60/303,125, filed July 5,2001, which is hereby incorporated herein by reference. This application is related to U. S. Patent Application filed July 3,2002, to Miki Brkovic, entitled: Inductor Current Sensing In Isolated Switching Regulators And Related Methods, hereby incorporated herein by reference.

1. Field of the Invention [0001] This invention relates generally to switch-mode regulators in general and, more particularly, to low output voltage switching regulators with isolation.

2. Background Discussion [0002] Switch-mode regulators are widely used to supply power to electronic devices, such as in computers, printers, telecommunication equipment, and other devices.

Such switch-mode regulators are available in variety of configurations for producing the desired output voltage or current from a source voltage with or without galvanic isolation.

The former are also known as an isolated power converters and the later are called a non- isolated power converters.

[0003] One of the more challenging loads for power supplies are microprocessors.

Because most of the microprocessors are implemented in complementary metal-oxide- semiconductor (CMOS) technology, the power dissipation of the microprocessor generally increases linearly with the clock frequency and to the square of the power supply voltage.

There are three common techniques used to reduce power dissipation: power supply voltage reduction, selective clock speed reduction and reducing capacitive loading of internal nodes within the microprocessor. The first two techniques may be used in combination and could be controlled by circuit designer. Even a small reduction in power supply voltage makes a significant reduction in power dissipation. Also, if the clock is removed or significantly slowed in portions of the microprocessor not being used at any given time, very little power is dissipated in those portions and the overall power dissipated is significantly reduced.

[0004] However, these power savings techniques come at a cost. Power supply current can swing widely-from hundreds of milli-amperes to over few tens amperes with the microprocessor unable to tolerate more than a few percent change in voltage. Further, the change in current can occur in tens of nanoseconds and may change in an order of magnitude. The power supply designed to supply the microprocessor must have a sufficient low impedance and tight regulation to supply such dynamic power consumption.

With output voltages approaching 1V or even sub-volt levels and load currents approaching hundred amperes, the power supplies are very difficult to make and control and still operate efficiently.

[0005] In addition, a dedicated power supply for the microprocessor has to be placed in close proximity to the microprocessor. Thus, the power supply must be small and efficient. To meet these requirements, a small DC-to-DC switching power regulator is usually used. The widely used switching regulator to convert a higher input voltage (usually 5V or 12V) to a lower output voltage level is"buck"regulator. In applications where input voltage, often referred as bus voltage, is greater than 12V (e. g. 24V or 48V) single non-isolated switching regulator, such as"buck"regulator, becomes very difficult to make small and efficient. In addition, in these applications a galvanic isolation is very often required thus switching regulator needs to have isolation. One of the most common approaches is to use two stage conversion. First stage conversion is provided using an isolated switching regulator in order to provide galvanic isolation and to step-down high voltage input bus (typically 48V) to lower voltage bus (5V or 3. 3V). The second stage is then realized using"buck"switching regulator. Obvious disadvantages of this approach are need for two switching regulators which increases overall cost and reduces overall efficiency.

[0006] Three kinds of feedback are generally used to control the operation of the regulator: voltage alone (with current limiting), voltage with peak current control, and voltage with the average current control. For reference see"Fueling the Megaprocessors- Empowering Dynamic Energy Management"by Bob Mammano, published by the Unitrode Corporation, 1996. The voltage with the average current control type of regulation is generally preferred over the other types for the described reasons. Regardless of which type of the feedback control is used, there is need for output current sensing either directly on indirectly. The most common approach is to the sensing resistor in series with the output inductor.

[0007] The circuit reconstructs the output inductor current as a differential voltage across the sensing resistor. Most integrated circuits using this approach regulates output voltage with current mode control and use the signal for output voltage feedback. The sensing resistor value must be on one side large enough to provide a sufficiently high voltage, usually tens of millivolts, to overcome input offset errors of the sense amplifier coupled to the sensing resistor and yet small enough to avoid excessive power dissipation.

Since the power dissipated in the sensing resistor increases with the square of current, this approach has the obvious efficiency drawback with high output current and low output voltage. For low voltage, high current applications, the value of the sensing resistor may be close or even higher than the on resistance of the power switch and inductor which are minimized for maximum efficiency. Thus, sensed signal is relatively small and requires use of more expensive either comparators or amplifiers. Further, the circuitry implementing the average current control technique is significantly more complicated than the circuitry of the other two techniques.

