Buck (or boost) converter with very fast output current transitions and low ripple voltage

[0001] In a buck converter, it is desired to have a very fast transient response, but also a very low ripple voltage. In present buck converters, these are conflicting goals.

[0002] In a multiphase buck converter running at precise duty cycles which have an inverse integer relationship with the number of phases and a precise synchronized phase relationship, it is possible to have zero output ripple.

[0003] If the output inductors are replaced with coupled inductors having nominally the same inductance, and the second windings of the inductors are wired in a loop with correct phasing, under steady state conditions the voltage drop around the loop will be zero. A small voltage injected into the loop can cause very fast transitions in current in the loop, and those currents will be reflected to the first windings of the inductors as a very rapid transition in the output current.

[0004] This converter has no dc voltage control capability, so it is preferably used with a pre-regulation stage as the control of the dc output voltage. Dynamic voltage control is very fast, as the current into the output capacitor can change very rapidly.

[0021] FIG. 1 shows a prior art buck converter circuit 1 having an input voltage Vi 3 and an output voltage Vo 15. Switches S1a 5 and S1b 7 are operated alternately to generate an average voltage on an inductor L19, and at steady state that average voltage appears on the output Vo 15, filtered by the output capacitor C111, as would be well known by one skilled in the art of converters. An output load 13 is shown as a variable resistor, as an illustration, not a limitation.

[0022] As a prelude to the introduction of this invention, consider that the buck converter 1 is operated at precisely 0.500 duty cycle. In this case the output voltage Vo is one half of the input voltage Vi, and the buck converter 1 will have no capacity to regulate the output voltage Vo. Accordingly, it is assumed that the input voltage Vi 1 is controlled by a pre-regulator (not shown), and the pre-regulator may be adjusted to control the output voltage Vo. In this respect, it is analogous to a common circuit arrangement in which a first stage regulator feeds a transformer circuit operating at “100 percent” duty cycle and a two to one ratio. The output voltage may be fed back to the first stage regulator.

[0023] The switches S1a 5 and S1b 7 (and all of the other switches in this specification) are preferably “solid state” switches, as examples, not limitations, MOSFETs, JFETs, bi-polar transistors, diodes, rectifiers, synchronous rectifiers and the like. One skilled in the art of power converters would know how to design and use solid state switches appropriate for the circuits, and the nature of the switches in not a point of novelty. Likewise, control circuits are needed to operate the solid state switches with the correct timing and in response to feedback and other control stimuli. One skilled in the art of power converters would know how to design and use such control circuits, and the nature of the control circuits is not a point of novelty. Accordingly, in this specification and the claims, a “switch” or “switching means” is a generic circuit element that can open and close with the correct timing. A “timing means” or “control means” is a controller means that can operate the switches with the appropriate timing. Generic switch symbols are used in the schematics, and the timing is as defined in the specification text and various graphs in the figures. One skilled in the art of power converter design would know how to use this information to make practical circuits.

[0024] FIG. 2 shows a two phase buck converter 21 having an input voltage Vi 23 and an output voltage Vo 41. Switches S1a 25, S1b 27, S2a 35 and S2b 37 generate average voltages on two inductors L129 and L239. The inductor currents il1 and il2 are combined as the converter output current iconv which, when filtered by a capacitor C131 becomes the output current iout to a load Ro 33, shown as a variable resistor as an illustration, not a limitation.

[0025] In FIG. 2, the duty cycles D1 and D2 are precisely 0.500 (except that they can be tweaked to maintain balance between the currents il1 and il2). Accordingly, the output voltage Vo is one half of the input voltage Vi, and the two phase buck converter 21 has no ability to regulate the dc output voltage. In a buck converter, the duty cycle is usually defined as the ratio of the time that the input switch is closed to the total period. Accordingly, in the present example, the first switch S1a 25 is closed 0.500 of the total period. Whenever S1a is open, S1b is closed, as is usual for a buck converter. S1a may be a MOSFET, as an example, not a limitation, and S1b may also be a MOSFET as a synchronous rectifier or it may be a Schottky rectifier, as examples, not limitations.

