Heavy-Ion Testing for Single-Event Effects

Susan Crain and Rocky Koga

The most reliable way to reproduce the space-particle environment on Earth is with a particle accelerator such as a cyclotron. Aerospace has conducted numerous tests measuring the susceptibility of microelectronic devices to single-event effects.

The liftoff of the Atlas Centaur launch vehicle seemed picture-perfect: The rocket completed its ascent and successfully deployed its payload to its intended orbit. What was not immediately apparent was that some bits in the computer memory were altered as the vehicle flew through a region of space dense with energetic protons. In this case, the errors were automatically detected and corrected by the computer—but could the launch team always count on such good fortune?

Events such as this have led to the realization that spaceborne microcircuits are vulnerable to galactic cosmic rays and trapped protons. Since the discovery of so-called "single-event upsets" in 1975, scientists have sought to characterize the space-radiation environment in greater detail and understand its interactions with microelectronics.

Test boards often accommodate several devices for testing, eliminating the need to vent the chamber to change the parts. The control software for the motion system logs the unique spatial information for each part so it is always centered in the beam line even when it is angled to achieve effective LETs.

Ideally, the study of space-radiation effects should be conducted in a manner that approximates, as closely as possible, the space-radiation environment. The most reliable test would use all of the same ion types that are found in space and allow measurement over a wide energy range for each. But such a test would be prohibitively expensive. A more practical approach is to use a medium-energy particle accelerator to simulate galactic cosmic rays and trapped protons in space-radiation environments.

The ability of an ionized particle to interact with materials is a function of its linear energy transfer (LET) value. LET is essentially the measure of ionizing energy deposited in a material per distance traveled, generally rendered in millions of electron volts per square centimeter per milligram (MeV-cm²/mg). For particles in space, the range of LET varies primarily from a few hundredths to just under 100 MeV-cm²/mg. Particles with low LET values are far more abundant than particles with high LET. Thus, in investigating a particular device, researchers seek to find the threshold value and to determine the magnitude of sensitivity at large LET values. Such an investigation requires an accelerator capable of generating many particles with different LET values.

The Facility

The choice of accelerator is based on its capability to produce ions with a reasonable particle range for a wide range of LET values. Other factors include the ease of use and cost of operation. Aerospace has traditionally used the 88-inch cyclotron at Lawrence Berkeley National Laboratory.

This cyclotron routinely and reliably accelerates ion species as light as protons and as heavy as gold. To achieve high energy without losing high intensity, it employs a sector-focused design. A process known as electron cyclotron resonance is used to generate the ion source; the ions are then injected into the cyclotron for acceleration. This technique allows continuous operation of the cyclotron for up to several weeks. Also important, it allows researchers to modify the ion intensity with the push of a button.

Ion	Energy (MeV)	LET (MeV-cm²/mg)	Range in silicon (microns)
¹¹B⁺³	108.2	0.89	323
¹⁸O⁺⁵	183.5	2.19	228
²²Ne⁺⁶	216.3	3.44	179
⁴⁰Ar⁺¹¹	400	9.88	129
⁵¹V⁺¹⁴	508.3	14.8	116
⁶⁵Cu⁺¹⁸	659.2	21.6	108
⁷³Ge⁺²⁰	724.7	25.37	104
⁸⁶Kr⁺²⁴	886	30.0	111
⁹⁸Mo⁺²⁷	983.6	38	102
¹³⁶Xe⁺³⁷	1330	53.7	104
¹³⁶Xe⁺³⁸	1403.4	53.6	110
Ten-MeV-per-nucleon particles are used more frequently with parts that cannot be easily delidded. Often, parts such as DRAMs need to be lapped from the back side of the die to avoid the lead frame, so the beam needs to have a greater range to pass through the sensitive regions. Berkeley is developing still more penetrating cocktails of ions.

The Berkeley cyclotron can produce several ion species of various LET values. A typical test run might use a half dozen different ion types ranging in mass from boron to xenon, each capable of penetrating to different depths within the target device. The ions can be switched in a matter of seconds, making single-event effects testing highly efficient.