[0008] Power inductors are known to have parasitic (or inherent) winding resistance, and therefore can be represented by an equivalent circuit of a series combination of an ideal inductor and a resistor. When direct (DC) current flows through the inductor (or a current having a DC component), a DC voltage drop is imposed across the inductor, which voltage is a product of the magnitude of the DC (component of the) current and the parasitic resistance of the inductor. Since such an inductor may already be present in the circuit, there is no an additional loss of efficiency in using the inductor for this purpose.

[0009] Parasitic resistance of the output inductor is used for current sensing as described in U. S. Pat. No. 5,465, 201, issued to Cohen, U. S. Pat. No. 5,877, 611, issued to Brkovic, U. S. Pat. No. 5,982, 160, issued to Walters et al. and U. S. Pat. No. 6,127, 814, issued to Goder, all of which patents are hereby incorporated herein by reference. In U. S.

Pat. No. 5,877, 611 and U. S. Pat. No. 5, 982, 160 load current dependant output voltage regulation employing inductor current sensing is proposed. Again, sensed signal is limited to product of inductor's winding resistance and inductor current and can be increased only by means of active amplification, which adds complexity, inaccuracy and mostly additional cost. In order to maximize efficiency of the converter, inductor's parasitic resistance (particular at high current applications) has to be minimized thus, the sensed signal is relatively small and requires use of more expensive either comparators or amplifiers. Very often error due to offset in comparator and/or amplifier is larger than variation in the winding resistance of the inductor (windings printed on the PCB).

[0010] Perhaps the most common approach to sensing the output inductor current indirectly in isolated topologies is to use sense resistor in series with power switches or current sense transformer. Use of sense resistor in single ended topologies, such as for example forward, flayback and others, as well as in full-bridge and push-pull topologies, allows that one end of sense resistor is coupled to GND pin of control chip, usually coupled to input return, which simplifies current sensing. On other hand, the sensing resistor value must be large enough to keep the sensed signal above the noise floor and yet small enough to avoid excessive power dissipation. In case of half-bridge converter, for example, this approach is not good since only one power switch is coupled to input return and sensed signal does not reflect current through second, floating power switch. Using sense resistor in return input path is also not good solution since sensed current is not exactly current through primary side switches but rather an input current of the converter smoothed by input capacitors. Also, sensed switch current differs from the output inductor current due to magnetizing current of isolation transformer which also varies with the input voltage. Using current sense transformer is not practical solution for two main reasons: it still measures the sum of the magnetizing and reflected output inductor current and it becomes difficult to implement in low profile high power density switching regulators.

SUMMARY OF THE INVENTION [0011] Therefore, it is an object of the present invention to provide an efficient isolated switching regulator having a voltage and current control technique.

[0012] It is another aspect of the invention to provide a switching regulator having a fast transient response with relatively simple control circuitry.

[0013] It is a further aspect of the invention to provide a switching regulator design that allows for parallel operation.

[0014] This and other aspects of the invention may be obtained generally in a computing system, a switching regulator for powering a load including microprocessor, the switching regulator comprising an input voltage source, a switching circuit and an transformer for coupling the input voltage source and an output stage, the switching circuit comprising at least two controllable power switches and a first and a second rectifier switch, the transformer comprising at least one primary winding coupled to the input voltage source via at least two controllable power switches and a first and a second secondary windings, the first and the second secondary winding having inherent resistances, the first and the second secondary windings coupled together in series at a first node, the first and the second rectifier switches coupled in series at a second node, the first secondary windings coupled to the first rectifier switch at a third node, the second secondary winding coupled to the second rectifier switch at fourth node, the output stage having an output stage input and a output stage output, the output stage input coupled to the first node and the second node, the output stage output coupled to a load circuit, the output stage input and the output stage output having a common node, the output stage comprising an inductor for providing output current to the load circuit, the inductor having a first terminal for coupling to the output stage input and a second terminal for coupling to the output stage output, a first resistor having a first and a second terminal, a second resistor having a first and a second terminal, a capacitor having a first and a second terminal, the first terminal of the first resistor coupled to the node three and the first terminal of the second resistor coupled to the node four, the second terminal of the first resistor coupled to the second terminal of the second resistor and the first terminal of the capacitor at a node five, the second terminal of the capacitor coupled to the second terminal of the inductor and the node five is coupled to an input of an error amplifier for controlling the switching circuit.

