Self driven active clamp

[0002] This invention relates to switching-type power converters and in particular forward and flyback-type converters that have synchronous output rectifiers. Converters with self-synchronized rectifiers refer to MOSFET devices used as rectifiers. The gates of these devices are driven by the secondary voltage signal of the power transformer. The use of so-called self-driven synchronous rectifiers in buck-driven converters is limited by the inefficiency of these rectifiers. The reset voltage for the transformer core in forward converters de-signed without an active clamp limits the conduction time of one of the MOSFETs in the synchronous rectifiers, thus pre-venting the converter from operating at maximum efficiency.

[0003] In the converter shown in FIG. 1A, a conventional for-ward topology of the prior art with an isolating power trans-former is combined with a self-synchronized synchronous rectifier. In such a rectifier, controlled devices are used with the control terminals (gates) being driven by the secondary winding of the power transformers.

[0004] A DC input voltage, V_IN, at input 100, 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 a synchronous rectifier including the MOSFET rec-tifier devices Q2110 and Q3112. Each rectifying device includes a (parasitic) body diode 120 and 122 respectively.

[0005] With the power switch Q1104 conducting, the input volt-age is applied across the primary winding 106. The second-ary winding 108 is oriented in polarity to the primary voltage with a current flow through the inductor 114, the load is connected to the output lead 118 and back through the MOSFET rectifier Q2110 to the secondary winding 108.

[0006] The current path provided by the conduction of the MOSFET rectifier Q3112 maintains continuity of the cur-rent flow in the inductor 114. An output filter capacitor 116 shunts the output of the converter.

[0007] The gate drive signals for the MOSFET rectifiers are pro-vided by the voltage appearing across the secondary winding 108.

[0008] In FIG. 1B, the voltage and current waveforms of the converter from FIG. 1A are shown graphically. Referring to FIG. 1B, as the pulse width modulator (PWM) 102, turn off MOSFET 104 disconnecting the primary winding 106 of the power transformer.

[0009] The drain voltage of Q1140 rises and due to parasitic capacitance of the input and output, assumes a sinusoidal wave-form from t0 to t1. During this period, t0-t1, the power transformer core resets itself (reset period). The top or sinusoidal portion of 140 appears as the drain voltage of Q2 and the gate of Q3 shown as VDQ2/VGQ3 in 142. During the reset period, t0-t1, IQ2146 drops to zero and all the current through the inductor 114 is supplied by Q3 shown in 148. Specifically IQ3148 assumes its maximum at t0 and then linearly decreases from t0 to t1. 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 a so-called “dead time” period in which no power transfer or core resetting takes place in the power transformer.

[0010] At t1(t) to t2, the inductor current shown in 150 continues to flow through the body diodes of Q3122 and Q2120. Most of the inductor current flows through 122 and a small portion through 120 due to the fact that the secondary winding of the power transformer is in series with diode 120. The forward voltage drop of the MOSFET body diodes is close to 1 volt, which is very high when compared to 0.1 volts when the MOSFET is on.

[0011] The dead time period, t1-t2, increases with increasing in-put voltage line V_IN. This large increase of the dead time, which assumes its maximum value at maximum input volt-age V_IN, results in the synchronous rectifier of FIG. 1 offering its worst efficiency at maximum input voltage.

[0012] At 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, the body diode is short circuited by the channel of Q2110 allowing all the secondary current to go through the channel.

[0013] 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.

[0014] During t2-t3, energy is stored at the output inductor 114, the charge is restored at output capacitor 116 and all the load current is carried through Q2110. The gate driving signal for Q2110 is shown in 144, which is the inverted lower part of the drain waveform of Q1140. The synchronous rectifiers driven directly from the secondary of the forward converter shown in FIG. 1A prohibit the converter from achieving maximum possible efficiency due to the operating nature of the unclaimed forward converter.

[0015] 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 and the converter's reliability making the converter unsuitable for high-power density. (Therefore there is a need for an improved version of a forward unclamped converter with synchronous rectifiers.)

[0016] Forward converter designs utilizing an “active clamp” to clamp the primary voltage generated during the reset period of the transformer core, when the converter must operate over more than 2:1 input voltage range, at high input voltage also enter into a dead time period.

[0017] Forward converters with an active clamp require more complex timing circuitry and additional power MOSFETs. Also, as it is pointed above when the converter operates over a wide input range, the self-driven synchronous rectifiers will encounter a dead time period.

[0018] An active clamp circuit is described by Vicor Corporation in its U.S. Pat. No. 4,441,146.

[0019] This invention combines a storage capacitor and control circuitry for charging and discharging the storage capacitor. The capacitor provides energy that switches or resets a transformer winding of a forward DC/DC converter. The drive signals for the control circuitry are taken from transformer windings. In virtually all DC/DC converters, a pulse width modulator signal drives an FET switch in the primary of a transformer. When that FET switch is turned OFF, the drain of the FET, and transformer connection to that primary switch, rises and through drive circuitry turns stores magnetic energy from the transformer in the storage capacitor. After the capacitor has reached its highest level the capacitor will start to discharge through a path provided, and thereby turns on a solid state switch that further discharges the storage capacitor through the transformer thereby resetting the transformer core. When the primary switch turns back ON, the solid state switch is turned OFF via the drive circuit and the capacitor is further discharged.

