Microstrip circuits are typically used to combine devices and circuits to improve the overall performance of a machine. They offer advantages of high reliability and integration; however, their cumbersome and complex processing process can easily cause breakage, bending, branching, and other inhomogeneity problems, thus causing adverse equipment operations. The classical microstrip-circuit defect-detection methods involve complex image post-processing, difficult to distinguish high-density alignments, long test cycles, and other shortcomings. Hence, a more efficient and convenient detection technology that can localize fault defects is required.

Time-domain reflection is a technique that involves injecting a pulse signal into a test sample, and the impedance-change position in the line will reflect a portion of the pulse signal. The type and location of impedance change can be determined based on the peak characteristics of the reflected signal. The classical pulse time-domain reflection technique features large signal jitters and low resolutions, whereas the rise time of terahertz pulses is on the order of picoseconds and the signal jitter is on the order of femtoseconds. Applying terahertz pulses in the time-domain reflection detection of defects in microstrip circuits allows one to detect defects with high resolution and accurate localization. In this study, different lengths of microstrip wires, different numbers of branching wires, and “back” bent wires were designed to simulate three different types of inhomogeneous structures typically observed in broken wires, T-branches, and continuous right-angle bending microstrip circuits, respectively. First, the terahertz pulse time-domain reflection signals of the three different types of inhomogeneous wires were obtained using the ADVANTEST TS9001TDR system, and the distance error between the inhomogeneous position calculated based on the pulse signals and the actual position was analyzed. Subsequently, simulation software was used to establish a pulse time-domain reflection model for the inhomogeneous microstrip wires. Finally, the theory of equivalent-circuit models was utilized to analyze equivalent circuits corresponding to the different types of inhomogeneous structures, based on which the occurrence mechanism of pulse reflection-signal characteristics was explained.

For different lengths of open wires, the upward positive pulse signal is reflected back to the location of the open-circuit fault. When the wire length is relatively short, the pulse signal is reflected multiple times between the test point and the fault location and gradually attenuates to 0. Meanwhile, the second and third reflection signals, which are two, four, and eight times the length of the wire of the first pulse time-domain reflection signal, coincide with the first reflection signal (Fig. 6). When the length of the wire is 50 or 100 mm, at the branch position of the wire, reflected signal with a peak downward will be generated. When a terahertz pulse signal is transmitted to the end of the wire, it is reflected back as an upward positive peak. Additional pulse signals are reflected when multiple branches exist; however, the fewer the number of branches, the more significant are the corresponding pulse time-domain reflection signals (Figs. 7 and 8). When multiple consecutive bends exist in the wire, negative pulse signals are reflected back at the bend locations, and the farther the bend location is from the test point, the weaker is the reflected signal (Fig. 9). Experimental results show that for the three different types of inhomogeneities, the greater the number of inhomogeneous locations on the wire and the longer the pulse transmission distance, the greater is the error (Fig. 10a). This is attributable to the experimental operation and the low precision of sample processing. The peak change states of the reflected signals in the time domain obtained from the simulation of the three different types of microstrip-circuit inhomogeneity models are consistent with the experimental results. Theoretical analysis shows that when a bend or T-branch occurs in the microstrip line, charges accumulate therein, which is equivalent to an increase in the capacitance, i.e., a decrease in the characteristic impedance, thus causing the reflection coefficient to be less than 0. Meanwhile, the reflection signal is shown as a downward reflection signal in the time domain, i.e., negative peaks at the location of impedance change, which is consistent with the experimental and simulation results.

The results show that the minimum errors between the impedance-change location calculated from the pulse correspondence time and the actual inhomogeneity-occurrence location are 1.2%, 0.2%, and 1.4%, and that a distance error of 10 μm can be detected. The terahertz pulse time-domain reflection technique can discriminate the type of microstrip-line inhomogeneity and locate it accurately; however, the detection effect for multiple inhomogeneities on the same microstrip line must be further optimized. Microstrip circuits, as a key device in radio equipment and modern integrated systems, contribute significantly to the rapid detection of microstrip-circuit wire inhomogeneity, thus providing a foundation for the research of more efficient detection of defects inside packaged chips.