Heavy ion single event effect (SEE) testing was performed at Brookhaven National Laboratories' twin Tandem van de Graaff accelerator on May 13-16, 1997 on the following 80486 microprocessors:
|MFR||Device||Date Code||Other Markings||Process|
During testing, the device under test (DUT) is exercised by cycling through several software routines (System test, paging, co-processor, external memory access, interrupts, and software performance). Clock speed is 25 MHz (50 MHz internal clock) and Vcc was 4.75V. Devices were tested both with and without internal cache. Three samples were tested.
Outputs were compared with a reference device, with miscompares in the data bus, address bus, or control lines counted as single event upsets (SEUs). Failure of the device to write to a test address was detected by a watchdog timer and counted as a "lockup" SEU, prompting a soft reset (power is not cycled). DUT current was monitored for both traditional single event latchup (SEL) and microlatch conditions.
Traditional SEL occurred when device current consumption (Icc) increased above the maximum current specified for the device, which was 1.2A for these devices. During microlatch, Icc may increase above the normal operating level, but not above the maximum specified for the device. Device operation halts, and a power reset is required to recover; during testing, DUTs were reset after ~1 second at the higher current, to verify that the condition was a microlatch, and not just a temporary spike.
Ions used for testing were:
|Ion||Energy in MeV||Linear Energy Transfer (LET) in MeV*cm2/mg at 0°|
All test runs were performed at either 0 or 38 from normal incidence of beam to the device, to vary the effective LET. Devices were tested at room temperature, ~ 25 C.
Fluxes: 2.5E3 to 2E4 particles/cm2/sec
Fluences: 1E6 particles/cm2
This test synopsis is not intended to be a detailed report, just a quick-look at the real-time data. All LETs are in MeV*cm2/mg.
SEU results were fairly similar to previous 80486 testing. Threshold for both miscompare and lockup SEUs was between 4.29 - 7.88 both with and without cache. Figure 1 shows SEU data.
Although none of these devices displayed typical SEL (up to a maximum tested LET of 37.3), unlike previous 80486 testing, all three did experience destructive conditions. Figures 2, 3, and 4 display DUT current consumption over the course of a test run, with samples taken ~ every 4 msec. Please note that at no time did the current reach the SEL limit of 1.2A. Figure 2 shows a typical test run where no microlatch was observed, for reference. The two spikes down to 0mA are board resets triggered by SEUs, and the typical current range is ~ 490-570mA. Figure 3 shows a test run where 9 microlatches occurred; the device was still functional following this test run; however, it failed following a microlatch on the next run. Figure 4 shows a test run where 6 microlatches occurred. The device was nonfunctional following the test run: it would lock up shortly after attempting to begin normal operations. Both of these test runs were at LET 37.3; a third device experienced a destructive event during its first run, at LET 26.6.
It is unknown if these destructive conditions were caused by the microlatches, or by other heavy ion-induced events such as gate or dielectric rupture. Analysis of the devices is ongoing.
Because these destructive conditions occurred at relatively high LETs of 26.6 and 37.3, they are expected to occur in the HST orbit ~ 1 every 3E11 years. While this is not a great concern for HST, more radiation-intensive orbits would require microlatch detection schemes and possibly redundant devices.
Please note the lot/date codes and other markings for these devices, along with other lots previously tested, which did not see the same destructive events:
|Tested||Lot/Date Code||Other Markings||Destructive Events|
Last Revised: February 03, 2010
Digital Engineering Institute
Web Grunt: Richard Katz