JPH0461435B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to semiconductor memory devices such as MOS dynamic memories.

[Conventional technology]

Figure 1 shows a so-called address multiplex method, in which the address signal is divided into two, a row address and a column address, and these are input in a time-division manner through the same input terminal, and the row address remains fixed. 1 shows a schematic circuit configuration of a dynamic memory using N-type MOS transistors according to the prior art, which has a function called page mode in which only column addresses are continuously changed. 1st
In the figure and other drawings, reference symbols with suffixes R and C are attached to circuit portions related to row selection and column selection operations, respectively. 1R and 1C are external control clocks, the former mainly controlling the start of the row selection operation, and the latter controlling the start of the column selection operation. 2R and 2C are circuits that receive the inputs of 1R and 1C, respectively, and generate a plurality of timing pulses necessary for the internal operation of the memory. In the figure, only representative outputs 11R, 12R, 13R, and 12C are shown, and the others are omitted. Circuit 2C is signal 11
It responds to signal 1C only under the condition that R is input. 3 consists of a plurality of signal lines for inputting row or column addresses consisting of a plurality of bits in parallel. The address buffer circuits 4R and 4C take in the row address and column address input in a time-sharing manner via the line 3 in accordance with the address buffer control signals 12R and 12C supplied from the circuits 2R and 2C, respectively. Row address signal 14R, its inverted signal 14 and internal column address signal 14
C and its inverted signal 14 are output. signal 14
R, 14R are row decoders (not shown), word lines
The signals 14C, 1 are supplied to the word line selection circuit 5R from a drive circuit (not shown) for W ₁ to W _n , and the other signals 14C, 1
4C is supplied to a bit line selection circuit 5C comprising a column decoder (not shown) and a drive circuit (not shown) for bit line selection lines Y ₁ to _Yo . Reference numeral 100 denotes a memory cell array section in which a pair of bit lines is connected as a so-called folded bit line.
A memory cell MC consisting of _a 1MOS transistor is arranged at one of two intersections between each of bit line pairs B _{1 -B o} and word lines W ₁ -W _n .
A dummy cell (not shown) is also connected to each bit line. 6R is a detection circuit for a minute signal from the memory cell MC, which is composed of transistors Q ₁ and Q ₂ and operates according to the instruction of the detection circuit drive signal 13R supplied by the circuit 2R. The gate circuit 101 has a pair of gates provided for each data line pair.
It has MOS transistors and connects the input/output data line pair I/O and the corresponding bit line pair in response to signals on the lines Y ₁ to _Yo . 6C is a detection circuit, 7C is an output amplification circuit, and 8 is an output terminal.
9 is a data input terminal, and 10C is a data input buffer. Note that each circuit in FIG. 1 is of a dynamic type. Hereinafter, the operation of the circuit shown in FIG. 1 will be explained with reference to FIG. 2.

First, when the row selection control clock 1R goes low, 12R clocks among the multiple internal clocks necessary for internal operations are generated by the circuit 2R and input via line 3 in synchronization with the falling edge of the signal 1R. The circuit 4R takes in the column address and generates internal address signals 14R and 14. Since this memory operates in a multi-address system, only the row address is first input to line 3. C1, C2, C3...Cj in Figure 2 are column addresses that will be input later.

Circuit 5R operates in response to internal address signals 14R and 14, and one of word lines _W1 to _Wn , for example _W1, is selected. In this way, a plurality of memory cells connected to the selected word line _W1 are read out. Each bit line is provided with a dummy cell (not shown), and the dummy cell connected to the bit line paired with the bit line connected to the selected memory cell is read by circuit 5R. In this way, minute signals are read out onto n bit line pairs. Thereafter, the signal 13R becomes a low potential, each detection circuit 6R operates, and the voltage of each data line pair is differentially amplified. With this operation, the row selection operation is generally completed.

When column select control clock 1C then goes low, circuit 2C generates signal 12C. In addition,
Signal 11R is an inverted signal of signal 1R, and circuit 2C is configured to respond to the falling edge of signal 1C only when signal 11R is at a high level. Signal 1C
The circuit 2C responds to the rise of the signal 11R regardless of the level of the signal 11R. Circuit 4C takes in column address C1 input via line 3 in synchronization with the fall of signal 1C in response to signal 12C, and generates internal address signals 14C and 14. signal 1
In response to signals 4C and 14, circuit 5C selects one of the bit line selection lines _Y1 to _Yo , for example _Y1 . As a result, MOS transistors Q ₃ and Q ₄ are turned on, and the signal on the data line pair B ₁ is transferred to the input/output data line pair I/O, differentially amplified by the detection circuit 6C, and its output is further output. The read data is amplified by the amplifier circuit 7C and outputted to the output terminal 8.

[Problem to be solved by the invention]

In the normal mode, both signals 1R and 1C are then returned to high potential, and the memory returns to its original standby state. At this time, the memory signal takes the level shown by the dotted line in FIG.

That is, when the signal 1R becomes high level, the circuit 2R is connected to a circuit related to the row selection operation, for example, 4.
R, 5R, 6R and the cell array section 100 are each in a standby state (that is, a precharge state)
It has a circuit (not shown) for supplying a signal. On the other hand, when the signal 1C becomes high level, the circuit 2C precharges the circuits involved in the column selection operation, such as the circuits 4C, 5C, 6C, 7C, 10C and the data line pair I/O, and puts them in a standby state. It has a circuit (not shown) for supplying a signal.

On the other hand, in page mode operation, the output terminal 8
After the output appears, as shown in FIG. 2, the signal 1R is kept at a low potential state, only the signal 1C is turned on and off, and only the column selection operation is performed continuously.

In this page mode, the signal 1R is in a low potential state, so the circuit related to the row selection operation remains in its previous state, that is, in this example, word line W ₁ is selected, and the detection circuit 6R remains in the operating state. It is. Therefore, when the signal 1C becomes a high potential state, only the circuits involved in the column selection operation, such as the circuits 2C, 4C, 5C, 6C, and 7C, go into a standby state at a predetermined timing and prepare for the next operation. Thereafter, when signal 1C goes low, circuits 2C and 4C operate in the same manner as described previously, circuit 4C takes in the next column address C2 input via line 3, and signals 14C and 14 are sent to circuit 5C. supply The circuit 5C selects one of the bit line pair selection lines Y ₁ -Y _o corresponding to the signals 14C and 14, and the signal of the corresponding bit line pair is transferred to the input/output data line pair I/O. , the data is output to the output terminal 8 via the circuit 7C. After that, the same operation continues and column addresses C3, C4...Cj
Data corresponding to is continuously output to terminal 8. At the end of page mode, signals 1C, 1
Both R are returned to high level and the memory is returned to its original standby state.

As mentioned above, in page mode, row selection operations are not repeated, so faster operations are possible than normal, and the access time in this case is equal to the time t _CA from column address input to data output. , this time _tCA is approximately 1/2 to 2/3 of the access time (time from row address input to data output) _tRA during normal operation.

Furthermore, the maximum number of data j that can be continuously read in page mode is, in principle, equal to the number n of bit line pairs that can be specified by a column address, assuming that data from memory cells at different addresses is always read. Normally, in multi-address memory, the number n of bit line pairs and the number m of word lines are made equal, so if the storage capacity of the entire memory is N, then
j=√N. This value is a theoretical value and can be changed as appropriate in relation to other characteristics, but usually j=
The number ranges from several + to several hundred, and in page mode, this amount of different data can be read out consecutively in the above access time.

However, even in the page mode described above, the access speed is slow for use as the main memory of a computer.

An object of the present invention is to provide a memory capable of successively reading data from two memory cells in one memory array connected to different word lines.

