Charge-pump closely coupled to switching converter

[0002] Cross reference is made to the following patent applications, each assigned to the same assignee, the teachings of which are incorporated herein by reference:

[0003] This invention relates generally to integrated circuits, and more specifically to a charge pump circuit for a switching converter.

[0004] As logic integrated circuits (ICs) have migrated to lower working voltages in the search for lower power consumption and higher operating frequencies, and as overall system sizes have continued to decrease, IC power supply designs with smaller size and higher efficiency are in demand. Switch mode power supplies, or DC to DC converters, as they are sometimes called, are frequently used in IC circuits such as digital signal processors (DSPs) and mixed signal analog circuits, etc., to efficiently convert an input supply voltage to voltage levels appropriate for internal circuitry as well as external circuitry that the IC is coupled to. For instance, a 2.8 volt supply voltage provided to a BiCMOS IC may need to be increased internally to 5.0 volts to operate internal CMOS circuitry. As appliances and circuit designs continue to decrease in size, the need for smaller and more efficient IC switch mode power converters increases.

[0005] The present invention achieves technical advantages as a charge pump adapted for use in a circuit, the charge pump having switches driven by a voltage higher than an input voltage to the circuit, resulting in the ability to use smaller NMOS FETs as switches, saving semiconductor real estate.

[0006] In one embodiment, a circuit is adapted to receive an input voltage, including a means of producing a second voltage higher than the input voltage, the means being integral to the circuit. The circuit includes a plurality of switches coupled together in a charge pump arrangement, including a first switch, where at least the first switch is driven by the second voltage. The first switch may be driven by a first driver, and another of the switches may be a second switch driven by a second driver. The charge pump arrangement may be adapted to provide a bootstrap voltage at a bootstrap voltage terminal and a drive voltage for a switch mode power supply. The higher voltage may be obtained from a voltage at a node between an additional switch and an additional capacitor, or alternatively, the higher voltage may be obtained from the bootstrap voltage.

[0007] In another embodiment, a switch mode power converter powered by a charge pump circuit includes a first, second, third, fourth and fifth switch coupled together in a charge pump circuit. The first switch is connected to an input voltage terminal and is driven by a first driver. The second switch is connected to an input voltage terminal and is driven by a second driver. The charge pump arrangement provides a bootstrap voltage at a boot voltage terminal for a third driver and a drive voltage at a drive voltage terminal to a fourth driver. A first capacitor is coupled between the drive voltage terminal and a ground terminal. The first transistor is driven by the first driver and is coupled to the input voltage terminal and a phase node. The second capacitor is connected at one end to the bootstrap voltage terminal and at the other end to the phase node. A second transistor is coupled to the output of the second driver, the phase node and the ground terminal. At least the first driver is powered by a voltage higher than the input voltage at the input voltage terminal. The higher voltage may be obtained from a voltage at a node between an additional switch and an additional capacitor, or alternatively, the higher voltage may be obtained from the bootstrap voltage.

[0008] Also disclosed is a method of reducing the lateral semiconductor area of a charge pump circuit. The method includes the step of providing a plurality of switches coupled together in a charge pump arrangement, where the charge pump arrangement is adapted to provide a bootstrap voltage at a boot voltage terminal for a switch mode power supply. The plurality of switches includes at least a first switch and a second switch coupled to an input voltage terminal. The first and second switches are driven by a first driver and a second driver, respectively. The method includes the step of powering at least first driver with a voltage higher than input voltage. The higher voltage may be obtained from a voltage at a node between an additional switch and an additional capacitor, or alternatively, the higher voltage may be obtained from the bootstrap voltage.

[0009] Advantages of the present invention include at least one driver of a switch in the charge pump circuit being powered by a voltage higher than the input voltage of the circuit, which allows NMOS FETs to be used for the majority of the switches in the charge pump circuit, resulting in semiconductor real estate savings of one-half to on-third. Another advantage of the present invention is that noise generated by the charge pump is inherently synchronous to the main switching converter, which reduces the likelihood that charge pump noise will interfere with the main converter. A further advantage of the present invention is that the main load of the charge pump is the gate charge required for the main switching converter power FETs, synchronizing the charge pump to the main converter, making the charge pump output approximately track the demand. Also, the second embodiment of the present invention requires no additional components to achieve a bootstrap voltage higher than the input voltage powering the driver of the first switch.

