As advances continue to be made in the area of
semiconductor memory devices, high capacity and low cost will be increasingly important. In particular, it will be necessary to create memory devices for which the testing of the device must be minimized in order to minimize costs. Current memory manufacturing costs are significant and will grow as the capacity of the devices grows—the higher the memory's capacity, the more storage locations that must be tested, and the longer the testing operation will take. The cost of the testing can be calculated by dividing the amortized cost of the
test equipment by the number of devices tested. As memory devices enter the
Gigabyte range and larger, the number of devices that can be tested by a given piece of
test equipment will go down. As a result, the cost per unit attributable to testing will rise. The present invention is a means and a method for accessing streams of data stored within a memory device so as to minimize the cost of device testing and thereby the cost of the device itself. By incorporating error-correcting bits in the
data stream, the
individual data bits need not be tested. Then, by accessing the memory locations so as to avoid having
error correcting techniques fail due to common memory device faults, testing costs can be significantly reduced while maintaining high device yields. A
sequential access of the memory device will access
data bits in a somewhat
diagonal path across the two-dimensional array. The key to the present invention is that the
sequential data stream stored in the device is stored such that
sequential access of that data will not dwell on a single (or a small number of) row or column line. In a three or more dimensional array, the
sequential data stream stored in the device is stored such that
sequential access of that data will not dwell on a single value in a given dimension for which bulk errors are considered likely.