The hardware defect of integrated circuit chip usually refers to the physical imperfection of the chip. Integrated circuit fault refers to the logic function error or abnormal operation of the circuit caused by the defect of the integrated circuit. The common factors leading to the failure of integrated circuit chip include the rapid decline of performance caused by the change of component parameters, the poor contact of components, the failure of signal line, the failure of equipment due to poor working environment, and so on. Circuit faults can be divided into hard faults and soft faults. The soft fault is temporary and will not cause permanent damage to the chip circuit. It usually appears randomly, causing the chip to work normally and abnormal. When dealing with such faults, the equipment can be restored to normal only by reconfiguring the system with the same configuration parameters when the fault occurs. The damage to the circuit caused by hard fault is permanent and can not be recovered without maintenance.
Generally, there are three necessary modules for IC chip fault detection: source excitation module, observation information acquisition module and detection module. The source excitation module is used to transmit the test vector to the integrated circuit chip to drive the chip into various working modes. Therefore, we usually hope that the test vector set can contain all possible input vectors as much as possible. The observation information acquisition module is responsible for collecting the information used for analysis and processing. The selection of observation information is very important for fault detection. It should contain as much fault characteristic information as possible and be easy to collect. The detection module is responsible for analyzing and processing the collected observation information, identifying the fault characteristics hidden in the observation information, so as to diagnose the mode of circuit fault.
The earliest circuit fault diagnosis methods mainly rely on some simple tools for test diagnosis, which greatly depends on the theoretical knowledge and experience of experts or technicians. Among these test methods, the most commonly used are four categories: virtual test, functional test, structural test and defect fault test. Virtual testing does not need to detect the actual chip, but only the simulated chip, which is suitable for testing before chip manufacturing. It can detect the faults in the chip design in time, but it does not consider the noise or difference in the actual manufacturing and operation of the chip. The function test determines whether the chip has fault according to whether the chip can complete the expected function in the test. This method is easy to implement, but it can not detect faults with non functional effects. Structural testing is an improvement of built-in testing. It combines scanning technology and is mostly used for fault detection of produced chips. The defect fault test is based on the chip actually produced. It can find out whether the fault is included by checking the production process quality of the chip. Defect fault testing requires high knowledge and experience of professional technicians. Chip manufacturers usually combine these four test technologies to ensure the reliability and safety of the whole process from design to production and then to application. However, for the increasingly complex circuit system, these early methods are more and more stretched. After continuous improvement and innovation, many new ideas and methods have come out one after another.
Voltage diagnosis appears earlier and is widely used. The observation information of voltage test is the logic output value of the circuit under test. This method obtains the logic output value of the corresponding circuit by inputting different test vectors into the circuit, and then compares the collected circuit logic output value with the expected logic output value of the circuit corresponding to the input vector, so as to detect whether the circuit can realize the expected logic function in the actual operation environment. This method is simple, but it is not suitable for large-scale integrated circuits with more redundancy. If the defect appears in the redundant part, it cannot be detected. Moreover, when the circuit scale is large, the test vector set will also grow exponentially, which will directly lead to problems such as difficult test vector generation and low diagnosis efficiency. In addition, if the fault only affects the circuit performance rather than the circuit logic function, the voltage diagnosis can not be detected. Because the voltage logic value output by the integrated circuit is not necessarily related to all nodes in the circuit, the voltage test can not detect the non functional failure of the integrated circuit. Therefore, in the early 1980s, the diagnosis technology based on the power supply current of integrated circuits was proposed. The power supply current is usually directly or indirectly related to all nodes in the circuit, so the current based diagnosis method can cover more circuit faults. However, the current diagnosis technology is not proposed to replace the voltage test, but to supplement it in order to improve the detection rate and coverage of fault diagnosis. Current diagnosis technology is divided into static current diagnosis and dynamic current diagnosis. The core of quiescent current diagnosis technology is to compare the power supply current of the circuit to be tested under stable operation with the preset threshold to determine whether there is a fault in the circuit to be tested. It can be seen that the selection of threshold is the key to determine the detection rate of this method. The early quiescent current diagnosis technology used a fixed threshold, but the fixed threshold can not adapt to the development of integrated circuit chips to deep submicron. Therefore, later generations have made continuous improvements in quiescent current detection methods, and successively proposed differential quiescent current detection technology, current ratio diagnosis method, quiescent current detection technology based on clustering technology and so on. Dynamic current diagnosis technology came out in the 1990s. The dynamic current can directly reflect the switching frequency of the internal voltage of the circuit during the state transition. The detection technology based on dynamic current can detect the faults that cannot be detected by the previous two methods, and further expand the fault coverage. With the development trend of integrated circuit technology, intelligent fault detection technology is gradually moving forward