JP2006228432A - Nonvolatile semiconductor memory - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a nonvolatile semiconductor memory in which rewriting in units of bytes can be performed. <P>SOLUTION: A memory cell array has a unit formed of one memory cell and one select transistor. One block has one control gate and memory cells connected to one control gate line form one page. A sense amplifier having a latch function is connected to a bit line. In rewriting data, data of the memory cell of one page is read to the sense amplifier. After data are superscribed on data in the sense amplifier and a page erase is performed, data of the sense amplifier is written in the memory cell of one page. Data rewriting in the units of bytes can be performed by superscription of the data in the sense amplifier. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a nonvolatile semiconductor memory that rewrites data in units of bytes.

  Conventionally, an EEPROM is known as a nonvolatile semiconductor memory that rewrites data in byte units.

  Non-Patent Document 1 proposes an EEPROM that uses FLOTOX (Floating Gate Tunnel Oxide) cells and rewrites data in byte units.

  65 is a plan view showing an example of a memory cell portion of an EEPROM capable of byte erasure, and FIG. 66 is a cross-sectional view taken along line LXVI-LXVI in FIG.

This EEPROM uses FLOTOX cells in the memory cell portion. The FLOTOX cell is characterized in that a tunnel oxide film 22a of about 10 nm is disposed between the N ⁺ drain 20a and the floating gate 21a, and an electric field is applied to the tunnel oxide film 22a to thereby form the N ⁺ drain 20a and the floating gate 21a. It is in the point of exchanging electric charge between them.

  The current flowing through the tunnel oxide film 22a is an FN tunnel current generated by an FN (Fowler-Nordheim) tunnel phenomenon.

  FIG. 67 shows an energy band diagram of the MOS capacitor portion.

When an electric field is applied to the MOS capacitor (N ⁺ drain-tunnel oxide film-floating gate), an FN tunnel current flows through the tunnel oxide film (SiO ₂ ) based on the equation (1).

I = S · α · E ² exp (−β / E) (1)
S: Area, E: Electric field α = q ³ /8πhΦB=6.94×10 ⁻⁷ [A / V ² ]
β = −4 (2 m) ^0.5 ΦB ^1.5 / 3hq
= 2.54 × 10 ⁸ [V / cm]
From this equation, it can be seen that the electric field at which the FN tunnel current starts to flow is about 10 [MV / cm]. This electric field theoretically corresponds to a case where a voltage of 10 [V] is applied to a tunnel oxide film of 10 [nm].

65 and 66, the capacitance ratio (coupling ratio) between the control gate 23a and the floating gate 21a when a voltage is applied between the N ⁺ drain 20a and the control gate 23a is 0.5.

In this case, in order to apply a voltage of 10 [V] to the tunnel oxide film 22a between the N ⁺ drain 20a and the floating gate 21a, a high voltage of 20 [V] is applied between the N ⁺ drain 20a and the control gate 23a. Must be applied.

For example, at the time of erasing, the N ⁺ drain 20a is set to 0 [V] and the control gate 23a is set to 20 [V] to move electrons from the N ⁺ drain 20a to the floating gate gate 21a. At the time of writing “1”, the N ⁺ drain 20a is set to 20 [V] and the control gate 23a is set to 0 [V] to move electrons from the floating gate 21a to the N ⁺ drain 20a.

  A drawback of the EEPROM using the FLOTOX cell is that, as shown in FIGS. 65 and 66, two elements of a memory cell and a selection transistor are required to store one bit.

  FIG. 68 shows another example of the memory cell portion of the EEPROM capable of byte erasure.

  This EEPROM is characterized in that a FLOTOX cell is used in the memory cell portion, and one byte control transistor Tr is provided for 8 bits (1 byte) of the memory cell.

The bias conditions in each mode in this EEPROM are as shown in Table 1.

  When such a memory cell portion is used, various malfunctions (disturbances) can be avoided. However, since 2+ (1/8) transistors are required to store 1 bit, there is a disadvantage that the cell area is increased and the cost cannot be reduced.

  A memory born to eliminate such drawbacks is a flash EEPROM. Conventional EEPROMs are very easy to use because data can be erased or written bit by bit.

  However, when a hard disk of a computer that requires a large storage capacity is composed of an EEPROM, this EEPROM does not need to have a function of erasing or writing data for each bit. This is because in a hard disk, data is erased or written in units of sectors (or blocks).

  Therefore, even if the rewriting function for each bit is eliminated, it is advantageous to achieve a large storage capacity by reducing the cell area and to reduce the cost of the product. Flash EEPROM was born.

  Details of the flash EEPROM are described in Non-Patent Document 2, for example.

  FIG. 69 shows the structure of the memory cell of the flash EEPROM.

  The memory cell of the flash EEPROM has a control gate and a floating gate similarly to the memory cell of the ultraviolet erasable EPROM. In the flash EEPROM, data is written by injecting hot electrons into the floating gate, like the ultraviolet erasable EPROM. Erasing is performed by extracting electrons from the floating gate using the FN tunnel phenomenon, as in the byte type EEPROM.

  In the flash EEPROM, the erase operation when the memory cells are individually viewed is the same as that of the byte type EEPROM, but the operation when the entire memory cell array is viewed is completely different from that of the byte type EEPROM. That is, the byte EEPROM erases data in byte units, whereas the flash EEPROM erases all bits at once. By adopting such an operation method, the flash EEPROM realizes a memory cell portion composed of one transistor per bit and achieves a large storage capacity.

  Note that data writing in the flash EEPROM can be performed bit by bit as in the ultraviolet erasable EPROM. In other words, the flash EEPROM and the ultraviolet erasable EPROM are the same in that erasing can be performed all at once and writing can be performed bit by bit.

  In order to realize a memory chip having a large storage capacity, a NAND flash EEPROM has been proposed based on the above-described flash EEPROM.

  Non-Patent Document 3 discloses a NAND flash EEPROM.

  As shown in FIGS. 70 and 71, in the memory cell array part of the NAND type EEPROM, a plurality of (for example, 16) memory cells are connected in series to form a NAND string, and one select transistor is connected to each of both ends. It is composed of NAND units.

  In the NAND type EEPROM, a bit line contact portion and a source line may be provided for one NAND unit instead of one memory cell, and a plurality of memory cells constituting the NAND string are adjacent to each other. Since the cells share one diffusion layer, the memory cell size per bit can be greatly reduced, and a memory chip having a large storage capacity can be realized.

  FIG. 72 shows a NOR type flash EEPROM. In the NOR type flash EEPROM, a 1-bit (one) memory cell is arranged between a bit line and a source line.

  From the viewpoint of cost, the above-described NAND flash EEPROM has a feature suitable for a file memory having a large storage capacity in which the cell size can be reduced compared to the NOR flash EEPROM, and the cost in bit units is low. Also, from the functional aspect, the NAND flash EEPROM has characteristics that the data rewriting speed is faster and the power consumption is lower than that of the NOR flash EEPROM.

  The functional feature of the NAND flash EEPROM is based on the data rewrite method. That is, in the case of a NAND flash EEPROM, writing and erasing are achieved by exchanging charges between the silicon substrate (channel) and the floating gate.

  In addition, the FN tunnel phenomenon is used for charge exchange. In other words, the current required for writing is an FN tunnel current that flows from the silicon substrate (channel) to the floating gate, and the consumption current of the NAND flash EEPROM is much higher than that of a NOR flash EEPROM that uses hot electrons for writing. Becomes smaller.

  In the case of a 64-megabit NAND flash EEPROM, writing in units of one page (512 bytes) can be performed in 200 [μs]. This writing time is shorter than the writing time for each block in the NOR type flash EEPROM.

Table 2 compares the characteristics of the NAND flash EEPROM and the characteristics of the NOR flash EEPROM.

  As shown in Table 2, the advantages and disadvantages of both memories are complementary to each other. For example, NAND flash EEPROM can be used for data reading on condition that rewriting is performed in a specific block data unit. In a digital camera having 300,000 pixels, a storage capacity of about 0.5 megabit is required for one shot of a photograph, and therefore, a NAND flash EEPROM is widely used.

  On the other hand, the NOR flash EEPROM is widely used as a memory for a control program of a mobile phone or the like because it can perform random access at a high speed of 100 [ns].

  Thus, in the field of nonvolatile semiconductor memory, it has evolved into EEPROM (conventional type), flash EEPROM, NAND type flash EEPROM, and in exchange for a rewrite function in units of bytes, the memory cell size is reduced, that is, 1 bit. The cost per unit (bit cost) has been reduced.

  However, the demand for data rewriting in byte units is increasing in the recent logic-embedded nonvolatile memory. For example, in the case of an IC card, when a part of data is rewritten by managing money such as income and expenditure, if the flash EEPROM is used, the amount of data to be rewritten becomes too large.

  Therefore, in order to eliminate such drawbacks, a byte EEPROM that can be rewritten in units of bytes is required. However, the byte type EEPROM has a large number of elements per bit as described above, which is disadvantageous for increasing the storage capacity and reducing the bit cost.

Currently, the mainstream of non-volatile semiconductor memories is flash EEPROM (NOR type, NAND type, etc.), so if a byte type EEPROM having the same process and rewriting method as a flash EEPROM is developed, an EEPROM that meets market demands Can be produced at low cost.
W. Johnson et al. "A 16 Kb Electrically Erasable Nonvolatile Memory," ISSCC Digest of Technical Papers, pp. 199 152-153, Feb. 1982. F. Masuoka et al. "A new Flash EEPROM cell using triple polytechnology technology," IEDM Technical Digest, pp. 464-467 Dec. 1984. F. Masuoka et al. "New ultra high density EPROM and Flash EEPROM with NAND structured cell," IEDM Technical Digest, pp. 552-555 Dec. 1987. A. Lancaster et al. "A 5V-Only EEPROM with Internal Program / Erase Control", IEEE International Solid-State Circuits Conference, pp. 164-165, Feb. 1983. W. D. Brown et al. , “Nonvolatile Semiconductor Memory technology”, IEEE Press Series on Microelectronic Systems StuTevsbury, Series Editor, p. 70, p. 212, p. 316, p. 326, p. 327, p. 344

  The present invention provides a novel non-volatile semiconductor memory that can be formed by the same process as that of a flash EEPROM, that can employ the same rewriting method as that of a flash EEPROM, and that can rewrite data in byte units.

  A nonvolatile semiconductor memory according to the present invention includes a memory cell array having a first memory cell unit including one memory cell and one select transistor, a bit line directly connected to the memory cell, and a bit line. The memory cell has a stack gate structure having a floating gate and a control gate, and data write / erase to / from the memory cell is performed by using the FN tunnel phenomenon. .

  According to the example of the present invention, it is possible to provide a novel non-volatile semiconductor memory that can be formed by the same process as that of the flash EEPROM, can adopt the same rewriting method as the flash EEPROM, and can rewrite data in byte units.

  The best mode for carrying out an example of the present invention will be described below in detail with reference to the drawings.

  FIG. 1 shows a memory cell of a byte type EEPROM of the present invention. FIG. 2 shows an equivalent circuit of the memory cell of FIG. FIG. 3 shows the overall circuit configuration of the memory cell array.

  The memory cell MC has a control gate and a floating gate, and has the same structure as the memory cell of the flash EEPROM. Select transistors ST1 and ST2 are respectively connected to both ends of the memory cell MC. The select transistor ST1 is connected to the bit line via the bit line contact portion BC, and the select transistor ST2 is connected to the source line SL.

  The memory cell MC and the select transistors ST1, ST2 constitute one memory cell unit, and the memory cell array is realized by arranging a plurality of memory cell units in an array.

  A plurality of memory cell units arranged in the row direction constitute one block. One control gate line CGL extending in the row direction is disposed in one block. Memory cells connected to one control gate line CGL are collectively called one page.

  The erase operation can be performed for each page. Each write and read operation to the memory cell can be performed simultaneously for one page by providing a sense amplifier having a latch function for each column. However, data input / output is performed serially for each bit, for example.

  Also, with such a configuration, data rewriting in units of bytes can be performed.

  From the viewpoint of structure, the byte type EEPROM of the present invention can be considered as one memory cell in one NAND unit in the NAND type flash EEPROM. However, the byte type EEPROM of the present invention is greatly different from the NAND type flash EEPROM in terms of function. This will be described in detail in the description of the operation.

  The structural advantages of the byte type EEPROM of the present invention will be described.

  The memory cell portion of the byte EEPROM of the present invention is different from the memory cell portion of the NAND flash EEPROM only in the number of memory cells constituting one unit. Therefore, in the byte type EEPROM of the present invention, the process of the NAND type flash EEPROM can be adopted as it is, so that the storage capacity can be increased and the production cost can be reduced even though the byte unit can be erased.

For example, when the design rule is 0.4 [μm], the area (short side length a × long side length b) of one memory cell has a short side length a of 1.2 [μm] and a long side length Since b is 3.2 [μm], it is 3.84 [μm ² ]. On the other hand, in the conventional byte type EEPROM as shown in FIGS. 65 and 66, when the design rule is 0.4 [μm], the area of one memory cell is 36 [μm ² ].

  In other words, with respect to the memory cell array part, even if it is simply calculated, the byte type EEPROM of the present invention can realize a storage capacity about 10 times that of the conventional byte type EEPROM.

  Further, since the byte type EEPROM of the present invention can be manufactured by the same process as the NAND type flash EEPROM, it can be easily applied to a logic mixed nonvolatile memory.

  In addition, since the memory cell of the byte type EEPROM of the present invention has the same structure as the memory cell of the NAND type flash EEPROM, the rewriting method of the flash EEPROM, that is, the rewriting using the FN tunnel phenomenon, can be seen for one memory cell. The method can be adopted as it is.

  However, when viewed as a whole memory cell array, the byte EEPROM of the present invention is different from the NAND flash EEPROM in that data rewriting (byte erasing) can be performed in units of bytes.

  Hereinafter, an erase operation, a write operation, and a read operation of the byte EEPROM according to the present invention will be sequentially described.

  During the erase operation, the ground potential is applied to the control gate (word line) CGL of the selected block, and the control gate CGL of the unselected block is set in a floating state.

  Thereafter, for example, erase pulses of 21 [V] and 3 [ms] are applied to the bulk. Here, the bulk is a well formed in the silicon substrate, and the memory cell MC and the select transistors SL1 and SL2 are all formed in this well.

  When the erase pulse is applied to the bulk, in the memory cell MC of the selected block, an erase voltage (21 [V]) is applied between the bulk and the control gate, and electrons in the floating gate are FN (Fowler-Nordheim) tunnel phenomenon. To move to the channel (well). As a result, the threshold voltage of the memory cell is about −3 [V].

  In the byte type EEPROM of the present invention, over-erasing in which the absolute value of the threshold voltage of the memory cell becomes extremely large in the erasing operation may not be a problem. Therefore, the erase operation is performed with a single erase pulse under the condition that the threshold voltage is about −3 [V], and the erase time (when verifying whether the threshold voltage is less than the predetermined value is performed) (Including the time required for).

  The reason why the byte type EEPROM of the present invention does not cause over-erasure is that the select transistors ST1 and ST2 are connected to both ends of one memory cell MC. That is, at the time of data reading, it is necessary to always turn off the non-selected memory cell and turn on or off the selected memory cell according to the data, but over-erasing turns on the non-selected memory cell. . If the select transistors ST1 and ST2 are provided, even if the non-selected memory cell is turned on, the data of the non-selected memory cell is not guided to the bit line, so there is no inconvenience regarding the operation of the memory.

  During the erase operation, the control gate CGL of the non-selected block is set in a floating state. Therefore, in the memory cell MC of the non-selected block, even if the bulk (well) potential rises, the potential of the control gate CGL also rises due to capacitive coupling between the control gate CGL and the bulk, so that data is erased. Absent.

  The control gate CGL is composed of polysilicon, a laminate of polysilicon and metal silicide, or the like. The control gate CGL is connected to the source of the word line drive MOS transistor via a metal wiring. Therefore, the junction capacity of the source of the word line drive transistor, the overlap capacity of the source and gate, the capacity between the control gate and the metal wiring, the capacity between the control gate and the bulk (well), etc. are connected to the control gate. Is done.

  Among these capacitances, the capacitance between the control gate and the bulk (well) is particularly large. That is, since the coupling ratio between the control gate and the bulk is as large as about 0.9, the FN tunnel current flows in the memory cell MC of the non-selected block due to the capacitive coupling between the control gate CGL and the bulk. Can be prevented.

  In the erase verify, for example, it is verified whether or not the threshold voltage of all the memory cells in the selected block is −1 [V] or less. In the present invention, as described above, since over-erasing does not become a problem, verification of over-erasing is not necessary. In addition, erasing can be performed under the condition that the threshold voltage can be reliably lowered to about −3 [V], and verification can be omitted.

  At the time of “0” write operation, the select transistor ST1 on the bit line side of the selected block is turned on, the select transistor ST2 on the source line side is turned off, and the memory cell for write execution (“0” write) is The bit line BLi is set to 0 [V], and the bit line BLi is set to the power supply potential VCC (for example, 3.3 [V]) for the memory cell in which writing is prohibited ("1" writing).

  A potential of 0 [V] is applied from the bit line BLi to the channel of the memory cell to be written through the select transistor ST1. Therefore, the channel potential of the memory cell in which writing is performed becomes the ground potential.

  When a write potential is applied to the selected word line (control gate), a large potential difference is generated between the floating gate and the channel of the memory cell to be written among the selected memory cells connected to the selected word line. . Therefore, in the memory cell in which writing is performed, electrons move from the channel to the floating gate due to the FN tunnel phenomenon.

