WO2007148354A1

WO2007148354A1 - Control device of a plurality of switching converters

Info

Publication number: WO2007148354A1
Application number: PCT/IT2006/000476
Authority: WO
Inventors: Claudio Adragna; Giuseppe Gattavari
Original assignee: Stmicroelectronics S.R.L.
Priority date: 2006-06-21
Filing date: 2006-06-21
Publication date: 2007-12-27
Also published as: US8604643B2; CN101473506B; EP2033289A1; CN101473506A; US20090152949A1

Abstract

A control device of a plurality of switching converters (Convl.ConvN) is disclosed; each converter comprises at least one power switch and is associated with a control circuit (Mod1 ...ModN) of the at least one power switch. The control device comprises means (100) suitable for comparing a signal (CTRL) representative of the load of the plurality of converters (ConvL.ConvN) with a plurality of reference signals (Vref1 ...Vref(N-l)) and suitable for enabling or disabling at least one of said plurality of control circuits (Mod1...ModN) in response to said comparison.

Description

"Control device of a plurality of switching converters."

*t* *i* *i* ^* *t*

DESCRIPTION

The present invention relates to a control device of a plurality of switching converters.

In the prior art, multiple converters are known; they are made by means of a parallel combination of two or more switching converters in any of the standard types (for example, buck, flyback, boost, etc.), typically the same for all, in such a way that they share the same voltage source and dispense power on the same load. If then in such converters control methods are actuated, that fundamentally consist of staggering in an appropriate manner the PWM pulse trains that control each converter, it is more proper to speak of "multiphase" converters.

Multiple and multiphase approaches are used when with a single converter it is impossible or economically disadvantageous to comply with design specifications. The most obvious situation in which such approaches may be suitable is at a high level of power and/or current. In fact, total power or current could equally be subdivided by a number n of converters, each one of which would be scaled to carry an N-th thereof. In addition to this, in the specific context of the multiphase approach, with an appropriate time control of the PWM pulse trains of the single converters it is possible to bestow on the totality thereof properties that are not detectable individually. For example, it is possible to minimise or even, in certain cases, to zero the ripple current at the input (or at the output) of the totality of converters, thereby optimising stress and thus minimising the bench cost of capacitors affected by optimisation; or, still with a suitable control method, the overall system can be made equivalent to one that works at a frequency that is the same as the sum of the individual frequencies, thereby enabling the dimensions of the magnetic parts to be minimised and dynamic performances to be obtained that are inconceivable with a single converter. Figure 1 shows a two-phase buck converter used to supply the modern processors present in desktop and notebook PCs, characterised by very low supply voltages (less than 1.8V), by high consumption (greater than 90 A) and by very high consumption dynamics (greater than 1 A/ns). Below, specific reference will be made to multiphase converters; nevertheless, it is pointed that all the remarks that will be made remain valid also in the simpler case of a multiple approach.

As already mentioned, a primary requirement that leads to the use of multiphase converters is the high power level. In this case, the maximum benefit from the use of a multiplicity of converters is derived when the system is called upon to work at full load, whereas with reduced loads to have many converters available leads to redundancy. Except for some cases in which redundancy is required in the supply system to ensure very high levels of service continuity, in general this constitutes a waste. Furthermore, at reduced loads, the loss of energy associated with the control (for example the driving of FET transistors), and a series of losses of energy regardless of the load (for example losses associated with loading and unloading stray capacitance of the power elements) begin to become significant and the conversion efficiency of the system (i.e. the ratio between the power returned to the load and the power absorbed by the input source) starts to deteriorate rapidly.

In many systems, which may have non-operational so-called standby conditions, characterised by an extremely reduced load for the converter that supplies them, conformity to voluntary standards or recommendations is requested that aim to regulate the reduction of the energy consumption of such appliances in the aforementioned conditions (e.g. EnergyStar, Energy2000, Blue Angel, etc.). hi this case, the reduction of energy losses mentioned above becomes essential for achieving conformity.

If, sometimes, it is not an easy task to ensure that the consumption of a single converter falls within the recommended limits, it can be easily imagined how this task is further aggravated by the presence of several converters. There is thus the need to extend already known techniques for single converters to multiple or multiphase converters or to complement them with new ones specific to such converters in such a way as to facilitate the task of the system designer.

Various techniques are known for minimising low or zero load consumption for single converters and all involve, substantially, the reduction of the operating frequency of the converter in the above conditions. It is obvious that in a multiphase converter composed by N single converters (namely an N-phase converter), any one of such techniques can be applied to each of the N converters of the totality. Thus if PinO is the input power absorbed by the single converter (for the sake of simplicity considered the same for all) in load conditions, for example zero, the power absorption in such conditions for the N-phase converter will be N-PinO. Although PinO is small, N-PinO could exceed the limits envisaged for the power class to which the N-phase converter belongs if N is large enough.

In view of the disclosed prior art, the object of the present invention is to provide a control device for a plurality of converters that enables the consumption thereof to be reduced. According to the present invention, this object is achieved by means of a control device of a plurality of switching converters, each converter comprising at least one power switch and being associated with a control circuit of said at least one power switch, characterised in that it comprises means suitable for comparing a signal representative of the load of the plurality of converters with a plurality of reference signals and suitable for enabling or disabling at least one of said plurality of control circuits in response to said comparison.

Owing to the present invention it is possible to make a control device for a plurality of switching converters that enables medium to low load conversion efficiency to be optimised, thus reducing the natural rate of reduction of efficiency as the load decreases. Furthermore, said control device enables energy consumption to be minimised at zero load, thus reducing it to that of a single converter and maintaining the time relations between the switching periods of the single converters, thus maintaining the additional properties bestowed on the plurality of converters (e.g. minimising current ripple).

