERTIES OF DC-TO-DC CONVERTERS FOR SWITCHED-MODE POWER SUPPLIES, Philips Application Information #472, 18 March 1975, pp. 8-10 [4] G. Wolf, MAINS ISOLATING SWITCH-MODE POWER SUPPLY, Philips Electronic Applications Bulleting, Vol. 32, No. 1, February 1973 [5] La Duca and Massey, IMPROVED SINGLE-ENDED REGULATED DC/DC CONVERTER CIRCUIT, IEEE Power Electronics Specialists Conference (PESC) record, June 1975, pp. 177-187 [6] Heinicke, Harald, APPARATUS FOR CONVERTING D.C. VOLTAGE, U.S. patent number 3,921,054, 18 November 1975 (1973 German filing) [7] Hamata and Katou, DC-TO-DC CONVERTER, U.S. patent number 3,935,526, 27 January 1976 (1972 Japanese filing) [8] Peterson, W.A., A FREQUENCY-STABILIZED FREE-RUNNING DC-TO-DC CONVERTER CIRCUIT EMPLOYING PULSE-WIDTH CONTROL REGULATION, IEEE PESC proceedings, June 1976, pp. 200-205 [9] Vermolen, J.V., NON-SATURATING ASYMMETRIC DC/DC CONVERTER, U.S. patent number 3,963,973, 15 June 1976 (1973 Dutch filing) [10] Lilienstein and Miller, THE BIASED TRANSFORMER DCTO-DC CONVERTER, IEEE PESC proceedings, June 1976, pp. 190-199 [11] Carsten, B., HIGH POWER SMPS REQUIRE INTRINSIC RELIABILITY, Power Conversion International (PCI) proceedings, September 1981, pp. 118-133 [12] Kuwabara and Miyachika, A VERY WIDE INPUT RANGE DC-DC CONVERTER, IEEE INTELEC proceedings, 1987, pp. 228-233 [13] Wittenbreder, Martin and Baggerly, A DUTY CYCLE EXTENSION TECHNIQUE FOR SINGLE ENDED FORWARD CONVERTERS, IEEE Applied Power Electronics Conference (APEC) proceedings, 1992, pp. 51-57 More references are available on-line at www.switchingpowermagazine.com/forwardconverters. These references, compiled by Rudy Severns, do not include all research. Additions to the list would be greatly appreciated, particularly those in Europe in the 1960s and 1970s.