Forward DC/DC converter with semi-synchronous rectification and improved efficiency

[0002] This invention relates to switching-type power converters and in particularly to forward and flyback-type DC/DC converters that use MOSFET devices used as synchronous output rectifiers with gates that are driven by the secondary voltage signal of the power transformer.

[0003] The inefficiency of MOSFET devices as self-driven synchronous rectifiers in prior art buck-driven converters has limited their use. Also, the reset voltage for the transformer core in forward converters designed without an active clamp limits the conduction time of one of the MOSFET devices in the synchronous rectifiers, thereby decreasing the efficiency of the converter.

[0004] In the prior art forward converter shown in FIG. 1A, an isolating power transformer is combined with a self-synchronized synchronous rectifier. In such a rectifier the gates of the MOSFET rectifiers Q2 and Q3 are driven by the secondary winding 108 of the power transformer.

[0005] A DC input voltage 100, Vin, is connected to the primary winding 106 of the power transformer by a MOSFET power switch Q1104. The secondary winding 108 is connected to an output lead 118 through an output filter inductor 114 and the synchronous rectifier MOSFET devices Q2110 and Q3112. Each rectifying MOSFET device includes an inherent body diode 120 and 122, respectively.

[0006] When Q1104 is conducting, the input voltage Vin 100 is applied directly across the primary winding 106. The secondary winding 108 provides a current flow through the inductor 114, a load connected to the output lead 118, and back through the MOSFET rectifier Q2110 to the secondary winding 108. Continuity of the current flow in the inductor 114 is maintained via the MOSFET rectifier Q3112. An output filter capacitor 116 shunts the output of the converter and the load (not shown).

[0007] FIG. 1B shows the voltage and current waveforms of the converter of FIG. 1A. When the pulse width modulator (PWM) 102 turns off MOSFET Q1104, disconnecting the primary winding 106 of the power transformer, the drain voltage of Q1140 rises and, due to parasitic capacitances, assumes a sinusoidal wave-form from t0 to t1. During this time t0-t1, the power transformer core resets itself (reset period). The sinusoidal portion of 140 appears at the drain of Q2 and the gate of Q3 as shown in the trace 142. During this reset period the current through Q2 IQ2 drops to zero and all the current through the inductor 114 is supplied by Q3 shown in trace 148. Specifically the current through Q3 IQ3148 assumes its maximum at t0 and then linearly decreases from t0 to t1.

[0008] Still referring to FIG. 1B, from t1 to t2, both MOSFET rectifiers Q2 and Q3 are off because the gate driving signals of both Q2144 and Q3142 are zero. The period from t1 to t2 is referred to as the “dead time” period in which no power transfer or transformer core resetting takes place. During this dead time, the inductor 114 current, IL, in trace 150 continues to flow through the body diodes of Q3122 and Q2120. Most of the inductor current flows through 122 since the secondary winding 108 is in series with diode 120.

[0009] There is a loss of efficiency during this dead time since the forward voltage drop of the MOSFET body diodes is about 1 volt as compared to 0.1 volts when the MOSFET is on. Moreover, as the input voltage, Vin, increases the dead time period increases and the efficiency of the converter decreases. At maximum input voltage, the inefficiency of the prior art DC converter is greatest.

[0010] At the end of the dead time, t2, the gate voltage of Q2110 is zero, the secondary momentarily forward biases the body diode 120 of Q2110 allowing the secondary current to flow. As soon as the dotted end of secondary 108 exceeds the gate threshold voltage of Q2110, Q2 turns on supplying all the secondary current.

[0011] From t2 to t3, Q1104 turns on again (the power transfer period of the conversion cycle) and Q2110 turns on connecting the secondary 108 of T1 through inductor 114 to the output terminal 118 and any load across Vo.

[0012] During t2-t3, load current is carried through the secondary 108 and Q2110. During this time energy is stored in the output inductor 114 and on the output capacitor 116. The gate driving signal for Q2110 is shown in trace 144, which is the inverted lower part of the drain waveform of Q1140.