[0010] In the drawings, which form an integral part of the specification and are to be read in conjunction therewith:

[0011] FIG. 1 is a block diagram of a switch mode converter powered from a charge pump;

[0012] FIG. 2 illustrates a block diagram of a schematic of a first embodiment of the present invention with additional switch S₆and capacitor C₃creating a voltage V₃that is used to power drivers X₁and X₂;

[0013] FIG. 3 is a timing diagram for the circuit shown in FIG. 2;

[0014] FIG. 4 is a more detailed view of portions of the circuit shown in FIG. 2, with switch S₆implemented as a diode;

[0015] FIG. 5 illustrates a second embodiment of the present invention with drivers X₁and X₂being powered by the bootstrap voltage V_boot; and

[0016] FIG. 6 is a timing diagram for the circuit shown in FIG. 5.

[0017] Like numerals and symbols are employed in different figures to designate similar components in various views unless otherwise indicated.

[0018] A problem with prior art switching converters with charge pumps is that the charge pump circuitry requires low resistance FETs in order to obtain a high efficiency. Some of the low resistance FETs must be PMOS, which require a large amount of area of silicon material, for example between 75,000 to 100,000 square microns. A charge pump circuit is needed in art having a high efficiency yet can utilize smaller FETs such as NMOS FETs which require one-half to two-thirds the area that PMOS FETs require, saving semiconductor real estate.

[0019] FIG. 1 illustrates a block diagram of a switch mode converter 10 powered from a charge pump 12, of related case filed herewith, entitled “Circuit and Method of Restoring a Boot Capacitor of a Switch Mode Power Converter” by Grant. The charge pump 12 initially runs from an internal oscillator (not shown), until the capacitors C₁and C₂are fully charged. Once both capacitors C₁and C₂are charged, the main switch mode converter is started. The switch mode converter alternately turns transistors MN₁and MN₂on, causing the phase node 20 to swing between 0 V and V_in. The transistors MN₁and MN₂are often field effect transistors (FETs). The phase node 20 is tied to one end of capacitor C₂. The other end of capacitor C₂provides the power supply for the driver 14 for transistor MN₁. When transistor MN₁begins to turn on, the phase node 20 rises, increasing the drive voltage for transistor MN₁. The charge required to turn on transistor MN₁is supplied from capacitor C₂. When the transistor MN₁turns off and transistor MN₁turns on, the phase node 20 is brought down to 0V and capacitor C₂is recharged by switch S₅.

[0020] The charge pump 12 may include a first driver X₁and a second driver X₂that are used to drive switches S₁and S₂, respectively. The drivers X₁and X₂are powered by input voltage Vin. The charge pump 12 is used to increase the supply voltage to a switching converter to allow the converter to drive external transistors, which may be inexpensive FETs with a high voltage threshold V_t. To achieve a charge pump 12 with a high efficiency, low resistance PMOS FETs are typically used in the art for the switches S₁, S₂, S₄, and S₅of the charge pump 12, which may require up to approximately twice the amount of semiconductor real estate that NMOS FETs require. Switch S₃is typically an NMOS FET.

[0021] FIG. 2 illustrates a block diagram of a first embodiment of the present invention, a circuit 30 including a charge pump 32 driving SMPS drivers 14 and 16. The charge pump comprises a first driver X₁and a second driver X₂, the drivers obtaining a clock signal by timing generator 34. Switches S₁and S₂are driven by the first driver X₁and second driver X₂, respectively. A switch S₃is coupled to capacitor C₀and a ground terminal 22. Switch S₄is coupled to C₀and S₁, as well as switch S₅. The signal from switch S₅is the voltage V_bootwhich is used to power driver 14. The signal V_drvat switch S₄is used to power driver 16. Signal V_drvis connected to capacitor C₁, which capacitor C₁is connected to ground terminal 22. Drivers 14 and 16 are controlled by SMPS control 18. Driver 14 is controllably coupled to transistor MN₁, and driver 16 is controllably coupled to transistor MN₂. A capacitor C₂is connected to the source of transistor MN₁at one end and to the node containing signal V_bootat the other end. The node at the source of transistor MN₁defines a phase node 20 at which the signal V_phaseis located. The drain of transistor MN₂is coupled to the phase node 20, and the source of transistor MN₂is connected to ground terminal 22. An inductor L₁is connected at one end to the phase node 20 and at the other end to an output terminal V_out. A load capacitor C_Land a load R_loadcomprising a processor, DSP, or logic IC, for example, are coupled across V_outand the ground terminal 22. The circuit 30 provides an output voltage V_outwhen an input voltage V_inis applied to the circuit.

