Semiconductor Memory

1. Field of the Invention

This invention relates to a semiconductor memory equipped with a chip selection signal terminal, and in particular to a semiconductor memory in which the access time as counted from the input of a chip selection signal is reduced.

2. Description of the Related Art

FIG. 6(a) is a block diagram of a conventional semiconductor memory including a memory cell array and a precharger/equalizer which are shown in detail in FIG. 6(b). CSrepresents the chip selection signal which is applied to a selection signal generating circuit 1. The H (High) level of this signal indicates the nonselecting condition, the memory circuit being in the stand-by condition, the inner selection signal CSAbeing at the H level. When signal CSAis at the H level, the operation of an address input buffer circuit 2 is inhibited and the operation of a decoder 4 is halted, keeping the level of the word and column select lines, which constitute the selection lines of memory cell array 6, at their lower level.

CTL is a control signal for circuit operation, controlling the operation of a precharger/equalizer 5 for the data lines for transmitting the data of memory cell array 6 as well as the operations of a data amplifier 7 and an output circuit 8. In the stand-by condition, signal CTL is at the H level, the data lines being precharged and equalized. At the same time, the operation of circuits 7 and 8 is stopped. When the signal CSis at the H level, the signal CSBis at the H level, keeping CTL at the H level. Further, a pulse signal generator 3 is provided so that operation may also be effected when changes occur in the address signals Ai supplied to circuit 2.

FIG. 7 shows concrete circuit examples of the selection signal generator 1, the address input buffer circuit 2 and the pulse signal generator 3. The signal CSAis input to input NOR gates 20 for the address signals, preventing changes in signals Ai from being transmitted to the inside of the circuit when CSAis at the H level. Any change in Ai when CSAis low (L level) is detected by the circuit 3 with respect to the rise and fall of the signal, a pulse signal ATD being generated for each address to perform OR operation of the signals ATD for all the addresses. The signal ATD associated with each address signal Ai (i=o . . . n) is supplied to a respective input device of OR circuit 3-2. When signal CSgoes low, the corresponding drop in CSBis delayed by the delay circuit DB of selection signal generator 1; when CSBdrops to the L (Low) level, a P-channel MOS transistor 3-1 is turned on, causing the OR circuit 3-2 to operate. When CSBis at the H level, P-channel MOS transistor 3-1 is turned off and current is not supplied to OR circuit 3-2.

Next, the operation of the circuit shown in FIG. 6(a) will be described with reference to the timing chart of FIG. 8. When the signal level of CSis changed from the H to the L level, CSAis changed to the L level, an address signal Ai being transmitted to the inside of the circuit through the associated input NOR gate 20. When Ai is at the L level, a change occurs in the inside address signal, generating an ATD pulse. CTL level is lowered at the termination of the ATD pulse; if the Ai's are all at the H level and no ATD pulse is generated, the CTL lowering time does not change since CSBis lowered simultaneously with the ATD pulse lowering. The address signals Ai are transmitted to decoder 4, and the signal level of the word lines and the column select lines which lead to the memory cells to be selected is raised. As is well known, the word lines are connected to the memory cells to select those memory cells which are in the ROW direction. The column select lines select the data lines which are connected to the column of the memory cell to be selected. The information stored in the memory cells then appears on the data lines, and is output to the output terminal through amplifier 7 and output circuit 8.

When an address signal Ai changes while CSremains at the L level, an ATD pulse is generated, causing the CTL to be raised to the H level so that precharging and equalization of the data lines are effected. Readout of data from the next address is then performed.

In conventional semiconductor memories with such a construction as described above, the CSaccess timing when the address signal Ai is at the L level is such that when CSchanges to the L level and the CSAlevel is consequently lowered, the output of the corresponding address input NOR gate 20 changes. In the case of CSaccess, access is started at this time. In the case of address access, the output of NOR gate 20 changes when the address signal Ai changes since CSAis already at the L level, thus starting access. Accordingly, the CSaccess is more delayed than the address access by the time between lowering of CSand the resultant lowering of CSA.

In the case of CSaccess, the data line is already precharged and equalized, access being started in this condition. Accordingly, there is no particular limit to the speed-up of the signal level rise of the word lines and the column select lines; the faster the rise, the better.