The characteristics and advantages of the present invention will be made more evident from the following detailed description of its embodiment thereof illustrated as non-limiting examples in the enclosing drawings, in which: figure 1 shows a buck two-phase converter according to the prior art; figure 2 shows a block diagram and a corresponding time diagram of a pulse-generating device for a multiphase converter; figure 3 is a block diagram of the control device according to the first embodiment of the present invention; figure 4 is a block diagram of the control device according to a second embodiment of the present invention; figure 5 is a block diagram of the control device according to a third embodiment of the present invention; figure 7 shows a possible embodiment of the block 10 of the circuits in figures 2-5; figures 7a and 7b show other possible embodiments of the block 10 of the circuits in figures 2-5; figure 8 shows a possible embodiment of the device 101 of the circuits in figures 3-5; figure 9 shows diagrams of output power Pout in function of the signal CTRL obtained with the circuit in figure 8; figure 10 shows another possible embodiment of the device 101 of the circuits in figures 3-5; figure 11 shows diagrams of output power Pout in function of the signal CTRL obtained with the circuit in figure 10; figure 12 shows a version of the embodiment of the circuit in figure 8; figure 13 shows a version of the embodiment of the circuit in figure 10; figure 14 shows another version of the circuit in figure 8; figures 15a and 15b show other embodiments of the circuits in figures 8 and 10; figure 16 shows the embodiment of the soft-stop function for a PWM modulator; figure 17 shows a version of the soft-stop function for PWM modulators; figure 18 shows an embodiment of the block 1 of the circuits in figure 3-5; figure 19 shows an embodiment of the block 2 of the circuits in figure 3-5; figure 20 shows another embodiment of the blocks 1 and 2 of the circuits in figure 3-5; figures 21-23 shows possible embodiments of the modulators of the circuit in figure 3; figures 22-26 shows possible embodiments of the modulators of the circuit in figure 5.

With reference to the block diagram and to the corresponding time diagram shown in figure 2, the block diagram comprises the blocks 1, a clock generator, and 2, which is a time division device; said blocks can be made by means of any of the pulse generating devices for a multiphase converter system according to the prior art.

If it is a system of fixed-frequency converters, the block 1 will be an oscillator that may be programmable by the user by means of the external passive components; if it is a system of hysteresis-controlled converters or self-oscillating converters the block 1 will consist of the master-slave designation system between the different converters. In each case the block 1 generates a signal CLK of a period Tsw that represents the time base of the overall system.

The block 2 accepts the signal CLK in input and generates, on the N outputs thereof signals Clkj, (j=l,2,... N) that are staggered in time by a period Tsw/N with respect to one another. If it is supposed that the signal

Clkl is synchronous with the signal CLK, each modulator Modi ...ModN will pilot by means of the signal PWMl... PWMN each converter Convl...CόnvN, more precisely the power switches of the single converter, in such a way that each single converter will operate at the frequency N/Tsw. In some practical embodiments the two blocks 1 and 2 considered may be physically indistinguishable from one another and the separation thereof into two functional blocks is simply conceptual.

The signal CTRL is an indicative signal of the load conditions of the plurality of converters Convl ... ConvN. The signal CTPvL is generated by the negative feedback control loop that regulates the output voltage Vout of the plurality of converters Convl ... ConvN by means of a block 10 and is supplied to all the modulators Modi ... ModN. The structure therefore depends on the control method used for the single converters and on the type of modulation used. Depending on the control, substantially "voltage mode" and "current mode" modulators are used with the variants thereof. In the former case, the input signals Rj (j = 1,...N) to the modulators are voltage ramps that are all the same as one another that are generated by the block 1 and are correlated in time with the signals ClIq . In the latter case the signals Rj are instant to instant voltage ramps that are proportionate to the current that flows in the power switch of the j-th converter. As far as modulation is concerned, there are "trailing-edge" modulators in which the start of the conduction cycle of the power switch is set by the signal CLK (typically pulsed) and the end of the comparison between CTRL and Rj, "leading- edge" modulators, where the signal CLK (typically pulsed) determines the end of the conduction cycle whereas the comparison between CTRL and Rj determines the start of the conduction cycle and "dual edge" modulators in which typically Clkj and Rj are identified in a triangular carrier and the comparison of the latter with CTRL determines both the start and the end of the conduction cycle. Thus at the output of the aforementioned blocks modulated PWM signals are present that through suitable interface circuits will drive the power switches of the single converters.

As CTRL, is common to all of them, the load of the plurality of converters Convl ...ConvN is distributed evenly between the single converters. It can thus be said, in an equivalent manner, that the CTRL level depends on the load level carried by each single converter, which is approximately the same as 1/N of the total.

With reference to figure 3 there is shown a block diagram of the control device according to the present invention. This device is applicable in general to all types, whether they are non-insulated or insulated (i.e. types in which there is an insulating transformer), both with a direct current input and corrected sinusoidal voltage such as PFCs (Power Factor Correctors), special switching converters typically of boost type that absorb from the alternating power supply a sinusoidal current in phase with the voltage. Also, the control device also applies to those N-phase converters the single components of which through the nature thereof, do not operate at a frequency fixed by a system oscillator but at a frequency depending on the operating conditions thereof, i.e. on input voltage and output current (for example, hysteresis-controlled converters and self-oscillating converters), and in which a slave system means that one of the converters ("master" converter) sets the switching frequency for all the others ("slave" converters).

