Kawasaki, Japan, April 30, 2008
Fujitsu Laboratories Ltd. today announced that, in collaboration with the Hawaii Observatory of the National Astronomical Observatory of Japan, an inter-university research institute that is part of Japan's National Institute of Natural Sciences, the development of a technique for quick and accurate field measurement of the rate of "soft errors(1)", which are erroneous operations that occur in advanced semiconductors and are caused by neutrons(2) from a type of cosmic ray.
The new technology enables quick and accurate determination of the rate at which soft errors in a LSI chip will be encountered, a rate that varies depending on such characteristics as the geometric latitude and altitude of the location, and the shield effect of buildings in which the LSI is employed. Through the accurate and fast assessment of soft error rates, developers of advanced semiconductors can be advised on the optimal countermeasures to take, depending on where and how the LSIs will be used.
Details of this technology will be presented at the International Reliability Physics Symposium (IRPS) being held from April 27 to May 1 in Phoenix, U.S.
Background
Figure 1: Site at which measurements were taken at: summit of Mauna Kea mountain
Larger View (460 KB)
Neutrons originating from secondary cosmic rays and alpha particles(3) originating from the breakdown of radioactive materials can exert an effect on memory and logic circuits of LSI, sometimes causing soft errors. Neutrons, in particular, account for 30 to 80 percent of all soft errors. Because the impact of soft errors increases as LSIs become more closely integrated, the development of countermeasures to address soft errors is becoming an increasingly important issue.
Due to the fact that energy spectra of neutrons(4) are influenced by such characteristics as geometric latitude and altitude coordinates as well as the construction (shield effect) of buildings, measurements of soft error rates vary by location. Thus, in order to determine appropriate countermeasures, it is essential that there be an accurate assessment of the soft error rate of the actual location where the LSIs will be used. Since conventional methods of assessing soft error rates using either computer simulations or particle-accelerator experiments may not reflect the energy spectra of neutrons at the actual location of use, soft error field experiments conducted at the actual location of use are becoming essential.
Technological Challenges
Conventional soft error field experiments were faced with problems such as the following:
- Measurement of neutron energy spectra at actual-use sites
The rate of soft errors depends on the energy spectra of neutrons in the atmosphere. Moreover, the spectrum varies depending on such variables as latitude, altitude, and type of building construction. Therefore, it had been difficult to accurately measure the neutron energy spectra existing at actual-use sites.
- Statistical precision problem
Even if 1000 memory chips are measured over the course of one year, it will only produce a range of a few to up to ten errors, making it difficult to obtain a statistically valid sampling.
Newly-developed Technology
Researchers from Fujitsu Laboratories and the National Astronomical Observatory of Japan developed a technology that measures quickly with a high degree of precision the rate of soft errors in advanced computer chips in a particular location.
Key features of this technology are as follows:
- Accurately assesses the relationship between the neutron energy spectrum and the soft error rate
Because measurements of the neutron detector, which accurately measures the intensity and energy spectra of neutrons, were taken simultaneously with measurements of the soft error measurement system, it was possible to accurately assess the relationship between the neutron energy spectrum and the soft error rate. Furthermore, by taking neutron measurements both inside and outside the building, Fujitsu Laboratories was able to evaluate the impact of the shield effect, from structures of buildings, on the soft error rate.
- More precise soft error assessments
In order to precisely assess soft errors, the researchers took measurements using the Subaru telescope(5) located atop Hawaii's Mauna Kea mountain, where the intensity of neutrons is ten times greater than that at sea level, thus making it possible to measure the neutrons energy spectrum and quickly collect statistical data using high-precision error rates.
Figure 2: Soft error rate simulation: value for Mauna Kea was standardized as 1. (See "Results" for further details)
Larger View (73 KB)
Results
By comparing measurements taken in Tokyo and the top of Mauna Kea, the researchers found the intensity of neutrons at the top of Mauna Kea to be 16 times the level in Tokyo. When, in addition, the effects of building composition were factored in, it was 7.4 times the level of Tokyo. When the soft error rates measured at the top of Mauna Kea were divided by this coefficient of 7.4, the amount basically corresponded to previous measurements for Tokyo as well as simulated rates for Tokyo.
Furthermore, in the measurements at the top of Mauna Kea, using a sample of 1,024 90nm SRAM chips, 36 soft errors were recorded over the course of approximately 2,400 hours of measurements. In a relatively short period of time, less than one-eighth the amount of time required for measurements in Tokyo, the researchers were able to obtain highly precise assessment results.
Future Developments
This technology will be applied to assess soft error rates of 65nm, 45nm, and 32nm LSI devices, thereby providing an important diagnostic tool for developing appropriate soft error countermeasures (such as determining whether such countermeasures are necessary, which materials to use, what circuit types to use) even for leading-edge LSIs in which soft errors are increasing.