High Voltage Generator Having a Self-Timed Clock Circuit and Charge Pump, and a Method Therefor

This invention relates to a voltage generator for low-power applications in integrated circuitry. It relates more particularly to a circuit for generating, controlling, and maintaining a high voltage, of the same polarity as the power source to the integrate circuit, for low-power applications in the integrated circuit.

Previously, a high voltage for an integrated circuit has been generated by free running oscillators and DC regulators driving a pump. However, these circuits can often draw considerable power. For low-power applications, it is desirable to reduce current and power in the circuit. One method to reduce the power is to use a low-frequency oscillator and to break the DC path in the voltage regulator circuit. However, power is still consumed in the oscillator and regulator circuit when they are turned on.

Pumped circuits have been used to provide a voltage to an integrated circuit substrate, where the substrate is pumped to a voltage opposite in polarity to the power supply voltage VCC. However, pumped circuits have not been particularly useful to generate high voltages of the same polarity as VCC, which high voltage may be used to drive word lines and other circuits. One main problem has been the power dissipation, described in the preceding paragraph. The art has often used bootstrapped circuits to provide high voltages.

An object of the present invention is to use pumped circuits which minimize power consumption for maintaining a high voltage called V.sub.CCP of a DRAM or other memory device by using separate standby and active circuitry.

Another object of the present invention is to provide a circuit having a low current standby circuit which continuously maintains V.sub.CCP of a DRAM and an active circuit which is activated whenever V.sub.CCP drops below a desired value and a word line of a DRAM is accessed.

Another object of the present invention is to provide a low current standby circuit which continuously maintains V.sub.CCP of a DRAM and an active circuit which is activated whenever a word line of a DRAM is accessed.

By using pumped circuits, the circuit allows a quicker response than would generally be available with bootstrapped circuits.

The present invention relates to a high-voltage regulator (or voltage-up converter) circuit for low-power applications in an integrated circuit. Specifically, the circuit uses charge pumps to provide a voltage V.sub.CCP, which has the same polarity as the power supply to the integrated circuit. Such voltage is useful, for example, when driving the access word lines in a DRAM (Dynamic Random Access Memory) or other memory devices. When the word lines of a DRAM cell are accessed, V.sub.CCP tends to drop. Therefore, it is necessary to maintain V.sub.CCP by pumping charge to a node. However, the present invention could find application in any circuit requiring a low power voltage generator for maintaining a high voltage at a particular node.

A high voltage generator according to the present invention does not rely wholly on a free-running oscillator. Preferably, no free-running oscillator is used. In the invented apparatus, at least one charge pump is used in combination with at least one self-timed clock circuit. A demand signal such as a clock enable signal corresponding to high voltage use (such as a row address strobe type of signal) or a pump signal from a voltage regulator preferably sets a latch in a self-timed clock circuit. Setting the latch initiates the generation of a first group of clock signals that are used by the charge pump for pumping charge to a capacitive node which is insulated from the substrate of the integrated circuit. Preferably at the end of the first group of clock signals, the latch is reset but concomitantly an input circuit to the latch is disabled from recognizing a new pump signal. Preferably the resetting of the latch causes the clock circuit to generate a second group of clock signals used in the charge pump to prepare fully for the next demand for pumping. Preferably at the end of the second group of clock signals, a full cycle of clock signals has been completed in a self-timed manner, and the input circuit to the latch is reenabled to recognize a subsequent demand signal. Thus, a self-timed interval having first and second consecutive portions is provided by the clock circuit once it has been initiated or activated.

According to method aspects of the present invention, a high voltage is generated having the same polarity as a power supply voltage for an integrated circuit, for use by circuits formed on or above a substrate of the integrated circuit. One feature of the method included is, in response to a demand signal, initiating the generation of a full cycle of clock signals over a predetermined interval, the full cycle including a plurality of first clock signals followed by a plurality of second clock signals. The full cycle is generated despite any termination in the demand signal. The method uses the first clock signals to cause a charge pump to pump charge having the same polarity as the power supply voltage to a capacitive node insulated from the substrate. The method uses the second clock signals to precharge the charge pump.

In the preferred embodiment, the method uses a self-timed clock circuit to produce the first and second clock pulses in a self-timed interval. Preferably this occurs by setting a logic circuit in the self-timed clock circuit, such as latch, to a first state. The setting of the latch initiates the generation of the first clock signals during a first portion of the self-timed interval. A signal is used in the preferred embodiment to disable an input to the latch so that a PUMP signal or other demand signal to the self-timed clock circuit will be ineffective until the second clock pulses have also been generated. Thus, at substantially the end of the first portion of the self-timed interval, an input to the latch is disabled, and the latch is reset. Preferably the resetting of the latch initiates the generation of the second clock pulses during a second portion of the self-timed interval. A signal at substantially the end of the second portion reenables the latch input so that a continued PUMP or other demand signal, or a new such signal, will initiate the generation of another full set of clock pulses.

The present invention combines separate standby and active pump circuits for maintaining V.sub.CCP under different circumstances. Each of these circuits operates independently of the other, but may operate simultaneously, to pump charge to maintain V.sub.CCP. The standby circuit is used to respond to power up and leakage current to maintain V.sub.CCP . To minimize power consumption, the current in the standby circuit is maintained as low as possible. Specifically, the standby circuit includes a low power standby voltage regulator circuit and low power standby pump which are sufficient to respond to a decreased V.sub.CCP due to a power-up condition or leakage current. The standby regulator monitors V.sub.CCP and enables a standby clock when V.sub.CCP drops below a desired value. The standby clock generates clock signals to operate the standby pump for pumping charge to restore V.sub.CCP. The standby clock is a self-timed clock and, therefore, eliminates the need for a constantly running oscillator. Preferably, the standby clock and standby pump are off when V.sub.CCP is at the desired voltage in order to reduce power consumption. Preferably the standby voltage regulator circuit is continuously operating and drawing current.

Because it is difficult for a circuit to draw low current and still react quickly to restore V.sub.CCP to the desired potential when there is a large current draw which lowers V.sub.CCP, it is beneficial to include a separate active circuit having an active regulator, one or more active clock circuits and associated active charge pumps. The active charge pumps may be responsive to the voltage V.sub.CCP as well as a signal RASBP which indicates when a circuit or device associated with the present invention is drawing current from V.sub.CCP. Because the active regulator circuit includes components which draw more current and the active charge pump is larger and able to pump V.sub.CCP to a desired potential faster, V.sub.CCP can be maintained at the desired potential when a large amount of current is drawn by a circuit associated with the present invention, such as when the word lines are driven and more current is drawn to lower V.sub.CCP.

