Qualifying a New Process or Product 


Semiconductor manufacturing involves processes that need a high degree of control.  As such, every significant change in these processes, or the introduction of any new process, must undergo proper qualification prior to its implementation.  This will ensure that the affected process flow will still be capable of consistently meeting the specifications of the form, fit, and function of the final products, even after the change.


Different process changes will have different effects on the process, so there can be no single, standard procedure for qualifying a change. The qualification procedure defined for a process change must therefore be custom-made for the change being qualified.  Furthermore, the qualification of a change usually requires that no degradation in the quality and reliability of the final products will be introduced, even if the change is being implemented to serve another purpose, e.g., productivity or cycle time.  Thus, in the context of this discussion, process qualification refers to the process of ensuring that the change will not result in any degradation in the quality and reliability of the final products.


Most semiconductor companies follow a basic flow when defining the qualification plan for a process change or a new process, namely:


1)  identify potential failure modes and failure mechanisms that the process change may bring about;

2)  subject samples to the appropriate reliability stresses to accelerate these potential failure mechanisms;

3)  test the samples to determine if they are still acceptable after completing the reliability stresses.  


This basic flow works on the premise that the process change may be qualified for implementation if the final products will not exhibit any failures associated with the change. 


The key to a good qualification plan is the excellent anticipation or prediction of what potential failure mechanisms the process change will trigger.  Once the potential failure mechanisms have been defined, the selection of the reliability tests will follow, based on the accelerating factors pertaining to the mechanisms. 


For instance, consider the process change wherein a new molding compound will be qualified.  What potential failure mechanisms would a change in molding compound bring about?  These would include:  

package cracking, because the molding compound may turn out to have thermomechanical characteristics that are less compatible with the intended package and application;

- die cracking, because the molding compound may turn out to be too stressful on the die;

- bond lifting/wire breaking, because the molding compound may turn out to be too stressful on the wires and bonds;

- die/wire/leadframe corrosion, because the molding compound may contain corrosive elements.


Given these failure mechanisms, the reliability engineer would, more or less, select the following reliability tests to accelerate these mechanisms: solder heat resistance test, temperature cycle or thermal shock test, and temperature humidity bias (THB) test.  The new molding compound may be considered as qualified only if the final products using this compound pass all the reliability tests conducted.


Reliability testing is statistical in nature, so the sampling scheme used for process qualification also has a significant bearing on the success of the qualification.  In theory, the greater the sample size used in reliability testing, the higher the confidence one can have in the reliability data generated.  Unfortunately, more samples mean higher costs, so the reliability engineer needs to strike a balance between extremes when designing his qualification plan. A good guideline is to base the sample size on the quality objectives of the company using industry-standard sampling methods such as LTPD and AQL sampling.


The sampling scheme should take into consideration process interactions and variabilities, as well as the failure rates acceptable to customers, versus the cost of samples and the cost of performing the reliability tests.  Getting qualification samples from several vehicle devices and several production lots is normal to ensure that the reliability assessment covers the extent of the variability of the process and products.


Failure analysis is an integral part of any process qualification.  Failures that are not related to the process change may be encountered during the qualification, so these must be differentiated from valid qualification failures.  Invalid failures or failures not related to the qualification should not affect the results of the qualification and, subsequently, the management decision on whether the process change must be implemented or not.  


However, valid qualification failures must be taken seriously, no matter how few, because it would be very expensive in the long run to implement a process change that is doomed to fail from the very start.


See also Reliability Engineering Reliability Modeling; Life Distributions; Failure Analysis; LTPD and AQL Sampling




Copyright 2001-2004 EESemi.com. All Rights Reserved.