  On the other hand, in the write-protected memory cell, the channel is charged to the power supply potential VCC and is set in a floating state. When a write potential is applied to the selected word line (control gate), the channel potential is automatically boosted by the series capacitive coupling of the control gate, floating gate, channel, and bulk (well).

  Therefore, a large potential difference does not occur between the floating gate and the channel of the write-protected memory cell connected to the selected word line, and electrons do not move from the channel to the floating gate.

  As described above, for the write-protected memory cell, the write potential is applied to the selected word line by increasing the coupling ratio between the control gate and the channel and sufficiently charging the channel. When this is done, the channel potential (write inhibit potential) can be made sufficiently high.

  The coupling ratio B between the control gate and the channel is calculated by the following equation.

B = Cox / (Cox + Cj)
Here, Cox is the total gate capacitance between the control gate and the channel, and Cj is the total junction capacitance between the source and drain of the memory cell.

  The channel capacity of the memory cell is the sum of the sum Cox of the gate capacity and the sum Cj of the junction capacity.

  Note that the overlap capacitance between the gate and the source of the select transistor and the capacitance between the bit line and the source / drain are neglected here because they are much smaller than the channel capacitance.

  In the read operation, after the bit line is charged to the precharge potential, 0 [V] is applied to the control gate (selected word line) of the selected memory cell as shown in FIGS. The power supply potential VCC is applied to the gates of the select transistors on both sides of the memory cell, and 0 [V] is applied to the gates of the select transistors on both sides of the unselected memory cell. At this time, the select transistors on both sides of the selected memory cell are turned on, and the select transistors on both sides of the non-selected memory cell are turned off.

  Among the selected memory cells, a memory cell in which data “1” is written, that is, an erased memory cell is in a depletion mode having a negative threshold voltage, and thus is turned on, and the potential of the bit line Go down. On the other hand, the memory cell in which data “0” is written has the threshold voltage in the positive enhancement mode, and thus is turned off, and the bit line potential is maintained at the precharge potential.

  As described above, the determination of data “0” and “1” is made based on whether or not a cell current flows from the bit line to the source line. The change in the potential of the bit line is amplified (detected) by the sense amplifier.

  According to the byte type EEPROM of the present invention, since the memory cell MC is sandwiched between the select transistors, the memory cell MC has the following advantages.

  First, when the read potential is set to 0 [V], as shown in FIG. 6, the threshold voltage distribution after erasing or writing becomes negative (data “1”) or positive (data “0”). It only has to be. That is, if a verify function for distinguishing between “1” and “0” is provided, a verify function for detecting over-erase or over-write need not be provided. This eliminates the need for complicated verification as in a conventional flash EEPROM. In the present invention, a normal read operation is possible even when the absolute value of the negative threshold voltage increases due to over-erasing or when the absolute value of the positive threshold voltage increases due to over-writing. Therefore, the electric field applied to the gate oxide film (tunnel oxide film) can be set high, and the erase time and write time can be shortened.

  Secondly, like the NAND flash EEPROM, erasing and writing are both performed by exchanging charges between the floating gate and the channel using the FN tunnel phenomenon. Therefore, the current consumption at the time of data rewriting can be kept very small, and the number of memory cells that can be rewritten simultaneously by one rewriting operation can be increased.

  Third, unlike the NAND flash EEPROM, the byte EEPROM of the present invention has only one memory cell between the select transistors. That is, since the selected memory cell and the non-selected memory cell are not mixed between the select transistors, it is not necessary to always turn on the non-selected memory cell to function as a pass transistor at the time of reading. Therefore, no measures are required to prevent overwriting.

  Further, since it is not necessary to always keep the non-selected memory cell in the ON state at the time of reading, when reading is performed with the control gate of the selected memory cell being 0 [V], the control gate of the non-selected memory cell is also 0 [V ], And it is not necessary to consider the read retention.

  That is, in the conventional NAND flash EEPROM, a plurality of memory cells are connected in series between the select transistors. Therefore, at the time of reading, the control gate of the selected memory cell is set to 0 [V], and the control gate of the non-selected memory cell is set. Vread (= 4.5V). This has caused the lead retention to shrink.

  In the present invention, since only one memory cell is connected between the select transistors, at the time of reading, the control gates of all the memory cells are set to 0 [V] and only the select transistors at both ends of the memory cell are turned on / off. Thus, selection / non-selection of the memory cell can be determined.

  Further, since the select transistor is connected between the bit line and the memory cell, it is not necessary to always keep the non-selected memory cell in the OFF state at the time of reading. Therefore, no measures for preventing over-erasure are required.

  In addition, when “0” is written, it is not necessary to apply an intermediate potential (a potential that is approximately ½ of the write potential) to the unselected word line (control gate). This is because a select transistor exists between the memory cell and the bit line, and there is only one memory cell between the select transistors.

  Further, since the erroneous writing can be prevented without applying an intermediate potential to the non-selected word line, the writing reliability is improved. In addition, rewriting can be performed in page units (or bit units). Even during reading, the cell current can be increased because there is no pass transistor. Therefore, high-speed reading is possible, and data retention characteristics at the time of reading are improved.

Table 3 shows the potentials of the select gate lines SSL and GSL, the control gate line (word line) CGL, the bit line BLi, the cell source line SL, and the cell P well in each of the above erase, write, and read operations. .

  In the erase operation, the control gate line CGL of the selected block is set to 0 [V], and the control gate line CGL of the non-selected block and all the select gate lines SSL and GSL are set to a floating state.

  In this state, when an erase potential Vera, for example, 21 [V] is applied to the cell P well, the potentials of all the select gate lines SSL and GSL in the floating state and the control gate lines CGL of the non-selected blocks are: By capacitive coupling with the cell P well, Vera × β (where β is a coupling ratio).

  Here, when β is 0.8, the potentials of all the select gate lines SSL and GSL in the floating state and the control gate line CGL of the non-selected block rise to 16.8 [V].

During the erase operation, the pn junction composed of the N ⁺ diffusion layer connected to the bit line BLi and the cell source line SL and the cell P well is biased in the forward direction. Therefore, the bit line BLi and the cell source line SL are charged to Vera-Vb. Vb is a built-in potential of a pn junction.

  In the write operation, the bit line BLi connected to the selected memory cell into which “1” data is written, that is, the bit line BLi connected to the selected memory cell maintaining the erased state is supplied with the power supply potential (for example, 3.3 [ V]) The bit line BLi that is set to VCC and connected to the selected memory cell to which the “0” data is written is set to 0 [V].

  The select gate line SSL on the bit line side of the selected block is set to the power supply potential VCC, the select gate line GSL on the cell source line side is set to 0 [V], and the control gate line CGL is set to the write potential (for example, 18 [V]) Vprog.

  The select gate lines SSL and GSL, the control gate line CGL and the cell P well in the non-selected block are set to 0 [V].

  The cell source line is set to 0 [V]. However, if the channel potential of the memory cell in which “1” data is written in the selected block is boosted by capacitive coupling with the control gate line CGL, and the leak current of the cell source line becomes a problem due to punch through, the cell The potential of the source line is preferably set to the power supply potential VCC.

  In the read operation, the select gate lines SSL and GSL of the selected block are set to the power supply potential VCC, and the control gate line CGL is set to 0 [V]. In the case of precharging the bit line before reading data, the bit line BLi is set to a precharge potential (eg, 1.2 [V]) VBL.

  Among the selected memory cells, those storing “1” data are turned on and the cell current flows, so that the bit line BLi is discharged to 0 [V]. On the other hand, the selected memory cell in which “0” data is stored is turned off and no cell current flows, so that the bit line BLi holds the precharge potential VBL.

  In the read operation, when it is desired to perform the read operation by applying the power supply potential (for example, 3.3 V) VCC to the control gate line CGL of the selected block, the threshold distribution of the memory cells may be set as shown in FIG. .

Table 4 shows the select gate lines SSL and GSL, the control gate line (word line) CGL, the bit line BLi, the cell source line SL, and the cell P in the erase, write, and read operations when the threshold distribution of FIG. The potential of the well is shown.

  In the present invention, as described above, since select transistors are provided at both ends of the memory cell, the base of the threshold distribution of the memory cell after erasure (“1” data) may extend from positive to negative.

  FIG. 8 shows the main part of the circuit block of the byte type EEPROM of the present invention.

  As described above, the EEPROM includes a memory cell array 11 in which three memory cell units each having one memory cell sandwiched between two select transistors are arranged in a matrix, and a plurality of EEPROMs are arranged in the row direction on the memory cell array 11. The control gate line 10a and the memory cell array 11 have a plurality of bit lines 10b arranged in the column direction.

  The row decoder 12 selects a row, that is, the control gate line 10a. Data of the memory cell connected to the selected control gate line 10a is input to a sense amplifier circuit 13 including a sense amplifier having a data latch function provided for each column. The column decoder 14 selects a column, that is, the bit line BLi.

  The data of the sense amplifier in the selected column is output to the outside of the memory chip via the data input / output buffer 18. Data input into the memory chip is latched by a sense amplifier having a latch function for the selected column via the data input / output buffer 18.

  The booster circuit 16 generates a high voltage necessary for a write operation and an erase operation. The control circuit 17 controls the operation of each circuit inside the memory chip and serves as an interface between the inside and the outside of the memory chip. The control circuit 17 includes sequence control means (for example, a programmable logic array) that controls erase, write, and read operations on the memory cells.

  FIG. 9 shows a sense amplifier having a latch function connected to one bit line BLi in the sense amplifier circuit 13 of FIG.

  The sense amplifier mainly includes a latch circuit 21 including two CMOS inverters I1 and I2 in which one output is the other input. The latch node Q of the latch circuit 21 is connected to the I / O line via the column selecting NMOS transistor M8. The latch node Q is connected to the bit line BLi via the sense amplifier cutoff NMOS transistor M4 and the bit line potential clamping NMOS transistor M1.

  A connection node of the NMOS transistors M1 and M4 becomes a sense node Nsense. A precharge PMOS transistor M2 and a discharge NMOS transistor M3 are connected to the sense node Nsense. The precharge PMOS transistor M2 charges the sense node Nsense for a predetermined period based on the precharge control signal Load. The discharge NMOS transistor M3 discharges the charge of the sense node Nsense based on the discharge control signal DCB.

  A reset NMOS transistor M5 for forcibly grounding the latch node Qb based on the control signal φL1 is connected to the latch node Qb of the latch circuit 21. A reset NMOS transistor M6 for forcibly grounding the latch node Q based on the control signal φL2 is connected to the latch node Q of the latch circuit 21.

  The common source of the reset NMOS transistors M5 and M6 is connected to the ground via the sense NMOS transistor M7 controlled by the potential of the sense node Nsense. The sense NMOS transistor M7 is used for resetting the latch circuit 21 together with the NMOS transistors M5 and M6.

  FIG. 10 shows a schematic flowchart of the rewrite operation in byte units of the byte type EEPROM of the present invention.

  The sequence operation shown in this flowchart is controlled by the control circuit 17 of FIG. Hereinafter, the data rewriting operation in units of bytes will be described according to this flowchart.

  In the byte-unit data rewrite mode, first, data for one page of a memory cell connected to the selected control gate line (word line) is read to the sense amplifier circuit (page reverse reading). Then, the data for one page is latched in the sense amplifier circuit (step ST1).

  Next, byte data corresponding to the column specified by the address is loaded. The loaded byte data is overwritten on the byte data to be rewritten out of the data for one page latched in the sense amplifier circuit (step ST2).

  Next, the data for one page of the memory cell connected to the selected control gate line is simultaneously erased (page erase) (step ST3). After erasure, erase verify is performed on each memory cell connected to the selected control gate line to verify whether erasure is completely performed or not erased (steps ST4 and ST5). ).

  Then, page erase and erase verify are repeatedly performed until the threshold values of all memory cells for one page are within the predetermined range, and the threshold values of all memory cells for one page are within the predetermined range (erase completion). If so, the operation proceeds to the next operation (steps ST3 to ST5).

  When only one sense amplifier circuit having a latch function exists for one bit line (when there is only one page), the data of the sense amplifier circuit may be destroyed depending on the result of erase verification. There is sex. Therefore, in such a case, the erasure is completed once without performing the erase verify.

  Thereafter, the data for one page latched in the sense amplifier circuit is simultaneously written into the memory cells connected to the selected control gate line (step ST6). After the writing, a write verify is performed to verify whether the writing is completely performed or the writing is not performed on each memory cell connected to the selected control gate line (steps ST7 and ST8). ).

  Then, page write and write verify are repeatedly performed until the threshold values of all memory cells for one page are within the predetermined range, and the threshold values of all memory cells for one page are within the predetermined range (write complete). If so, the data rewrite operation in byte units is terminated.

  Note that when a high write potential is used and one write is performed with one write pulse, the write verify can be omitted.

  11 to 14 show the state of the data of the selected memory cell and the node Qb (FIG. 9) of the sense amplifier circuit in the main steps of FIG.

  FIG. 11 shows a state where data for one page of the memory cell connected to the selected control gate line (word line) is read by the sense amplifier circuit (corresponding to step ST1).

  When the data in the memory cell is “0” (threshold voltage is positive), the charge of the bit line BLi is not discharged and the precharge potential is maintained. Therefore, the sense node Nsense in FIG. 9 becomes the power supply potential VCC. When the control signal φL2 is the power supply potential VCC, the node Q becomes the ground potential VSS, that is, “0”.

  Conversely, when the data in the memory cell is “1” (threshold voltage is negative), the charge on the bit line BLi is discharged. Therefore, the sense node Nsense in FIG. 9 becomes the ground potential VSS. When the control signal φL2 is set to the power supply potential VCC, the node Q becomes the power supply potential VCC, that is, “1”.

  FIG. 12 shows a state in which data is overwritten on byte data (8-bit data) designated by an address among data for one page latched by the sense amplifier circuit (in step ST2). Correspondence).

  FIG. 13 shows a state in which data of a memory cell connected to the selected control gate line (word line) is erased (page erase) (corresponding to step ST3). By page erasing, all data in the memory cells connected to the selected control gate line becomes “1”.

  FIG. 14 shows a state where one page of data latched in the sense amplifier circuit is written (page write) to the memory cell connected to the selected control gate line (word line) (step writing). Corresponding to ST6).

  As described above, in the operation of the memory cell array 11, data is rewritten in units of pages, but actually, data in units of bytes is rewritten.

  Next, a read operation for page write and write verify will be described in detail with a focus on the operation of the sense amplifier circuit of FIG. 9 with reference to the timing charts of FIGS.

  15 and 16 show respective portions obtained by dividing one timing chart into two, and t5 in FIG. 15 and t5 in FIG. 16 represent the same time. That is, the waveform in the latter half of FIG. 15 partially overlaps with the waveform in the first half of FIG.

  When a command instructing writing is input from the outside of the chip to the inside of the chip, the writing operation is started.

  First, in order to reset the sense node Nsense, the control signal DCB is set to the power supply potential VCC. At this time, the MOS transistor M3 is turned on, and the sense node Nsense is grounded (t1).

  When the control signal BLSHF is also set to the power supply potential VCC together with the control signal DCB, the MOS transistor M1 is turned on and the bit line BLi is grounded.

  Before loading the write data into the sense amplifier circuit, the data latch control signal φL1 is set to the power supply potential VCC, and the precharge control signal Load is set to the ground potential VSS. At this time, the MOS transistors M5 and M7 are turned on, the latch node Qb of the latch circuit 21 is forcibly grounded, and the data is reset. That is, in all the sense amplifiers of the sense amplifier circuit 20, the latch node Q of the latch circuit 21 becomes the power supply potential VCC and the latch node Qb becomes the ground potential VSS (t2).

  Next, write data is loaded from the I / O line, the data is latched in each latch circuit 21 of the sense amplifier circuit 20, and the nodes Q and Qb are set to "H" and "L" according to the load data. (T3).

  Specifically, in the latch circuit 21 of the sense amplifier corresponding to the memory cell to which “0” is written, “L” (= VSS) is given to the latch node Q, and “1” is written (write prohibited) memory cell. In the latch circuit 21 of the sense amplifier corresponding to “H”, “H” (= VCC) is applied to the latch node Q.

  Next, the control signals BLSHF and SBL become “H”, and charging of each bit line is started based on the data latched in each latch circuit 21 of the sense amplifier circuit 20 (t4).

  That is, the bit line BLi connected to the memory cell for performing “0” write is set to the ground potential VSS, and the bit line connected to the memory cell for “1” write (write inhibit) is charged to the power supply potential VCC. . The selected control gate line (word line) is set to the write voltage Vprog (about 20 [V]).

  By this operation, writing to one page of memory cells is performed.

  After the data writing is completed, a write verify for verifying whether the data writing is completed properly is started.

  First, reading for write verification is performed. This verify read operation is the same as a normal read operation.

  When the control signal DCB is set to the power supply potential VCC, the MOS transistor M3 is turned on and the sense node Nsense is forcibly grounded (t5).

  Subsequently, when the reference potential Vref (about 0.5 [V]) is applied to the selected control gate line CGL and the power supply potential VCC is applied to the select gate lines SSL and GSL, verify read is performed. (T6).

  At the time of reading, a bit line precharge type sensing method, a current detection type sensing method, or the like can be used. In the bit line precharge type sensing method, after the bit line BLi is precharged and brought into a floating state, the potential of the bit line is maintained or lowered according to the data of the memory cell. The current detection type sensing method will be described in detail below.