Additionally, multiple active pumps can be used depending upon the amount of current drawn. For example, two active pumps can be used for a 16 MEG generation DRAM having the option to refresh with 2K cycles (versus 4K cycles). Since 2K cycles refresh twice as many word lines, twice as much current is drawn to lower V.sub.CCP in this type of DRAM. Therefore, depending upon the size of pump employed, it may be beneficial to use two or more charge pumps. Although the multiple active pumps can be driven by single active clock, each active pump preferrably is driven by a separate active clock.

Preferably these active clocks are self-timed clock circuits.

Finally, in an alternate embodiment, the active regulator can be removed and the active pumps can be activated whenever a certain condition occurs in a circuit coupled to the voltage generator of the present invention. For example, charge pumps can be activated in a DRAM whenever RASBP signal goes low, signaling that the word lines of the DRAM are accessed. Because V.sub.CCP is often lowered when the word lines are accessed, it is possible to design the pump and related circuitry to react quickly enough to restore the charge that is lost within the time that RASBP is low.

In describing the preferred embodiment, reference is made to accompanying figures wherein:

FIG. 1 is a block diagram of an embodiment of the voltage generator according to the present invention.

FIG. 2 is a block diagram of the voltage regulator circuit 100 including standby regulator circuit 110 and active regulator circuit 200 shown in FIG. 1.

FIG. 3 is a detailed schematic of standby regulator circuit 110 (shown in FIG. 2) for the standby circuit.

FIG. 4 is a detailed schematic of the standby master clock circuit of the standby clock circuit for generating a standby master clock signal.

FIG. 5 is a block diagram of the clock circuit for both the standby clock and the active clock for generating clock signals for the standby pump and the active pump, respectively.

FIG. 6 is a schematic diagram of the pump circuit used for the standby pump and the active pump.

FIG. 7 is a timing diagram for the input and output signals and voltages at various nodes of the standby pump circuit.

FIG. 8 is a detailed schematic of the active regulator circuit 200 shown in FIG. 2 for the active circuit.

FIG. 9 is a detailed schematic of the active master clock circuit of the active clock circuit for generating a master clock signal.

FIG. 10 is a timing diagram for the input and output signals and voltages at various nodes of the active pump circuit.

FIG. 11 is a block diagram of a second embodiment of the voltage regulator circuit having no active regulator.

FIG. 12 is a detailed schematic of the master clock circuit for the active pump for the second embodiment.

First Embodiment

FIG. 1 is a block diagram of a voltage generator circuit 10 according to a first embodiment of the present invention. The voltage generator circuit of the present invention is used for generating, controlling, and maintaining a high voltage V.sub.CCP for low power applications in an integrated circuit, such as a DRAM or other memory device. Although the present invention is described in reference to a DRAM by way of example, this circuit could be employed in other circuits where the voltage at a node may drop and must be maintained at a certain voltage.

Circuit 10 produces its output voltage V.sub.CCP at a capacitive output node 12. Inputs 14 and 16 receive signals RASPB and 2KREFPAD, respectively. These are described infra. The voltage at output node 12 is used to drive word lines in an array of memory cells, for example, as represented by circuits 18 and 20 controlled by switches (e.g., decoders) S1 and S2 in FIG. 1. In a typical DRAM, the voltage V.sub.CCP will be low at initial power-up, as well as through leakage current. This voltage drop at V.sub.CCP is shown by a constant current source 22 in the block diagram of FIG. 1 representing the continuous current leakage. A voltage regulator circuit 100, including standby regulator 110 and active regulator 200, monitors the voltage V.sub.CCP to determine whether V.sub.CCP has dropped too low. When V.sub.CCP has dropped below a desired voltage, the standby regulator activates a standby clock 300 which outputs clock signals to drive a standby pump 478. Because standby regulator 110 is constantly running, it is designed to generally draw low current to minimize power consumption in the circuit. The standby circuits react quickly enough and pump enough current to maintain V.sub.CCP at a desired voltage as a result of leakage current 22.

However, a circuit may experience greater current loss other than merely leakage current and therefore may require an active pump. For example, when the word lines of a DRAM are driven, current is often drawn from node 12, as by operation of circuits 18 or 20, and this lowers the voltage V.sub.CCP. Switches S1 and S2 are closed when RASBP goes low to show the current typically drawn in a DRAM when word lines are accessed or driven. Note that the constant current sources shown in the circuit are merely representative of current loss resulting from leakage or accessing the word lines.

A low RASBP enables active regulator circuit 200 to monitor V.sub.CCP. If V.sub.CCP is low when the word lines are accessed, then the active regulator enables the active clock 500 to generate clock signals for active pump 478'. Generally, the active regulator 200 responds more quickly than the standby regulator 110 for pumping charge to maintain V.sub.CCP when word lines are accessed. Because active regulator 200 draws more current and consumes more power, preferably it is activated only if V.sub.CCP falls or has fallen below a desired value when the word lines are accessed. In addition, active pump 478' is a larger pump (i.e. able to pump more current than the standby pump) to pump charge to maintain V.sub.CCP more quickly when the word lines are accessed. Further, the active clock circuit 500 is a self-timed circuit and, therefore, eliminates the need for a free running oscillator. The self-timed feature will be discussed in more detail infra.

The 2KREFPAD input 16 is included to activate optionally a second active pump 478'. The second active pump and associated clock circuitry is employed when a 2K refresh cycle for the DRAM is used. When a 2K refresh cycle is used, twice as many word lines are accessed, drawing more current to lower V.sub.CCP. While any number of active pumps could be used, these pumps (and related circuits) are preferably identical. Similarly, a single active self-timed clock 500 could drive multiple active pumps, or preferably, a separate active clock could drive each pump. Therefore, the description of active circuitry infra will refer to only one active pump and one active clock circuit.

Although both the active and standby circuits work in conjunction and may or may not operate at the same time, they will be described separately for ease of understanding.

The Regulator Circuit

Turning now to FIG. 2, a block diagram for regulator circuit 100, which monitors V.sub.CCP for both the standby circuits and active circuits, is shown. Regulator 100 includes a first resistor divider network 112, having resistors 113, 114 and 115 and an output node N1. Network 112 generates a reference signal V.sub.CCPREF, which is proportional to the voltage V.sub.CCP, In addition, a second resistor divider network 116, including resistors 117 and 118 and an output node N2, generates a reference voltage V.sub.CCREF, which is proportional to a power supply voltage V.sub.CC.

The voltage divider outputs V.sub.CCREF and V.sub.CCPREF are applied to positive and negative (non-inverting and inverting) inputs respectively of a differential amplifier 120 of the standby voltage regulator 110. When V.sub.CCREF exceeds V.sub.CCPREF, the output of the differential amplifier 120 is high, outputting a high PUMP signal after being passed through a series of inverters 140, 150, a latch 170 and an inverter 190. The standby circuit is designed to limit power in the circuit, and to this end, inverters 140, 150, latch 170 and inverter 190 are included and do limit current in the circuit. The details of the inverters and latch will be described in reference to FIG. 3, infra.

