Semiconductor memory device and refresh period controlling method
Disclosed is a memory device including an error rate measurement circuit and a control circuit. The error rate measurement circuit, carrying a BIST circuit, reads out and writes data for an area for monitor bits every refresh period to detect an error rate (error count) with the refresh period. The control circuit performs control for elongating and shortening the refresh period so that a desired error rate will be achieved. The BIST circuit issues an internal command and an internal address and drives the DRAM from inside. The BIST circuit writes and reads out desired data, compares the monitor bits to expected values (error decision) and counts the errors.
Latest Patents:
This application is a division of application Ser. No. 11/152,762, filed Jun. 15, 2005, now pending, and based on Japanese Patent Application No. 2004-181734, filed Jun. 18, 2004, by Yutaka Ito and Takeshi Hashimoto, which are incorporated herein by reference in their entirety. This application claims only subject matter disclosed in the parent application and therefore presents no new matter.
FIELD OF THE INVENTIONThis invention relates to a semiconductor memory device and, more particularly, to a semiconductor memory device in need of refresh for data retention, and to a refresh period controlling method.
BACKGROUND OF THE INVENTIONCapability of pause refresh of a memory device is widely changed responsive to ambient temperature. Temperature compensation of a self-refresh period is the technique of controlling the refresh period for coping with the temperature dependency of the capability of pause refresh. If the technique of temperature compensation of the self-refresh period is brought to completion, it is possible to reduce the data retention current by one order of magnitude at ambient temperature as compared with that at elevated temperatures (upper limit of the guaranteed temperature). It is however technically difficult to design a circuit showing extensive temperature traceability suited to a device, such that there lacks an efficacious circuit designing technique.
The temperature compensation technique, so far proposed, may roughly be classified into the following two types:
(a) the technique in which temperature-dependent device parameters, such as the current of the MOS transistor, are detected by some means and converted into the refresh period, for example, the thermometer technique of monitoring the temperature to control the refresh period, or the technique of monitoring the cell leakage characteristics to control the refresh period, termed the ‘cell leakage monitor system’ in the present specification; and
(b) the technique of directly reading and writing plural memory cells that are for data retention, and determining the refresh period from the status of occurrence status of error.
Recently, a mobile SDRAM (Synchronous DRAM), characterized by the low data retention current, has been presented to the market. The products of this type basically uses the above technique (a).
Typical of the above technique (a) is the thermometer (temperature sensor) circuit. The refresh period controlling method, employing the thermometer circuit (thermometer system), detects e.g. minute temperature change of the level of the reference voltage (Vref) in a temperature sensor device (temperature sensor) and, based on the prevailing temperature information, elongates the refresh period by carrying out frequency-division, such as by doubling, quadrupling and the like, of the fundamental period of a timer counter.
However, since the variations in the level of generation of the reference voltage (Vref) in a thermometer are inherently independent of variations in the capability of refresh of the DRAM device, it is mandatory with the temperature system to match (trim) these two sorts of the variations at plural temperature points. This poses a serious problem in a wafer test step (the step of applying a probe on a pad of a wafer chip to carry out an electrical test of a die).
In the practical manufacturing process for DRAM devices, only temperature trimming for two points (high temperature point and low temperature point) is feasible from the perspective of test cost. As a result, fails tend to be produced at a mid temperature range so that a large operating margin needs to be provided as the variations are taken into account. With the result that it is not possible to achieve marked temperature traceability.
On the other hand, the techniques pertinent to (b) are disclosed in e.g. the Patent Publications 1 and 2 and in Non-Patent Publication 1.
The Patent Publication 1 discloses a standby current decreasing method (data retention current decreasing method) in a memory system employing the DRAM. As shown in
As the configuration of a DRAM device, provided with the ECC (error-checking and correction) encoding circuit and the decoding circuit as well as the refresh period control circuit, reference is made to the description of Patent Publications 2 and 4, indicated hereinbelow.
The operation of the system shown in
During the operation for data retention, all bit refresh→pause (e.g. refresh interval) is repeated. Each time all bit refresh is carried out, all data are read to an error correction decoding circuit (error correction circuit) to carry out error detection and correction. If there is no error, the refresh period is extended (elongated) by a preset multiplication factor k1 and if there is any error, the row in error is copied. Or, the refresh period is shortened in dependence upon the number of errors. By repetition of the above sequence of operations, the refresh period is allowed to converge to an optimum value.
On the other hand, the number of errors detected (error rate) is controlled to be within the gamut of the error correction capability of the error correction decoding circuit (error correction circuit).
In the Non-Patent Publication 1, this technique was substantiated by evaluation on a real operating apparatus. This technique, if grasped as being equipped in a DRAM, may be summarized as follows:
In a data retention mode, carrying an on-chip ECC (realizing low power dissipation by power control), all data are read to an ECC decoding circuit to monitor the error rate.
The refresh cycle is shortened and elongated in case the error rate measured is higher and lower than a preset value, respectively. By so doing, refresh period temperature compensation may be enabled as the error rate is maintained at lower than the ability for correction.
As the configuration for variably controlling the refresh period based on the error rate of the memory cells, there is disclosed in e.g. Patent Publication 4 a semiconductor integrated circuit device including an error rate selection circuit booted for the data retention mode to read plural data held in a dynamic memory circuit to generate check bits for error detection and correction for storage in a supplementary memory circuit. Plural items of data and check bits associated therewith are read by an ECC circuit, at a constant refresh period, to effect error detection and correction. A first detection signal indicating the absence of error is integrated, a second detection signal, indicating the presence of error, is integrated with a weighting factor larger than that of the first detection signal, the detection signals are summed in the manner of subtracting the first stated integrated value, the refresh period is elongated a predetermined length of time when the integrated value exceeds a predetermined value and the refresh period is shortened a predetermined length of time when the integrated value has become smaller than a predetermined value.
There is also known a configuration in which, with the data retention mode, tail bits inferior in the capability (real power) of pause refresh are corrected for errors by the ECC circuit and masked, and the refresh period is elongated up to one second to realize the saving in power usage (Patent Publication 2). As for temperature dependency of pause refresh characteristics (temperature dependency of tail bits and normal bits), see the above Patent Publication 2 (e.g.
[Patent Publication 1]
JP Patent Kokai Publication JP-A-11-213659 [Patent Publication 2] JP Patent Kokai Publication JP-P2002-056671A [Patent Publication 3] JP Patent Kokai Publication JP-2004-152378A [Patent Publication 4] JP Patent Kokai Publication JP-P2002-025229A [Non-Patent Publication 1]DFT'99 (1999 IEEE International Symposium on Defect and Fault Capability in VLSI systems), pp. 311-318
SUMMARY OF THE INVENTIONAs a technique for temperature compensation, provided with an on-chip ECC circuit, realizing a data retention mode of low power dissipation through power control, realizing the refresh of a long refresh period, reducing the data retention current and aimed to reduce the current consumption further in a low temperature range, there may further be provided a thermometer circuit and a cell leakage monitor circuit. The data retention mode in which the on-chip ECC circuit is provided, the parity information for the memory cells is generated in the ECC circuit to carry out power control, and in which the cells suffering from refresh defects are restored by error correction by the ECC circuit to achieve refresh with a long refresh period and low current consumption, is herein termed a ‘super self refresh mode’, abbreviated below to SSR mode. Regarding the semiconductor memory device, having an on-chip ECC circuit, realizing the data retention mode of the low power dissipation through power control, and also realizing the refresh with a long refresh period, see the description of the above Patent Publication 2.