  At time t6, the control signal BLSHF is clamped from the boosted potential VCC + α to the potential VCC-α, and reading is performed by the balance between the memory cell current flowing through the MOS transistor M1 and the current of the MOS transistor M2 that charges the sense node Nsense. When the potential of the bit line BLi rises to, for example, 0.9 [V], the MOS transistor M1 is cut off, and the sense node Nsense becomes the power supply potential VCC.

  After the sense node Nsense becomes “H” (= VCC), the latch control signal φL1 is set to the power supply potential VCC, and the MOS transistor M5 is turned on (t7). When the sense node Nsense is at the power supply potential VCC (in the case of a sense amplifier connected to a memory cell whose threshold is higher than the verify potential Vref), the MOS transistor M7 is turned on, the latch node Qb is at the ground potential VSS, and the latch node Q is at It becomes the power supply potential VCC.

  When the ground potential VSS is loaded to the latch node Q and data is normally written, the latch data of the latch circuit 21 is inverted. When the writing to the memory cell is insufficient, the sense node Nsense remains “L” (= VSS) in the verify reading, so that the data inversion of the latch circuit 21 does not occur and the latch node Q maintains VSS. In the sense amplifier connected to the write-inhibited memory cell, the latch node Q is at the power supply potential VCC, so there is no data inversion.

  When there is an insufficiently written memory cell, that is, when there is a sense amplifier in which data inversion of the latch circuit 21 does not occur, writing and verify reading are repeatedly performed. When the potentials of the latch nodes Q of all the sense amplifiers for one page become the power supply potential VCC, the writing is completed.

  Next, the data rewriting operation in units of bytes will be described in detail with a focus on the operation of the sense amplifier circuit of FIG. 9 with reference to the timing chart of FIG.

  When a command for instructing byte rewriting is input from the outside of the chip to the inside of the chip, the byte rewriting operation starts.

  First, a reverse read operation of already written data is started for one page of memory cells connected to the selected control gate line (word line).

  The reverse read operation is the same as the read operation.

  First, the data latch control signal φL1 is set to the power supply potential VCC, and the precharge control signal Load is set to the ground potential VSS. At this time, the MOS transistors M5 and M7 are turned on, the latch node Qb of the latch circuit 21 is forcibly grounded, and the data is reset. That is, the latch nodes Q of all the latch circuits 21 of the sense amplifier circuit are set to the power supply potential VCC, and the latch nodes Qb are set to the ground potential VSS (t1).

  Next, the control signal DCB is set to the power supply potential VCC. At this time, the MOS transistor M3 is turned on, and the sense node Nsense is forcibly grounded (t2). Subsequently, when VSS (= 0 V) is applied to the selected control gate line CGL and the power supply potential VCC is applied to the select gate lines SSL and GSL, reading is performed (t3).

  After the sense node Nsense becomes “H” (= VCC), the latch control signal φL2 becomes the power supply potential VCC, and the MOS transistor M6 is turned on (t4). When the sense node Nsense is at the power supply potential VCC (that is, when the sense amplifier is connected to a memory cell in which data “0” is written and the threshold voltage is higher than VSS), the MOS transistor M7 is turned on and the latch node Q Is at ground potential VSS, and latch node Qb is at power supply potential VCC.

  Next, the control signal DCB is set to the power supply potential VCC, the control signal BLSHF is set to the power supply potential VCC or the potential VCC + α, and the bit line BLi and the sense node Nsense are reset (t5).

  Thereafter, byte data is loaded into the latch circuit 21 of the sense amplifier circuit 20 designated by the column address, and the nodes Q and Qb are set to “H” and “L” according to the byte data (t6).

  Byte data input from the outside of the chip is overwritten on predetermined data of the page data written in the latch circuit 21.

  Thereafter, a page erase operation is performed on the memory cell connected to the selected control gate line.

  The control gate line of the selected block is set to the ground potential VSS, and the control gate line and all the select gate lines of the non-selected block are set to a floating state. When the erase voltage Vera is applied to the cell P well, the select gate line in the floating state and the control gate line of the non-selected block are set to Vera × β (β is a coupling ratio) due to capacitive coupling with the cell P well. Boosted.

The bit line BLi and the cell source line SL are connected to the N ⁺ layer in the cell P well. When the pn junction between the N ⁺ layer and the cell P well is forward-biased, the bit line BLi and the cell source line SL are charged to Vera-Vb, respectively (t7). However, Vb is a built-in potential of a pn junction.

  Thereafter, erase verify is performed to confirm that all the memory cells of the selected page are in an erased state, that is, the threshold voltage of the memory cell has become negative. Based on the data stored in the latch circuit 21, a write operation and a write verify operation are performed on the memory cells of the selected page.

  In FIG. 17, operations after the erase verify are omitted.

  FIG. 18 shows an example in which a part of the NAND flash EEPROM memory cell array is used as the byte EEPROM memory cell array of the present invention.

  The byte EEPROM memory cell array according to the present invention can be considered as one memory cell between two select transistors in the NAND flash EEPROM memory cell array. Therefore, the EEPROM as in this example can be easily realized.

  In the EEPROM of this example, two types of memory cell units having different configurations are connected to one bit line BLi. That is, in the first memory cell unit, a plurality of memory cells (for example, 4, 8, 16, 32, etc.) are connected between two select transistors, and the second memory cell unit has two select transistors. One memory cell is connected between the transistors.

  In selecting the control gate line (word line), a drive circuit may be provided separately in the first memory cell unit region and the second memory cell unit region, and if it can be shared, The drive circuits in both regions may be combined into one.

  In place of the NAND flash EEPROM memory cell array of FIG. 18, the following memory cell array may be employed.

  The memory cell array shown in FIG. 19 is an AND type flash EEPROM memory cell array. The memory cell array shown in FIG. 22 is a memory cell array of DINOR type flash EEPROM.

  The AND unit of the AND type flash EEPROM of FIG. 19 has a plurality of memory cells connected in parallel between the sub bit line and the sub source line. The sub bit line is connected to the main bit line via the drain side select transistor. The sub source line is connected to the main source line via the source side select transistor.

  For example, in the case of a 64-megabit AND type flash EEPROM, one AND unit includes 128 memory cells (m = 128) and two select transistors.

  This memory cell array is characterized in that bit lines (data lines) and source lines are hierarchized. Each of the bit line and the source line includes a main wiring and a sub wiring, and the sub wiring has a pseudo contactless structure formed of a diffusion layer.

  Data write / erase to / from the memory cell is performed by an FN (Fowler-Nordheim) tunnel current.

  As shown in FIG. 20, data writing is performed by extracting electrons from the floating gate to the drain using an FN tunnel current. As shown in FIG. 21, data is erased by injecting electrons from the substrate (entire channel surface) to the floating gate using an FN tunnel current.

  The DINOR (Divided Bit Line NOR) type flash EEPROM of FIG. 22 is capable of a single power supply operation like the NAND type flash EEPROM, has a high rewrite speed, a small memory cell size, and a NOR type flash. Like EEPROM, it has the feature that high-speed random access is possible.

  The memory cell unit of the DINOR type flash EEPROM has a main bit line and a sub bit line in the memory cell array in a hierarchical structure, and is therefore almost the same as the AND type AND unit in size. The structure of the memory cell is a stacked gate type, similar to the structure of the NOR type flash EEPROM or NAND type flash EEPROM, and the drain of the memory cell is connected to a sub-bit line formed of polysilicon.

  For example, in the case of a 16 megabit DINOR type flash EEPROM, 64 memory cells are connected to the sub bit line. If the contact to the memory cell is achieved by a so-called buried contact of polysilicon and a diffusion layer, the memory cell size can be reduced.

  The data write / erase mechanism for the memory cell is the same as that of the AND type flash EEPROM, and is performed by an FN (Fowler-Nordheim) tunnel current.

  That is, data writing to the memory cell is performed by drawing out the electrons of the floating gate to the drain using the FN tunnel current. Data is erased by injecting electrons from the substrate (entire channel surface) to the floating gate using an FN tunnel current.

  The sense amplifier circuit of FIG. 9 is connected to the main bit line of FIG. 19 and FIG. 22, and data rewriting in units of bytes is executed based on the flowchart of FIG.

  As described above, even in the EEPROM having the memory cell array as shown in FIGS. 18, 19 and 22, the rewriting method as shown in the flowchart of FIG. Data rewriting in byte units is possible.

  Further, in the byte type EEPROM memory cell of the present invention, the select transistor on the bit line side can be omitted, and a memory cell unit can be constituted by one memory cell transistor and one source side select transistor. In this case, at the time of data writing, based on the data of the sense amplifier circuit, a write-prohibiting intermediate voltage Vm, which is about 1/2 of the write voltage Vprog, is applied to the write-inhibited bit line.

  Conventionally, a memory cell called a SONOS (silicon-oxide-nitride-oxide-silicon) cell is known. This memory cell is characterized in that data (“0” or “1”) is specified by the amount of electrons trapped in the silicon nitride film immediately below the gate electrode (word line).

  The SONOS cell is disclosed in Non-Patent Document 4, for example.

  The memory cell unit of Non-Patent Document 4 includes one memory cell and two select transistors sandwiching the memory cell. Further, this Non-Patent Document 4 points out that data rewriting in byte units is possible in the SONOS cell (see “LOAD-LATCHES-ROW-ERASE operation” p.164, left column, lines 31 to 40).

  However, Non-Patent Document 4 does not specifically disclose the data rewrite operation in byte units. That is, it is unclear how to actually rewrite data in byte units. The memory cell disclosed in Non-Patent Document 4 has a structure in which the gate of the memory cell and the gate of the select transistor overlap, and does not have a stack gate structure like a flash EEPROM.

  Further, according to the present invention, a remarkable effect different from the effect of the conventional NAND flash EEPROM and the effect of the memory cell of Non-Patent Document 4 can be obtained.

  That is, the threshold distribution of the memory cell after writing or erasing is as shown in FIG. 6 or FIG. 7, for example, as described above. Here, in the conventional NAND flash EEPROM, the upper and lower limits of the threshold distribution of data “1” and “0” are determined, and the threshold distribution of each data must be within a predetermined range by verification. In addition, since the memory cell of Non-Patent Document 4 has a SONOS structure and the amount of trapped electrons in the silicon nitride film is determined to some extent, it is difficult to freely shift the threshold distribution of the memory cell (about this) For example, refer nonpatent literature 5.).

  On the other hand, according to the present invention, for example, the threshold distribution of data “1” and the threshold distribution of data “0” in FIG. 6 or FIG. Can be separated. That is, by increasing the margin of the threshold distribution of data “1” and the threshold distribution of data “0”, writing and erasing can be sufficiently performed and erroneous reading can be prevented. In addition, since there is no upper limit and lower limit of the threshold distribution of each data, verification is not necessary, and so-called one-time writing and erasing are possible.

  FIG. 23 shows an example of a circuit block of the byte type EEPROM of the present invention. FIG. 24 shows a part of the memory cell array 11 of FIG.

  The circuit block of this example is applied to an EEPROM having the memory cell array of FIG. 3, and approximates a circuit block of a NAND type EEPROM.

  In the present invention, since the memory cell unit is composed of three elements of one memory cell and two select transistors sandwiching the memory cell, one block BLKi (i = 0, 1,... N) Memory cells connected to the control gate line CGL, that is, memory cells for one page are arranged.

  The control gate / select gate driver 12c is provided corresponding to one block BLKi (i = 0, 1,... N), that is, one control gate line CGL (one page). Each driver 12c includes a booster circuit. The predecoder 12a and the row decoder 12b are also provided corresponding to one block BLKi, that is, one control gate line CGL (one page).

  The row address signal is input to the predecoder 12 a via the address register 19. Then, one row (or one block) is selected by the predecoder 12a and the row decoder 12b. When the selected block is BLKi, for example, the driver 12c applies a predetermined potential corresponding to the operation mode to the control gate line CGL and the select gate lines SSL and GSL in the selected block BLKi (see Tables 3 and 4).

  The sense amplifier circuit 13 having a latch function latches read data and write data. Read data (output data) is output to the outside of the memory chip via the column selection circuit 15 and the input / output buffer 18. Write data (input data) is latched by the sense amplifier circuit 13 having a latch function via the input / output buffer 18 and the column selection circuit 15.

  The command signal is input to the command decoder 26 via the data input / output buffer 18 and the command register 25. The control circuit 17 receives signals such as an output signal of the command decoder 26, a command latch enable signal CLE, a chip enable signal / CE, and a write enable signal / WE.

  The signal generation circuit (boost circuit) 27 generates a potential to be applied to the control gate line CGL and the select gate lines SSL and GSL under the control of the control circuit 17, and supplies this potential to the control gate / select gate driver 12c.

  FIG. 25 shows another example of the circuit block of the byte type EEPROM of the present invention. FIG. 26 shows a part of the memory cell array 11 of FIG.

  The circuit block of this example is applied to an EEPROM having the memory cell array of FIG.

  The memory cell array is composed of a 3-tracell unit 11-0 in which memory cell units according to the present invention are arranged and a NAND cell unit 11-1 in which NAND cell units are arranged.

  The 3-tracell unit 11-0 has a memory cell unit composed of three elements of one memory cell and two select transistors sandwiching the memory cell, and is divided into n blocks BLK0, BLK1,... BLKn. The NAND cell unit 11-1 has a NAND cell unit including a plurality of (4, 8, 16, etc.) memory cells connected in series and two select transistors sandwiching the memory cell, and includes m blocks BLK0, BLK1. , ... are divided into BLKm.

  In each block BLKi (i = 0, 1,... N) of the three tracell unit 11-0, memory cells connected to one control gate line CGL, that is, memory cells for one page are arranged. On the other hand, memory cells connected to a plurality of control gate lines CGL, that is, memory cells for a plurality of pages are arranged in each block BLKi (i = 0, 1,... M) of the NAND cell unit 11-1. Is done.

  In the 3-tracell unit 11-0, the control gate / select gate driver 12c is provided corresponding to one block BLKi, that is, one control gate line CGL (one page). Each driver 12c includes a booster circuit. The predecoder 12a and the row decoder 12b are also provided corresponding to one block BLKi, that is, one control gate line CGL (one page).

  In the NAND cell unit 11-1, the control gate / select gate driver 12c is provided corresponding to one block BLKi including a plurality of control gate lines CGL0,... CGL7 (a plurality of pages). Each driver 12c includes a booster circuit. The predecoder 12a and the row decoder 12b are also provided corresponding to one block BLKi including a plurality of control gate lines CGL0,... CGL7 (a plurality of pages).

  The row address signal is input to the predecoder 12 a via the address register 19. Then, one row (or one block) of the 3-tracell unit 11-0 or the NAND cell unit 11-1 is selected by the predecoder 12a and the row decoder 12b.

  The sense amplifier circuit 13 having a latch function latches read data and write data. Read data (output data) is output to the outside of the memory chip via the column selection circuit 15 and the input / output buffer 18. Write data (input data) is latched by the sense amplifier circuit 13 having a latch function via the input / output buffer 18 and the column selection circuit 15.

  The command signal is input to the command decoder 26 via the data input / output buffer 18 and the command register 25. The control circuit 17 receives signals such as an output signal of the command decoder 26, a command latch enable signal CLE, a chip enable signal / CE, and a write enable signal / WE.

  The signal generation circuit (boost circuit) 27 generates a potential to be applied to the control gate line CGL and the select gate lines SSL and GSL under the control of the control circuit 17, and supplies this potential to the control gate / select gate driver 12c.

  FIG. 27 shows a data rewrite operation in units of bytes applied to the EEPROM of FIGS.

  This rewriting operation is a summary of the rewriting operations shown in FIGS.

The byte-unit data rewrite operation of the present invention includes the following four main steps.
i. Data is read backward from one page of memory cells in the selected block, and this is held in a sense amplifier circuit having a latch function.
ii. The byte data is overwritten on the data held in the sense amplifier circuit having the latch function.
iii. Erase the data in one page of memory cells in the selected block.
iv. Write the data held in the sense amplifier circuit having the latch function to the memory cells for one page in the selected block.

  Through the above steps, it is possible to provide a non-volatile semiconductor memory that can be manufactured in the same process as the flash EEPROM and can rewrite data in byte units even though the same rewriting method is applied. (Normally, the data in the memory cell is erased in batch units without the above step i.), The data cannot be rewritten in byte units, but can be written in byte units after erasure.) .

  Here, the number of data rewrites in the EEPROM memory cell to which the byte-by-byte data rewrite method shown in FIG. 27 is applied will be considered.

  When data rewriting for 1 byte is performed by the rewriting method of FIG. 27, one reverse read operation, erase operation, and write operation are performed on the data for one page in the selected block. That is, in the selected block, one reverse read operation, erase operation, and write operation are performed even for memory cells that are not subjected to data rewrite.

  Therefore, for example, when all the data in one page is rewritten, the page read, erase, and write counts when the data for one page is rewritten for each byte by the rewrite method of FIG. The number of page reading, erasing, and writing in the case of overwriting is substantially larger by the number of bytes included in one page.

  For example, when one page consists of 64 bytes, one page read, erase, and write operations are sufficient when rewriting data for one page at a time, but 64 when rewriting data for one page every byte. One page read, erase, and write operations are required.

  As described above, in the data rewrite method in units of bytes shown in FIG. 27, when data rewrite for 1 byte is performed, one reverse read operation, erase operation, and write for one page of data in the selected block are performed. Operation is performed. Therefore, the number of page reading, erasing, and writing when rewriting data for one page by the method of the present invention is at most 1 than the number of page reading, erasing, and writing when rewriting data for one page at a time. Increased by the number of bytes contained in the page.

