Altitude and underground real-time SER tests of embedded SRAM

Real-time SER tests have been performed on embedded SRAMs in a 0.13-μm process technology. 511 samples of a product IC have been tested at two locations. First, a “mountain test” was performed, at an altitude of 3450 m, followed by a “cave test” at an underground location. The resulting real-time SER data have been combined to determine the neutron-SER and alpha-SER contributions under nominal conditions. The resulting neutron-SER is in very good agreement with data from accelerated SER tests. Also, the real-time alpha-SER agrees within 90% confidence limits with the accelerated SER result. However, the deviation in the case of alpha-SER is larger than for neutron-SER, which can be attributed to uncertainties in the alpha-particle flux observed by the product die. Radiation-induced soft errors are generally assumed to be random external events, resulting in a constant upset rate. From a statistical point of view, this corresponds to an exponential distribution. In the present work, statistical analysis is applied to the measured real-time SER data. It is shown that the exponential distribution indeed fits the mountain-SER data very well. In contrast, the cave-SER data are fitted rather poorly by the exponential distribution.