KR20020086746A - Synchronous flash memory with concurrent write and read operation - Google Patents

Abstract

Synchronous flash memory includes an array of nonvolatile memory cells. The memory array is arranged in a matrix, and can also be arranged in addressable blocks. The data communication connection is used for bidirectional data communication with external devices such as a processor or other memory controller. The write latch is coupled between the data buffer and the memory array to latch data provided to the data communication connection. The memory may write data to one location, such as a memory array block, while the data is read from a second location, such as a second memory array block. Write and read operations are performed on the common addressable rows of the array block.

Description

SYNCHRONOUS FLASH MEMORY WITH CONCURRENT WRITE AND READ OPERATION}

The memory device is typically provided with internal storage areas of the computer. The term memory relates to data storage in the form of integrated circuit chips. There are several different types of memories. One type is random-access memory (RAM). It is usually used as main memory in computer environment. RAM is associated with read and write memory; That is, data may be written to or read from the RAM. This is in contrast to ROM, which only allows reading data. Most RAM is volatile, which means that a steady flow of electricity is required to maintain its contents. As soon as power is turned off, whatever data in RAM is lost.

Computers almost always contain a small amount of read-only memory (ROM) that holds instructions for running the computer. Unlike RAM, it cannot be written to ROM. Electrically erasable programmable read-only memory (EEPROM) is a specific type of nonvolatile ROM that can be erased by exposure to electrical charge. Like other types of ROM, EEPROM is typically not as fast as RAM. The EEPROM includes a plurality of memory cells having electrically insulated gates (floating gates). Data is stored in memory cells in the form of charges in floating gates. Charge is transferred to or removed from the floating gates by programming and erase operations, respectively.

Another type of nonvolatile memory is flash memory. Flash memory is a type of EEPROM that can be erased and reprogrammed in blocks instead of one byte at a time. Many modern PCS have a BIOS stored on a flash memory chip that can be easily updated when needed. The BIOS is often referred to as the flash BIOS. Flash memory is also popular in modems because it is standardized so that modem manufacturers can support new protocols.

Typical flash memory includes a memory array comprising a plurality of memory cells arranged in a row and column fashion. Each memory cell includes a floating gate field effect transistor capable of retaining charge. Cells are typically grouped into blocks. Each of the cells in the block can be electrically programmed in any manner by charging the floating gate. Charge can be removed from the floating gate by a block erase operation. Cell data is determined by the presence or absence of charge in the floating gate.

Synchronous DRAM (SDRAM) is a type of DRAM that can operate at much higher clock speeds than conventional DRAM memory. SDRAM is synchronized to the CPU bus and is approximately 3 times faster than conventional fast page mode (FPM) RAM, and 100 MHZ approximately 2 times faster than high-speed Extended Data Ouput (EDO) DRAM and Burst Extended Data Output (BEDO) DRAM. Can be operated. SDRAM is quickly accessed, but volatile. Many computer systems are designed to operate using SDRAM, but it is more beneficial to use nonvolatile memory.

For the reasons described above, and for other reasons described below that will be apparent to those skilled in the art upon reading and understanding the specification, non-volatile memory that can operate in a manner similar to SDRAM operation in the art. You need a device. In particular, there is a need for a synchronous nonvolatile memory that allows simultaneous read and write operations simultaneously.

<Summary>

The above and other problems associated with memory devices will be addressed by the present invention and will be understood by reading and studying the following specification.

In one embodiment, the nonvolatile memory includes an array of nonvolatile memory cells arranged in addressable blocks, and a read / write circuit coupled to the array. The read / write circuit writes the first data in the first addressable block and simultaneously reads the second data from the second addressable block. The memory may include a latch for storing the first data prior to writing to the memory array.

In another embodiment, a method of operating a nonvolatile memory includes writing first data to a first location within a memory array of the nonvolatile memory; And reading the second data from the second location in the memory array of the nonvolatile memory at about the same time.

In another embodiment, a method of operating a memory device provided includes receiving first externally supplied data, storing first data in a write latch, and first data from the write latch in a first memory array of a memory. Writing to the location, and reading the second data from the second location in the memory array of the memory at about the same time.

In yet another embodiment, a method of operating a nonvolatile memory device includes receiving first data from a first external processor coupled to a memory, storing the first data in a write latch, and deleting the first data from the write latch. Writing to a first location within the memory array of the volatile memory device. The method also reads second data from a second location in the memory array of the nonvolatile memory device at about the same time and outputs the second data to a second external processor coupled to the memory device.

The present invention relates generally to nonvolatile memory devices and in particular to synchronous nonvolatile flash memory.

1A is a block diagram of a synchronous flash memory of the present invention.

1B is an integrated circuit pin interconnect diagram of one embodiment of the present invention.

1C is a diagram of an integrated circuit interconnect bump grid array of one embodiment of the present invention.

2A and 2B show a mode register of one embodiment of the present invention.

3 is a diagram illustrating read operations with CAS latency of one, two and three clock cycles.

4 is a diagram for activating a specific row of a memory bank of one embodiment of the present invention.

5 shows timing between an active command and a read or write command.

6 shows a read command.

7 illustrates timing of successive read bursts in one embodiment of the invention.

Figure 8 illustrates random read accesses within a page of one embodiment of the present invention.

9 is a diagram showing a read operation followed by a write operation.

10 is a diagram illustrating a read burst operation that is terminated using a burst end command according to an embodiment of the present invention.

11 is a diagram showing a write command.

12 is a diagram showing a recording followed by a read operation.

13 is a diagram illustrating a power-down operation of one embodiment of the present invention.

14 is a diagram illustrating a clock interrupt operation during burst reading.

FIG. 15 is a diagram illustrating a memory address map of one embodiment of a memory having two boot sectors. FIG.

16 is a flowchart of a self-timing recording sequence, in accordance with an embodiment of the present invention.

17 is a flowchart of a complete write state-check sequence in accordance with an embodiment of the present invention.

18 is a flowchart of a self-timed block erase sequence in accordance with an embodiment of the present invention.

19 is a flowchart of a complete block erase state-check sequence in accordance with an embodiment of the present invention.

20 is a flowchart of a block protection sequence in accordance with an embodiment of the present invention.

21 is a flowchart of a complete block state-check sequence in accordance with an embodiment of the present invention.

22 is a flowchart of a device protection sequence in accordance with an embodiment of the present invention.

23 is a flowchart of a device unprotected sequence, according to one embodiment of the invention.

24 is a diagram illustrating timing of initialization and load mode register operations.

25 is a diagram illustrating timing of a clock interrupt mode operation.

Fig. 26 is a diagram showing the timing of the burst read operation.

27 is a diagram illustrating timing of replacement bank read accesses.

Fig. 28 is a diagram showing the timing of all page burst read operations.

29 is a diagram showing timing of a burst read operation using a data mask signal.

30 is a diagram showing the timing of a write operation followed by reading from different banks.

31 is a diagram showing the timing of a write operation followed by a read of the same bank.

32 is a diagram showing a memory system of the present invention.

Figure 33 illustrates a multi-processor system of the present invention.

In the following detailed description of embodiments of the invention, reference is made to the accompanying drawings, which form a part thereof and in which are shown by way of illustration specific embodiments in which the invention may be implemented. The embodiments are described in sufficient detail to enable those skilled in the art to implement the invention, other embodiments may be used, and logical, mechanical, and electrical changes may be made within the spirit and scope of the invention. Will know. Accordingly, the following detailed description is not to be taken in a limiting sense, and the scope of the present invention is limited only by the claims.

The following detailed description is divided into two main sections. The first section is the Interface Functional Description detailing the compatibility with SDRAM memory. The second main section is the Functional Description, which describes the flash architecture function commands.