[0020] Bipolar or N, or P type FET transistors may be used in the control circuitry, as is known in the art.

[0021] The solid state switch is turned ON after the switch in the primary is turned OFF, and the solid state switch is turned OFF after the primary switch is turned ON.

[0022] For a more complete understanding of the present invention and the advantages it offers, please reference the following descriptions with the accompanying drawings.

[0023] FIG. 1A is a schematic of a forward-type converter of the prior art, but employing a synchronous rectifier.

[0024] FIG. 1B is the voltage and current waveforms for the forward converter presented in FIG. 1A.

[0025] FIGS. 2A & 2B are the schematic and associated wave-forms, respectively, for describing the present invention.•

[0026] FIG. 3 is the schematic of the present invention implemented with gate voltage limiters for the synchronous rectifiers.

[0027] FIG. 4 is the schematic of the present invention with gate voltage limiters (FIG. 3) specifically designed for low output voltage or wide input voltage ranges with a boost in the secondary synchronous rectifier driving voltage.

[0028] FIGS. 5A & 5B are schematics and associated wave-forms, respectively, of the present invention with overlap-ping synchronous rectifiers driving signal independent delay control without voltage limiters or secondary boosting windings.•

[0029] FIG. 6 is a block diagram of a self-driven active clamp utilizing either N or P channel MOSFET.

[0030] FIGS. 7A and 7B are schematics and associated wave-forms of the forward converter with improved synchronous rectification and N channel self-driven active clamp.•

[0031] FIG. 8 is a schematic of a two-version P channel self-driven active clamp.

[0032] FIGS. 2A & 2B present a forward converter with semi-selfdriven synchronous MOSFET rectifiers. The circuit is derived from the circuit in FIG. 1A by adding 4 discrete low-cost parts D1220, Q4221, T2215, C1208 and a delay circuit 204. The voltage and current waveforms for the for-ward converter in FIG. 2A are shown graphically in FIG. 2B. The operation of the circuit in FIG. 2A is the same as described for the circuit in FIG. 1A with the exception of the driving signal for the synchronous MOSFET rectifier Q3222.

[0033] Referring to both FIGS. 2A & 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.

[0034] Diode D1220 and the parasitic 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 50-100 ns 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 and is delayed by a small amount of time through delay 204 50-100 ns before arriving at the gate of Q1206.

[0035] Assuming the propagation delay through C1208 and T2215 is equal to that of power transformer T1, the delay of 204 must be equal to that of Q4221 plus Q3222. The delay of 204 is needed to turn Q3222 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 with the series-added gate resistor forms a low-pass filter thus delaying the output of PWM 202.

[0036] The semi-self-driven synchronous MOSFET rectifier pre-sented in FIG. 2A and described above forces Q3 to re-main on during the reset and the dead time periods of the power transformer in a conventional (without active clamp) forward DC/DC converter. The body diodes of the synchronous MOSFET rectifier are off most of the time and the efficiency is improved by 8% to 9% in comparison to the self-driven converter given in FIG. 1A.

[0037] Referring to FIG. 2B, the current waveforms of Q2218 and Q3222 are given as IQ2240 and IQ3238, respectively. By comparing the current waveforms given in FIG. 1B for the self-driven synchronous MOSFET rectifiers of IQ2 and IQ3 to the same current waveforms given in FIG. 2A for the present invention, it is obvious the current discontinuities during the dead time period (t1-t2) due to the body diode conduction are removed from the current waveforms in FIG. 2B.

[0038] NOTE: If the I_O.R drop in the transistor channel is higher than the forward drop of the parasitic diodes, these diodes will be forced to conduct.

[0039] 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 T1212 exceeds the maximum gate voltage of the MOSFET synchronous rectifiers Q2, 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; 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 in T1 or LO and can be adjusted to any volt-age level (be it DC or pulsating). Also Q5230, Q6232 can be bipolar NPN Darlington transistors, such as the Philips BST50, or single NPN with parallel diode oriented as the body diode in the N channel MOSFETs.

[0040] A 50-watt converter based on the design given in FIG. 3 was constructed in a 27″.1″.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—0.005 ohm on channel resistance.30VDS MOSFET I.R. (International Rectifiers); D1—1A Schottky 60V diode; Q4—VN 3205 Supertex; Q5, Q6—IRF 7341 Dual N Channel MOSFET.

[0041] A multi-layer (20 layer) printed circuit board is used for the power transformer T1 and inductor LO winds, while the two outer surfaces are populated with the SMD components.

[0042] A 3.3V_OUT15-amp converter using 48V_INwas tested. Its efficiency was 89% and by short-circuiting D1, the self-driven synchronous MOSFET rectifier design in FIG. 1A was implemented and tested to offer efficiency of 81%. The multi-layer board is connected to the load and V_INthrough edge pins, which can be surface mounted or though hole.

