A new image-based diagnostic method is proposed for an E-beam tester. In this method, a static fault imaging (SFI) technique and a navigation map for fault tracing are introduced. The SFI technique, in which a dc E-beam is used, enables fast acquisition of images without any additional hardware. Then, guided by the navigation map derived from CAD data, marginal timing faults can be easily pinpointed. The proposed method is applied to a 19k-gate CMOS logic LSI circuit and a marginal timing fault is located in about 1/10 the turn around time (TAT) of the conventional waveform-based method. o d u c a E-beam testers are indispensable for diagnosing marginal timing faults, in which a device under test (DUT) passes or fails conditionally.However, the conventional waveform-based approach to E-beam diagnosis is very inefficient, because many intemal wires in the DUT have to be probed one by one with reference to a netlist[ I] [2] to observe the waveforms. As DUTs increase in scale and complexity, the probing count becomes enormous in the waveform-based method.An image-based method called dynamic fault imaging [3][4] has been proposed as a substitute for the waveform-based method. Since many intemal nets in the DUT can be observed at a glance with this technique, a faulty area can be quickly localized with very little diagnostic effort [4]. So far, however, no practical implementation of the image-based method has been realized, because (a) there has been no established guideline for the image-based fault tracing, and (b) it takes considerable time to acquire dynamic fault images by a stroboscopic technique. To improve the stroboscopic image acquisition rate, a hardware technique called burst mode imaging has been proposed[S]. However, this technique requires expensive additional hardware.Here we propose a viable practical approach involving two innovations: static fault imaging using a dc E-beam, and a CAD-data-based navigation map for fault tracing. This paper describes the diagnostic procedure using these two techniques, a trial application to a 19k-gate LSI circuit, and a TAT evaluation of the trial.
Timing fault obse rvation method
Conventional fault imagingIn the event of a marginal timing fault, the DUT either conditionally passes or fails in response to changing the clock rate. The fault image is the difference between the two E-beam images acquired separately under both passing and failing conditions so that only the faulty-state wires can be imaged. In conventional fault imaging (dynamic fault imaging), the E-beam image is acquired by a stroboscopic technique, such as shown in Fig. l(a).Since the E-beam is pulsed once for every test pattem sequence, the average E-beam current becomes so small that prolonged observation time is necessary for S/N improvement. With this approach, the image observation time (Td) can be estimated with the following equation:Tim: image observation time of the dc ETc,Tp: clock rate and pulse width (Tc = Tp) Np: test pattem number beam The observation time (Td) is proportional to Np. Th...