[Means to solve the problem]

one memory array having a plurality of word lines, a plurality of data line pairs, and a plurality of memory cells arranged at desired intersections of the plurality of word lines and the plurality of data line pairs; a common data line pair connected to the line pair; a connecting means connected between the plurality of data line pairs and the common data line pair; and a data holding means having an input connected to the common data line pair. , a data output means having an input connected to the output of the data holding means; a disconnection means disposed between the common data line and the input of the data holding means; and a first memory in the memory array. comprising a first word line connected to the cell, a second word line connected to the memory cell, and selection means for selecting the first word line and the second word line, While the data of the first memory cell held in the data holding means is being outputted by the data output means, the second memory cell is transferred by selecting the second word line. The data of the first and second memory cells belonging to the one memory array can be read out in succession by reading the data of at least the second memory cell to the data line pair connected to the second memory cell.

[Effect]

Two memory cells connected to different word lines in one memory array can be accessed consecutively in a short time.

〔Example〕

The present invention will be illustrated below with reference to Examples.

(1) Continuous mode In Figure 3, the same reference numbers as in Figure 1 indicate the same items as in Figure 1. Memory cell array 100 consists of an array of memory cells having the same structure as in FIG. In this embodiment, four input/output data line pairs I/O to I/O are provided, and four bit line pairs are simultaneously selected from cell array 100 during column selection operation. Therefore, the cell array 100 has four blocks 100 to 100.
Each block has the same i bit line pairs. The bit line pairs of block j (1≦j≦4) are represented by numbers Bj1 to Bi. The column addresses of bit line pairs Bk to Bk (1≦k≦i) are assigned so that they are the same except for the lower two bits.

In this embodiment as well, the address multiplexing method is used as in FIG.

The address buffer circuit 4CA takes in only the upper bits other than the lower two bits of the column address input via the line 3, and outputs the corresponding internal column address signal 14CA and its inverted signal 14. So circuit 4C in Figure 1
different from.

Along with this, the bit line selection circuit 5CA is
This bit line selection circuit differs from the bit line selection circuit shown in FIG. 1 in that it responds to internal address signals 14CA and 14. Note that, for the sake of simplicity, a signal line connecting the bit line selection circuit 5CA and the gate circuit 101 is not shown.

Furthermore, four data line pairs I/O to I/O
Detection circuits 6C to 6C are connected to the circuit 201, and a circuit 201 for selecting the outputs of these circuits is provided.
, a circuit 5ZS for controlling this, a circuit 7CS for amplifying the output 202 of the circuit 201, and a buffer 4 for giving the address to be selected to the circuit 5ZS.
The memory of FIG. 3 differs from that of FIG. 1 in that it is provided with C'S, a circuit 2C' that generates a pulse to start it, a circuit 203 that selects write data, and a write data buffer 10CS.

Incidentally, the buffer 4C'S, the selection circuit 5ZS, and the output amplification circuit 7CS are constituted by static type circuits, and the respective circuit configurations are shown in FIGS. 4A to 4C. Buffer 10CS is also a static type. Circuits other than these are dynamic type.

Furthermore, since the output amplification type circuit 7C in FIG. ). In FIG. 3, since the output amplifier circuit 7CS is a static type, there is no need to supply this signal from the circuit 2CA to the circuit 7CS.
Circuit 2C in FIG. 1 in that it does not have this supply circuit.
The only difference is that

The circuit 2C' is a circuit that outputs an inverted signal 12C' of the signal 1C every time the level of the signal 1C is inverted.

In FIG. 3, the detection circuit 6R of FIG. 1 is used as is, but for the sake of simplicity, it is assumed that it is not shown and is included in the memory cell array section 100.

The operation of the embodiment will be described below with reference to FIG.

The row selection operation in response to signal 1R is performed in exactly the same manner as in FIG. 1, based on the row address.
Then, in response to signal 1C, the column address C1
A column selection operation is performed based on .

Column address C1 almost synchronously with the falling edge of signal 1C or before the falling edge of signal 1C.
is input to line 3, which is input to buffer 4CA. Buffer 4CA takes in the upper bits other than the lower 2 bits of address C1 at the rising edge of signal 12C generated in response to signal 1C, and internal address signals 14CA, 1
4CA, and then the output does not change even if the address on line 3 changes until signal 1C goes high and buffer 4CA is precharged.

Bit line selection circuit 5CA responds to internal address signals 14CA, 14 to select block 1.
One bit line pair for every 00~, e.g.
The gate circuit 101 is controlled to simultaneously select B1, B1, B1, and B1, and a signal is sent to the data line pairs I/O to I/O. These signals are differentially amplified by detection circuits 6C to 6C, respectively, and MOS transistors Q ₅ to
It is supplied to a selection circuit 201 consisting of _Q8 . 2
C' is a circuit that generates a plurality of timing pulses in response to signal 1C for operation according to the present invention (hereinafter referred to as continuous mode operation). In the figure, the representative signal 1C is the output.
Only the inverted signal 12C' of is shown, and the others are omitted. Address buffer 4C'S outputs internal address signal 14C' and its inverted signal 14' in response to the least significant two bits of column address C1 input via line 3 when signal 12C' is at high level. The output circuit constitutes a static type circuit.

FIG. 4A shows an example of a part related to 1 bit of address in address buffer 4C'S, where Q ₁₁ ,
Q ₁₄ is driven by MOS transistor, Q ₁₂ and Q ₁₃ are loaded
It is a two-stage inverter circuit using MOS transistors. Signal 14C' is a 1-bit non-inverted signal of the address input to line 3, and 14C' is an inverted signal of this address. Here, the gates of the load transistors Q ₁₂ and Q ₁₃ are controlled by the signal 12C' when the signal 1C is not input, that is, in the standby state.
This is to turn off the MOS transistor and reduce power consumption.

The portion of the buffer 4C' relating to the other bit of the column address is constructed in exactly the same manner. Although buffer 4C'S is a static type circuit, it starts operating from the moment signal 12C' becomes high potential, so the lower two bits of the first column address C1 are taken in in synchronization with signal 12C'. It is done. When the signal 12C' is held at a high potential, the outputs 14C' and 14' change after a circuit-specific delay time (1 to several nanoseconds) in response to a change in the address input from the line 3.

The decoder 5ZS selects one of the lines Z to Z according to the output of the buffer 4C'S. Here, a case is illustrated in which Z is selected according to address C1.

Figure 4B shows the output line Z of the decoder 5ZS.
shows the part that selects the transistor Q ₁₂ ,
A NOR circuit is configured for the lower two bits of the column address input to the gate of _Q16 .
When both inputs are in a low potential state, a high potential is output to the output Z via the load transistor _Q17 . Since this circuit is also of a static type, when the signal 12C' is at a high level, the output changes immediately in response to a change in the level of the input address.

A portion of the decoder 5ZS that selects the output line Z is similarly configured. The reason why the gate of load transistor _Q17 is controlled by signal 12C' in FIG. 4B is the same as in FIG. 4A.

When the signals of the data line pairs I/O to I/O are differentially amplified by the detection circuits 6C to 6C, the decoder 5ZS has already selected the line Z corresponding to the column address C1,
The output of circuit 6C is selected by MOS transistor _Q5 and supplied via line 202 to output amplification circuit 7CS.

As shown in Figure 4C, the output amplifier circuit 7CS
consists of an inverter circuit consisting of MOS transistors Q ₁₈ and Q ₁₉ and a push-pull circuit consisting of Q ₂₀ and Q ₂₁ . This circuit is also static type, and after a delay time inherent in the circuit, the line 202
Output the above signal to terminal 8. The signal 12C' is applied to the fourth load transistor _Q19 .
This is for the same reason as in Figure A.

In this way, as in the conventional case, after the signal 1R or 1C becomes low level, t _RA , respectively.
After the t _CA time has elapsed, the first data corresponding to address C1 is output to terminal 8.

After that, the signals 1R and 1C are maintained at a low potential, and the original operating state is maintained. Therefore,
Four data read from four blocks of the memory are held in the data line pairs I/O to I/O, and the detection circuits 6C to 6C also output signals obtained by amplifying these four data.

