How do you measure 500 GHz?

On June 20, 2006, IBM and the Georgia Institute of Technology demonstrated the first silicon-based transistor that could operate at 500 GHz (Ref. 1). To achieve that speed, a team of researchers led by Professor John D. Cressler had to cool the SiGe chip to 4.5 K (–451°F). But how did they know that the transistor could switch at a half terahertz when no test equipment exists that can measure at such a frequency? To find out, I spoke with Cressler by phone.

PhD student Ram Krithivasan (seated) and Professor John Cressler oversee a SiGe chip mounted in a cryogenic test station. Courtesy of Georgia Institute of Technology.

While you can't directly measure speeds to 500 GHz, you can infer how a device will respond at frequencies higher than today's test equipment can measure. To perform the measurement, Cressler and his team measured S-parameters with a vector-network analyzer. They excited the device to 35 GHz and measured its gain. Then, they extrapolated the gain versus frequency to find the device's unity-gain frequency.

“It's a standard technique for measuring high-speed devices,” noted Cressler. “We assumed that the gain will roll off at 20 dB/decade. If we can prove that roll off through direct measurements at frequencies that our equipment can cover, then we're highly confident that the extrapolation is correct.”

Each measurement point on the device takes several days to make. Researchers must calibrate the system at room temperature using on-wafer calibration structures, then cool the device to 77 K and then to 4.5 K and make each measurement. They make measurements at other temperatures as well. They repeat the calibration and measurement steps at each frequency.