The above techniques have their proper merits and demerits, such that, for realization of the data retention current of the order of 10 μA (ambient temperature), new temperature compensation techniques must be re-examined, beginning from their principles.
Accordingly, it is an object of the present invention to provide a semiconductor memory device and a refresh period controlling method in which there is exploited a temperature compensation technique whereby marked reduction in the data retention current may be achieved.
The present invention which attains the above and other objects may be summarized as follows.
A semiconductor memory device in accordance with one aspect of the present invention comprises a memory array including a plurality of memory cells which are in need of refresh operations for data retention, a circuit for writing preset data in a plurality of predetermined memory cells, termed monitor cells, in the memory array, a circuit for performing control for reading data from the plural monitor cells, in which the predetermined data have been written, when a preset period equal to or shorter than the refresh period has lapsed, and a circuit for comparing the data read from the monitor cell to the preset data to measure an error count or an error rate and for variably controlling the refresh period based on measured results of the error count or the error rate.
In the semiconductor memory device according to the present invention, the refresh is self-refresh, and the refresh period is controlled so that the error count or error rate as measured during the self-refresh will be within a desired setting range.
The semiconductor memory device of the present invention further comprises an error correction circuit, and the refresh being self-refresh. The self-refresh period may be controlled so that the error count or error rate as measured during the self-refresh will be within a range correctable by the error correction circuit.
The semiconductor memory device of the present invention may further comprise a BIST (built-in self test) circuit for generating an internal command and an internal address for writing and reading of preset data for the monitor cell and for controlling the comparison with expected value data.
In the semiconductor memory device according to the present invention, the monitor cells are provided in a monitor area provided separately from a usual memory area in the memory array, and the error count or the error rate is obtained by writing physical or logical data ‘1’ in the monitor area, by reading out the ‘1’ data from the monitor cells at a period equal to or shorter than the refresh period, and by observing the read data.
In the semiconductor memory device according to the present invention, the monitor area may have an unused area of a redundancy restoration memory.
In the semiconductor memory device according to the present invention, the monitor cells may be memory cell for one or more words selected by a word line distinct from word lines provided in a usual memory area.
A semiconductor memory device in accordance with another aspect of the invention includes a storage device having stored therein an internal address of the monitor cell. The monitor cell is accessed by the internal address from the storage device during writing and reading of data for the monitor cell. In the semiconductor memory device in another aspect of the invention, the period of measuring the error count or the error rate is set so as to be shorter than the refresh period for improving temperature follow-up characteristics.
In the semiconductor memory device according to the present invention, when the measured error count or error rate has exceeded the desired set value, error correction decoding for the memory array is carried out to lower the error rate to a desired set value. Control is managed for enabling all errors to be ultimately corrected by the decoding operation at the time of exiting from the data retention mode.
The semiconductor memory device according to the present invention further includes an ECC (error-checking and correction) encoding/decoding circuit. Data of the memory array is read at the time of entry to a data retention mode, the error detection correction encoding/decoding circuit generating check bits for error detection and correction for storage in a preset check area. The refresh period is set to a longer value within the range allowed for error generation by an error correction operation employing check bits. The ECC encoding/decoding circuit performs error bit correction and decoding, using the check bits, before returning to the normal operation in exiting from the data retention mode.
The semiconductor memory device according to the present invention further comprises an ECC (error-checking and correction) encoding/decoding circuit. Data of the memory array is read, at the time of entry to a data retention operation mode, the error detection correction encoding/decoding circuit generating check bits for error detection and correction and storing the check bit generated in a check area. Then,
(A1) an expected value data ‘1’ is written in the monitor cell;
(A2) all cells of the memory array are refreshed;
(A3) the operation is paused during the refresh time (at least a portion of the internal power supply circuit is turned off);
(A4) the internal power supply circuit is turned on and data is read from the monitor cell and compared with the expected value data ‘1’ to measure an error; and
(A5) when the error rate is larger than a preset upper limit value, the refresh period is shortened, whereas, when the error rate is not larger than a preset lower limit value, the refresh period is elongated and, when the error rate is intermediate between the preset upper and lower limit values, the refresh period is not changed.
The operations from (A1) to (A5) are repeated until exiting from the data retention mode. When exiting from the data retention mode, error correction is carried out in the error detection correction encoding/decoding circuit before proceeding to the normal operation.
In the semiconductor memory device according to the present invention, control of the self-refresh period may be such that several bits or more inferior in the capability of pause refresh to the usual memory area are used as monitor cells. These several bits, inferior in the capability of pause refresh to the usual memory area, are obtained by carrying out the restoration by pause refresh for M or more bits and by selecting, out of these bits, approximately M/10 having the worst capability. The more the number of the monitor bits, the higher is the precision in error rate judgement. In actuality, M is 50 or more bits.
A method for controlling the refresh period for a semiconductor memory device, in another aspect of the present invention, comprises the steps of writing expected value data in a plurality of memory cells in a memory array including an array of memory cells, each being in need of a refresh operation for retention of data, reading data from the plural memory cells, in which the expected value data have been written, at a period equal to or shorter than the refresh period, comparing the read data to the expected value data to measure an error count or an error rate, and variably controlling the refresh period based on measured results of the error count or error rate for elongating, shortening or maintaining the refresh period.
The meritorious effects of the present invention are summarized as follows.
According to the present invention, the memory cells, in which the expected value data have been written, are read with a period equal to or shorter than a refresh period, and compared with the expected value data to measure an error count or an error rate, and in which the refresh period is variably controlled based on the measured results, so that temperature compensation characteristics may be improved appreciably.
Moreover, according to the present invention, the trimming adjustment in the manufacture process may be dispensed with to reduce the cost, while stable temperature adjustment may be achieved extremely readily.
Still other objects and advantages of the present invention will become readily apparent to those skilled in this art from the following detailed description in conjunction with the accompanying drawings wherein only the preferred embodiments of the invention are shown and described, simply by way of illustration of the best mode contemplated of carrying out this invention. As will be realized, the invention is capable of other and different embodiments, and its several details are capable of modifications in various obvious respects, all without departing from the invention. Accordingly, the drawing and description are to be regarded as illustrative in nature, and not as restrictive.
Referring to the drawings, the present invention will be described hereinafter in more detail. According to the present invention, temperature control for the refresh period is as shown in
A refresh address generating circuit 107, responsive to an output of the frequency divider circuit 106 (refresh trigger signal), generates and outputs a refresh address for refreshing the memory cells selected by the refresh address. That is, the data signal of the memory cell, output by the sense amplifier to a bit line, is amplified, and the so amplified data signal is re-stored from the bit line to the memory cell.