At the timing when the operation of the output amplifier circuit 7CS is completed and the data is output, the next column address is
C2 is input via line 3. This column address C2 differs from column address C1 only in its lower two bits. In response to the lower two bits of address C2, output 14C' of circuit 4C'S,
14C' changes and the address is set by circuit 5ZS.
The output line corresponding to the lower 2 bits of C2, for example Z
is selected. This results in transistor Q ₆
is turned on, and the contents of the detection circuit 6C are output as data to the terminal 8 through the output amplifier circuit 7CS. From now on, input column addresses C3 and C4 every time the operation of output amplifier circuit 7CS is completed.
By repeating the same operation, the corresponding data,
are output sequentially. During this time, the signal 12C remains at high level, so the upper bits of addresses C2 to C4 are not taken in by buffer 4CA, and its outputs 14CA and 14 are output from addresses C2 to C4.
remains as for C1. Therefore, this shows that the upper bits of addresses C2 to C4 do not need to be input from line 3. Therefore, in Figure 5, the address
The signal on line 3 is illustrated assuming that the upper bits of C2 to C4 are not input.

After this continuous mode ends, signals 1C and 1R are returned to high level and the memory returns to the standby state.
In other words, static type circuits 4C'S, 5ZS, 7
The CSs are put into a standby state by their input signal 12C' going low, and each of the other dynamic circuits of the memory are precharged by a signal supplied from either circuit 2R or 2CA.

According to the embodiment described above, the access time in continuous mode, that is, after the second and subsequent column addresses C2 to C4 are input, data to
The time t _ZSA until output is the circuit 4C′S,
Because it is determined by the operating speed of a few circuits, 5ZS and 7CS, and unlike dynamic type circuits, these circuits are static type circuits that do not require precharge, so compared to the page mode access time t _CA of conventional memory, 1/2～1/
5, which enables high-speed continuous operation. In addition, this high-speed operation cycle time t _ZSC
The access time t is also almost the same as _ZSA , and is reduced to 1/2 to 1/5 of the conventional access time.

The read operation has been described above, but the write operation is also performed from the data input 9 to the data input buffer 10CS, as shown in FIG.
data line pairs I/O to I/
A pair of differential write data is continuously supplied to O, and high-speed continuous writing is performed.

(2) Combination of continuous mode and page mode In the above embodiment, when reading/writing four or more different data, after handling the four data in continuous mode, as shown in FIG. Then, it is necessary to return the signals 1R and 1C to high potential, return all circuits to the standby state, and restart continuous mode operation. Therefore, the continuous mode is only executed intermittently, leaving room for further improvements in speed when reading large amounts of data. Several embodiments capable of continuous mode operation with multiple data will be described below. FIG. 6 shows an embodiment of a memory operating in a combined continuous mode and paged mode, and the same reference numerals in FIG. 6 as in FIG. 3 refer to the same elements as in FIG. Note that in FIG. 6, parts related to data writing are not shown for the sake of simplicity.

Figure 6 differs from Figure 3 mainly in the following points.

Disconnecting MOS transistors _Q27 to _Q34 and dynamic type latch circuits 6C'' to 6C'' for temporarily storing data are provided, and circuit 2C'A is used in place of circuit 2C' in the circuit shown in FIG. .

Various configurations are possible for the latch circuits 6C'' to 6C'', one example of which will be explained later in FIG. 13.

Circuit 2C'A outputs its inverted signal 12C' in response to the first falling edge of signal 1C.
It is the same as circuit 2C' in FIG. 3, but differs from circuit 2C' in FIG. 3 in that it does not respond to subsequent changes in the level of signal 1C while signal 11R is at a high level. Furthermore, circuit 2C'A differs from circuit 2C' in FIG. 3 in that it outputs signal 15C which becomes high level after a predetermined period has elapsed after the fall of signal 1C.

Transistors Q ₂₇ ~ _{Q 34} are the detection circuit 6C ~
The detection data of 6C is the latch circuit 6C''~6
After being taken in by C'', it is turned off under the control of signal 15C, and has the function of separating latch circuits 6C'' to 6C'' from detection circuits 6C to 6C.

The operation of the memory shown in FIG. 6 will be explained with reference to FIG.

The operation until data is read out to the data line pair I/O to I/O by the first address C1 of the first set of addresses is the same as in the embodiment shown in FIG.

The detection circuits 6C to 6C differentially amplify the voltages of the corresponding data line pairs I/O to, and output a pair of signals of different levels as detection data according to the amplification results.

At the time when the operational amplification operation by the detection circuits 6C to 6C is completed, the signal 15C becomes a high potential state, and the latch circuits 6C'' to 6C'' are connected to the detection circuits 6C to 6C via the transistors _Q27 to _Q34 .
are latched in a state corresponding to a pair of signals output from each. Latch circuit 6C''~6C
One of the outputs of ``, for example 6C'' is selected by a selection circuit responsive to address C1,
It is output as data from the output amplifier circuit 7CS. After this, by sequentially changing the column address input via line 3 to addresses C2 to C4, it becomes possible to output data ~ in continuous mode based on the outputs of latch circuits 6C'' to 6C''. .

In this embodiment, in order to start the page mode operation in parallel with the continuous mode operation, the signal 1C is set to a high potential state after the latch operation to the latch circuits 6C'' to 6C'' is completed. As a result,
By returning the signal 15C to the original low potential by the circuit 2C'A and turning off the transistors _Q27 to _Q34 , the latch circuits 6C'' to 6C'' are disconnected from the detection circuits 6C to 6C. As in the case of the mode, in response to the high level of the signal 1C, the circuit 2CA selects the circuits related to the column selection operation, that is, the buffer 4CA, the bit line selection circuit 5CA, and the data line pair I/O.
~, returns the detection circuits 6C to 6C to the standby state of the memory.

When the return operation of this column selection circuit is started, address C2 is selected regardless of this return operation.
In order to perform continuous mode operation for , the next column address C2 is input via the line 3 at the timing when the output amplifier circuit 7CS outputs data. However, it is only necessary to input the lower two bits of address C2. This is because the buffer 4CA is in a standby state in response to the rise of the signal 1C, and is therefore in a state in which it does not respond to the address on the line 3. Therefore, there is no need to input the upper bits of address C2. This will be explained later in the subsequent example addresses C3 , C4
The same is true when inputting , and as a result, the address in the upper bits of each address
This means that only the upper bits of C1 need to be input. Also, in order to enable continuous mode operation even while signal 1C is at a high level, circuit 2
C'A is the signal 1C while the signal 11R is low level.
The signal 12C' is held at a high level even when the signal 12C' returns to a high level. In this way, continuous mode operation is performed based on the lower two bits of address C2 in parallel with the return operation of the column selection circuit, and data is read out. Assuming that the circuit recovery operation related to the column selection operation is completed at the time when the output amplifier circuit 7CS outputs data and starts continuous mode operation based on address C3, in order to immediately start the next column selection operation from this point. From this point on, it is desired to set the signal 1C to a low potential again and then start capturing the first address C1' of the second set of addresses C1', C2', . To do this, it is necessary to simultaneously import address C1' and address C3 via line 3. Since only the lower two bits of the column address need to be used for continuous mode operation, the lower two bits of address C3 are
The bits are sent from the outside, and the upper bits other than the lower 2 bits of address C1' are input via the remaining line 3.

When the continuous operation using the lower two bits of address C3 is completed, only the lower two bits of address C4 are similarly input via line 3. During this period, the memory performs a column selection operation in response to the upper bits of address C1' from when the signal 12C returns to high level, the voltage of the input/output data line pair I/O~ changes, and the detection circuit 6C~ works. Assuming that the output amplifier circuit 7CS completes outputting the data for addresses C3 and C4 after returning the signal IC to a low level and before the operation of the detection circuit 6C~ is completed, the signal will be output again when the operation of the detection circuit 6C~ is completed. 1C is set to high level, and signal 15C is set to high level for a certain period of time. As a result, data on I/O lines I/O~ read out based on the upper bits of address C1' is taken into latch circuits 6C''~6C''. In this way, continuous operation using addresses C1', C2', . . . is started, and data 1', 2', . . . are read out to terminal 8.

