The present invention relates to semiconductor memory designs, and more particularly to improved word line driver designs for flash memory.
The memory matrix architecture most widely utilized in the construction of semiconductor integrated, non-volatile memory devices is the NOR type. In this type of architecture, memory cells belonging to one row have their gate terminals in common, while memory cells belonging to one column have their drain terminals in common The source terminals are, on the other hand, shared by all the cells of one sector. A portion of a NOR matrix, reprinted from U.S. Pat. No. 6,515,911, the entirety of which in incorporated by reference herein, is shown in
Each memory location is identified by a given row and given column and found at the intersection thereof. Each memory cell comprises a floating-gate transistor which has drain and source conduction terminals. As known in the art, the source, drain and gate terminals are biased accordingly to perform read, program and erase operations.
A prerequisite of non-volatile memories of the flash EEPROM type is that the information stored therein should be erased by groups or packages of bits. The erase operation is the only operation that involves biasing of the source terminal, and since all the cells have this terminal in common, they can be written into and read from in an independent manner but must be erased simultaneously.
Particularly with flash memories, the erase operation is performed by sectors, in the sense that all the cells that run to the same source line must be erased simultaneously. Within a non-volatile memory matrix, the sectors can be organized either into rows or columns. In a row type of organization, the size of a sector is given by the number of rows that it contains. The architecture of the storage device is designed to fit the number and size of the sectors in order to optimize the circuit area consumption, as well as the device performance and reliability.
A single bit line is not shared by all of the sectors because of a problem known as “drain stress.” Therefore, each sector is arranged to include a specific group of columns referred to as the “local bit lines.” Local bit lines are individually connected, via a pass transistor, to a main metallization connection referred to as the “main bit line.” Each sector is assigned a local group of pass transistors which are only turned “on” in the addressed sector, so that the cells of the other sectors need not be affected by drain stress.
Shown schematically in
Another prior art architecture shown in
In view of the shortcomings of these prior art architectures, the '911 patent proposes embodiments of hierarchical row decode. In one described embodiment, a circuit device is provided capable of carrying out a hierarchical form of row decoding for semiconductor memory devices of the non-volatile type having a matrix of memory cells with sectors organized into columns. Each sector has a specific group of local word lines individually connected to a main word line running through all of the sectors which have rows in common. The '911 patent describes a three transistor structure for carrying out the hierarchical row decoding.
The word line driver design for the row decoders of these NOR architectures has become increasingly more important since the periphery transistors of the driver design cannot shrink proportionately with reductions in the size of the cell dimensions, as the driver transistors must be able to sustain the bias conditions of legacy generations. Therefore, as cell sizes reduce, the word line driver occupies a greater amount of the overall circuit layout area. While there are advantages to hierarchical row decoding schemes such as those proposed in the '911 patent, these schemes consume valuable space. Therefore, improved word line driver designs are desired.
A non-volatile memory device is provided. The device includes a memory array having a plurality of memory cells organized as sectors, each sector having a main word line associated with a plurality of local word lines, each local word line coupled to the main word line by a respective local word line driver circuit, each of the local word line driver circuits consisting of a first MOS transistor coupled between the respective main word line and a respective local word line and a second MOS transistor coupled between the respective local word line and a first biasing terminal.
The above and other features of the present invention will be better understood from the following detailed description of the preferred embodiments of the invention that is provided in connection with the accompanying drawings.
The accompanying drawings illustrate preferred embodiments of the invention, as well as other information pertinent to the disclosure, in which:
As those in the art will recognize, word line driver circuits are used to boost selected word lines to a desired voltage. Word line driver circuits also provide final decoding of the row or X address of the selected core cell. Typically, each word line has an accompanying word line driver circuit. With improvements in device layout and in process technologies, the core cells in a memory array are laid out at increasingly finer pitches. As word lines are placed closer together, limitations are placed on the size of the word line driver circuits. The word line driver designs described herein provide for reduced word line driver size by limiting the number of operational components within the word line drivers. The exemplary biasing conditions proposed herein illustrate the operability of the designs.
For NMOS cell designs, the P-well bias is raised to perform erasing. The word line in the selected sector is biased to a negative voltage. Wth serial flash, when performing sector erase (as opposed to block erase), the de-selected sectors in a selected block suffer erase disturbance from the common P-well bias shared with the cells of the selected sector. The word line driver design is important for reducing this erase disturbance.
Turning to
The gate of transistor M1 receives control signal GMn[m]. The gate of transistor M2 receives control signal GNn[M].
In embodiments, during programming, the de-selected local word line is biased at 0V. More preferably, the de-selected word line can be biased with a negative voltage, such as −0.5 V or −1V, to reduce possible leakage current of de-selected cells with the shared bit line, i.e., LWL0[1:255] if LWL0[0] is selected for programming.
