JPH0991972A - Nonvolatile multilevel memory - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain a nonvolatile multilevel memory in which a selection can be made between the reliability of data and the high density recording for each data by varying the memory resolution in same memory cell array depending on the type of data. SOLUTION: Voice data of 4 bit unit is outputted from an ADPCM decoder 2 while an address data of 1 bit unit storing a voice data is outputted from an address controller 9. A number of bit conversion circuit 13 converts the address data of 1 bit unit into a 4 bit address data of same level which is inputted to a second multiplexer along with the 4 bit voice data. Output from the second multiplexer is selected based on a switching signal NTSEL delivered from the address controller 9 and a selected data is delivered to a read/write circuit. Consequently, the voice data is stored in an EEPROM memory cell 3 with memory resolution of 16 values while the address data is stored with memory resolution of 2 values.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a nonvolatile multi-level memory device using an EEPROM or the like capable of storing multi-level information.

[0002]

2. Description of the Related Art EEPR with floating gate
In a non-volatile memory such as an OM, it has been conventionally practiced to change the threshold level by controlling the amount of charges injected into the floating gate and store the analog amount and multi-valued information in the memory cell.

For example, in Japanese Patent Application Laid-Open No. 4-500576, while an input analog signal is sampled and held by an analog sample and hold circuit, a high voltage write pulse is supplied to a nonvolatile memory cell to charge the floating gate. After the injection, the analog amount corresponding to the injected charge is read out, compared with the analog signal that was sampled and held, and the supply of the write pulse is repeated until both analog amounts match, corresponding to the input analog voltage The analog amount to be recorded is recorded in a memory cell.

Further, in Japanese Patent Publication No. 4-57294,
A sense amplifier that latches the input digital data with the data latch circuit and reads the multi-value storage state of the memory cell and outputs the digital value corresponding to the storage state is provided. The output of this sense amplifier and the data held by the data latch circuit The comparator is compared with each other, and the writing operation of the multi-valued information to the memory cell is continued until both contents match.

In either case, the storage resolution of analog quantity or multi-valued information was constant.

[0006]

When a multi-valued memory is used, n-bit digital data such as a voice signal can be stored in one memory cell as multi-valued information of the n-th power of 2. Therefore, the digital value remains unchanged. The memory capacity can be much reduced as compared with the case of binary storage. by the way,
When voice data or the like is stored in a multi-valued memory, it may be desirable to store address information indicating where in the memory the data is stored for later reading. Original data such as voice data is not a big problem because the voice changes only slightly even if some errors occur due to writing and reading in the memory, and it is rather a problem of storage for the purpose of reducing the storage capacity. It is desired that the resolution be high. On the other hand, the address information is data that requires extremely high reliability because the read position itself changes if an error occurs.

However, in the conventional example, since the storage resolution is always constant, the data for which high density recording such as voice is desired and the data such as address data for which reliability is indispensable have the same resolution. I had to remember. For this reason, there is a contradictory problem that the reliability of address data and the like decreases when the resolution is increased, and the recording density of voice data and the like decreases when the resolution is decreased.

[0008]

According to the present invention, in a nonvolatile multi-valued memory device capable of storing a plurality of types of digital data as multi-valued information in a non-volatile memory cell array, a storage resolution according to the type of digital data to be stored. The above-mentioned problem is solved by providing a switching circuit for switching.

In the present invention, the plurality of types of digital data are first type digital data and address data indicating a storage address of the first type digital data, and the storage resolution of the address data is the first type. It is characterized in that it is lower than the storage resolution of the seed digital data. In the present invention, the first-type digital data has n bits (n:
A data generation circuit for sequentially outputting in units of 2 or more),
An address controller that sequentially outputs the address data in units of m bits (m: an integer of m <n) and generates a switching signal in accordance with the type of data to be written, and an n-bit digital data that is input corresponds to the address controller. A write circuit for writing to the nonvolatile memory cell array as value information, wherein the switching circuit outputs the n-bit (n: integer of 2 or more) first type digital data when the switching signal is at a first level. A switching circuit which outputs the signal to the writing circuit as it is, converts the m-bit digital data into n-bit digital data with the upper m bits of the digital data and outputs the digital data when the switching signal is at the second level. And