  In order to prevent such an increase in the number of page read, erase, and write operations, after a single page read in the rewrite method of FIG. 27, multiple bytes of data are overwritten to reduce the number of page read, erase, and write operations. You can also.

  However, in the following, a non-volatile semiconductor memory that can reduce the number of page reading, erasing, and writing while maintaining rewriting in units of bytes by means other than overwriting of data of a plurality of bytes will be described.

  FIG. 28 shows an improved example of the byte type EEPROM of FIG.

  In the present invention, the memory cell array 11 is composed of a plurality of blocks BLKi-j (i = 0, 1,... N; j = 0, 1, 2, 3) arranged in a matrix in the row direction and the column direction. ing.

  In the example described so far, as shown in FIGS. 23 and 25, the block BLKi is arranged only in the column direction, and the memory cells for one page connected to one control gate line CGL are always It existed in the same block BLKi. In the present invention, one page of memory cells is divided into a plurality of multiples of a multiple of 1 byte (8 bits), and a plurality of blocks are also arranged in the row direction.

  Specifically, when one page is composed of memory cells of k (k is a positive number) bytes, if one block is composed of memory cells of r (r is a positive number, r ≦ k) bytes, The number of blocks in the row direction is k / r. In this example, the number of blocks in the row direction is four. In this case, for example, one block is composed of 16-byte memory cells, and one page is composed of 64-byte memory cells.

  The main control gate / select gate driver 12c is provided corresponding to four blocks BLKi-j in the row direction, that is, one control gate line CGL (one page). Each driver 12c includes a booster circuit. The predecoder 12a and the row decoder 12b are also provided corresponding to four blocks BLKi-j, that is, one control gate line CGL (one page).

  The sub control gate driver 28 is provided corresponding to each block BLKi-j.

  The row address signal is input to the predecoder 12 a and the subdecoder 29 via the address register 19. The four blocks BLKi-j in one row are selected by the predecoder 12a and the row decoder 12b. Further, the sub-decoder 29 selects one of the four selected blocks BLKi-j.

  Note that the sub-decoder 29 has a function of selecting a plurality of blocks in one selected row or all blocks (four blocks in this example) in one selected row. Also good.

  In the present invention, data can be read, erased and written in units of blocks. That is, it is not necessary to read out data for one page to the sense amplifier circuit having a latch function in the data rewrite operation in byte units. Therefore, in the present invention, the number of page read, erase, and write operations can be reduced in the byte-unit data rewrite operation, compared to the examples of FIGS. 23 and 25, and the substantial page rewrite characteristics can be improved.

For example, consider a case where the page rewrite characteristic (number of rewrites) of an EEPROM composed of memory cells in which one page is k (k is a positive number) bytes is 1 × 10 ⁶ times.

In the example of FIGS. 23 and 25, since page read, erase, and write operations are required k times to rewrite the data for one page, the page rewrite characteristic is substantially (1 / k) × 10. Decrease to ⁶ times.

In the present invention, one page is divided into k / r (r is a positive number, r ≦ k) blocks, each block is composed of r-byte memory cells, and data can be read, erased and written in units of blocks. Therefore, page read, erase, and write operations for rewriting data for one page are substantially (1 / r) × 10 ⁶ times.

Specifically, for example, when one page is composed of 64 bytes, the page rewriting characteristics of the example of FIGS. 23 and 25 are 1.7 × 10 ⁴ times. On the other hand, when one page is composed of 8 blocks and one block is composed of 8 bytes, the page rewriting characteristic of the present invention is 1.3 × 10 ⁵ times, which is more than the example of FIGS. The actual rewriting characteristic is improved by one digit.

In the case of the present invention, by configuring one block from one byte, the substantial rewrite characteristic can be made 1 × 10 ⁶ times at the maximum.

  When the selected block is BLKi-j, the main control gate / select gate driver i gives a predetermined potential corresponding to the operation mode to the control gate line CGL and the select gate lines SSL, GSL in the selected block BLKi-j (Table). 3 and 4).

  The sense amplifier circuit 13 having a latch function latches read data and write data. Read data (output data) is output to the outside of the memory chip via the column selection circuit 15 and the input / output buffer 18. Write data (input data) is latched by the sense amplifier circuit 13 having a latch function via the input / output buffer 18 and the column selection circuit 15.

  The command signal is input to the command decoder 26 via the data input / output buffer 18 and the command register 25. The control circuit 17 receives signals such as an output signal of the command decoder 26, a command latch enable signal CLE, a chip enable signal / CE, and a write enable signal / WE.

  The signal generation circuit (boost circuit) 27 generates a potential to be applied to the control gate line CGL and the select gate lines SSL and GSL under the control of the control circuit 17, and supplies this potential to the control gate / select gate driver 12c.

  FIG. 29 shows an example of the configuration of the predecoder PDi.

In this example, the number of rows, that is, the number of control gate lines CGL (number of blocks) is assumed to be 1024 (2 ¹⁰ ). In this case, one control gate line CGL can be selected by 10-bit row address signals a1, a2,.

  The row address signals a1, a2, a3 are input to the NAND circuit 30-1, the row address signals a4, a5, a6 are input to the NAND circuit 30-2, and the row address signals a7, a8, a9, a10 are Input to the NAND circuit 30-3. The output signal of the NAND circuit 30-1 becomes a signal D via the inverter 31-1, and the output signal of the NAND circuit 30-2 becomes a signal E via the inverter 31-2. The output signal becomes the signal F via the inverter 31-3.

  Different row address signals a1, a2,..., A10 are input to each predecoder PDi. Only the output signals D, E, and F of the predecoder PDi belonging to one selected row are all “1”.

  FIG. 30 shows an example of the configuration of the row decoder RDi and the main control gate / select gate driver i.

  The row decoder RDi includes a NAND circuit 32 and an inverter 33. Output signals D, E, and F of the predecoder PDi are input to the NAND circuit.

  The main control gate / select gate driver i includes a booster circuit 34 and N-channel MOS transistors 35-1, 35-2, 35-3 as drive circuits.

  In the main control gate / select gate driver i belonging to the selected row, the power supply potential VCC or the boosted potential is applied to the gates of the N-channel MOS transistors 35-1, 35-2, 35-3.

  For example, at the time of data writing, in the driver i belonging to the selected row, the output potential VB of the booster circuit 34 becomes the boosted potential Vprog, and the N-channel MOS transistors 35-1, 35-2, and 35-3 are turned on. On the other hand, in the signal generation circuit 27, SS (= VCC), CG (= Vprog), and GS (= 0V) are generated. These potentials SS, CG, GS are transmitted to the main control gate line CGLi and the select gate lines SSLi, GSLi in the selected row via the N-channel MOS transistors 35-1, 35-2, 35-3. .

  At the time of data erasure, in driver i belonging to the selected row, output potential VB of booster circuit 34 becomes power supply potential VCC, and N-channel MOS transistors 35-1, 35-2, 35-3 are turned on. On the other hand, in the signal generation circuit 27, SS (= VCC), CG (= 0V), and GS (= VCC) are generated. These potentials SS, CG, GS are transmitted to the main control gate line CGLi and the select gate lines SSLi, GSLi in the selected row via the N-channel MOS transistors 35-1, 35-2, 35-3. .

  Since the select gate lines SSLi and GSLi are subsequently floated, when the erase potential Vera is applied to the P well, the potentials of the select gate lines SSLi and GSLi are set to the P well and the select gate lines SSLi and GSLi. It rises to Vera + α by capacitive coupling.

  At the time of data reading, in the driver i belonging to the selected row, the output potential VB of the booster circuit 34 becomes the power supply potential VCC or VCC + α (α is a value equal to or higher than the threshold voltage of the N channel transistor), and the N channel MOS transistor 35-1. , 35-2, 35-3 are turned on. On the other hand, in the signal generation circuit 27, SS (= VCC), CG (= 0V or VCC), and GS (= VCC) are generated. These potentials SS, CG, GS are transmitted to the main control gate line CGLi and the select gate lines SSLi, GSLi in the selected row via the N-channel MOS transistors 35-1, 35-2, 35-3. .

  In the main control gate / select gate driver i belonging to the non-selected row, the ground potential is applied to the gates of the N-channel MOS transistors 35-1, 35-2, 35-3. -2 and 35-3 are turned off. Therefore, the main control gate line CGLi and the select gate lines SSLi and GSLi in the non-selected row are all in a floating state.

  Note that VSS (0 V) may be applied to the select gate lines SSLi and GSLi in the non-selected rows when reading data. In this case, for example, ground MOS transistors are connected to all the select gate lines SSLi and GSLi, and on / off of the ground MOS transistors is controlled depending on whether row (or block) is selected.

  FIG. 31 shows an example of the configuration of a plurality of blocks and sub-control gate drivers arranged in one row.

  In this example, a case will be described in which four blocks BLKi-0, BLKi-1, BLKi-2, and BLKi-3 are arranged in one row, corresponding to the circuit block of FIG.

  In each block BLKi-j (j = 0, 1, 2, 3), sub-control gate lines CGLi-0, CGLi-1, CGLi-2, and CGLi-3 are arranged, respectively. The sub control gate line CGLi-j (j = 0, 1, 2, 3) is connected to a memory cell that is a multiple of 1 byte (for example, 16 bytes) arranged in the block BLKi-j.

  The sub control gate line CGLi-j is connected to the main control gate line CGLi via an N channel MOS transistor 36-j as a drive circuit constituting the sub control gate driver 28.

  On / off of the N-channel MOS transistor 36-j is controlled by the sub-decoder 29. The subdecoder 29 has a function of selecting one N-channel MOS transistor 36-j (one block).

  Note that the sub-decoder 29 may have a function of selecting a plurality or all of the N-channel MOS transistors 36-j (a plurality or all of the blocks).

  At the time of data writing, in the selected block BLKi-j in the selected row, since Vprog is applied to the gate of the N channel MOS transistor 36-j, the N channel MOS transistor 36-j is turned on. Therefore, the high potential Vprog for writing is transmitted from the main control gate line CGLi to the sub control gate line CGLi-j in the selected block BLKi-j.

  At the time of erasing data, VCC is applied to the gate of the N channel MOS transistor 36-j in the selected block BLKi-j in the selected row, so that the N channel MOS transistor 36-j is turned on. Therefore, the ground potential is transmitted from the main control gate line CGLi to the sub control gate line CGLi-j in the selected block BLKi-j.

  At the time of data reading, VCC is applied to the gate of the N channel MOS transistor 36-j in the selected block BLKi-j in the selected row, so that the N channel MOS transistor 36-j is turned on. Therefore, the ground potential or the power supply potential VCC is transmitted from the main control gate line CGLi to the sub control gate line CGLi-j in the selected block BLKi-j (see Tables 3 and 4).

  On the other hand, in the non-selected block BLKi-j in the selected row, since the ground potential is applied to the gate of the N-channel MOS transistor 36-j, the N-channel MOS transistor 36-j is turned off. That is, the sub control gate line CGLi-j in the non-selected block BLKi-j is in a floating state.

  Here, in the selected row, a plurality of sub control gate lines CGLi-j are arranged immediately below the main control gate line CGLi. Accordingly, when a predetermined potential is applied to the main control gate line CGLi at the time of writing, erasing, and reading, the potential of the sub control gate line CGLi-j in the non-selected block BLKi-j may change due to capacitive coupling. There is.

  However, the change in the potential of the sub control gate line CGLi-j in the non-selected block BLKi-j does not give any inconvenience to the write, erase, and read operations.

  The select gate lines SSLi and GSLi in the selected row are common to all the blocks BLKi-j in the selected row.

  Therefore, at the time of data writing, the ground potential or the power supply potential VCC is applied to the select gate lines SSLi and GSLi of all the blocks BLKi-j in the selected row via the N-channel MOS transistors 35-1 and 35-3. The At the time of data erasing, VCC is applied to select gate lines SSLi and GSLi of all blocks BLKi-j in the selected row via N-channel MOS transistors 35-1 and 35-3. At the time of data reading, the power supply potential VCC is applied to the select gate lines SSLi and GSLi of all the blocks BLKi-j in the selected row via the N-channel MOS transistors 35-1 and 35-3 (Table 3 and Table 3). (See Table 4).

  FIG. 32 shows a first example of a data rewrite operation in units of bytes applied to the EEPROM of FIGS.

The byte-unit data rewrite operation of the present invention includes the following four main steps.
i. Data is reversely read from the memory cells in the selected block and held in a sense amplifier circuit having a latch function.
ii. The byte data is overwritten on the data held in the sense amplifier circuit having the latch function.
iii. Erase the data in the memory cells in the selected block.
iv. Write the data held in the sense amplifier circuit having the latch function to the memory cell in the selected block.

  The feature of the data rewrite operation in byte units according to the present invention is that reverse reading of data for one page in the selected row is performed when performing data rewrite in byte units, as is apparent from the rewrite operation in FIG. Rather, only the data of the selected block BLKi-j in the selected row (data that is a multiple of 1 byte) is reversely read. That is, since it is not necessary to perform reverse reading on the data of the memory cells in the non-selected block in the selected row, unnecessary reading, erasing, and writing operations on the memory cells that are not subjected to data rewriting can be eliminated.

  Therefore, if the same data is rewritten, the rewrite operation of the present invention can reduce the number of page read, erase, and write operations compared with the rewrite operation of FIG. Can be improved.

  As described above, according to the present invention, data can be rewritten in units of bytes without deteriorating rewrite characteristics even though the same process as that of the flash EEPROM can be manufactured and the same rewrite method is applied. Is possible.

  FIG. 33 shows a second example of the data rewrite operation in units of bytes applied to the EEPROM of FIGS.

The byte-unit data rewrite operation of the present invention includes the following four main steps.
i. Data is reversely read from one page of memory cells in the selected row and held in a sense amplifier circuit having a latch function.
ii. The byte data is overwritten on the data held in the sense amplifier circuit having the latch function.
iii. Erase the data in the memory cells in the selected block.
iv. Write the data held in the sense amplifier circuit having the latch function to the memory cell in the selected block.

  The byte-unit data rewrite operation of the present invention is characterized in that reverse reading is performed on one page of memory cells as compared with the rewrite operation of FIG. That is, in the present invention, the data of the memory cell for one page is reversely read, but erasing and writing are performed only for the selected block in the selected row. For this reason, unnecessary erase and write operations on the data of the memory cells of the non-selected block in the selected row can be eliminated.

  In this case, in the reverse read operation, all the blocks BLKi-j in the selected row are multiple-selected by the sub-decoder so that all the blocks BLKi-j in the selected row are selected.

  Compared with the rewriting operation of FIG. 27, the rewriting operation of the present invention can reduce the number of page erasing and writing operations, and can improve the substantial page rewriting characteristic (the number of rewriting operations).

  As described above, according to the present invention, data can be rewritten in units of bytes without deteriorating rewrite characteristics even though the same process as that of the flash EEPROM can be manufactured and the same rewrite method is applied. Is possible.

  FIG. 34 shows an example of the well layout in the memory cell array region.

  In a flash EEPROM, all memory cell units (memory cells and select transistors) are usually formed in one well (for example, a twin well, ie, a p-type well in an n-type well formed in a p-type substrate). The However, in the present invention, the sub control gate driver is arranged between the memory cell units. The sub-control gate driver has a role of transmitting a high potential to the sub-control gate, and if this is formed in the same well as the memory cell, the threshold increases due to the back gate bias effect, or the sub-control gate driver operates according to the well potential. May become unstable.

  Therefore, in this example, common wells are provided in the column-direction blocks BLKi-j, and the row-direction blocks BLKi-j are arranged in different wells. In this case, the sub-control gate driver is formed outside the well, that is, on the p-type substrate, and the above-described problem can be avoided.

  It should be noted that all memory cell units and sub-control gate drivers can be arranged in one well by devising the potential applied to the well during writing and erasing.

  However, in this case, an increase in the threshold due to the back gate bias effect cannot be avoided.

  FIG. 35 shows another example of the configuration of a plurality of blocks and sub-control gate drivers arranged in one row.

  This example is a modification of the circuit of FIG. 31, and is characterized by the connection relationship of N-channel MOS transistors 36-0, 36-1, 36-2, and 36-3.

  In each block BLKi-j (j = 0, 1, 2, 3), sub-control gate lines CGLi-0, CGLi-1, CGLi-2, and CGLi-3 are arranged, respectively. The sub control gate line CGLi-j (j = 0, 1, 2, 3) is connected to a memory cell that is a multiple of 1 byte (for example, 16 bytes) arranged in the block BLKi-j.

  The sub control gate line CGLi-j is connected to the sub decoder 29 via an N channel MOS transistor 36-j as a drive circuit constituting the sub control gate driver 28, respectively.

  On / off of the N-channel MOS transistor 36-j is determined by the potential of the main control gate line CGLi. In the selected row, since the boosted potential Vprog or the power supply potential VCC is applied to the main control gate line CGLi, all the N channel MOS transistors 36-0, 36-1, 36-2, 36-3 in the selected row are turned on. It becomes a state.

  At the time of data writing, a high potential Vprog for writing is supplied from the sub decoder 29 to the sub control gate line CGLi-j of the selected block BLKi-j. A ground potential is supplied from the sub decoder 29 to the sub control gate line CGLi-j of the non-selected block BLKi-j.

  At the time of data erasure, the ground potential is supplied from the sub decoder 29 to the sub control gate line CGLi-j of the selected block BLKi-j. VCC is supplied from the sub decoder 29 to the sub control gate line CGLi-j of the non-selected block BLKi-j.