Interface function technology

1A, a block diagram of one embodiment of the present invention is shown. The memory device 100 includes a nonvolatile flash memory cell array 102. The array is arranged in a plurality of addressable banks. In one embodiment, the memory includes memory banks 104, 106, 108 and 110. Each memory bank includes addressable sectors of memory cells. Data stored in memory can be accessed using externally provided location addresses received by address register 112. The addresses are decoded using the row address multiplex circuit 114. The addresses are also decoded using the bank control logic 116 and the row address latch and decode circuit 118. To access a suitable column of memory, the column address counter and latch circuit 120 couples the received addresses to the column decode circuit 122. Circuit 124 provides input / output gating, data mask logic, read data latch circuits, and write driver circuits. Data is input through data input registers 126 and output through data output registers 128. Command execution logic 130 is provided to control the basic operations of the memory device. State machine 132 is also provided to control certain operations executed in memory arrays and cells. Status register 134 and identification register 136 may also be provided for outputting data.

1B is a diagram illustrating interconnect pin assignment in one embodiment of the present invention. Memory package 150 has 54 interconnect pins. The pin configuration is very similar to the effective SDRAM packages. Two interconnects specific to the present invention are RP # 152 and Vccp 154. Although the present invention may share interconnect labels that appear identical to SDRAM, the functionality of the signals provided to the interconnects is not the same as SDRAM unless described herein and described herein. FIG. 1C illustrates one embodiment of a memory package 160 with bump connections instead of the pin connections of FIG. 1C. Thus, the present invention is not limited to any particular package configuration.

Prior to describing the operating features of the memory device, the interconnect pins and respective signals are described in more detail. The input clock connection is used to provide a clock signal CLK. The clock signal can be driven by the system clock, and all synchronous flash memory input signals are sampled at the positive edge of CLK. CLK also increments the internal burst counter and controls the output register.

The input clock enable (CKE) connection is used to activate (HIGH state) and deactivate (LOW state) the CLK signal input. By disabling the clock input, it provides POWER-DOWN and STANDBY operations (all memory banks are idle), ACTIVE POWER-DWON (memory rows are ACTIVE in other banks), or CLOCK SUSPEND operation (burst / access in progress). The CKE is synchronous except after the device is in power-down modes, and the CKE is asynchronous until it exits the same mode. Input buffers, including CLK, are disabled during power-down mode to provide low reserve power. In systems where power-down modes (not RP # deep power-down) are not required, the CKE may be HIGH.

The chip select (CS #) input connection provides a signal to enable (LOW register) and disable (HIGH register) the command decoder provided to the command execution logic. All commands are masked when CS # is registered as HIGH. In addition, CS # provides multiple banks for selecting an external bank of the system, and CS # may be considered part of the command code; It doesn't have to be.

Input Commands for RAS #, CAS # and WE # (along with CAS #, CS #) Input connections define commands executed by the memory as described in detail below. Input / output mask (DQM) connections are used to provide input mask signals for write access and output enable signals for read access. Input data is masked when the DQM is sampled HIGH during the WRITE cycle. When the DQM is sampled HIGH during the READ cycle (after a 2-clock latency), the output buffers are in high impedance (High-Z) state. DQML corresponds to data connections DQ0-DQ7 and DQMH corresponds to data connections DQ8-DQ15. DQML and DQMH are considered to be in the same state when described as DQM.

Address inputs 133 are mainly used to provide address signals. In this embodiment, the memory has twelve lines A0-A11. Other signals may be provided over address connections as described below. Address inputs are sampled during the ACTIVE command (row-address A0-A11) and READ / WRITE command (column-address A0-A7) to select one location of each memory bank. The address inputs are also used to provide an opcode during the LOAD COMMAND REGISTER operation described below. Address lines A0-A11 are also used to enter the mode settings during the LOAD MODE REGISTER operation.

Input reset / power-down (RP #) connection 140 is used for reset and power-down operations. At initial device power-up, a 100 ms delay is required after RP # transitions from LOW to HIGH in one embodiment for internal device initialization prior to issuing an executable command. The RP # signal clears the status register, sets the internal state machine (ISM) 132 to array read mode, and puts the device into deep power-down mode when low. During power-down, all input connections, including CS # 142, become "Don't Care" and all outputs are in High-Z state. When the RP # signal is equal to the VHH voltage (5V), all protection modes are ignored during WRITE and ERASE. The RP # signal also causes the device protection bit to be set to 1 (protected) when going to VHH and the block protection bits of the 16 bit register to be set to 0 (unprotected) at locations 0 and 15. The guard bits are described in more detail below. During all other modes of operation, RP # is held high.

Bank address input connections, BA0 and BA1, define the bank to which the ACTIVE, READ, WRITE, or BLOCK PROTECT command is applied. DQ0-DQ15 connections 143 are data bus connections used for bidirectional data communication. 1B, the VCCQ connection is used to provide isolated power to the DQ connections to improve noise immunity. In one embodiment, VCCQ = Vcc or 1.8V ± 0.15V. The VSSQ connection is used to provide isolated ground to the DQ for improved noise immunity. The VCC connection provides a 3V power supply. Ground connection is provided via the Vss connection. Another optional voltage is provided over the VCCP connection 144. The VCCP connection can be externally a VCC, supplying current during device initialization, WRITE, and ERASE operations. That is, writing or erasing the memory device can be performed using the VCCP voltage, and all other operations can be performed with the VCC voltage. The Vccp connection is coupled to the high voltage switch / pump circuit 145.

The following sections describe in more detail the operation of the synchronous flash memory. One embodiment of the invention is a nonvolatile, electrically sector-erasable (flash), programmable read only memory comprising 67,108,864 bits organized into 4,194,304 words in 16 bits. Other population densities are also contemplated, and the invention is not limited to the density of this example. Each memory bank is organized into four independent erasable blocks (16 in total). To ensure that the critical firmware is protected from accidental erase or overwrite, the memory may include sixteen 256K-word hardware and software lockable blocks. The four-bank architecture of memory supports concurrent operations.

Read access to any bank may occur concurrently with background WRITE or ERASE operations for any other bank. The synchronous flash memory has a synchronous interface (all signals are registered at the positive edge of the clock signal, CLK). Read accesses to the memory may be burst oriented. That is, memory access starts at the selected location and continues for the programmed number of locations in the programmed sequence. Read accesses start with the registration of the ACTIVE command and follow for the READ command. Address bits registered at the same time as the ACTIVE command are used to select the bank and row to be accessed. Address bits registered at the same time as the READ command are used to select the starting column location and bank for burst access.

Synchronous flash memory provides an option for burst termination during programmable page burst lengths of 1, 2, 4 or 8 locations or for all pages. In addition, synchronous flash memory uses an internal pipeline architecture to achieve high speed operation.

Synchronous flash memory can operate in low power memory systems such as a system operating at 3 volts. Deep power-down mode is provided with power-saving redundancy mode. All inputs and outputs are low voltage transistor-transistor logic (LVTTL) compatible. Synchronous flash memory significantly advances flash operation performance, including the ability to synchronous burst data at high data transfer rates with automatic column address generation and the random change of column addresses in each clock cycle during burst access.

In general, a synchronous flash memory is constructed similar to a multi-bank DRAM that operates at a low voltage and includes a synchronous interface. Each bank is organized into rows and columns. Prior to normal operation, the synchronous flash memory is initialized. The following sections provide detailed information about device initialization, register definitions, command descriptions, and device operation.

