Adjusting access of non-volatile semiconductor memory based on access time

Abstract

A non-volatile semiconductor memory is disclosed comprising a first memory device and control circuitry operable to issue an access command to the first memory device. A command status is requested from the first memory device after a status delay. When the command status indicates the first memory device has completed the command, a first access time of the memory device is measured. An access sequence of the first memory device is then modified in response to the access time.

Description

BACKGROUND

Description of the Related Art

A non-volatile semiconductor memory may be employed as mass storage for a computer system (e.g., desktop, laptop, portable, etc.) or a consumer device (e.g., music player, cell phone, camera, etc.) or other suitable application. The non-volatile semiconductor memory may comprise one or more memory devices (such as a flash memory) and control circuitry for accessing each memory device. Each memory device is coupled to an address bus and a data bus, as well as a number of interface control lines. When issuing a write command or an erase command to a memory device, the control circuitry configures the address bus and control lines, and then transfers command data (and user data for a write operation) over the data bus. After receiving the command, the memory device executes the command internally over an “access time”. To determine when the memory device is finished executing the command, the control circuitry typically polls the memory device by issuing a “command status” request wherein the memory device transmits the contents of a command status register to the control circuitry.

FIG. 1A illustrates a problem with the prior art sequence for writing or erasing a memory device. The waveform of FIG. 1A represents the ready/busy line of the memory device. After transferring (serial data in) the command data (and user data for a write operation), the control circuitry delays for a predetermined time (STATUS_DELAY) and then begins polling the memory device for the command status register. When the command status register indicates the previous command has completed successfully, the control circuitry may transmit a next command to the memory device. Depending on the speed of the particular memory device, the status delay time (STATUS_DELAY) may be significantly shorter than the actual access time, and therefore the control circuitry may begin polling for the status command register well before the memory device has finished processing the command. This may be undesirable since excessive polling for the status command register consumes power in battery operated host devices.

FIG. 1B illustrates another problem with the prior art sequence when writing to a first and second memory device over single channel. In the example shown in FIG. 1B, the second memory device has a shorter access time than the first memory device. After transmitting command data to the first memory device, command data is immediately transmitted to the second memory device during the access time of the first memory device. When the command status register of the first memory device indicates the command has completed successfully, a next command is transmitted to the first memory device. However, while transferring the next command to the first memory device, the second memory device may also be ready to receive a next command but must wait for the control circuitry to finish transmitting the next command to the first memory device (i.e., there is a collision in accessing the shared bus). This WAIT TIME shown in FIG. 1B is undesirable because it reduces the overall performance (throughput) of the non-volatile semiconductor memory.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1A illustrates a prior art access of a memory device wherein a status delay may be significantly shorter than the access time which increases the number of command status requests sent to the memory device.

FIG. 1B illustrates a prior art access of two memory devices over a shared bus wherein a collision may occur due to the different access times of the memory devices.

FIG. 2A shows a non-volatile semiconductor memory according to an embodiment of the present invention comprising a first and second memory device accessed by control circuitry.

FIG. 2B is a flow diagram executed by the control circuitry according to an embodiment of the present invention wherein the access time of the memory devices is measured in order to modify an access sequence of the memory devices.

FIG. 3A shows a prior art erase operation of a memory device wherein a status delay is significantly shorter than the access time which increases the number of command status requests sent to the memory device.

FIG. 3B illustrates an embodiment of the present invention wherein the access sequence of the memory devices is modified by adjusting the status delay in response to the measured access times, thereby reducing the number of command status requests sent to the memory devices.

FIG. 3C illustrates an embodiment of the present invention wherein the access sequence of the memory devices is modified by modifying an order in which commands are transmitted to the memory devices, thereby improving performance by reducing or avoiding wait times due to collisions.

FIG. 4 is a flow diagram according to an embodiment of the present invention wherein the access sequence of the memory devices is modified by adjusting the status delay in response to the measured access times.

FIG. 5 is a flow diagram according to an embodiment of the present invention wherein the access sequence of the memory devices is modified by modifying an order in which commands are transmitted to the memory devices.

FIG. 6A shows an embodiment of the present invention wherein the control circuitry accesses each memory device over a separate, dedicated bus.

FIG. 6B shows an embodiment of the present invention wherein the control circuitry accesses each memory device over a shared bus.

DETAILED DESCRIPTION OF EMBODIMENTS OF THE INVENTION

FIG. 2A shows a non-volatile semiconductor memory 2 according to an embodiment of the present invention comprising a first memory device 4A and control circuitry 6 operable to execute the flow diagram shown in FIG. 2B. An access command is issued to the first memory device (step 8). A command status is requested from the first memory device after a status delay (step 10). When the command status indicates the first memory device has completed the command (step 12), a first access time of the memory device is measured (step 14). An access sequence of the first memory device is then modified in response to the access time (step 16).