The beam diameter is about 7.6 centimeters, within which the target position is determined by a laser targeting system. The beam may be directed to a small section of a microcircuit or to a large detector. The ion flux range is between a few particles to a few hundred thousand particles per square centimeter per second. A low flux is used for sensitive devices, and a high flux is used to check for rare events. A surface-barrier detector for energy measurement and a position-sensitive detector serve to identify ion species, energy, and uniformity. A diagnostic/dosimetry apparatus verifies that the beam is suitable for the type of testing being performed.

The irradiation chamber measures 96.5 X 99 X 116.8 centimeters. Vacuum is controlled by a high-capacity system of pumps capable of evacuating the chamber in about four minutes. This makes sample changes quick and easy. A mechanized, remote-controlled system moves individual test samples in and out of the beam and changes beam-exposure angles. Changing the beam-exposure angle effectively changes the charge deposition in the sensitive region of a microcircuit. Charge deposition is related to the concept of "effective LET," which is calculated by multiplying the LET of the incident ion by the secant of the angle between the incident beam and the chip-surface normal.

Test Methodology

The facility at Lawrence Berkeley National Laboratory has been used to test all kinds of devices and circuits. In the past, some electronics manufacturers maintained separate production lines for radiation-hardened devices, and the cyclotron was used to examine these parts. With the subsequent increase in commercial space systems, designers sought to use cheaper off-the-shelf devices, and the cyclotron was used to assess their potential for particular missions. More recently, the cyclotron has been used to evaluate a technique known as "radiation hardening by design," which uses specific design principles to increase the radiation resistance of components produced via standard commercial foundries.

The Aerospace single-event effects testing program has investigated both military and commercial products. Often, a commercial device will be tested to determine whether it can pass as a rad-hard product according to military specifications. Other testing efforts involve the characterization of board-level circuits for space systems using commercially available parts.

The chamber is designed to accommodate large systems as well as single boards. This instrument is using the monoenergetic particle beams to calibrate its detectors before flight.

Ground testing of devices for use in military, commercial, and research efforts is done using specially designed testers. The process involves exposing a part to a particle beam while monitoring its function. By counting the number of upsets and knowing how many particles passed through the part, investigators can calculate the likelihood that a particle strike will cause a single-event effect. Such calculations may be used to produce a set of sensitivity curves for a microcircuit type, which can in turn be used to estimate the upset rate of the microcircuits for various orbits. A microcircuit may respond differently depending on factors such as case temperature, clock speed, and cumulative total dose. In addition, the vulnerability for one microcircuit type to different types of single-event effects varies at different energy values for heavy ions and protons. These are some of the many parameters that must be carefully monitored.

In general, a device is first tested for destructive single-event effects such as latchup, burnout, and gate rupture. If the device does not display latchup, for example, or if the onset for latchup is at a high enough LET value to be tolerable for the particular mission, then the device will be tested for nondestructive effects.

When assessing risk, the designer or program manager needs to consider the single-event effect data in the context of the circuit features and the intended mission. For instance, single-event latchup can be mitigated (rendered nondestructive) by watchdog circuits that cycle power when a current limit is reached; thus, a somewhat sensitive device might be considered suitable for a given mission provided that such watchdog circuits are included. However, because microelectronics are getting more complex, with denser and larger designs, such circuits might not be feasible. Some complex device architectures divide the circuits into sections powered independently using different supply voltages. In such cases, setting an appropriate current limit becomes a challenge, and may prohibit the use of a device in a particular orbit.

For nondestructive effects in a complex microcircuit, fully characterizing a device type takes about 12 to 16 hours of beam time. If the part is vulnerable to destructive effects such as gate rupture and burnout, the testing can take even longer. In more complex devices, single-event upset sensitivity in different areas of the circuit may vary, and the effects might have different onsets with respect to effective LET. Fortunately, the single-event upset will usually have a different signature for the different circuit elements. Separation of the effects can happen, but the time it takes to characterize the device increases.

Testing Innovations

Through the years, Aerospace has developed specialized testers for characterizing a wide variety of devices. The most recent is the Aerospace Single Event Tester (ASSET), which provides a general-purpose interface for evaluating the single-event effect susceptibility of a wide variety of complex microcircuits. It employs two general test methods—the memory test and the sequence test.