In the present invention, the switching circuit is a conversion circuit for converting the m-bit (m = 1) digital data into n-bit digital data in which all bits have the same level as the m-bit digital data. It is characterized by including.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows a schematic block diagram when the present invention is applied to a voice recording / reproducing apparatus. First, 1 is an AD converter that samples an input analog audio signal at a predetermined sampling period and sequentially converts it into 12-bit digital audio data, and 2 is an input 12-bit digital audio data that is sequentially 4-bit digital An ADPCM encoder for encoding and outputting compressed data VODAT, an EEPROM cell array 3 having a floating gate and capable of storing multilevel information, 40, 41, 42, ...
Is a plurality of read / write circuits R / that write the inputted 4-bit digital data as multi-valued information into the EEPROM cell array 3 and read out the written multi-valued information.
W, 50, 51, 52, ... Are provided for each read / write circuit, and are X address decoders for designating the X address of the EEPROM cell array 3, and 6, 7 are for word lines WL for designating the Y address of the EEPROM cell array 3. And a Y address decoder for the source line SL, 8 a microcomputer interface circuit for interpreting a command from a microcomputer, 9 a read / write circuit 40, 41, 42, ... And an X address decoder based on a command from the microcomputer interface circuit 8. The address controller 10 supplies the X address ADRX and the control signal to 50, 51, 52, ... And the Y address ADRY to the Y address decoders 6 and 7, and 10 is a 4-bit digital compressed data RDAT read from the EEPROM cell array 3. 12-bit digital audio ADPCM decoder for decoding the data, 11 is a DA converter that converts 12-bit digital audio data DA into an analog audio signal, the converted analog signal is sounded as a sound from a speaker or the like (not shown).

In this device, as shown in FIG.
The EPROM cell array 3 has a first area 31 for storing the audio data VODAT output from the ADPCM encoder 2 and a second area 32 for storing a start address and a stop address indicating the start and end addresses of the stored audio data. are doing. Then, in order to store such start and stop addresses, the address controller 9 outputs the start address and the stop address as address data in the write mode. However, unlike the audio data output every 4 bits, the address data is output every 1 bit, and X, Y
The total 20-bit address is output 20 times. The address controller 9 has a 4-bit down counter 90, and sequentially outputs the 4-bit down count data DWDAT output from the down counter 90 in the read mode.

Further, the address controller 9 switches the read / write control signal R / W which becomes H level in the read mode and becomes L level in the write mode, and H level when the audio data is written and becomes L level when the address data is written. The signal NTSEL is output. The control signal R / W, the switching signal NTSEL, the 1-bit address data, and the 4-bit audio data VODAT are input to the switching circuit 12.

The switching circuit 12 converts the address data output in 1-bit units into 4-bit address data ADDA.
A bit number conversion circuit 13 for converting into T, the converted 4-bit address data ADDAT and 4-bit down-count data DWDAT are input, and either one is selectively output according to a read / write control signal R / W. The 4-bit data output from the first multiplexer 14 and the first multiplexer 15 and the 4-bit data output from the ADPCM encoder 2
Bit audio data VODAT is input and switching signal NT
It is composed of a second multiplexer 15 which selects and outputs any one of 4-bit data according to SEL. The bit number conversion circuit 13 is, for example, as shown in FIG.
As shown in FIG. 5, a configuration is adopted which includes five inverters 121 to 125 and outputs the input 1-bit address data and 4-bit address data in which all bits are at the same level.

The operation of the apparatus shown in FIG. 1 will be described in detail below. First, when a voice data write command is given from the microcomputer interface 8 to the address controller 9, the write mode is set, and the address controller 9 sets the read / write control signal R / W to the L level. The output of the conversion circuit 13 is selected. Next, the address controller 9 sets the switching signal NTSEL to the H level and
X address ADRX indicating the address to be written
And the Y address ADRY are sequentially output. In the second multiplexer 15, in response to the switching signal NTSEL being at the H level, the audio data V input in units of 4 bits.
ODAT is selected, and a plurality of read / write circuits 40, 4 are selected.
1, 42 ... are sequentially output.

