by Pete Singer, Editor-in-Chief, Solid State Technology
May 19, 2008 – Cascade Microtech (Beaverton, OR) has unveiled a new test system designed to measure flicker noise in ICs, seen as a barrier to lowering device operating voltages as geometries shrink — and a strategic initiative to integrate measurements systems to provide users with measurement accuracy assurance.
Flicker (1/f) noise, which occurs in all semiconductors, can compromise device performance by causing jitter or phase noise in communications devices (resulting in high bit error rates), or random retention errors in flash memory or soft errors in SRAM. At smaller geometries, flicker noise has become a key contributor to overall noise in high-performance devices — in 2003 the ITRS amended a previous observation by reporting that operating voltages actually increase as device geometries shrink. Flicker noise is the gating barrier to reducing device operating voltages, so accurately characterizing it is increasingly urgent. (And today’s market pressures allow even less room for the types of performance degradation flicker noise can cause: higher serial data rates means less jitter margin, and the low standby power demanded by handhelds results in lower signal-to-noise ratios.)
Unfortunately, accurately characterizing flicker noise is “virtually impossible” with the traditional approach of “piecing together” a system in which device noise thresholds are becoming undetectable, says Cascade Microtech. Flicker noise is characterized by a corner frequency (fc) beyond which other types of noise become dominant; because extrinsic background noise often obscures the ultralow-noise frequency corner, measurement and extraction tasks are typically difficult to achieve with precision. Moreover, says the company, such background noise is often introduced by the very systems attempting to measure flicker!
Today’s new MOS communications devices generally have a higher corner frequency than previous-generation, bipolar devices approaching 30MHz. Device flicker noise levels are higher and therefore intercept the thermal noise at a higher frequency. And therein lies the challenge, according to Cascade Microtech. When conducting noise measurements at these higher corner frequencies, the device load resistance and load capacitance form an RC circuit, resulting in a low-pass filter, which limits bandwidth and subsequently the ability to detect the higher corner frequency.
Adding to the difficulty of realizing accurate flicker noise data is the fact that the low-level corner frequency is often obscured by other extrinsic background noise brought on from external sources — a symphony of noise made even more complex when connecting multiple instruments and probes. Optimizing these instruments and probes for the lowest possible noise is necessary for the best results, but achieving such an optimized environment has proven to be difficult.
Cascade says its new Edge system is the only flicker noise measurement system certified to provide accurate measurements from 1Hz-30MHz. In contrast to traditional “rack and stack” flicker noise measurement systems that typically comprise elements from up to five different suppliers, the Edge is designed to be turnkey, with integrated wafer probe station, instruments, software and accessories (see image). Service and support include site surveys, pre- and post-measurement optimization, and continuing application support.
Singapore’s Chartered Semiconductor Manufacturing deployed the first Edge tool, which has a stated pricetag of $1.2M. The foundry has “confidence in the measurements this system produces,” said Michael Cheng, senior director of the foundry’s device technology division, in a statement, adding that the tool provides “substantially more information on the devices’ flicker noise performance” critical in designing low-noise, high-performance circuits.
The Edge measurement tool is also the first piece of Cascade’s new initiative to develop integrated measurement systems for on-wafer semiconductor device characterization and process development, rather than “bolting together” systems from different vendors. Each system will be designed “from the ground up” to optimize measurement capability and accuracy. More technologies/tools are expected in the coming months.
For Chartered, such an integrated measurement system from one vendor offers a more secure investment, providing value in terms of “guaranteed measurement accuracy and single-vendor support,” noted Sanford Chu, VP, device technology division, in a statement. — P.S.