[0015] In one embodiment of the present invention a third resistor is disposed in parallel with the capacitor for additional adjustment of the droop voltage.

[0016] In one embodiment of the present invention a first impedance is coupled between the fifth node and an inverting input of the error amplifier and a second impedance is disposed between the output and inverting input of the error amplifier.

[0017] In one embodiment of the present invention a fourth resistor is connected to the error amplifier inverting input for adjusting output voltage to a reference voltage.

[0018] In one embodiment of the present invention a third impedance is coupled between the output stage output and inverting input of said error amplifier. In one embodiment, the first impedance provides DC coupling so the output voltage is load dependant while, the third impedance provides only AC coupling from the output voltage of the switching regulator for improved transient response. In yet another embodiment, the first impedance provides only AC coupling while, the third impedance provides DC coupling from the output voltage of the switching regulator for load independent regulation. By selecting the first impedance to provide only AC coupling, current feedback is fed into the error amplifier only during the load current transients thus providing more stable transient response.

[0019] In one embodiment of the present invention a switching regulator with load dependent regulation is parallel-coupled to a plurality of switching regulators sharing a common reference voltage and common output coupled to the load.

BRIEF DESCRIPTION OF THE DRAWING [0020] The objects, advantages and features of the invention will be more clearly perceived from the following detailed description, when read in conjunction with the accompanying drawing, in which: FIG. 1 is an exemplary schematic diagram of exemplary paralleled half-bridge switching regulators, each with voltage and current control; FIG. 2 is a partial circuit diagram of one embodiment for further improved transient response of a switching regulator from FIG. 1.

FIG. 3 illustrates effect of different values of resistors and capacitor in current sensor on the transient response of a switching regulator shown in FIG. 1.

DETAILED DESCRIPTION [0021] With reference now to FIG. 1, there is shown an exemplary computing system 5, having a switching regulator 10 for powering a load 15 including a microprocessor 16. The switching regulator 10 is a half-bridge regulator, shown as an example, having power switches 101,102, rectifier switches 103 and 104, an input capacitors 11 and 12, an output capacitor 109, a transformer 200, an inductor 300 and an error amplifier 110. The error amplifier 110 has an input for controlling the power switches, 101 and 102, and rectifier switches, 103 and 104. Transformer 200 has one primary winding 203 and two secondary windings 201 and 202 coupled in series at a node 9. Secondary winding 201 is coupled to the rectifier switch 103 at a node 5 and secondary winding 202 is coupled to the rectifier switch 104 at a node 6. The inductor 300 is coupled between node 9 and output capacitor 109 at a node 8. A first resistor 105 is coupled between node 5 and to node 7. A second resistor 106 is coupled between node 6 and node 7. A capacitor 107 is coupled between node 7 and node 8. Node 7 is in turn coupled to the input of the amplifier 110. Thus, first resistor 105, second resistor 106 and capacitor 107 combine to be the feedback path for controlling the switching regulator 10.

[0022] In more detail, the switching regulator 10, here a half-bridge regulator, takes an input voltage VIN and converts it to a lower voltage with galvanic isolation for use by load 15. The load 15 is illustrated here as a microprocessor 13 with an exemplary one of a plurality of bypass capacitors 17 and inductance 18 (shown as a lumped inductance) representing the distributed inductance of the power supply printed circuit board (PCB) traces.