The memory device 4A in the embodiment of FIG. 2A may comprise any suitable device, such as a NAND flash memory. A typical NAND flash memory is accessed by transmitting a corresponding address and transmitting/receiving data over an address/data bus (ADDR/DATA), and asserting appropriate control lines (e.g., address latch enable (ALE), command latch enable (CLE), chip enable (CE), etc.). In the embodiment of FIG. 2A, the non-volatile semiconductor memory 2 comprises multiple memory devices (e.g., 4A, and 4B) which are accessed by the control circuitry 6 by transmitting the appropriate address over the address/data bus (which may be a parallel or serial bus). In one embodiment, the control circuitry 6 measures the access time of each memory device in FIG. 2B by initializing a counter (step 18) which is incremented (step 20) until the access time of all of the memory devices has been measured (step 22).

FIG. 3A shows a prior art erase operation of a memory device wherein after transmitting an erase command, the control circuitry waits for a nominal status delay and then begins polling the memory device by requesting a command status from the memory device. However, the nominal status delay may be significantly shorter than the actual access time of the memory device, and therefore the control circuitry may begin polling for the command status sooner than needed resulting in undesirable power dissipation. In an embodiment of the present invention shown in FIG. 3B, the control circuitry measures the access time of the memory device and then modifies the access sequence of the memory device by adjusting the status delay in response to the access time. In this manner, the control circuitry begins polling for the command status at a more optimal time, which reduces power consumption by reducing the number of requests for the command status as illustrated in FIG. 3B.

In one embodiment, the access time needed to perform an erase command is different from the access time needed to perform a write command. Accordingly, in one embodiment the status delay is optimized relative to different types of access commands (e.g., a status delay optimized for erase commands and a status delay optimized for write commands).

FIG. 3C shows an embodiment of the present invention wherein the access sequence of the memory devices is modified by selecting an order in which write commands are transmitted in response to the access times. For example, in one embodiment data is written to the memory devices using a “striping” technique so as to optimize performance. The data blocks of a write command are stored sequentially, one block at a time to each of the selected memory devices in a particular order. In this manner, the access times of each memory devices can overlap while concurrently processing the write commands rather than storing each block sequentially to an individual memory device which would require the control circuitry to wait for the access time of the memory device before transmitting the next data block.

In the example of FIG. 3C, the second memory device has a shorter access time than the first memory device, and therefore a first write command and corresponding data block is transmitted to the second memory device. During the access time of the second memory device, the control circuitry transmits a second write command and corresponding data block to the first memory device. When the second memory device is finished processing the first write command, it can immediately receive a new write command from the control circuitry 6. Similarly, when the first memory device is finished processing the second write command, it can immediately receive a new write command from the control circuitry. Accordingly, this embodiment may reduce or eliminate the wait time shown in the prior art order of FIG. 1B by reducing or avoiding collisions when attempting to access two or more of the memory devices over the shared bus at the same time.

FIG. 4 is a flow diagram according to an embodiment of the present invention which extends on the flow diagram of FIG. 2B, wherein the step of modifying the access sequence of the memory device (step 16 of FIG. 2B) comprises adjusting the status delay for the memory device (step 24) as illustrated in FIG. 3B. FIG. 5 is a flow diagram according to an embodiment of the present invention which extends on the flow diagram of FIG. 2B, wherein the step of modifying the access sequence of the memory device (step 16 of FIG. 2B) comprises modifying an order in which commands are transmitted to the memory device (step 26) as illustrated in FIG. 3C.

FIG. 6A shows a non-volatile semiconductor memory 2 according an embodiment of the present invention wherein the control circuitry 6 accesses each memory device (e.g., 4A and 4B) over separate, dedicated buses 28A and 28B. In this embodiment it's unnecessary to order the transmitted commands relative to the access times since there is never a collision attempting to simultaneously access a shared bus. However, this embodiment still benefits from optimizing the status delay in order to minimize the number of requests for the command status as illustrated in FIG. 3B.

FIG. 6B shows a non-volatile semiconductor memory 2 according an embodiment of the present invention wherein the control circuitry 6 accesses each memory device (e.g., 4A and 4B) over a shared bus 28. Sharing the bus 28 is illustrated conceptually in FIG. 6B as configuring a multiplexer 30 to access a desired memory device. However, sharing the bus may be implemented in any suitable manner, such as a wired-OR implementation with dedicated chip enable (CE) lines that enable the desired memory device being accessed. This embodiment benefits both from optimizing the status delay to reduce the number of command status requests, as well as ordering the transmitted commands relative to the access times in order to reduce or avoid collisions.