A plurality of read / write circuits 40, 41, 42
......, 4-bit audio data VOD that is sequentially input
ATs are sequentially fetched and held in a data register provided in each read / write circuit. Then, when the reading into the predetermined number of read / write circuits is completed, the predetermined number of read / write circuits simultaneously execute the write operation, and the held 4-bit audio data VODAT is converted into 16 bits.
Converted to discrete analog quantity of values, then converted 16
EEP the analog amount of the value through the X address decoder
Writing to each memory cell of the ROM cell array 3.

Therefore, in this writing operation, the storage resolution of the audio data VODAT is "16". When the writing of the voice data is completed in this way, the microcomputer interface 8 issues a write stop command, and the address controller 9 responds to this by issuing a switching signal NTSE.
Then, L is set to L level, and then the output of the start address and stop address in which the voice data is stored as address data is started in 1-bit units. This 1-bit address data is converted by the bit number conversion circuit 13 into "1111" if it is "1" and "000" if it is "0".
All bits are converted into 4-bit address data ADDAT having the same level as the input data, such as "0". In the second multiplexer 15, the output of the bit number conversion circuit 13 is selected in response to the switching signal NTSEL becoming L level, so that the 4-bit data of “1111” or “0000” is input to the read / write circuit. It will be.

That is, in this case, the data "111"
Data is EEP in binary corresponding to "1" and "0000"
Since the data is stored in each memory cell of the ROM cell array 3, the storage resolution is "2", which is lower than the audio data "16". When writing this address data, the addresses ADRX and ADRY corresponding to the second area 32 of the EEPROM cell array 3 are output from the address controller 9.

On the other hand, when a read command is given from the microcomputer interface 8, the read mode is entered and the address controller 9 sets the read / write control signal R / W to the H level, so that the first multiplexer 14 causes the 4-bit down count data DWDAT. Will be selected. Further, the address controller 9 sequentially outputs the down count data DWDAT, and the switching signal NT
Down count data DWDA with SEL at L level
T is output through the second multiplexer 15. The output of the down-count data DWDAT is for AD-converting the multivalued information read by the read / write circuit into 4-bit digital data, and this operation will be described in detail later.

In this reading, the address controller 9 first specifies the addresses ADRX, ADRY corresponding to the second area 32 of the EEPROM cell array 3 and executes the read operation to the read / write circuits 40, 41, 42, .... Then, the start address and the stop address stored in the second area 32 are read. in this case,
The read / write circuit outputs 4-bit data RDAT by reading, but only the most significant bit D3 thereof is input to the address controller 9 and this bit information is taken in as address data. That is, even if the read data has any value in the range of “1000” to “1111”, the fetched address data is “1” and the read data has any value in the range of “0000” to “0111”. If so, the address data to be fetched will be "0". Therefore, as described above, the storage resolution of the address data is "2".

When the reading of the start address and the stop address is completed, the address controller 9 outputs the above-mentioned down count data DWDAT, and sequentially designates the read addresses ADRX and ADRY from the read start address to the stop address for reading. Since the read operations are executed by the write circuits 40, 41, 42, ..., 16-valued multivalued information stored in the first area 31 of the EEPROM cell array is read out as 4-bit audio data, and the ADPCM decoder 10 is read. Is output. Then, decompression processing is performed here, the original digital audio data of 12 bits is decoded, and DA of the next stage is used.
The digital audio data decoded by the converter 11 is converted into an analog audio signal and output. In this case, 16
Since the discrete analog amount of the value is converted into the original 4-bit digital data, the storage resolution becomes "16".

As described above, one EEPROM
In the cell array 3, since the voice data is stored with high resolution, high density recording can be realized, and the address data is stored with low resolution, so that the reliability of the data can be secured. Next, the read / write circuits 40, 41, 42, ...
A specific configuration of the above will be described with reference to FIG.