When the continuous operation using address C2′ ends and the next continuous operation starts using address C3′,
In order to perform continuous operation for the third set of addresses C1''..., the signal 1C is made low and the upper line of line 3 is connected to the upper bit of the first address C1'' of this third set of addresses. is input.

The same operation is repeated thereafter.

In this way, data is read out continuously in a mode that is a combination of continuous mode and page mode. When reading of this data is completed, both signals 1C and 1R are set to high level, and all circuits in the memory are returned to a standby state.

As mentioned above, in this embodiment, the signal 1C
Each time the detection circuit 6C becomes a low level, the detection circuit 6C~ performs the operation until the operation is completed, and each time four pieces of data are generated by the above operation, these are successively outputted to the terminal 8. In this way, data can be extracted continuously from the terminal 8 without interruption.

Although the above operation is only for reading, it goes without saying that writing can be performed in the same way. In the case of writing, it would be undesirable if the address of the bit to be written changes during operation, so the write address may be input in the next page mode cycle.

Note that in this embodiment, the first address of the second set of addresses is taken in the upper bits of C1' when the lower two bits of address C3 are taken in, but this may depend on the operating speed of the memory or the design. This may vary depending on the situation and is not limited to this example. Further, it goes without saying that the number k of consecutive reads is not limited to four, and can be changed in various ways. In addition, if a relationship of t _CC k・t _ZSC is established between the cycle time t _CC in page mode and the cycle time t _ZSC in continuous operation, k
More than one piece of data can be retrieved consecutively. In addition,
Even if t _CC >k·t _ZSC , there will only be a slight time difference, and this will not impair the effectiveness of this embodiment.

According to this embodiment, the amount of data that can be continuously read/written at high speed is j×k, where j is the number of page modes, which is significantly larger than in the previously described embodiments. That is, this embodiment enables continuous read/write at 1/2 to 1/5 of the operating format of the conventional page mode.

The operation in the combination mode of the continuous mode and the page mode described above can also be realized in a memory configured in a dynamic type circuit.

Before explaining this embodiment, an outline of a memory that consists only of dynamic type circuits and operates only in a continuous mode like the memory shown in FIG. 3 will be explained.

FIG. 8 shows a buffer 4C' and a selection circuit 5 in which the buffer 4C'S, the selection circuit 5ZS, and the output amplification circuit 7CS in FIG. 3 are of dynamic type, respectively.
It differs from the memory shown in Figure 3 mainly in that it is replaced by an output amplifier circuit 7C, and that the pulse generation circuits 2CA and 2C' in Figure 3 are replaced with pulse generation circuits 2CD and 2C'B, respectively. .
Note that the input data buffer 10CS shown in FIG. 3 is also replaced with a dynamic type circuit.
In FIG. 8, parts related to data writing are not shown for the sake of simplicity.

Circuit 2CD outputs signal 1 only when signal 1R is at low level, i.e. signal 11R is at high level.
In response to the falling edge of C, the inverted signal 1 of signal 1C
It is the same as the circuit 2CA in Fig. 3 in that it outputs 2C, but in response to the rise of signal 1C, it inverts the level of signal 12 and precharges the circuits related to the column selection operation. This circuit differs from the circuit 2CA in FIG. 3 in that the operation of generating the signal 11R is performed only when the signal 11R is at a low level. Circuit 2C'B is the same as circuit 2C' in Figure 3 in that it changes the level of inverted output 12C' every time the level of signal 1C is inverted, but every time signal 1C rises, buffer 4C' is selected. The main difference from the circuit 2C' of FIG. 3 is that the circuit 5Z and the output amplifier circuit 7C generate a precharge signal for putting the output amplifier circuit 7C into a standby state.

As can be seen from the time chart in FIG. 9, the operation of the memory shown in FIG. 8 includes the column selection operation using the column address and the row selection operation using the row address C1. It is exactly the same as the memory in Figure 3.

In this memory, the signal 1C is raised when the output amplifier circuit 7C outputs data.
Along with this, circuit 2C'B provides buffer 4
C', the selection circuit 5Z, and the output amplifier circuit are precharged and returned to the standby state. At this time, signal 11
Since R is at a high level, circuit 2CD receives signal 1
There is no response to the rise of C. Therefore,
Buffer 4CA, bit line selection circuit 5C, and detection circuit 6C are not precharged and maintain their previous states.

When the first data is output from the output amplifier circuit 7C, the signal 1C rises, and in response to this rise, the circuit 2C'B precharges the circuits 4C', 5Z, and 7C related to the continuous mode and goes into standby. It generates a signal (the signal line for this purpose is not shown) and sets the signal 12C' to a low level. Before these circuits return to the standby state, the lower two bits of the next column address C2 are input via line 3. At this time, even if the upper bits of address C2 are input through line 3, circuit 4CA is in a state where it does not respond to this, so there is no point in inputting these upper bits through line 3. it is obvious. When the return to the above standby state is completed, the signal 1C is set to a low level. In response, circuit 2
C'B stops sending out the precharge signal and makes signal 12C' high level. Batsuhua 4C is
At the rising edge of signal 12C', the lower two bits of address C2 are taken in and corresponding internal address signals 14C' and 14' are output. After this,
As in the case of address C1, the output of the detection circuit 6C is selected by the selection circuit 5Z, and data is output from the output amplifier circuit 7C. Thereafter, the lower two bits of addresses C3 and C4 are sequentially input in the same manner, and the data is sequentially input.
is output. Thereafter, both signals 1C and 1R are set to high level, and circuit 2R precharges circuit 5R, cell array 100, etc. related to row selection in response to the rise of signal 1R, and returns to a standby state. At this time, the signal 11R becomes low level, and the circuit 2CD responds to the low level of the signal 11R and the high level of the signal 1C, and the circuit 4 related to column selection
A signal is generated to precharge CA, 5C, 6C~, and further the signal 12C is set to a low level.

In this way, data can be read out in continuous mode using a dynamic type circuit. It is different from the continuous mode memory using a static type circuit shown in Figure 3 in that the circuit related to the continuous mode is precharged and returned to the standby state each time one piece of data is output. good. Therefore, a memory that operates in a combination mode of continuous mode and page mode can also be easily constructed based on FIG. 6, as shown in FIG.

The memory in FIG. 10 uses the dynamic type buffer 4C', selection circuit 5Z, and output amplification circuit 7C used in FIG.
The pulse generating circuit 2CA used in FIG. 6 and the circuit 2CE whose details are shown in FIG. 11 are used instead of the pulse generating circuit 2CD shown in the figure.
In place of the pulse generating circuit 2C'A shown in the figure, a pulse generating circuit 2C'D which generates the signal 12C' etc. in response to the signal 1C and a circuit 2C'D which generates the signal 15C in response to the output of the circuit 2CE are used. It differs from the memory shown in FIG. 6 only in that the memory shown in FIG.

Circuit 2CE is a circuit that outputs signal 1C' in response to signal 1C when signal 1R is at a low level, and signal 1C' falls in response to the first falling edge of signal 1C as shown in Figure 13. (period), the signal 1C including this first falling edge is 4
It falls once within the period of falling (period~). Note that, as will be described later, the signal 1C' only needs to fall a quarter of the total number of falls of the signal 1C, and falling during the period V is not necessarily necessary. Here, 4 or 1/4 respectively represents the number k of data read in continuous mode or its reciprocal.