In figure 3 there is shown a device 100 suitable for controlling the PWM modulators 1...N associated with the power switches of the corresponding converters. The device 100 comprises a set of comparators Cl ... Cq, with l≤q≤N-1 possibly with hysteresis, to an input of which the signal CTRL is applied that is indicative of the load conditions of the plurality of converters Convl ...ConvN and to the other inputs the reference voltages Vrefi (i = 1,... q) are applied the values of which are monotonally not in ascending order from 1 to q. Preferably, it is assumed that the outputs of these comparators are at the high logic level if CTRL >Vrefi and at the low logic level otherwise. The device 100 comprises the monitoring device 101 that receive at the input the outputs of the comparators C 1...Cq, and, optionally, additional control variables Vk (k = 1, 2,...p). The device 101, on the basis of the logic status of all the inputs thereof, provides ENj signals (j

= 1,2,...N-I) that are able to enable or disable the operation of N-I of the N PWM modulators. When the j-th modulator is disabled this means that the output thereof is always such as to always keep the power switch off that is controlled by the latter; this means saying that also the j-th converter is disabled. Furthermore, the device 101, on the basis of the number M of signals Enj at the low logic state and therefore the number of disabled modulators/converters acts through the signals line ADJ_Φ on the block 2 in such a way that the Clkj signals of the active outputs N-M are staggered in time by Tsw/(N-M). Optionally, in systems in which the current ripple of the single converter depends on the operating frequency thereof, the device 101 or alternatively the block 2 will act by means of the signals line ADJJT on the block 1 in such a way that the period thereof is Tsw(N-M)/N, thus every single converter will continue to operate at the same frequency and thus at the same current ripple level. Thus the N-phase converter arising from the plurality of converters

Convl ...Convn operates with all the active N phases for as long as the load is maintained above a certain level for which the CTRL signal is such that CTRL > Vrefl (with CTRL > Vrefq). If the load is such that CTRL < Vrefli (h = I₅... q-1) ( with CTRL < Vref(q-(h+l) ) so that h of the n comparators Ci have outputs at the low logic state, M (M = 0,... N-I) of the N converters, designated by the logic of the device 100 are switched off, and the staggering in time between the Clkj signals of the N-M active converters becomes Tsw/(N-M). Optionally, depending on the type of single converters and the operating mode thereof (assumed to be known a priori), the frequency of the CLK signal is increased in such a way that the switching period of the single converter is Tsw(N-M)/N. It should be noted explicitly that in response to the switching off of one or more converters, as each of the remaining ones has to carry a greater quantity of power, the value of the signal CTRL increases, the reason for which the Ci comparators can be provided with an appropriately set hysteresis.

At the moment in which the load is such that CTRL < Vrefq (CTRL < Vrefl) only the N-th converter will be active and to the latter known techniques can be applied for the single converters to minimise low-load or zero consumption. If the load increases in such a way that CTRL > Vrefh (h = 1,... q-1) (CTRL > Vref(q-(h+l)) (h= 1,... q-1)) the Ms (M-I,... N-I) the switched-off converters are progressively enabled again until, returning to full load CTRL > Vrefl (CTRL > Vrefq) all Ns will be operative again.

In figure 4 there is shown a control device according to a second embodiment of the present invention; said device is useful in the case of PWM modulators of "trailing edge" type, i.e. in which the conduction cycles of the power switch are initiated by the signals Cllcj and terminated by the comparison between CTRL and Rj. This differs from the latter substantially through the lack of signals ENj (j = 1,2,... N-I) and through the fact that the disabling function of the PWM modulators and therefore of the associated converters is entrusted to the block 2. The latter in fact disables the j-th modulator simply by no longer sending the corresponding clock signal Clkj, so that the power switch commanded by it is not switched on.

In figure 5 there is shown a control device according to a third embodiment of the present invention; said device in which each PWM modulator communicates with the block 2 with a signal Φj (j = 1,2,... N-I) instead of the line ADJ_Φ coining from the device 101. In this case any action on the block 1 is of necessity actuated through the block 2.

In figure 6 there is shown a possible embodiment of the block 10 of figures 2-5 in which the converters are not insulated. The block 10 comprises a series of two resistances Rl and R2 arranged between the voltage Vout and ground GMD, an error amplifier 300 having a non- inverting input terminal connected to a voltage supply Vref300 and the inverting input terminal connected to the voltage V300=R2*Vout/(Rl+R2) and a feedback network 301 arranged between the output and the inverting input terminal of the amplifier 300.

In figures 7a and 7b there are shown possible embodiments of the block 10 of the figures 2-5 in the case of insulated converters. The voltage Vout is an input on a side 401 of an optocoupler 400 and the other side of the optocoupler that is connected to a resistance R400 connected to a voltage generator V400 (figure 7a) or to a current generator 1400 connected in parallel to the resistance R400 (figure 7b).

Subsequently, there is disclosed a practical embodiment for the event that the N-phase converter is not redundant, i.e. that N-I phases are not sufficient to carry the full load, or in other words, each converter is dimensioned to carry at most 1/N of total power. For the sake of the simplicity όf^3φOMtroTI7^"let^~N-^"37^~q-2^~p-^^~and the signal CTRL that is representative of the load conditions be exactly proportional to the load (CTRL = K-Pout) and, to fix the concepts, let the dynamism thereof be comprised between 0 and 3 V. With comparators with hysteresis, by suitably selecting hysteresis of the comparators, it is in principle possible to use the outputs of said comparators directly as signals ENl and EN2, as shown in figure 8. The relationship between Pout and CTRL obtained with this circuit is illustrated in figure 9 and can be described as follows. Let a start from maximum load condition (nominal power) be supposed, so all containers are switched on. The representative point of this condition is CTRL=3V, Pout=l 00%, the outputs of both comparators are high, so EN1=EN2=1. By decreasing trie load the representative point will move along the line 50 to the point CTRL=Vrefl=1.5V, Pout=50%, where the output of the comparator Cl gets low and thus becomes ENl=O and the converter Convl is switched off, leaving the converters Conv2 and Conv3 active. As a consequence of this, each converter which, before the triggering of the Cl, carried 16.7% nominal power, will now have to carry 25% nominal power, so the output voltage regulating system will move the value CTRL from 1.5V to 2.25V, it is necessary to take this power, as shown by the representative point that passes from line 50 to line 51 through a horizontal portion with constant power.