In FIG. 4, reference numeral 20 denotes a D flip-flop, which is output from the second multiplexer 15 as 4
A 4-bit data register for fetching and holding bit digital data, 21 is a reference voltage Vref from V0 to V15
(V0 <V1 <... <V14 <V15) 16-step voltage dividing circuit for dividing the voltage, and 22 for the data register 20
A decoder which decodes the contents of the above and selectively outputs any one of the voltages V0 to V15 corresponding to the contents, 23 inputs the analog voltage Vdec output from the decoder 22 to the non-inverting terminal +, and the memory of the EEPROM 3 A comparator for inputting the voltage Vm read from the cell 60 to the inverting terminal − and comparing the two voltages, 24 outputs the output of the comparator 23 as it is while the timing clock RWCK4 is at the H level, and outputs L
Latches the output of the comparator when falling to the level
Latch circuit that outputs the output latched during the level period, 2
Reference numeral 5 is an output buffer for outputting the contents of the data register 20, and the resistance division circuit 21 and the decoder 22 constitute a DA converter during a write operation.

The memory cell 60 of the EEPROM 3 is a split gate type cell provided with a floating gate FG, and writing is performed by injecting charges into the floating gate FG to perform writing.
The erase is performed by extracting the charge injected into G. Each memory cell 60 has its drain D connected to bit lines BL1, BL2,..., Its source S connected to source lines SL1, SL2,..., And its control gate CG connected to word lines WL1, WL2,.
It is connected to the. Each bit line BL1, BL2,.
Any one of the lines is selected by the X address decoder 50 that decodes the X address ADRX [8: 5] of the upper 4 bits and connected to the inverting terminal of the comparator 23. The word lines WL1, WL2, ... And the source lines SL1, SL2 ,.
It is connected to Y address decoders 6 and 7 for decoding the address [10: 0], and various bias voltages are supplied from the second bias generation circuit 400 to these decoders. This bias voltage includes a high voltage bias Vhv1 for writing and a high voltage bias Vhv2 for erasing.

The address decoders 50, 6 and 7 include
RWCK3, RWCK4, WBE as timing signals
Has been entered. The terms drain and source here are based on the operating state at the time of reading. Three types of bias voltages VBH, VBLH, VBLL (VBH>VBLH>) supplied to the bit lines BL1, BL2, ....
VBLL) is output from the first bias generation circuit 500, and a P-channel MOS transistor 26, an N-channel MOS transistor 27, and an N-channel MOS transistor 28 are inserted as switches in the supply lines of these bias voltages, respectively. . An analog switch 29 that is turned on only at the time of writing is connected to the output side of these transistors, and the output of this analog switch 29 is an input / output line 30 to the X address decoder 100.
It is connected to the. P-channel MOS transistor 26
The output of the AND gate 31 which receives the output COMP of the latch circuit 24 at one end is applied to the gate of, and the outputs of the AND gates 32 and 33 are applied to the N-channel MOS transistors 27 and 28, respectively. The outputs of the AND gate 31 are commonly input to one ends of the AND gates 32 and 33, and a signal obtained by inverting the upper bit D1 to the data register 20 by the inverter 34 is input to the other end of the AND gate 32. The upper bit D1 to the data register 20 is directly input to the other end of the AND gate 33.

Further, a read bias generation circuit 35 constituted by a resistance division circuit is provided for reading the analog amount written in the memory cell 60 as a voltage, and the voltage dividing point P thereof is turned on only at the time of comparison. N channel MO
Via the S transistor 36, the X address decoder 10
0 is connected to the input / output line 30. An N-channel MOS transistor 37 that is turned on by a control signal WBE is inserted between the input / output line 30 and the ground to supply a ground potential to the bit lines BL1, BL2,.

By the way, the read / write circuit shown in FIG. 4 manages eight memory cells in the X address direction as one block, and each block is a block for detecting that its own block is selected. Selector 600
Is arranged. The block No. shown in FIG. In the block of 0, the block selector 600 is configured by an AND gate that detects that the lower 6 bits of the X address ADRX [5: 0] are all “0”.