It would be advantageous, if, in a conventional semiconductor memory, the CSBdrop could occur at a shorter time after a CSdrop, so that the width of the ATD pulse diminished, thereby speeding up the operation of the address input buffer 2 and of the decoder 4. However, in a conventional semiconductor memory, reducing the width of the ATD pulse results in the pulse width of the CTL being also diminished. In the case of address access, however, there is a danger of the wrong data being written to the memory cell selected next, thereby destroying the stored data, unless the data lines are kept in the reset condition for a sufficient length of time through precharging and equalization by the CTL since the data of the previous cycle remains on the data line. Further, it is not sufficient for the operation of data amplifier 7 and of output circuit 8 to be inhibited for a short period of time by CTL since that will cause amplifier 7 to be operated before the data appears on the data line to a sufficient degree, which will cause malfunctions due to noises, resulting in a delayed access time.

Thus, it has been impossible in conventional semiconductor memories to speed up the CSaccess by diminishing the width of the ATD pulse since that involves the above-mentioned problem at the time of address access.

It is therefore an object of the invention to alleviate the above-mentioned problem encountered in conventional semiconductor memories.

A more specific object of this invention is to provide a semiconductor memory in which a high-speed CSaccess is realized and a stable address-access operation is maintained.

The above and other objects are achieved, according to the present invention, by a semiconductor memory actuatable in response to a chip selection signal and address signals, comprising:

a memory cell array having data lines;

a selection signal generating circuit connected for generating inner selection signals in response to the chip selection signal, each inner selection signal having an associated chip selecting state;

a control signal generating circuit connected for detecting changes in the address signals and generating a first control signal containing a pulse in response to each address signal change; and

a pulse width changing circuit connected for deriving from the first control signal a second control signal containing pulses for controlling the data lines of the memory cell array, each pulse of the second control signal corresponding to a respective pulse of the first control signal, wherein the pulse width changing circuit comprises means for causing each pulse of the second control signal to have a longer duration than the corresponding pulse of the first control signal when the inner selection signals are in the chip-selecting state.

FIG. 1 is a block diagram of a semiconductor memory in accordance with this invention.

FIG. 2 is a circuit diagram showing embodiments of the pulse signal generating circuit, the pulse width changing circuit and the selection line control circuit of the memory of FIG. 1.

FIG. 3 is a circuit diagram showing other embodiments of the pulse width changing circuit and the selection line control circuit of the memory of FIG. 1;

FIG. 4 is a circuit diagram showing an embodiment of the decoder of the memory of FIG. 1.

FIG. 5 is a timing chart illustrating the operation of a semiconductor memory in accordance with this invention.

FIGS. 6(a) and 6(b) are block diagrams illustrating a conventional semiconductor memory.

FIG. 7 is a circuit diagram showing a selection signal generating circuit, an input buffer circuit and a pulse signal generating circuit in the conventional semiconductor memory of FIGS. 6.

FIG. 8 is a timing chart illustrating the operation of the conventional semiconductor memory of FIGS. 6 and 7.

Embodiments of this invention will now be described with reference to the attached drawings. FIG. 1 is a block diagram of a semiconductor memory in accordance with this invention. FIG. 5 is a timing chart illustrating the operation of the circuit shown in FIG. 1. CSdenotes the chip selection signal input to the chip selection signal terminal of the semiconductor memory, the signal being in the chip selecting condition when it is at the L (low) level, and in the nonselecting condition when it is at the H (high) level. When CSis input to a selection signal generating circuit 1', inner selection signals CSA, CSBand CSCare generated, CSCbeing transmitted to a pulse width changing circuit 9 and a selection line control circuit 10. A signal CTL1 is generated by a pulse signal generating circuit 3 and is input to circuits 9 and 10. A signal CTL2, produced by circuit 9, controls the operation of a precharger/equalizer 5, a data amplifier 7 and an output circuit 8. The word and column selection line selecting operation is conducted by a decoder 4 for decoding address signals Ai. A selection line control signal WCTL, produced by circuit 10, effects active/inactive control of decoder 4.

As in the conventional memory shown in FIG. 7, the selection signal generating circuit 1' generates the signal CSB. However, the delay produced by delay circuit DB' of FIG. 1 is set shorter so that the CSBdrop may be faster than in conventional memories. The embodiment shown in FIG. 1 further includes an address input buffer circuit 2, the basic construction of which is the same as the conventional example shown in FIG. 7, except for the transistor size which is designed to be larger so as to realize a high-speed operation. Likewise, the basic construction of the pulse signal generating circuit 3 is the same as in the conventional memory shown in FIG. 7, except for the fact that the number of inverter stages is reduced so as to diminish the width of the ATD pulse, thereby reducing the delay.