At the moment in which the output of Cl gets low it becomes Vrefl=2.5V, so that the signal CTRL, although it increases, will not change the output of Cl, which remains low. In order to get Cl to reswitch and to reactivate the converter Convl, CTRL has to exceed the threshold of 2.5V, i.e. the load, equal to 50% of nominal value, has to become greater than 55.6% of nominal value. By decreasing the load further, the representative point moves on the line 51, to the point CTRL=Vref2=lV, Pout=22.2%, where also the output of C2 gets low and therefore in addition to ENl=O, it also becomes EN2=0 and the converter Conv2 is switched off, leaving only the converter Conv3 active. As a consequence of this, the converter Conv3 that before the triggering of the comparator C2 carried the value 11.1% of nominal power, has to now carry this load on its own and to do so the regulating system will take the CTRL value from IV to 2V, as shown by the representative point that goes from line 51 to line 52 through a horizontal portion at constant power.

At the moment at which the output C2 gets low it becomes Vref2=2.5V, so CTRL, although it increases, will not change the output C2, which remains low. In order to reswitch C2 and reactivate the converter Conv2, CTRL has to exceed the 2.5V threshold, i.e. the load that is equal to 22.2% of the nominal value has to become 27.8% greater than nominal value. By decreasing the load further, the representative point will move along the line 52. To the converter Conv3, at the moment in which the load falls below a suitable value, known techniques will be applied to minimise losses of power thereof, depending on type.

By increasing the load, the aforementioned lines 50-52 are travelled along in the opposite direction: as soon as the load reaches 27.8% of the total (CTRL=2.5V), the output C2 is high and EN2=1 and the converter Conv2 will be enabled again; the load for each converter being halved in this way, CTRL will decrease from 2.5V to approximately 1.25V, leaving C2 stably high (the threshold Vref2 is returned to IV when C2 is triggered) and the representative point on line 51 is indicated. By increasing the load still further, the point will rise along the line 51 until, by becoming 55.6% greater than the nominal value (CTRL=2.5V), also the output Cl becomes high again, ENl=I is obtained and also the converter Convl is enabled again. In this way, the load for each converter becomes 18.5% and becomes CTRL=I.67V, leaving Cl stably high (the threshold Vrefl is returned to 1.5V when Cl is triggered) and the representative point is returned to line 50.

With a suitable structure of blocks 1 and 2 the signals ENl and EN2 can be used as lines AD J_Φ and, possibly, ADJ_T.

A suitable choice of hysteresis enables comparators to be used without hysteresis. However, in this case, the logic that controls the signals ENl and EN2 cannot be simply a combination as in the preceding case as it is necessary to keep a record of how many and which converters are switched off to determine the action that the trigger of the comparators has to produce.

In our case, as two converters can be switched on or off, we shall have two "status variables" ENl, EN2, which we shall assume are the outputs of two flip-flops FFl, FF2, of the edge-triggered set-reset type. Let Sl and S2 be the respective Set signals and. Rl, R2 the respective Reset signals. Cl and C2 indicate the logic outputs of the two homonymous comparators, referring respectively to the greater reference voltage Vrefl and to the lesser reference voltage Vref2. On the basis of this choice, it should be observed that if Cl=I, of necessity C2=l, just as if ENl=I it is also EN2=1.

By using a positive logic, the methodology that is the object of the present invention can be translated by the following Boolean expressions: Rl=I if C2=0 AND ENl=I, R2=l if C2=0 AND ENl=O AND EN2=1, S2=l if Cl=I AND ENl=O AND EN2=0, Sl=I if Cl=I AND ENl=O AND EN2=1. Said logic is achieved by means of circuitry 120 in which a port

AND 121 has the signals Cl at the input, Q of the flip-flop FF2 and negated Q of the flip-flop FFl and at the output the signal set Sl of FFl, a port AND 122 having at the input signals C2 and Q of FFl and at the output the signal reset Rl of FFl, a port AND 123 having at the input the signals Cl, negated Q of FFl and negated Q of FF2 and at the output the signal set S2 of FF2 and a port AND 124 having at the input the signals C2, negated Q of FFl and Q of FF2 and providing at the output the reset signal R2 of FF2.

The circuit that achieves the aforementioned Boolean functions is illustrated in figure 10, (for the sake of simplicity, the inputs of C2 are exchanged in such a way as to save an inverter) whilst the relationship between Pout and CTRL obtained with this circuit is illustrated in figure 11 and can be disclosed as follows.

Let it be supposed that the starting point is maximum load condition (nominal power), so all converters are switched on. The representative point of this condition is CTRL=3V, Pout=100% and Cl=I, C2=0 is obtained and

EN1=EN2=1. By decreasing the load, the representative point will move along the line 60; when the load reaches 83.3% of nominal load it becomes Cl=O but this does not cause any variation to the flip-flop outputs, it inhibits only the set signals. By further diminishing the load to the point CTRL=Vref2=lV, Pout=33.3%, it becomes C2=l so that with ENl=I Rl=I and FFl is reset, so that it becomes ENl=O and the converter Convl is switched off, leaving the converters Conv2 and Conv3 active. Consequently, each converter that, before C2 was triggered, carried 11.1% of nominal power, now has to carry 16.7% of nominal power, so that the voltage output regulating system will shift the CTRL value from IV to 1.5V, which is necessary to carry to said power, as shown by the representative point that moves from line 60 to line 61 through a constant-power horizontal portion. Again there is C2=0 but, to reactivate the converter 1, Cl=I, i.e. CTRL has to exceed the threshold by 2.5V, i.e. the load, currently at 33.3% of nominal value, has to become greater than 55.6% of nominal \ alue.