According to the present invention, described above, it is unnecessary to read check bits by the ECC decoding circuit to detect errors for measuring the error rate, contrary to the case of the aforementioned Patent Publication 2.
The present invention may be applied to a usual self refresh system in which encoding/decoding by the ECC encoding/decoding is not carried out. In this case, approximately 10 refresh restoration bits may be monitored.
By way of comparison with the present invention, such a configuration in which the refresh period is controlled by a cell leakage monitor circuit, referred to below as the ‘cell leakage monitor system’, and such a configuration in which the refresh period is controlled based on temperature detection, referred to below as the ‘temperature system’, will now be described.
As a principle, the refresh period may be adjusted continuously (that is, in an analog method), depending on the temperature. Moreover, during the time other than the refresh period, only the cell leakage monitor circuit may be in operation, so that the data retention current not larger than 10 μA may be achieved. Although the cell leakage monitor system may be said to be an ideal system, it exhibits marked process dependency, while the matching of the period (absolute value) is of a problem.
Hence, the temperature points detected depend on samples and are in need of trimming, even though there is no risk of the period being severely beside the mark. The result is that the margin of the refresh period needs to be set to a larger value, such that, in the worst high-temperature state, the refresh period tends to be shorter than in the conventional product exhibiting no temperature dependency. In addition, since the reference voltage (Vref) generating circuit and an amplifier circuit are needed at all times, it is difficult to achieve the data retention current not larger than 20 μA.
In comparison with the cell leakage monitor circuit and with the thermometer circuit, the present invention, described with reference to
The refresh period can be elongated and moreover the data retention current can be lowered to 10 μA or less. That is, the DC component of the refresh current can be lowered to 10 μA or less.
(a) If conversely the thermometer circuit is to be realized, the reference voltage (Vref) generating circuit, exhibiting temperature dependency, the reference voltage (Vref) generating circuit, not exhibiting temperature dependency, and a differential amplifier circuit for amplifying the results of comparison of the reference voltage Vref, are needed, so that the DC current is caused to flow at all times. For generating the reference voltage Vref, the data retention current equal to or larger than 20 μA is needed.
(b) According to the present invention, there is no process dependency, while trimming adjustment is scarcely needed. Consequently, stable temperature compensation may readily be achieved.
Conversely, for realizing the thermometer circuit, trimming for absorbing variations in the detected temperature is needed. In actuality, there is variation on the order of 30° C. in the detected temperature and, after trimming, there persists the variation of the order of, for example, 10° C.
The refresh capability of the device has marked temperature dependency, so that the temperature difference of 10° C. causes the refresh capability to be changed by a factor only slightly smaller than two. Thus, the refresh period has to be set to a shorter value (in a margin increasing direction). Conversely, there are occasions wherein the data retention current itself is increased as a result of the temperature compensation.
(c) Since the data of the real memory is read, and the refresh period is set depending on the state of errors, the refresh period may be elongated (adjusted) depending on the error of retention data of the memory circuit to achieve temperature compensation.
(d) However, in actuality, the configuration of reading out all data of the memory array to monitor the error state cannot be put to practical use as the technique for temperature compensation.
That is, since the error rate of the totality of data needs to be monitored, all bit READ needs to be carried out for each refresh. The base (AC component) of the data retention current is increased by a factor of two or more and hence low current consumption cannot be achieved in a high temperature range.
(e) The above problem (d) may be avoided by diminishing the memory area for interest. Thus, a correct value (error rate) cannot be obtained, such that refresh period setting may be random. For example, if the retention data is 0 data, no failure occurs.
The present invention improves the shortcomings of the above-described conventional cell leakage monitor or thermometer system to realize the temperature compensation system aimed to lower the power, as the merits of the two systems are exploited. Specifically, for exploiting the basic operation of the conventional technique and eliminating the shortcomings thereof, the following three requirements are added and the subject of monitoring the error rate is limited to a newly provided monitor bit area.
Referring to
The monitor cell for monitoring the error rate, according to the present invention, will now be described.
According to the present invention, data ‘1’ is written in the monitor bit 103 during the refresh operation in the data retention mode (SSR mode) of low power dissipation. The data is read every pause refresh (power off during the pause period) to monitor the error rate. Similarly to the check bits (parity bits) 102, the monitor bit 103 can be redundancy-restored (a fail cell is replaced by a redundancy cell during wafer test) and specified address is allocated to the restored bit. There are provided a number of monitor bits 103 corresponding to the number of the memory cells connected to the word line selected by an X-decoder 104. This word line is distinct from the word line for information or check bits. Meanwhile, the monitor bits 103 are selected by the internal address (X address and Y address), generated by the BIST circuit, not shown.
Alternatively, an unused area (which passed e.g. a wafer test) of the redundancy restoration memory may be allocated as monitor bits.
The error rate calculating circuit (ERATE MONITOR) 105 is provided in parallel with the error detection correction encoding/decoding circuit (ECC CODEC) to generate write data for the monitor bits as well as to measure the error rate (number of failure bits) of the read data.
According to the present invention, in which the expected value compared with the read data from the monitor cell is known, the error rate calculating circuit (ERATE MONITOR) 105 may be composed by a simplified configuration composed of a counter circuit (COUNTER) and an adder (ADDER), as shown for example in
In case where the present invention is applied to a DRAM, carrying the data retention mode (SSR mode) of low power dissipation, with the ECC-CODEC of 1 set/4 banks, the error rate calculating circuit (ERATE MONITOR, also referred to as an ‘error monitor circuit’) 105 is attached to the internal I/O bus, as is the encoding/decoding circuit (ECC CODEC). The error rate calculating circuit (ERATE MONITOR) 105 executes error rate count and write operation, in addition to the usual read and write operations.
For monitoring the error rate during the self refresh operation, the present invention includes an ECC controller (see CONTROLLER 6 of
The BIST circuit generates internal commands (e.g. ACT, READ, WRITE, PRE), generates an internal address of the monitor bit to output the so generated address to the interfacing circuit of the SDRAM, generates an address signal for controlling the outputting of the expected value data to the I/O bus and the sampling of the read data from the I/O bus for the error rate calculating circuit (ERATE MONITOR), for generating an address signal for generating the expected value data, and acquiring an error rate from the error rate counter circuit (ERATE MONITOR) to output a refresh period control signal (UP, DOWN, KEEP) to the refresh control circuit (see
According to the present invention, in which measurement of the error count or the error rate is limited to pause refresh failure (‘1’ failure) of the monitor bit area, as a small capacity memory, the current consumption in monitoring the error rate is reduced to the write and read current for the small capacity memory (monitor cell). The write and read current for the monitor cell becomes negligibly smaller than the refresh current.
The monitor bit area 103 of
There is no vital difference in the probability of occurrence of tail bits of the pause refresh failure (‘1’ failure) of conventional products and products of many manufacturers. This probability is on the order of 0.001% in the lowest case. Thus, if the memory capacity is approximately 256 kbits or more, the error rate as necessary may be monitored with the small memory area of, for example 32 kbits. It is noted that addition of 32 Kbits of monitor bits only increases 1/(256×32)=0.12% for the 256 M cell array.
