Does Signal Integrity Engineering have a Future?

(1.4 MB)

What do SI engineers do? What are their deliverables? Almost everyone says they need an SI engineer, but no one seems to agree on exactly what they do. Some SI practicioneers describe their collection of magic tricks, and along with their horror stories of people who did not listen to them, peddle their wears to whoever will buy. Other people have their own horror stories of dealing with SI engineers who iterated their way to success, obtaining a dozen or more attempts before claiming victory.

So what is the current state of Signal Integrity engineering, and how did it begin, and where is it likely to be in ten years? This paper will describe all this and more, including the authors’ personal experiences and interviews with other engineers and managers. Funny, sad, and angry, these stories are all part of the emerging discipline of signal integrity. This paper will provide several optimistic future directions that SI can branch into, and what is needed to realize each of these directions.

Network Class

(583 KB)

Network class processors with more than 1000 high speed I/Os are becoming more important in the telecommunications backbone.  This class of processor has higher off chip switching  bandwidths, larger buss widths and a greater number of different busses compared with microprocessors. Because of the off chip features, the package signal integrity requirements are more severe.

The combination of higher bandwidth, mostly differential signal lines and flip chip attach make the electrical characterization of these packages particularly challenging. We show how to fully characterize a package using a unique top-bottom probing system and multi-port differential S parameter test system.

Differential Bus Design

(1.8 MB)

•         Historically, differential interconnects were often twisted wire pairs — the desired differential signal propagated well, & common mode noise propagated poorly.

•         Current designs use stripline, with many ground planes.

•         Common mode signals now propagate very well.

•         Differential signaling is mystic—if same thing is done to both traces, the differential signal will always get through.

•         This is a bad assumption.

Slammer

(720 KB)

The increasing power requirements of current and future microprocessor designs present both a tremendous DC and transient challenge to the wafer probing hardware.. Power supply, power path, package, test and wafer probing hardware technology is being pushed to new electrical and physical limits which require innovative solutions to assure successful delivery of the power. It appears we have a paradigm shift with signal integrity taking a temporary back seat to power integrity considerations.