The synchronous flash is powered up and initialized in a predetermined manner. After power is applied (at the same time) to VCC, VCCQ and VCCP, and the clock signal is stable, RP # 140 goes from a LOW state to a HIGH state. After RP # transitions to HIGH, a delay such as a 100ms delay is required to complete internal device initialization. After the delay time has elapsed, the memory enters array read mode and waits for a mode register programming or executable command. After initial programming of nonvolatile mode register 147 (NVMode register), the contents are automatically loaded into volatile mode register 148 during initialization. The device powers up in its programmed state and does not require reloading of the nonvolatile mode register 147 prior to issuing operational commands. This is described in more detail below.

The mode register 148 is used to define a particular mode of operation of the synchronous flash memory. The definition includes selection of burst length, burst type, CAS latency and operating mode as shown in FIGS. 2A and 2B. The mode register holds the stored information until programmed and reprogrammed with the LOAD MODE REGISTER command. The contents of the mode register may be copied to the NVMode register 147. The NVMode register setting automatically loads the mode register 148 during initialization. The ERASE NVMODE REGISTER and WRITE NVMODE REGISTER command sequences are described in detail below. Those skilled in the art will appreciate that the SDRAM requires that the mode register be externally loaded during each initialization operation. The present invention allows the default mode to be stored in the NV mode register 147. The contents of the NV mode register are then copied to the volatile mode register 148 for access during memory operation.

In this embodiment, the mode register bits M0-M2 specify the burst length, M3 specifies the burst type (sequential or interleaving), M4-M6 specifies the CAS latency, M7 and M8 specify the operating mode, and M9 It is set to 1 and M10 and M11 are reserved. Since WRITE bursts are not currently implemented, M9 is set to logic 1 and write accesses are single location (non-burst) accesses. The mode register must be loaded when all banks are idle, and the controller must wait a specified time before starting further operation.

As shown in Table 1, read accesses to synchronous flash memory can be burst oriented and the burst length can be programmed. The burst length determines the maximum number of column locations that can be automatically accessed for a given READ command. Burst lengths of 1, 2, 4 or 8 locations are useful for sequential interleaving burst types, and all page bursts are useful for sequential type. A full page burst can be used with the BURST TERMINATE command to generate arbitrary burst lengths. That is, the burst can optionally end to provide custom length bursts. When a READ command is issued, a block of columns equal to the burst length is selected efficiently. All accesses to the burst occur within the block, which means that if the burst reaches a boundary, it will wrap within the block. The block is uniquely selected by A1-A7 when the burst length is set to 2, by A2-A7 when the burst length is set by 4 and by A3-A7 when the burst length is set by 8. The remaining (lowest) address bit (s) is used to select a starting location within the block. All page bursts cycle within the page when the boundary is reached.

Access within a given burst may be programmed to be sequential or interleaved; This is related to the burst type and is selected via bit M3. As shown in Table 1, the order of access within a burst is determined by the burst length, burst type, and starting column address.

The column address strobe (CAS) latency is the delay of clock cycles between registration of the READ command and validation of the first piece of output data of the DQ connections. The latency can be set in one, two, or three clock cycles. For example, if a READ command is registered at clock edge n and the latency is m clocks, then the data is validated by clock edge n + m. The DQ connections will begin driving data as a result of a clock edge one cycle earlier (n + m-1), and once the associated access time is satisfied, the data is valid by clock edge n + m. For example, if the clock cycle time is assumed that all relevant access times are satisfied, then if the READ command is registered at T0 and the latency is programmed at 2 clocks, then the DQs will start driving after T1 as shown in FIG. The data is made valid by T2. 3 illustrates example operating frequencies in which different clock latency settings may be used. Normal operation mode is selected by setting M7 and M8 to 0, and the programmed burst length is applied to the READ burst.

The following truth tables provide detailed information on the operation commands of the embodiment of the memory of the present invention. A description of the command is provided, which follows Truth Table 2.

Truth Table 1

Interface Commands and DQM Operation

Name (function) CS # RAS # CAS # WE # DQM ADDR DQ COMMNAD INHIBIT (NOP) H X X X X X X NO OPERATOIN (NOP) L H H H X X X ACTIVE (select banks and activate rows) L L H H X Bank / row X READ (bank, column select, and READ burst initiated) L H L H X Bank / column X WRITE (bank, column selection and WRITE initiated) L H L L X Bank / column available BURST TERMINATE L H H L X X active ACTIVE TERMINATE L L H L X X X LOAD COMMAND REGISTER L L L H X Com code X LOAD MODE REGISTER L L L L X Op code X Write Enable / Output Enable - - - - L - active Write-protection / output High-Z - - - - H - High-z

Truth Table 2

Flash memory command sequence

Regardless of whether the CLK signal is enabled or not, the COMMNAD INHIBIT function prevents new commands from being executed by the synchronous flash memory. Synchronous flash memory is effectively deselected, but operations already in progress are not affected.

The NO OPERATOIN (NOP) command is used to execute NOP for the selected synchronous flash memory (CS # is LOW). This prevents undesired commands from being registered during idle or standby states, and operations already in progress are not affected.

Mode register data is loaded via inputs A0-A11. The LOAD MODE REGISTER command can only be issued when all array banks are idle, and the next executable command cannot be issued until a predetermined time delay (MRD) is reached. The data in the NVMode register 147 is automatically loaded into the mode register 148 at power-up initialization and is the default data unless dynamically changed by the LOAD MODE REGISTER command.

The ACTIVE command is used to open (activate) a row of a specific array bank for the next access. The value of the BA0, BA1 inputs selects a bank, and the address provided through inputs A0-A11 selects a row. The row remains active for access until the next ACTIVE command, power-down or RESET.

The READ command is used to initiate a burst read access for the active row. The value of the BA0, BA1 inputs selects the bank, and the address provided via inputs A0-A7 selects the starting column location. Read data appears in the DQs under the influence of the logic level of the data mask (DQM) input that appears two clocks earlier. If a given DQM signal is registered as HIGH, the corresponding DQs become High-Z (high impedance) after two clocks; If the DQM signal is registered low, the DQs will provide valid data. Thus, the DQM input can be used to mask the output during the read operation.

The WRITE command is used to initiate a single-location write access to the active row. For the WRITE command, the WRITE SETUP command must be prioritized. The value of the BA0, BA1 inputs selects the bank, and the address provided via inputs A0-A7 selects the column location. The input data represented by the DQ is written to the memory array under the influence of the DQM input logic level, which coincides with the data. When the predetermined DQM signal is registered as LOW, corresponding data is written to the memory; If the DQM signal is registered HIGH, the corresponding data inputs are ignored and WRITE for the word / column location is not performed. The WRITE command when DQM is HIGH is considered NOP.

The ACTIVE TERMINATE command is not needed for synchronous flash memory and may be provided to end the read in a manner similar to the SDRAM PRECHARGE command. The ACTIVE TERMINATE command may be issued to end an ongoing BURST READ, and may or may not be banked.

The BURST TERMINATE command is used to truncate bursts of fixed length or all pages. The most recently registered READ command is truncated before the BURST TERMINATE command. BURST TERMINATE is not banked.

The load command register operation is used to initiate flash memory control commands for Command Execution Logic (CEL) 130. The CEL receives and interprets commands for the device. The commands control the operation and read path of the internal state machine 132 (ie, memory array 102, ID register 136 or status register 134).

Before a READ or WRITE command can be issued to a bank in synchronous flash memory, a row of the bank must be "open". This is accomplished through an ACTIVE command (defined by CS #, WE #, RAS #, CAS #) that selects both the bank and the row to be activated, see FIG.

