Best practices for revising test programs
Test programs for ATE eventually need to be updated to fix bugs, optimize test algorithms, update test limits, add or remove tests and many other reasons. For purposes of data analysis over a long period using yield management systems like yieldHUB, it is important that the test program updates ensures that the data over multiple versions of the test program can still be associated on a per-test basis reliably.
We at yieldHUB have encountered numerous practices by customers that often require additional steps at data processing to ensure that the users can still analyze the data conveniently. Here are some of the best practices that we have observed and recommend that you do it for your test programs:
1. Keep a version number in the test program name and use consistent names. For example: MFGYH2218_FT_01 and when revised this becomes MFGYH2218_FT_02. In yieldHUB, programs are unique based on the whole length of the program name. This way, it is very easy to distinguish between program versions. It is also very easy to put together multiple versions of a program for analysis since you can search for the test program name without the version number.
2. Ensure that the test numbers are unique. Every time a test program is revised, this should be a routine check, especially if new tests are added. Unique test numbers allow for automated handling of retesting of tests within a single die. It also avoids confusion on which tests is being referred to. For efficiency in database queries, test numbers may be used as index fields so it is really ideal if they are unique in a test program.
3. Keep the test number the same across multiple versions of the test program. This is very important for long term data analysis. Using test numbers for reference, indexing and connecting tests across test program revisions is very efficient on the database.
4. Keep the test names unique and the same across multiple test program versions. Although not as critical as keeping the test numbers unique, to avoid confusion among the users, especially the new ones, test names should be kept unique. It is also possible to use test names as the basis for generating unique test numbers using algorithms like MD5 or CRC16. This is especially true for certain testers that can only generate a series of numbers as test numbers and the test developer has no control on which test number gets assigned.
5. Keep the sequence of tests the same as much as possible. This helps the test list look the same across versions. Of course, if tests are added or removed, the sequence will differ but it is ideal to keep the sequence the same as much as possible.
Some of our customers who were unfortunate to have test systems that make it impossible to control the test numbers needed to pay for custom work to assign consistent and unique test numbers to their data.
Another customer does not add revision numbers on its test program name and the result is multiple support tickets from confused users when they see unexpected data or do not see the expected data. This increased the number of hours of support they need to spend.
These best practices will help you keep your YMS ready to analyze your data even across several months of data. It helps reduce the cost of getting your data online on a YMS and help reduce the cost of maintaining your data.