By further decreasing the load, the representative point moves on the line 61, until, yet again, CTRL=Vref2=lV, but with Pout=22.2%. Again C2=l will occur, as ENl=O and EN2=1 becomes R2=l and FF2 is reset in such a way that in addition to ENl=O, also EN2=0 and the converter Conv2 is switched off, leaving only converter Conv3 active. In consequence thereof, the converter Conv3 that, before C2 was triggered carried 11.1% of nominal power, it will now have to carry this load alone, and to do this, the regulating system will take the CTRL value from IV to 2 V, as shown by the representative point that passes from line 61 to line 62 through a horizontal constant power portion. C2=0 is obtained again, but in order to reactivate the converter Conv2 it is necessary for Cl=I, i.e. it is necessary for CTRL to exceed threshold of 2.5V, i.e. it is necessary for the load, currently equal to 22.2% of nominal value, to become 27.8% of nominal value.

By further decreasing the load, the representative point will move along the line 62. To the converter Conv3, at the moment at which the load falls below a suitable value, the known techniques will be applied for minimising the power losses thereof, depending on type.

By increasing the load, aforementioned lines 60-62 are travelled along in the reverse directions: as soon as the load reaches 27.8% of the total (CTRL=2.5V), Cl=I occurs, so that as EN1=EN2=O it becomes S2=l, FF2 is set, it becomes EN2=1 and the converter Conv2 will be enabled again. Consequently, by halving the load for each converter, CTRL will decrease from 2.5V to approximately 1.25V, and thus Cl=O still obtains. By still increasing the load, the point will rise along the blue line until, by becoming greater than 55.6% of the nominal value (CTRD=2.5V), yet again Cl=I is obtained, so that, as ENl=O, EN2=1 it becomes Sl=I, FFl is set, it becomes ENl=I and the converter 1 will be enabled again. In this way the load for each converter becomes 18.5% and becomes CTRL=I.67V and returns the representative point to the line 60. Also in this case, with a suitable structure of the blocks 1 and 2 the same signals ENl and EN2 can be used as lines ADJ_Φ and, possibly, ADJJT.

When one or more ^• converters are switched on or switched off, the signal CTRL does not pass from the initial value to the new equilibrium value dictated by the regulation loop of output voltage according to grade variations, but through transistors that, in most cases are of the damped oscillating type. During these transition phases, therefore, CTRL has undershoots or overshoots with respect to what will be the new equilibrium value, the entity of which is often proportional to the amplitude of the transient, and may thus exceed the amplitude of the hysteresis or thresholds of one or more comparators. This may cause two effects: first, a load variation that is of such an entity as not to trigger comparators and thus cause the converters to switch on/off (in the hypothesis of considering only the initial and final initial and final CTRL values), which may temporarily take CTRL below or above one of the thresholds and cause an undesired switch-on or switch-off; secondly, following a load variation of an entity such as to cause one or more comparators to trigger, exceeding overshoot/undershoot the hysteresis or the thresholds of the comparators, the latter will switch back to the preceding status. Before the aforementioned phenomena may lead to system instability, inasmuch as the comparators continue to switch between one state and the other, causing continuous switching on and off of the individual converters, which is obviously undesirable or hazardous.

In order to eliminate or minimise the possibility of triggering this instability, in the practical embodiment of the circuit, starting from the circuits shown in figures 8 and 10, alternatively or conjunctly it is necessary to add the additional logic that is able to reject the aforementioned overshoot/undershoots and ensure that said undershoot/overshoots are minimised or eliminated completely.

For the circuit in figure 8 logic signals have to be generated that are able to mask temporary CTRL swings beyond certain thresholds. This is the equivalent of introducing delays. With regard to this, it should be noted that even a long delay is tolerable before switching off a converter (efficiency is a question of regular operation) but at switch-on it has to be rather short: following a sudden load increase beyond what the active number of converters can bear, an excessive delay at switch-on could cause a temporary loss of regulation of the output voltage.

A practical example of an embodiment based on the circuit in figure 8 that, by maintaining the functionality thereof illustrated in figure 9, immunises it with respect to the transient oscillations of CTRL is shown in figure 12. hi this circuit, after each switching, each of the comparators is deactivated for a time Tmask fixed by a monostable multivibrator 500 that is sensitive to both edges of the input thereof, in such a way as to freeze the state thereof. The time Tmask, obviously, has to be sufficiently long to mask the undershoots/overshoots of CTRL but not so long as to delay possible restarts following rapid and frequent variations in loads that are such as to temporarily lose regulation of output voltage.

Another example of an embodiment based on the circuit in figure 8 is shown in figure 13 (only for the comparator Cl, on C2 the same circuit would have to be replicated). Therein, the hysteresis of the comparator, after each switching, is increased for a time Tmask in such a way that the undershoot/overshoots of CTRL do not make it reswitch. This is possible by means of two monostable multivibrators MvI and Mv2, the first with an input that is sensitive to negative edges and the other with an input that is sensitive to the positive edges. When Cl goes low to switch off the converter Convl, after switch-off CTRL will suddenly increase with a probable overshoot, so that the threshold of Cl is temporarily moved to the value Vmax (for example by 5 V) by the switch-on of the pull-up Ql controlled by the monostable MvI. If on the other hand Cl goes up to again switch on the converter Convl, after voltage is switched on again CTRL will suddenly decrease with a probable undershoot, so that the threshold of Cl is temporarily taken to zero by the switch-on of the pull-down Q2 controlled by the monostable Mv2.

Also in this case the time Tmask has to be sufficiently long to mask the undershoots/overshoots of CTRL but not too much as to delay possible new switch-ons following rapid and frequent variation of the loads such as to temporarily lose the regulation of the output voltage.