After opening a row (after issuing an ACTIVE command), a READ or WRITE command can be issued to the row under the influence of the time period (tRCD) specification, where tRCD (MIN) is divided by the clock period and passed to the next integer. It is rounded up to determine the first clock edge after the ACTIVE command that a READ or WRITE command can be entered. For example, a tRCD specification result of 30 ns according to a 90 MHz clock (11.11 ns period) is 2.7 clocks, rounded to three. This is reflected in FIG. 5, which covers any case where 2 <tRCD (MIN) / tCK ≦ 3. (The same procedure is used to convert different specification limits from time units into clock cycles.)

If the minimum time interval between successive ACTIVE commands for the same bank is defined as tRC, the next ACTIVE command for a different row in the same bank can be issued without closing the previous active row.

The next ACTIVE command for the other bank can be issued while the first bank is being accessed, thereby reducing the total row access overhead. The minimum time interval between successive ACTIVE commands for different banks is defined as time period tRRD.

As shown in FIG. 6, READ bursts are initiated by a READ command (defined by CS #, WE #, RAS #, CAS #). Start column and bank addresses are provided by the READ command. During READ bursts, a valid data-out element from the starting column address may be available following the CAS latency after the READ command. Each next data-out element is validated by the next positive clock edge. Upon burst completion, if it is assumed that no other commands have been initiated, the DQ goes into the High-Z state. The previous page burst continues until it ends (at the end of the page, cycles to column 0 and continues). The data from any READ burst may be truncated by the next READ command, followed by the data from the next READ command immediately after the data from the fixed length READ burst. In other cases, a continuous flow of data can be maintained. The first data element from the new burst follows the last element of the completed burst or the last desired data element of the longer burst in truncation. A new READ command must be issued in x cycles before the clock edge for which the last desired data element is valid, where x is CAS latency-1. This is shown in FIG. 7 for CAS latencies of 1, 2 and 3; Data element n + 3 is the last element of a burst of four or the last desired element of a longer burst. Synchronous flash memory uses a pipeline architecture. Therefore, no 2n rules related to the prefetch architecture are needed. The READ command can be initiated in any clock cycle following the previous READ command. Full speed, random read accesses within the page may be performed as shown in FIG. 8, or each next READ may be executed for a different bank.

Data from any READ burst can be passed by the next WRITE command (WRITE commands must be preceded by WRITE SETUP), and data from a fixed-length READ burst is from the next WRITE command (affected by the bus return limit). The data of can be immediately followed. If I / O contention can be prevented, WRITE may be initiated at the clock edge immediately after the last (or last desired) data element from the READ burst. In certain system designs, it may be possible for a device driving input data to go low-Z before the synchronous flash memory DQs go high-Z. In this case, at least a single-cycle delay occurs between the last read data and the WRITE command.

The DQM input is used to prevent I / O contention as shown in FIG. The DQM signal must be asserted (HIGH) at at least two clocks before the WRITE command (DQM latency is two clocks for output buffers) to suppress data-out from READ. When the WRITE command is registered, the DQs go into the High-Z state (or remain in the High-Z state) regardless of the state of the DQM signal. The DQM signal must be de-asserted before the WRITE command (DQM latency is zero clocks for input buffers) to ensure that the written data is not masked. 9 illustrates the case where the clock frequency prevents bus contention without adding a NOP cycle.

READ bursts of fixed length or previous page may be truncated by ACTIVE TERMINATE (may or may not be bank) or BURST TERMINATE (not bank) commands. An ACTIVE TERMINATE or BURST TERMINATE command is issued in x cycles before the clock edge where the last desired data element is valid, where x is equal to CAS latency-1. This is shown in FIG. 10 for each possible CAS latency; Data element n + 3 is the last desired element of a burst of four or the last desired element of a longer burst.

Single-location WRITE is initiated by a WRITE command (defined by CS #, WE #, RAS #, CAS #) as shown in FIG. Start column and bank addresses are provided by the WRITE command. Once the WRITE command is registered, the READ command can be executed as defined by truth table 4 and truth table 5. One example is shown in FIG. 12. During WRITE, valid data-in is registered at the same time as the WRITE command.

Unlike SDRAM, synchronous flash does not require a PRECHARGE command to deactivate open rows in a particular bank or open rows in all banks. The ACTIVE TERMINATE command is similar to the BURST TERMINATE command; However, ACTIVE TERMINATE may or may not be banked. During the ACTIVE TERMINATE command, input A10 is asserted HIGH, ending BURST READ in any bank. When A10 is low during the ACTIVE TERMINATE command, BA0 and BA1 determine which bank will initiate the end operation. ACTIVE TERMINATE is considered NOP for banks not addressed by A10, BA0, BA1.

When there is no access in progress, the clock is enabled and a power-down occurs when the CKE is registered LOW simultaneously with NOP or COMMAND INHIBIT. Power-down disables input and output buffers (except CKE) after internal state machine operations (including WRITE operation) are completed to save power while in the standby state.

The power-down state is terminated by registering NOP or COMMAND INHIBIT and CKE as HIGH at the desired clock edge (satisfied with tCKS). See FIG. 13 for an example power-down operation.

Clock interrupt mode occurs when thermal access / burst is in progress and CKE is registered low. In the clock stop mode, the internal clock is deactivated to "freeze" the synchronization logic. For each positive clock edge whose CKE is sampled low, the next internal positive clock edge is stopped. Unless the clock is interrupted, any command or data present on the input pins is ignored at the time of the interrupted internal clock edge, any data present on the DQ pins remains driven, and the burst counters do not increment. (See the example in FIG. 14). The clock stop mode ends by registering the CKE as HIGH; The internal clock and associated operation are resumed for the next positive clock edge.

The burst read / single write mode is the default mode in one embodiment. All WRITE commands cause access to a single column location (burst of 1), and READ commands access columns according to the programmed burst length and sequence. Truth Table 3 below illustrates memory operations using the CKE signal.

Truth Table 3-CKE

Truth Table 4-Commands for Current State Bank n-Bank n

Truth Table 5-Commands for Current State Bank n-Bank m

Function technology

Synchronous flash memory uses a number of features that are ideally suited for code storage and execute-in-place applications on the SDRAM bus. The memory array is divided into individual erase blocks. Each block may be erased without affecting the data stored in the other blocks. The memory blocks are read, written and erased by issuing commands to the command execution logic 130 (CEL). CEL controls the operation of the internal state machine 132 (ISM), which fully controls all ERASENVMODE REGISTER, WRITE NVMODE REGISTER, WRITE, BLOCK ERASE, BLOCK PROTECT, DEVICE PROTECT, UNPROTECT ALL BLOCKS, and VERIFY operations. ISM 132 protects each memory location from over-erasure and optimizes each memory location for maximum data retention. In addition, the ISM greatly simplifies the control needed to record the devices of an internal programmer or external programmer.

Synchronous flash memory is organized into 16 independently erasable memory blocks, allowing memory locations to be erased without affecting the remaining memory data. Any block may be hardware-protected against accidental erase or write. The protection block requires the RP # pin to be driven to VHH (relatively high voltage) before it is modified. 256K-word blocks of locations 0 and 15 may have additional hardware protection. If the PROTECT BLOCK command has been executed for these blocks, the UNPROTECT ALL BLOCKS command will unlock all blocks except for blocks of locations 0 and 15, unless the RP # pin is in the VHH state. If an unintentional power loss or system reset occurs, this provides additional security for critical code during in-system firmware update.

Power-down initialization, ERASE, WRITE and PROTECT timings are simplified using ISM to control all programming algorithms in the memory array. ISM guarantees protection against over-erasure and optimizes the recording margin for each cell. During WRITE operation, the ISM automatically increments and monitors WRITE attempts, verifies the write margin for each memory cell and updates the ISM status register. When a BLOCK ERASE operation is performed, the ISM automatically overwrites the entire address block (removes over-eraser), increments, monitors ERSAE attempts, and sets the bits in the ISM status register.