[0013] Driving the MOSFET synchronous rectifiers directly from the transformer secondary reduces the efficiency of the converter. Even though the self-driven synchronous rectifiers in the forward converter are simple and low-cost, the inherent low efficiency reduces the maximum possible power density of the converter, raises the converter's temperature, thereby lowering its reliability and making the converter unsuitable for high-power density applications.

[0014] Therefore there is a need for an efficiency improved version of a forward unclamped converter with synchronous rectifiers.

[0015] Forward converter designs using an “active clamp” to clamp the primary voltage generated during the dead time period also suffer from inefficiencies, especially when the input voltage is at its maximum. Typically, DC converters must operate over a 2:1 input voltage range. Forward converters with an active clamp require more complex timing circuitry and additional power MOSFETs. A prior art active clamp circuit is described U.S. Pat. No. 4,441,146 assigned to Vicor Corporation..

[0016] In view of the foregoing background discussion, the present invention provides the combining of primary and secondary signals in an unclamped forward converter to generate the gate drive signals for MOSFET synchronous rectifiers. During the power transfer period, a first MOSFET, which connects the secondary of the transformer to load return (usually ground), is directly driven from the secondary. However, a second MOSFET synchronous rectifier is driven from a combination of the transformer secondary and a second pulse transformer. The second MOSFET rectifier is turned on and remains on during the reset and dead time period of the conversion, its gate drive signal starts at the beginning of the reset period and terminates prior to the start of the power transfer period.

[0017] The advantage is having the second MOSFET is on and supplying current rather than having the current travel through its intrinsic diode. Therefore there is better conversion efficiency due to the lower voltage drop across the on MOSFET as compared to its intrinsic diode.

[0018] For a more complete understanding of the present invention and its advantages, please reference the following descriptions with the accompanying drawings.

[0019] FIG. 1A is a schematic of a forward-type converter of the prior art employing a synchronous rectifier.

[0020] FIG. 1B is the voltage and current waveforms for the forward converter presented in FIG. 1A.

[0021] FIGS. 2A and 2B are a schematic and associated wave-forms, respectively describing an embodiment of the present invention.

[0022] FIG. 3 is a schematic of an embodiment of the present invention implemented with gate voltage limiters for the synchronous rectifiers.

[0023] FIG. 4 is the schematic of an embodiment of the present invention with gate voltage limiters (FIG. 3) specifically with a boost in the secondary synchronous rectifier driving voltage to accommodate low output voltage or wide input voltage ranges.

[0024] FIGS. 5A and 5B are schematics and associated wave-forms, respectively, of an embodiment of the present invention with gates driven from a combination of input and output control signals.

[0025] FIG. 6 is a block diagram of a self-driven active clamp utilizing either N or P channel MOSFET.

[0026] FIGS. 7A and 7B are schematics and associated wave-forms of the forward converter with improved synchronous rectification and an N channel self-driven active clamp.

[0027] FIG. 8 is a schematic of two versions of a P channel self-driven active clamp.

[0028] FIGS. 2A and 2B present a forward converter with semi-self driven synchronous MOSFET rectifiers. The circuit is derived from the circuit in FIG. 1A by adding four discrete low-cost parts D1220, Q4221, T2215, C1208 and a delay circuit 204. The voltage and current waveforms for the forward converter in FIG. 2A are shown graphically by the traces in FIG. 2B. The operation of the circuit in FIG. 2A is similar to that described for the circuit in FIG. 1A but with a different driving signal for the synchronous MOSFET rectifier Q3220.

[0029] Referring to both FIGS. 2A and 2B, at t0 the power transformer T1 begins its core reset cycle. At the same time, the gate driving signal of Q2218 goes low (waveform 234), the drain of Q2218 (waveform 232), starts going high and through D1220 charges the gate to source capacitor of MOSFET Q3222 turning it on. Diode D1220 and the gate to source capacitance of Q3222 act as a voltage peak detector for the drain voltage of Q2218 waveform 232. The gate of Q3222 will remain charged (Q3 on) during the reset and the dead time period up to a few nanoseconds prior to t2 (power transferring period of T1). At t2 minus the delay 304, the gate charge of Q3222 is removed by the low-power MOSFET Q4221. The secondary 216 of the pulse transformer 215 drives the gate of Q4221. The primary 214 of T2215 is AC-coupled to the output of the PWM 202 through capacitor C1208. As the new power transferring cycle begins (t2 to t3), the output of PWM 202 goes high. The output is delayed 204 before arriving at the gate of Q1206.

