Power supplies with magnetic amplifier voltage regulation

Abstract

 A power supply including a saturable reactor voltage regulator using a core 124 formed of relatively inexpensive magnetically soft material with a B-H loop which is poor in squareness. The reset point of the saturable reactor core is established by a clamping circuit, including a transistor switch Q₂ and diode 134 in series, which effectively clamps a short circuit across the reactor winding when the core reaches a desired reset point on its B-H loop. The proper time of actuation of the clamping circuit is selected by a control circuit which includes a comparator 138 for generating a control signal for actuating the switch whenever an error voltage V_e derived from the output voltage exceeds the voltage level of a triangular wave developed by integrating a replica of the input pulsations. The saturable reactor is driven to saturation at a time during positive pulsations related to the position of the reset point.
In one embodiment, the clamping circuit is connected directly across the reactor winding, while in a second embodiment, the clamping circuit is connected across a secondary reactor winding 226. A bias winding 228 for the reactor may be provided to shift the B-H loop of the core to the right to increase the range of adjustment.

Description

[0001] This invention relates to power supplies and, more particularly, to power supplies of the switching converter type using magnetic amplifier or saturable reactor voltage regulating means.

[0002] Switching power supplies are frequently used to provide a plurality of separate outputs, and independent control of the outputs is often required. This has usually been done in the past by packaging multiple power stages in one unit and using a separate control loop for each output. A separate switching power stage and transformer are needed for each controlled output. This solution works well but is expensive.

[0003] Another solution, involving the use of a magnetic ampli­fier, which is described in the article "Switch Mode Converter Using High-Frequency Magnetic Amplifier" by Hiramatsu, Harada and Ninomiya appearing in Power Con­version International for March - April 1980 at pages 75 - ­82, allows control on the secondary side of the transformer. Thus, one transformer delivers multiple, independently controlled, outputs with large cost and size savings.

[0004] As will be explained more fully below in connection with Fig. 1, prior art converters of this type use a switch transistor to drive the primary of a transformer which has two output windings. One of the output windings feeds a rectifier and filter to supply a first output voltage which is sensed by a control circuit which adjusts the operating duty cycle of the switch to maintain the output voltage at a preset value. The second output winding is connected to a rectifier and filter through a saturable reactor having a core with a square B-H loop. When the core is not saturated, the reactor exhibits a high impedance and prevents the voltage in the second output winding from reaching the rectifier and filter. The voltage will cause the core to saturate after a period of time determined by the starting point on the B-H loop and the applied voltage. When the core saturates, the reactor switches to a low impedance value permitting the voltage to be applied to the rectifier and filter. Between pulses, a control circuit forces the reactor to reset with a current which is poled in a direction opposite to the direction of the current during the active conduction period. The reset point is adjusted in response to the output voltage to maintain the output voltage at a preset value.

[0005] In the article of Hiramatsu the reset current flows through the load and is superimposed by the persisting current through the filter inductance. This makes control difficult. Furthermore, there are cost problems associated with the construction of the saturable reactor. A square loop material with a high degree of squareness is required. This is usually obtained by using metal tapewound cores of permalloy which are expensive. Square loop ferrites may also be used, but the available ferrite materials are quite lossy, leading to heat problems. To maintain squareness, an ungapped magnetic structure, usually a toroidal core, is required. These are expensive to wind and difficult to mount.

[0006] It has been suggested in US patent No. 2 753 518 that a magnetic amplifier voltage control system use relatively inexpensive substantially zero remanent, moderately low permeability cores with a B-H characteristic having poor squareness. In order to control the direct current power to a load, the reset point of the core is controlled by adjusting a current supplied during a control half-cycle in a control circuit including a variable resistance, a rectifier and a control winding for the saturable reactor.

[0007] It is also known from US patent Nos. 2 054 496, 2 638 571 and 3 182 249 to control the current through a reactor by controlling the application of a short circuit across the reactor or across a winding coupled magnetically with the reactor.