The 8-bit ISM status register 134 allows the external processor 200 to monitor the status of the ISM during WRITE, ERSAE, and PROTECT operations. One bit SR7 of the 8-bit status register is globally set and cleared by the ISM. The bit indicates whether the ISM is in use by an ERSAE, WRITE or PROTECT task. The additional error information is set to three different bits SR3, SR4 and SR5: write and protect block errors, erase and unprotect all block errors, and device protection errors. Status register bits SR0, SR1 and SR2 provide detailed information about the ongoing ISM operation. The user can monitor whether device-level or bank-level ISM operations (including which banks are under ISM control) are in progress. Six bits SR3-SR5 must be cleared by the host system. The status register is described in more detail below with reference to Table 2.

CEL 130 receives and interprets commands for the device. The commands control the operation and read path of the ISM (ie, memory array, device configuration or status register). Commands may be issued to the CEL while the ISM is active.

To allow for maximum power conservation, synchronous flash memory features a very low current deep power-down mode. To enter the mode, RP # pin 140 (reset / power-down) is VSS ± 0.2V. To prevent accidental RESET, RP # must be held at Vss for 100ns before the device enters reset mode. When RP # is held at Vss, the device is in deep power-down mode. After the device is in deep power-down mode, the transition from LOW to HIGH of RP # will result in a device power-up initialization sequence as described herein. The transition from LOW to HIGH of RP # after entering the reset mode but before entering the deep power-down mode requires a 1 ms delay before issuing an executable command. When the device is in deep power-down mode, all buffers except the RP # buffer are disabled and the current draw goes low, up to 50µs, for example at 3.3V VCC. The RP # input must remain at Vss during deep power-down. Once in the RESET mode, the status register 134 is cleared and the ISM 132 is set to the array read mode.

The synchronous flash memory array architecture is designed so that sectors are erased without disturbing the rest of the array. The array is divided into 16 addressable "blocks" that can be erased independently. By erasing blocks rather than the entire array, total device endurance is enhanced along with system flexibility. Only the ERSAE and BLOCK PROTECT functions are block oriented. The sixteen addressable blocks are equally divided into four banks 104, 106, 108 and 110 of each of the four blocks. Four banks have simultaneous read-write functions. An ISM WRITE or ERSAE operation for any bank may occur concurrently with a READ operation for any other bank. Status register 134 may be polled to determine which bank is in ISM operation. Synchronous flash memory has a single background operation ISM to control power-up initialization, ERSAE, WRITE and PROTECT operations. Only one ISM operation can occur at any time; However, certain other commands may be executed during the ISM operation, including READ operations. Operation commands controlled by the ISM are defined as bank-level operation or device-level operation. WRITE and ERSAE are bank-level ISM operations. After the ISM bank operation is initiated, a READ for any location in the bank may output invalid data, while a READ for any other bank will read the array. The READ STATUS REGISTER command outputs the contents of the status register 134. The ISM status bit will indicate when the ISM operation is complete (SR7 = 1). When the ISM operation completes, the bank automatically enters array read mode. ERASE NVMODE REGISTER, WRITE NVMODE REGISTER, BLOCK PROTECT, DEVICE PROTECT, and UNPROTECT ALL BLOCKS are device-level ISM operations. Once ISM device-level operation has been initiated, a READ for any bank will output the contents of the array. The READ STATUS REGISTER command may be issued to determine the completion of an ISM operation. When SR = 1, the ISM operation is completed and the next ISM operation may begin. As discussed below, any block can be protected from unintentional ERSAE or WRITE by hardware circuitry that requires the RP # pin to be driven to VHH before WRITE or ERASE is initiated.

Any block may be hardware-protected to provide extra security for the most sensitive portions of the firmware. During WRITE or ERASE of the hardware protection block, the RP # pin must be held to VHH until WRITE or ERASE is completed. Any WRITE or ERASE attempt to the protected block when RP # = VHH is prevented and a write or erase error is caused. Blocks of locations 0 and 15 may have additional hardware protection to prevent accidental WRITE or ERASE operation. In this embodiment, the blocks cannot be software-locked via the UNPROTECT ALL BLOCKS command unless RP # = VHH. The protection status of any block can be checked by reading the block protection bit by the READ STATUS REGISTER command. In addition, to protect the block, a 3-cycle command sequence must be issued with a block address.

Synchronous flash memory may be characterized by three different types of READs. Depending on the mode, the READ operation generates data from one of the memory array, status registers, or device configuration registers. READ to the device configuration register or status register must be preceded by an LCR-ACTIVE cycle and the burst length of the data out is defined by the mode register settings. A READ that does not precede the next READ or LCR-ACTIVE cycle reads the array. However, some differences exist and are described in the sections below.

The READ command for any bank outputs the contents of the memory array. While a WRITE or ERSAE ISM operation is occurring, a READ for any location in the bank under ISM control may output invalid data. During an existing RESET operation, the device automatically enters array read mode.

The READ execution of status register 134 requires the same input sequence as when reading the array, except that the LCR READ STATUS REGISTER 70H cycle must precede the ACTIVE READ cycles. The burst length of the status register data-out is defined by the mode register 148. The status register contents are updated and latched on the next positive clock edge, subject to CAS latency. The device automatically enters array read mode for the next READs.

Reading any device configuration registers 136 requires the same input sequence as when reading the status register except that certain addresses must be issued. WE # must be HIGH and DQM and CS # must be LOW. To read the manufacturer compatibility ID, the addresses must be at 000000H and to read the device ID, the addresses must be at 000001H. Any block protection bits are read at the third address location in each erase block xx0002H and the device protection bits are read from location 000003H.

The DQ pins are used to enter data into the array. The address pins are used to specify an address location during the LOAD COMMAND REGISTER cycle or to enter a command in the CEL. Command inputs input an 8-bit command to CEL to control the operating mode of the device. WRITE is used to enter data into the memory array. The following section describes both types of inputs.

In order to execute command input, DQM must be LOW and CS # and WE # must be LOW. Address pins or DQ pins are used to input commands. Address pins not used for command input are "Don't Care" and must be held in a stable state. An 8-bit command is input at DQ0-DQ7 or A0-A7 and latched at the positive clock edge.

WRITE for the memory array sets the desired bits to logic zeros but cannot change a given bit from logic one to logic zero. Setting any bits to logic 1 requires the entire block to be erased. To run WRITE, DQM must be LOW, CS # and WE # must be LOW, and VCCP must be VCC. Writing to the protection block also requires the RP # pin to go to VHH. A0-A11 provide the address to be written, and the data to be written to the array is input at the DQ pins. Data and addresses are latched on the rising edge of the clock. WRITE must be preceded by the WRITE SETUP command.

To simplify the writing of memory blocks, synchronous flash uses an ISM that controls all internal algorithms for WRITE and ERASE cycles. An 8-bit command set is used to control the device. See truth tables 1 and 2 for a list of valid commands.

The 8-bit ISM status register 134 (see Table 2) is polled to check for ERASE NVMODE REGISTER, WRITE NVMODE REGISTER, WRITE, ERASE, BLOCK PROTECT, DEVICE PROTECT or UNPROTECT ALL BLOCKS completion or any related errors. Completion of the ISM operation can be monitored by issuing a READ STATUS REGISTER 70H command. The contents of the status register are output to DQ0-DQ7 and updated at the next positive clock edge (affected by CAS latency) for a fixed burst length as defined by the mode register settings. ISM operation is completed when SR7 = 1. All of the defined bits are set by the ISM, and only the ISM status bits are reset by the ISM. Clear / unprotected blocks, write / protect blocks, and device protection must be cleared using the CLEAR STATUS REGISTER (50H) command. This allows the user to choose when to poll and clear the status register. For example, the host system may execute multiple WRITE operations before checking the status register instead of checking after each individual WRITE. If an RP # signal or power down is asserted, the device clears the status register.