[0022] In the preferred embodiment of the circuit 30 shown in FIG. 2, an additional switch S₆is connected at one end to the bootstrap capacitor C₂and also to an additional capacitor C₃which is connected to the ground terminal 22 at the other end. The voltage at a node V₃between capacitors C₃and switch S₆is used to power the drivers X₁and X₂of the charge pump 32. When the end of capacitor C₂connected to V_bootis high, some of the charge from node V_bootis transferred through switch S₆to capacitor C₃. The voltage at node V₃is approximately equal to (V_inplus V_drv) and is used to power drivers X₁and X₂for switches S₁and S₂, in the charge pump 32. Because V₃is V_drvhigher than V_in, switches S₁and S₂are preferably NMOS devices, rather than being PMOS as in the prior art. This is advantageous because NMOS devices require the lowest amount of lateral semiconductor area for their required Rds_on. Using an NMOS FET may require 50,000 square microns versus the 75,000 to 100,000 square microns required by a PMOS FET. This results in a semiconductor real estate savings of 25,000 to 50,000 square microns, for each PMOS FET replaced by an NMOS FET. Switch S₃is preferably NMOS and may be driven directly from V_drv. Depending on the minimum level of V_in, it may also be possible to make S₄an NMOS FET driven off V₃because the gate drive of S₄will be V_inabove V_drv. However, if V_inis too low to allow sufficient gate drive for switch S₄, a PMOS FET maybe used. Preferably, switches S₁, S₂, S₃and S₄are NMOS FETs to save the most amount of semiconductor real estate, although any one or several of them may also be PMOS, as the circuit design requires. Switch S₅is preferably a PMOS device.

[0023] Preferably in order to reduce diode losses internal to FETs, switches S₁, S₂, S₃, S₄, S₅, and S₆are MOS transistors. The sequencing of the MOS FETs is very closely tied to the switching of the main converter. At initial power up, the main converter is not running, and the charge pump 32 pumps V_drvand V_bootup to their target voltages. The phase node 20 is initially at 0 V. The target voltage of bootstrap voltage V_bootis approximately equal to V_drv. Also at initial power up, the charge pump 32 does not have V₃available yet, so the charge pump 32 runs at reduced efficiency. However, the main load on the charge pump 32, the main load being the main converter comprising SMPS drivers 14 and 16, is not running, so full efficiency is not yet required. An auxiliary clock is preferably an external oscillator and is represented by the timing generator 34 generates the timing for the charge pump 32. Once voltages V_drvand V_bootare precharged sufficiently to start the main converter, the control of the timing of the charge pump 32 is handed over to the main converter and the main converter is then allowed to start. Once the main converter starts, the phase node 20 begins switching, so voltage V_bootnow switches between V_drvand (V_drv+V_in), and V₃equals approximately (V_drv+V_in). The charge pump 32 is able to run off a higher supply in accordance with the present invention, achieving full efficiency operation.

[0024] FIG. 3 shows a timing diagram for the circuit 30 of FIG. 2 once the system has been running for some time, with the x-axis representing time. At time “a,” switch S₆is open and V_phase(signal 42) at the phase node 20 rises from 0 V to V_in. V_boot(signal 44) is pushed up via the capacitor C₂from approximately V_drvto (V_drv+V_in). At time “b” switch S₆closes (signal 46), allowing capacitor C₃to charge from V_boot, shown at V₃(signal 47). At time “c”, the charge pump 32 reaches the point in its sequence when switch S₁needs to be closed (signal 48). Because V_bootis at approximately (V_drv+V_in), V_bootcan be used to drive an NMOS FET rather than a PMOS FET of the prior art to perform the switching function for switch S₁. In the first embodiment where S₆and capacitor C₃are present, the voltage V₃on capacitor C₃is used to power the drivers X₁and/or X₂that drive the gates of switches S₁and S₂, respectively, which switches are preferably NMOS FETs.

[0025] At time “d”, switch S₁is opened (signal 48) before V_bootfalls again to V_drv. At time “e”, switch S₆is opened (signal 46) before V_bootfalls to V_drv. At time “f”, V_phaseand V_bootfall (signals 42 and 44, respectively). In the preferred embodiment, capacitor C₃and switch S₆are present so the voltage on V₃(at signal 47) is held roughly equal to the peak voltage on V_boot(signal 44), or (V_drv+V_in). At time “g” switch S₂is closed (signal 50) and according to the preferred embodiment the voltage on V₃(signal 47) is used to drive an NMOS FET for switch S₂. Voltage V₃(signal 47) fluctuates from approximately (V_drv+V_in−200 mV) to (V_drv+V_in). At time “h”, S₂is closed again (signal 50) and the cycle is ready to restart.