[0026] To achieve the objectives of this invention, S2a 35 has the same timing as S1b 27, and S2b has the same timing as S1a 25. That is, whenever S1a 25 is closed, S2a 35 is open, and vice versa. With this condition, the converter output current iconv has no ripple component, as shown in FIG. 3. The inductor currents il1 and il2 each have complimentary triangle current wave forms, and their sum is a constant current. At steady state conditions, this equals the output current iout. It should be noted that the ideal duty cycle of 0.500 is for an ideal circuit. In a practical circuit there may be imbalances in component values, voltage drops, circuit impedances and so forth which may cause one current il1 or il2 to drift up or down, and there will be a complimentary drift in the other. This can be corrected by tweaking the duty cycle somewhat, perhaps to 0.490 for one input switch and 0.510 for the other, as an illustration, perhaps more or perhaps less, and this would be expected to vary with operating conditions. However for tweaking the currents to attain current balance while maintaining zero ripple in the converter output current iconv, the switches are nonetheless operated synchronously as described above.

[0027] Unfortunately, this circuit 21 has very poor load regulation, as shown in FIG. 4. While this could be improved using feedback control, this invention teaches a better method of improving load regulation. To complete the discussion, output regulation could be improved through changing the input voltage Vi to cause the inductor currents il1 and il2 to ramp up faster, or the duty cycles could be increased to achieve the same result. However, it would take a large change in the input voltage to make much of an improvement, and any change in the duty cycle relationship of 0.500 would greatly increase the ripple current.

[0028] FIG. 5 shows a two phase buck converter circuit 51 with is similar to the two phase buck converter 21 of FIG. 2 except that the two inductors L129 and L239 have been replaced with coupled inductors T159 and T269. The converter 51 has an input voltage Vi 53 and an output voltage Vo 73. Four switches S1a 55, S1b 57, S2a 65 and S2b 67 are controlled at a precise duty cycle of 0.500 to provide an average voltage to the couple inductors T159 and T269. Currents il1 and il2 combine as the converter current iconv which is then filtered by an output capacitor C161 to a load Ro 63, shown as a variable resistor for discussion purposes.

[0029] The second windings of the coupled inductors T159 and T269 are connected as shown in FIG. 5, where the usual dot convention is used to show the correct phasing. An inductor L 71 may just represent the combined secondary leakage inductance or it may be a purposefully added component.

[0030] With the precise operation of the switches as discussed with FIG. 2, the voltage around the secondary loop of the coupled inductors T159 and T269 is zero. This can be seen in FIG. 5 where the polarity of the voltage across the inductors is shown for the switch positions shown. The voltage across each of the inductors T159 and T269 is one half the input voltage Vi (which also equals the output voltage Vo) with polarity as shown. In the secondary loop, the equal and opposite voltages cancel. The same is voltage cancellation holds true when the switches change state as described in the discussion of FIG. 2 although the polarity in each secondary will be opposite.

[0031] It can be seen by inspection, however, that if a large current could be induced into the secondary loop, it would reflect to the converter output current iconv additively, equally in each of the coupled inductors T159 and T269, given their phase relationship as shown by the dots. It is well known that the current through an inductor cannot change instantaneously or even very fast, but that refers to the net ampere turns in the inductor, or the magneto-motive force. The current in one winding can change arbitrarily fast, however, if there is an equal and opposite change in one or more other windings such that the net ampere turns in the inductor is maintained.

[0032] The benefits of the converter 51 of FIG. 5 can be seen in FIG. 6. It is assumed that the output load Ro 63 has undergone a step change as shown by the curve iout. It can be seen that the current iconv follows extremely rapidly. This can be seen in greater detail in FIGS. 7 and 8, which show, respectively, a step increase and a step decrease in load. In FIG. 7, it can be seen that the inductor currents il1 and il2 have complementary triangular wave forms at steady state. At the transient, however, both ramp up rapidly so that the total converter current iconv will closely follow the output current iout. Similarly, in FIG. 8, both il1 and il2 ramp down rapidly so that the total converter current iconv will closely follow the output current iout.

[0033] There are several ways that this can be accomplished, but they have in common that a voltage is induced or injected into the secondary loop. In the converter 51 of FIG. 5, this may be accomplished by inserting a segment into the timing pulses during which the switches S1a 55 and S2a 65 are momentarily in phase, either both closed (with their compliments S1b 57 and S2b 67 both open) to increase the current rapidly or with both open (with their compliments S1b 57 and S2b 67 both closed) to decrease the current rapidly. It can be seen by inspection that if both switches S1a 55 and S1b 65 are in the same state, the voltages no longer cancel in the secondary loop, they add, and the polarity depends upon which state that the switches S1a 55 and S2a 65 are in. The total voltage around the loop will equal the input voltage Vi, and the impedance of the loop is very small. In FIG. 5, the impedance is just the inductor L 71, and the loop current will rise (or fall) at a rate di/dt equal to Vi/L.