Exemplary operating conditions for this word line driver 10 when used with parallel flash for sector-by-sector erase, where the sectors do not share a common P-well (See
The following table shows operating conditions for both parallel flash (
The two foregoing tables show two alternative programming conditions, designated Program-1 and Program-2. Soft program conditions are also shown. During programming of local word line LWL0[0], the de-selected word lines (LWL0[1:255] and LWLn[0:255]) are biased at ground (0V) (Program-1 condition). In Program 2 conditions, local word line LWL0[1:255] of the selected sector 0 are biased at a negative voltage, such as −0.5V or −1.0V, to reduce the possible cell leakage current on the de-selected cell with the shared bit line. For GM0[0] and GN0[0], “8V.fwdarw.0V” means 8V is applied initially to pass −0.5V to all word lines controlled by GM0[0] and GN0[0]. Then, 0V is applied to pass 8V to the selected word line. These conditions assume the NMOS threshold voltage Vth>0.5V.
In the table above, the block erase of the serial flash is the same as the sector erase of the serial flash except for the biasing of the main word lines MWL0[1:15] associated with local word lines LWL0[16:255]. During sector erasing of the serial flash, the de-selected sector can have a positive voltage on its word lines, such as 2.5 V, to reduce erase disturbance. The P-well of the NOR cells will be biased between 6V and 8V generally. The positive word line bias reduces the voltage drop from the P-well to the de-selected word lines, thereby reducing erase disturbance.
Those in the art will understand that soft programming is used after erasing to correct the over-erased cells, i.e., cells where the threshold voltage is too low, for example, below 1V, to have a higher threshold. During soft program, the word line voltage is set to 0V or a negative voltage such as −0.5V or −1.0 V rather than 8.0 V. Soft programming is also known in the art as over-erase correction.
NMOS transistor M2 preferably is a triple well NMOS transistor, since a negative voltage is applied as signal VNEG during erase and when using the biasing conditions of Program-2 and Soft Program-1. The NMOS bulk must be biased at the most negative voltage. If a normal NMOS is used, the bulk is at VSS and is p-type. When a negative voltage is input to the word line, the p-n junction on the bulk to n+ source/drain would be turned on. Turn-on is avoided using a triple well NMOS.
Exemplary bias conditions for the embodiment of
As shown in
The use of two transistors in each word line driver saves large amounts of area verses prior art three or more transistor designs. In embodiments where two NMOS transistors are used, there are penalties, but these penalties are outweighed by the area savings. For example, during programming, the NMOS gate must be biased at a voltage value GM0[0] (e.g., 10.5V) that is larger than main word line voltage MWL0[0] (e.g., 8V) so that the full main word line voltage level can be passed to the local word lines. The NMOS transistor M3 has a threshold voltage, Vth. The local word line LWL0[0] will be biased at a maximum level of GM0[0] minus Vth, depending on the voltage level of the main word line. If the main word line voltage is less than this maximum number, the full voltage can pass to the local word line. Therefore, GM0[0] should be at least Vth+MWL0[0] (i.e., the voltage level of the main word line). This is not an issue if a PMOS transistor is selected to pass the voltage. The PMOS gate is biased at GM0[0] that is a negative voltage (e.g., −2V) or ground so as to fully pass the main word line voltage. In the dual NMOS embodiment, a separate voltage circuit is used to provide a voltage that exceeds the main word line voltage, but this signal is a global signal and any area consumed by this high voltage circuit is much smaller than the area consumed by the use of third transistor or other transistors in each local word line driver of the prior art. Various designs for circuits for providing this high voltage are known per se to those in the art and need not be detailed herein. By way of example, like other voltages that are higher than VCC, this voltage can be generated by pump and regulator circuits. Though not by way of limitation, examples of charge pump circuits are described in U.S. Pat. No. 5,793,679 to Caser et al. and U.S. Publication No. 2005/0207236 A1 to Yamazoe et al.
It should apparent to those in the art that the word line driver design described herein is incorporated into an integrated circuit having the NOR cell memory array and other circuit components, including control logic, address decoder circuitry such as row and column decoders, and other circuit components or modules familiar to those in the art.
Exemplary applications for the flash memory described herein incorporating the exemplary word line driver include, but are not limited to, digital audio players, digital cameras, mobile telephones, USB flash drives (thumb drives), SPI serial flash and gaming memory cards.
Although the invention has been described in terms of exemplary embodiments, it is not limited thereto. Rather, the appended claims should be construed broadly to include other variants and embodiments of the invention that may be made by those skilled in the art without departing from the scope and range of equivalents of the invention.
This application is a continuation of U.S. patent application Ser. No. 14/933,264, filed on Nov. 5, 2015, which is a continuation of U.S. patent application Ser. No. 13/895,591 filed May 16, 2013, now issued as U.S. Pat. No. 9,214,233 on Dec. 15, 2015, which is a continuation of U.S. patent application Ser. No. 13/477,431, filed May 22, 2012, now U.S. Pat. No. 8,456,922, which is a continuation of U.S. patent application Ser. No. 11/610,573 filed Dec. 14, 2006, now U.S. Pat. No. 8,189,396, which are hereby incorporated by reference.
Number | Date | Country | |
---|---|---|---|
Parent | 14933264 | Nov 2015 | US |
Child | 16583352 | US | |
Parent | 13895591 | May 2013 | US |
Child | 14933264 | US | |
Parent | 13477431 | May 2012 | US |
Child | 13895591 | US | |
Parent | 11610573 | Dec 2006 | US |
Child | 13477431 | US |