[0030] Assuming the propagation delay through C1208 and T2215 is equal to that of power transformer T1, the delay of 204 is arranged equal to the delay through Q4221 and Q3222. In this case, Q3 will be turned off just prior to t2 or the beginning of the power transfer period t2 to t3. To those familiar with the art, it should be obvious that the delay 204 can be implemented by adding a small series resistor from the output of PWM 202 to the gate of Q1, a fixed delay buffer or a combination of both. The parasitic gate to source capacitor of Q1 with the series added gate resistor forms a low-pass filter thus delaying the output of PWM 202. Typically, the delay 304 is about 50 to 100 ns.

[0031] The circuit of FIG. 2A as described above forces Q3 to remain on during the reset and the dead time periods described above. The body diodes of the synchronous MOSFET rectifier are off most of the time and the efficiency is improved by about 8% to 9% in comparison to the self-driven converter given in FIG. 1A.

[0032] Referring to FIG. 2B, the current waveforms of Q2218 and Q3222 are shown as IQ2240 and IQ3238, respectively. By comparing the current waveforms given in FIG. 1B for IQ2 and IQ3 to the same current waveforms in FIG. 2A for present invention, the MOSFET rectifier body diode conduction during the dead time period (t1-t2) is eliminated in FIG. 2B.

[0033] NOTE: If the I.R drop in the transistor channel is higher than the forward drop of the parasitic diodes, these diodes will conduct.

[0034] In FIG. 3, one of the preferred embodiments of the invention is shown. Two low-power MOSFETs are used as gate voltage limiters for Q2218 and Q3222, in case the secondary voltage of T1222 exceeds the maximum gate voltage of the MOSFET synchronous rectifiers Q2 and Q3. Specifically Q5230 is inserted between the cathode of D1220 and the gate resistor of Q3222. The drain terminal of Q5230 is connected to the cathode of D1220 and the source of Q5230 is connected at the node of the drain of Q4221 and the gate resistor of Q3222. The gates Q5230 and Q6232 are driven from a control voltage VC 234. Voltage limiter Q6232 has its source connected to the series gate resistor of Q2218 and its drain of the node connecting the secondary of the power transformer T1212 to the drain of Q3222 and inductor LO 224. The control voltage VC 234 can be generated by a secondary winding(not shown) in T1 or LO and can be adjusted to any voltage level (be it DC or pulsating). Also Q5230 and Q6232 can be bipolar NPN Darlinton transistors, such as the Philips BST50, or a single NPN with a parallel diode oriented as the body diode in the N channel MOSFETs.

[0035] A 50-watt converter based on the FIG. 3 design was constructed within a 2″ by 1″.by 0.4″ case using the following key components: PWM—UCC3807D 8-pin current mode; T1 Core—UG42011 Magnetics with 0.120″ window; LO Core—Philips 3F3 material; T2 core—Toroid 40502-TC Magnetics W or H material; Q2, Q3—5μ 30VDS MOSFET I.R. (International Rectifiers); D1—1A Schottky 60V diode; Q4—VN 3205 Supertex; Q5, Q6—IRF 7341 Dual N Channel MOSFET.

[0036] In one example of the present invention a twenty layer printed circuit board is used for the power transformer T1 and inductor L0 winds, while the two outer surfaces are populated with the surface mounted devices (SMD). The twenty layer board is connected to the load and the input voltage through edge pins, which are mounted via through holes or surface mount.

[0037] A 15-amp converter with 48 volts input and a 3.3 volt output was built according to the present invention and tested. Its efficiency was 89%. The self-driven synchronous MOSFET rectifier design in FIG. 1A was built and tested with an efficiency of 81%.

[0038] For output voltages below 2 volts, the embodiment in FIG. 4 can be used. In FIG. 4, two auxiliary secondary winding 234 and 236 connected in series with the power secondary 212 are used to generate the needed gate voltages for Q3222 and Q2218.