[0008] It is the object of the present invention to address the problem of a costly saturable reactor and provide a power supply using a magnetic amplifier having a core formed of relatively inexpensive magnetically soft material with a B-H characteristic which is poor in squareness and having improved control circuitry for the operation of the magnetic amplifier to provide effective voltage regulation.

[0009] To this end, the present invention contemplates the provision of a power supply in which the reset point of the saturable core is established by providing clamping means, including a transistor switch and diode in series, effectively to clamp a short circuit across the reactor winding when the core reaches a desired reset point on its B-H loop. In oder to select the proper time of actuation of the clamping means, a control circuit includes a comparator for generating a control signal for actuating the switch whenever an error voltage derived from the output voltage of the power supply and a reference voltage exceeds the voltage level of a triangular wave developed by integrating a replica of the input pulsations of the power supply.

[0010] The power supply of the invention includes a source of alternating positive and negative voltage pulse- tions, such as square pulses of the type typically provided by a switching inverter, which are applied through a transformer to a secondary winding. The saturable reactor winding, which is connected between the secondary winding and rectifier and filter means providing the direct current output voltage, is driven to saturation at a time during the positive pul- sations related to the position of the reset point. By controlling the position of the reset point, the control circuit controls the duty cycle of the voltage applied to the rectifier means. Since the operation of the control circuit depends on the magnitude of the error voltage, the circuit operates to maintain the output voltage at a preset value.

[0011] In order to insure that the short circuit is applied only during negative pulsations when the core is being reset, the diode in series with the transistor switch is poled to block current during positive pulsations.

[0012] The control circuit includes an integrator and a baseline clipper diode which prevents the generated triangular wave from becoming negative. The error voltage is generated by an error circuit comprising a differential amplifier, the inputs of which receive the output voltage and a reference voltage.

[0013] The above-described technique for controlling voltage may be used in a switching converter having a plurality of secondary windings each feeding a rectifier-filter. The output of one of the rectifier filters may be controlled as described above, while another output may be used to control a pulse width modulator controlling the duty cycle of the switching transistor driving the transformer primary winding.

[0014] The clamping switch may be connected directly across the reactor winding or may be effectively connected across the reactor winding by being connected across a secondary reactor winding coupled to the reactor winding. A bias winding may also be provided to apply a bias signal for shifting the B-H characteristic of the core to the right to increase the range of adjustment.

[0015] These and other objects, features and advantages of the invention will be more fully appreciated with reference to the accompanying figures, in which:

Fig. 1 is a schematic circuit diagram of a power converter of the prior art using a magnetic amplifier voltage regulator;

Fig. 2 is a hysteresis characteristic of the core member of the magnetic amplifier of the circuit of Fig. 1;

Fig. 3 is a schematic circuit diagram of an embodiment of the power supply circuit of the present invention;

Fig. 4 is a schematic circuit diagram of the error circuit used in the circuit of Fig. 3;

Fig. 5 is a hysteresis characteristic of the core member of the magnetic amplifier used in the circuit of Fig. 3;

Fig. 6 includes a set of voltage curves illustrating the operation of the circuit of Fig. 3;

Fig. 7 is a schematic circuit diagram of a second embodiment of a power supply circuit of the present invention; and

Fig. 8 is a hysteresis characteristic of the core member of the magnetic amplifier used in the circuit of Fig. 7.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0016] A prior art converter circuit is shown in Fig. 1. In this circuit a switching inverter 10 includes a direct current supply 11, which drives a primary winding 12 of power transformer T₁ through a transistor switch Q₁. The transformer has two secondary windings N_s1 and N_s2. The voltage appearing across winding N_s1 is rectified and filtered in the conventional manner. The rectifier 14, includes a series diode 16 and a shunt diode 18; and a low pass filter 20 includes a series inductor L₁ and a shunt capacitor C₁ to remove alternating current ripple components to provide a first output direct current voltage V_o1. A control circuit 21 is a conventional pulse width modulator error circuit providing a control pulse on output 22. The control pulse, the width of which is a function of the difference between the magnitude of output voltage V_o1 and a reference voltage, controls the switching of tran­sistor switch Q₁ and thus adjusts the duty cycle of inverter 10. This adjustment maintains output voltage V_o1 at a preset value, voltage V_o1 being proportional to the duty cycle of inverter 10.