In Figure 11, 202 and 203 are 1/4 of the signal 1C.
(i.e., 1/k), and an example using a well-known JK type flip-flop is shown here. Of course, it is also possible to use other types of flip-flops, such as a D-type flip-flop. In addition, the above JK
The flip-flop used is one whose state is reversed at the falling edge of 1C input as the clock pulse Cp (Negative Edge Trigger).
Type). 204 is an SR that recognizes when the operation starts
This is a type flip-flop, and the state is inverted at the rising edge of the S and R input signals. Note that a function for resetting (or resetting) each flip-flop to its initial state is omitted so that even if memory operation is interrupted midway, normal operation will resume in the next cycle. The same applies to the following examples. 205 is an inverter, 206~
209 is an AND circuit, and 210 is an OR circuit.

Non-inverting output 218 of flip-flop 204
rises when both signals 1R and 1C become high potential (at the end of operation), starts operation (1R is low potential) and falls when signal 1C rises for the first time. The gates 208 to 210 perform logic operations on the inverted output 212 of the flip-flop 202, the non-inverted output 213 of the flip-flop 203, the signal 218, and the inverted output 219 of the flip-flop 204.
form C′.

As a result, as shown in FIG.
the first low level period and signal 1C is 3
The signal 1C' becomes low level during the period from when it becomes low level at the +4αth time (α=0, 1, 2, . . . ) until the signal 1C starts to go low level at the 5+4αth time.

This signal 1C' is input to circuit 2CA. In Fig. 6, the signal 1C was input to the circuit 2CA, but in Fig. 10, this signal is changed to the signal 1C.
C' is input to circuit 2CA.

Circuit 2C'D differs from circuit 2C'B in FIG. 8 only in that it does not have a circuit part that generates signal 15C, and circuit 2C'E differs from circuit 2CA in FIG.
The circuit 2CE is connected to the circuit 2CE so as to generate the signal 15C in response to the output 1C' of the circuit 2CE.

Now, FIG. 13 is one example of the configuration of the latch circuit 6C'', and it goes without saying that the other latch circuits 6C'' to 6C'' are similarly configured. As in the 6th
It can also be applied to memory of figures. As shown in FIG. 13, transistors Q _L1 , Q _L2 and capacitors C _L1 ,
Consists of C _L2 . Here, when the signal 15C becomes a high potential, transistors Q ₂₈ and Q ₂₉ turn on, and the output signal 61 is transmitted to the node, and when the signal 15C becomes a low potential, the transistors Q 28 and Q 29 turn on.
Q ₂₈ and Q ₂₉ are turned off, and the above signals are confined in the nodes and held in the form of charges in the capacitors _CL1 and _CL2 , respectively. That is, 6
Latch the output signal of 1. At this time,,
Each signal of is an inverted signal of the other, and according to the signal of , the transistor
When either Q _L1 or Q _L2 is turned on and is at a high potential (i.e., is a low potential), the transistor Q _L1 is on and Q _L2 is off, and the high potential is applied to 201, while is is at a low potential (i.e., is high potential), transistor Q _L2 is off, Q _L2
is turned on, and a low potential is output to 201.

As explained above, the transistors Q _L1 ,
Q _L2 is never turned on at the same time, so there is no unnecessary power consumption. Also, signal 1
The latched signal is changed only by 5C, and no special signal is required to return the circuit to the standby state. In addition, in order to operate this latch circuit normally, 6C~
Of course, the 6C circuit needs to have the driving ability necessary for charging and discharging the capacitors C _L1 and C _L2 .

The operation of the memory shown in FIG. 10 will be explained with reference to FIG. 14.

When signal 1R is at low level, signal 1C
When the signal becomes low level for the first time, the signal 1C' becomes low level in synchronization with it. In response to the first falling edge of signal 1C', a column selection operation is performed in exactly the same manner as in the case of FIG.
The data detected in "~6C" is set. On the other hand, in response to the rise and fall of signal 1C, circuit 2C'D interrupts precharging of buffer 4C', selection circuit 5Z, and output amplifier circuit 7C, and outputs signal 12.
Raise C′ to a high level. In response to the rise of signal 12C', continuous mode operation based on the lower two bits of addresses C1 to C4 is started in exactly the same way as in the case of FIG. 8, and data is transferred to terminal 8.
Read from. In this case, MOS transistor
In order to turn on Q ₂₇ and Q ₃₄ when data is read to the data line pair I/O~, the signal 15 is turned on.
A circuit 2C'E is provided for setting C to a high level in synchronization with the start of the column selection operation. To run page mode in parallel with this continuous mode,
When the signal 1C first rises, the signal 1C' rises, and in response, the circuit 2CA performs the circuits 4CA, 5CA, 6C to 6C related to the column selection operation.
Outputs a signal to return to pre-charge standby state.

During this return-to-standby operation, signal 1C is repeatedly changed and continuous mode operation based on addresses C2-C4 continues in the same manner as in FIG. Here column address C3
It is assumed that the above-mentioned return operation is completed before starting the continuous mode operation based on . When inputting the lower two bits of address C3 via line 3,
The upper bit of the first address C1' of the next set of four addresses C1' to C4' is input via the upper line of line 3. Thereafter, when the signal 1C rises, the column selection operation using this address C1' is started. At this time, address C3
Continuous mode operation is carried out in parallel. Thereafter, the page mode and continuous mode are executed in parallel as in the case of FIG. In the case of Fig. 10, since circuits 4C', 5Z, and 7C related to continuous mode operation are dynamic type circuits, each time continuous mode operation is completed in response to the lower 7th two bits of each column address C1 to C4. The operation of the memory in FIG. 10 differs from that in FIG. 6 in that it is necessary to precharge these circuits to a standby state by circuit 2C'D as in FIG. 8.

Therefore, the memory shown in Fig. 10 has a slower operating speed than the memory shown in Fig. 6 due to the time required for this precharge operation, but since all the circuits are dynamic type, it consumes less power than the memory shown in Fig. 6. It can be done. This also applies to the comparison of the memories in FIGS. 3 and 6.

(3) Row continuous mode With the above embodiment, as mentioned earlier, it has become possible to handle j×k data continuously at high speed, but this amount of data is limited to one row selection address. limited to the range specified. The next embodiment expands the above concept, that is, operates other circuits during continuous operation, and simply operates continuously without interruption, to perform both row selection and column selection operations, and all data in the memory is I will explain what can be read out continuously at high speed. FIG. 15 shows an example of this, and, like the example of FIG. 10, it is an example of a memory configured with a dynamic type circuit.

In the same figure, a circuit 2CF whose details are shown in FIG. 16 is used in place of the pulse generating circuit 2CE shown in FIG. This memory differs from the memory shown in FIG. 10 in that it is input instead of the signals 1C' and 1R shown in FIG.

Circuit 2CF is a circuit that outputs signals 1R' and 1C'' in response to signal 1C when signal 1R is at a low level, and signal 1R' is the first rising edge of signal 1R, as shown in FIG. In response to the falling edge, the signal 1C falls (periods I to R), and falls once within a period in which the signal 1C falls four times (periods -R to -R), including this first falling edge.

Similarly to 1R', signal 1C'' also falls in response to the first fall of 1C (period -C), and once within the period in which signal 1C falls four times, including this first fall. As is clear from FIG. 17, the period during which the signal 1R' falls (periods -C to -R) is at a low level is two cycles of 1C, whereas the signal 1C'' is different in that it corresponds to one cycle of signal 1C.
Note that this time relationship is based on high-speed continuous operation ( C1
˜C4) is an example where k=4, and it goes without saying that it can be changed as appropriate depending on the number of k.

Furthermore, the signals 1R' and 1C'' only need to fall 1/4 of the total number of falling circuits of the signal 1C, and the falling periods -R and C are not necessarily required. /4 represents k or 1/k, respectively.

In FIG. 16, the same parts as in the 2CE circuit shown in FIG. 11 are indicated by the same numbers, and the difference is that the AND circuit 222 and the OR circuit 210 are replaced with a three-input OR circuit 210'.