According to the present invention, in which the error rate of data written in the memory cell is directly monitored, the ideal temperature compensation, similar to that of the aforementioned Patent Publication 1 (see
According to the present invention, the refresh period is controlled based on the error rate of the error rate calculating circuit 105, so that the reference voltage Vref is unneeded, and hence the data retention current may be set to 10 μA or less.
The present invention may also be applied to the usual self-refresh. Approximately 10 refresh restoration bits are monitored.
With the semiconductor memory device, according to the present invention, the following operation is carried out after entry of self refresh and the SSR mode on command input.
(1) All bits are read to generate parity bits (this is the same encoding operation as that of the above Patent Publication 2).
(2) ‘1’ is written in the monitor bits and all bits are burst-refreshed.
(3) After a set pause period, ‘1’ monitor bits are read and the error rate is counted. Meanwhile, the error rate count corresponds to the count of the number of the errors generated.
(4) If the error rate is larger than the upper limit value e1, the refresh period is made shorter than the current value. If the error rate is lower than the lower setting limit value, the refresh period is made longer than the current value. If none of the above applies, the refresh period is not changed.
The operation then goes back to (2) above to repeat the refresh and error rate monitor operation. By this control algorithm, the refresh period can be variably controlled within the range of the correction capability of the on-chip ECC, as the error rate is suppressed.
(5) On exit command input, all bits inclusive of parity bits are read, and the error bits are corrected, after which reversion is made to the normal mode (the same decoding operation as that of the aforementioned Patent Publication 2). The present invention will now be described in detail with reference to preferred embodiments thereof.
EmbodimentThe sense amplifier 203 receives and amplifies a small potential difference, appearing on the complementary bit lines on data read from the memory cells. The I/O gating write driver 204 includes a column switch MOS transistor for selecting the complementary bit lines for connection to the complementary I/O lines. The column switch MOS transistor is selectively activated responsive to the decoded results of the column address signal by the column decoder 205. Each memory array 200 of the banks 0 to 3 is similarly provided with a row decoder (ROW DECODER & LATCH), a sense amplifier, an I/O gating write driver, and a column decoder.
The I/O line is provided common to the memory banks and connected to an output terminal of a data input register 210 and to an input terminal of a data output register 211. Terminals D0 to DQ15 (16 bits) are data input/output terminals that receive or output 16-bit data D0 to D15 (of which D0 to D7 are lower bytes and D8 to D15 are upper bytes). DQML and DQMU are control signals for byte controlling, used for controlling the input/output buffer. Specifically, the control signals DQML and DQMU control the input buffer for upper bytes and that for lower bytes, respectively.
Address bits A0 to A14 of 15-bit address signal, supplied from the address input terminal, are temporally held by an address register (ADD REG) 213. Out of the above address signals, entered sequentially, the row address signal for selecting the memory cell is sent to the row decoder 201 of the memory bank, via a row address multiplexer (ROW ADD MUX) 206, and thence supplied to the row decoder 201 of the memory bank. Address bits A13 and A14 are allocated to bank select signals BA0 and BA1 for selecting the memory bank. The signals BA0 and BA1 are sent to a bank control circuit (BANK CONTROL LOGIC) 212, where a selection signal for the four memory banks is generated and sent to the row decoder 201. The column address signal is held in a column address counter/latch (COLUMN ADDRESS CONTROL/LATCH) 207.
A refresh counter (REFRESH COUNTER) 208 generates a row address for self-refresh. For example, with the memory capacity of e.g. 256 Mbits, a 9 bit address signal is valid for a×8 bit configuration. The column address counter 207 is supplied with an input column address signal as preset data and, in the burst mode, specified by a command, as later described, outputs the column address signal, as the preset data, or a sequentially incremented value of the column address signal, to the column decoder 205 of each memory bank.
A control logic (CONTROL LOGIC) 209 forms an SRAM interfacing circuit and includes a command decoder (COMMANDDEC) 2091, a self-refresh control (REF CONTROL) 2092 and a mode register (MODE REGISTER) 2093. The mode register 2093 holds various operating mode information.
The command decoder 2091 receives and decodes outer and internal commands, depending on the operating mode.
The self-refresh control circuit 2092 is a self-refresh control block and performs refresh operations and period control thereof.
Only the row decoder 201 associated with the bank as specified by the bank control circuit (BANK CONTROL LOGIC) 212 is in operation to perform the word line selecting operation.
In the example shown in
The clock enable signal CKE is a signal which commands the validity of the next clock signal. When the clock enable signal CKE is high or low, the rising edge of the next clock signal is valid or invalid, respectively.
If, in the read mode, there is provided an output enable signal OE controlling the output enable for the data output register (DATA OUTPUT REGISTER) 211, the signal OE is also output to the control circuit 209. If the signal is e.g. in a high level, the output of the data output register 211 is in a high output impedance state.
The row address signal is defined by the levels of address bits A0 to A12 in a row address strobe/bank active command cycle synchronized with the rising edge of the clock signal (internal clock signal).
The address bits A13 and A14 (BA0 and BA1 of
If, in an SDRAM, a burst operation is going on in a given memory bank, another memory bank is specified in the interim and a row address strobe/bank active command is supplied, the operation of the row address system in said another memory bank is enabled without affecting the operation going on in the given memory bank. Thus, except if data collision occurs in e.g. 16-bit data input/output terminal DQ0 to DQ15, the internal operation may be started in advance by issuing a pre-charge command and a row address strobe/bank active command to a memory bank different from the memory bank processed by the command being executed during execution of the command the processing of which has not as yet come to a close.
Referring to
The reference numeral 2 is an internal command signal from the ECC controller 6 to the SDRAM interface, or a JOB end signal. For example, when the encoding or decoding operation is finished, a READY signal is output.
The reference numeral 3 is an internal address signal supplied from the ECC controller 6 to the address register 213.
The reference numeral 4 is an operation mode signal (CODEC MODE) of an ECC coding/decoding circuit (ECC CODEC) 7. The signal is controlled from the ECC controller depending on the operation. There are e.g. four modes of initialization (INIT), syndrome calculation (SYNDROME), outputting of parity bits (PARITY) and error position detection.
The reference numeral 5 is an error detection signal, and an error position detection signal (ERROR-LOCATION DETECT) from the ECC coding/decoding circuit (ECC CODEC) 7 to the ECC controller 6.
The reference numeral 6 is an ECC controller (1 set/4 banks). The ECC controller 6 outputs an internal command and an internal address to control the internal operation of the SDRAM, as well as to control the ECC codec 7 depending on the operations. The ECC controller also controls an error rate calculating circuit (ERATE MONITOR) 11, in the error monitor operation, depending on the operations.
The reference numeral 7 is an ECC coding/decoding circuit (ECC CODEC) and includes a coding circuit and a decoding circuit to carry out syndrome calculation, parity bit calculation and error detection/correction.
The reference numeral 9 is a period change signal (e.g. UP/DOWN signal) from the ECC controller 6 to the self-refresh control circuit 2092.
