Patent Application
20060285387
1. Field of the Invention
The present invention relates to the non-volatile memory field. More specifically, the present invention relates to the erasing of a flash memory device With NAND architecture.
2. Description of the Related Art
The flash memory devices are non-volatile memories wherein each single cell may be programmed electrically, but a large number of cells, forming a sector, are erased at the same time. Typically, the cells consist of floating gate MOS transistors which store a logic value defined by their threshold voltage (which depends on the electric charge stored on the floating gate). Particularly, in a flash memory with NAND architecture, the cells are grouped in strings, each one consisting of a set of cells that are connected in series. The main advantage of such architecture is the reduced area occupation, essentially due to both the reduction of the contacts number and the reduced size of each cell. This makes the NAND memories particularly advantageous in a number of applications such as memory cards, memories of digital video-cameras and of audio recorders.
In order to maintain reduced sizes, the NAND memories implement a decoding system that is able to apply positive voltages only (i.e., greater or equal to zero) to the various cells selectively. Moreover, the cells are generally erased by applying a single blind erasing pulse (which reduces the threshold voltages of the cells below a reference reading voltage).
A problem of the NAND memories is due to the capacitive coupling between the floating gates of adjacent cells; such effect makes the threshold voltage of a cell dependent non-uniquely on the electric charge stored in its floating gate but even on the electric charges stored in the floating gates of the adjacent cells. The problem is particularly acute in the NAND memories because of their high integration. Such effect modifies the threshold voltages of the cells whenever the adjacent cells are programmed. The suffered variation increases with the amplitude of the gap of the threshold voltages of the adjacent cells (for example, for a gap of 2.5V the variation of the threshold voltage is equal to about 130 mV, while for a gap of 7V the variation of the threshold voltage is equal to about 360 mV). Thus, the problem is particularly acute when the adjacent cells that are programmed start from a very low original threshold voltage. Such effect is emphasized by the single pulse erasing procedure; in fact, the erasing pulse has to be dimensioned in order to guarantee the erasing in the worst conditions, so that it normally brings the threshold voltages of the erased cells to very low values. The above described variation of the threshold voltages can cause reading errors. Such effect is experienced in the standard NAND memories but it is more and more limiting in the multi-level NAND memories wherein each cell stores multiple bits (since the margins which are used for discriminating the different stored logic values are reduced).
In order to limit the capacitive coupling effect, a sequential programming method of the pages into which the memory is logically divided is often adopted; in such case, the pages have to be programmed in succession according to their physical order. This operative way removes the capacitive coupling effect due to the preceding pages. In fact, the preceding pages cannot change the threshold voltages of the cells of the current page because they are not modified any longer after their programming. However, such technique does not remove the capacitive coupling effects of the other cells which are programmed successively.
The patent U.S. Pat. No. 6,661,711 B2 discloses a method for shifting the distribution of the threshold voltages of the erased cells. In such case, the cells are programmed until their threshold voltage does not exceed a predetermined value. Such operation is indiscriminately performed over all the cells, with the exception of some cells (in the example at issue, eight), which the threshold voltage exceeds an acceptable maximum value. In such a way, a shifting of the distribution of the threshold voltages is obtained. Nevertheless, the proposed solution substantially maintains the same shape of the distribution, which generally has a rather long tail due to the cells having a threshold voltage more negative.
One embodiment of the present invention proposes a solution that is based on the idea of compacting the distribution of the threshold voltages of the erased cells.
One embodiment of the present invention proposes a flash memory device with NAND architecture. The memory devices include a matrix of memory cells (each one having a programmable threshold voltage). The matrix includes one or more sectors which are individually erasable. The matrix is arranged in a plurality of rows and columns. The cells of each row are connected to a corresponding word line and the cells of each column are arranged in a plurality of strings of cells connected in series; the strings of each column are connected to a corresponding bit line. The memory device further includes means for erasing the cells of a selected sector; means is further provided for restoring the threshold voltage of the erased cells. The means for restoring act in succession on each of a plurality of blocks of the sector; the block includes, for each one of a set (one or more) of selected bit lines, a group of cells connected to a set (one or more) of selected word lines. The means for restoring includes means for reading each group with respect to a limit value (for example, −3V) exceeding a reading reference value (for example, 0V). Means is further provided for programming only each group wherein the threshold voltage of at least one cell of the group does not reach said limit value. Further means is used for stopping the restoring in response to the reaching of the limit value by at least one sub-set of the groups.