[0039] The voltage limiters Q5230 and Q6232 can be removed if the input voltage to the converter varies only by a ratio of 2:1, e.g., from 72 volts to 36 volts. Both designs in FIGS. 3 and 4 offer wide input voltage range and minimum parts count, which reduces costs.

[0040] A schematic and timing waveform are shown in FIGS. 5A and 5B. This converter with synchronous MOSFET rectifiers Q2318, Q3322 is a semi-self-driven synchronous rectifier forward converter where the gate driving signal is generated by combining input and output control signals.

[0041] Bipolar transistors Q4, Q5, Q6, and Q7 are used as edge trigger devices to charge and discharge the gate to source capacitance of the synchronous rectifiers Q2318, Q3322. Specifically, at t0 the output of PWM 302 goes low and is delayed 304 before reaching Q1. In parallel, it travels via C1308 and T2315 and arrives at the node formed by capacitors 348, 344 and resistor 346.

[0042] The low going edge of the PWM 302 output via capacitor 348 momentarily turns on Q5352 which in turn turns off Q2318. The delay 304 is arranged so that Q1 is turned off at the same time as Q2. Also, at t0, the same low going edge of the PWM momentarily turns on Q6332 through capacitor 334 turning on Q3322. Both PNP transistors Q5352 and Q6332 must be able to carry the needed current for discharging and charging the synchronous MOSFET rectifier's parasitic capacitances, such as gate to source and drain to gate, etc.

[0043] At t1, the drain voltage of Q2318 (FIG. 5B VDQ2382) drops below a control voltage V_cc. Q2318 is turned on again by the low-going edge, trace 382, at its drain. This low going voltage is transmitted via capacitor 357 and PNP transistor Q4360.

[0044] The diode 354 prevents any current flow back flow when VDQ2382 exceeds V_cc329. As in the other preferred embodiments of the present invention, during the dead time period (t1 to t2) both MOSFET rectifiers Q2318 and Q3322 are on and both carry the inductor 324 current. As before the body diode of Q3 is not conducting during the dead time. At t2 the power transferring cycle begins, Q2318 was turned on at t1. The momentary turn-on of the body diode of Q2 described for the self-synchronized synchronous rectifiers in FIG. 1A is eliminated. At t2 the positive edge of the output pulse of PWM 302 through capacitor 308, transformer T2315 and capacitor 344 momentarily turns on Q7340 and turns off Q3322.

[0045] It should be noted that all the bipolar transistors have an off-state bias and at turn-on the converter will operate through the body diodes of the synchronous rectifiers for a few cycles until V_ccreaches voltage levels above the MOSFET rectifiers' threshold voltage.

[0046] The circuit design in FIG. 5A is very flexible. The propagation delay of the output pulse of PWM 302 can be tailored for either the positive or negative edge of the pulse through the R-C resistor networks for each and every bipolar transistor. Base or gate series resistors (not shown) can also be adjusted with a parallel small signal-switching diode.

[0047] The invention in FIG. 5A offers high efficiency for an unclamped forward converter. It eliminates the need for gate voltage limiters and runs cooler than prior art designs, thereby improving reliability.

[0048] The pulse transformer T2315 can be eliminated if input to output isolation is not required. It can be replaced with a high-voltage capacitor diode combination or any other fast-isolation signal buffer. In other preferred embodiments, the bipolar transistor drivers and their associated components can be constructed in an integrated circuit package that will further improve reliability and power density.

[0049] SELF-DRIVEN ACTIVE CLAMP

[0050] In prior art, the active clamp in DC/DC converters has been used to reduce the voltage stress of the primary transistor switch in forward/flyback converters by allowing the magnetic current to circulate and reset the power transformer core. An additional benefit of the active clamp is the improvement of the efficiency of the converter because part of the energy required to reset the power transformer core is returned back to the input source powering the converter.

[0051] In prior art, the active clamp consists of a power MOSFET connected in series with the storage capacitor. The transistor is driven by a control signal generated from the PWM, located at either in-put or output section of the converter. Depending on the location of the active clamp and the type of MOSFET (N channel, P channel), additional components and specific timing between the PWM main power transistor driving signal and that of the active clamp is required. But, such particular circuit designs are well known in the art.