[0034] This can be seen graphically in FIGS. 9 and 10, which show, respectively, the same step changes in current as in FIGS. 7 and 8. In FIG. 9, the both switches S1a 55 and S2a are on (closed) simultaneously as the current rises. In FIG. 9, both are off (open) simultaneously as the current falls. It should be noted, however, that to maintain the complementary triangular wave forms after the transient, the time that both switch are in the same state, both on or both off, is added to the switching period. The graphs in this specification are the output graphs of SPICE simulations, and it was discovered early on that this was necessary to maintain circuit symmetry, that is il1 and il2 being equal on average.

[0035] FIG. 11 shows a converter 101 comprising a voltage pre-regulator 105 having an input voltage Vi 103 and an output voltage Vm 107, and a two phase buck converter of this invention 111 that is the two phase buck converter 51 of FIG. 5 with the addition of a voltage injection circuit 113 to inject a voltage Vc into the secondary loop to control the output current iconv of the converter 101 during transients of the load 73.

[0036] The voltage pre-regulator 105 has an input voltage Vi 103 that may be an unregulated voltage bus. It is contemplated that the voltage pre-regulator may be a buck converter of conventional design, but any voltage regulation means may be used. It is also contemplated that the pre-regulator 105 will control the output voltage Vo 73, with appropriate compensation to ensure overall loop stability, as would be known to one skilled in the art of power converters. The output of the pre-regulator 105 is an intermediate voltage Vm 107 which will be nominally equal to two times the output voltage Vo 73.

[0037] At steady state conditions, the voltage source 113 Vc is zero, but it can controlled, positive or negative, to change the converter output current iconv very rapidly in either direction. The voltage Vc will induce a change of the current is around the loop di/dt equal to Vc/L, and it will reflect equally to both inductor currents il1 and il2 so that it is doubled in the output current iconv of the converter.

[0038] FIG. 12 shows a possible voltage injection circuit 115 for use in the two phase buck converter 111 of FIG. 11. A full bridge circuit comprising four switches S3a 117, S3b 119, S3c 121 and S3d 121 may be operated to inject a control voltage Vc 125 of either polarity into the secondary loop comprising the coupled inductors T159 and T269 and the inductor L 71. In FIG. 12, the switches S3a 117 and S3c 121 are shown closed. It is necessary for the loop to be a closed circuit so that the secondary loop current is can circulate.

[0039] In an ideal circuit, the circulating current in the secondary loop would be loss-less and its magnitude would not matter. A practical circuit would have losses, so it would be desirable for the current in the secondary loop to decay to zero after a transient. This will tend to happen if there is resistance in the secondary loop, and additional resistance can be added to the secondary loop to make the decay faster and more precise. As the current in the secondary loop decays, this will be reflected to the inductor currents il1 and il2, but the change can be fairly slow by design, slow enough so that the total converter current iconv can be maintained by tweaking the intermediate voltage Vm under feedback control.

[0040] In FIG. 5, the situation is different. While the same effect is present, superimposed on it is an effect of the secondary loop voltage. If the intermediate voltage Vm is not equal to exactly two times the output voltage Vo, a difference voltage will appear in the secondary loop. While this voltage is much smaller than the effect of having the switches S1a and S2a in phase, or of injecting a voltage Vc, it will nonetheless cause the current to rise in the secondary loop and that will be reflected to the inductor currents il1 and il2. Thus this is an alternative method of current control, and one that is less sensitive, yet still much improved over response of the circuit of FIG. 2.

[0041] For comparison, consider that the in-phase switching method is used. In this case, the net voltage around the secondary loop is plus or minus the intermediate voltage Vm. If an injected voltage is used, the net voltage around the secondary loop is plus or minus the control voltage Vc. If the intermediate voltage Vm is raised or lowered to be different than two times the output voltage, then the net voltage around the secondary loop is plus or minus the difference voltage Vm−2 Vo.

[0042] FIG. 13 shows the input current wave form iin for a step increase in load current. While there is a small saw-tooth ripple, it is a continuous current. It increases very rapidly, however, as the load current increases. With reference to FIG. 11, the current iin is the current out of the pre-regulator 105. To accommodate the very large step change in current, the pre-regulator 105 must have a low impedance output, as by having a large output capacitor and a good loop response.