FIG. 2 shows embodiments of the selection signal generating circuit 1', the pulse width changing circuit 9 and the selection line control circuit 10. A drop in the signal CSCis delayed by a delay circuit DC. When CSCis at the H level in circuit 9, the level of CTL1 determines the level of CTL2 output through a NOR gate 91 and an inverter 92. When the CSCis at the L level, CTL1 is input to NOR gate 91 not only directly but also through a three-stage inverter and a NOR gate 90, OR operation being performed at NOR gate 91. Thus, the output of NOR gate 90 is delayed as a result of the signal passing through four gates in total, CTL2 then appearing as a signal having a correspondingly increased pulse width. When at the H level, CTL2 precharges and equalizes the data lines, like signal CTL in FIG. 6(b) inhibiting the amplification and output of data.

To avoid malfunction, the pulse width of CTL2 at the time of an address change should be sufficiently large; it must be set to a value approximately equal to the pulse width of the ATD and CTL in conventional memories. Since, according to the present invention, the ATD pulse width is relatively small, the pulse width changing circuit 9 serves to augment the pulse width. That is, the circuit 9 augments the pulse width when address access is performed with CSCat the L level. When CSaccess is performed with CSCstarting at the H level, CSCdoes not delay the fall of CTL2 but it is caused to change quickly.

When, in the selection line control circuit 10, CSCis at the H level, WCTL is held at the L level. When CSCis at the L level, WCTL is constituted by CTL1, which is output through a NAND gate 100 and an inverter 101.

Thus, when CSCis at the H level, precharging and equalization are only performed for a short period of time, the word and column select lines not being deactivated. When CSCis at the L level, the long pulse of CTL 2 causes precharging and equalization to be performed for a long period, the word and column select lines being temporarily deactivated by WCTL so as to avoid malfunction at the time of address change.

The circuit diagram of FIG. 3 shows other embodiments of the pulse width changing circuit 9' and the selection line control circuit 10'. When, in the example shown in FIG. 3, an IC chip is nonselected, i.e., when CSis at the H level, CTL2 must be at the H level before precharging and equalization can be performed. Further, WCTL must be at the H level for the signal level of the word and column select lines to fall. In circuit 9', CSBis input to a NOR gate 93, and, in circuit 10', CSAis input to a NAND gate 103 through an inverter 102. Thanks to this arrangement, it is not necessary to input CSBto circuit 3 or CSAto the circuit 4 as in the embodiment shown in FIG. 1.

In other words, in the embodiment shown in FIG. 1, CSBis input to pulse signal generating circuit 3 to control CTL1, whereas in that shown in FIG. 3, CSBis input to the NOR gate 93 along with CTL1, the control by CSBbeing performed in circuit 9'. Further, while in the embodiment shown in FIG. 1 decoder 4 is controlled by CSA, it is controlled by the WCTL when the circuit shown in FIG. 3 is used. Consequently, in the case of the example shown in FIG. 3, WCTL remains at the H level until the CSAfalls, inhibiting operation of decoder 4.

The circuit diagram of FIG. 4 shows an embodiment of the decoder 4' of this invention. In order to deactivate all the word lines by means of WCTL, WCTL is input to NAND gates 40 of decoder 4' through an inverter 41. When WCTL is at the H Level, the outputs of the NAND gates 40 are all at the H level, so that the word lines Wi are all at the L level. This also applies to the column select lines.

The operation of the circuit shown in FIG. 1 will now be described with reference to the timing chart of FIG. 5. When CSis at the H level, it implies that the memory is in the condition where the chip is nonselected. At this time, the signals CSA, CSBand CSCare all at the H level. When CSis lowered to the L level, i.e., the selection condition is indicated, CSAis first lowered to the L level and the address signals Ai are output through address input buffer circuit 2, the address signals Ai being transmitted to the inside to generate an ATD pulse in pulse generating circuit 3. However, since CSBremains temporarily at the H level, CTL1 does not change and remains at the H level even when the ATD pulse is generated. When CTL1 is at the H level, CTL2 is also at the H level. CTL2 causes the precharger/equalizer 5 to precharge and equalize the data lines, the operation of the data amplifier 7 and that of the output circuit 8 being inhibited.