In a preferred embodiment of the invention, the reading is performed by applying a negative voltage to a common body region of the cells of the sector (so as to obtain a rising of their apparent threshold voltage equal to the limit value).
In a specific implementation, the reading is performed at the single cell level (with the restoring that is stopped in response to the reaching of the limit value by all the cells of the block).
Alternatively, the same operation is performed at the string level (with the restoring that is stopped in response to the reaching of the limit value by a predetermined number of strings, for example, only one).
Advantageously, the cells of the sector are programmed before the erasing.
A still further embodiment of the present invention proposes erasing the cells of the sector in a controlled way.
As a further improvement, the cells are verified after the restoring (and a reduced erasing pulse is applied when it is necessary).
In a preferred embodiment of the invention, the verify of the erasing is performed by applying a negative voltage to the body region as well (so as to obtain a rising of the threshold voltage of the cells equal to a reference reading value with a margin).
A way to improve the solution is by enabling the means for restoring in succession with two or more limit values (closer and closer to the reading reference value).
A further embodiment of the present invention proposes a corresponding erasing method.
The invention itself as well as further features and the advantages thereof will be best understood by reference to the following detailed description, given purely by way of a non-restrictive indication, to be read in conjunction with the accompanying drawings. In this respect, it is expressly intended that such figures are not necessarily in scale and that, unless otherwise indicated, they are intended simply to disclose the described structures and procedures conceptually. In particular:
a-4c detail several phases of the erasing procedure;
a-5d are an exemplary representation of several distributions of the threshold voltage of the erased cells.
With reference in particular to
In an erased condition each cell 110 has a low threshold voltage (to which a logic level 1 is typically associated). The cell 110 is programmed by injecting electrons into its floating-gate; in this condition the cell 110 has a high threshold voltage (to which a logic level 0 is typically associated). Therefore, when the cell 110 is biased for reading, it is conductive if erased or it is non-conductive if programmed.
The cells 110 are arranged in a plurality of rows and columns. The memory device 100 has NAND type architecture. In this configuration, each sector 115 includes a plurality of strings 125 that are connected in parallel; each string 125 is formed by a set of cells 110 (typically 16) that are connected in series between two select transistors 120s and 120d. In detail, an intermediate cell 110 of the string 125 has a drain terminal connected to the source terminal of an upper cell 110 and a source terminal connected to the drain terminal of a lower cell 110. The first cell (lowest) 110 of the string 125 has a source terminal connected to the drain terminal of the select transistor 120s, while the last (uppermost) cell 110 of the string 125 has a drain terminal connected to the source terminal of the select transistor 120d. Moreover, the select transistor 120s has a source terminal connected to the source terminal of the select transistor 120s of a lower string and the transistor 120d has a drain terminal connected to the drain terminal of the select transistor 120d of an upper string. The gate terminals of the cells 110 of each row are connected to a corresponding word line WL. The drain terminals of the select transistors 120d of the strings of each column are connected to a corresponding bit line BL. The gate terminals of the select transistors 120d positioned on a same row are connected to a corresponding drain select line DSL; likewise, the gate terminals of the select transistors 120s positioned on a same row are connected to a corresponding source select line SSL. Moreover, the source terminals of all the select transistors 120s are connected to a common source line SL, which is typically maintained at a reference voltage (or ground).
The cells 110 belonging to each row are logically grouped in two pages; in detail, a first page is formed by the cells 110 having an even position and a second page is formed by the cells 110 having an odd position. The memory device 100 receives an address ADR for selecting the desired page. A portion of the address ADR is supplied to a row decoder 130r that selects the word line WL of the desired page, and the drain select line DSL and the source select line SSL of the corresponding strings. Another portion of the address ADR is supplied to a column decoder 130c; the column decoder 130c connects the bit lines BL of the desired page, that is, the even ones or the odd ones, to a page buffer 140 (which is used for reading and for writing the cells 110 of the selected page).