The device ID, manufacturer compatibility ID, device protection state and block protection state can all be read by issuing a READ DEVICE CONFIGURATION (90H) command. In order to read the desired register, a specific address must be declared. See Table 3 for a more detailed description of the various device configuration registers 136.

Commands can be issued that cause the device to enter different modes of operation. Each mode has certain operations that can be executed during that mode. Some modes require that the command sequence be written before it is reached. The following section describes the characteristics of each mode, and truth tables 1 and 2 list all the command sequences needed to carry out the desired operation. The read-write function causes a background operation write or erase to be executed for any bank, while simultaneously reading any other bank. In the write operation, the LCR-ACTIVE-WRITE command sequences of truth tables 1 and 2 must be completed in successive clock cycles. However, to simplify synchronous flash controller operation, a limited number of NOPs or COMMAND INHIBITs can be issued through the command sequence. For further protection, the command sequences must have the same bank address for three cycles. If the bank address is changed during the LCR-ACTIVE-WRITE command sequence, or if the command sequences are not contiguous (NOPs and COMMAND INHIBITs, NOPs and COMMAND INHIBITs are allowed), write and erase status bits SR4 and SR5. Setting and operation is prohibited.

At power-up and before any operational commands are issued to the device, the synchronous flash is initialized. After power is applied (at the same time) to VCC, VCCQ and VCCP and the clock is stable, RP # transitions from LOW to HIGH. A delay (100 ms delay in one embodiment) is required after RP # transitions to high to complete internal device initialization. The device is in array read mode upon completion of device initialization, and an execute command can be issued to the device.

To read each of the device ID, manufacturer compatibility ID, device protection bit and block protection bits, a READ DEVICE CONFIGURATION 90H command is issued. In this mode, specific addresses are issued to read the desired information. Manufacturer compatibility ID is read at 000000H; The device ID is read at 000001H. Manufacturer compatibility ID and device ID are output from DQ0-DQ7. The device protection bit is read at 000003H; Each of the block protection bits is read at the third address location within each block xx0002H. Device and block protection bits are output at DQ0.

Three consecutive commands of consecutive clock edges are needed to enter data into the array (NOPs and COMMAND INHIBITs are allowed between cycles). In the first cycle, the LOAD COMMAND REGISTER command is provided with WRITE SETUP 40H at A0-A7, and bank addresses are issued at BA0, BA1. The next command is ACTIVE, which activates the row address and confirms the bank address. The third cycle is WRITE, during which start columns, bank addresses and data are issued. The ISM status bit is set at the following clock edges (affected by CAS latency). While the ISM is executing WRITE, the ISM status bit SR7 is at zero. READ operations on banks under ISM control may generate invalid data. When the ISM status bit SR7 is set to logic 1, WRITE is complete and the bank is in array read mode and will wait for executable commands. Writing to the hardware-protection blocks requires the RP # pin to be set to VHH before the third cycle (WRITE) and RP # must be held to VHH until the ISM WRITE operation is completed. The write and erase status bits SR4 and SR5 are set if the LCR-ACTIVE-WRITE command sequence is not completed in successive cycles or the bank address is changed during any of the three cycles. After ISM initiates WRITE, it cannot be stopped except by RESET or part power down. Execution during WRITE may corrupt the data being recorded.

ERASE sequence execution sets all bits in the block to logical one. The command sequence required to run ERASE is similar to WRITE. To provide additional security against accidental block erase, three consecutive command sequences at consecutive clock edges are needed to initiate the ERASE of the block. In the first cycle, LOAD COMMANDREGISTER is provided with ERASE SETUP 20H at A0-A7, and the bank address of the block to be erased is issued at BA0, BA1. The next command is ACTIVE, where A10, A11, BA0, BA1 provide the array read mode of the block to be erased. The third cycle is WRITE, during which the ERASE CONFIRM (DOH) is provided at DQ0-DQ7 and the bank address is reissued. The ISM status bit is set at the following clock edges (affected by CAS latency). After ERASE CONFIRM (DOH) is issued, the ISM initiates an ERASE of the address block. Any READ operation on the bank in which the address block is present may output invalid data. When the ERASE operation is complete, the bank goes into array read mode and waits for executable commands. Clearing for the hardware-protection blocks requires that the RP # pin be set to VHH before the third cycle (WRITE) and RP # must be held to VHH until ERASE is complete (SR7 = 1). If the LCR-ACTIVE-WRITE command sequence does not complete in consecutive cycles (NOPs and COMMAND INHIBITs are allowed between cycles) or if the bank address is changed during one or more of the command cycles, the write and erase status bits SR4. And SR5) are set and the operation is prohibited.

The contents of the mode register 148 may be copied to the NVMode register 147 by the WRITE NVMODE REGISTER command. Before writing to the NVMode register, the ERASE NVMODE REGISTER command sequence must complete to set all bits of the NVMode register to logical one. The command sequence required to execute ERASE NVMODE REGISTER and WRITE NVMODE REGISTER is similar to WRITE. See truth table 2 for more information about the LCR-ACTIVE-WRITE commands required to run ERASE NVMODE REGISTER and WRITE NVMODE REGISTER. After the WRITE cycle of the ERASE NVMODE REGISTER or WRITE NVMODE REGISTER command sequence is registered, a READ command may be issued to the array. The new WRITE operation is not allowed until the current ISM operation is completed and SR7 = 1.

Execution of the BLOCK PROTECT sequence enables a first level of software / hardware-protection for a given block. The memory includes a 16-bit register with 1 bit corresponding to 16 protectable blocks. The memory also has a register that provides the device bit used to protect the entire device from write and erase operations. The command sequence required to execute BLOCK PROTECT is similar to WRITE. To provide additional security against accidental block protection, successive command cycles are needed to initiate a BLOCK PROTECT. In the first cycle, the PROTECT SETUP (60H) command is issued to the LOAD COMMAND REGISTER at A0-A7, and the bank address of the block to be protected is issued at BA0, BA1. The next command is ACTIVE, which activates the row of the block to be protected and checks the bank address. The third cycle is WRITE, during which the BLOCK PROTECT CONFIRM (01H) is provided at DQ0-DQ7 and the bank address is reissued. The ISM status bit is set at the following clock edges (affected by CAS latency). The ISM then initiates a PROTECT operation. If the LCR-ACTIVE-WRITE is not completed in consecutive cycles (NOPs and COMMAND INHIBITs are allowed between cycles) or the bank address is changed, the write and erase status bits SR4 and SR5 are set and the operation is inhibited. . When the ISM status bit SR7 is set to logic 1, PROTECT is complete and the bank is in array read mode and will wait for executable commands. If the block protect bit is set to 1 (protect), it can only be reset to 0 when the UNPROTECT ALL BLOCKS command is issued. The UNPROTECT ALL BLOCKS command sequence is similar to the BLOCK PROTECT command; In the third cycle, WRITE is issued a UNPROTECT ALL BLOCKS CONFIRM (D0H) command and the addresses are "Don't Care". See truth table 2 for additional information. The blocks of locations 0 and 15 have additional security. If the block protection bits of locations 0 and 15 are set to 1 (protected), then each of the RP # s is VHH before the third cycle of the UNPROTECT operation and is held by VHH until the operation is completed (SR7 = 1). The bit can be reset to zero. In addition, if the device protection bit is set, RP # must be set to VHH before the third cycle and held to VHH until the BLOCK PROTECT or UNPROTECT ALL BLOCKS operation is completed. To check the protection status of the block, a READ DEVICE CONFIGURATION (90H) command may be issued.