The same concepts used to immunize the circuit in figure 8 can be used for the circuit in figure 10. In figure 14 there is shown an example that, in some way, combines both techniques. Two monostable multivibrators for each line (MlJL and Ml_2 for ENl and M2_l and M2_2 for EN2) are sensitive two-by-two to the negative edges (MlJL and M2_l) and to the positive edges (Ml_2 and M2_2). With the same methods disclosed with regard to the circuit in figure 13, the outputs of the comparators Cl and C2 are masked for a time Tmask after a switch-off of a converter and after a switch-on resp ectively .

The fundamental limitation to the approach considered so far lies in the compromise on the duration of time Tmask, to be set during design of the integrated control device so as to cover the greatest number of possible application situations without, however, having the certainty that with loads having wide and frequent variations there are no malfunctions due to the fact that the monitoring system is in practice inhibited for a certain time after each intervention.

The complete minimisation or elimination of the CTRL undershoots and overshoots during transitions can be performed by means of sequencing of the progressive switch-on and off.

On the basis of the former criterion, following a load decrease that is such that more than a converter has to be switched off, said converters are switched off one by one, inserting a delay between a switch-off and another. In this way, the total transient is divided into several transients of lesser amplitude, so that also the overshoots and undershoots have less amplitude.

In figures 15a and 15b there is shown, by way of example, how this can be achieved in the circuits of figure 8 and figure 10.

With respect to the circuit in figure 8, a port OR has been added and a delay cell Td (Figure 15a); the delay cell Td has in input the signal El and supplies an output signal that is in input to to a port OR 600 with the signal

C2. The signal EN2 will get low when the C2 output is low, provided that ENl has been low for at least a time that is the equivalent of Td. With respect to the circuit in figure 10 a delay cell Td has been added (Figure 15b) having at the input the negated output Q of the FF flip-flop 1 and being suitable for supply the output signal to the port AND that supplies the signal

R2.

The progressive switch-off criterion does not involve the device 100 as considered so far but intervenes on the Modulators PWM. The basic idea is not to suddenly switch off the j-th converter when the signal ENj gets low but to reduce progressively the power carried by the latter from the value that determined the transition to zero of ENj, so as to make gradual the consequent increase of CTRL and to minimise, if not to eliminate completely, the consequent overshoots. A certain gradualness in restarts, in practice the restart known as soft-start, but very brief to prevent the regulation of output voltage from being lost, will contribute to reducing the consequent undershoots.

As the actual switch-off of the j-th converter is delayed with respect to the transition to zero of ENj, the action of adapting the block 2 to the new number of active converters can no longer be performed directly by the device 100 as in the preceding cases but will be performed by the j-th modulator that, by means of the line Φj, communicates complete switch-off. The system in which this operating method is used is the one illustrated in figure 5.

To provide an example of. an embodiment of said function that will be called soft-stop for short, it is appropriate to refer to a preset structure of a

PWM modulator and will choose the structure of the PWM controller that is most widely used commercially, the UC3842. In figure 16 the soft-stop circuit is shown, which is identified by the square dotted together with the PWM modulator of UC3842. The fluctuation constituted by the transistors Ql, Q2 clamps the voltage on the emitter terminal of Q2 at a voltage equal to that present on the capacity C 16, short-circuiting to the ground the current that the circuit connected to the Q2 can dispense. In order to limit this current, the resistance 2R of the divider R-2R has been doubled that is typical of the modulator of the UC3842 in R+R and the first resistance R has been placed at the output of the non-inverting buffer Bl, used not to alter the signal level CTRL that is also supplied to the other modulators.

If ENj is high, the generator 12 is switched off, the capacity C is kept loaded by the generator Il at a voltage near Vbus (Vbus-Vbe, according to a typical embodiment of the generator II), so the transistor Ql is conducting, the emitter of the transistor Q2 clamps Vbus-Vbe. Having selected Vbus in a suitable manner, in these conditions, the voltage on the emitter of Q2 is less than Vbus-Vbe for all the values CTRL can assume, so Q2 is disallowed and the voltage Vcs_ref at the inverting input of the comparator PWM COMPl is linked to the value of CTRL in this manner: V_Csre_f=(CTRL-2*Vbe)/3 if CTRL≤3V+2Vbe and

if CTRL≥3 V+2Vbe.

In these conditions, the transistor Q3 is saturated and the transistor Q4 is disallowed, so that to the port AND a high logic input is applied and thus the clock pulses Clkj coming from the block 1 pass and go to set the PWM latch FF, causing the switch-on of the power switch of the j-th converter. The end of conduction occurs at the moment at which the signal Rj becomes the same as the level Vcs_ref: hi fact, the output of the comparator COMPl becomes high and resets the PWM latch FF. In the specific case, the signal Rj is a voltage that is proportional to the current that traverses the power switch of the j-th converter, so the level Vcsr_ef defines the current that traverses the power switch at the end of the conduction, and thus the power carried by the converter.

If, instead of the block Q3+Q4 with the corresponding polarisation resistances, the comparator COMPl were present, the output of the latter would be high and exactly the same normal operating situation would be obtained.