Further, in FIG. 4, 38 is a sampling clock RWCK2 and a latch enable signal LATEN.
To input the output BSEL of the block selector 600 and N
An AND gate 39 is a NAND gate for inputting the timing clock RWCK3, the read enable signal REAEN2 and the output COMP, and 40 is a block selector 60
0 output BSEL and read enable signal REAEN2
, 41 is a NAND gate for inputting the outputs of the two NAND gates 38 and 39, 42
Is an AND gate for inputting a timing clock RWCK3 and a write enable signal WREN2, and 43 is a read enable signal REAEN2 and a write enable signal W
An OR gate 44 for inputting RIEN2 and an A 44 for inputting the timing clock RWCK4 and the output of the OR gate 43
An ND gate, the output of the NAND gate 41 is applied to the clock terminal of a D flip-flop constituting the data register 20, the output of the NAND gate 40 is applied as an on / off control signal of the output buffer 25, and the output of the AND gate 42 is applied. This signal is applied as an on / off control signal for the analog switch 29, and the output of the AND gate 44 is applied to the N-channel M
The voltage is applied to the gate of the OS transistor 36.

The write operation and read operation of the read / write circuit shown in FIG. 4 will be described below with reference to the timing charts of FIGS. Memory cell 6
Bias conditions in each operation state of 0 are as shown in FIG. First, in the write mode, a latch period for latching data in the data register 20 is entered prior to the actual write operation. During this period, 4-bit digital data D3, D2, D1, D0 is input line 4
EEPR to send data to and write data to
The addresses ADRX and ADRY of the OM6 are sent from the address generation circuit 10 and the signal L indicating the latch mode is output.
ATEN goes high. The lower 6 bits ADRX [5: 0] of the output X address are the same as the block No. of its own. , The output of the block selector 600 becomes H level, and therefore the sampling pulse RWC
At the rising edge of K2, the output of the NAND gate 38 becomes L level and the output of the NAND gate 41 also becomes L level. Therefore, a clock is applied to the clock terminal CK of the D flip-flop constituting the data register 20, and the input data D1 and D0 are taken into the data register 20.

When the capturing is completed, the signal WBE becomes H level, the N-channel MOS transistor 37 is turned on, and the input / output line 30 becomes the ground potential 0V. In the X address decoder 100, since the bit line selected by the X address ADRX [8: 5] is connected to the input / output line 30, the bit line BL becomes 0V.
On the other hand, a high voltage bias Vhv2 for erasing is applied to the selected word line WL by the Y address decoder 250, and the Y address decoder 20 is applied to the source line SL.
Since 0 to 0 V is applied, the selected memory cell is in the erased state. That is, the charge to the floating gate FG of the memory cell 60 is drawn.

After such erasing, the actual writing operation is started. In the write operation period, as shown in FIG. 5C, the signal WRIEN2 becomes H level, and therefore the output of the AND gate 42 becomes H level while the clock RWCK3 is H level as shown in FIG. Furthermore, since the latch circuit 24 is initially set to the H level, the output of the AND gate 31 also becomes the H level. Therefore,
When the analog switch 29 is turned on, the P-channel M
The OS transistor 26 turns off.

If the most significant bit D3 of the input data is "0", the output of the AND gate 32 becomes H level, the N channel MOS transistor 27 is turned on, and the bias voltage is changed as shown in FIG. VBLH is supplied to the selected bit line BL via the analog switch 29, the input / output line 30, and the X address decoder 50. On the contrary, if the most significant bit D3 of the input data is "1", the output of the AND gate 33 becomes H level,
The N-channel MOS transistor 28 is turned on, the bias voltage VBLL is changed to the analog switch 29, the input / output line 3
It is supplied to the selected bit line BL via the 0, X address decoder 50.

While the clock RWCK3 is at H level,
Source line S selected by Y address decoder 6
Since the high voltage Vhv1 is supplied to L (FIG. 5C) and VB2 is supplied to the word line WL selected by the Y address decoder 7 (FIG. 5C), the write bias condition shown in FIG. Writing to the cell 60 is executed. That is, the floating gate F of the memory cell 60
The injection of electric charge into G is started.