[0017] The second output winding N_s2 of transformer T₁ is connected to an identical rectifier 14' and low pass filter 20' through a saturable reactor L_s1 which comprises a reactor winding 22 and a saturable core 24 formed of a material, such as tape wound permalloy, and an ungapped toroidal structure providing a highly square characteristic. The square B-H hysteresis loop of core 24 is shown in Fig. 2 in which, in the usual manner, B re­presents magnetic flux density and H signifies magnetizing force. At the start of a pulse, the core is reset to a point a in the left hand plane, which consists of the upper left and lower left quadrants, of the characteristic. When transistor Q1 becomes conductive, a positive voltage appears across winding N_s2. At this time saturble reactor L_s1 is not saturated, and reactor winding 22 exhibits a high impedance preventing the voltage across winding N_s2 from being applied to rectifier 14' and filter 20'. The voltage applied to reactor L_s1 will cause core 24 to saturate after a period of time determined by the starting point a and the magnitude of the applied voltage. Core 24 will move along the hysteresis loop from point a to point b and then to point c in the upper right quadrant of the characteristic. The core will then be saturated, and the reactor will be switched to a low impedance value. The voltage across winding N_s2 will now be applied to rectifier 14' and low pass filter 20' supplying direct current output voltage V_o2 to output terminals 26. When transistor switch Q₁ turns off, the positive voltage on secondary winding N_s2 is removed; and core 24 returns to point d on its B-H loop. Between pulses, a reset control circuit 28 senses output voltage V_o2 and generates a control current which is a function of the difference between output voltage V_o2 and a reference voltage. The reset control current, which is opposite the polarity to the current through reactor winding 22 during the period of conduction of transistor switch Q₁, forces reactor core 24 to reset to point a of the characteristic so that the reactor will be ready for the next pulse. The controlled range of adjustment of the flux density is shown as B in Fig. 2. Saturable reactor L_s1 acts to shrink the pulse from transistor Q₁ by an amount controlled by the location of the reset point a to maintain output voltage V_o2 at a preset value.

[0018] Although the prior art circuit of Fig. 1 is effective, it requires a saturable reactor core made of a square loop material having a high degree of squareness. Expensive metal tape-wound cores using permalloy or lossy square loop ferrites may be used, and ungapped toroidal structures which are expensive to wind and difficult to mount are needed.

[0019] An embodiment of the present invention, which avoids the use of expensive high squareness reactor cores, is shown in Fig. 3. As in the circuit of Fig. 1, an inverter 110 includes a direct current supply 111 which drives a primary winding 112 of a transformer T₂ through a transistor switch Q₁. Transistor Q₁ is turned on (become conductive) in response to a signal applied to its base electrode on lead 22 from pulse width control 21. Inverter 110 thus operates as a switching inverter, generating a square wave, the pulse width of which is responsive to the pulse width control circuit 21.

[0020] On the secondary side of transformer T₂, a first output winding N₁ developes a voltage whenever transistor Q₁ is conducting. This voltage is rectified in rectifier 14 which includes a series diode 16 and a shunt diode 18. The rectified voltage is then passed through a low pass filter 20, which includes a series inductor L₁ and a shunt capacitor C₁, to remove the A>C> ripple component and apply a direct current output voltage V_o1 across output terminals 23. The output voltage V_o1 is applied to pulse width control circuit 21 which compares it to a reference voltage to develop a pulse width control signal in a manner known in the art. This pulse width control signal, as explained above, is connected to the base electrode of transistor Q₁ to control the duty cycle of inverter 110 and maintain output voltage V_o1 at a present value.