Flip-flops 202-204 perform the same operations as previously described, and gates 208-210', 222, 2
24 performs a logical operation, and the already explained signal 1
R′, 1C″ is formed. As a result, 1R′ is the first
As shown in Figure 7, the period -R from when the signal 1R goes low until the first rise of the signal 1C, and when the signal 1C rises for the (3+4α)th time (α=0, 1, 2-). After reaching a low level,
1C becomes low level during the period -R to -R until it starts to become low level for the (5+4α)th time. Also, the signal 1C'' has an initial low level period of -C and 1C of 4+
After becoming low level for the 4α time, 1C is 5+
The level becomes low during the period -C to -C until it starts to become low level at the 4αth time.

The signal 1R' formed by the above is the circuit 2
R, the signal 1C'' is input to the circuit 2CA. That is, in FIG. 10, the signal 1R is input to the circuit 2R,
While signal 1C′ is input to circuit 2CA,
In Fig. 15, signal 1R' is connected to circuit 2R;
C'' is input to circuit 2CA.

Figure 18 shows detailed operation waveforms of this memory. In contrast to the memory in Figure 10, where continuous mode and page mode were operated in parallel, this memory operates in continuous mode and normal row mode. This memory differs from the memory shown in FIG. 10 in that the column selection operations are performed in parallel and continuously. transistor
Up to the circuits Q ₂₇ to _{Q 34} , a normal row and column selection memory operation is performed, and thereafter a continuous mode is performed in parallel and continuously.

When signal 1R goes low, 1R' goes low, and in response, a row selection operation based on the address input is performed in the same manner as in FIG. Next, when 1C becomes low level, 1C'' becomes low level, and the address is changed as in Figure 10.
A column selection operation is performed based on the upper bit of C1, and the detected data is set in latch circuits 6C''-6C''. After that, when 1C rises for the first time, 1R' and 1C'' rise,
The circuits 2R and 2CA that responded to this cause the next row,
In order to prepare for the column selection operation, the circuits related to these operations are made similar to those shown in FIG. 10, and the return operation to the standby state is executed. On the other hand, in response to signal 1C, continuous mode operation based on the lower two bits of C1 to C4 is performed in the same manner as in FIG.

Here, it is assumed that, as in FIG. 10, the recovery operation of the circuits related to the row and column selection operations described above is completed before the continuous mode operation based on C3 is started.
When address C3 is input, 1R′ becomes 1
In response to C falling, a row select operation is initiated based on the row select address ' of the next set of four addresses input on line 3 at the same time as C3. At this time, continuous mode operation by C3 is performed in parallel. Then the address
When C4 is input, 1C'' responds to 1C and, in the same way as the falling edge, column selection operation based on column selection address ' is started. At this time, continuous mode operation by C4 is performed in parallel. When the row and column selection operations based on ' and ' are completed in this way, the data detected in 6C'' to 6C'' are set in the same way as before.

Thereafter, similarly, row and column selection operations and continuous mode operations are performed in parallel.

Now, in this embodiment, if the row address is input at a phase such as C3, it becomes impossible to input the address necessary for continuous operation to input at C3. There is no problem if you configure less memory. In addition, if it is not necessary to have the same number of both, when inputting C2, input C3 at once using the upper bit of input line 3. That is, in the embodiments described so far, the addresses of continuous operations were input sequentially, but in this method they are input all at once.

Furthermore, regarding the method of addressing data, in the embodiments described above, the data to be retrieved consecutively has a common row address and only differs in column address. However, this is the essential point of the present invention. For example, the column addresses are common and only the row addresses differ, so
It goes without saying that the addresses C1 to C4 can be changed in any of the embodiments, such as by inputting them as row addresses or by using a mixture of row and column addresses.

As described here, according to the embodiment, continuous operation is possible without any limit on the number of data (within the total capacity of the memory). This allows the memory to be used as if it were a high-speed shift register. Although we have described the dynamic type method here, it goes without saying that based on the same idea, it is possible to combine not only page mode but also normal memory operation and continuous operation in the static type as explained in Figure 6. .

(4) Modification In the continuous mode in the above embodiment, the order in which the four data are read can be specified randomly using the lower two bits of addresses C1 to C4, but it is also possible to configure this order to be fixed in advance. It is.

For this purpose, for example, the selection circuit 5 in FIG.
It is sufficient to use a decoder 5ZA configured to select output lines in order from output line Z every time the input signal 12C' becomes high level. The selection circuit 5ZA may be, for example, a four-stage shift register in which a pulse for instructing selection is sequentially transferred each time the signal 12C' is input, and each stage is directly connected to the line Z~. , or divide the signal 12C' and connect it to the line Z
There is a flip-flop circuit that outputs a signal for sequentially selecting .about.Z. As another example of reading out four pieces of data in a fixed order in continuous mode as shown in FIG. Therefore, using a four-stage shift register SR that performs a shift operation,
The output may be connected to the output amplification circuit 7C. With this arrangement, data is transferred to the output amplifier circuit 7C in a fixed order every time the signal 12C' is generated, and data can be continuously taken out from the output terminal 8.

In the above example, the order of the four data handled in continuous mode is fixed, so the lower two bits of the example address required to specify this order in Figure 8 etc. are no longer necessary, and the memory input/output terminal (package pin count). Note that in order to specify the first data of the four data to be read in continuous mode, only the lower two bits of the column address of the first data are input, and after that, the three data following this first data are fixed. Decoder 5Z and selection circuit 20 so as to read in order
1 can also be configured. For example, the 19th
The shift register SR shown in the figure may be configured to circulate periodically so that the data is output from a designated portion of the first data.

Since the data order to be outputted in these modified sequences is fixed in advance, higher speed operation is possible than in the embodiment shown in FIG. 8 described above.

Now, in each of the above embodiments, the read and write operations were performed individually, but by simple improvement, operations consisting of various combinations of read and write operations can be made possible. For example, it is possible to perform image operations simultaneously, or to write only to some addresses during continuous operations. These will be explained below based on examples.

In FIG. 20, a signal 1W is an external control clock for controlling reading/writing, and here reading is performed in a high potential state and writing is performed in a low potential state. 2
Similar to the pulse generation circuits 2R and 2C (for example, see FIG. 1), W is a circuit that generates multiple timing pulses necessary for internal memory operations, and is mainly used for controlling read/write operations. supplied to Here, the signal 12W supplied to buffers G to G, which will be described next.
is shown as a representative example. Batsuhua G～G
takes the logical product of the signal 12W and the outputs Z~Z of the decoder 5Z (FIG. 8) described above, and selects the selection MOS transistors _Q23 ~ of the selection circuit 203.
By the control of this circuit, the input data coming from the input terminal 9 through the buffer 10C is connected to the common input/output data line pair I/O by the AND circuit that controls _Q26.
supplied to one of the In addition, in the figure, for the sake of simplicity, signals such as the common input/output data line pair I/O~ and the write data line 204 are shown as one line, and each data line pair I/O is shown as one line.
The selection MOS transistor for ~ is also Q ₂₃ ~
Only one item is shown as shown in Q ₂₆ .

Referring to the operating waveforms in Figure 21, C1 in the figure
〜C4 During the low level period of signal 1C, the column address C is changed as shown in FIG.
1 to C4 are input. When the signal 1W is at a low potential, the circuit 2W generates an inverted signal 12W of the signal 1C in synchronization with the fall of the signal 1C.
Now, signal 12W and signals Z~Z are ANDed by AND circuits G~G, and signal 12W is
When W occurs, at that point Batsuhua 4C (8th
One of the lines Z' to Z' is selected corresponding to the lower two bits of the column address input in FIG. The write data is transferred to one of the write data, and the voltage of the data line pair is changed depending on the write data. Thereafter, data is written into the memory cell in the same manner as in the prior art based on the voltage of this data line pair.

When signal 1W is at high level, signal 12W is at low level and no writing is performed. Therefore, either writing or reading can be performed in continuous mode by simply changing the level of signal 1W.