  At the time of data reading, a read potential (ground potential or power supply potential VCC) is supplied from the sub decoder 29 to the sub control gate line CGLi-j of the selected block BLKi-j. The ground potential is supplied from the sub decoder 29 to the sub control gate line CGLi-j of the non-selected block BLKi-j (see Tables 3 and 4).

  On the other hand, in the block BLKi-j in the non-selected row, since the ground potential is applied to the gate of the N channel MOS transistor 36-j, the N channel MOS transistor 36-j is turned off.

  At the time of data writing, the ground potential or the power supply potential VCC is applied to the select gate lines SSLi and GSLi of all the blocks BLKi-j in the selected row via the N-channel MOS transistors 35-1 and 35-3. The At the time of data erasing, VCC is applied to select gate lines SSLi and GSLi of all blocks BLKi-j in the selected row via N-channel MOS transistors 35-1 and 35-3. At the time of data reading, the power supply potential VCC is applied to the select gate lines SSLi, GSLi of all the blocks BLKi-j in the selected row via the N-channel MOS transistors 35-1, 35-3 (Table 3 and Table 3). (See Table 4).

  Even in such a configuration, it is possible to perform read, erase or write operations in units of blocks, and naturally, the data rewrite method in units of bytes of FIGS. 32 and 33 can be applied.

  Therefore, unnecessary read, erase, and write operations on memory cells that are not subjected to data rewrite can be eliminated, and substantial page rewrite characteristics (number of rewrites) can be improved.

  FIG. 36 shows an improved example of the byte type EEPROM of FIG. FIG. 37 shows only two rows adjacent to each other in the memory cell array 11 of FIG.

  In the example of FIG. 28, the predecoder 12a, the row decoder 12b, and the main control gate / select gate driver 12c are arranged together at one end in the row direction of the memory cell array 11.

  In contrast, according to the present invention, the predecoder 12a, the row decoder 12b, and the main control gate / select gate driver 12c are arranged at one end and the other end of the memory cell array 11 in the row direction.

  For example, predecoders PD0, PD2,... For selecting even-numbered rows and row decoders RD0, RD2,. And row decoders RD1, RD3,... Are arranged at the other end of the memory cell array 11 in the row direction. Further, main control gate / select gate drivers 0, 2,... For applying a predetermined potential to even-numbered rows are arranged at one end in the row direction of the memory cell array 11, and a main control gate for applying a predetermined potential to odd-numbered rows. The select gate drivers 1, 3,... Are arranged at the other end in the row direction of the memory cell array 11.

  Thus, the layout of the predecoder 12a, the row decoder 12b, and the main control gate / select gate driver 12c can be easily determined at the time of circuit design.

  That is, the main control gate / select gate driver 12c generates, for example, a high potential for writing and transmits this to the main control gate line CGLi, so that the circuit size tends to increase. Therefore, if the predecoder 12a, the row decoder 12b, and the main control gate / select gate driver 12c are arranged together only at one end in the row direction of the memory cell array 11, it is very difficult to determine the layout of these circuits on the chip. Become.

  As described above, if the predecoder 12a, the row decoder 12b, and the main control gate / select gate driver 12c are arranged at one end and the other end in the row direction of the memory cell array 11, the space on the chip can be effectively utilized. The circuit block can be stored compactly on the chip.

  As shown in the figure, the drive circuit for driving the control gate line CGLi and the drive circuit for driving the select gate lines SSLi, GSLi in the same block BLKi-j are both collectively referred to as a driver i. It arrange | positions at one end or other end of.

  As a result, the timing difference between the signal applied to the memory cell in the selected block BLKi-j and the signal applied to the select transistor is eliminated, and malfunction during writing and reading can be prevented, thus improving reliability.

  Furthermore, in this example, it is desirable to drive two select gate lines SSLi and GSLi and one control gate line CGL at the same time. Further, since the area of the high breakdown voltage transistor (driver) becomes large, if the select gate lines SSLi and GSLi and the control gate line CGL are arranged as one set, the pattern on the chip becomes uniform. Therefore, it is possible to prevent the thinning of the word line due to the electroloading effect that occurs when the pattern is not uniform.

  38 and 39 show an improved example of the sense amplifier circuit.

  The example of FIG. 38 is a configuration example when a differential sense amplifier is used. In this case, 1-bit data may be stored in two memory cell units as complementary data. Further, since data reading is performed by detecting a slight difference between the signal amounts (potentials) output from the two memory cell units and increasing this difference, high-speed reading is possible.

  In addition, since two memory cell units are paired and 1-bit data is stored in the pair of memory cell units, even if the rewrite characteristics of one memory cell unit deteriorate due to repeated data rewrite operations, the other memory cell unit If the rewriting characteristics of the cell unit are good, the reliability is not lowered.

  The example of FIG. 39 is an example in which one sense amplifier circuit is commonly connected to a plurality of (for example, two) bit lines. In this case, for example, data rewriting in units of bytes in the block BLKi-j is performed in two steps. That is, the first rewrite is performed on the memory cell units connected to the even-numbered bit lines, and the second rewrite is performed on the memory cell units connected to the odd-numbered bit lines.

  When the sense amplifier circuit of this example is used, when data is read out to one bit line, the other bit line is set to a fixed potential (for example, ground potential) (shield bit line reading method). As a result, problems such as erroneous writing in unselected cells during reading can be avoided. Further, the EEPROM of this example can be applied when multi-value data is stored in one memory cell unit.

  FIG. 40 shows an improved example of the byte type EEPROM of FIG.

  In the example of FIG. 28, the memory cell array 11 is made up of a plurality of blocks BLKi-j (i = 0, 1,... N; j = 0, 1, 2, 3) arranged in rows and columns. Configured. In the present invention, on the premise of this, the number of main control gate drivers (including booster circuits) occupying a large area on the chip is reduced to facilitate the layout of circuit blocks on the chip.

  In this example, the number of blocks in the column direction is n (for example, 1024), and the number of blocks in the row direction is four. In this case, for example, one block is composed of 16-byte memory cells, and one page is composed of 64-byte memory cells.

  The main control gate driver 37 is provided corresponding to a plurality of rows, in this example, two rows, that is, two main control gate lines CGL (2 pages) adjacent to each other. That is, in the present invention, two main control gate lines CGL are driven by one main control gate driver 37. Each main control gate driver includes a booster circuit.

  The sub control gate driver 28 is provided corresponding to each block BLKi-j.

  The select gate driver 38 is provided corresponding to one row, that is, one control gate line CGL (one page). The predecoder 12a and the row decoder 12b are also provided corresponding to one row, that is, one control gate line CGL.

  The row address signal is input to the predecoder 12 a and the subdecoder 29 via the address register 19. The four blocks BLKi-j in one row are selected by the predecoder 12a and the row decoder 12b. Further, the sub-decoder 29 selects one of the four selected blocks BLKi-j.

  Note that the sub-decoder 29 has a function of selecting a plurality of blocks in one selected row or all blocks (four blocks in this example) in one selected row. Also good.

  In the present invention, as in the example of FIG. 28, data can be read, erased and written in units of blocks. Therefore, in the data rewriting operation in byte units, it is not necessary to read out data for one page to the sense amplifier circuit having a latch function, and the substantial page rewriting characteristics can be improved.

  In the present invention, for example, when the selected block is BLKi-j, the main control gate driver 37 applies the two main control gate lines CGLi and CGLi + 1 of the row to which the selected block BLKi-j belongs and the row adjacent thereto to A predetermined potential corresponding to the operation mode is applied. That is, since one main control gate driver 37 is provided in common for the two main control gate lines CGLi and CGLi + 1, the number of main control gate drivers 37 can be reduced, the layout is simplified, and the load at the time of circuit design is increased. Mitigation can be achieved.

  The select gate driver 38 applies a predetermined potential corresponding to the operation mode to the select gate lines SSL and GSL of the row to which the selected block BLKi-j belongs.

  The sense amplifier circuit 13 having a latch function latches read data and write data. Read data (output data) is output to the outside of the memory chip via the column selection circuit 15 and the input / output buffer 18. Write data (input data) is latched by the sense amplifier circuit 13 having a latch function via the input / output buffer 18 and the column selection circuit 15.

  The command signal is input to the command decoder 26 via the data input / output buffer 18 and the command register 25. The control circuit 17 receives signals such as an output signal of the command decoder 26, a command latch enable signal CLE, a chip enable signal / CE, and a write enable signal / WE.

  The signal generation circuit (boost circuit) 27 generates a potential to be applied to the control gate line CGL and the select gate lines SSL, GSL under the control of the control circuit 17, and this potential is supplied to the main control gate driver 37 and the select gate driver 38. Supply.

  FIG. 41 shows an example of the configuration of the predecoder PDi.

In this example, the number of rows, that is, the number of control gate lines CGL (number of blocks) is assumed to be 1024 (2 ¹⁰ ). In this case, one row can be selected by 10-bit row address signals a1, a2,.

  The row address signals a2, a3, and a4 are input to the NAND circuit 30-1, the row address signals a5, a6, and a7 are input to the NAND circuit 30-2, and the row address signals a8, a9, and a10 are input to the NAND circuit. 30-3. The output signal of the NAND circuit 30-1 becomes a signal D via the inverter 31-1, and the output signal of the NAND circuit 30-2 becomes a signal E via the inverter 31-2. The output signal becomes the signal F via the inverter 31-3.

  Different row address signals a1, a2,..., A10 are input to each predecoder PDi. Only the output signals a1, D, E, F of the predecoder PDi belonging to one selected row are all “1”.

  FIG. 42 shows an example of the configuration of the row decoder RDi, the main control gate driver 37, and the select gate driver 38.

  The row decoder RDi includes a NAND circuit 32 and an inverter 33. Output signals D, E, and F of the predecoder PDi are input to the NAND circuit.

  The select gate driver 38 includes N channel MOS transistors 35-1 and 35-3 as drive circuits. In the selected row, the output signal of the row decoder RDi becomes VCC, so that the N-channel MOS transistors 35-1 and 35-3 are turned on. Therefore, the signals SS and GS generated by the signal generation circuit 27 are supplied to the select gate lines SSLi and GSLi.

  The main control gate driver 37 includes a decode circuit 39, a booster circuit 34, and an N-channel MOS transistor 35-2 as a drive circuit.

  In the main control gate driver 37 provided in common for the selected row and the adjacent row, the output signal of the decode circuit 39 becomes VCC. Further, depending on the operation mode, the booster circuit is activated or deactivated, and the power supply potential VCC or the boosted potential is applied to the gate of the N-channel MOS transistor 35-2.

  For example, when writing data, in the main control gate driver 37 provided in common for the selected row and the adjacent row, the output potential VB of the booster circuit 34 becomes the boosted potential Vprog, and the N-channel MOS transistor 35-2 is Turns on. On the other hand, CG (= Vprog) generated by the signal generation circuit 27 is transmitted to the selected row and main control gate lines CGLi and CGLi + 1 of the row adjacent thereto via the N-channel MOS transistor 35-2.

  At the time of data erasing, in the main control gate driver 37 provided in common for the selected row and the adjacent row, the output potential VB of the booster circuit 34 becomes the power supply potential VCC, and the N-channel MOS transistor 35-2 Turns on. On the other hand, CG (= 0 V) generated by the signal generation circuit 27 is transmitted to the selected row and the main control gate lines CGLi and CGLi + 1 of the row adjacent thereto via the N-channel MOS transistor 35-2.

  At the time of data writing, in the main control gate driver 37 provided in common for the selected row and the adjacent row, the output potential VB of the booster circuit 34 becomes the power supply potential VCC, and the N-channel MOS transistor 35-2 Turns on. On the other hand, CG (= 0 V or VCC) generated by the signal generation circuit 27 is transmitted to the selected row and the main control gate lines CGLi and CGLi + 1 of the row adjacent thereto via the N-channel MOS transistor 35-2. .

  In main control gate driver 37 provided in common to two non-selected rows adjacent to each other, output signal VB of booster circuit 34 becomes the ground potential, and this ground potential is applied to the gate of N-channel MOS transistor 35-2. Is done. Therefore, N channel MOS transistor 35-2 is turned off.

  FIG. 43 shows an example of the configuration of a plurality of blocks and sub-control gate drivers arranged in two adjacent rows.

  In this example, a case will be described in which four blocks are arranged in one row corresponding to the circuit block of FIG.

  In each block BLKi-j, BLK (i + 1) -j, sub-control gate lines CGLi-j, CGL (i + 1) -j are arranged (j = 0, 1, 2, 3), respectively. The sub control gate line CGLi-j is connected to a memory cell of a positive multiple of 1 byte (for example, 16 bytes) arranged in each block BLKi-j, and the sub control gate line CGL (i + 1) -j is Each is connected to a memory cell of a positive multiple of 1 byte (for example, 16 bytes) arranged in the block BLK (i + 1) -j.

  The sub control gate line CGLi-j is connected to the main control gate line CGLi via an N channel MOS transistor 36-j as a drive circuit constituting the sub control gate driver 28. Sub-control gate line CGL (i + 1) -j is connected to main control gate line CGLi + 1 via N-channel MOS transistor 40-j as a drive circuit constituting sub-control gate driver 28, respectively.

  On / off of the N-channel MOS transistors 36-j and 40-j is controlled by the sub-decoder 29. The subdecoder 29 has a function of selecting one N-channel MOS transistor 36-j (one block). For example, when the block BLKi-1 is selected, the N-channel MOS transistor 36-1 is turned on. At this time, the main control gate line CGLi and the sub control gate line CGLi-1 are electrically connected.

  Note that the sub-decoder 29 may have a function of selecting a plurality of or all N-channel MOS transistors in one row.

  Also in the EEPROM of the present invention, the memory cell array is composed of a plurality of blocks arranged in a matrix in the row direction and the column direction, and data can be read, erased, and written in units of blocks. Therefore, also in the present invention, the data rewriting operation in units of bytes shown in FIGS. 32 and 33 can be applied. That is, when performing data rewriting in units of bytes, it is possible to read only the data of the selected block in the selected row (data that is a multiple of 1 byte) without reading out data for one page in the selected row.

  Therefore, unnecessary read, erase, and write operations on memory cells that are not subjected to data rewrite can be eliminated, and substantial page rewrite characteristics (number of rewrites) can be improved.

  In the present invention, one main control gate driver (including a booster circuit) is commonly used for a plurality of (for example, two) rows adjacent to each other. Therefore, the width of the main control gate driver having a large size in the column direction can be made larger than the width of one row, and the layout of the main control gate driver can be easily performed at the time of circuit design.

  Further, at the time of writing, since the high potential Vprog is applied to the main control gate line of the selected row and the power supply potential VCC is applied to the select gate line, only a plurality of main control gate drivers that must output the high potential Vprog are provided. The select gate driver is arranged for each row.

  In this case, for example, at the time of writing, the high potential Vprog is applied to the two main control gate lines, but this high potential Vprog is transmitted only to the sub control gate lines in the selected block selected by the sub decoder. Therefore, there is no problem in operation.

  FIG. 44 shows an example of arrangement of sub-decoders.

  In the present invention, one page of memory cells in the memory cell array 11 is divided into a plurality of blocks, and a plurality of blocks BLKi-j are provided in the row direction. A sub control gate driver 28 is disposed between the blocks BLKi-j in the row direction. The sense amplifier circuit 13 is provided corresponding to the block BLKi-j arranged in the row direction.

  Therefore, a space is formed at a position between the sense amplifier circuits 13 and corresponding to the sub control gate driver 28. In this example, the sub-decoder 29 is arranged in this space.

  When the sub-decoder 29 is arranged at a plurality of locations corresponding to the sub-control gate driver 28 as in this example, the space on the chip is used more effectively than when the sub-decoder 29 is arranged at one location. Can contribute to reducing the chip size.

  45 to 47 show an example of an EEPROM to which the present invention can be applied.

  In the example of FIG. 45, sense amplifier circuits 13A and 13B, column selection circuits 15A and 15B, and data input / output buffers 18A and 18B having a latch function are arranged at both ends in the column direction of the memory cell array. In this example, the memory cell array is composed of three tracell units (see FIG. 26) 11-0 and a NAND cell unit 11-1. Of course, the memory cell array may be composed of only three tracell units.

  According to this example, since circuits for reading and writing operations such as a sense amplifier circuit are arranged at both ends in the column direction of the memory cell array, the layout of these circuits becomes easy, and the load at the time of circuit design can be reduced. .

  In the example of FIG. 46, the memory cell array is composed of a three-tracell unit 11-0 and a NAND cell unit 11-1, and the three-tracell unit 11-0 is arranged on the sense amplifier circuit 13 side. The cell is used as a cache memory.

  According to this example, since the data of the NAND cell unit 11-1 can be temporarily stored in the 3 tracell unit (cache memory) in units of blocks, the data can be read at high speed.

  In the example of FIG. 47, a plurality of memory circuits 42 a and 42 b are arranged in one chip 41. Each of the memory circuits 42a and 42b can perform a read operation, a write operation, and an erase operation independently of each other. Thus, for example, the memory circuit 42b can perform a write operation while the memory circuit 42a is performing a read operation. The EEPROM of the present invention is used for at least one of the memory circuits 42a and 42b.

  According to this example, since two different operations can be performed simultaneously, data processing can be performed efficiently.

  By the way, in the invention described so far, a cell unit including one memory cell having a stack gate structure and two select transistors connected to both ends of the memory cell has been a main component.

  As described above, according to such a cell unit, various features such as data rewriting in byte (or page) units can be obtained.