Next, the clock RWCK3 falls,
When the clock RWCK4 becomes H level as shown in FIG.
When the output of the AND gate 42 is at L level, the AND gate 44
Becomes an H level, the analog switch 29 is turned off, the N-channel MOS transistor 36 is turned on,
The voltage dividing point P of the read bias generation circuit 35 is connected to the input / output line 30. The potential at the voltage dividing point P is N channel MO
When the S transistor 36 is off, the voltage VREFM is set slightly higher than V4. Further, in this state, the word line WL selected by the Y address decoder 7 is selected.
Is applied to the source line SL and 0 V is applied to the source line SL from the Y address decoder 6, so that the selected memory cell 60 is in a read state. Therefore, the voltage Vm corresponding to the charges injected into the floating gate FG of the selected memory cell is obtained in the input / output line 30, and this voltage Vm is compared with the output voltage Vdec from the decoder 22 in the comparator 23.

In the decoder 22, one of the 16 levels of voltage V0 to V15 from the resistance division circuit 21 is selected in accordance with the data latched in the data register 20, and the analog voltage of the comparator 23 is not inverted. It is output to the terminal. If Vdec> Vm as a result of the comparison, the output of the comparator 23 maintains the H level, and the clock RWCK3 described above is used.
Based on the clock RWCK4 and the read and compare operations based on the clock RWCK4 are repeated. By repeating the write operation, the amount of charges injected into the floating gate FG increases, and the read voltage Vm increases as shown in FIG. Then, when Vdec ≦ Vm, as shown in FIG. 5, the output of the comparator 23 is inverted to L level, and the output COMP of the latch circuit 24 also becomes L level. Therefore, the output of the AND gate 31 is inverted from the H level to the L level, the P-channel MOS transistor 26 is turned on,
Further, the outputs of the AND gates 32 and 33 become L level, and the two N channel MOS transistors 27 and 28 are turned off. Therefore, when the clock RWCK3 next becomes H level, the bias voltage VBH is supplied to the bit line BL of the memory cell via the analog switch 29 (see FIG. 5C). That is, the write bias condition shown in FIG. 7 is broken and the write operation is stopped.

As described above, in the write mode, a 16-value analog amount corresponding to 4-bit input digital data is stored in the selected memory cell 60. Next, the operation in the read mode will be described with reference to FIG. In the read mode, first, the signal XSET (FIG. 6C) becomes the H level, so that the initial value all “1” is set in the data register 20 (FIG. 6E), and from the decoder 22 as shown in FIG. To
The analog voltage V15 corresponding to all "1" is output. Therefore, when the clock RWCK4 becomes the H level as shown in FIG. 5, the bias condition for the memory cell 60 becomes exactly the same as that in the read operation in the write mode, so that it corresponds to the charge injected into the floating gate of the selected memory cell. The voltage Vm to be applied is obtained at the inverting terminal of the comparator 23, and this voltage Vm is compared with the voltage V4 from the decoder 22. If Vm> V4 as a result of the comparison, the output COMP of the comparator 23 and the latch circuit 24 becomes L level, so the output of the NAND gate 39 becomes H level, and at this time, the output of the NAND gate 38 is fixed at H level. Therefore, the output of the NAND gate 41 becomes the L level, and the latch operation is not performed thereafter, and all “1” s are retained in the data register 20.

On the other hand, if the comparison result is Vm ≦ V4, the output COMP of the comparator 23 and the latch circuit 24 becomes H level, so that when the clock RWCK3 becomes H level as shown in FIG. Output becomes L level, so that a clock signal is output from the NAND gate 41 to the data register 20, and the data supplied to the data input line 45 is latched in the data register 20. The data input line 45 is read from the down counter 90 shown in FIG.
“0”, “1101”, “1100”, ………, “000
1 ”,“ 0000 ”data“ D3, D2, D1, D
Since "0" is sequentially output each time the clock RWCK4 falls, the data "111" is followed by the data "111".
"0" will be latched in the data register 20 as shown in FIG. Then, the output Vdec of the decoder 22
6 decreases to the voltage V14 as shown in FIG. 6 and when the clock RWCK4 becomes the H level again, the voltage Vm and the voltage V14 corresponding to the analog amount read from the memory cell.
Are compared. If Vm> V14, the comparator 2
3 and the output COMP of the latch circuit 24 are inverted to the L level, and "1110" is held in the data register 20 without performing the latch operation thereafter. Result of comparison Vm ≦
At the time of V14, the output COMP of the comparator 23 and the latch circuit 24 maintains the H level, so the next data "11"
01 ”is latched in the data register 20, and the comparator 23
Then, the voltages V13 and Vm are compared. By this comparison, V
If m> V13, the content of the data register 20 is "11.
It is fixed to 01 ”and if Vm ≦ V13, the next data“ 1100 ”is further latched in the data latch 20 and the voltages Vm and V12 are compared. When the above operation is repeated and "0111" is latched in the data register 10 and Vm> V7 in the comparison, the output COMP of the comparator 23 and the latch circuit 24 is inverted to the L level,
The contents of the data register 10 are "01" as shown in FIG.
It is fixed to 11 ".