The memory device 100 further includes a power management unit (PMU) 150 and a control unit 160. The PMU 150 supplies the voltages that are used for performing the various operations on the memory device 100, for example, from −20V to 20V (referred to as a whole with Vin); the voltages Vin are generated (for example, through charge pumps) starting from a power supply voltage Vcc supplied from the outside (typically, 3.3 or 5V). The control unit 160 provides the control signals (referred to as a whole with Sc) that are used for driving the various components of the memory device 100.
Moving now to
Each cell 110 is formed by an n-type drain diffusion 221 and by a n-type source diffusion 222, both realized in the body region 220 (of p-type). A floating gate 223 and a control gate 225 are stacked over a channel formed between the drain diffusion 221 and the source diffusion 222. The floating gate 223 is not accessible from the outside whereas the control gate 225 is connected to the corresponding word line WL. The select transistors 120s and 120d are likewise formed by a drain diffusion 221 and by a source diffusion 222. A control gate 228 is insulated from the channel (between the diffusions 221 and 222) by means of an oxide layer. The control gates 228 of the select transistors 120s and 120d are connected to the corresponding select source line SSL and to the corresponding drain select line DSL, respectively. As shown in the figure, the drain diffusion 221 and the source diffusion 222 of each cell 110 or select transistor 120s, 120d are in common with the source diffusion 222 and the drain diffusion 221 of an adjacent element 110, 120s, 120d, respectively.
With reference to
The block 310 is followed by the block 320, wherein all the cells of the sector are erased (as described in detail in the following with reference to
As previously indicated,
With reference to
The block is determined by the bit lines of a desired page; for each bit line the block includes a group of cells connected to one or more selected word lines. In particular, each group is formed by a single cell (performing a page level reading) or by an entire string (performing a string level reading).
In the first case, (page level reading) the select lines of the strings corresponding to the selected page are brought to the voltage Vcc. The word line of the selected page is brought to the voltage 0V whereas the other word lines of the corresponding strings are brought to a voltage Vpass (for example, 20 V) so that the respective cells are always conductive (irrespective of their threshold voltage). A (negative) voltage is applied to the body region, in order to produce an increasing equal to the value VLa of the apparent threshold voltage of the cells due to the body effect. Thus, by using the conventional reference reading value (0V), each cell of the selected page will be conductive if its (real) threshold voltage is lower then the limit value VLa and not conductive otherwise; this allows discriminating such cells on the basis of the read logic value (1 or 0, respectively). It should be noted that in such way the negative voltage (used for reading the cells with respect to the limit value VLa) is applied to the body region of the whole sector, so that it is not necessary any decoding of negative voltages (for driving the selected word lines). This allows continuing on using a decoding system for positive (or null) voltages only, with a reduced area occupation on the semiconductor material chip.
In the case of string level reading, the select lines of the selected strings are likewise brought to the voltage Vcc; however, all the word lines of the selected strings are now brought to the voltage 0V whereas the body region is brought to the negative voltage used for reading the cells with respect to the limit value VLa (by using a conventional reference reading value 0V). In such a way, for each bit line there is read the logic value 1 if the threshold voltage of all the cells of the corresponding string is lower than the limit value VLa (i.e., the cells are conductive) or the logic value 0 if the threshold voltage of at least one cell of the string is greater than the limit value VLa (i.e., such cell is not conductive).
Thus, the method verifies at the block 449a if the reading has a negative result, i.e., there is at least one group (cell or string) which has provided a logic value 1 (since the cell or all the cells of the string have a threshold voltage lower than the limit value VLa, respectively). In such case, the flow of activity moves to the block 451a, wherein a programming pulse is applied to the groups of cells (single cell or string) which do not satisfy the above-mentioned condition. In particular, the body region and the corresponding select and bit lines are brought to the voltage 0V. Thus, a voltage pulse is applied to the desired word lines; for example, such pulse has a length typically equal to a ten of μs and an initial value of 6V which is incremented by a predetermined amount (typically, 50 mV) during each iteration of the operation on the same block. Thus, the method returns to the block 446a for repeating the above described operations.
As soon as the reading of all the cells of the block (in the case of page level reading) or the reading of a predetermined number of strings of the block, for example one (in the case of string level reading) provides the logic value 0 (block 449a), the method exits the above described loop and it moves to the block 454a. During such phase there is verified whether all the blocks of the sector have been compacted. In the negative case, the flow of activity returns to the block 446a for repeating the same operations on a next block.