[0043] For output voltages VO below 2 volts, the embodiment in FIG. 4 can be used. In FIG. 4, two auxiliary secondary winding 234, 236 connected in series with the power secondary 212 are used to generate the needed gate voltages for Q3222, Q2218.

[0044] The voltage limiters Q5230, Q6232 can be removed if the input voltage to the converter varies only by a ratio of 2:1 (36 to 72 volts (High line V_IN/Low line V_IN)). Both designs in FIGS. 3 and 4 offer wide input voltage range and minimum parts count, which results in low cost.

[0045] A schematic and timing waveform is given in FIGS. 5A & 5B. The 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.

[0046] Bipolar transistors Q4, Q5, Q6, 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 as it is delayed by delay 304 while it propagates through C1308 and T2315 and arrives at the node formed by capacitors 348, 344 and resistor 346.

[0047] The negative-going edge of the PWM 302 pulse through capacitor 348 momentarily turns on Q5352 which in turn discharges Q2318 and turns it off at the same time Q1 is turned off by the same PWM pulse going through delay 304. Also, at t0, the same negative pulse 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 capacitance such as gate to source, drain to gate, etc.

[0048] At t1, the drain voltage of Q2318 (FIG. 5B VDQ2382) drops below V_CC. Q2318 is turned on again by the negative-going edge of its drain voltage waveform 382 through capacitor 357 and PNP transistor Q4360.

[0049] The diode 354 prevents any current flow back to its anode node when VDQ2382 exceeds V_CC329. During the dead time period (t1 to t2) both MOSFET rectifiers Q2318 and Q3322 are on and both carry the inductor 324 current. The IR drop in Q3322 is reduced during the dead time period (t1-t2). At t2 the power transferring cycle begins, Q2318 was turned on at t1 assisting Q3322 and waits for the next power transferring cycle, which in this case, starts at t2. 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.

[0050] 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.

[0051] Even though it was not described herein, the 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 through series a gate resistor in parallel with a small signal-switching diode.

[0052] The benefits of the invention in FIG. 5A offers the high-est possible efficiency for an unclamped forward converter. It eliminates the need for gate voltage limiters and has better reliability because it runs cooler. However, it does require more parts and area.

[0053] It is also obvious that 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. If the bipolar transistor drivers and their associated components can be replaced by an integrated circuit (e.g. an 8-pin IC), its reliability and power density will be further improved.

[0054] In prior art, the active clamp in DC/DC converters have been used to reduce the voltage stress of the primary transistor switch in forward/flyback converters (see references) 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 of part of the energy required to reset the power transformer core is returned back to the input source powering the converter.

[0055] 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 main control device, pulse width modulator (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.

[0056] The self-driven active clamp (SDAC) offers all the benefits of the prior art with less components, lower cost and higher reliability. As in prior art, the active clamp can be an N channel or P channel MOSFET and be placed at either input or output section of the converter and will operate with any power conversion topology.

[0057] In FIG. 6, a block diagram is given of both N channel and P channel (dotted line) self-driven 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.

[0058] In FIG. 7A, a schematic of an N channel SDAC is given, while in FIG. 7B the associated voltage and current wave-forms are presented. Referring to FIG. 7A, the operation of the SDAC can be described as follows: 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 diode of Q2502 (not shown) starts charging CA 504. At the same time, the gate to source ca-pacitors of Q2502 and Q3510 are charged to VD 518 volt-age, Q2502 through the parasitic diode of Q3510 and Q3 through the zener diode 506.

[0059] After VC 520 reaches its peak value due to the leakage current path formed by R1508 and zener 506, current starts to flow from VC 520 back to 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 core. Transistor Q3510 turns off Q2 at the end of the resetting period of the transformer core. The zener diode 506 also protects the gates of Q3510 and Q2502.

[0060] 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 resistor R1508, R2512 are very large compared to the on resistance of Q2502, Q3510.

[0061] At the beginning of the on period when VD 518 drops to zero, 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.

[0062] During the on time of Q1510, capacitor CA 504 continues to discharge through R1 and zener diode 506 to ground through Q1500. During the on period Q3510 must remain on to keep Q2502 off. For those familiar with the art, it will be obvious that Q2502, Q3510 can be replaced by a bipolar transistor with different bias scheme.

[0063] In FIG. 7B, the voltage and current waveforms from FIG. 6A are shown. Specifically, VD 540 is the drain voltage of Q1500 in FIG. 7A. 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.

[0064] 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 consists of CA620 auxiliary winding 610 in series with capacitor 612, resistor 616 and a P channel MOSFET 614 with its parasitic diode shown.

[0065] In the first version, the inverter 604, MOSFET 606 and any signal condition required can be implemented on a single semiconductor die. 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 converter with improved synchronous rectification discussed previously (FIGS. 2 through 4).

[0066] The major difference between 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 444146 (page 5, line 36-40). All control signals for SDAC are derived from the signal generated from the primary switch; it is not needed for specific delays, isolation transformers or specialized drivers as is specified in reference 6. It should also be obvious that even if the SDAC allows the converter to assume a dead time period, use of the improved synchronous rectification embodiments outlined here can bring about the realization of a lowcost, high-efficiency forward converter.