Congratulations to ECE faculty Prof. Chris Kim on being awarded the Semiconductor Research Corporation (SRC) Technical Excellence Award for 2016. The award recognizes his work in the invention of a new class of compact on-chip sensors called “silicon odometers” that can accurately and efficiently measure circuit aging effects.
Wear and tear of systems and devices is a concept most of us are familiar with. Usually there are warning signs or messages that allow us to take preventive or corrective action. But what about wear and tear that happens quietly, behind the scenes, and the only notice the user gets is when the device fails? A figurative coughing and sputtering followed by failure? Microprocessors are such devices; transistors, their key component, do wear down over time and adversely affect the performance of microprocessors. And currently there is no way to indicate the wear and tear they have undergone.
Some of the contributing factors to transistor degradation are their arrangement in any given circuit, and the voltages and temperatures they are exposed to. Given the extent of variability of these factors, it is difficult to predict the decline in performance and to measure the effects of aging in transistors. Compounding these challenges are problems of costs and time associated with testing or measuring the aging effects.
So why is measuring and predicting the aging of a transistor an imperative now? Advancements in microprocessor technology such as increased transistors (to the tune of billions), extremely high clock speeds, and the resulting exposure to high heat make for faster decline in the performance of transistors.
However, improvements in measurement and prediction of the aging process in a transistor would mean that chip makers can let microprocessors run faster, instead of following the current industry practice of setting the clock speed low enough that the microprocessor will operate smoothly through its intended lifespan.
Prof. Chris Kim’s group has developed a solution to this problem. Based on a pair of ring oscillators, the silicon odometer measures the beat frequency of the ring oscillators. The odometer measures the difference in frequency of an unstressed ring oscillator versus a stressed ring oscillator (one at a higher operating voltage). Measuring the beat frequency affords the measurement of extremely minuscule changes in transistor switching times.
Prof. Kim was awarded two SRC grants to work on his silicon odometer designs. Over the span of two projects, his team has experimentally demonstrated 15 different odometer designs in 130nm, 65nm and 32nm technologies. Tiny delay shifts as small as 1ps can be measured within a microsecond (critical for preventing unwanted recovery) using a standard serial interface. Another key advantage of the silicon odometer is that any change in operating voltage or temperature will have an equal impact on both oscillators.
Such on-chip circuit aging sensors could soon be designed into microprocessors, to monitor clock speeds and voltages, allowing for adjustments that let the microprocessors function at their peak levels. The semiconductor industry is poised to embraced this idea, and the direct technology transfers to SRC member companies such as IBM, Texas Instruments, and GlobalFoundries by Prof. Kim’s students are indicative of the impact of the group’s work.
The following students contributed to Prof. Kim’s research:
John Keane, Ph.D.
Tony Kim, Ph.D.
Xiaofei Wang, Ph.D.
Pulkit Jain, Ph.D.
Won Ho Choi, Ph.D.
Qianying Tang, Ph.D. candidate
Chen Zhou, Ph.D. candidate
Established in 1991 by the Semiconductor Research Corporation’s Board of Directors, the award seeks to encourage and provide recognition to researchers conducting work in areas of significance to SRC members. Among other factors, nominations are reviewed for their creativity, relevance, and value to the semiconductor industry. The award is presented in fall every year.