  However, when the cell unit is composed of three transistors (only one memory cell), the cell size per memory cell is larger than that of a normal NAND flash EEPROM. It is not always advantageous to increase the capacity.

  Therefore, in the following invention, a novel cell unit structure or data writing method capable of reducing the cell size per memory cell while maintaining data rewriting in byte (or page) units will be described.

  First, a conventional NAND flash EEPROM will be examined.

  A conventional NAND flash EEPROM has a structure most suitable for reducing the cell size per memory cell, for example, because 16 memory cells connected in series are arranged in one memory cell unit. is doing.

  However, with such a structure, the feature of reducing the cell size is obtained, but the feature of rewriting data in byte (or page) units is lost.

  The reason why the conventional NAND flash EEPROM cannot rewrite data in byte (or page) units will be described.

  In order to understand the reason why data cannot be rewritten in units of bytes (or pages) in the NAND flash EEPROM, it is first necessary to understand the data rewrite operation of the NAND flash EEPROM.

  The data rewrite operation of the NAND flash EEPROM is performed in units of blocks.

  First, batch erase of data (operation for extracting electrons from the floating gate and lowering the threshold value) is performed on all memory cells of the NAND cell unit in the selected block. Thereafter, for example, data writing is sequentially executed in page units from the memory cell on the source side to the memory cell on the drain side of the NAND cell unit in the selected block.

A specific data write operation will be described with reference to FIGS.
In this example, it is assumed that data is written to a memory cell connected to the control gate line CGL1.

  First, 0 V is applied to the source side (source line side) select gate line GSL to cut off the source side select transistor. Further, the power supply potential VCC is applied to the drain side (bit line side) select gate line SSL to turn on the drain side select transistor.

  Further, the potential of the bit line to which the memory cell M1 for performing “0” writing (injecting electrons into the floating gate and raising the threshold) is set to 0 V, and “1” writing (operation for maintaining the erased state) is performed. The potential of the bit line connected to the memory cell M2 to be performed is set to the power supply potential VCC.

  At this time, the channel potential of all the memory cells in the NAND cell unit including the memory cell M1 to which “0” is written becomes 0 V, and all the memory cells in the NAND cell unit including the memory cell M2 to which “1” is written are written. The channel potential is precharged to VCC-Vth (Vth is the threshold voltage of the select transistor). Thereafter, the select transistor on the drain side (bit line side) in the NAND cell unit including the memory cell M2 to which “1” is written is cut off.

  Thereafter, the potential of the control gate line (selection) CGL1 rises from 0V to the power supply potential VCC (eg, 3.3V) and from the power supply potential VCC to the write potential Vprog (eg, 18V). Further, the potentials of the control gate lines (non-selected) CGL0, CGL2,... CGL15 rise from 0V to the power supply potential VCC and from the power supply potential VCC to Vpass (VCC <Vpass (for example, 9V) <Vprog).

  At this time, since the channel potential is 0 V in the memory cell M1 to which “0” is written, a high voltage is applied to the tunnel insulating film between the floating gate and the channel, and electrons are transferred from the channel to the floating gate. Moving. On the other hand, in the memory cell M2 to which “1” is written, since the channel is in a floating state, the channel potential rises to Vch due to capacitive coupling. Therefore, in the memory cell M1 in which “1” is written, a high voltage is not applied to the tunnel insulating film between the floating gate and the channel, and the erased state is maintained.

  Here, the Vpass applied to the non-selected control gate lines CGL0, CGL2,... CGL15 in the selected block will be considered.

  Writing “1” to the memory cell M2 is achieved by suppressing the injection of electrons to the floating gate of the memory cell M2 and maintaining the erased state in the memory cell M2 during data writing. In order to maintain the memory cell M2 in the erased state, at the time of data writing, the channel potential of each memory cell in the NAND cell unit including the memory cell M2 is sufficiently increased by capacitive coupling, and the floating gate of the memory cell M2 The voltage applied to the tunnel insulating film between the channels may be relaxed.

  Meanwhile, the channel potential of each memory cell in the NAND cell unit including the memory cell M2 depends on Vpass applied to the non-selected control gate lines CGL0, CGL2,. Therefore, the higher Vpass is, the higher the channel potential of each memory cell in the NAND cell unit including the memory cell M2 is, so that erroneous writing to the memory cell M2 is prevented.

  However, when Vpass is increased, erroneous writing is likely to occur in the non-selected memory cell M3 in the NAND cell unit including the memory cell M1 to which “0” is written.

  That is, the channel potential of each memory cell in the NAND cell unit including the memory cell M1 is maintained at 0V. For this reason, when Vpass is close to the write potential Vprog, “0” is also written to the unselected memory cell M3. Therefore, in order to prevent erroneous writing to unselected memory cells in the NAND cell unit including the memory cell M1, it is necessary to make Vpass as low as possible.

  As described above, Vpass applied to the non-selected control gate lines CGL0, CGL2,... CGL15 in the selected block is not required to be too high or too low. The optimum value is set such that VCC <Vpass (for example, 9V) <Vprog so that “0” is not written to the memory cell M3.

  The data rewriting operation of the NAND flash EEPROM has been described in detail above. Therefore, the following explains why the data rewriting operation of the NAND flash EEPROM is not performed in units of bytes (or pages).

  Assume that the data rewrite operation is performed in units of bytes (or pages) in the NAND flash EEPROM.

  In this case, the same control gate line, for example, the control gate line CGL1 is repeatedly selected, and only the memory cell connected to the control gate line CGL1 is repeatedly rewritten. Conceivable. In such a situation, the operation of erasing data in the memory cell connected to the control gate line CGL1 and the operation of writing data to the memory cell connected to the control gate line CGL1 are repeatedly performed.

  However, at this time, Vpass is repeatedly applied to the control gates of the non-selected memory cells in the selected block at the time of data writing.

  Therefore, in the NAND flash EEPROM, when the data rewrite operation in units of bytes (or pages) is repeated many times, the threshold value of the non-selected memory cells in the selected block gradually rises due to Vpass (electrons gradually enter the floating gate). There is a possibility that erroneous writing occurs.

  In order to eliminate this possibility, it is necessary to lower Vpass or change to block-based rewriting.

  However, as described above, Vpass is applied to a selected memory cell in which “1” is written or a non-selected memory cell in the same cell unit as a memory cell in which “0” is written in one data write operation. , “0” writing (erroneous writing) is set to an optimum value so that it is practically impossible to lower this value.

  Therefore, as a result, in the NAND flash EEPROM, data rewriting in byte (or page) units becomes impossible, and data rewriting in block units is performed.

  In the following, as in a NAND flash EEPROM, the cell size per memory cell can be reduced, and data rewriting in units of bytes (or pages) can be achieved by lowering Vpass. A simple cell unit structure or data writing method will be described.

  FIG. 50 shows a memory cell unit of the byte type EEPROM of the present invention. FIG. 51 shows an equivalent circuit of the memory cell of FIG.

  The memory cells MC1 and MC2 have a control gate and a floating gate, and have the same structure as the memory cell of the flash EEPROM. Memory cells MC1 and MC2 are connected in series with each other, and select transistors ST1 and ST2 are respectively connected to both ends thereof. The select transistor ST1 is connected to the bit line via the bit line contact portion BC, and the select transistor ST2 is connected to the source line SL.

  Memory cells MC1 and MC2 and select transistors ST1 and ST2 constitute one memory cell unit, and a memory cell array is realized by arranging a plurality of memory cell units in an array.

  The memory cell unit of the present invention can be considered as two NAND memory cells in a NAND flash EEPROM (2 NAND cells).

  However, in the present invention, the number of memory cells in the memory cell unit is not limited to two. For example, as long as the conditions described later are satisfied, a plurality (three, four, five, etc.) is set. can do. In some cases, the number of memory cells in the memory cell unit may be set to 16 as in the conventional NAND flash EEPROM.

  The structural advantages of the byte type EEPROM of the present invention will be described.

  The structure of the memory cell part of the byte type EEPROM of the present invention is the same as the structure of the memory cell part of the NAND type flash EEPROM. However, normally, the number of memory cells in the cell unit of the byte type EEPROM of the present invention is smaller than the number of memory cells (for example, 16) in the cell unit of the NAND type flash EEPROM.

  Therefore, in the byte type EEPROM of the present invention, the NAND type flash EEPROM process can be adopted as it is, so that although the byte unit can be erased (this will be described later), the storage capacity can be increased. In addition, production costs can be reduced.

  For example, in the present invention, a case where the number of memory cells in a cell unit is two is considered.

In the present invention, when the design rule is 0.4 [μm], the short side length a of the two memory cells is 1.2 [μm] and the long side length b is 3.96 [μm]. The area per memory cell ([short side length a × long side length b] / 2) is 2.376 [μm ² ]. On the other hand, in a NAND flash EEPROM (16 NAND cells) in which a cell unit is composed of 16 memory cells, when the design rule is 0.4 [μm], the area per memory cell is 1.095 [μm ^2]. ].

  That is, when the memory cell unit (2 NAND cell) of the present invention is employed, the area per memory cell is about twice the area per memory cell of 16 NAND cells.

In the conventional byte type EEPROM as shown in FIGS. 65 and 66, when the design rule is 0.4 [μm], the area per memory cell is 36 [μm ² ]. Also, when a cell unit (3 tracell or 1 NAND cell) in which one memory cell is sandwiched between two select transistors is adopted, if the design rule is 0.4 [μm], the area per memory cell is 3.84 [μm ² ].

  That is, the area per memory cell of the memory cell unit (2 NAND cell) of the present invention can be made smaller than that of a conventional byte EEPROM or 1 NAND cell.

Table 5 shows a comparison of areas per memory cell according to the structure of the memory cell unit.

  As is apparent from this table, the area per memory cell of the memory cell unit (2 NAND cell) of the present invention is not as large as that of the NAND flash EEPROM (16 NAND cell), but 1 NAND cell (3 tracell) It will be about 60%.

  Therefore, according to the cell unit structure of the present invention, the reduction of the memory cell area can contribute to the increase of the memory capacity of the byte type EEPROM, the reduction of the chip area, the reduction of the manufacturing cost, and the like.

  Further, since the byte type EEPROM of the present invention can be manufactured by the same process as the NAND type flash EEPROM, it can be easily applied to a logic mixed nonvolatile memory.

  In addition, since the memory cell of the byte type EEPROM of the present invention has the same structure as the memory cell of the NAND type flash EEPROM, the rewriting method of the flash EEPROM, that is, the rewriting using the FN tunnel phenomenon, can be seen for one memory cell. The method can be adopted as it is. Therefore, in addition to a reduction in manufacturing cost, it is possible to reduce development cost.

  By the way, according to the cell unit structure of the present invention, a plurality of (for example, two, three,...) Memory cells are connected between two select transistors. Therefore, similarly to the NAND flash EEPROM, when rewriting data in units of bytes (or pages) is repeatedly performed, there arises a problem of erroneous writing to unselected memory cells in a selected block in which Vpass is applied to the control gate.

This problem is solved as follows.
In a NAND flash EEPROM, Vpass is not selected in the same cell unit as a memory cell that performs “1” programming (maintains an erased state) or a memory cell that performs “0” programming in one programming operation. The optimum value is set on condition that “0” writing (erroneous writing) does not occur in the memory cell.

  The optimum value is determined completely independent of the power supply potential VCC and the potential Vread applied to the control gate of the non-selected memory cell at the time of reading. Usually, VCC (for example, 3.3 V) <Vpass ( For example, 9V) <Vprog (for example, 18V) was set.

  In the present invention, Vpass is set to the power supply potential VCC (for example, 3.3 V) or the potential Vread (for example, 4.5 V) applied to the control gate of the non-selected memory cell at the time of reading.

  These VCC and Vread are lower than the value of Vpass (for example, 9 V) used in the NAND flash EEPROM.

  That is, in the present invention, Vpass is set to the power supply potential VCC or the potential Vread to be applied to the control gate of the non-selected memory cell at the time of reading, that is, by making it lower than the value of Vpass used in the NAND flash EEPROM. This prevents the problem of erroneous writing of unselected memory cells in the selected block when data rewriting in units of bytes (or pages) is repeated.

  Further, in the present invention, by setting Vpass to VCC or Vread, it is not necessary to newly provide a circuit for generating Vpass, so that the configuration of the control gate driver is simplified, the control gate driver can be reduced, and the layout can be reduced. Effects such as simplification, design and shortening the development period can be obtained.

  On the other hand, in the present invention, Vpass is set to the power supply potential VCC or the potential Vread to be applied to the control gate of the non-selected memory cell at the time of reading, so that “1” write is performed in one data write operation (erase state). The question arises that the channel potential of the memory cell may not be sufficiently increased.

  Therefore, in the present invention, the number of memory cells in the cell unit, the initial potential of the channel of the memory cell that performs “1” writing, and the memory cell's channel potential so that the channel potential of the memory cell that performs “1” writing increases sufficiently. Sets coupling ratio between control gate and channel.

  For example, assuming that the initial potential of the channel of the memory cell in which “1” is written and the coupling ratio between the control gate and the channel of the memory cell are the same as those of the NAND flash EEPROM, FIG. 50 and FIG. Thus, if the number of memory cells in the cell unit is two, the channel potential of the memory cell to be written “1” can be raised to the same level as that of the NAND flash EEPROM (this point will be described later). This will be described in detail in the description of the data write operation to be described.)

  Thus, according to the present invention, firstly, since the cell unit structure is exactly the same as that of the NAND flash EEPROM, it is possible to achieve a reduction in cell size, an increase in memory capacity, a reduction in cost, and the like.

  Second, the potential Vpass applied to the unselected control gate line in the selected block at the time of data writing is set to the power supply potential VCC or the potential Vread applied to the unselected control gate line at the time of reading. Therefore, the problem of erroneous writing of unselected memory cells in the selected block can be prevented, and data rewriting in byte (or page) units becomes possible.

  Thirdly, even if Vpass is set to VCC or Vread, the number of memory cells in the cell unit and the memory for performing “1” write so that the channel potential of the memory cell for performing “1” write is sufficiently increased. The initial potential of the cell channel and the coupling ratio between the control gate of the memory cell and the channel are set to appropriate values. It is also possible to prevent erroneous writing to the memory cell that performs “1” writing.

  Hereinafter, an erase operation, a write operation, and a read operation of the byte EEPROM according to the present invention will be sequentially described.

・ Erase operation
As shown in FIG. 52, the ground potential VSS is applied to the selected control gate line (word line) CGL11 in the selected block, and the unselected control gate line (word line) CGL12 in the selected block enters a floating state. In addition, the select gate lines SSL1 and GSL1 in the selected block and the control gate lines (word lines) CGL21 and CGL22 and the select gate lines SSL2 and GSL2 in the non-selected block are also in a floating state.

  Thereafter, for example, erase pulses of 21 [V] and 3 [ms] are applied to the bulk (cell P well). At this time, in the memory cell connected to the selected control gate line CGL11 in the selected block, an erase voltage (21 [V]) is applied between the bulk and the control gate line, and electrons in the floating gate are FN (Fowler-Nordheim). ) Moves to the bulk due to the tunnel phenomenon.

  As a result, the threshold voltage of the memory cell connected to the selected control gate line CGL11 in the selected block is about −3 [V]. Here, the selected memory cell is erased by one erase pulse so that the threshold voltage is about −3 [V].

  On the other hand, the unselected control gate line CGL12 in the selected block and the control gate lines CGL21 and CGL22 in the unselected block are set in a floating state.

  Therefore, for example, when an erase pulse of 21 [V] and 3 [ms] is applied to the bulk (cell P well), the control gate lines CGL12 and CGL21 are caused by capacitive coupling between the control gate line in the floating state and the bulk. , CGL22 also increases in potential.

  Here, considering the coupling ratio between the control gate lines CGL12, CGL21, and CGL22 and the bulk, the control gate lines CGL12, CGL21, and CGL22 include a drive circuit (source of the MOS transistor), the drive circuit, and the control gate line (polysilicon). Metal wiring for connecting the silicon layer), silicide for forming the control gate line, and the like are connected.

  The coupling ratio depends on the capacitance parasitic on the control gate lines CGL12, CGL21, and CGL22 in the floating state. The capacitance includes a source junction capacitance of a MOS transistor as a drive circuit, an overlap capacitance between a source and a gate, a capacitance of a polysilicon layer and a metal wiring in a field region, a capacitance of a control gate line and a bulk (cell P well), and the like. included.

  However, most of the parasitic capacitances of the control gate lines CGL12, CGL21, and CGL22 are occupied by the capacitances of the control gate line and the bulk (cell P well).

  That is, the coupling ratio between the control gate lines CGL12, CGL21, and CGL22 and the bulk is a large value, for example, 0.9, and when the bulk potential increases, the potentials of the control gate lines CGL12, CGL21, and CGL22 also increase. To do.

Therefore, the occurrence of the FN tunnel phenomenon can be prevented in the memory cells connected to the non-selected control gate line CGL12 in the selected block and the memory cells connected to the control gate lines CGL21 and CGL22 in the non-selected block.
Thus, the erase operation is completed.

  After the erase operation, for example, an erase verify operation is performed to verify whether or not the threshold voltages of all the memory cells connected to the selected control gate line CGL11 in the selected block are less than −1 [V]. .

・ Write operation
As shown in FIG. 53, a case will be described in which writing is performed on a memory cell connected to the control gate line CGL1. It is assumed that all the memory cells to be written are in an erased state.