IfENj is driven low because the voltage CTRL has become less than the reference level that marks switch-off of the j-th converter, the generator 12 is switched on and Cl 6 is discharged with the current 12-11. This difference is deliberately small in such a way as to discharge Cl 6 in the course of some milliseconds. Thus, the voltage on the Q2 base decreases the reserve thereof (maintaining a Vbe below) and as soon as there is a Vbe below the voltage on the emitter Q2 starts conduction and forces voltage on the following emitter to follow within a few millivolts the voltage on Cl 6. Consequently, the level Vcs_ref is not correlated to CTRL and progressively decreases with the same dV/dt as the voltage on C 16, thus progressively diminishing the power carried by the converter. At the moment at which C becomes the same as 2-Vbe, Vcs_ref is cancelled by the effect of the diodes Dl and D2 and therefore ideally the output of COMPl should be high and the flip-flop set reset FF should always remain reset, thereby switching off the converter. Through the effect of the voltage offset of the comparator COMPl, however, the output of the latter might not always be high and the clock pulses Clkj could still set FF and provide very short pulses at the output of the latter. In order to prevent this, as soon as the voltage on the capacitor Cl 6 falls below 2-Vbe and the voltage on the transistor bases Q2 and Q3 becomes less than Vbe, Q3 is disallowed, Q4 switches on and at the input of the port AND there is a low signal that blocks the pulses Clkj . This is also the signal Φj that goes down and acts on the block 2. The discharge of C16 continues until there is typically a voltage Vbe thereupon, so also Ql is at the disallowed limit (also because on the emitter thereof there is a small positive voltage that is due to the base current of Q2).

If instead of the block Q3+Q4 with the corresponding polarisation resistances there were the comparator COMPl the output of the latter would be low as soon as the voltage on C16 becomes less than Vbe+0.5V and therefore the voltage on the base of Q2 becomes less than 0.5V and exactly the same situation would be obtained.

If RNj now had to become high again because the load has increased, the generator 12 would be switched off and the capacitor Cl 6 would be loaded starting from approximately Vbe with the current II, which is considerably greater than 12-11 so the voltage on C16 increases much more rapidly than it does in descent. As soon as the voltage on C16 reaches 2-Vbe, Q3 starts conduction and Q4 is disallowed, so that the pulses Clkj can still set FF; furthermore the voltage Vcs_ref becomes greater than zero and then the converter in fact restarts and becomes able to provide a power value at the moment at which the voltage on Cl 6 becomes such as to disallow Q2.

In the considered example, the capacity Cl 6 is inside the control device and is affected by the value limitations of the integrated capacities. If switch-off times were required that are such as to make necessary capacities of a non-integrable value, Cl 6 should be external. In an N-phase controller N-I capacity and N-I pin dedicated to this function would be required. It is more advantageous to use a sole timing capacity in common between all the converters and a logic that controls an analogue multiplexer 700 that runs the timer on the converter during the switch-off phase, as shown in figure 17. For the sake of a complete exposition, in figures 18 and 19, there are shown respectively an example of an embodiment of the block 1 with adjustable frequency and an example of an embodiment of the block 2.

In figure 18 there is shown an oscillating block, the internal structure of which may be of any prior-art type, in which the timing elements (in the example, the capacitors Ci, C₂ and C₃ which are assumed to be the same as one as one another) can be disconnected by means of the controlled switches in such a way as to modify the oscillation frequency thereof. In the example under consideration, when a sole converter is operational, all the capacitors are connected and the oscillation frequency is f_OSo; if the functioning converters are two, Ci is disconnected by the corresponding switch and oscillation frequency becomes 2-f_osc; when all the converters are operational Ci and C₂ are disconnected by the corresponding switches and oscillation frequency becomes 3-f_osc.

In figure 19 there is shown a loop counter comprising three flip-flops of type D 701-703 in which states Qi, Q₂, Q₃ mask the clock signal CLK generated by the circuit in figure 18. Of these outputs, only one is at logic level one and the position thereof slides forwards by a place at the trailing edge of each clock pulse.

The loop (i.e. the output Q of the third flip-flop) is closed on the input D of the former when all the converters are active, i.e. ENl (Φl) and EN2 (Φ2) are both at logic level one, in such a way as to count for three and thus distribute the pulses CLK in sequence to the various outputs Clkl, Clk2, Clk3. It should be remembered that in these conditions the oscillator is operating at a frequency 3-f_0Sc, so each converter operates at the frequency f_osc- The loop is on the other hand closed on the input D of the second flip- flop when ENl (Φl) and at the logic level zero and EN2 (Φ₂) at the logic level one, whilst the input D of the first flip-flop is forced to zero, so as to count for two and thus distribute the pulses CLK in sequence to the sole outputs Clk2 and Clk3 and disable CM . In these conditions the oscillator is operating at a frequency 2-f_osc, so each converter also operates at the frequency f_0Sc. When only the converter Conv3 is operational, the inputs D of the first two flip-flops are forced to zero and that of the third flip-flop is forced to one in such a way that also the output thereof is one and that accordingly CIk₃ =CLK, which in these conditions is a frequency signal f_osc. Lastly, it should be noted that signals ENj (Φj) are asynchronous with respect to CLK and therefore, to ensure an always coherent transition between a condition and the other of enabling/disabling of the converters and variation to oscillator frequency, it may be advantageous for said signals to be able to cause variation to the closing of the counter loop and timing of the oscillator only upon completion of a count cycle, i.e. only when the output Q₃ of the third flip-flop is high. There will therefore be delays to switch-off and switch-on that at the most may amount to two clock cycles and which are thus negligible. A system composed by the blocks 1 and 2 that operate according to the last remark is illustrated in figure 20. In this, the signals ENl (Φl) and EN2 (Φ2) are taken to the given inputs of two other flip-flops of type D 801-802, called synchronisation flip-flops, that make the datum visible at the output only at the pulse CLK that means that Q₃ goes to logic state one. If Q₃ is always in logic state one because only the converter Conv3 is operational, the clock pulse will provide the edge that activates the synchronisation flip-flops.