Run the DEVICE PROTECT sequence to set the device protection bit to 1 and prevent the block protection bit from changing. The command sequence required to execute DEVICE PROTECT is similar to WRITE. Three consecutive command cycles are needed to initiate the DEVICE PROTECT sequence. In the first cycle, LOAD COMMAND REGISTER is provided with PROTECT SETUP 60H at A0-A7, and bank addresses are issued at BA0, BA1. The bank address is "Don't Care" but the same bank address must be used for all three cycles. The next command is ACTIVE. The third cycle is WRITE, during which the DEVICE PROTECT (F1H) command is issued at DQ0-DQ7 and RP # goes to VHH. The ISM status bit is set at the following clock edges (affected by CAS latency). Executable commands may be issued to the device. RP # must be held to VHH until the WRITE operation completes (SR7 = 1). New WRITE operations are not allowed until the current ISM operation is completed. Once the device protection bit is set, it cannot be reset to zero. If the device protection bit is set to 1, BLOCK PROTECT or BLOCK UNPROTECT is prevented as long as RP # remains VHH during any operation. The device protection bit does not affect WRITE or ERASE operations. See Table 4 for more information on block and device protection operations.

After the ISM status bit (SR7) is set, device / bank (SR0), device protection (SR3), bank A0 (SR1), bank A1 (SR2), write / protect block (SR4), and erase / unprotect (SR5) states. The bits may be checked. If the status bit or the combined status bit of one of SR3, SR4, SR5 is set, an error occurred during operation. ISM cannot reset the SR3, SR4 or SR5 bits. To clear the bits, a CLEAR STATUS REGISTER (50H) command must be provided. Table 5 lists the combinations of errors.

Synchronous flash memory is designed and manufactured to meet advanced code and data storage requirements. To ensure this level of reliability, the VCCP must be at Vcc during WRITE or ERASE cycles. Operation outside of the threshold may reduce the number of WRITE and ERASE cycles that may be executed in the device. Each block is designed and processed for at least 100,000-WRITE / ERASE-cycle endurance.

Synchronous flash memory provides several power-saving functions that may be used in array read mode to conserve power. Deep power-down mode is enabled by bringing RP # to VSS ± 0.2V. The current draw (ICC) in this mode is low at up to 50mA. When CS # is HIGH, the device is in active standby mode. In this mode, the current is low with up to 30 mA ICC current. When CS # goes high during a write, erase, or protect operation, the ISM continues the WRITE operation, and the device consumes active Iccp power until the operation is complete.

Referring to FIG. 16, a flowchart of a self-time sequence in accordance with one embodiment of the present invention is shown. The sequence includes loading a command register (code 40H), receiving an active command and an array of rows, and receiving a write command and a column address. Then a sequence for status register polling is provided to determine if writing is complete. Polling monitors status register bit 7 (SR7) to determine if it is set to 1. A selection status check may be included. When writing is complete, the array is in array read mode.

Referring to FIG. 17, a flowchart of a state-check sequence in accordance with one embodiment of the present invention is shown. The sequence checks for status register bit 4 (SR4) to determine if it is set to zero. If SR4 is 1, there is an error during the write operation. Examine status register bit 3 (SR3) to determine if the sequence is set to zero. If SR3 is 1, there is an invalid write error during the write operation.

Referring to FIG. 18, a flowchart of a self-time block erase sequence in accordance with one embodiment of the present invention is shown. The sequence includes loading a command register (code 20H) and receiving an active command and an array of rows. The memory then determines if the block is protected. If not protected, the memory executes a write operation (D0H) on the block and monitors the status register for completion. A selection status check can be performed and the memory is in array read mode. If the block is protected, erasure is not allowed until the RP # signal is at a high voltage (VHH).

19 illustrates a flowchart of a complete block erase state-check sequence, in accordance with an embodiment of the present invention. The sequence monitors the status register to determine if a command sequence error has occurred (SR4 or SR5 = 1). When SR3 is set to 1, an invalid erase or unprotected error has occurred. Finally, if SR5 is set to 1, a block erase or unprotected error occurred.

20 is a flowchart of a block protection sequence in accordance with an embodiment of the present invention. The sequence includes loading a command register (code 60H) and receiving an active command and a row address. The memory then determines if the block is protected. If not protected, the memory executes a write operation (01H) on the block and monitors the status register for completion. A selection status check can be performed and the memory is in array read mode. If the block is protected, erasure is not allowed unless the RP # signal is at a high voltage (VHH).

Referring to Figure 21, a flowchart of a complete block state-checking sequence in accordance with one embodiment of the present invention is shown. The sequence monitors status register bits 3, 4 and 5 to determine if an error has been detected.

22 is a flowchart of a device protection sequence in accordance with an embodiment of the present invention. The sequence includes loading a command register (code 60H) and receiving an active command and a row address. The memory then determines if RP # is VHH. The memory executes the write operation F1H and monitors the status register for completion. A selection status check can be performed and the memory is in array read mode.

23 is a flowchart of a block unprotected sequence according to an embodiment of the present invention. The sequence includes loading a command register (code 60H) and receiving an active command and a row address. The memory then determines whether the memory device is protected. If not protected, the memory determines if boot locations (blocks 0 and 15) are protected. If the blocks are not protected, the memory executes a write operation D0H for the block and monitors the status register for completion. A selection status check can be performed and the memory is in array read mode. Once the device is protected, erasure is not allowed unless the RP # signal is at a high voltage (VHH). Similarly, if boot locations are protected, the memory determines whether all blocks are unprotected.

24 illustrates the timing of initialization and load mode register operations. The mode register is programmed by providing a load mode register command and providing an opcode on the address lines. The opcode is loaded into the mode register. As described above, the contents of the nonvolatile mode register are automatically loaded into the mode register upon power-up and no load mode register operation may be required.

FIG. 25 shows the timing of a clock interrupt mode operation, and FIG. 26 shows the timing of another burst read operation. 27 shows the timing of replacement bank read accesses. Here an active command is needed to change the bank addresses. The previous page burst read operation is shown in FIG. Note that all page bursts require a termination command unless they terminate themselves.

29 shows timing of a read operation using a data mask signal. The DQM signal is used to mask the data output so that Dout m + 1 is not provided over the DQ connections.

Referring to Fig. 30, the timing of a write operation followed by a read for different banks is shown. In this operation, writing is performed for bank a and the next read is performed for bank b. The same row is accessed in each bank.

Referring to Fig. 31, the timing of a write operation followed by a read on the same bank is shown. In this operation, writing is performed for bank a and the next read is also performed for bank a. Different rows are accessed during the read operation, and the memory must wait before the write operation is completed. This is different from the read of Fig. 30 in which the read is not delayed due to the write operation.

Synchronous flash memory provides latency free write operations. This is different from SDRAM, which requires the system to provide latency for the write operation, only the read operation is the same. The write operation does not take from the system bus as many cycles as the SDRAM takes, thus improving system read throughput. See Figure 12 where write data Din is provided in the same clock cycle as the write command and column address. Clock cycle T1 in FIG. 12 need not be a NOP command (see FIG. 30). The read command can be provided at the next clock cycle after the write data. Thus, while the read operation requires the read DQ connection to remain valid for a predetermined number of clock cycles after the write command (latency), the DQ connection can be used immediately after the write command is provided (without latency). Thus, the present invention allows zero bus turn around capability. This is virtually different from SDRAM, where a large number of waits are required on the bus when going between read and write operations. Synchronous flash offers these two characteristics and can improve bus throughput.