[0023] The switching regulator 10 includes power switches 101,102, rectifier switches 103 and 104, an input capacitors 11 and 12, an output capacitor 109, a transformer 200 and an inductor 300 (which includes an inherent DC resistance Rw, discussed below). Input voltage VIN is split with capacitors 11 and 12 coupled in series at a node B. Power switches are coupled in series at a node A and across the input voltage VIN. Transformer 200 has one primary winding 203 (with number of turns Np) coupled between node A and node B and two secondary windings 201 and 202 (which include an inherent DC resistance RSl and RS2, and leakage inductance LSl and Læ) respectively and discussed below) coupled in series and to the inductor 300 at a node 9. Secondary winding 201 is also coupled to the rectifier switch 103 at a node 5 and secondary winding 202 is also coupled to the rectifier switch 104 at a node 6. The inductor 300 is also coupled to the output capacitor 109 at a node 8. Drive waveforms for switches are such that when power switch 101 is on, power switch 102 and rectifier switch 104 are both off and switch 103 is on. During on time of power switch 101 the power is delivered from VIN to the load 15 and load current flows through power switch 101, primary winding 203, rectifier switch 103, secondary winding 201, and output inductor 300. In contrast, when power switch 102 is on, power switch 101 and rectifier switch 103 are off and rectifier switch 104 is on.

During on time of power switch 102 the power is delivered from VIN to the load 15 and load current flows through power switch 102, primary winding 203, rectifier switch 104, secondary winding 202 and output inductor 300. When both power switches 101 and 102 are off, rectifier switches 103 and 104 are both on, load current is supplied from inductor 300 and splits in between rectifier switches 103 and 104, and secondary windings 201 and 202. As a consequence, during off time (non-conducting time) of power switches 101 and 102, windings of the transformer 200 are shorted. A first resistor 105 is coupled between node 5 and to node 7 and a second resistor 106 is coupled between node 6 and node 7. A capacitor 107 is coupled between node 7 and node 8 and node 7 is in turn coupled to an error amplifier 110, having in combination impedances Zl, Za and operational amplifier 111. Impedances Zl, Z2 may include reactive elements to achieve optimum compensation to the overall operation of the regulator 10.

[0024] The output of amplifier 110 drives a modulator 112. The modulator 112 (shown here as a conventional pulse-width modulator) generates signal for adjusting an operating parameter of the driver circuitry 113. The drive circuitry 113 than controls switches 101,102, 201 and 202, as described above, and adjusts operating parameter of the regulator to regulate output voltage Vo. Those of skill in art will readily appreciate the various constructions and operations of the modulator 112 and drive circuitry 113. The modulator could be, for example, pulse width modulator (PWM) as well as any other type of modulator operating at constant or variable switching frequency.

[0025] The combination of resistors 105 and 106, and capacitor 107 serve to provide to the error amplifier 110 signals representing the output voltage Vo and output inductor's current Io from the regulator 10. The output current is substantially determined by the voltage drop across the resistance Rw of inductor 300 and resistance Rsi and Rs2 of windings 201 and 202, respectively. DC voltage across capacitor 107, is VF = (Rw + RS/2) Io. Voltage at node 7 is approximately Vo + Io (Rw+Rs/2). Since voltage at node 7 is regulated by the closed loop nature of the regulator 10 to voltage VREF, the output voltage Vo is maintained to be substantially equal to VREF-Io (Rw + RS/2). Consequently, the output impedance is approximately the sum of resistances Rw + Rs/2. Accordingly, the output voltage Vo droops with increasing current Io. Magnitude of droop is than determined by sum of resistances Rw and Rs/2 rather than only by resistance Rw as in prior art.

[0026] The main advantage of the invention is that the voltage VF across capacitor 107 is (l+0. 5*RS/RW) times higher than in prior art U. S. Pat. No. 5, 877, 611, for the same DC resistance of the output inductor and load current, which patent is hereby incorporated herein by reference. In order to demonstrate why increase in voltage VF by 0. 5*Rs/RL is indeed significant improvement over prior art consider the following example: Io = 40 A, Rw = 0.5 mg and Rs = lmQ. Maximum droop voltage in prior art will be 20 mV while in the present invention will be 40 mV, thus twice bigger without any additional active circuitry for amplification. It is obvious that invention provides larger droop voltage and consequently reduces power dissipation in the microprocessor.

