A Novel Test Flow for One-Time-Programming Applications of NROM Technology

An Efficient Test and Repair Flow for Yield Enhancement of One-Time-Programming NROM-Based ROMs

“…The margin-read-1 (0) test is used to test a memory cell to see if it can store the quality value of 1 (0). The detail steps of the conventional test flow can be found in [13].…”

“…Therefore, most yield loss occurs in the wafer test phase rather than the package test phase. To cure this dilemma, we will use the observation that memory bits failing the margin-read-1 test are still able to store the value 0 [13], [14]. Therefore, the fault replacement analysis is moved on to the package test phase as shown in Fig.…”

“…That is, they fail the margin-read-1 test (cannot store value 1). Fortunately, a NROM cell which cannot correctly store the bit value 1 still has very high probability to correctly store binary value 0 [13], [14]. If the data word to be written into a faulty memory word containing a single defect bit and the fault effect cannot be masked, then we can complement the data word and then program it into the faulty memory word.…”

“…In this table, we assume that the percentages of faulty cell, faulty row, and faulty column are set to 96%, 2%, and 2%, respectively. This setting is based on the statistics of foundry data presented in [14]. The row "conventional" denotes that merely the fault replacement technique is used.…”

“…According to the statistics from the memory foundry, the single defective bits occupy more Manuscript than 50% of all the fault types. Moreover, it has 96.4% of probability that the number of failed addresses in a memory row is one [13]. According to these facts, besides the traditional fault replacement techniques, we also use the fault masking techniques to further increase the fabrication yield.…”