Referring to FIG. 32, the system 300 of the present invention includes a synchronous memory 302 having an internal write latch 304 used to store write data received at the DQ input 306. The write latch is coupled to the memory array 310. Again, the memory array may be arranged in a number of addressable blocks. Data can be written to one block while a read operation is performed on another block. The memory cells of the array may be nonvolatile memory cells. The data communication connection 306 is used for bidirectional data communication with an external device, such as the processor 320 or another memory controller.

The data buffer 330 may be coupled to a data communication connection to manage bidirectional data communication. This buffer can typically be a FIFO or pipelined input / output buffer circuit. The write latch is coupled between the data buffer and the memory array to latch the data provided on the data communication connection. As a result, control circuitry is provided to manage the read and write operations performed on the array.

By latching the input write data, the data bus 306 (data bus of DQ) can be released and the write operation can be performed using the latched data. Subsequent write operations to the memory can be prevented while the first write operation is performed. However, the bus is effective for performing a read operation on memory immediately.

The present invention cannot be confused with a conventional input / output buffer structure. That is, compared to conventional memory devices using input buffers on the DQ input path and output buffers on the DQ output path, the clock latency used for both read and write operations remains the same. The present invention may include an input / output buffer for providing an interface by an DQ connection and an external processor. Additional write latches allow the memory to separate write paths / operations into one area of memory, while allowing read operations to areas of another memory.

Simultaneous Read and Write Operations

Previous flash memory devices have very limited concurrent operation capabilities. In other words, previous flash memory typically blocked reading from the memory while a write operation was being performed. Some memory devices have allowed reading during writing by pausing an ongoing write operation and then reading the array. However, other flash memories could allow read-simultaneous-write by providing a limited sector that can be written while the remaining sectors of the memory are valid for reading. The purpose of such flash memory is to eliminate the need for an individual EEPROM in the system. Limited sector space provides EEPROM location within flash, leaving the rest of the memory for flash operation. However, such flash memory does not allow reads to be performed to the entirety of the memory array while a write operation can be performed.

The present invention provides a flash array arranged in a bank structure similar to SDRAM. In one embodiment, the 64M synchronous flash is divided into four 16M banks with the same addressing as 64M SDRAM. These banks are also divided into smaller addressable 4M sectors that may be deleted and programmed. The memory allows simultaneous reads and writes per bank. Thus, one bank can be written while simultaneous reads are performed in some other bank. As is known to those skilled in the art, the SDRAM has one common row open in each bank. Read and write operations can be performed sequentially through open rows and array banks.

This synchronous flash memory has a bank structure similar to SDRAM which allows one row to be opened in each bank while one of the banks is being written. Referring to Figure 33, one embodiment of a processing system 400 of the present invention is shown. Synchronous flash memory 410 is coupled to multiple processors 440, 442, 444, 446 via a bidirectional data bus 435. The memory includes an array of nonvolatile memory cells arranged in a number of banks 412, 414, 416, 418. Read / write circuit 430 is generally described as managing an array of data communications as an array. In operation, a processor, such as processor 440, may initiate the write operation of row 420 of array bank 412. While the write operation is being performed, a second processor, such as processor 442, may read data from row 420 of the second array bank. This allows four processors to operate synchronous flash memory devices independently. The invention is not limited to four banks or four processors. One or more processors may perform write operations to one bank while reading data from the remaining memory array banks (see FIG. 32). Thus, simultaneous write and multiple simultaneous read operations can proceed.

conclusion

A synchronous flash memory has been described that includes a nonvolatile memory cell array. The memory array may be arranged in a matrix, and further arranged in addressable blocks. The data communication connection is used for bidirectional data communication with external device (s) such as a processor or other memory controller. The write latch is coupled between the data buffer and the memory array to latch data provided to the data communication connection. The memory may write data to one location, such as a memory array block, while the data is read from a second location, such as a second memory array block. Write and read operations are performed on the common addressable rows of the array block.

Claims (23)

Non-volatile memory, A nonvolatile memory cell array arranged into a plurality of addressable banks; And Read / write circuit coupled to the array Including; The read / write circuit writes first data in a first bank of the plurality of addressable banks and simultaneously reads second data from a second bank of the plurality of addressable banks. Nonvolatile Memory. The method of claim 1, And said read / write circuit comprises a write latch for storing said first data provided to external data communication connections. The method of claim 1, And the plurality of addressable banks includes four blocks. The method of claim 1, And the read / write circuit receives the first data from a first external processor, and the read / write circuit provides the second data in response to a second external processor. The method of claim 1, The read / write circuit writes the first data to an array row of the first bank of the plurality of addressable banks, and simultaneously reads the first data from the array row of the second bank of the plurality of addressable banks. And non-volatile memory. The method of claim 1, And the nonvolatile memory is a synchronous nonvolatile memory. Non-volatile memory, A nonvolatile memory cell array arranged into a plurality of addressable banks; A write latch for storing first data provided to external data communication connections; And Read / write circuit coupled to the array Including, The read / write circuit writes the first data in a first bank of the plurality of addressable banks and simultaneously reads second data from a second bank of the plurality of addressable banks. Non-volatile memory characterized by. The method of claim 7, wherein And the read / write circuit receives the first data from a first external processor, and the read / write circuit provides the second data in response to a second external processor. In the processing system, A processor; And Non-volatile memory coupled to the processor Including, The nonvolatile memory, A nonvolatile memory cell array arranged into a plurality of addressable banks, and A read / write circuit coupled to the array, The read / write circuit writes first data provided by the processor to a first bank of the plurality of addressable banks, and reads second data from a second bank of the plurality of addressable banks. And provide the second data to the processor. The method of claim 9, And a write latch for storing the first data. In the processing system, First and second processors; And Non-volatile memory coupled to the first and second processors Including, The nonvolatile memory, A nonvolatile memory cell array arranged in a plurality of addressable banks, A data latch coupled to the array for storing write data, and A read / write circuit coupled to the array, The read / write circuit writes first data provided by the first processor to a first bank of the plurality of addressable banks, and simultaneously writes second data from a second bank of the plurality of addressable banks. Reads and provides the second data to the second processor. In the non-volatile memory operation method, Writing first data to a first bank location within a memory array of the nonvolatile memory; And Reading second data from a second bank location of the memory array of the non-volatile memory at about the same time. Non-volatile memory operation method comprising a. The method of claim 12, And the first data is provided to the nonvolatile memory through an external processor. The method of claim 12, And the first data is provided to the nonvolatile memory through the first external processor, and the method further comprises outputting the second data to a second external processor. The method of claim 12, And accessing the common addressable rows of the first and second blank locations prior to the writing of the first data and the reading of the second data. The method of claim 12, The writing of the first data may include storing the first data in a latch circuit before writing the first data to the first bank location in the memory array. How memory works. In the memory device operating method, Receiving first externally provided data; Storing the first data in a write latch; Writing the first data from the write latch to a first bank location in a memory array of the memory; And Reading second data from a second bank location in the memory array of the memory at about the same time. The method of operating a memory device comprising a. The method of claim 17, And wherein said first data is provided by an external processor. The method of claim 18, And outputting the second data to the external processor. The method of claim 17, And wherein the first data is provided by a first external processor, the method further comprises outputting the second data to a second external processor. A method of operating a nonvolatile memory device, Receiving first data from a first external processor coupled to the memory; Storing the first data in a write latch; Writing the first data from the write latch to a first location in a memory array of the nonvolatile memory device; Reading second data from a second location in the memory array of the nonvolatile memory device at about the same time; And Outputting the second data to a second external processor coupled to the memory device And operating the non- volatile memory device. The method of claim 21, And wherein the first and second locations are first and second addressable blocks of the memory array. The method of claim 22, And wherein the first data is written to a row of the first block and the second data is read from a common row of the second block.