The reference numeral 11 is an error rate calculating circuit (ERATE MONITOR) for comparing a read data from the monitor cell with an expected value to count errors. The error rate calculating circuit also outputs the expected value as write data to the memory array.
The reference numeral 12 is an operating mode signal (MODE) to the error rate calculating circuit 11. The signal is controlled from the ECC controller 6 and has e.g. an expected value READ mode and an expected value WRITE mode. The error rate calculating circuit 11, that generates the expected value responsive to an address, is also equipped with a data scrambler (see
The reference numeral 13 is an error rate (FBIT) from the error rate calculating circuit 11. It is a binary number made up of several bits and is output from the error rate calculating circuit 11 to the ECC controller 6.
The reference numeral 14 is a memory array of a bank memory and the reference numerals 15 and 16 denote a monitor field and a parity field, respectively.
The present embodiment uses a cyclic code for ECC (error-checking and correction) mounted on-chip with the DRAM, and diminishes the stand-by current due to correction of refresh malfunctions (masking) during self-refresh. Specifically, an ECC codec (coding/decoding circuit) 7 is provided for each of the memory banks 0 to 3. The ECC codec 7 receives a mode (CODEC MODE) signal from the ECC controller 6 to correct the read data for errors and notifies the error detection information (ERROR LOCATYION DETECT) to the ECC controller 6. The ECC controller 6 also performs control of varying the refresh period in the self refresh control circuit 2093 based on the error rate.
The error rate calculating circuit 11 receives a mode (MODE) signal from the ECC controller 6 and, in the error monitor mode, measures the error of the monitor bits (Fail Bit) read from the I/O gating write driver 204 to notify the ECC controller 6 of the error.
If the cyclic code is used as the ECC circuit, the circuit of the coder/decoder (CODEC) forming the ECC is also reduced in size. Consequently, the code length may be increased, and the number of parity bits may be decreased, so that the increase in the chip area by the DRAM on-chip ECC may be suppressed to the smallest possible value.
Meanwhile, plural cycles are needed for generating parity bits or for correcting error bits. Consequently, the ECC operation by the ECC codec (generation of parity bits and error detecting/correcting operation) in the present embodiment is carried out in case the access speed is not of vital concern, such as when the self refresh is entered for the data retention mode and when reversion is made to the normal mode from the self refresh mode.
In the present embodiment, low power dissipation may be achieved by restricted ECC operation. With the ECC codec 7, pause refresh malfunctioning bits may be corrected, and hence the ECC circuit is not in operation during self refresh in the data retention mode. Thus, even if error bits are generated, these error bits are allowed to stand, without being corrected, during the self refresh period in the data retention mode. It is at the time of exiting from the data retention mode to the normal mode that error detection/correction is carried out by the ECC decoding circuit of the memory array.
In the configuration shown in
Although not shown, an internal power generating circuit is provided in
Meanwhile, in the embodiment shown in
For storage of the monitor bits, a separate memory is needed, as is the memory for parity bits.
When an unused area of the redundancy restoration memory is used for the monitor bits, the area that may be arranged for the area of the monitor bits differs from chip to chip. Thus, one set is provided on the internal I/O bus, as shown in
The super self-refresh mode is entered from the normal operation, responsive to a command.
The parity (check bit) for error detection/correction is generated (step S101).
‘1’ is written in the monitor cell (monitor bit field) of the memory array (step S102).
All cells of the memory cell array is refreshed (step S103).
For example, a pause is made for one second (power off). In the power off state, the internal power supply circuit is partially turned off.
‘1’ is read from the monitor cell (step S105).
When the error rate is larger than an upper limit e1 , the refresh period tREF is shortened (step S107).
tREF←tREF/k1
When the error rate is not larger than a lower limit e2, the refresh period tREF is elongated (step S108).
tREF←tREF×2
If an error rate is not larger than e1 and larger than e2, the refresh period is not changed.
In exiting from the SSR mode, the ECC codec corrects errors, using check bits of the memory array, to correct the errors of the memory cells. The ECC codec then proceeds to the normal operation.
In the present embodiment, error measurement is made every refresh period. For coping with rapid temperature change during the pause period, the refresh period is reduced to a short value of 1/k1 in case the error rate (number of counts) is larger than a set value (e1 ). At this time, rapid period control of ½ or ¼ is used.
If, on the other hand, the error rate (number of counts) is smaller than the set value (e2), the refresh period is elongated by a factor of k2. It is noted that low speed period control is performed by setting k2 e.g. by a factor of 1.1 or 1.2.
In case of a simplified control by multiplying the period by 2 or 4 by a frequency dividing counter, no detection is made during the pause period, so that, when the temperature rise has commenced, the period tends to be too long, and hence k2 is increased by a factor of 1.1 or 1.2.
The timer circuit, determining the refresh period, is required to have this period controlling function. The timer circuit will be described subsequently (see
This flowchart shows the operation of an embodiment for monitoring the error with a period shorter than the refresh period. The pause period is successively increased to monitor the error rate. However, no period change is made at each error monitor point. The period change is carried out by each cycle of refresh for all cells.
The monitor period for the monitor bit error rate is set to 1/n of the refresh period. Referring to
The encoding circuit of the ECC codec 7 generates the parity of the memory cells of the memory cell array.
‘1’ is written in the monitor cell (step S202).
All of the cells of the memory array are refreshed (step S203).
A 1/n pause of the refresh period (power down) is made (step S204).
‘1’ is read from the monitor cell (step S205).
If the error rate is larger than the set value e1 , the refresh period tREF is shortened (step S207).
tREF←tREF/k1
If, with the error rate not higher than e2, the number of times of error monitor is less than n, that is, error check is to be performed further, pause processing of the step S204 is performed. If, with the error rate not higher than e2, n times of the error monitor are finished, the refresh period tREF is elongated (step S209).
tREF←tREF×k2
If the error rate is not larger than e1 and larger e2, the refresh period is not changed.
When exiting from the super self refresh mode, error correction is made and transition is made to the normal operation.
When desired to improve follow-up characteristics to changes in temperature, the error rate is monitored at a period shorter than the refresh period. For example, the refresh for all cells operation is carried out after carrying out error rate monitoring twice. If n=2 and the refresh period is 1 sec, the pause period is 500 msec.
After the pause time, the error monitor for monitor bits is carried out (the monitor area is read and the error rate is determined). If the error rate is larger than the set value, processing moves to period shortening and refreshing for all cells, without regard to the number of times of error rate monitoring, and the number of times of pause is also reset.
When the error rate is smaller than the set value e2, and the current error monitoring is the first one, reversion is directly made to the pause period without doing anything. After this pause, the second error monitor operation is again performed. At this time, the error monitor operation is carried out for the pause period twice as long and transfer is made to period change according to the error rate and to the refresh for all cells operation. By so doing, the error rate may be monitored with the one-half of the refresh period (where n=2). However, in order to render this effective, a monitor bit area twice as large is required.
That is, there is meaning only when the error monitor memory area of the second monitor operation is the memory area of monitor bits distinct from that of the first monitor operation. It is because the error monitor operation (read operation of the monitor area) performed results in the refresh (re-write) operation being carried out for the memory cells in question. If the second error monitor operation is performed for the same memory cell, the pause time for the memory cell is scarcely changed from that for the first error monitor operation.