Active regulator circuit 200 is also shown in FIG. 2 to contrast how standby regulator circuit 110 and active regulator circuit 200 compare the reference voltages at nodes N1 and N2 to output signals to enable the standby and active clock circuits, respectively, which drive the standby and active pump circuits, In addition to receiving V.sub.CCREF and V.sub.CCPREF, active regulator circuit 200 receives an ENREGB signal and is enabled only under certain circumstances, Active regulator circuit 200 employs a differential amplifier 220 which also receives reference voltages V.sub.CCREF and V.sub.CCREF from resistor divider networks 112 and 116.

The active regulator circuit also includes an inverter 202 and an enabling transistor 210, Enabling transistor 210 is responsive to the ENREGB signal and enables or disables the differential amplifier 220 depending on whether word lines are being accessed. An enabling transistor is not used with differential amplifier 120 of the standby circuit which outputs a signal depending only upon the voltages at nodes N1 and N2, A switch 240 is also coupled to an inverter 202, Switch 240 controls whether the output of differential amplifier 220 is passed to a Schmitt trigger inverter 260 and an inverter 270 to output a PUMPA signal, The active regulator circuit 200 will be described in detail in reference to the active circuit.

Standby Circuits

Turning now to FIG. 3, having shown the general configuration of the voltage regulator circuit 100 in FIG. 2, additional features of the standby regulator circuit 110 which reduce current loss and power consumption are shown in the detailed schematic diagram. The preferred dimensions of the transistors in FIG. 3 (and the remaining figures) can be found in Appendix A, although other dimensions could be used within the scope of the invention. Also, the preferred devices shown in the schematic diagrams are those well known in the art (e.g. for example, transistor 114 of FIG. 3 is a P channel enhancement type transistor with the substrate coupled to the source), although other devices could be used. For example, all transistors are preferably enhancement type transistors, although depletion type transistors could be used.

In order to reduce power consumption, the standby voltage regulator circuit 110 of FIG. 3 is designed to generally operate on low current and limit current loss in the circuit. In the preferred standby regulator circuit, the resistor networks 112 and 116 include transistors configured as resistors (i.e., transistors each having the gate tied to the source or drain). Because the current through the transistors is a function of the ratio of the channel width/length ratio, the channel lengths of these enhancement transistors are generally long to limit current through the transistor networks.

However, by choosing long transistors in the resistor divider networks, the voltages at nodes N1 and N2 respectively will change slowly. Accordingly, capacitors 111 and 119 can be added at nodes N1 and N2 respectively to cause the voltage at the nodes to change more rapidly. Preferably, capacitor 111a is 2 pf and capacitor 111b is 1 pf. Capacitors 119 are preferably 1 pf. Although the transistors are preferably formed using metal and oxide layers on the device, transistors configured as capacitors could be used.

The differential amplifier circuit 120 preferably consists of a conventional configuration of transistors 122-128 and is used to compare V.sub.CCREF at a node N2 with V.sub.CCPREF at a node N1 to output a PUMP signal. The preferred dimensions of the transistors 122-128 (shown in Appendix A) of the differential amplifier circuit 120 are specifically chosen to keep current in the circuit low, and therefore minimize power consumption. Specifically, the width/length ratio of the transistors of the differential amplifier circuit 120 is preferably kept small by selecting long channel lengths to limit the current flow.

Although the long channel transistors of the differential amplifier circuit limit the current flow, the circuit will generally react slowly and will output a slow changing signal at node N3 (i.e. a signal which does not have a sharp transition between low and high states). A slow changing signal can often lead to current loss in a circuit, particularly when input to a two transistor inverter. Specifically, both transistors of the inverter will be on for a certain period of time, leading to "through current" or "crowbar current" through the series of transistors. Therefore, there is often current loss from V.sub.CC to ground by way of a two transistor inverter during the transition.

To compensate for the slow changing signal at node N3, transistors 130-136 are added to differential amplifier circuit 120. Generally, the sources of transistors 130 and 132 are coupled together at V.sub.CC, while the drains of transistors 130 and 132 are also coupled together to transistor 122. The gate electrode of transistors 130 and 132 are coupled together and are coupled to the gates of transistors 122, 124 and 134. Similarly, the sources of transistors 134 and 136 are coupled together at V.sub.CC, while the drains of transistor 134 and 136 are coupled together at transistor 124. The gate of transistor 136 is coupled to the output of the differential amplifier 120.

The operation of the conventional configuration of transistors 122-128 of differential amplifier circuit 120 is well known in the art. Transistors 130-136 are added to drive the output more quickly without drawing additional current. Particularly, these transistors drive the output V.sub.OUT at node N3 of the differential amplifier circuit 120 when output V.sub.OUT is at a transition point (i.e. a point between outputting a high and low signal).

For example, as V.sub.CCPREF at node N1 reaches a higher potential, transistor 126 begins to turn on harder, pulling node N3 low. Since node N3 is coupled to the gate of transistor 136, which is a p channel FET, it begins to turn on harder. Its source-drain path couples VCC to the source of transistor 124, which now reaches a higher potential and, therefore, turns on transistor 124 harder. This raises the potential at the drain of FET 124 and the gate electrodes of transistors 122, 130, and 132. This permits the output at node N3 to be pulled low even more quickly.

Conversely, circuit 120 acts to drive the output at node N3 high quickly as V.sub.CCPREF (node N1) begins to drop below V.sub.CCREF (node N2). Specifically, as V.sub.CCPREF begins to drop, FET 126 becomes less conductive, permitting the voltage at node N3 to begin to rise. When this happens, p channel transistor 136 becomes less conductive, which lowers the voltage at the source of transistor 124. As a result, the voltage at the drain of transistor 124 is lowered, turning on transistor 122, 130, and 132 harder and driving the output at node N3 high more quickly.

While transistor 130-136 are added to drive the output V.sub.OUT of the differential amplifier 120 more quickly, they are also designed to limit the capacitance at output node N3. It is beneficial to limit the capacitances at node N3 because it generally takes a longer time and more charge to drive a node coupled to a large capacitive load. Specifically, because the capacitance of a transistor is directly dependent upon the channel length, a long channel length will create a large capacitive load at the node coupled to the gate of the transistor (according to the equation: C=c.multidot.l.multidot.w, where c is the capacitance of the gate oxide per unit area, l is the length of the channel and w is the width of the channel). Therefore, transistor 136, which is coupled to node N3, is generally chosen to have a short channel length and width to limit the capacitance at the output.