The process of measuring the access times of each memory device and modifying the access sequence may be performed at any suitable time. In one embodiment, the access times are measured and the access sequence modified during a manufacturing procedure of the non-volatile semiconductor memory. In other embodiments, the access times may vary over the lifetime of the non-volatile semiconductor memory 2 and/or affected by environmental conditions, such as temperature. Therefore, in one embodiment the access times may be measured whenever the non-volatile semiconductor memory is powered on, or measured periodically over the lifetime of the non-volatile semiconductor memory, or measured in response to detected changes in environmental conditions, or measured in response to a command received from a host system.

The non-volatile semiconductor memory 2 of the present invention may be employed in any suitable application, such as in a solid state disk drive (SSD), or in a memory card or memory stick. In addition, the non-volatile semiconductor memory 2 may employ any suitable memory devices 4A and 4B that may exhibit different access times (e.g., different access times for erase commands and/or for write commands).

Claims

1. A non-volatile semiconductor memory comprising a first memory device, a second memory device, and control circuitry operable to: measure a first access time of the first memory device following issuance of a first access command to the first memory device;measure a second access time of the second memory device following issuance of a second access command to the second memory device, wherein the second access command is issued after the first access command according to an access order of the first and second memory devices, wherein the access order defines a sequence of accessing the first and second memory devices for consecutive access commands;modify the access order of the first and second memory devices for subsequent consecutive access commands based on the measured first and second access times; andset a delay period of the subsequent consecutive access command associated with the first memory device, wherein the delay period is based on the first access time and indicates when a polling for a status of the subsequent consecutive access command begins after the subsequent consecutive access command is issued.

2. The non-volatile semiconductor memory as recited in claim 1, wherein the control circuitry is further operable to access the first and second memory devices over a shared bus.

3. The non-volatile semiconductor memory as recited in claim 2, wherein the control circuitry is further operable to modify the access order of the first and second memory devices for subsequent access commands by modifying an order in which the control circuitry transmits access commands to the first and second memory devices over the shared bus.

4. The non-volatile semiconductor memory as recited in claim 3, wherein the control circuitry is further operable to: first select one of the first and second memory devices in response to the first and second access times;transmit a third access command to the first selected memory device;second select a different one of the first and second memory devices in response to the first and second access times; andafter transmitting the third access command to the first selected memory device and during an access time associated with the first selected memory device for the third access command, transmit a fourth access command to the second selected memory device.

5. The non-volatile semiconductor memory as recited in claim 4, wherein the access time of the first selected memory device is less than the access time of the second selected memory device.

6. A method of operating a non-volatile semiconductor memory comprising a first memory device and a second memory device, the method comprising: issuing an access command to the first memory device;requesting a command status from the first memory device after a status delay;when the command status indicates the first memory device has completed the access command, measuring a first access time of the first memory device;measuring a second access time of the second memory device;modifying, based on the first and second access times, an order in which subsequent access commands are issued to the first and second memory devices such that the first memory device is before the second memory device in the order if the first access time is less than the second access time and the second memory device is before the first memory device in the order if the second access time is less than the first access time, wherein the order defines a sequence of accessing the first and second memory devices for consecutive access commands; andsetting a delay period of the subsequent consecutive access command associated with the first memory device, wherein the delay period is based on the first access time and indicates when a polling for a status of the subsequent consecutive access command begins after the subsequent consecutive access command is issued.

7. The method as recited in claim 6, further comprising accessing the first and second memory devices over a shared bus.

8. The method as recited in claim 7, further comprising modifying the order in which the subsequent access commands are issued to the first and second memory devices by modifying an order in which commands are transmitted to the first and second memory devices over the shared bus.

9. The method as recited in claim 8, further comprising: first selecting one of the first and second memory devices in response to the first and second access times;transmitting a third access command to the first selected memory device;second selecting a different one of the first and second memory devices in response to the first and second access times; andafter transmitting the third access command to the first selected memory device and during an access time associated with the first selected memory device for the third access command, transmitting a fourth access command to the second selected memory device.

10. The method as recited in claim 9, wherein the access time of the first selected memory device is less than the access time of the second selected memory device.

11. The non-volatile semiconductor memory as recited in claim 1, wherein modifying the access order further comprises: modify the access order such that the first memory device is before the second memory device if the first access time is less than the second access time; andmodify the access order such that the second memory device is before the first memory device if the second access time is less than the first access time.

12. A method comprising: determining a first access sequence for a first memory device and a second memory device, wherein the first memory device is before the second memory device in the first access sequence;determining a first access time of the first memory device;determining a second access time of the second memory device;determining that the second access time is less than the first access time; modifying, in response to determining the second access time is less than the first access time, the first access sequence to a second access sequence, wherein the second memory device is before the first memory device in the second access sequence, and the second access sequence defines a sequence of accessing the first and second memory devices for consecutive access commands; andsetting a delay period of the consecutive access command associated with the first memory device, wherein the delay period is based on the first access time and indicates when a polling for a status of the consecutive access command begins after the consecutive access command is issued.