A further possibility for switching off converters consists of modifying the circuitry of the modulators, by changing the gain thereof, as visible in figures 21-23 and 24-26 that correspond to the respective circuit diagrams of figures 3 and 5. The modulators of said figures comprise a part of the modulator in figure 16 without the soft-stop function and differ from one another through the increase in the signal in input to at the inverting terminal of the comparator COMPl (figures 21 and 24) or the decrease of the signal in input to the non-inverting terminal of the comparator COMPl (figures 22, 23, 25 and 26). In the former case (figures 21 and 24) a level (N-M)/N of the signal Rj is carried with a circuitry comprising a resistance Ra connected to the signal Rj and to the inverted terminal of the comparator COMPl, another resistance Ra connected to the non-inverting terminal of the comparator COMPl and to a switch Sl connected to ground and piloted by the negated signal Enj-1; the circuitry comprises a buffer Bl having the signal CTRL at the input and having the output connected to a series of a resistance R, two diodes Dl and D2, another resistance R connected to a terminal of a parallel of another resistance R and a IV Zener diode and connected to the non- inverting terminal of the comparator COMPl. The output of the comparator COMPl is the reset signal R of a set reset flip-flop FF the set signal S of which is in figure 21 the output of a port AND having at the input the signal Enj-1 and Clkj whilst in figure 24 it is the signal Clkj; the output signal Q of the flip-flop FF is the signal PWMj. hi the second case (figures 22, 23, 25 and 26) a circuitry is added with switches and resistances placed parallel or serially with respect to the two diodes Dl and D2; the switches are controlled by the signals Enj-1 in such a way as to short-circuit a greater number of resistances as the number of converters that are switched off increases.

In figures 22 and 25 a resistance having the value 2/3 of R and a switch Sl are placed parallel to the series of two resistances R; the switch si is driven by the negated signal Enj-1 and the sole difference between the circuits in figures 22 and 25 is that the signal S of the flip-flop FF is the signal at the output of a port AND at the input the signals Enj-1 and Clkj in the circuit in figure 22 whilst it is only the signal Clkj in the circuit in figure 25. In figures 23 and 26 a resistance having the value 4/3 of R and a switch Sl are placed parallel to a resistance having the value 4*R placed parallel to the IV Zener diode; the switch Sl is piloted by the negated signal Enj-1 and the only difference between the circuits of figures 23 and 26 is that the signal S of the flip-flop FF is the signal at the output of a port AND having the signals Enj-1 and Clkj at the input in the circuit in figure 23 whilst it is only the signal Clkj in the circuit in figure 26.

Claims

1. Control device of a plurality of switching converters (ConvL.ConvN), each converter comprising at least one power switch and being associated with a control circuit (Modl ...ModN) of said at least one power switch, characterised in that it comprises means (100) suitable for comparing a signal (CTRL) representative of the load of the plurality of converters (ConvL.ConvN) with a plurality of reference signals (Vrefl ...Vref(N-l)) and suitable for enabling (ENl ... EN(N-I) or disabling at least one of said plurality of control circuits (Modi ...ModN) in response to said comparison.

2. Device according to claim 1, characterised in that it comprises further means (1, 2) suitable for generating a plurality of signals (CIkL .. CIkN) staggered temporally by a given period of time (Tsw/N), said plurality of signals (CM ...CIkN) being at the input of said plurality of control circuits (Modi ... ModN).

3. Device according to claim 2, characterised in that said means (100) acts on said further means (1, 2) to modify said given period of staggered time (Tsw/N) in function of the number (M) of disabled control circuits.

4. Device according to claim 1, characterised in that said means (100) comprises a plurality of comparators (Cl ...C(N-I), each comparator having at the input said signal (CTRL) representative of the load of the plurality of converters (ConvL.ConvN) and a reference signal of said plurality of reference signals (Vrefl ...Vref(N-l)).

5. Device according to claim 4, characterised in that said comparators (C 1...C(N-I) are comparators with hysteresis, each of said comparators is suitable for disabling a control circuit if said representative signal of the load of the plurality of converters is less than the corresponding reference signal and to enable it if said signal representative of the load of the plurality of converters is the same as or greater than the total value given by the corresponding reference signal and the hysteresis.

6. Device according to claim 4 or 5, characterised in that the reference signals (Vrefl ...Vref(N-l)) of said plurality of reference signals have values that differ between one another.

7. Device according to claim 4, characterised in that it comprises other means (500) suitable for deactivating each comparator for a preset period of time (Tmask) after switching thereof.

8. Device according to claim 5, characterised in that it comprises a circuit block (MvI, Mv2) that is capable of increasing or decreasing by a given value the hysteresis value of each comparator after switching thereof.

9. Device according to claim 4, characterised in that it comprises a plurality of delay elements (Td, 600) arranged at the outputs of the comparators (C 1...C(N-I) and such as to obtain switch-off in sequence of the converters of the plurality of converters.

10. Device according to any preceding claim, characterised in that said means (100) acts directly on said plurality of control circuits for enabling or disabling at least a circuit of said plurality of control circuits.

11. Device according to any preceding claim, characterised in that said means (100) acts on said further means (I₃ 2) to enable or disable at least one of said plurality of control circuits.

12. Device according to claim 10 as appended to claim 2, characterised in that it comprises a connecting line (Φ_j) between said plurality of control circuits (Modi ...ModN) and said further means (I₃ 2).

13. Device according to claim 12, characterised in that it comprises a plurality of circuitries (Q3, Q4), one for each control circuit, and each one of them being capable of reducing progressively to zero the power carried by the corresponding converter once disabling has been received by said means.

14. Device according to claim 1, characterised in that said control circuits (Modi ...ModN) comprise elements suitable for changing the gain of said control circuits in response to signals supplied by said means (100) of said control device.

15. Control apparatus for a plurality of switching converters (ConvL.ConvN), each converter comprising at least one power switch, said apparatus comprising a plurality of control circuits (Modi ...ModN) associated with said plurality of converters, said apparatus comprising a control device as defined in claims 1 to 14.