To further compensate for the slow changing output V.sub.OUT of the differential amplifier 120 at node N3 (i.e., the slow changing output being a result of the differential amplifier which is designed to minimize the current in the circuit), inverters 140 and 160 and latch 170 are specifically designed to reduce the "through current" in the inverters and to output a PUMP signal (i.e. clock enabling signal) having a sharp transition. The switching time of transistors is generally dependent on the channel width/length ratio. Therefore, the channel widths and lengths can be chosen to establish when the transistors of the inverters will switch. Because the widths of the transistors in inverter circuit 140 are the same, the switching time will be a function of their channel lengths. The channel lengths of the transistors 144, 146, 156 and 158 of circuit 140 are chosen so that the voltages at nodes N4 and N5 (and therefore transistors 162 and 164 of circuit 160) switch at an appropriate time.

Specifically, when the voltage at node N3 transitions from a high signal to a low signal to change the voltage at node N6 from high to low, it is beneficial to turn off transistor 162 before turning on transistor 164 to limit through current from V.sub.CC to ground in inverter 160. To accomplish the switching of the transistors of inverter 160 in this order, the width/length ratios of the transistors of inverter 140 are selectively chosen. Specifically, the width/length ratio of transistor 144 is generally greater than that of transistor 156. As a result, transistor 144 turns on before transistor 156 turns on. Therefore, node N4 is pulled high before node N5 is pulled high to ensure that transistor 162 switches before transistor 164.

In contrast, when V.sub.out transitions from a low signal to a high signal, it is beneficial to turn off transistor 164 before turning on transistor 162 to limit through current from V.sub.CC to ground. Therefore, the width/length of transistors 146 and 158 are selectively chosen. Specifically, the width/length ratio of transistor 158 is generally greater than that of transistor 146. As a result, transistor 158 turns on before transistor 146 turns on. Therefore, node N5 is pulled low before node N4 is pulled low to ensure that transistor 164 turns off before transistor 162 turns on. Accordingly, the selection of channel lengths of the transistors of inverter circuit 140 reduces any current path from V.sub.CC to ground in inverter 160 during either transition.

To further limit through current in the voltage regulator circuit, transistors 148 and 154 are added to inverter 140 to eliminate any through current in inverters 143 and 153. Because of the "current-limiting" transistors 148 and 154 have long channel lengths, nodes N4 and N5 will not be pulled rapidly to ground and V.sub.CC respectively. Therefore, a path from V.sub.CC to will exist while driving transistors 162 and 164, but the current will be reduced in the series of transistors 144-148 and 154-158 by the long channel lengths of transistors 148 and 154.

Also, transistors 148 and 154 are selectively positioned to ensure that the timing of the switching of the transistors as described above is not affected. When node N3 transitions from high to low, it is preferable that the inverter 143 drive N4 high very quickly by turning on transistor 144 to shut off transistor 162. Therefore, "current-limiting" transistor 148 is added only to the n-channel side of the inverter 143 to limit the effect of transistor 148 on transistor 144. Specifically, because it is beneficial for transistor 144 to switch quickly and drive node N4 high quickly, no current limiting transistor is added to the p-channel side of inverter 143 (between transistor 144 and V.sub.CC) . While "current-limiting" transistor 148 does effect the timing of driving node N4 low when node N3 transitions from low to high, the capacitive coupling of transistor 148 to node N4 is large enough to pull node N4 low enough to turn on transistor 162 at an appropriate time.

Similarly, when node N3 transitions from low to high, it is preferred that inverter 153 drive node N5 low very quickly by turning on transistor 158 to shut off transistor 164. Hence, the "current-limiting" transistor 154 is added only to the p-channel side of inverter 153 to limit the effect of transistor 154 on transistor 158. A current limiting transistor is not added in series between transistor 158 and ground in order to avoid limiting the rate at which transistor 158 pulls N5 to ground. Although "current-limiting" transistor 154 does affect the timing of driving node N5 high when node N3 transitions from high to low, the capacitive coupling of transistor 154 to node N5 is large enough to pull node N5 high enough to turn on transistor 164 at an appropriate time.

It should be understood that while the preferred embodiment employs a specific differential amplifier circuit 120 shown in FIG. 3, other circuits which compare two signals to provide an output could be used within the scope of the present invention.

Finally, the standby regulator circuit 110 includes a latch 170 to maintain the output of the PUMP signal during the period when neither transistor 162 nor transistor 164 of inverter 160 is on (as discussed above to prevent through current). The inverted PUMP signal is fed back via line 185 to transistor network 174-180 to maintain the original voltage at node N6 until the state of N6 changes when invertor 160 drives node N6.

In order to reduce the power consumption of the oscillator, latch 170 itself is designed to limit the through current while also providing a weak latch. Transistors 162 and 164 are generally small transistors (i.e., small widths and lengths) to limit the current required from inverter 140 to switch inverter 160. However, because of their size, they generally cannot drive a large capacitive load at node N6. Too large a capacitance at node N6 is undesirable because this capacitance will cause the voltage at the node to switch slowly. Therefore, a weak latch is provided at node N6 by selecting transistors 174 and 176 to have smaller widths and lengths, preferably on the order of 1-2 .mu.m, to limit the capacitance.

However, because transistors 174 and 176 of the latch have widths and lengths which are about equal, they also have a large width/length ratio and drive greater current. Therefore, "current-limiting" transistors 178 and 180 are included. These transistors have long channel lengths (and therefore small width/length ratios) to limit the current in the latch 170 and reduce power consumption in the circuit.

Clock Circuits

Turning now to FIG. 4, having shown the detailed schematic of the standby regulator circuit 110, the description of the remaining circuitry of the standby circuit (i.e. the standby clock circuits and the standby pump circuit) will be in described in reference to the various cycles of the pump. The master clock circuit 301 for the standby clock 300 is shown in FIG. 4. Circuit 301 receives the PUMP signal at an input and produces a master clock output signal MCLKS. As will become apparent in reference to the charge pump circuit (shown in FIG. 9), the pump circuit pumps on a two-cycle principle: a precharge cycle and a pumping cycle. Therefore, the clock circuit must generate appropriate signals during these cycles. The clock signals generated during the pumping cycle will be described first, followed by the clock signals generated during the precharge cycle.

The operation of the standby master clock circuit 301 of FIG. 4 which generates the master clock signal MCLKS for the standby circuit will be described first, followed by the operation of a clock circuit 445 shown in FIG. 5. As described in reference to the block diagram of FIG. 2, when the constantly running standby regulator 110 (FIG. 3) detects that V.sub.CCPREF has fallen below a selected value, the standby regulator will output a high PUMP signal. The PUMP signal is a demand signal applied to the FIG. 4 circuit and is shown as an input thereto (center of FIG. 4). At the beginning of the precharge pump cycle, a high PUMP signal drives the gate of a transistor 312 high. Initially, as described infra, FET 314 is on, so that when FET 312 turns on, it pulls the gates of transistors 304 and 306 low. Transistors 304 and 306 form an inverter having an output node "START," which thus is pulled high. The output MCLKS of master clock circuit 301 is high after START is transmitted through a sequence of inverters 316 and 322. Although preferably a high pump signal and MCLKS signal are generated by the circuits to drive the charge pump, the circuits could be configured to generate active low pump and MCLKS signals within the scope of the present invention.

