Patent Application
20050099875
This invention generally concerns memory arrays, and more particularly, techniques for identifying and substituting for defective locations in memory arrays.
Memory arrays consist of a number of storage locations or memory cells. Memory arrays are generally coupled to a storage control unit for accessing storage locations (cells) in order to store information in, or to retrieve information from, the memory array. Memory arrays are generally constructed in a reliable and robust manner. However, some random number of memory locations may be or become defective. It is desirable to avoid storing data to, or retrieving data from, defective memory locations. In order to avoid accessing defective memory locations, a memory map is generally devised to map addresses for defective memory array cells to another memory array or another portion of the memory array that is known to be operative. This memory array or portion is commonly referred to as a spare or a redundant memory array. The memory map is stored in another memory array, known as the map memory array. The size required for the spare and map memory arrays depends on the number of defective memory cells to be replaced. If the primary or first memory array requires that all locations be spared, the spare memory array will need to be as large as the first memory array and the map memory array will also need to be large. Since relatively few locations in the first memory array usually are defective, the size of the spare memory array is usually far less than the size of the primary or first memory array. It is desirable, of course, to make the spare and map memory arrays as small as possible, as unused locations dedicated to spare or map memory are wasted and this wastes resources. In order to maximize the efficient use of memory resources, there is need for a method and corresponding apparatus for reducing the amount of data to be stored in the map memory array.
Briefly stated, in a first embodiment, the present invention contemplates a method for identifying defective cells in a memory array. The method includes steps of receiving a request for accessing an address and analyzing the address to determine when the address matches an address stored in a temporary memory array. When the address does not match any address stored in the temporary memory array, the method includes steps of sending a wait instruction, analyzing the address to determine which portion of compressed data stored in a map memory array to decompress and decompressing the portion of compressed data to provide expanded data. The method also includes steps of writing the expanded data to the temporary memory array and comparing the expanded data to the address to determine when the address corresponds to an expanded datum of the expanded data.
In a second embodiment, the present invention includes a method for accessing a memory array. The method includes steps of requesting an address for one or more memory array cells comprising a first memory array and comparing the address to decompressed data describing defective memory array cells in the first memory array to determine when the address and a datum from the decompressed data match. The method includes a step of routing the address to a second memory array when the address and a datum from the decompressed data match.
In a third embodiment, the present invention includes a method for accessing a memory array. The method includes steps of receiving a memory array access request including a requested address, generating a first hash code from the requested address and comparing the first hash code to hash codes for decompressed addresses stored in a temporary memory array. When a match is found between a hash code for a decompressed address and the first hash code, the method includes a step of routing the memory array address request to a spare memory array.
Another embodiment of the present invention includes a memory control circuit. The memory control circuit includes a storage control unit coupled to a bus. The storage control unit accesses memory array units to retrieve data from a first memory array unit in response to memory array access requests delivered via the bus. The memory control circuit also includes a first memory array coupled to the storage control unit. The first memory array is for temporarily storing data. The memory control circuit additionally includes a second memory array coupled to the storage control unit. The second memory array provides cells for replacing cells determined to be defective in the first memory array. The memory control circuit further includes a data compression circuit that compresses data describing memory array addresses corresponding to cells determined to be defective in the first memory array to provide compressed addresses. The data compression circuit also decompresses compressed addresses to provide decompressed addresses. The data compression circuit couples to the storage control unit.
The storage control unit 20 couples via the bus 17 to a collection of memory devices including ROM 31, RAM 35 (i.e., the primary memory or first memory array), a spare memory array 39, a map memory array 41 and a temporary memory array 43. The spare memory array 39, map memory array 41 and/or temporary memory array 43 may be included within the storage control unit 20, may comprise sections of RAM 35 reserved for this purpose or may comprise separate memory units.
The storage control unit 20 is coupled via bus 17 to the map memory array 41 and the temporary memory array 43. In operation, the CPU 12 operates on data from the I/O device 15 in response to commands retrieved from memory, either from ROM 31 or RAM 35 via the storage control unit 20, and also stores data via the storage control unit 20 into RAM 35. Memory cells in each of the memories ROM 31, RAM 35, spare memory array 39, map memory array 41 and temporary memory array 43 include a regular arrangement having a fixed number of dimensions. These dimensions are ordered in some fashion, e.g., page, row and column. Each of these dimensions are then subdivided and numbered, for example, page 1, 2, 3 and the like. This common scheme specifies a regular structure, repeating for some interval. This also allows for easy accessing of data stored in the memory array as the page number, row number and column number are all that is necessary to access memory cells for information storage or retrieval. Because of the regular structure of the memory array, creating a map to direct retrieval or storage to spare memory cell locations in the spare memory array 39 for replacement of defective memory cells in, e.g., RAM 35, is also simplified.
For example, if page 3, row 16, column 32 is the location of a defective storage cell, the map may need only the numbers 3, 16, 32 in order to provide a replacement memory cell. Replacement locations are normally assigned on a first come, first served basis. Thus, for every entry in the map memory array 41, a sequential replacement location may be assigned in the spare memory array 39. While this method allows efficient and easy access to replacement memory cells in the spare memory array 39, certain types of defects consume large amounts of the map memory array 41 storage space.