[0021] A second output winding N₂ of transformed T₂ develops a voltage V₁ in response to current conducted through primary winding 112. A saturable reactor L_s2 include a reactor winding 122 and a saturable reactor core 124. Winding N₂ is connected to a rectifier 114, again comprising a series diode 116 and a shunt diode 118, through reactor winding 122. The rectified voltage is then applied through low pass filter 120, which includes series inductor L₂ and shunt capacitor C₂, to provide a direct current output voltage V_o2 across output terminals 126.

[0022] In accordance with the present invention, a clamping circuit 130 is connected across reactor winding 122 and includes a clamping transistor Q₂ and a diode 134. As will be explained below, clamping transistor Q₂ is actuated to clamp a short circuit across reactor winding 122 at a desired reset point on the B-H hysteresis characteristic of saturable core 124. The control signal applied to the base electrode of transistor Q₂ is obtained from a control circuit 136 and specifically from a comparator 138 is derived from an auxiliary winding N_F of transformer T₂. The voltage appearing across winding N_F includes information on the timing and voltage of the input pulse wave applied through the transformer. Winding N_F typically may be the same winding used for feed-forward compensation (not shown). If no feed forward compensation is provided, the voltage provided by winding N_F might instead be obtained from any other winding, such as windings N₁ or N₂ of the transformer. The voltage from winding N_F is applied to an integer 140 including a series resistor R and a shunt capacitor C. The capacitor is shunted by a diode D₁, which functions as a baseline clipper to keep the signal V₃, which is applied to one input terminal of comparator 138, positive.

[0023] An error circuit 150, shown in detail in Fig. 4, develops an error signal V_e from output voltage V_o2 and applies it to the other input terminal of comparator 138. As seen in Fig. 4, error circuit 150 includes a differential amplifier 42. A reference voltage V_ref is applied to a first input terminal 43 of amplifier 42. Output voltage V_o2 is applied across input terminals 44 and 45, the latter of which is grounded. Terminal 44 connects voltage V_o2 through a series resistor R₁ to a second input terminal 46 of differential amplifier 42. Resistor R₁ is shunter by a resistor R₂ and a capacitor C₃ in series, and a resistor R₃ and capacitor C₄ in series form a feedback circuit for amplifier 42. Impedances R₁, R₂ and C₃ and R₃ and C₄ are frequency shaping and compensation networks. The reference voltage V_ref is preferably selected to be of such magnitude that the error voltage V_e wil always be of positive polarity.

[0024] Because it is not necessary for the saturable core 124 to have a square hysteresis characteristic, it may be formed of a wide variety of low cost, magnetically soft materials and may be formed in physical shapes which have small gaps. Low remanenent core materials, such as Stackpole 24B or Ferroxcube 3C8 )a ferrite material) may be used. Such materials provide a B-H hysteresis characteristic which is poor in squareness as illustrated by the B-H characteristic shown in Fig. 5. Saturable reactor L_s2 is thus much less expensive than saturable reactor L_s1 of the prior art circuit of Fig. 1.

[0025] In the operation of the power supply circuit of Fig. 3, saturable core 124 will be at rest point e of the hysteresis loop when transistor Q₁ switches on. The core then travels the path e-f-g and saturates. At the end of the pulse, transistor Q₁ turns off. The voltage of secondary winding N₂ then reverses during the reset period of transformer T₂. This reverse voltage brings the core from point g back toward the remanent flux density B_r along the upper branch of the loop. When the core reaches point e, transistor Q₂ switches on to clamp a short circuit across reactor winding 122. The current in winding 122 now circulates through transistor Q₂ and core 124 stays at reset point e waiting for the next pulse. The magnetic flux density falls an amount designated as B' in Fig. 5. The available range of adjustment is designated by B_A, the distance between the saturation point g and the remanent flux density B_R. Since reset point e is in the same (upper right) quadrant of the hysteresist characteristic as the saturation point g, the core operates entirely within a single quadrant making it unnecessary to use a forcing current of reverse polarity to reset the core as is required in the prior art circuit of Fig. 1.