Because the MCLKS signal is used to enable a clock circuit 445 (shown in FIG. 5) which generates clock signals to drive the standby pump (shown in FIG. 6) during different cycles, MCLKS varies during the different cycles. Therefore, a feedback operation of standby master clock circuit 301 is provided, and circuit 301 is designed to change the MCLKS signal (i.e., between a binary low and binary high signal). This aspect of the circuit will be described first, followed by a description of the clock circuit 445 which receives the MCLKS signal.

While the pumping cycle is completed when MCLKS is high, MCLKS is fed back through circuit 301 of FIG. 4 to generate a low MCLKS during the precharge cycle to reset the inputs and voltages at the nodes of the pump circuit for the next pumping cycle. After passing through inverters 328 and 334, the MCLKS signal follows two feedback paths. Two paths are designed to establish the proper voltages at node START for the precharge and pumping cycles. Depending on the state of MCLKS, the MCLKS signal generally will be transmitted through the circuit quickly in one of the paths, while a delay will be provided in the other path. Generally, a first feedback path exists through an inverter 342, a delay circuit 350, and gates 370, 380 and 390. A second feedback path is provided by a delay circuit 400 (having inputs at the gate electrodes of transistors 404 and 412) and inverter 430 and 440. Because the signal from the faster path (i.e. either the first or second path, depending upon the state of MCLKS) reaches a latch 302 (next to the pump input) first and, therefore, establishes a condition of the latch, the fast path will be discussed first, followed by the slow path.

When MCLKS is high (for example, during the pumping cycle), the signal is passed through inverters 328 and 334 and a high signal is input to inverter 342. The output of inverter 342 is low, turning on a transistor 354 and turning off a transistor 352. Because there is no path from a node N1 to ground through transistors 352 and 360, capacitor 364 is pulled to V.sub.CC via the source-drain path of FET 354. As a result, capacitor 364 charges quickly, providing the fast path through the circuit to latch 302. Specifically, the voltage at node N1 is transmitted through Schmitt Trigger inverter 370 and inverters 380 and 390 to output a low signal at the gate of a transistor 314, turning it off. With transistor 314 off, the circuit ignores the PUMP signal at the gate of transistor 312, even if the PUMP signal is high. It will be seen that when PUMP went high, with transistor 314 on, the input at the left side of latch went low, drove START high, and was inverted again and supplied back to the input at the gates of transistors 304 and 306. The latch will eventually be reset into the other state when a transistor 450 becomes conductive, pulling START low. START cannot go high again until transistor 304 turns on, which cannot occur until PUMP is high and transistor 314 is on. From this, it will be appreciated that turning off transistor 314 at this time allows further actions to occur before a new START signal is generated. This is part of the self-timed feature of the circuit. Accordingly, an oscillator is not needed to provide timing for the clock circuit.

With regard to the slow path, the output of inverter 334 is input to delay circuit 400 (which provides the slow path of the circuit in this case). Because the output of inverter 334 is high, transistor 404 is off and transistor 412 is on. Therefore, a path from a capacitor 402 through transistor 406 and resistor 410 is provided to ground, leading to the discharge of capacitor 402. Because capacitor 402 is relatively large, preferably 1.2 pf, and resistor 410 is also large, preferably 10K.OMEGA., this discharge of the capacitor provides a long delay (preferably 40-60 nanoseconds). The delay provided by discharging the capacitor is established to ensure the pump circuit 478 completes the pumping cycle (described in detail in reference to the pump circuit).

The voltage at node N2 of the capacitor 402 is input to a Schmitt Trigger inverter 420 and inverters 430 and 440 to provide an output to the gate of a transistor 450. The high output at the gate of the n-channel transistor 450 turns it on, thereby pulling the node START to ground and inverting the latch 302. The low voltage at the node START is transferred through inverters 316 and 322 to provide a low MCLKS output signal which starts the precharge cycle of the pumping circuit to prepare the circuit for the next pumping cycle. Even though START is now low, it cannot be brought high via latch 302 because the fast path caused transistor 314 to turn off. Since the path to ground for the gate of p-channel transistor 304 goes through the source-drain path of transistor 314, the precharge cycle will continue.

When master clock MCLKS is low, MCLKS is passed through inverters 328 and 334, providing a low output of inverter 334. The output of inverter 342 is high, turning off transistor 354 and turning on transistor 352. Because a path from capacitor 364 to ground is provided by way of transistors 360 and 352, allowing for the discharge of capacitor 364 (i.e., node N1), the delay through delay circuit 350 is long, while the delay through delay circuit 400 is short. Therefore, the passage of the low MCLKS through delay circuit 350 will be described after the short path through delay circuit 400.

The low output signal of inverter 334 is input to transistors 404 and 412 of delay circuit 400. The low signal turns on transistor 404 and turns off transistor 412. Therefore, the capacitor 402 is coupled to V.sub.CC via transistor 404 and the capacitor charges quickly, pulling the voltage at node N2 high. The voltage at node N2 is transmitted through Schmitt trigger inverter 420 and inverters 430 and 440. A low output of inverter 440 is input to transistor 450. The low signal at the gate of transistor 450 turns off the transistor, thereby releasing the clamp to ground, and allows the pumping cycle to begin again after (1) the signal passes through the slow path (i.e., delay circuit 350) to turn on transistor 314, and (2) the circuit receives an active pump signal at transistor 312 to reset latch 302.

When the low MCLKS passes through delay circuit 350 (i.e., during the precharge cycle), the discharge of capacitor 364 generally provides a delay. Specifically, the low output of inverter 334 is input to inverter 342. The output of inverter 342 is input to delay circuit 350. The high input turns on transistor 352 and turns off transistor 354. As a result, a path is provided from capacitor 364 to ground by way of transistor 360, a resistor 362 (preferably 10K.OMEGA. and transistor 352. A transistor 356 and resistor 358 (preferably 10K.OMEGA.) are added to maintain a constant current in the path from node N1 to ground and, therefore, establish a set delay. During this delay, the inputs to the charge pump circuit and voltages at various nodes of the charge pump are reset for a subsequent precharge and pumping cycle.

As the voltage at node N3 goes low as capacitor 364 discharges, the voltage is transmitted through Schmitt trigger 370, inverter 380 and inverter 390 to provide a high signal at the gate of transistor 314. A high signal at the gate of the n-channel transistor 314 turns on the transistor and allows transistor 312 to receive the pump signal for the subsequent pumping cycle.