The data written in a monitor bit is data ‘1’ and is mainly aimed to detect pause refresh failure and need not be physical data ‘1’.
For example, if data scrambling (see
The parity is generated in the coding circuit of the ECC codec (step S301).
All of the cells of the memory array are refreshed (step S302).
With the error check mode (Yes of step S303), ‘1’ is written in the monitor cell (Yes of step S304). When error decision has been carried out n times, processing moves to a step S302.
1/n pause of the refresh period is made (power off) (step S305).
‘1’ is read from the monitor cell (step S306).
The error rate is judged (step S307). If the error rate is larger than the set value e1 , the refresh period is set so that tREF←tREF×k1 (step S308) and then processing moves to a step S303.
If the error rate is not larger than e2, tREF←tREF×k2 (step S309) and then processing moves to the step S303.
If the error rate is not larger than e1 and larger than e2, processing moves to a step S303 without changing the refresh period.
When exiting from the super self refresh mode, the decoding circuit of the ECC codec carries out the error correction. Then, processing moves to the normal operation (step S310).
The parity is generated in the coding circuit of the ECC codec (step 401).
‘2’ is written in the monitor cell (step S402).
All of the cells of the memory array are refreshed (step S403).
If the mode is the error check mode (Yes of step S404), 1/n pause of the refresh period is made (power off) (step S405). If an error monitor operation is carried out n times (No of step S404), processing moves to a step S402.
‘1’ is read from the monitor cell (step S406).
The error rate is judged (step S407). If the error rate is larger than the set value e1 , the refresh period is set so that tREF←tREF/k1 (step S408) and then processing moves to a step S404.
If the error rate is not larger than e2, tREF←tREF×k2 (step S409) and then processing moves to the step S404.
If the error rate is larger than e1 and larger than e2, processing moves to a step S404 without changing the refresh period.
When exiting from the super self refresh mode, the decoding circuit of the ECC codec carries out the error correction. Then, processing moves to the normal operation (step S410).
The operating state moves from the idle state to the active state on receipt of the active command.
From the idle state, the operating state undergoes a transition to mode register setting by command input (mode register setting). After the mode register setting, the state transition back to the idle state is executed.
From the idle state, parity is appended by a command input of the SSR entry, followed by the transition to SSR. Error correction is executed by a command input of SSR exit. The state transition back to the idle state is performed.
From the idle state, the operating state undergoes a transition to SR by a command input of the SR (self refresh) entry and returns to the idle state by a command input for SR exit.
From the idle state, the operating state undergoes a transition to DPD (Deep Power Down) by a command input to DPD entry, and goes to the idle state by a command input of the DDR exit.
The ECC controller 6 accesses the ROM 214, having stored therein a usual self refresh accommodating monitor bit address, and outputs an internal address from the monitor bit address to carry out the read-write operation.
An adder (ADDER) 1706, supplied with a fail bit count (Fail Bit Count: 4 bits) from 16 I/O lines, adds the fail bit counts every clock ACLK and outputs the result of addition (total sum) associated with the 16 I/O lines.
Meanwhile, a data scramble circuit 1701 is supplied with TPH, internal address IYA (i) and IXA (j), supplied from BIST (Built In Self Test) controller, not shown (see
BISTR (BIST read), BISTW (BIST write), DCLK (clock), EINT (initialization), ECLK (clock) and ACLK (clock of the adder) are supplied from a BIST controller enclosed in the DRAM. TM79 is a control signal supplied from a control logic 209 of
In
Referring to
Responsive to this period UP signal, an UP value (+3) is summed to the limit value which is thereby increased to 22 (hexadecimal). The clock signals CLK1 and CLK2 are then clocked. As the UP and DOWN values, ⅛ (½ multiplied by itself three times) of the LIMIT value is found and ½ is found, respectively.
That is, when LIMIT is 1f (hexadecimal), a 12-bit counter (2508 of
Responsive to the high level of the signal (MONITOR PERIOD), the signal MREAD (MONITOR READ) is activated. UP_VAL is set to 03 and DOEN_VAL is set to −0f. Responsive to the high level UP of the period control signal, UP_VAL is entered to a 13-bit adder (2504 of
On the other hand, an output value 22 of the 13 bit adder (2504 of
The ECC block 4105 is responsive to the internal clock signal ICLK, encoding control signal ENCODE and decoding control signal DECODE to output an internal address, an internal command and a ready signal READY. The ECC block 4105 is also supplied from the ECC codec with ERROR and LOCATION to output a signal INIT, commanding the initialization, a signal PARITY, commanding the generation of parity, a signal SYNDROME, commanding the syndrome calculation, a signal CORRECT, commanding error correction, and a codec enable signal CODECE.
An address generator 4104 of the BIST block 4101 outputs an internal address signal IXA. The command generator 4103 outputs the read signal BITSR, write signal BISTW, initialization signal EINT, clock signal ACLK, DCLK and ECLK to the error rate calculating circuit (ERATE MONITOR).
The error rate check circuit (ERATE CHECKER) 4102 receives EBIT and OVER from the error rate calculating circuit (ERATE MONITOR) to output UP, DOWN and KEEP to the self-refresh control circuit (SELF REFRESH CONTROL).
When EBIT is not larger than e2, an output of the comparator 4204 is high and is sampled by a register 4207. When OVER is at a low level, UP is at a high level.
When EBIT is not less than e1 , an output of the comparator 4203 is at a high level and is sampled by a register 4205. The signal DOWN is at a high level. When an overflow signal OVER is at a high level, the signal DOWN is at a high level, and the refresh period is shortened.
Referring to
Also, in the fuse ROM of
On the other hand, in the fuse ROM of
By the MREAD (MONITOR READ), the initializing signal EFNIT goes high, error count is cleared and EBIT is cleared to 0, as shown in
An ACT command and an internal address XA(0) are supplied, a monitor cell is read by a READ command and YA(0) (CAS latency CL=2) and the signal BISTR is set to a high level. An output 1 of MA (OUTPUT REGISTER) is propagated to a comparator circuit (1802 of
The read data of the internal address YA(1) is set to 1, which differs from the expected value 1, so that the ERROR is at a high level, and FBIT at 16 I/O, which is 2 (output of the adder 1804), is output to the BIST controller. When the reading of address data of YA (511) comes to a close, a precharge command is entered. 198 of FBIT is compared with the clock signal JCLK with e1 and e2. Since FBIT is smaller than e2, an UP signal is output and control is carried out for elongating the timer period. When reading from the monitor bit has come to a close, READY (one-shot) is output and MREAD is set to a low level.
An X-address XA(1), associated with the monitor point 1, is entered by an ACTV command, a Y address YA(1) is entered with the READ command and the data (MA) of the monitor bit, read at the latency CL1, is compared with the expected value (in this case, ERROR=0). In this manner, the monitor bit of address (random) stored in a monitor bit address storage ROM (14 of
The MWRITE signal then goes high and the operation of writing ‘1’ to the monitor bit area of, for example, 32 Kbits, starts.