[0043] FIG. 14 shows a five phase buck converter 131 having an input voltage Vi 133 and an output voltage Vo 183. Five input switches S1a 135 to S5a 143 and five catch switches S1b 145 to S5b 153 control the average voltage on five coupled inductors T1155 to T5163. When filtered by an output capacitor C1165 the output current is taken to a load 167 shown for illustration purposes as a variable resistor.

[0044] The secondary loop of the five coupled inductors T1155 to T5163 is connected in series, phased as indicated by the dots. An inductor L 181 may represent the total leakage inductance of the loop, or it may be a purposefully added component. There are two resistors R1177 and R2179 in the secondary loop as well. These are used as voltage injection points for a simple voltage injection circuit comprising a control voltage source Vp 169 and two switches Sca 171 and Scb 173 connected to a return. In operation, either Sca 171 or Scb 173 can be closed to inject voltage into the loop and the polarity is controlled by which switch Sca 171 or Scb 173 is closed. This very simple circuit resembles a push pull drive circuit, and its ground referenced control (if MOSFETs are used) is very simple. The resistors R1177 and R2179 be bypassed by closing an optional switch Scc 175 when neither Sca 171 nor Scb 173 is closed.

[0045] This circuit is similar in operation to the circuit 111 described above in reference to FIG. 11. With five phases, there are several ripple free operating conditions, occurring when the duty cycle is equal to m/5, where m is the number of input switches S#a operated so as to be on at the same time, and n is the number of phases. Thus duty cycles of 1/5, 2/5, 3/5, 4/5 and 5/5 are “legal” and will result in ripple free operation. They also result in a net zero voltage around the secondary loop at steady state.

[0046] Two duty cycles are shown in the text on the schematic of FIG. 14, and are practical as processor power converter output stages. Assuming an output voltage in the order of 1 volt, the input voltage Vi can be 2.5 volts or 5.0 volts nominal. Recently a number of papers have suggested that a two stage power converter stage is more efficient for a processor power converter than a single stage operating from 12 volts, as an illustration, not a limitation. Usually, a loosely regulated fixed voltage is generated using a buck converter to power a multiphase output stage buck converter. The same components can be used in this circuit, with the modification of the coupled inductors and voltage injection circuits. Additionally, the control for the first stage may be controlled by feedback from the output.

[0047] In a practical scenario, the voltage on the output capacitor may be sensed to detect a change in output current, or it may be directly measured. Rapid current changes are accommodated by injecting a voltage in the secondary loop to rapidly change the converter current as necessary to maintain the output voltage Vo 183 in regulation. However, the five phase buck converter operating at a fixed duty cycle cannot control the dc output voltage, so a slower feedback loop may control the input voltage Vi 133 so as to fine tune the output voltage Vo 183.

[0048] FIG. 15 shows a multiphase boost converter 201 having an input voltage Vi 203 and an output voltage Vo 225. Four boost switches S1205 to S4211 charge four coupled inductors T1227 to T4233 when any of them is closed, and four rectifiers Cr1213 to Cr4219 transfer the current to an output storage capacitor C1221 when any of the switches S1205 to S4211 is open. An output load 223 is shown as a variable resistor.

[0049] The boost converter is a buck converter operated in reverse, (and recitation of a buck converter in this specification and the claims includes the equivalent “reverse buck” or boost converter) so it is natural to apply the teachings of this invention to a boost converter as well. When operated at a duty-cycle that has an inverse integral relationship to the number of phases, the input ripple current will be zero, as will the voltage around the secondary loop of the coupled inductors T1227 to T4233. (Understanding this requires some sorting out, as the duty cycle is defined differently in the boost converter, and the input and output are swapped). Thus an input stage can be constructed with very low ripple, yet very fast dynamic response by injecting a voltage into the secondary loop using one or more of the three methods described above. In FIG. 15, the voltage is injected from a voltage Vp 241 through a pair of full bridge switches S5a 235 and S5b 237. At steady state conditions, the switches S5a 235 and S5b 237 are closed through, to close the secondary loop with no injected voltage. If both switches are in positions A, one polarity of voltage is applied to the secondary loop, and if both switches are in position B, the other polarity is applied to the secondary loop. An inductor L 239 may represent the stray inductance of the loop or it may be a purposefully installed component.