As soon as all the cells of the sector have been compacted, the method moves to the block 443b wherein a second limit value VLb equal to the actual desired value (i.e., −3V in the example at issue) is chosen. Thus, the method proceeds with a further compacting at the blocks 446b-454b having operative modes analogous to the preceding one (blocks 446a-454a). In such a way a fine compacting of the distribution of the threshold voltages of the erased cells is performed. It should be noted that in such case the variation of the threshold voltages has a small value (thanks to the preceding coarse compacting); thus, the capacitive coupling effect on the already compacted cells by the subsequent programming operations can be considered negligible.
As soon as all the cells of the sector have been compacted (block 454b), the method ends to the block 457.
The
Considering now the
Moving to the
Likewise, in
Notwithstanding, such choice offers a high processing speed (to the detriment of lower accuracy).
The reduction of the disturb, due to the capacitive coupling among the floating gates of adjacent cells, obtained by means of the above described solution is shown schematically in
When an (erased but not compacted) adjacent page is programmed, the distribution of the threshold voltages spreads (550) due to the high increment of the threshold voltage of the cells of the adjacent page having a starting value very negative. On the contrary, if the adjacent page has been compacted, such effect is much more reduced; thus, the distribution of the threshold voltages spreads less (curve 560).
Naturally, in order to satisfy local and specific requirements, a person skilled in the art may apply to the solution described above many modifications and alterations. Particularly, although the present invention has been described with a certain degree of particularity with reference to preferred embodiments thereof, it should be understood that various omissions, substitutions and changes in the form and details as well as other embodiments are possible; moreover, it is expressly intended that specific elements and/or method steps described in connection with any disclosed embodiment of the invention may be incorporated in any other embodiment as a general matter of design choice.
For example, similar considerations apply if the memory device has a different structure or includes equivalent components. Moreover, the numerical examples are merely indicative, and they have not to be considered in a limitative way; in particular, the reference reading value can be different from 0V as well as the limit value can be different from −3V/−3.5V. Moreover, it should be specified that the programming and/or erasing pulses that are applied to the cells of the sector need not necessarily to have length and width equal to the ones indicated in the preceding description. Likewise, the maximum tolerable number of iterations of each loop described in some phases of the proposed solution can also be different from a phase to the other and it can also results equal to some tens.
It is emphasized that, although for the sake of simplicity of the description reference has been made to a two-level memory, it is evident that the proposed solution is particularly advantageous in a multi-level memory.
Similar considerations apply if the n-type and p-type regions are reversed and consequently voltages of opposite sign are used; for example, it is not excluded the application of the proposed solution to a structure using a negative (i.e., lower or equal than zero) row decoding, whereas suitable biasing positive voltages are applied to the body region for reading the cells with respect to positive values. In any case, the proposed solution can also be applied with a row decoding that manages biasing voltages of both signs.
Moreover, nothing prevents each group to be formed by more cells and/or strings.
Likewise, any other criterion can be used for stopping the restoring; for example, it is not excluded that the restoring is interrupted in response to the reaching of the limit value by a predetermined number of cells of the page or by all the read strings.
In addition, it is possible that the method operates by omitting the initial programming phase.
With regards the erasing phase, alternatively to the erasing pulse proposed in the preceding description, it is not excluded to use of a single blind erasing pulse.
Moreover, an implementation without any erase verify after the compacting is contemplated. In any case, such verify can also be performed with respect to the effective reading reference value (without any margin); alternatively, it is possible to perform the erase verify phase at the level of the whole sector (by biasing all the word lines to the reference reading value).
In addition, it is possible that a general variant of the proposed solution provides three or more iterations of the compacting loop. However, it is not essential to introduce more restoring loops; in fact, in some cases only one loop can be sufficient for the proposed purposes.
All of the above U.S. patents, U.S. patent application publications, U.S. patent applications, foreign patents, foreign patent applications and non-patent publications referred to in this specification and/or listed in the Application Data Sheet, are incorporated herein by reference, in their entirety.
| Number | Date | Country | Kind |
|---|---|---|---|
| 05104742.1 | Jun 2005 | EP | regional |