CTL1 falls together with CSB. Since CSCis still at the H level when CTL1 falls, the inputs of the NOR gate 91 or NOR gate 93 are all at the L level, so CTL2 also falls following CTL1. This terminates the precharging and equalization, causing amplifier 7 and output circuit 8 to be activated. Since at this time WCTL remains at the low level, the NAND gates 40 in FIG. 4 are not inhibited by WCTL. Accordingly, the output of input buffer circuit 2 is decoded by decoder 4 ', thereby performing word line selection. Column select line selection is also performed in the same manner. Accordingly, the word line signal level rises in accordance with the operational speed of the address input buffer 2 and decoder 4', thereby performing memory cell selection. After that, CSCis lowered to the L level.

When Ai changes following data readout from an address and address access is started, an ATD pulse is generated in pulse signal generating circuit 3, CTL1 also changing following the ATD to become a pulse at the H level. Since at this time CSCand CSBare already at the L level, the pulse width changing circuit 9 or 9' operates. The pulse rise of CTL1 causes a rise of CTL2 through the NOR gate 91 or 93. On the other hand, fall of CTL1 pulse is input to NOR gate 91 or 93 after being delayed through the three-staged inverter and NOR gate 90. As a result, the fall of CTL2 is delayed. This is, as will be appreciated from FIG. 5, the CTL1 pulse is transformed into a longer pulse, CTL2.

At this time, the selection line control circuit 10 or 10' also operates, a pulse appearing on WCTL. When, as shown in FIG. 4, WCTL attains the H level, all the NAND gates 40 output H level signals, thereby lowering the level of all the word lines Wi to the L level. Since the word lines perform memory cell selection at the H level, all of them are temporarily put in the nonselecting condition at the time of address change. As a result, when address change occurs, no memory cell is selected to bring about an unstable condition in the chips, thereby improving memory reliability.

As will be appreciated from FIG. 5, the CTL2 pulse is long at the time of address change, so that CSaccess takes less time than address access in respect of the time between the ATD pulse rise and the access completion. In the case of address access, the CTL2 pulse width must be of the same duration as in conventional cases so as to avoid malfunction. In view of this, at the time of the address change shown in FIG. 5, the CTL2 pulse width is set at approximately the same value as the CTL pulse of conventional memories. At the same time, the ATD pulse width is diminished and the operational speed of the address input buffer circuit and that of the decoder circuit increased so that CSBmay fall sooner. Thus, in accordance with this invention, the CSBfall is earlier and the ATD pulse is shorter than in the conventional CSaccess, so that precharging and equalization are terminated earlier. By speeding up the operation of the address input buffer circuit and that of the decoder circuit, the CSaccess becomes faster. On the other hand, since a short ATD pulse is transformed into a sufficiently long one at the time of address access, the precharging and equalization of the data lines are positively performed, thereby avoiding malfunction.

As described above, this invention makes it possible to speed up access in the time of CSaccess while securing a sufficient length of time for precharging and equalization at the time of address access. At the time of CSaccess, the signal rise in the word lines and the column select lines can be made faster by speeding up the operation of the address input buffer circuit and that of the decoder circuit. Furthermore, since the operation of the amplifier and that of the output circuit are started with the precharging and equalization of the data lines promptly completed, the entire circuit can operate at a high speed, simultaneously with the rise of the word lines or that of the column select lines.

In addition, since a sufficient length of time can be secured for precharging and equalization at the time of address access, and consequently the word lines and the column select lines only rise after the data lines have attained a stable condition, amplification and output can be effected with stability without destroying the data in the memory cells. Furthermore, since all the word lines are temporarily put in the inactive condition at the time of address access, malfunction at the time of access can be avoided.

Since, in a semiconductor memory according to the present invention, buffer circuit 2 and pulse signal generating circuit 3 can operate faster than corresponding prior art circuits, such prior art circuits differ from those employed in the practice of the present invention in that the prior circuits have at least one of the following:

two or four additional inverters in each channel;

inverters having 30-50% longer gates;

inverters having gate areas which are larger by approximately 40 .mu..sup.2, with correspondingly larger capacitances.

This application relates to subject matter disclosed in Japanese Patent Application No. 094122/88, filed on Apr. 15, 1988, the disclosure of which is incorporated herein by reference.

While the description above refers to particular embodiments of the present invention, it will be understood that many modifications may be made without departing from the spirit thereof. The accompanying claims are intended to cover such modifications as would fall within the true scope and spirit of the present invention.

The presently disclosed embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims, rather than the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.