[0050] As an analog to the alternative current control of FIG. 5, a much smaller differential voltage will be applied to the loop if the input and output voltages Vi and Vo differ from ideal, so a slower transition can be achieved by tweaking Vo. As it would not usually be desirable to have extremely fast transients in the input current, this may be the preferred mode of control for an input boost circuit, an it will simplify the circuit by eliminating the voltage injection circuitry.

[0051] The four phase boost converter was chosen for FIG. 15 to illustrate another characteristic of this circuit. If the input voltage is applied with all of the switches S1205 through S4211 open, the output voltage will tend to charge resonantly to two times the input voltage. If that is the desired output voltage, the switches can then be operated at 0.500 duty cycle to maintain that voltage.

[0052] If the switches are operated with a 0.500 duty cycle when the voltage Vi is applied, the output capacitor will tend to resonantly charge to four times the input voltage. If that is the desired output voltage, the duty cycle can then be changed to 0.750 to maintain that voltage.

[0053] FIG. 16 shows a power converter circuit 301 having an input voltage Vi 303 and an output voltage Vo 305 that, with appropriate controls, may power a microprocessor μ307. The output voltage Vo may be in the order of 1.0 volts at 100 amperes, as an illustration, not a limitation. A buck pre-regulator 309 provides an intermediate voltage Vm 311 to a four phase buck converter 313 of this invention. The switches in this FIG. 16 are shown as MOSFETs, as an illustration, not a limitation.

[0054] The buck pre-regulator 309 comprises an input switch Q1315, a switch to return Q2317, an inductor L1319 and a filter capacitor C1321. Because the four phase buck converter 313 operating at a fixed duty cycle has no ability to control its output voltage, the duty cycle of the buck pre-regulator 309 is modulated by feedback to control the dc output voltage Vo 305 by modulating Vm 311.

[0055] The four phase buck converter 313 comprises four input switches Q3323 to Q6329, four switches to return Q7331 to Q10337, four coupled inductors T1339 to T4345 and an output filter capacitor C2347. The switches are operated at a fixed duty cycle of 0.250 in an inverse integral relationship for zero ripple operation. Accordingly, the output voltage Vo 305 is one fourth of the intermediate voltage Vm 311.

[0056] The second windings of the coupled inductors T1329 to T4345 are connected in series, phased as shown using the dot convention. Also in series with the secondary loop is an inductor I2349 and a voltage injection circuit 351. The voltage injection circuit operates from a power source Vc 353 and comprises a full bridge switch comprising four switches Q11355 to Q14361. The operation of this voltage injection circuit is similar to that described for FIG. 12.

[0057] The output current could also be changed rapidly by altering the phasing of the switches as taught in FIG. 5.

[0058] In operation, micro-processors can have very fast transitions of current, and if they occur, the voltage injection circuit operates so as to cause a very rapid increase or decrease in the secondary loop current which will be reflected to the output times four (the number of phases). Voltage corrections at dc and for slower current transitions may be accommodated by modulating the voltage Vm out of the buck pre-regulator 309.

[0059] It is noteworthy that the current balance in the inductors is not assured. Slight impedance differences or timing irregularities may cause the various inductor currents to drift up or down as in prior art multi-phase buck converters. In the present example, the duty cycle is 0.250 nominally, or one fourth. Accordingly, one and only one of the input switches Q323 to Q6329 will be on at any instant. Thus the current in any phase can be determined with a single current detecting circuit in the input, sampled when the phase in question is “on”. Differences are corrected by tweaking the duty cycle timing slightly.

[0060] For this specification and the claims, a multiphase buck converter is operated in an “inverse integral relationship for zero ripple” if the duty cycle of each phase is equal (except for tweaking) and at every instant the same number of input switches is always on. In the present example, the duty cycle of each phase is 0.250. One of the input switches is always on and three of the switches to return are always on, cycling in an interleaved fashion. If there are n phases (in this example, four), and exactly m input switches are always on (in this example, one), the duty cycle has a inverse integral relationship m/n (1/4 in this example) for zero ripple. Four a four phase buck converter, the duty cycle may be 1/4, 2/4, 3/4 or 4/4 for ripple free operation. (4/4 is a degenerate case, but the relationship is valid). In a boost converter, the duty cycle is defined differently (1-D) and the input to output is reversed, but the zero ripple state (on the input) still occurs when the boost duty cycle is m/n, or 0 (the degenerate but valid case), 1/4, 1/2 or 3/4 (4/4 is not allowed, it is a short circuit on the input), yielding an output voltage, respectively, of Vi, 4/3 Vi, 2 Vi or 4 Vi.