This delay for the precharge cycle (set by delay circuit 350) is generally shorter than the delay for the pumping cycle (set by delay circuit 400) by providing a smaller capacitor which is discharged. A long delay during the precharge cycle is not required because the internal voltages in the pump reach precharge levels more quickly. This is because the gate-to-source voltages are higher and can react or provide more current at a faster rate.

To recapitulate, the standby master clock circuit 301 shown in FIG. 4 is responsive to an active PUMP signal (or clock enabling signal from the standby regulator circuit shown in FIG. 3 when V.sub.CCPREF drops below a certain value) and generates the MCLKS signal. MCLKS is fed back through the circuit and varies during different cycles. MCLKS is input to the clock circuit 445 shown in FIG. 5. As MCLKS transitions high or low, clock circuit 445 generates different clock signals during the different cycles to drive the charge pump. The circuit is self-timed. It completes the full cycle for pumping and for precharge once initiated by the active PUMP signal even if that signal terminates during the full cycle.

Turning now to FIG. 5, clock circuit 445 is shown. Clock circuit 445 receives MCLKS at an input and outputs signals CP1 through CP5. This circuit shows the preferred configuration of gates to generate the signals to drive the charge pump in FIG. 6. The signals generated by clock circuit 445 are shown in the timing diagram of FIG. 7. Clock circuit 445 uses a series of gates and feedback paths to generate the outputs shown in the timing diagram of FIG. 7. The operation of each gate shown in FIG. 5 is well known in the art. Generally, each signal is a part of a feedback loop to provide an inverted output depending on the status of the signals. As a result, the outputs transition based upon the state of the master clock signal as well as the previous signal.

The configuration of the gates is designed to establish the signals shown in FIG. 7. One skilled in the art could follow the MCLKS input signal to generate the output signals in FIG. 7 at any period based upon the previous input signals. While the preferred clock circuit is shown in FIG. 7, any other circuit or other means of generating signals (CP1-CP4) shown in FIG. 7 could be used within the scope of the present invention to generate the input signals to the pump circuits.

The function of the clock signals generated by the circuit of FIG. 5 will be described in detail in reference to the operation of the charge pump of FIG. 6. The clock signals are shown as a function of time in FIG. 7. The structure of the charge pump circuit will be described in detail, followed by a detailed description of the operation of the charge pump.

Pump Circuit

Turning now to FIG. 6, the charge pump circuit is shown. The charge pump preferred for use the present invention is disclosed in co-pending U.S. application Ser. No. 07/964,003, filed on Oct. 20, 1992 and owned by the assignees of the present invention (attorney docket number UM 104), the entire specification of which is incorporated herein by reference. Pump 478 receives at several inputs the clock signals CP1 through CP5, and provides an output V.sub.CCP while the operation of the pump is shown having a 2.7 volt V.sub.CC, some other value of V.sub.CC could be used. The use of a 2.7 volt V.sub.CC is simply illustrative.

The charge pump of the present invention is a hybrid two-stage charge pump having a first stage 477. Charge pumped to a node N4 during the first stage of a pump cycle (by first stage 477) is used to enable the charge pump to pump charge from a node N1 to V.sub.CCP during a second stage of the pumping cycle. Generally, signal CP1 is used to drive nodes N1 and N2 to a greater potential by way of capacitors 479 and 480. Signals CP2, CP3, CP4 and CP5 are generally used to regulate the operation of the remaining transistors in the circuit to pump charge to V.sub.CCP during the pumping cycle, and to reset the nodes during the precharge cycle.

Referring to the pump circuit of FIG. 6 and the associated timing diagram of FIG. 7, the operation of the pump circuit will be described. The input signal to the pump circuit and the voltages at various nodes are shown at t.sub.0 when the pump circuit is in a quiescent state. As stated earlier, the input signals CP1 to CP5 are established by the clock circuit 445 shown in FIG. 5, while the voltage at the nodes of the pump circuit in FIG. 6 are determined by the clock signals CP1-CP5. Initially, CP2 is high (V.sub.CC), turning on transistors 486 and 491 to maintain nodes N1 and N2 at V.sub.CC.

At a time t.sub.1, PUMP signal goes high (after the standby voltage regular circuit 110 shown in FIG. 3 detects a drop in V.sub.CCP). As described in reference to the standby master clock circuit 301 of FIG. 4, the MCLKS signal also goes high. As MCLKS is passed through the clock circuit 445 shown in FIG. 5, CP2 goes low (0 volts), turning off transistors 486 and 491 to prevent any transfer of charge from node N1 by way of transistor 486 to V.sub.CC and to prevent any transfer of charge from node N2 by way of transistor 491 to V.sub.CC. CP4 also goes high to turn on transistor 490 and to allow transfer of charge from N2 to N4 (when CP1 goes high at t.sub.2). Also, CP5 goes low to turn off transistor 488 and unclamp node N4 from V.sub.CC.

At a time t.sub.2, CP1 goes high, driving the voltage at nodes N1 and N2 to approximately 2 V.sub.CC (or 5.4 volts for a 2.7 volt V.sub.CC) . Because transistor 490 is on, node N4 approaches 2 V.sub.CC (or approximately 2 V.sub.CC minus V.sub.t or 3 V.sub.CC /2). This transfer of charge is necessary to turn on transistor 489 at a time t.sub.4.

At a time t.sub.3, CP4 goes low to turn off transistor 490 and to end the first stage of the pumping cycle of node N4 by preventing a transfer of charge from node N4 to node N2.

At time T.sub.4, CP3 goes high to start the second stage of the pumping cycle. A high CP3 signal further increases the voltage at N4 and, turns on transistor 489 to allow the transfer of charge from node N1 to V.sub.CCP. As is apparent in FIG. 7, the charge at node N1 and V.sub.CCP begin to equalize.

At a time t.sub.5, the pump signal may or may not go low (as shown by the dotted line), depending on whether V.sub.CCP has been restored to its desired value. The dashed line designating a high pump signal is shown in FIG. 7 and will be discussed further in regard to the end of the precharge cycle (at time t.sub.10).

At a time t.sub.6 (regardless of whether V.sub.CCP is restored to its desired value), the pumping cycle is complete as the high MCLKS signal is fed back through the master clock circuit 301 (discussed in FIG. 4). As MCLKS goes low, CP3 goes low to end the pump cycle.

The pump must then reset the inputs (CP1-CP5) and the voltages at the nodes (N1-N5) to their quiescent values for the next precharge and pumping cycle. At a time t.sub.7, CP5 transitions high to couple node N4 to V.sub.CC. At a time t.sub.8, CP1 transitions low to drive nodes N1 and N2 to V.sub.CC. At a time t.sub.9, CP2 goes high to couple nodes N1 and N2 to V.sub.CC and maintain the voltage at the node.