After the end of the 11WRITE operation, an end signal (READY signal) is output, as in the encoding operation, and the operation of the ECC controller ceases.
The state of the chip changes in succession to all bit refresh→pause period (internal power supply in cessation, power-off period).
After the pause period has come to a close, the MREAD signal supplied to the ECC controller goes high to cause the ECC controller to carry out 1READ operation and error count operation for the monitor bit area.
Since the error monitor period is one-half the refresh period, error count occurs twice during all bit refresh.
After the error count operation has come to a close, an end signal (READY signal) is output as a pulse to halt the operation.
Responsive to the result of the error rate count, the period control signal is changed UP/DOWN/KEEP.
In the present embodiment, the decision is for UP.
Then, transition is made 1 WRITE operation→all bit refresh→pause period for the monitor bit area, in order to repeat the error monitor operation.
When an SSR exit command is entered, the operation is interrupted to transfer to the decoding operation (error correcting operation). The signal DECODE goes high with respect to the ECC controller. The error correction is carried out as the internal command and the internal address are controlled.
When the error correction has come to a close, an end signal (READY signal) is output as a pulse to go back to an Idle state (normal state) and to an external command accepting state.
The technique of the present invention is not limited to entry to the SSR mode, parity generation/storage by the ECC, refresh, pause, power on, exit or error correction and may naturally be applied to usual self-refresh not performing parity generation or error correction by the ECC circuit.
As pause refresh restoration bits, approximately 100 bits are provided. About 10 bits with the worst capability are used as monitor bits and the addresses thereof are recorded in a ROM circuit, such as a fuse circuit which is cut off by a laser. During the self refresh cycle, it is sufficient to carry out the control algorithm of the present embodiment (see
Meanwhile, in the case of usual self-refresh, the d.c. components cannot be decreased by power down of the internal power supply. Hence, the effect on low power dissipation tends to be saturated in approximately one second. Thus, when the period has reached e.g. 2 seconds, the period is no longer extended but is fixed.
With the present embodiment, the data retention current in the SSR mode dependent on ambient temperatures may be realized as the capability in a high temperature range as before is maintained, by ideal temperature compensation of the refresh period, such that the data retention current may be reduced to as low as 10 μA at ambient temperature. Although the present invention has been described with reference to the above-described embodiments, it is noted that the present invention is not limited to the configuration of these embodiments and may naturally comprise various changes or corrections that may occur to those skilled in the art within the scope of the invention.
It should be noted that other objects, features and aspects of the present invention will become apparent in the entire disclosure and that modifications may be done without departing the gist and scope of the present invention as disclosed herein and claimed as appended herewith.
Also it should be noted that any combination of the disclosed and/or claimed elements, matters and/or items may fall under the modifications aforementioned.
Claims
1. A semiconductor memory device comprising:
- a memory array including a plurality of memory cells, each of said memory cells being in need of a refresh operation for retention of data;
- a circuit for writing preset data in predetermined plural memory cells, referred to as monitor cells, provided in said memory array;
- a circuit for performing control for reading data from said plural monitor cells, in which said preset data have been written, with the lapse of time equal to or shorter than a refresh period;
- a circuit for comparing the data read from said monitor cell with said preset data to measure an error count or an error rate and for variably controlling the refresh period based on a measured result of said error count or the error rate; and
- an error correction decoder,
- wherein
- when the measured error count or error rate has exceeded the preset value as determined by said circuit for comparing the data read from said monitor cell with said preset data, error correction decoding for said memory array is carried out by said error correction decoder to lower the error rate to a desired value.
2. The semiconductor memory device according to claim 1, wherein
- said memory cell includes:
- a cell transistor having a terminal connected to a word line for being controlled on/off, said cell transistor having a first signal terminal connected to a bit line; and
- a capacitor for storing data therein by having one end thereof connected to a second signal terminal of said cell transistor;
- said preset data being of such a value as to cause electric charge to be stored in said capacitor in the selected memory cell.
3. The semiconductor memory device according to claim 1, wherein
- said refresh is self-refresh; and wherein
- said semiconductor memory device further comprises
- a circuit for controlling said refresh period for said self-refresh so that said error count or error rate as measured during said self-refresh is within a desired setting range.
4. The semiconductor memory device according to claim 1, wherein
- said refresh is self-refresh; and wherein
- said semiconductor memory device further comprises:
- an error correction circuit; and
- a circuit for controlling said self-refresh period so that said error count or error rate as measured during said self-refresh is within a range correctable by said error correction circuit.
5. The semiconductor memory device according to claim 1, further comprising
- a test control circuit for generating an internal command and an internal address for writing and reading of preset data for said monitor cells and for controlling the comparison of the read data with expected value data.
6. The semiconductor memory device according to claim 1, wherein
- said plural monitor cells are provided in a monitor area provided separately from a usual memory area in said memory array.
7. The semiconductor memory device according to claim 6, wherein
- said error count or said error rate is measured by said circuit for writing preset data in predetermined plural memory cells writing physical or logical data ‘1’ in the monitor cell of said monitor area and by said circuit for performing control for reading data from said plural cells reading said ‘1’ data from said monitor cell at a period equal to or shorter than said refresh period.
8. The semiconductor memory device according to claim 6, wherein
- said monitor area has an unused area of a redundancy restoration memory.
9. The semiconductor memory device according to claim 6, wherein
- said monitor area includes, as said monitor cells, the memory cells connected to one or more word lines and selected by said word line distinct from word lines provided in a usual memory area in said memory array.
10. The semiconductor memory device according to claim 1, wherein
- said monitor cell is a memory cell of an optional address in said memory array;
- there being provided a memory device having stored therein an internal address of said monitor cell; and wherein
- said monitor cell is accessed by the internal address from said memory device during writing and read of data for said monitor cell.
11. The semiconductor memory device according to claim 1, further comprising:
- an ECC (error-checking and correction) encoding/decoding circuit for performing error checking and correction of data stored in said memory array;
- wherein
- data of said memory array is read at the time of entry to a data retention operation mode, said error detection correction encoding/decoding circuit generating check bits for error detection and correction for storage in a preset check area of said memory array;
- said circuit for comparing the data read from said monitor cell with said preset data setting said refresh period to a longer value within the range allowed for error generation by an error correction operation employing check bits;
- said error detection correction encoding/decoding circuit performing an error bit correction and decoding of said memory array, using said check bits, by said error detection correction encoding/decoding circuit, before reversion to the normal operation in exiting from the data retention mode.