For example, when signal 1W is held at a low level between the inputs of column addresses C1 to C4, as shown by the solid line in FIG. 21, a write is performed based on addresses C1 to C4;
As shown by the chain line in Fig. 21, if the signal 1W is set to low level only when addresses C1 and C3 are input, continuous mode is started in which reading based on addresses C2 and C4 based on addresses C1 to C3 is mixed. It will be held in

Furthermore, if the signal 1W is input with a certain time delay from the signal 1C, a so-called read-modify-write operation is also possible, in which data is read from a certain memory cell and then written to the same memory cell. . It is easy to understand that when this operation is possible, it means that read/write operations for each memory cell can be performed simultaneously.

In addition, in FIG. 20, when performing a write operation, it may be necessary to electrically disconnect between the data line pair I/O ~ and the detection circuit 6C depending on the circuit configuration of the memory. , the circuit 6C should have that function, or the second
For switches as shown by the broken line in Figure 0.
MOSTQ ₂₂ may be provided.

Furthermore, in the read-modify-write operation described above, there is also a method in which writing to four memory cells is performed simultaneously. FIG. 22 shows an example of this, in which each data line pair I/O~
The selection circuit 203 sequentially writes write data to the latch circuits (or flip-flops) 10C' to 10C' provided corresponding to the latch circuits 10C' to 10C', and after the writing to the latch circuits 10C' to 10C' is completed, these data are controlled by the signal 12W'. Write data is transferred in parallel to the common input/output data lines I/O~ to perform writing. Here signal 12
W' is generated by circuit 2W.

In Fig. 20, when reading and writing to address C1, it is necessary to perform the write operation after completing the read operation of the common input/output data line pair I/O~, so it may take some time depending on the memory design. Although it may be expected that the speed will be slower, there is no problem in this embodiment because writing is performed on the common input/output data line for which reading has already been completed.

Furthermore, in the above embodiment, the so-called folded bit line format in which the bit lines are folded over each other has been described, but in the so-called folded bit line format in which the bit lines are arranged to spread left and right across the detection amplifier circuit 6R.
It is also applicable to Open bit linear format memory. In addition, although the explanation here deals with data handled as continuous operation that has a fixed row address and only a different column address, it can also be applied to data that has a fixed column address and different row addresses, or a combination of both addresses. It is possible. Also, in FIG. 17, signal 1
Although a memory that uses C and 1R to perform continuous mode and page mode has been disclosed, if a certain rule is established in the method of supplying 1C, it is not necessary to use signal 1R. For example, if signal 1C is input only once, only the operation related to row address selection will be performed, and then a refresh operation specific to dynamic memory will be performed, and if signal 1C is input twice in succession, normal read/write operation will be performed. If a rule such as 1R is established, the signal 1R becomes unnecessary, which is effective in reducing the number of pins of the package that accommodates the memory chip.

Furthermore, although the case where the input/output terminals 8 and 9 are individually provided is described here, the present invention is also applicable to a memory commonly used for input/output of one terminal, and conversely, the case where the input/output terminals 8 and 9 are provided separately is also applicable. It goes without saying that the present invention can be similarly applied to a memory in which a plurality of each are prepared.

(5) Cell array arrangement In the embodiments described so far, the memory cell array is consolidated into one memory cell array. In concrete memories, the word line is divided into several parts to minimize the delay time of the word line, or the bit line is divided into several parts to reduce the parasitic capacitance of the bit line and increase the read signal from the memory cell. It becomes necessary to split the line. Therefore, an example of a memory in which the memory cell array is divided into several arrays will be described below. The following examples are shown in Figures 3, 6, 8, 10, and 13.
It is applicable to any of the embodiments described in the figures. Therefore, only the part related to the array arrangement will be explained below. Further, in the following, reference numbers with subscripts such as L, R, etc. refer to the same reference numbers as those without subscripts in the above embodiments.

In Figure 23, only the bit line is divided into two.
Arrays 100L and 100R each consist of 100L and 100R.
It is divided into 4 blocks of 100L or 100R to 100R.

8 input/output data line pairs I/O+ to I/O
Each of L, I/OR to I/OR is provided corresponding to one block. Detection circuits 6CL to 6C are connected to each input/output data line pair.
L and one of 6CR to 6CR are connected.

Word line selection circuits 5RL and 5RR are provided corresponding to each array, and select one word line of the corresponding array in response to a row address. In this way, one word line is selected in each of the left and right arrays 100L and 100R. The bit line pair selection circuit 5CA is provided between the two arrays, and controls the gate circuit 101L in response to the upper bits other than the lower two bits of the column address to select the array 10.
One bit line pair is selected from each block of 0L, and similarly, one bit line pair is selected from each block of array 100R.
One of the four bit line pairs corresponding to each of the four bit line pairs is selected from each block in array 100R. Eight outputs, including four outputs from the array with the word line selected in this way, are detected by the sensing circuits 6CR-6CR, 6CL-6.
It is amplified by CL. Of the eight outputs from the detection circuit 6CL, etc., the array 100L or 100
A selection circuit 300 selects four outputs corresponding to any one of R based on the lowest one bit of the row address, and inputs the selection circuit 201 for continuous mode. As shown in FIG. 6, in order to operate in both page mode and continuous mode, a latch 6C is connected between the two selection circuits 300 and 201.
What is necessary is to provide ``~6C'' and MOS transistors _Q27 ~ _Q34 .

This embodiment enables continuous mode operation when the data line is divided into two.

In FIG. 24, four input/output data line pairs I/O~ and detection circuits 6C to 6C are provided for the same two arrays as in FIG. 3.

For each block, an intermediate data line pair AL(R) ~
A switch circuit 301R consisting of MOS transistors Q ₃₅ to _{Q 42} and MOS transistors Q ₄₃ to connect each intermediate data line pair to the corresponding input/output data line pair I/O.
A switch circuit 301L consisting of _Q50 is provided.

Arrays 100L and 100R as in Figure 23
Intermediate data line pairs A to A are selected from the left array 100L or the right array 100R by a bit line selection circuit 5CA (not shown) provided between them.
Four data are read to L or AR to AR. Assuming that the row addresses of the word lines of arrays 100L and 100R are even and odd numbers, respectively,
The least significant bit of the row address and its inverted bit are applied to lines 302R and 302L, and one of selection circuits 301L and 301R is turned on. In this way, the four outputs from either one of the two arrays are connected to the four data line pairs I/O~
and is detected by four detection circuits 6C-.

According to this embodiment, the number of input/output data line pairs and detection circuits required for continuous mode operation is k, that is, four in this case, and the chip area does not increase. Furthermore, the connections between each bit line pair and the intermediate data line pairs AL(R) to AL(R) are the same and simple as in the prior art, and pattern design is also facilitated.

In Figure 25, both the word line and data line are divided into two.
In other words, it indicates a memory in which the array is divided into four,
The word lines of arrays 100L and 100R in FIG. 24 are connected to blocks 100L and 100L, respectively.
This corresponds to the case where it is divided between 100R and 100R.

Word line selection circuits 5RL and 5RR are provided between the divided word lines, and along with the division of the word lines, gate circuits 101L and 101R and switch circuits 301L and 201R are arranged vertically in two directions.
It is divided. The same applies to the bit line pair selection circuit not shown here.

In order to simplify the drawing, here, intermediate data line pair AL(R) to AL(R), input/output data line pair I/O, and other signals that form a set of two lines are also represented by one line. It shows.

FIG. 26 shows an example in which the word lines are divided into two and the data lines are divided into four, that is, the whole is divided into eight.

FIG. 26 shows the cell array 100 shown in FIG. 25 when both the word line and the data line are divided into two, and the cell array 100 having the same configuration.
0 is provided and the input/output data line pair I/O~ common to cell arrays 100, 100 is connected to a first portion provided between both cell arrays in a direction parallel to the word line, and to block 100R in cell array 100. 100R and between blocks 100L and 100L of cell array 100, a second line is provided in a direction parallel to the data line.
This part, this second part, and the selection circuit 301L
or 301R.