In the case where errors are not in a random distribution, e.g., burst errors where multiple cells are centered around a particular location, or the loss of an entire dimension in the memory array, such as the loss of an entire row or column, significant efficiencies may be gained in utilization of the map memory array 41 using data compression to reduce the map data size required to store the addresses of the defective memory cells in the map memory array 41 prior to storage. For example, if page 0, row 10, column 13 in RAM memory array 35 had 192 consecutive defective locations on it, starting with the first location, then a simple data compression saving 191 map locations is effected by representing the defects in the RAM memory array 35 with the record 0, 10, 13, 192. This is a relatively primitive approach to data compression, however, it illustrates the basic principle. When more sophisticated compression algorithms are used, greater compression efficiency may be achieved.
In the operation of the computer system 10 of
One way of accomplishing this is to use a hash coding algorithm. During data compression within storage control unit 20, a code word may be generated that points to each defective cell address from the RAM 35 mapped in the map memory array 41. A cyclic code or some other robust process may be used to quickly produce a unique code word for each defective cell address in the primary memory. This code word represents a “hash code” and may be used as an index into the decompressed addresses and their associated hash codes stored in temporary memory array 43.
One way to generate hash codes is to treat the address data as a contiguous series of bits, i.e., as a number. This number is then divided by a prime number, such as 11. The most significant bits of the hash code provide an estimate of the address data and is useful as a quick index to a given number of replacement records in the map memory array 41.
When a hash code matching one of the hash codes stored in the temporary memory array 43 is produced by analysis of the address for the accessed memory cell site, the matching record in the temporary memory array 43 is used to redirect access to the spare memory array 39. When the comparison of the hash code for the address to which access is sought to the hash codes stored in the temporary memory array 43 produces a “miss” i.e., the defective cell addresses in the primary memory array 35 are not available in expanded form in the temporary memory array 43, a wait request may be issued to the CPU 12 while the data stored in the map memory array 41 are decompressed by the storage control unit 20.
Alternatively, an estimate may be made as to the range of memory array addresses required and some of the compressed data from the map memory array 41 may be uncompressed by the storage control unit 20 and then stored in the temporary memory array 43. A least-recently-used algorithm may be used to specify replacement cell addresses from the temporary memory array 43.
In a preferred embodiment, the storage control unit 20 of
A hash code register stack 209 stores hash codes for uncompressed address data from the map memory array 41 that are stored in the temporary memory array 43. The hash code register stack 209 provides quick reference to the storage addresses corresponding to defective memory locations in RAM 35 that have been most recently accessed and includes records for hash codes corresponding to memory locations that are currently stored in an uncompressed state, for example, in temporary memory array 43. An error detection code generator 211 provides and stores a check code every time a memory location is accessed for a write operation. When the same memory location is read, another check code is generated and is compared with the earlier check code generated during the write operation. When the two code words differ, an error has occurred. The error check operation may be implemented using any known method, including methods as simple as parity checking or as thorough as the Reed-Solomon technique.
The storage control unit 20 then analyzes the address to determine if it is the address of a location known to have failed, using a hash code that is calculated when a memory cell test indicates a defective memory cell. When there is a match between the hash code for the current address to which access is sought and a stored hash code in the storage control unit 20, the storage control unit 20 fetches the relocation record from the relocation record register 207 using the using the address as a key, obtains the address in the spare memory array assigned to the requested address and substitutes this spare memory array address for the requested address. When there is no match between the hash code for the current address to which access is sought and a stored hash code in the storage control unit, the current address is treated as an address for a defective memory cell as described with reference to the initial memory testing (see
While
When the query task 60 determines that the last cell in the array has been tested, defective memory array cell address data are compressed in step 64, for example, via the storage control unit 20 of
The process 50 thus tests, for example, all memory locations in RAM 35 during a power on or reset operation, or any other type of reset operation (or in response to a memory test command). The process 50 gleans information describing defective memory array cell address data for RAM 35, compresses those data and then stores the compressed data in the map memory array 41.
When the query task 76 determines that a requested address is for a defective memory array cell, a spare memory array address is obtained from the temporary memory array 43 in step 80, and a cell in the spare memory array 39 is accessed in step 82. Control then passes to the query task 84 to determine whether or not all memory array access requests have been executed.
When the address that is requested is not in the range of the temporary addresses stored in the temporary memory array 43, an optional “wait” signal may be passed in step 92 to the unit requesting memory access, such as the CPU 12 of
The invention as described may allow a memory array to be operated more efficiently by reducing the number of memory cells needed in order to store a map of defective memory cells in the memory array. It may be integrated into a memory chip, implemented in software or by discrete components or it may be employed together with memory devices in a larger system.
From the foregoing it will be appreciated that, although specific embodiments of the invention have been described herein for purposes of illustration, various modifications may be made without deviating from the spirit and scope of the invention. Accordingly, the invention is not limited except as by the appended claims.
| Number | Date | Country | |
|---|---|---|---|
| Parent | 09695756 | Oct 2000 | US |
| Child | 11007502 | Dec 2004 | US |
| Parent | 09012036 | Jan 1998 | US |
| Child | 09695756 | Oct 2000 | US |