[0026] The operation of control circuit 136 will be understood from the voltage curves of Fig. 6. The curve V₁ represents the voltage V₁ from secondary winding N₂ as indicated on Fig. 3. From this curve, it is seen that V₁ has a positive magnitude V_F during the forward conduction period of winding N₂ and a negative magnitude V_R during the recovery period of transformer T₂. The voltage magnitudes V_F and V_R are usually, but not necessarily, equal. V₁ is held at magnitude V_F for a time D₁t_cyc, where D₁ represents the duty cycle of the main output voltage V_o1 and t_cyc represents the period of the switching inverter-regulator - ­that is, the time for one switch cycle of transistor Q₁. Thus, t_cyc is equal to the inverse of the switching frequency f_sw

[0027] Curve V₂ represents the voltage V₂ appearing at the output side of reactor winding 122 and is thus also the input voltage supplied to rectifier 116. Voltage V₂ has a magnitude V'_F for a time D₂t_cyc, where D₂ at a value of the control loop is to maintain duty cycle D₂ at a value which will keep output voltage D_o2 at its desired value. The main control is effected by the delay td₂, the delay between the onset of the positive pulses of voltage waves V₁ and V₂. This delay results from the operation of saturable core 124. At the time of the onset of positive pulse V_F of input voltage wave V₁, the core is at its reset point e and is not saturated. Winding 122 therefore presents a high impedance to the applied voltage blocking the start of the corresponding positive pulse V'_F of voltage V₂ on the output side of the reactor winding. When the core reaches point g on its hysteresis loop, the core saturates and the impedance of reactor winding 122 becomes low permitting the reactor winding to apply the pulse V'_F of voltage wave V₂ to the output side of the reactor.

[0028] Delay td₂ is a direct function of the clamp delay td₁ as shown by the relationship:

[0029] Neglecting diode drops, output voltage V_o2 is given by

[0030] The desired relationship between delay td₁ and V_e is ob­tained by control circuit 136. Resistor R and capacitor C form integrater 140 which provides voltage V₃. If this circuit is treated as an ideal integrator, the slopes of the curve for V₃ will be as seen in Fig. 6. Diode D₁ acts as a baseline clipper to keep the triangular wave signal positive. The generated triangular wave is compared in comparator 138 with error voltage V_e. Whenever the tri­angular wave voltage V₃ is less than the error voltage V_e, comparator 138 provides positive output signal on output lead 139. This output signal is applied to the base electrode of clamping transistor Q₂ causing transistor Q₂ to become conductive. During the forward conduction period V'_F, diode 134 is poled to block conduction through clamping circuit 130. However, when the forward conduction period V'_F ends at a time coinciding with the maximum point M of the triangular wave of voltage V₃, diode 134 no longer blocks conduction through circuit 130. Thus, when the falling triangular wave of voltage V₃ crosses the value of error voltage V_e at point P, transistor Q₂ becomes conductive and clamping curcuit 130 applies a short circuit clamp across reactor winding 122. The short circuit current circulates in the loop formed by inductor winding 122, transistor Q₂ and diode 134; and core 124 is held at reset point e.

[0031] The delay td₁ is governed by the equation:

where N₂ and N_F represent the number of turns of windings N₂ and N_F, respectively, and τ is the constant of integrator 40, being equal to the product of the resistance of resistor R and the capacitance of capacitor C. Substituting the expression for td₁ given in equation (3) in equation (2) and simplifying, we have:

[0032] It is to be noted from equation (4) that output voltage V_o2 is now a function of a single variable, the error voltage V_e; all of the other parameters of equation (4) are fixed.

[0033] As shown in Fig. 6, the rising slope of the triangular wave of voltage V₃ is defined by the expression

and the declining slope by the expression

[0034] In the above analysis, integrator 40 is treated as an ideal integrator. Some error is introduced by the approximate nature of the assumed integrator operation. This error may be held to an acceptable value by keeping τ equal to or greater than t_cyc.