At a time t.sub.10, MCLKS has passed through master clock circuit 301 to turn on transistor 314, making latch 302 ready to respond to an active pump signal. If enough charge had not been pumped as of t.sub.5 (i.e., the PUMP signal had not gone low as indicated by the dashed line), pumping would begin at time t.sub.10 as MCLKS goes high (as shown in the dashed line). The precharge cycle would again begin as shown at time t.sub.1. Otherwise, the circuit remains in a quiescent state until a time t.sub.11, when V.sub.CCP has again dropped and an active PUMP signal is generated by the standby regulator 110 to begin the pumping cycle again.

To recapitulate, the standby circuit of the present invention monitors V.sub.CCP and activates the standby charge pump to maintain a desired voltage at V.sub.CCP. Generally, the standby regulator circuit, clock circuit and standby pump are designed to consume low power. Specifically, only the standby regulator circuit is constantly operating to monitor V.sub.CCP and is specifically designed to minimize power consumption. The clock circuit and pump circuits are enabled and draw current only when V.sub.CCP has dropped below a desired value, due typically to leakage current or initial power up.

Active Circuit

Having shown the standby circuit which continuously monitors V.sub.CCP to determine if V.sub.CCP has fallen due to leakage current, the active circuits (500) will now be described. Generally, active circuit 500 includes active regulator 200 (FIG. 8), active master clock 501 (FIG. 9), clock circuit 445 (FIG. 5) and pump circuit 478' (FIG. 6). As mentioned earlier, the active circuits operate independently of (although simultaneously with) the standby circuits and are enabled when the word lines of a DRAM are accessed, as indicated by a RASBP signal. If V.sub.CCP drops below a desired value when the word lines are accessed and the active circuit is enabled, the active pump (which is generally a larger pump than the standby pump) will pump charge to V.sub.CCP to maintain the desired voltage at V.sub.CCP.

Turning to FIG. 8, active regulator circuit 200 (also shown in block form in FIGS. 1 and 2) also regulates the voltages V.sub.CCPREF and V.sub.CCREF at nodes N1 and N2, respectively, to determine if V.sub.CCP has fallen below a desired value. Unlike the standby regulator 110 which activates standby clock 300 whenever V.sub.CCP drops below a desired value, the active regulator 200 also receives an ENREGB signal, which enables or disables the active regulator circuit 200. The state of ENREGB signal is determined, for example, by whether the word lines of a DRAM are driven and, therefore, the circuit is drawing current which lowers V.sub.CCP.

Specifically, a low ENREGB signal is inverted by inverter 202 and drives the gate of transistor 210 to enable differential amplifier 220. Differential amplifier 220 includes the conventional configuration of transistors 222-228. The operation of the differential amplifier is well known in the art. However, the differential amplifier further includes transistors 232 and 234 to enable the amplifier to respond more quickly to provide an output at V.sub.OUT. Transistors 232 and 234 generally serve the same functions as transistors 130 to 136 of standby differential amplifier 120 (FIG. 3). However, because long channel transistors are not used in this differential amplifier (unlike those of the standby differential amplifier 120) to limit current, the capacitance at the output node V.sub.OUT is not as great. Therefore, transistors 232 and 234 could be coupled in parallel to transistors 222 and 224, respectively, to more quickly drive the output without increasing the capacitance at the output node V.sub.OUT.

A switch 240 is also responsive to the ENREGB signal to allow the output of the differential amplifier 220 to be output as a PUMPA signal (similar to the PUMP signal of the standby master clock circuit). Specifically, a low ENREGB signal is twice inverted by inverters 202 and 241, so node N7 is also low and is coupled to the gate of transistor 208, turning off the transistor. The low output of inverter 241 is inputted to inverter 249 and is also applied to the gate of transistor 252, which turns on. The output of inverter 249 turns on transistor 250, providing a path from the output of the differential amplifier (V.sub.OUT) to the output at PUMPA by way of inverters 260 and 270. Specifically, the output V.sub.OUT of differential amplifier 220 is transmitted through the switch 240 to node N8. Node N8 is coupled to a Schmitt trigger inverter 260, the output of which is inverted by an inverter 270 to provide the output signal PUMPA. The PUMPA signal is used by the active master clock circuit 501 (shown in FIG. 9) to generate a MCLK signal.

Turning now to FIG. 9, the active master clock circuit 500 outputs a high MCLK signal when an input RASBP is low and an input PUMPL (i.e. PUMPA previously latched as PUMPL) is high. Generally, a high RASBP drives ENREGB low which enables the active voltage regulator described supra in FIG. 8 to monitor V.sub.CCP. If V.sub.CCP has dropped below a desired value, the active voltage regulator will output an active PUMPA signal (which is input to the active master clock circuit 501 of FIG. 9). When PUMPL is high, MCLK is high to enable a clock circuit to drive the pump circuit. Note that the same clock circuit 445 shown in FIG. 5 is used to generate clock signals CP1-CP5 to drive the active pump circuit 478'. Also, the active pump circuit 478' is identical in structure to the standby pump circuit. However, transistors having different channel widths and lengths are used in the active pump circuit to drive more current to maintain the voltage at V.sub.CCP. The preferred values of the channel lengths and widths of corresponding transistors employed in the active pump are shown in parenthesis in Appendix A.

In addition, the operation of clock circuit 445 and the active pump circuit 478' is identical to that of the standby circuit. Specifically, clock circuit 445 is responsive to the MCLK signal (similar to MCLKS of the standby master clock circuit 301) to generate clock signals to operate the active pump circuit 478'. The relationship of the MCLK signal to the clock circuit 445 and the active pump circuit 478' has been described in detail in reference to the standby circuit and will not be repeated here. Therefore, only the role of the RASBP signal (indicating that the word lines of a DRAM are being accessed) and other signals in the active master clock circuit 501 to generate a MCLK signal will be discussed in detail.

The detailed operation of active master clock circuit 501 of FIG. 9 is now described. Because RASBP can be either high or low (indicating whether the word lines of a DRAM are being accessed and the active pump should be activated if V.sub.CCP drops below a given value), the operation of the circuit will be described separately in both states. When RASBP transitions from high to low (i.e. the word lines will be driven), delay circuit 502 provides a low RASBPP at node N2 after a small delay. RASBPP is a signal which remains low for an additional period of time longer than RASBP, as will be explained infra. When RASBP goes low, a transistor 506 is turned off, decoupling the path from a current source 509 to ground. A low RASBP signal also turns on a transistor 504, which couples a node N1 to V.sub.CC. Because FET 504 presents little resistance between node N1 and V.sub.CC, capacitor 508 charges almost immediately and node N1 transitions quickly to V.sub.CC. This high state at node N1 provides a low RASBPP after passing through a sequence of inverters 520, 530 and 536.

