Boundary Scan Meets Functional Test

Estimated reading time: 3 minutes

One of the most serious problems in testing is the continuous decline in physical access. Across a range of markets, users unanimously report slumps of up to 20% within the last five years, which inevitably leads to discussions on test methods that can be used in future. JTAG boundary scan has gained a good reputation in this context, but this process alone cannot provide a solution to all problems either, and partners are needed. The main focus of interest is the functional test. This report analyses some fundamental aspects of this combination and introduces new technological solution approaches for embedded functional tests, as well as their practical implementation.

The Interplay Between Structure and Function

No matter which of the common electrical testing strategies are examined, they all fall into two basic categories—structure test and functional test. According to Table 1, both adhere to completely different principles to trigger and diagnose a fault. While the structure test attempts to test the design through access on the element level, the functional test aims to test connected components.

Table 1:  Characterization of fundamental test principles

However, the success of the structure test is based on securing the required access to the elements, and techniques such as ICT, MDA or FPT are becoming less and less useful in this context. The functional test on the other hand typically uses the native connections of a Unit Under Test (UUT) as the basis for access, thereby circumventing this problem.

As a result, it is no surprise that users are again increasingly focusing on this technology. The so-called cluster test is a subgroup of the functional test. Cluster tests are basically the result of insufficient access, as they include the testing of a partial circuit via the test points that are available at the edge of the partition. The logical consequence of diminishing test access is ever increasing cluster size, as far as the complete loss of meaningful clustering.

The question if the structure test makes the functional test redundant or vice versa, is also a frequent, hotly debated topic. The main reason behind this is the objective to test as little as possible due to costs, and where testing cannot be avoided, it would be beneficial to at least agree on one primary method.

Several fundamental factors play a role in answering this question. These are discussed in the following.

A structure test can prove that the integrity of the circuit is present on an electronic physical level. The tests are typically of a mainly static nature, partially at-speed. It can also provide precise fault diagnostics in this context. It does not, however, test any functions and is therefore not able to detect a malfunction of the circuit in operating mode. This limits the fault coverage, when measured against the entire range of possible faults. Thus, structure tests are ideal as indirect process sensors, but by no means replace the functional test.

The functional test on the other hand provides high fault coverage in principle, and can also detect anomalies in operating mode, provided it is carried out with the corresponding dynamic. However, the test patterns used also exert a very strong influence on the outcome, as faults must be stimulated in order to be detected. The test can also detect structural faults if they effect the function, but the fault diagnoses are very unclear. If structural faults are not stimulated by functions, this generally leads to fault slippage with potentially serious consequences at a later stage. This also affects parametric faults on individual components, open pins with floating signal level, missing pull resistors or incorrectly mounted decoupling capacitors, to list but a few of the problems. For this reason, functional tests in production are particularly suitable for providing basic proof that the selected operation modes are fulfilled, such as those carried out in the end-of-line (EoL) test. However, they are by no means a replacement for the structure test.

To read this entire article, which appeared in the December 2016 issue of SMT Magazine, click here.