[0035] When the positive pulse V_F of applied voltage wave V₁ ceases, the triangular wave V₃ has reached its apex M and the positive pulse V'_F of voltage wave V₂ also cases. Becuse the fall of positive pulse V_F of applied voltage V₁ brings core 124 back from saturation, reactor L_s2 again presents a high impedance, blocking the negative pulse V'_R of voltage wave V₂. When, however, V₃ falls below error voltage V_e, the clamping transistor Q₂ is again actuated causing transistor Q₂ to become conductive. Diode 134 does not block the reverse pulse V'_R, and the clamp effectively short circuits reactor winding 122. This permits the reverse pulse V'_R to appear at the output side of the reactor.

[0036] As the reverse pulse V_R of voltage wave V₁ is applied, core 124 is driven along its characteristic from saturation point g toward its remanent point B_R. When the core reaches reset point e, triangular wave V₃ crosses the value of error voltage V_e. Comparator 138 provides an actuating signal on the lead 139 to the base electrode of clamping transistor Q₂. Claping transistor Q₂ is therefore actuated, and the short circuit across winding 122 clamps core 124 at reset point e until the next positive pulse V_F is applied as the current in the reactor winding circulates through diode 134 and transistor Q₂.

[0037] The embodimentof Fig. 7 incorporates two modifications of the circuit of Fig. 3. First, the clamping circuit is no longer connected directly across the reactor winding, but is, instead, connected across a secondary winding inductively coupled to the reactor winding. Second, the adjustment range of the circuit is increased by providing biasing means to shift the hysteresis characteristic of the core to the right.

[0038] As seen in Fig. 7, a self-excited, inverter 200, as shown, for example, in the aforementioned Hiramatsu et al article, generates a square wave to drive primary winding 202 of a transformer T₃. It is to be understood, however, that a switching inverter as shown in the prior art circuit of Fig. 1 or the embodiment of Fig. 3 could be used to drive the transformer. A square wave voltage is induced in secondary winding 204 of transformer T₃. A saturable reactor L_s3, which is used to regulate the output voltage V_o3, includes a reactor winding 222, a reactor core 224, a secondary winding 226 and a bias winding 228. Reactor winding 222 connects secondary winding 204 to a rectifier 214 and a low pass filter 220. Rectifier 214 includes series and shunt diodes 216 and 218, and filter 220 includes series inductor L₃ and shunt capacitor C₅. Output voltage V_o3 appears across output terminals 221 on the output side of filter 220. An error circuit 240, which may correspond to the circuit of Fig. 4, develops error voltage V_e and applies it to one input of comparator 238. The other input of comparator 238 is received from an integrator and clipper circuit 250, identical to the integrator 140 and diode clipper D₁ of the embodiment of Fig. 3. An auxiliary winding 206 on transformer T₃ provides a sample of the input voltage from transformer T₃ to integrator 250, but this sample could also be taken from across another winding, such as winding 204, of the transformer.

[0039] Comparator 238 provides an output signal on lead 239 whenever the magnitude of the triangular wave from integrator and clipper 250 is less than the error voltage V_e. This output signal is applied on lead 239 to the base electrode of clamping transistor Q₃ of clamping circuit 230 causing the transistor to become conductive. During the positive pulse in winding 204, diode 234 blocks the clamping circuit from applying a short circuit across a reactor secondary winding 226 inductively coupled to reactor winding 222. Diode 234 is poled to permit conduction through transistor Q₃ on the reverse wave appearing in winding 204; a short circuit is then clamped across secondary winding 226, effectively clamping a short circuit across reactor winding 222 as a current induced from reactor winding 222 circulates in the loop including winding 226, diode 234 and transistor Q₃.

[0040] Core 224 of saturable reactor L_s3 may be identical to the core 124 of the embodiment of Fig. 3. As explained above, the core may be made of magnetically soft material and be formed with small gaps. Such cores are relatively inexpensive and have hysteresis characteristics which are poor in squareness.