RASBPP signal is input to a number of circuits which control the output MCLK. Generally, the timing of the circuit will be understood in reference to the diagram of FIG. 10. Typically, a constant current source is used to discharge a capacitor to set a delay, although other means of setting the delay could be used. Note generally in FIG. 10 that when RASBP transitions low (i.e. at a time t.sub.1) RASBPP transitions low after only a very short delay. However, when RASBP transitions from low to high at a time t.sub.5 (discussed in detail infra), RASBPP does not transition high until time t.sub.6 after a delay 3 (shown in FIG. 10). This delay is used to ensure that the pumping cycle lasts sufficiently long enough to pump charge and maintain the desired voltage V.sub.CCP.

In the preferred embodiment, when the RASBPP signal transitions from high to low, transistor 546 of the delay circuit 540 is turned off, decoupling the path from current source 549 to ground. Also, transistor 544 is turned on, coupling the capacitor 548 at node N3 to V.sub.CC. This capacitor charges quickly, providing a high output at node N3 to inverters 560 and 570.

The output of inverter 570 and signal RASBPP are input to a NAND gate 580 at nodes N4 and N5 respectively. The output of the NAND gate (NAND OUT) is therefore high. This high output of the NAND gate provides a high ENREGB signal after NAND OUT passes through inverters 590 and 596. Therefore, when RASBP is low, active master clock circuit 500 disables the active voltage regulator circuit 200 (shown in FIG. 8) and regulator 200 then no longer monitors V.sub.CCP or transmits a high PUMPA signal if V.sub.CCP has dropped below a desired value.

The output (NAND OUT) of NAND gate 580 also coupled to a switch 600. This switch generally allows the PUMPA input signal to be latched by latch 610 (as PUMPL). Inverters 590 and 596 are coupled to the output of NAND gate 580 to add a short delay to ensure that switch 600 is disabled before ENREGB goes high to disable the voltage regulator circuit. This ensures that the correct PUMPA signal is latched.

Latch circuit 610 is coupled to an inverter 620, the output of which is a signal PUMPL. When RASBPP is low, switch 600 prevents the PUMPA signal from being transferred to transistor 626. Therefore the value of PUMPL will be determined by the previously latched PUMPA signal (i.e. when RASBPP was previously high). PUMPL is input to a transistor 626. The latch 630 will be allowed to receive PUMPL when transistor 640 is turned on by a low RASBPP signal passed through inverter 642. When RASBPP is low, transistor 648 is turned off and the output of latch 630 is determined by PUMPL.

The output of latch 630 is passed through inverters 650 and 660 to generate the output signal MCLK. If PUMPL is high and RASBPP low, the START signal will be high, which is output through inverters 650 and 660 to provide a high MCLK. If PUMPL is low, the START signal will be low to provide a low MCLK.

Next, the operation of active master clock 501 will be described as RASBP transitions from low to high at time t.sub.5. When this occurs, transistor 504 is turned off and transistor 506 is turned on. Therefore, capacitor 508 discharges to ground through constant current source 509. Therefore, a delay is provided (designated as delay 3 in FIG. 10). As stated earlier, this delay is to ensure that the pumping circuit 478' remains in the pumping cycle for a sufficient time to adequately pump V.sub.CCP to a desired voltage. As capacitor 508 discharges, the low signal at node N1 is transmitted through inverters 520, 530 and 536. The output of inverter 536 provides a high RASBPP signal at a time t.sub.6 after capacitor 508 is discharged (i.e. after delay 3).

The high RASBPP signal is input to delay circuit 540. The high signal turns off transistor 544 and turns on transistor 546, providing a path from capacitor 548 at node N3 to ground by way of current source 549. As the capacitor 548 discharges at node N3 a delay is provided. During this delay, RASBPP is high while the output of inverter 570 is high. Both RASBPP and the output of inverter 570 are input to NAND 580. As a result, the output (NAND OUT) of the NAND gate is low, opening switch 600 and providing a low ENREGB. After transistor 548 discharges, the output of inverter 570 is low, driving the output of the NAND gate (and ENREGB) high.

This delay (designated as delay 4 in FIG. 10) is provided to ensure that ENREGB remains low for a predetermined time for the voltage regulator circuit to compare V.sub.CCPREF with V.sub.CCREF to determine whether to pump on the next cycle where RASBP goes low. Also, as stated earlier, inverters 590 and 596 are added to the output of the NAND gate 580 to ensure that ENREGB stays low when switch 600 is closed (i.e. when the output NAND OUT of the NAND gate 580 goes low) to ensure that the correct PUMPA signal is latched by latch 610 when RASBP is high.

At the same time that the high RASBPP signal generates the ENREGB signal to determine whether to pump when RASBP goes low, the high RASBPP also generates a low MCLK signal (at time t.sub.7 in FIG. 10). Specifically, the latch 630 will be allowed to receive PUMPL only when transistor 640 is turned on by a low RASBPP signal passed through inverter 642. Because RASBPP is high, the output of inverter 642 is low, turning off transistor 640 to ignore the PUMPL signal, providing a self-timing clock which eliminates the requirement for an oscillator. Also, the gate of transistor 648 is high, turning on the transistor to drive the START node low and to generate a low MCLKS signal.

When RASBPP goes low at a later time, switch 600 is closed, and the circuit is ready to pump based upon the previously determined pumping signal (PUMPL) which is input to the gate of transistor 626 to output an appropriate MCLK signal. Therefore, MCLK will drive the active charge pump only when the appropriate conditions exist (i.e. when the RASBP signal indicates that the word lines are activated, drawing charge from V.sub.CCP, and when PUMPA (latched as PUMPL) has gone high indicating that V.sub.CCP has dropped.)

Second Embodiment

In a second embodiment, the active regulator circuit can be eliminated as shown in the block diagram of FIG. 11 and the charge pump can pump whenever RASBP goes low. The only difference in this embodiment compared to the first embodiment can be found in the master clock circuit 501. Like circuit components in FIG. 12 are numbered the same as those in FIG. 9. The primary difference in the circuit is that the circuit does not require an active regulator. Rather, a pump signal (i.e. PUMPA or PUMPL) is pinned at V.sub.CC and the output MCLK of this master clock circuit is high whenever RASBP goes low. The operation of this circuit is identical to the circuit in FIG. 9, excluding any requirement to generate a PUMPA signal, a PUMPL signal and an ENREGB signal, and therefore, is not described again in detail here.

While this invention has been described with reference to an illustrative embodiment, it will be understood that this description is not intended to be construed in a limiting sense, but is intended to include any substitutes within the scope of the invention. Various modifications of the illustrative embodiment, as well as other embodiments, will become apparent to those skilled in the art upon reference to this description. The invention is intended to be set forth in the depending claims.