Selection circuits 301L and 301R in cell arrays 100 and 100 receive two of the row addresses from lines 302L and 302R, respectively.
Bits are given. For example, cell array 1
The lowest two bits of the row address of the left word line group, right word line group, right word line group and right word line group in cell array 100 are respectively
Assume 00, 10, 01, 11. cell array 100
Inner lines 302L, 302R, cell array 100
A high level signal is applied to the internal lines 302L and 302R when the externally applied addresses are 00, 10, 01, and 11, respectively.

Note that instead of the embodiment, the input/output data line pair I/
It is also possible to extend the second portion of O~ further to the right and provide the detection circuit 6C~ there. In this case, input/output data line pair I/
The above-mentioned first part of O~ is unnecessary. In addition, as a part provided in the cell arrays 100, 100 and extending in the word line direction as shown in FIG. 25, an input/output data line pair I /O~ can also be configured. In this case, it goes without saying that the first and second parts described above become unnecessary.

FIG. 27 shows the positions of selection circuits 301L and 301R in cell arrays 100 and 100 in the word line direction in blocks 100L and 100.
26th in that the positions in the data line direction are between the selection circuits 5RL and 5RR.
An example different from the figure is shown. This position has a relatively large amount of space due to the layout design,
The layout design of the selection circuits 301L and 301R becomes easier.

Application examples of the present invention in various memory cell array configurations have been described above. The method introduced here in which an intermediate input/output data line pair is provided between the bit line pair and the common input/output data line pair, and this is selected by a switch, contributes to reducing the parasitic capacitance of the input/output data line pair, and allows continuous mode operation. It is applicable not only to conventional memory but also to conventional memory.

FIG. 28 shows an example of this, in which all bit line pairs are B1-Bi, B1-Bi, B1
It is divided into four blocks each consisting of ~Bi, B1~Bi, and intermediate input/output data line pairs A~A correspond to each block.
are provided, and intermediate input/output data line pairs A to A
A selection circuit 30 consisting of transistors Q ₅₁ to _{Q 58} for connecting the input/output data line pair I/O to a common input/output data line pair I/O.
The main difference from FIG. 1 is that, as in FIG. 3, a bit line selection circuit 5CA is provided which responds to the upper bits of the column address except for the lower two bits. Bit line selection circuit 5CA is gate circuit 1
01 to select one bit line pair from each block and connect the selected bit line pair to one corresponding intermediate input/output data line pair. Selection circuit 3
Of the four pairs of transistors in 01, only one pair is
It is turned on by a circuit (not shown) responsive to the lower two bits of the column address. In this way, only one desired bit line is connected to the common data line pair I/O.

Now, the most dominant parasitic capacitance of the common input/output data line pair I/O is the gate circuit 101.
This is the depletion layer capacitance that occurs between the source or drain diffusion layer and the silicon substrate of the MOS transistor (see Figure 1), which is a component of the MOS transistor.

In this embodiment, all of the gate circuits 101
Only 1/4 of the MOS transistors are connected to the data line pair 301 at the same time. Therefore, the gate circuit 10
In this embodiment, the parasitic capacitance due to the MOS transistor in the MOS transistor 1 is reduced to 1/4 of that of the conventional one, so the parasitic capacitance is significantly reduced, and the operation related to the input/output data line pair and the I/O line can be accelerated. Become. As is clear from the above description, the layouts of FIGS. 24 to 27 can also be applied to a memory having a pair of common input/output data lines as shown in FIG. 28. In addition to the bit line pair to be selected, three bit line pairs are selected in FIG. 28, and these are connected to the three corresponding intermediate data line pairs.

Each of these three intermediate data line pairs is connected to a sense amplifier (not shown) provided for each bit line pair.
The operation of these sense amplifiers can be slow because they are driven solely by In order to avoid this, the bit line pair selection circuit 5CA selects bit line pairs B1 to B in response to all bits of the column address.
It is sufficient to use a circuit (ie, the same circuit as circuit 5C in FIG. 1) that controls the gate circuit so as to select only one of i and i.

〔Effect of the invention〕

As explained above, according to the present invention, it is possible to provide a memory that reads data at high speed.

[Brief explanation of drawings]

Fig. 1 is a schematic circuit diagram of a dynamic memory using conventional MOS transistors, Fig. 2 is a time chart showing the operation of the memory shown in Fig. 1, and Fig. 3 is a partial diagram of a static type circuit according to the present invention. 4A is a circuit configuration diagram of a buffer used in the circuit of FIG. 3, FIG. 4B is a configuration diagram of a selection circuit used in the circuit of FIG. 3, and FIG. 4C is a configuration diagram of a selection circuit used in the circuit of FIG. The configuration diagram of the output amplification circuit used in the circuit shown in FIG.
6 is a time chart showing the operation of the circuit shown in FIG. 6. FIG. 6 is an embodiment of the present invention that operates in a combination of continuous mode and page mode. FIG. 7 is a time chart showing the operation of the memory shown in FIG. 6. 9 is a block diagram of a memory consisting only of dynamic circuits operating in continuous mode, FIG. 9 is a time chart showing the operation of the memory of FIG. 8, and FIG. 10 is a diagram of the present invention operating in a combination of continuous mode and page mode. Example, 1st
1 is a block diagram of a pulse generation circuit used in the memory of FIG. 10, FIG. 12 is a time chart showing the operation of the circuit of FIG. 11, and FIG. 13 is a block diagram of a latch circuit used in the memory of FIG. Figure 14 is the 10th
Time chart showing the operation of the memory shown in Fig. 15.
The figure shows an embodiment of the present invention that performs continuous mode, page mode, and row selection operation in succession; FIG. 16 is a block diagram of a pulse generation circuit used in the memory of FIG.
Figure 7 is a time chart of the operation of the circuit in Figure 16.
18 is a time chart of the operation of the memory shown in FIG. 15, FIG. 19 is a modification of the selection circuit for continuous mode operation, FIG. 20 is a modification of the data write circuit, and FIG. 21 is a modification of the selection circuit for continuous mode operation. The time chart of the circuit, FIG. 22 is another modification of the data writing circuit,
FIG. 23 shows a layout of a memory according to the invention, FIG. 24 shows another layout of a memory according to the invention, FIG. 25 shows yet another layout of a memory according to the invention, and FIG. 26 shows a layout of a memory according to the invention. shows yet another layout of memory according to the second
FIG. 7 shows still another layout of the memory according to the invention, and FIG. 28 shows still another layout of the memory according to the invention.

Claims (1)

[Claims] 1. A memory array having a plurality of word lines, a plurality of data line pairs, and a plurality of memory cells arranged at desired intersections of the plurality of word lines and the plurality of data line pairs. , a common data line pair connected to the plurality of data line pairs, a connecting means connected between the plurality of data line pairs and the common data line pair, and an input connected to the common data line pair. a data holding means having an input connected to the output of the data holding means; a disconnecting means disposed between the common data line pair and the input of the data holding means; a selection means for selecting a first word line connected to a first memory cell of the memory array and a second word line connected to a second memory cell of the memory array; By selecting the first word line, the data of the first memory cell is read out to at least the data line pair to which the first memory cell is connected, and then read out to the common data line pair. , while the data output means is outputting the data of the first memory cell held in the data holding means, the selection means selects the second word line to output the data of the first memory cell. The data of the second memory cell is read out to at least the data line pair to which the second memory cell is connected, and then the data is read out to the common data line pair and held in the data holding means. A semiconductor memory device, wherein the data output means outputs the data of the cell, thereby making it possible to successively read data of each of the first and second memory cells belonging to the one memory array. 2. The semiconductor memory device according to claim 1, wherein the first and second memory cells each include one transistor and one capacitor. 3. The semiconductor memory device according to claim 1 or 2, wherein the data holding means includes a latch circuit. 4. Claim 1, wherein the data output means includes a static type amplifier circuit.
3. The semiconductor memory device according to any one of Items 1 to 3.