[0026] FIG. 4 illustrates in more detail portions of the circuit 30 shown in FIG. 2. The gates of switches S₁and S₂, which preferably comprise NMOS FETs, are driven by drivers X₁and X₂respectively. Switch S₃also preferably comprises an NMOS FET. A Schottky diode implemented for switch S₆, although a PMOS FET may also be used for switch S6, as shown in FIG. 2. Capacitor C₃is preferably a capacitor of approximately 0.1 μF for example.

[0027] FIG. 5 shows a second embodiment of the present invention where driver X₁is powered from V_bootrather than V_in. This is advantageous because one less pin is required on the device, compared with the first embodiment, and an external capacitor C₃is not required as in the first embodiment. However, switch S₂does not derive benefit from an increased gate drive and is powered by Vin as in the prior art. In this configuration, S₂cannot be powered by V_boot, because V_bootis not in the high state when switch S₂is required to be on. An additional advantage of the second embodiment is that an additional switch S₆is not required, as in the first embodiment. However, the second embodiment is the less preferred implementation of the present invention because V₃does not exist and switch S₂must be implemented as a less efficient and real estate consuming PMOS FET device.

[0028] FIG. 6 shows a corresponding timing diagram 60 for the second embodiment. The timing of the second embodiment is similar to that of the first embodiment, without S₆and V₃shown. At time “a,” V_phase(signal 42) at the phase node 20 rises from 0V to V_in. V_boot(signal 44) is pushed up via the capacitor C₂from approximately V_drvto (V_drv+V_in). At time “c”, the charge pump 32 reaches the point in its sequence when switch S₁needs to be closed (signal 48). Because V_bootis at approximately (V_drv+V_in), V_bootcan be used to drive an NMOS FET rather than a PMOS FET of the prior art to perform the switching function for switch S₁. The voltage V_booton bootstrap capacitor C₂is used to power the driver X₁that drives the gate of switch S₁, which switch is preferably an NMOS FET. At time “d”, switch S₁is opened (signal 48) before V_bootfalls again to V_drv. At time “f”, V_paseand V_bootfall (signals 42 and 44, respectively). At time “g” switch S₂is closed (signal 50). At time “h” S₂is closed again (signal 50) and the cycle is ready to restart.

[0029] One advantage of the present invention is that at least one driver of a switch in the charge pump circuit is powered by a voltage higher than the input voltage of the circuit. This allows NMOS FETs to be used for the majority of switches S₁, S₂, S₃, S₄and S₅in the charge pump circuit. Because NMOS FETs require less semiconductor area than the PMOS FETS used for switches in the prior art, the present invention saves semiconductor real estate.

[0030] Another advantage of the present invention is that noise generated by the charge pump is inherently synchronous to the main switching converter, which reduces the likelihood that charge pump noise will interfere with the main converter. This is particularly advantageous if the main converter is running fixed frequency to reduce interference, as it is often the case in sampled data system in communication systems, where the charge pump noise lies in the same frequency bands as the main converter noise.

[0031] Another advantage of the present invention is that the main load of the charge pump is the gate charge required for the main switching converter power FETs, which synchronizes the charge pump to the main converter, making the charge pump output approximately track the demand.

[0032] An advantage of the second embodiment of the present invention is that no additional components are required to achieve a voltage, Vboot, higher than voltage V_inpowering the driver X₁of the first switch S₁. This allows the use of a semiconductor real estate saving PMOS FET for switch S₁.

[0033] While this invention has been described with reference to illustrative embodiments, this description is not intended to be construed in a limiting sense. Various modifications and combinations of the illustrative embodiments, as well as other embodiments of the invention, will be apparent to persons skilled in the art upon reference to the description. For example, preferably four of the five switches within the charge pump are implemented as NMOS FET devices in the present invention. However, fewer than four switches may be implemented as NMOS devices, as is advantageous in each particular application. The charge pump circuit of the present invention may be utilized with a power converter in a wide variety of applications, such as digital signal processors, microprocessors, telecommunications applications, mobile devices and systems, laptops and personal computers, mixed signal and analog devices, and any lower power electrical application, in general.