[0052] A self-driven active clamp (SDAC) offers all the benefits of the prior art with fewer components, lower cost and higher reliability. As in prior art, the active clamp can be an N channel or P channel MOSFET; it can be placed at either input or output section of the converter; and it will operate with virtually any power conversion circuit designs.

[0053] In FIG. 6, a block diagram is given for both N channel and P channel (dotted line) self-driven active clamps. The N channel SDAC 404 and the P channel SDAC 408 are shown as a 2-terminal device connected in series to storage capacitor CA 402 and CA 406, respectively.

[0054] In FIG. 7A, a schematic of an N channel SDAC is given, and the associated voltage and current waveforms traces are presented in FIG. 7B. Referring to FIG. 7A, after Q1500 turns off, the voltage at node VD 518 starts to increase from zero volts. When VD 518 becomes one diode drop more positive than VC 520, the parasitic body diode of Q2502 (not shown) starts charging CA 504. At the same time, the gate to source capacitors of Q2502 and Q3510 are charged to VD 518 voltage, Q2502 through the parasitic diode of Q3510 and Q3 through the zener diode 506.

[0055] After VC 520 reaches its peak value, determined by the leakage current path formed by R1508 and zener 506, current starts to flow from VC 520 back to the positive Vin terminal 516 through the primary of T1514 discharging CA 504. When the threshold of Q2502 is reached, the channel of Q2 starts to conduct current from CA 504 and resets the transformer core. Transistor Q3510 turns off Q2 at the end of the transformer core resetting period. The zener diode 506 also protects the gates of Q3510 and Q2502.

[0056] Resistor R2512 biases the gate of Q2 depending on the input voltage, the selected value of CA 504, switching frequency of the converter and the parasitic capacitors of Q2502 and Q3510. The bias resistors R1508 and R2512 are large compared to the on resistance of Q2502 and Q3510.

[0057] At the beginning of the on period when VD 518 drops to zero volts due to Q1 being on, the gate to source voltage of Q3 is clamped by the zener diode 506 turning Q3510 completely on and Q2 off. The gate to source capacitance of Q3 is also discharged through Q1500.

[0058] During the on time of Q1510, capacitor CA 504 continues to discharge through R1 and zener diode 506 to ground through Q1500. During this on period, Q3510 must remain on to keep Q2502 off. For those familiar with the art, it will be obvious that Q2502 and Q3510 can be replaced by bipolar transistors with appropriate biases.

[0059] In FIG. 7B, the voltage and current waveforms from FIG. 7A are shown. Specifically, VD 540 is the drain voltage of Q1500, the current through Q1500 is given as IQ1542, the current through CA 504 is given as ICA 544, and the voltage VC 520 is given as VC 546.

[0060] In FIG. 8, two different versions of a P channel SDAC are presented: the first version consists of CA 602 inverter 604 and P channel MOSFET 606 with its parasitic diode shown. The second version (dotted) consists of CA620 auxiliary winding 610 in series with capacitor 612, resistor 616 and a P channel MOSFET 614 with its parasitic diode shown.

[0061] In the second version, the signal inversion for the gate signal drive of MOSFET 614 is accomplished through the auxiliary transformer winding 610, while capacitor 612 and resistor 616 perform the scaling of the gate signal. Both N and P channel SDACs can be used with any of the preferred embodiments of the forward converters discussed previously with respect to FIGS. 2, 3 and 4.

[0062] A distinction between the inventive SDAC and the prior art is that the driving signal for the SDAC is not turned on “prior to the on period of the primary switch 10 and closed after this period,” as stated in U.S. Pat. No. 4,441,146 on page 5, lines 36-40. All control signals for the inventive SDAC are derived from the signal generated from the primary switch. It should also be noted that even if the SDAC has a dead time period, the inventive synchronous rectification embodiments disclosed herein will improve efficiency and reduce costs.

[0063] It should be understood that above-described embodiments are being presented herein as examples and that many variations and alternatives thereof are possible. Accordingly, the present invention should be viewed broadly as being defined only as set forth in the hereinafter appended claims.