Summary of the Data Retention
As time goes by, the Data stored in flash media will suffer from Retention errors. As mentioned in the previous article “Analysis on the Root cause of Flash Error”, Data Retention errors are mainly caused by two reasons:
1. One of the reasons for the error of Data Retention is that FG's time-dependent dielectric Breakdown (TDDB) leads to the increase in low-field leakage rheology, and the increase of leakage current leads to the weakening of the Cell's Retention threshold voltage transferability. Thus, errors in Data Retention are generated. 2. Another reason for Data Retention is that Erase and Program also cause the oxide to collect charge, which affects the threshold voltage of the cell. When the charge is out of the well, the threshold drifts and Bit reverses.
Data Retention errors are closely correlated with Program/Erase times. The more P/E cycles, the higher the Data Retention error rate. Meanwhile, Data Retention is directly correlated with ECC strength. Even if bit flipping occurs, some errors can still be corrected by the error correction algorithm. ECC of different intensities has an extremely important influence on Data Retention.
Therefore, the evaluation of Data Retention must be based on the conditions of the P/E cycle and ECC, and it is not rigorous to answer the Data Retention of flash memory under a single condition. Strictly speaking, how many Program/Erase sessions and how long data can be stored with what kind of ECC.
Factors affecting Data Retention include: 1. Number of P/E cycles; 2. ECC strength; 3. Temperature environment; 4. Others (such as radiation, light, etc.).
If the special environment in Article 4 is not considered, the evaluation of Data Retention needs to take into account the other three conditions. Among them, a high-temperature environment can also be used to accelerate the evaluation of Data Retention.
For the testing department, it is far from enough to only evaluate the data storage without P/E status. It is also necessary to evaluate the storage time under different P/E times so that the R&D personnel can correctly choose the ECC strength and objectively evaluate the storage time of data.
It is necessary to divide a certain number of P/E cycles into different intervals, and establish a mathematical model based on ECC, P/E times, and bit error rate based on the test results, so as to comprehensively evaluate the Data Retention of Nand. For example, MLC can be evaluated at 500 times /1000 times /1500 times /2000 times /2500 times /2800 times /3000 times Program/Erase as well as at ECC strength specified and higher than ECC strength specified Often has more practical application significance.
The closer Nand Flash is to the critical value of the P/E Cycle, the greater the influence of P/E on its Data Retention will be each time. Therefore, when approaching the critical value, the P/E Cycle interval can be narrowed, for example, Data Retention will be measured every 200 P/E cycles.
According to JEDEC standards, Data Retention of Nand Flash can be tested under high-temperature ambient acceleration conditions.
Reread the data previously written to Nand (important! Don’t have to write original data comparing because may occur in the process of writing data errors, if compared with write data directly, maybe the wrong conclusion), and then save as a separate file, over time, or temperature changes will Nand writes data compared with separate files, to see if there is an error occurs, This testing can be accelerated under high-temperature conditions.
Basic Setting: ECC1 is set according to the specification (e.g., 24bit/1024Byte), and the other three groups can be evaluated with higher strength ECC in which ECC range errors will occur.
For the new Nand Flash, there is no influence on the P/E cycle. Therefore, parameters can be set and measured directly.
① Set the number of blocks to be tested, the number of times to be read, and the pattern of the Block to be written (for example, a set of pseudo-random numbers seeded 456789AB);
② At room temperature, the tool will Erase 456789AB as the seed of the pseudo-random Program to the specified Block, then read the data, save the data file, and create a file for comparison.
③ Set up a result file, all comparison results will be automatically saved to the result file;
④ Then set a time, such as 1000 hours, to start the verification of Data Retention. If any error occurs, the Data in Nand will be different from the specified comparison file.
⑤ Errors under various ECC conditions will be shown in graphs and data as shown below.
⑥ Put the Nand Flash written into the pseudo-random number into the 120°C high-temperature boxes for 34 minutes and 13 seconds and then take it out to see if any error occurs. If no error occurs, the data can be saved for one year. Cycle the test in turn until there is an error.
The P/E cycle test of Nand Flash should be added to the P/E cycle test of Nand Flash.
After each P/E cycle (as shown in the figure above, 500 P/E cycles is one cycle), repeated actions 1-6 can be used to evaluate Data Retention at different P/E times and under different ECC conditions.