  First, the source side select gate line GSL in the selected block becomes the ground potential VSS, and the drain side select gate line SSL becomes the power supply potential VCC. As a result, the source side select transistors ST21 and ST22 are cut off, and the drain side select transistors ST11 and ST12 are turned on.

  In addition, the potential of the bit line BL to which the memory cell MC11 to which “0” is written is connected is set to VSS, and the potential of the bit line BL to which the memory cell (write-inhibited cell) MC12 to which “1” is written is connected. Set to VCC. In addition, the potentials of the control gate lines CGL1 and CGL2 are set to the ground potential VSS. At this time, the channel potentials of the memory cells MC11 and MC21 become the ground potential VSS, and the channels of the memory cells MC12 and MC22 are precharged to VCC-Vth (Vth is the threshold voltage of the select transistor ST12).

  Thereafter, the potentials of the control gate lines CGL1 and CGL2 are set to the power supply potential VCC (for example, 3.3V) or the potential Vread (for example, 4.5V) to be applied to the non-selected control gate line at the time of reading. Further, the potential of the selection control gate line CGL1 rises from VCC or Vread to the write potential Vprog (for example, 18V).

  At this time, in the selected memory cell MC11, since a large potential difference occurs between the channel (= VSS) and the control gate line CGL1 (= Vprog), electrons are injected from the channel into the floating gate due to the FN tunnel phenomenon. Thereby, “0” writing to the selected memory cell MC11 is completed.

  The initial potential of the channel of the selected memory cell MC12 before applying a high potential to the control gate line, that is, before channel boosting is set to VCC-Vth and is in a floating state. Therefore, after that, when the potential of the selected control gate line CGL1 becomes Vprog and the potential of the non-selected control gate line CGL2 becomes VCC or Vread, the channel potential of the selected memory cell MC12 also automatically rises due to capacitive coupling.

  That is, in the selected memory cell MC12, the potential difference between the channel (= Vch) and the control gate line CGL1 (= Vprog) is reduced, and injection of electrons into the floating gate due to the FN tunnel phenomenon is suppressed. Thereby, “1” writing to the selected memory cell MC12 is completed.

  By the way, in order to execute “1” write to the selected memory cell (write-inhibited cell) MC12, the channel potential (write-inhibit potential) Vch of the selected memory cell MC12 is sufficiently increased and erroneous write (“0” write). It is necessary to prevent this from occurring.

  The channel potential Vch of the memory cell MC12 after the channel boost is mainly the initial potential of the channel of the memory cell MC12 before the channel boost, the coupling ratio between the control gates of the memory cells MC12 and MC22, and the memory cells in the cell unit. (In this example, 2).

  Thus, for example, when the number of memory cells in the cell unit is fixed, the channel potential Vch of the memory cell MC12 is obtained by setting the initial channel potential of the memory cell MC12 and the coupling ratio between the control gates of the memory cells MC12 and MC22 and the channel. By increasing it, it can be raised sufficiently.

The control gate / channel coupling ratio B of the memory cell is calculated by the following equation.
B = Cox / (Cox + Cj)
Here, Cox is the total gate capacitance between the control gate and the channel of the memory cell, and Cj is the total junction capacitance of the source region and the drain region of the memory cell.

  Further, the channel capacity of the memory cell can be approximately represented by the sum of Cox and Cj. That is, the channel capacity of the memory cell includes, in addition to Cox and Cj, an overlap capacity between the control gate and the source region, a capacity between the bit line and the source area, and a capacity between the bit line and the drain region. However, these capacities are very small compared to Cox and Cj and can be ignored.

  Next, with regard to the byte type EEPROM of the present invention and the conventional NAND type flash EEPROM, the value of the channel potential (write inhibit potential) of the memory cell that performs “1” write will be specifically examined.

  For example, as shown in FIGS. 50 and 51, the byte type EEPROM of the present invention has a structure in which two memory cells are arranged in one cell unit.

In this case, the channel potential Vch is
Vch = Vini + (Vprog−VCC) × B
+ (Vpass-VCC) × B
B = Cox / (2 × Cox + 3 × Cj)
(Note: When memory cell is 2, diffusion layer (source / drain) is 3)
It becomes.

Here, when Cox = Cj = 1, the coupling ratio B is 0.2. In the present invention, Vpass = VCC. When the power supply potential VCC is 3 [V], the initial channel potential Vini is 2 [V], and the write potential Vprog is 16 [V], the channel potential Vch is
Vch = 2 + (16-3) × 0.2 = 4.6 [V]
It becomes.

On the other hand, the channel potential Vch of the NAND flash EEPROM is
Vch = Vini + (15/16) × (Vpass−VCC) × B
+ (1/16) × (Vprog−VCC) × B
B = 16 × Cox / (16 × Cox + 17 × Cj)
(Note: When memory cell is 16, diffusion layer (source / drain) is 17)
It becomes.

  Here, the cell unit of the NAND flash EEPROM is composed of 16 memory cells connected in series, and Vprog is applied to one memory cell and Vpass is applied to the remaining 15 memory cells.

When Cox = Cj = 1, the coupling ratio B is 0.48. When the power supply potential VCC is 3 [V], the initial channel potential Vini is 2 [V], the write potential Vprog is 16 [V], and Vpass is 8 [V], the channel potential Vch is
Vch = 2 + (15/16) × (8-3) × 0.48
+ (1/16) × (16-3) × 0.48
= 4.64 [V]
It becomes.

  As described above, in the byte type EEPROM of the present invention, for example, even when Vpass is set to the power supply potential VCC (or Vread), the number of memory cells in the cell unit is set to two, so that it is the same as the NAND type flash EEPROM. A write inhibit potential (channel potential of the “1” write cell) can be obtained.

  That is, in the present invention, by setting Vpass to VCC (or Vread), the voltage between the control gate and the channel of the non-selected memory cell connected to the non-selected control gate line can be relaxed. Data rewriting in units of bytes (or pages) can be repeatedly performed without erroneous writing.

  In the present invention, even if Vpass is set to VCC (or Vread), the same write inhibit potential as that of the NAND flash EEPROM can be obtained. Therefore, the write inhibit cell (“1” write cell connected to the selected control gate line). ) Can also be prevented.

・ Read operation
As shown in FIG. 54, after the bit line BL is charged to the precharge potential, 0 [V] is applied to the selected control gate line CGL11 in the selected block, and the non-selected control gate line CGL12 in the selected block and A power supply potential VCC (for example, 3.3 V) or a read potential Vread (for example, 4.5 V) is applied to the select gate lines SSL1 and GSL1, respectively.
Further, 0 [V] is applied to the control gate lines CGL21 and CGL22 and the select gate lines SSL2 and GSL2 in the non-selected block.

  At this time, the select transistors in the selected block are turned on, and the select transistors in the non-selected block are turned off. In addition, unselected memory cells in the selected block are turned on regardless of the data value (see FIG. 6 for the threshold distribution of the memory cells).

  The selected memory cell in the selected block is turned on or off depending on the data value.

  As shown in FIG. 55, when “1” data is written in the selected memory cell, that is, when the selected memory cell is in the erased state, the threshold voltage of the selected memory cell becomes the depletion mode. Yes. Therefore, a cell current flows through the selected memory cell, and the potential of the bit line BL is lowered.

  On the contrary, when “0” data is written in the selected memory cell, the threshold voltage of the selected memory cell is in the positive enhancement mode. Therefore, no cell current flows through the selected memory cell, and the potential of the bit line BL is maintained at the precharge potential.

  As described above, the determination of data “0” and “1” is made based on whether or not a cell current flows from the bit line to the source line. The change in the potential of the bit line is amplified (detected) by the sense amplifier.

The distinction between data “0” and “1” is made, for example, based on whether or not negative charges are stored in the floating gate of the memory cell.
That is, when negative charges are stored in the floating gate, the threshold voltage of the memory cell becomes high, and the memory cell becomes an enhancement type. On the other hand, when negative charge is not stored in the floating gate, the threshold voltage of the memory cell is less than 0 [V], and the memory cell becomes a depletion type.

Table 6 shows the potentials of the select gate lines SSL and GSL, the control gate line (word line) CGL, the bit line BLi, the cell source line SL, and the cell P well in each of the above-described erase, write, and read operations. .

  In the erase operation, the selected control gate line CGL in the selected block is set to 0 [V], and the unselected control gate line CGL in the selected block, the control gate line CGL in the unselected block, and all the selected gate lines. SSL and GSL are set in a floating state.

  In this state, when an erase potential Vera, for example, 21 [V] is applied to the cell P well, the potentials of all the select gate lines SSL and GSL in the floating state and the potentials of the unselected control gate lines CGL are Due to the capacitive coupling with the well, it rises to Vera × β (β is a coupling ratio).

  Here, when β is 0.8, the potentials of all the select gate lines SSL and GSL in the floating state and the potentials of the non-selected control gate lines CGL rise to 16.8 [V].

During the erase operation, the pn junction composed of the N ⁺ diffusion layer connected to the bit line BLi and the cell source line SL and the cell P well is biased in the forward direction. Therefore, the bit line BLi and the cell source line SL are charged to Vera-Vb. Vb is a built-in potential of a pn junction.

  In the write operation, the bit line BLi connected to the selected memory cell into which “1” data is written, that is, the bit line BLi connected to the selected memory cell maintaining the erased state is supplied with the power supply potential (for example, 3.3 [ V]) The bit line BLi that is set to VCC and connected to the selected memory cell to which the “0” data is written is set to 0 [V].

  The select gate line SSL on the bit line side in the selected block is set to the power supply potential VCC, the select gate line GSL on the cell source line side is set to 0 [V], and the unselected control gate line CGL is set to VCC or Vread (for example, 4.5 [V]) is set, and the selection control gate line CGL is set to a write potential (for example, 18 [V]) Vprog.

  The select gate lines SSL and GSL, the control gate line CGL, and the cell P well in the non-selected block are set to 0 [V].

  The cell source line is set to 0 [V]. However, if the channel potential of the memory cell in which “1” data is written in the selected block is boosted by capacitive coupling with the control gate line CGL, and the leak current of the cell source line becomes a problem due to punch through, the cell The potential of the source line is preferably set to the power supply potential VCC.

  In the read operation, the select gate lines SSL and GSL and the non-selected control gate line CGL in the selected block are set to the power supply potential VCC (for example, 3.3 V) or the read potential Vread (for example, 4.5 V) and selected. The control gate line CGL is set to 0 [V]. In the case of precharging the bit line before reading data, the bit line BLi is set to a precharge potential (eg, 1.2 [V]) VBL.

  Since the selected memory cell storing “1” data is turned on, a cell current flows through the selected memory cell, and the bit line BLi is discharged to 0 [V]. On the other hand, since the selected memory cell in which “0” data is stored is turned off, no cell current flows through the selected memory cell, and the bit line BLi holds the precharge potential VBL.

  FIG. 56 shows the main part of the circuit block of the byte type EEPROM of the present invention.

  As described above, the EEPROM includes, for example, a memory cell array 11 in which four memory cell units each having two memory cells sandwiched between two select transistors are arranged in a matrix, and a plurality of memory cell units 11 in the row direction on the memory cell array 11. A plurality of control gate lines 10 a and a plurality of bit lines 10 b arranged in the column direction on the memory cell array 11 are provided.

  The row decoder 12 selects a row, that is, the control gate line 10a. Data of the memory cell connected to the selected control gate line 10a is input to a sense amplifier circuit 13 including a sense amplifier having a data latch function provided for each column. The column decoder 14 selects a column, that is, the bit line BLi.

  The data of the sense amplifier in the selected column is output to the outside of the memory chip via the data input / output buffer 18. Data input into the memory chip is latched by a sense amplifier having a latch function for the selected column via the data input / output buffer 18.

  The booster circuit 16 generates a high voltage necessary for a write operation and an erase operation. The control circuit 17 controls the operation of each circuit inside the memory chip and serves as an interface between the inside and the outside of the memory chip. The control circuit 17 includes sequence control means (for example, a programmable logic array) that controls erase, write, and read operations on the memory cells.

  FIG. 57 shows a configuration of the memory cell array 11 of FIG.

  In this example, the memory cell unit is composed of a NAND cell composed of two memory cells connected in series, and two select transistors connected to each of both ends thereof. The memory cell is composed of a MOSFET having a so-called stack structure in which a floating gate and a control gate are stacked.

  One block is constituted by a plurality of memory cell units in the row direction, and one page is constituted by a plurality of memory cells connected to one control gate line CGL.

  In the present invention, erasing, writing, and reading can be performed in units of pages. Further, in the present invention, data rewriting in units of bytes is possible by adopting a rewriting method described later.

  58 shows a sense amplifier having a latch function connected to one bit line BLi in the sense amplifier circuit 13 of FIG.

  The sense amplifier mainly includes a latch circuit 21 including two CMOS inverters I1 and I2 in which one output is the other input. The latch node Q of the latch circuit 21 is connected to the I / O line via the column selecting NMOS transistor M8. The latch node Q is connected to the bit line BLi via the sense amplifier cutoff NMOS transistor M4 and the bit line potential clamping NMOS transistor M1.

  A connection node of the NMOS transistors M1 and M4 becomes a sense node Nsense. A precharge PMOS transistor M2 and a discharge NMOS transistor M3 are connected to the sense node Nsense. The precharge PMOS transistor M2 charges the sense node Nsense for a predetermined period based on the precharge control signal Load. The discharge NMOS transistor M3 discharges the charge of the sense node Nsense based on the discharge control signal DCB.

  A reset NMOS transistor M5 for forcibly grounding the latch node Qb based on the control signal φL1 is connected to the latch node Qb of the latch circuit 21. A reset NMOS transistor M6 for forcibly grounding the latch node Q based on the control signal φL2 is connected to the latch node Q of the latch circuit 21.

  The common source of the reset NMOS transistors M5 and M6 is connected to the ground via the sense NMOS transistor M7 controlled by the potential of the sense node Nsense. The sense NMOS transistor M7 is used for resetting the latch circuit 21 together with the NMOS transistors M5 and M6.

  FIG. 59 is a flowchart showing a schematic sequence control of the rewrite operation in byte units of the byte type EEPROM of the present invention.

  This sequence control is performed by, for example, the control circuit 17 in FIG. The rewriting operation in units of bytes will be briefly described in accordance with this flowchart as follows.

  In the byte-unit data rewrite mode, first, data for one page of a memory cell connected to the selected control gate line (word line) is read to the sense amplifier circuit (page reverse reading). Then, the data for one page is latched in the sense amplifier circuit (step ST1).

  Next, byte data corresponding to the column specified by the address is loaded. The loaded byte data is overwritten on the byte data to be rewritten out of the data for one page latched in the sense amplifier circuit (step ST2).

  Next, the data for one page of the memory cell connected to the selected control gate line is simultaneously erased (page erase) (step ST3). After erasure, erase verify is performed on each memory cell connected to the selected control gate line to verify whether erasure is completely performed or not erased (steps ST4 and ST5). ).

  Then, page erase and erase verify are repeatedly performed until the threshold values of all memory cells for one page are within the predetermined range, and the threshold values of all memory cells for one page are within the predetermined range (erase completion). If so, the operation proceeds to the next operation (steps ST3 to ST5).

  When only one sense amplifier circuit having a latch function exists for one bit line (when there is only one page), the data of the sense amplifier circuit may be destroyed depending on the result of erase verification. There is sex. Therefore, in such a case, the erasure is completed once without performing the erase verify.

  Thereafter, the data for one page latched in the sense amplifier circuit is simultaneously written into the memory cells connected to the selected control gate line (step ST6). After the writing, a write verify is performed to verify whether the writing is completely performed or the writing is not performed on each memory cell connected to the selected control gate line (steps ST7 and ST8). ).

  Then, page write and write verify are repeatedly performed until the threshold values of all memory cells for one page are within the predetermined range, and the threshold values of all memory cells for one page are within the predetermined range (write complete). If so, the data rewrite operation in byte units is terminated.

  Note that when a high write potential is used and one write is performed with one write pulse, the write verify can be omitted.

  FIG. 60 shows the state of the data of the selected memory cell and the node Qb (FIG. 58) of the sense amplifier circuit in the main steps of FIG.

  FIG. 5A shows a state where data for one page of a memory cell connected to the selected control gate line (word line) is read by the sense amplifier circuit (corresponding to step ST1).

  When the data in the memory cell is “0” (threshold voltage is positive), the charge of the bit line BLi is not discharged and the precharge potential is maintained. Therefore, the sense node Nsense in FIG. 58 becomes the power supply potential VCC. When the control signal φL2 is the power supply potential VCC, the node Q becomes the ground potential VSS, that is, “0”.

  Conversely, when the data in the memory cell is “1” (threshold voltage is negative), the charge on the bit line BLi is discharged. Therefore, the sense node Nsense in FIG. 58 becomes the ground potential VSS. When the control signal φL2 is set to the power supply potential VCC, the node Q becomes the power supply potential VCC, that is, “1”.

  FIG. 4B shows a state in which data is overwritten on byte data (8-bit data) designated by an address among data for one page latched in the sense amplifier circuit ( Corresponding to step ST2).

  FIG. 6C shows a state where data of a memory cell connected to the selected control gate line (word line) is erased (page erase) (corresponding to step ST3). By page erasing, all data in the memory cells connected to the selected control gate line becomes “1”.

  FIG. 4D shows a state in which one page of data latched in the sense amplifier circuit is written (page writing) to the memory cell connected to the selected control gate line (word line). (Corresponding to step ST6).

  As described above, in the operation of the memory cell array 11, data is rewritten in units of pages, but actually, data in units of bytes is rewritten.