As described above, the voltage Vm corresponding to the analog amount read from the memory cell is stored in the data register 2
0, resistance dividing circuit 21, decoder 22, comparator 23, N
The signal is AD-converted by the AND gate 39 and the NAND gate 41 and transferred to the outside via the output buffer 25.

[0039]

According to the present invention, the storage resolution can be changed according to the type of data. Therefore, while the reliability of data of a certain type can be secured in the same memory cell array, It is possible to satisfy the contradictory requirements of realizing high-density recording for data of different types.

[Brief description of drawings]

FIG. 1 is an overall block diagram of a voice recording / playback apparatus to which the present invention is applied.

FIG. 2 is a circuit diagram showing a specific configuration of a bit number conversion circuit.

FIG. 3 is an explanatory diagram showing a data storage structure of an EEPROM.

FIG. 4 is a circuit diagram showing a specific configuration of a read / write circuit according to the present invention.

FIG. 5 is a timing chart showing the operation of the read / write circuit of the present invention in the write mode.

FIG. 6 is a timing chart showing an operation in a read mode of the read / write circuit according to the present invention.

FIG. 7 is a diagram showing a bias condition of a memory cell in the present invention.

[Explanation of symbols]

1 AD converter 2 ADPCM encoder 3 EEPROM cell array 40, 41, 42, ... Read / write circuit 50, 51, 52, ... X address decoder 6 Y address decoder (SL) 7 Y address decoder (WL) 8 Microcomputer interface Circuit 9 Address Controller 10 ADPCM Decoder 11 DA Converter 12 Switching Circuit 13 Bit Number Conversion Circuit 14 First Multiplexer 15 Second Multiplexer 20 Data Register 21 Resistance Dividing Circuit 22 Decoder 23 Comparator 24 Latch Circuit 25 Output Buffer 26 P-Channel MOS Transistor 27 , 28, 36, 37 N-channel MOS transistor 29 Analog switch 60 Memory cell 90 Down counter 400 Second bias generation circuit 500 First via Generation circuit 600 block selector

Claims (4)

[Claims] 1. A non-volatile multi-valued memory device capable of storing a plurality of types of digital data as multi-valued information in a non-volatile memory cell array, comprising a switching circuit for switching the storage resolution according to the type of digital data to be stored. The non-volatile multi-valued memory device according to claim 1. 2. The plurality of types of digital data are the first
7. A type of digital data and address data indicating a storage address of the first type of digital data, wherein the storage resolution of the address data is lower than the storage resolution of the first type of digital data. 2. The nonvolatile multi-valued memory device according to 1. 3. A data generation circuit for sequentially outputting the first type digital data in units of n bits (n: an integer of 2 or more), and the address data in units of m bits (an integer of m: m <n). An address controller that sequentially outputs and generates a switching signal according to the type of data to be written, and a write circuit that writes input n-bit digital data as corresponding multi-valued information into the nonvolatile memory cell array are provided. The switching circuit has n bits (n: an integer of 2 or more) when the switching signal is at the first level.
Of the first type of digital data are directly output to the write circuit, and when the switching signal is at the second level, the m-bit digital data is converted into n-bit digital data having the upper m bits as the digital data. 3. The non-volatile multi-valued memory device according to claim 2, comprising a switching circuit for outputting. 4. The switching circuit comprises the m bits (m =
4. The nonvolatile multi-valued memory device according to claim 3, further comprising a conversion circuit for converting the digital data of 1) into n-bit digital data in which all bits have the same level as m-bit digital data.