12. The semiconductor memory device according to claim 1, further comprising:
- an ECC (error-checking and correction) encoding/decoding circuit for performing error checking and correction of data stored in said memory array;
- wherein data of said memory array is read at the time of entry to a data retention operation mode;
- said error detection correction encoding/decoding circuit generates check bits for error detection and correction and stores the check bits generated in a preset check area of said memory array;
- (A1) an expected value data ‘1’ is written in said monitor cell by said circuit for writing preset data in predetermined plural memory cells;
- (A2) all cells of said memory array are refreshed by a refresh circuit connected to said memory array;
- (A3) the operation is paused during the refresh time by said refresh circuit;
- (A4) data is read from said monitor cell by said circuit for performing control for reading data, and compared with the expected value data ‘1’ by said circuit for comparing the data read from said monitor circuit with said preset data to measure an error; and
- (A5) when said error rate is larger than a preset upper limit value, the refresh period is shortened by said circuit for comparing the data read from said monitor circuit with said preset data, when said error rate is not larger than a preset lower limit value, the refresh period is elongated by said circuit for comparing the data read from said monitor circuit with said preset data and, when said error rate is intermediate between said preset upper and lower limit values, the refresh period is not changed by said circuit for comparing the data read from said monitor circuit with said preset data;
- the operations from (A1) to (A5) are repeated until exiting from said data retention mode; and wherein
- when exiting from the data retention mode, error correction is carried out in said error detection correction encoding/decoding circuit before proceeding to the normal operation.
13. The semiconductor memory device according to claim 1, further comprising:
- an ECC (error-checking and correction) encoding/decoding circuit for performing error checking and correction of data stored in said memory array;
- wherein data of said memory array is read at the time of entry to a data retention mode,
- said ECC encoding/decoding circuit generates check bits for error detection and correction for storage in a preset check area of said memory array;
- (A1) an expected value data ‘1’ is written in said monitor cell by said circuit for writing preset data in predetermined plural memory cells;
- (A2) all cells of said memory array are refreshed by a refresh circuit connected to said memory array, the pause period is made shorter than the refresh period and not larger than a preset number (m) times of error check operations are carried out; that is,
- (A3) at least a part of the internal power supply is turned off by a timer to make a pause for a preset pause period;
- (A4) data is read from said monitor cell by said circuit for performing control for reading data, and compared with the expected value data ‘1’ by said circuit for comparing the data read from said monitor circuit with said preset data to measure an error;
- (A5) when said error rate is larger than a preset upper limit value, the refresh period is shortened by said circuit for comparing the data read from said monitor circuit with said preset data and, when said error rate is intermediate between said preset upper and lower limit values, the refresh period is not changed by said circuit for comparing the data read from said monitor circuit with said preset data and the processing as from (A1) is commenced; and
- (A6) if, after m times of the error check operations, said error rate is not larger than a preset lower limit value, the refresh period is elongated by said circuit for comparing the data read from said monitor circuit with said preset data; wherein the operations of (A1) until A(6) are repeated until exiting from the data retention mode; and wherein
- when exiting from the data retention mode, error correction is carried out in said ECC encoding/decoding circuit before proceeding to the normal operation.
14. The semiconductor memory device according to claim 1, further comprising:
- an ECC (error-checking and correction) encoding/decoding circuit for performing error checking and correction of data stored in said memory array; wherein data of said memory array is read at the time of entry to a data retention mode, said ECC encoding/decoding circuit generating check bits for error detection and correction for storage in a preset check area of said memory array;
- (A1) all cells of said memory array are refreshed by a refresh circuit and the pause period is made shorter than the refresh period;
- (A2) it is checked by said ECC encoding/decoding circuit whether or not an error check has been carried out a preset number (m) of times;
- if the number of times is less than m, then
- (A3) all cells of said memory array are refreshed by said refresh circuit;
- (A4) pause is made during the pause period by said refresh circuit, based on a timer value provided by a timer circuit;
- (A5) data is read from said monitor cell by said circuit for performing control for reading data, and compared with the expected value data ‘1’ by said circuit for comparing the data read from said monitor circuit with said preset data to measure an error;
- (A6) when said error rate is larger than a preset upper limit value, the refresh period is shortened by said circuit for comparing the data read from said monitor circuit with said preset data, when said error rate is intermediate between said preset upper and lower limit values, the refresh period is not changed by said circuit for comparing the data read from said monitor circuit with said preset data and, when said error rate is not larger than a preset lower limit value, the refresh period is elongated by said circuit for comparing the data read from said monitor circuit with said preset data and the processing as from (A2) is started;
- after the end of m times of the error check, the processing as from (A1) above is performed;
- the operations of (A1) until A(6) are repeated until exiting from the data retention mode; and wherein
- when exiting from the data retention mode, error correction is carried out in said ECC encoding/decoding circuit before proceeding to the normal operation.
15. The semiconductor memory device according to claim 1, further comprising:
- an ECC (error-checking and correction) encoding/decoding circuit for performing error checking and correction of data stored in said memory array; wherein data of said memory array is read at the time of entry to a data retention mode, said ECC encoding/decoding circuit generating check bits for error detection and correction for storage in a check area of said memory array;
- (A1) an expected value data ‘1’ is written in said monitor cell by said circuit for writing preset data in predetermined plural memory cells;
- (A2) all cells of said memory array are refreshed by a refresh circuit; the pause period is made shorter than a refresh period;
- (A3) it is judged whether or not by said ECC encoding/decoding circuit whether error check has been carried out a preset number (m) of times;
- if the number of times is less than m,
- (A4) a pause is made for a preset pause period by said refresh circuit, based on a timer value provided by a timer circuit;
- (A5) data is read from said monitor cell by said circuit for performing control for reading data, and compared with the expected value data ‘1’ by said circuit for comparing the data read from said monitor circuit with said preset data to measure an error;
- (A6) when said error rate is larger than a preset upper limit value, the refresh period is shortened by said circuit for comparing the data read from said monitor circuit with said preset data, when said error rate is intermediate between said preset upper and lower limit values, the refresh period is not changed by said circuit for comparing the data read from said monitor circuit with said preset data; when said error rate is not larger than a preset lower limit value, the refresh period is elongated by said circuit for comparing the data read from said monitor circuit with said preset data and the processing as from (A3) is commenced;
- the processing as from (A1) is carried out after the end of m times of error check operations;
- the operations of (A1) until A(6) are repeated until exiting from the data retention mode; and wherein
- when exiting from the data retention mode, error correction is carried out in said ECC encoding/decoding circuit before proceeding to the normal operation.
16. The semiconductor memory device according to claim 1, wherein
- plural storage locations of said memory array inferior in the capability of the pause refresh to the usual memory area are used as said monitor cell.
17. The semiconductor memory device according to claim 16, wherein
- storage locations of said memory array inferior in the capability of the pause refresh than the usual memory area are those bits obtained on carrying out bit restoration by pause refresh for a preset number (M) of bits and selecting approximately one-tenth of said preset number M of bits of the worst capability.
18. The semiconductor memory device according to claim 1, wherein said an error correction decoder comprises:
- a plurality of half-adder circuits;
- a plurality of 2-bits adder circuits configured to receive outputs from a respective pair of said plurality of half-adder circuits; and
- a plurality of 3-bits adder circuits configured to receive outputs from a respective pair of said plurality of 2-bits adder circuits.
Type: Application
Filed: Jan 9, 2009
Publication Date: Jul 30, 2009
Applicant:
Inventors: Yutaka Ito (Tokyo), Takeshi Hashimoto (Tokyo)
Application Number: 12/318,840
International Classification: H03M 13/05 (20060101); G06F 11/00 (20060101); G06F 11/07 (20060101);