Next, a read operation for page write and write verify will be described in detail with a focus on the operation of the sense amplifier circuit of FIG. 58 with reference to the timing charts of FIGS.
61 to 63 are obtained by dividing one timing chart into a plurality of parts.

  When a command instructing writing is input from the outside of the chip to the inside of the chip, the writing operation is started.

  First, in order to reset the sense node Nsense, the control signal DCB is set to the power supply potential VCC. At this time, the MOS transistor M3 is turned on, and the sense node Nsense is grounded (t1).

  When the control signal BLSHF is also set to the power supply potential VCC together with the control signal DCB, the MOS transistor M1 is turned on and the bit line BLi is grounded.

  Before loading the write data into the sense amplifier circuit, the data latch control signal φL1 is set to the power supply potential VCC, and the precharge control signal Load is set to the ground potential VSS. At this time, the MOS transistors M5 and M7 are turned on, the latch node Qb of the latch circuit 21 is forcibly grounded, and the data is reset. That is, in all the sense amplifiers of the sense amplifier circuit 20, the latch node Q of the latch circuit 21 becomes the power supply potential VCC and the latch node Qb becomes the ground potential VSS (t2).

  Next, write data is loaded from the I / O line, the data is latched in each latch circuit 21 of the sense amplifier circuit 20, and the nodes Q and Qb are set to "H" and "L" according to the load data. (T3).

  Specifically, in the latch circuit 21 of the sense amplifier corresponding to the memory cell to which “0” is written, “L” (= VSS) is given to the latch node Q, and “1” is written (write prohibited) memory cell. In the latch circuit 21 of the sense amplifier corresponding to “H”, “H” (= VCC) is applied to the latch node Q.

  Next, the control signals BLSHF and SBL become “H”, and charging of each bit line is started based on the data latched in each latch circuit 21 of the sense amplifier circuit 20 (t4).

  That is, the bit line BLi connected to the memory cell for performing “0” write is set to the ground potential VSS, and the bit line connected to the memory cell for “1” write (write inhibit) is charged to the power supply potential VCC. . The selected control gate line (word line) is set to the write voltage Vprog (about 20 [V]). At this time, the non-selected control gate line (word line) is not supplied with Vpass (for example, 8 [V]), but is supplied to the power supply potential VCC (for example, 3.3 [V]) or the unselected memory cell at the time of reading. The read potential Vread (for example, 4.5 [V]) is set.

  By this operation, writing to one page of memory cells is performed.

  After the data writing is completed, a write verify for verifying whether the data writing is completed properly is started.

  First, reading for write verification is performed. This verify read operation is the same as a normal read operation.

  When the control signal DCB is set to the power supply potential VCC, the MOS transistor M3 is turned on and the sense node Nsense is forcibly grounded (t5).

  Subsequently, a reference potential Vref (about 0.5 [V]) is applied to the selected control gate line CGL, and the memory cell is not connected to the unselected control gate line CGL regardless of the data stored in the memory cell. A read potential Vread (for example, 4.5 [V]) for turning on the cell is applied. Further, the power supply potential VCC is applied to the select gate lines SSL and GSL. Thereby, verify reading is performed (t6).

  At the time of reading, a bit line precharge type sensing method, a current detection type sensing method, or the like can be used. In the bit line precharge type sensing method, after the bit line BLi is precharged and brought into a floating state, the potential of the bit line is maintained or lowered according to the data of the memory cell. The current detection type sensing method will be described in detail below.

  At time t6, the control signal BLSHF is clamped from the boosted potential VCC + α to the potential VCC-α, and reading is performed by the balance between the memory cell current flowing through the MOS transistor M1 and the current of the MOS transistor M2 that charges the sense node Nsense. When the potential of the bit line BLi rises to, for example, 0.9 [V], the MOS transistor M1 is cut off, and the sense node Nsense becomes the power supply potential VCC.

  After the sense node Nsense becomes “H” (= VCC), the latch control signal φL1 is set to the power supply potential VCC, and the MOS transistor M5 is turned on (t7). When the sense node Nsense is at the power supply potential VCC (in the case of a sense amplifier connected to a memory cell whose threshold is higher than the verify potential Vref), the MOS transistor M7 is turned on, the latch node Qb is at the ground potential VSS, and the latch node Q is at It becomes the power supply potential VCC.

  When the ground potential VSS is loaded to the latch node Q and data is normally written, the latch data of the latch circuit 21 is inverted. When the writing to the memory cell is insufficient, the sense node Nsense remains “L” (= VSS) in the verify reading, so that the data inversion of the latch circuit 21 does not occur and the latch node Q maintains VSS. In the sense amplifier connected to the write-inhibited memory cell, the latch node Q is at the power supply potential VCC, so there is no data inversion.

  When there is an insufficiently written memory cell, that is, when there is a sense amplifier in which data inversion of the latch circuit 21 does not occur, writing and verify reading are repeatedly performed. When the potentials of the latch nodes Q of all the sense amplifiers for one page become the power supply potential VCC, the writing is completed.

  Next, the data rewriting operation in units of bytes will be described in detail with a focus on the operation of the sense amplifier circuit of FIG. 58 with reference to the timing chart of FIG.

  When a command for instructing byte rewriting is input from the outside of the chip to the inside of the chip, the byte rewriting operation starts.

  First, a reverse read operation of already written data is started for one page of memory cells connected to the selected control gate line (word line).

  The reverse read operation is the same as the read operation.

  First, the data latch control signal φL1 is set to the power supply potential VCC, and the precharge control signal Load is set to the ground potential VSS. At this time, the MOS transistors M5 and M7 are turned on, the latch node Qb of the latch circuit 21 is forcibly grounded, and the data is reset. That is, the latch nodes Q of all the latch circuits 21 of the sense amplifier circuit are set to the power supply potential VCC, and the latch nodes Qb are set to the ground potential VSS (t1).

  Next, the control signal DCB is set to the power supply potential VCC. At this time, the MOS transistor M3 is turned on, and the sense node Nsense is forcibly grounded (t2). Subsequently, when VSS (= 0 V) is applied to the selected control gate line CGL and the power supply potential VCC is applied to the select gate lines SSL and GSL, reading is performed (t3).

  After the sense node Nsense becomes “H” (= VCC), the latch control signal φL2 becomes the power supply potential VCC, and the MOS transistor M6 is turned on (t4). When the sense node Nsense is at the power supply potential VCC (that is, when the sense amplifier is connected to a memory cell in which data “0” is written and the threshold voltage is higher than VSS), the MOS transistor M7 is turned on and the latch node Q Is at ground potential VSS, and latch node Qb is at power supply potential VCC.

  Next, the control signal DCB is set to the power supply potential VCC, the control signal BLSHF is set to the power supply potential VCC or the potential VCC + α, and the bit line BLi and the sense node Nsense are reset (t5).

  Thereafter, byte data is loaded into the latch circuit 21 of the sense amplifier circuit 20 designated by the column address, and the nodes Q and Qb are set to “H” and “L” according to the byte data (t6).

  Byte data input from the outside of the chip is overwritten on predetermined data of the page data written in the latch circuit 21.

  Thereafter, a page erase operation is performed on the memory cell connected to the selected control gate line.

  The control gate line of the selected block is set to the ground potential VSS, and the control gate line and all the select gate lines of the non-selected block are set to a floating state. When the erase voltage Vera is applied to the cell P well, the select gate line in the floating state and the control gate line of the non-selected block are set to Vera × β (β is a coupling ratio) due to capacitive coupling with the cell P well. Boosted.

The bit line BLi and the cell source line SL are connected to the N ⁺ layer in the cell P well. When the pn junction between the N ⁺ layer and the cell P well is forward-biased, the bit line BLi and the cell source line SL are charged to Vera-Vb, respectively (t7). However, Vb is a built-in potential of a pn junction.

  Thereafter, erase verify is performed to confirm that all the memory cells of the selected page are in an erased state, that is, the threshold voltage of the memory cell has become negative. Based on the data stored in the latch circuit 21, a write operation and a write verify operation are performed on the memory cells of the selected page.

  In FIG. 63, operations after the erase verify are omitted.

  FIG. 64 shows an example in which a part of a NAND flash EEPROM memory cell array is used as a byte EEPROM memory cell array of the present invention.

  The byte EEPROM memory cell array according to the present invention can be considered as two memory cells between two select transistors in the NAND flash EEPROM memory cell array. Therefore, the EEPROM as in this example can be easily realized.

  In the EEPROM of this example, two types of memory cell units having different configurations are connected to one bit line BLi. That is, in the first memory cell unit, a plurality of memory cells (for example, 8, 16, 32, etc.) are connected between two select transistors, and the second memory cell unit has two select transistors. Two memory cells are connected between them.

  In selecting the control gate line (word line), a drive circuit may be provided separately in the first memory cell unit region and the second memory cell unit region, and if it can be shared, The drive circuits in both regions may be combined into one.

  With such a configuration, it is possible to rewrite data in units of bytes for a part of the memory cell array.

  Note that, instead of the NAND flash EPROM memory cell array of FIG. 64, a memory cell array such as an AND flash EEPROM or a DINOR flash EEPROM may be employed.

  As described above, according to the byte type EEPROM of the present invention, (1) since the memory cell unit is composed of one stack type memory cell sandwiched between two select transistors, the same process as the flash EEPROM is performed. In addition, the same rewriting method as that of the flash EEPROM can be adopted, and data rewriting in byte units can be made possible.

  In addition, (2) VCC or Vread instead of Vpass is applied to the control gate of the non-selected memory cell at the time of writing, and a plurality of (for example, two) stacked types in which the memory cell unit is sandwiched between two select transistors If the memory cell is used, in addition to the above effects, the size per memory cell can be further reduced.

Table 7 shows a comparison of the effects of a normal NAND flash EEPROM and the byte EEPROM of the present invention.

  Further, according to the byte type EEPROM of the present invention, (3) the memory cell array is composed of a plurality of blocks arranged in a matrix, and operations such as reading, erasing and writing can be performed in units of blocks. Even in unit data rewriting, substantial data rewriting characteristics are not deteriorated.

The figure which shows the memory cell unit of the byte type EEPROM of this invention. The figure which shows the equivalent circuit of FIG. The figure which shows the memory cell array of the byte type EEPROM of this invention. The figure which shows the relationship between the gate voltage according to the data of a memory cell, and a cell current. The figure which shows the electric potential given to a memory cell unit at the time of data reading. The figure which shows an example of the threshold distribution according to the data of a memory cell. The figure which shows the other example of the threshold value distribution according to the data of a memory cell. The block diagram which shows the principal part of the byte type EEPROM of this invention. FIG. 9 is a diagram showing an example of the configuration of the sense amplifier circuit of FIG. 8. The flowchart which shows the rewriting operation | movement of the byte unit of this invention. The figure which shows the mode at the time of reverse reading of the memory cell data of the sequence of FIG. The figure which shows the state at the time of the overwrite of the byte data of the sequence of FIG. The figure which shows the mode at the time of page deletion of the sequence of FIG. The figure which shows the mode at the time of page writing of the sequence of FIG. The wave form diagram which shows the data rewriting operation | movement of the page unit of this invention. The wave form diagram which shows the data rewriting operation | movement of the page unit of this invention. The wave form diagram which shows the data rewrite operation | movement of the byte unit of this invention. The figure which shows the modification of the memory cell array of the byte-type EEPROM of this invention. The figure which shows the modification of the memory cell array of the byte-type EEPROM of this invention. The figure which shows the mode at the time of the write-in operation | movement of a stack gate type memory cell. The figure which shows the mode at the time of the erase operation of a stack gate type memory cell. The figure which shows the modification of the memory cell array of the byte-type EEPROM of this invention. The figure which shows an example of the byte type EEPROM of this invention. FIG. 24 is a diagram showing the memory cell array of FIG. 23. The figure which shows the other example of the byte-type EEPROM of this invention. FIG. 26 is a diagram showing the memory cell array of FIG. 25. The figure which shows the data rewrite operation | movement of the byte unit of this invention. The figure which shows the modification of EEPROM of FIG. The figure which shows an example of the predecoder of FIG. FIG. 29 is a diagram showing an example of the row decoder and driver of FIG. 28. FIG. 29 is a diagram showing one row of the memory cell array in FIG. 28. The figure which shows the data rewrite operation | movement of the byte unit of this invention. The figure which shows the data rewrite operation | movement of the byte unit of this invention. The figure which shows an example of arrangement | positioning of the well in a memory cell array area | region. FIG. 32 is a diagram showing a modification of the memory cell array in FIG. 31. FIG. 29 is a diagram showing a modification of the EEPROM in FIG. 28. FIG. 37 is a diagram showing two rows adjacent to each other in the memory cell array of FIG. 36. The figure which shows the example of the system using a differential type sense amplifier. The figure which shows the example of the system which provided one sense amplifier in the several bit line. FIG. 29 is a diagram showing a modification of the EEPROM in FIG. 28. FIG. 41 is a diagram illustrating an example of the predecoder of FIG. 40. FIG. 41 is a diagram illustrating an example of the row decoder and driver of FIG. 40. FIG. 41 is a diagram showing two rows adjacent to each other in the memory cell array of FIG. 40. The figure which shows the example of arrangement | positioning of a subdecoder. The figure which shows an example of EEPROM to which this invention is applied. The figure which shows an example of EEPROM to which this invention is applied. The figure which shows an example of EEPROM to which this invention is applied. The figure which shows about the disturbance at the time of writing of NAND type EEPROM. The wave form diagram which shows the data write-in operation | movement of NAND type EEPROM. The figure which shows the memory cell unit of the byte type EEPROM of this invention. The figure which shows the equivalent circuit of FIG. FIG. 5 is a diagram showing potentials applied to a memory cell unit during an erasing operation. FIG. 10 shows potentials applied to a memory cell unit during a write operation. FIG. 10 shows potentials applied to a memory cell unit during a read operation. The figure which shows the relationship between the gate voltage according to the data of a memory cell, and a cell current. The block diagram which shows the principal part of the byte type EEPROM of this invention. 57 is a diagram showing a circuit configuration of the memory cell array in FIG. 56. FIG. FIG. 57 is a diagram showing an example of a configuration of a sense amplifier circuit of FIG. 56. The flowchart which shows the rewriting operation | movement of the byte unit of this invention. The figure which shows the mode of the node Qb of the sense amplifier at the time of byte rewriting. The wave form diagram which shows the data rewriting operation | movement of the page unit of this invention. The wave form diagram which shows the data rewriting operation | movement of the page unit of this invention. The wave form diagram which shows the data rewrite operation | movement of the byte unit of this invention. The figure which shows the modification of the memory cell array of the byte-type EEPROM of this invention. The figure which shows the memory cell of the conventional byte type EEPROM. FIG. 66 is a sectional view taken along line LXVI-LXVI in FIG. 65. The energy band figure which shows the mechanism of FN tunnel current. The figure which shows the memory cell of the conventional byte type EEPROM. The figure which shows the basic structure of the memory cell of the conventional byte type EEPROM. The figure which shows the NAND unit of NAND type flash EEPROM. The figure which shows the equivalent circuit of FIG. The figure which shows the memory cell of NOR type flash EEPROM.

Explanation of symbols

11: memory cell array,
11-0: 3 trussels,
11-1: NAND cell part,
12, 12b: row decoder,
12a: predecoder,
12c: control gate / select gate driver,
13: Sense amplifier circuit 14: Column decoder
15: Column gate (switch),
16: Booster circuit,
17: control circuit,
18: Data input / output buffer,
20: Sense amplifier,
21: Latch circuit,
25: Command register,
26: Command decoder,
27: signal generation circuit,
28: Sub-control gate driver,
29: Sub-decoder,
30-1, ... 30-3, 32: NAND circuit,
31-1, ... 31-3, 33: inverter,
34: Booster circuit,
35-1, ... 35-3, 36-0, ... 36-3, 40-0, ... 40-3: N-channel MOS transistors,
37: Main control gate driver,
38: Select gate driver,
39: decoding circuit,
41: Semiconductor chip,
42a, 42b: memory circuit,
M1 to M8: MISFET,
I1, I2: inverter,
MC: memory cell,
ST1, ST2: Select transistor,
BC: Bit line contact portion,
SL: Source line
CGL: Control gate line (word line),
SSL, GSL: Select gate line,
BLi: Bit line.

Claims (6)

A memory cell array having a first memory cell unit composed of one memory cell and one select transistor;
A bit line directly connected to the memory cell;
A sense amplifier having a latch function connected to the bit line;
The non-volatile semiconductor memory according to claim 1, wherein the memory cell has a stack gate structure having a floating gate and a control gate, and data writing / erasing with respect to the memory cell is performed using an FN tunnel phenomenon.   The memory cell array includes a second memory cell unit including a plurality of memory cells, and both the first and second memory cell units are connected to the bit line. 2. The non-volatile semiconductor memory according to 1.   The non-volatile semiconductor memory according to claim 2, wherein the second memory cell unit is one of a NAND unit, an AND unit, and a DINOR unit.   The nonvolatile semiconductor memory according to claim 1, wherein the select transistor has the same structure as the memory cell. The nonvolatile semiconductor memory according to claim 1,
Data for one page in the memory cell array is read to a sense amplifier for one page, and data of less than one page to be rewritten out of the data for one page in the sense amplifier for one page is stored. Nonvolatile semiconductor, further comprising means for overwriting and erasing one page of data in the memory cell array and then writing the data of the sense amplifier for one page after the overwriting into the memory cell array memory.   6. The nonvolatile semiconductor memory according to claim 5, wherein the data of less than one page is byte data.

Images