Information in this section describes general characteristics of the SET response characteristics of the device, and may not be accurate for all use cases or conditions. In-circuit results vary according to application specifics. TI’s customers are responsible for determination of components for their purposes, and validating and testing design implementation to confirm system functionality.

The data suggest the rate at which the OPA4H014-SEP exhibits SET events, and the magnitude of those events, is a function of several factors. These include supply voltage, output loading, beam flux, ion energy, and temperature. Signal voltages, gain, and noise bandwidth are theorized as potential secondary contributors. No op-amp channel was observed to be significantly "better" or "worse", in terms of transient count, than any other. Differences in event count channel-to-channel were largely caused by differences in the oscilloscope cards used, as verified through A↔B site swaps. The output exhibited a higher tendency to shift towards the high supply rail than to the low supply rail.

Generally, when the OPA4H014-SEP experiences an SET, the output typically shifts (but not always) towards the high supply rail. These events appear as sudden spikes and are usually resolved within 10µs of the trigger event. Events where the output shifted by more than ±100mV were recorded by the oscilloscope cards. A small percentage of these captures show measurable undershoot or overshoot behavior after the initial spike as the output settles. Some instances of the output ringing (whether due to noise or to an SET) were also recorded. Appendix A and Appendix B show notable oscilloscope captures.

Note that the OPA4H014-SEP also experiences transient events of less than 100mV. These most likely dominate event counts, but are more difficult to measure accurately. As a result, this study focuses on only events more than 100mV in magnitude. Testing at MSU and TAMU has shown that the beam area is an electrically noisy environment, which can lead to false trigger events. Implementing filters on the device to reject noise can lead to reductions in SET count or impact the magnitude of those events. As a result, the device was tested with the full noise gain-bandwidth to serve as a worst-case analysis.

Half of the devices evaluated were tested with ion energy in descending order, from 45.8MeV-cm² / mg to 19.3MeV-cm² / mg. The other devices were tested with ion energy in largely ascending order, from 2.68MeV-cm² / mg to 18.9MeV-cm² /mg. The event counts per device as the two groups approach the same nominal LETeff differs significantly. Two possible mechanisms (not mutually exclusive) are proposed to explain this.

The first hypothesis attributes this change to differences in ion flux per run. While all runs continued until the same nominal fluence of 1 × 10⁷ ions / cm², some runs used higher or lower flux rates than the nominal 1 × 10⁵ ions/s-cm². A correlation between high flux rates and a high event count per device, per run can be inferred from the data. Due to facility limitations, the second group of devices were mostly tested around 2 × 10⁴ ions / s-cm² for the copper ion. In comparison, the first group of devices saw a flux of 3.2-5.2 × 10⁴ ions / s-cm², correlating with higher event counts. This hypothesis implies that as flux rate increases, the susceptibility of the device to SETs also increases.

The second hypothesis attributes the change to cumulative total ionizing dose (TID) damage. The OPA4H014-SEP is specified to 30krad(Si), and characterized to 50krad(Si), for high-dose-rate (HDR) TID. However, the devices used in this testing were exposed to far higher accumulated dosages. By the conclusion of testing DUT 3 experienced approximately 100krad(Si), and DUTs 3 and 5 experienced 117krad(Si). The devices were still functional despite the high stress, but showed higher event counts for copper ions than the subset of devices tested in ascending order. Those devices experienced approximately 5krad(Si) and experienced relatively fewer transient events. A possible link between accumulated TID damage and SET susceptibility has not been previously explored. This hypothesis implies that as more TID damage is accumulated over an operational lifetime of a device, the susceptibility to SETs increases.

Note that other factors such as the time between decap and testing (time the die is exposed to air), annealing time between runs, and simple device-to-device variation can also play a potential role in the differing event counts. Isolating any single factor is difficult due to the complexities and practical challenges of the testing.