In memory-test mode, the system treats the device-under-test like a typical memory with some control lines, address lines, and data lines. The tester writes a known pattern to an array of addresses while the device is not being irradiated, and then reads it back to ensure successful writing. Then, while the device is exposed to the particle beam, the tester continuously reads the memory locations and compares them to what was written. Any discrepancy in a bit location is counted as an error. The tester communicates the error, the address location, and the cycle count to the host computer. Later, this stored information is analyzed for single-bit upsets, addressing errors, multiple-bit upsets, and stuck-bit errors. The error is corrected in the device, and the test continues. The flux of the beam is kept low enough to keep the error-handling process manageable.

This graph charts the sensitivity of a 256-megabit synchronous dynamic random-access memory (SDRAM) to heavy ions at 4.5 MeV per nucleon. The testing regimen looked for single-event upsets (SEU), single-event functional interrupts (SEFI), and stuck-bit effects. SEFI included errors to a patch in the memory, errors to consecutive address locations, and errors accompanying an increase in bias current (view larger image).

The sequence-test mode is used for a broader type of test. In this case, a sequence of patterns is stored temporally while the device is undergoing a normal function, without irradiation. This recorded pattern is then compared with the device outputs during exposure to the particle beam. This is the mode used to test more complicated microelectronics such as microprocessors, digital signal processors, and field-programmable gate arrays. In this way, the device can be running specialized programs designed to exercise particular sections such as the arithmetic logic unit or the cache, or application-specific programs. The tester can monitor up to 512 signals at 7 megahertz and can analyze patterns up to 64,000 words deep. The test protocol is set in firmware on ASSET under control from a host computer interfaced via an in-house specialized parallel bus protocol.

Recently, dynamic random-access memories (DRAMs) have attracted the attention of space system designers because of their high storage capacity. These and other complex devices such as synchronous DRAMs, microprocessors, digital-signal processors, flash memories, and even analog-to-digital and digital-to-analog converters have control registers that can be upset by radiation. Error detection and correction schemes can help mitigate single-event upset in memories, but if the control circuits experience single-event upset, then the function of the device can be completely impaired. The designer using a DRAM for a space application has many choices about how to implement operational modes such as "idle" and "refresh." The selected implementation can affect the device's radiation sensitivity. Aerospace has used ASSET to evaluate such devices. For one program, Aerospace exhaustively tested synchronous DRAMs not just from a number of manufacturers for comparison, but also in many different configurations, identifying the most robust scheme for writing, reading, and refreshing. Based on this data, the customer was able to redesign the control circuit and successfully implement a high-capacity memory design.

ASSET can distribute four different power supplies to as many as 32 different devices at a time. The supplies are floating with respect to ground, allowing for inverted (negative) voltages as required. ASSET can also power thermoelectric coolers or heaters and monitor temperature. A cold plate is available for cooling of the test devices such as emitter-coupled logic parts or high-power devices like power converters. The devices to be tested are built onto daughter cards with a standard interface to the test head. This allows many different devices with varying power, control, and interface requirements to be tested with the same basic system. The entire apparatus was designed to fit inside a vacuum chamber with the devices under test, enhancing signal integrity by eliminating long cables.

Aerospace is updating ASSET to make it more portable. It will be about the size of a tackle box and will be faster and better able to handle low-voltage devices. It has also been designed to accommodate even more varieties of DRAMs, flash memories, and other memory types.

The motion system on which the microelectronics being studied are mounted allows the user easy access. The control software will position the parts normal to the beam at the start of a run. The beam enters the chamber through the lower hole at the back wall. The chamber has several ports with various connector types available on interchangeable flanges.

Future Trends

The first heavy-ion tests at Berkeley in 1979 immediately led to the discovery of single-event latchup. Aerospace investigators were the first to identify several other kinds of single-event phenomena in various types of microcircuits. They include single-event snapback, single-event transients, single-word multiple-bit upset, and stuck-bits effect.

Despite this knowledge, microcircuits occasionally experience anomalies in space—often because of a lack of preflight investigation of radiation effects. When this happens, Aerospace may be called in to assist the anomaly investigation. Such efforts are necessary postlaunch activities; however, the trend is to assess the sensitivity of microcircuits to single-event effects prior to deployment in space. Designers and program managers are increasingly aware that a systematic investigation of all microcircuits is essential to ensure mission success—and prevention of single-event effects through component testing at development stages is perhaps the most cost-effective approach. As the microcircuits in space systems grow ever more complex, ground-based heavy-ion testing of spaceborne microcircuits becomes all the more essential.