[0027] As noted above, the output voltage Vo decreases with increasing output current Io. At zero load current (Io = 0), the output voltage Vo = VREF. When load current increases, the output voltage Vo decreases thus, behaving as a voltage source with open circuit voltage equal to Vo and output DC resistance (Rw + Rs/2). A resistor 108 is added across capacitor 107 to reduce voltage across capacitor 107, if an additional adjustment of droop voltage is require. In this case, the output voltage Vo is approximately VREF-lo (Rw + RS/2)R107/(R107+RS/2) where Rio ? is the resistance of resistor 107.

[0028] A further adaptation is the addition of resistor 114 to the error amplifier 110. Resistor 114 combined with impedance Z1 allows the output voltage Vo to be scaled to the reference voltage VREF and can be set accordingly.

[0029] The choice of resistors 105 and 106, and capacitor 107 affects the response of the switching regulator to transient in output current. A first time constant, inherent to power stage of switching regulator, is defined by the leakage inductances Lsl and LS2 of secondary windings 201 and 202, respectively and inductance Lo of inductor 300 and their DC resistances Rsi, RS2 and Rw, respectively. A second time constant is defined by resistances Rios and R106 of resistors 105 and 106, respectively and capacitance CF of capacitor 107. The first time constant is defined as ratio of equivalent resistance REL = Rw + Rios)) Rio6 and an equivalent inductance LEL = Lo + Lsi 11 LS2, i. e. REL/LEL where, symbol represents parallel combination of inductances Ls, and LS2, and resistors R105 and Rios, respectively. In most practical realization of the converters the leakage inductances, Lsi and LS2, are more than orders of magnitude smaller that inductance Lo in which case the equivalent inductance is approximately LEL zLo and the first time constant is approximately-ri = REL/Lo. The second time constant is defined as product of equivalent resistance R105 # R106 and capacitance CF of capacitor 107, T2 = CF (R105 # R106).

[0030] Relative ration between the first and second time constant affects the response of the switching regulator to transients in output current as illustrated in FIG. 2.

In most cases, the flattest response may be the most desirable in which case the second time constant is chosen to be substantially equal to the first time constant, T2 = T1- [00311 It is known to those of skill in the art that while current mode control provides the most stable transient response, the voltage mode control provides the fastest transient response but very often with undesirable oscillations in the output voltage. In one embodiment of the present invention, an impedance Z3 is added to error amplifier 110, as shown in FIG. 3. The output voltage Vo and current dependant voltage at node 7 are fed into operational amplifier 111 and compared with voltage VnEF. Impedances Zi, Z2 and Z3 are chosen such that optimum transient response is achieved for given application.

Impedance Z1 feeds voltage proportional to output current (needed for both droop regulation and stable transient response) into operational amplifier 110 while, impedance Z2 feeds output voltage Vo (needed for fast response) into operational amplifier 110.

[0032] In one embodiment of present invention impedance Z1 is chosen to provide only AC coupling from node 7 to the input of operational amplifier 111 while, impedance Z2 provides DC coupling from the output Vo to the input of operational amplifier 111. The output voltage regulation is load independent and current feedback through impedance Z1 affects output voltage regulation loop only during the load current transients providing more stable response.

[0033] In one embodiment of present invention impedance Zi is chosen to provide DC coupling from node 7 to the input of operational amplifier 111 while, impedance Z3 provides only AC coupling from the output Vo to the input of operational amplifier 111.

The output voltage regulation is load dependant but now with faster transient response than without impedance Z3.

[0034] With load dependant output voltage regulation several converters can be connected in parallel with inherent current sharing provided through output resistance Rw+Rs/2 or portion of that if resistor 108 is used, as illustrated in FIG. 1. Multiple regulators 10 can operate in parallel to provide more current to the load 15 than one regulator 10 can provide individually. Regulator can either share a common reference voltage VREF (as shown in FIG. 1) or have independent reference voltage (not shown in FIG. 1).

[0035] Many modification and other embodiments of the invention will come to the mind of one skilled in the art having the benefit of the teachings presented in the foregoing descriptions and the associated drawings. Embodiments of the invention herein are shown in a isolated half-bridge converter. The invention may also be applied to other similar converter topologies, e. g. push-pull converters, full-bridge converters, their derivatives and the like.

[0036] Therefore, it is to be understood that the invention is not to be limited to the specific embodiments disclosed, and that modifications and embodiments are intended to be included within the scope of the appended claims.