[0041] As shown in Fig. 8, the effective B-H loop of core 224 is shifted to the right to increase the available flux swing. This is accomplished through the use of bias winding 228 connected across output voltage V_o3 through inductor L₄ and resistor R₄. Because the reset point e may be adjusted as far as remanent flux density B'_R over an available range of adjustment B'_A which is much larger than the available range of adjustment B_A for the embodiment of Fig. 3 (see Fig. 5), the use of bias winding 228 permits a wider range of voltage control.

[0042] The circuit of Fig. 7 otherwise operates in the same manner as the circuit as Fig. 3. The output voltage V_o3 is regulated by adjusting the position of reset point e of the hysteresis characteristic of core 224 in response to the magnitude of error voltage V_e. Core 224 is reset during the reverse wave when clamping transistor Q₃ becomes conductive. The core is then clamped at its reset point e. The reset point e, in turn determines the duty cycle of voltage V₂ and thus the magnitude of output voltage V_o3.

Claims

1. A regulated power supply circuit with a source of alternate positive and negative voltage pulsatons;
with a saturable reactor having a reactor winding and a saturable core, said reactor winding being coupled between said source and an output terminal;
with rectifier means poled to couple current through said reactor winding during pulsations of one polarity, said core being driven to saturation during said one of said pulsations;
means for resetting said core during the other of said pulsations to hold said core at a reset point on its B-H characteristic, said reset point determining the time of saturation during the next of said one pulsations,
characterized in that said means for resetting comprise clamping means with switch means (Q₂, Fig. 3) connected effectively across said reactor winding to clamp a short circuit across said reactor winding and control means to close said switch means to clamp said core at said reset point.

2. A power supply as recited in claim 1, wherein said switch means is connected across a reactor secondary winding (226, Fig. 7) coupled to said reactor winding.

3. A power supply as recited in claim 1, wherein said core is formed of a magnetically soft material with a B-H characteristic which is poor in squareness.

4. A power supply as recited in claim 1, wherein said core operates entirely within the upper right quadrant of its B-H characteristic.

5. A power supply as recited in claim 1, wherein said clamping means further comprises a diode (134) in series with said switch means (Q₂), said diode being poled to prevent conduction through said switch means during said one pul­sation and to permit conduction during said switch means during said other pulsation.

6. A power supply as recited in claim 1, wherein said clamping means comprises control means, said control means comprising an error circuit (150, 240) deriving an error voltage from said output terminal, an integrator circuit (140, 250) integrating a signal derived from said source (110) to provide a wave related to the timing and voltage of said pulsations, and com­parator means (138, 238) for developing a control signal for actuating said switch means when said error voltage is greater than the voltage of said wave.

7. A power supply as recited in claim 6, wherein said source comprises a transformer (T₂), said signal derived from said source being taken from a winding (N_F) of said transformer.

8. A power supply as recited in claim 6, wherein said wave is a triangular wave.

9. A power supply as recited in claim 6, wherein said integrator circuit further comprises a baseline clipper diode (D₁) to keep the wave positive.

10. A power supply as recited in claim 6, wherein said error circuit comprises a differential amplifier having a first input coupled to said output terminal and a second input connected to a source of reference voltage.

11. A power supply as recited in claim 1, wherein said source comprises a transformer (T₂) having a primary winding (112) and a plurality of secon­dary windings, one of said secondary windings (N₂) being connected to said reactor winding and another secondary winding (N₁) being connected to second rectifier means (16, 18) to provide an output voltage (V01) to a second output ter­minal (23) and wherein said second output ter­minal is coupled to a pulse width modulator (21) generating a control pulse, to a second transistor switch (Q₁), said control pulse controlling the duty cycle of said source (110).

12. A power supply as recited in claim 1, further comprising bias means (Fig. 7) to shift said core B-H characteristic to the right to increase the range of adjustment of said reset point.

13. A power supply as recited in claim 12, wherein said bias means comprises an additional winding (228) coupled to said reactor winding (122), said additional winding being connected across a source of direct current voltage (V_o3).

Drawing