Ircona, IBIST, and a great Signal Integrity story

Last week I had the privilege of speaking with the staff of Ircona (www.ircona.com), a firm with special expertise in high-speed processor-based designs and BIOS customization. Based in West Dublin and a development partner for Phoenix Technologies, Ircona has recently used our Intel® Interconnect Built-In Self Test (IBIST) tool to troubleshoot some very elusive signal integrity issues on a high-end server design.

We at ASSET are always interested in speaking with customers who use our technologies in innovative ways, and we’re especially pleased when we save our customers time and money. The Ircona team shares the same core values. In a recent design they were working on for a large HPC firm, IBIST caught some unforeseen crosstalk on a high-speed signal which was not being probed using oscilloscope-based tests. The issue was caught early in the prototype stage subsequent to weeks of HSPICE simulation and rectified with a simple board re-spin. This assured rapid time-to-market and avoided potential field failures after volume deployment. Ircona shared with us some of their insights into using legacy oscilloscope-based I/O validation versus the non-intrusive IBIST tools:

  • The ScanWorks® Intel IBIST tool is very simple to set up and use compared to traditional ‘scopes. There’s no wrestling with figuring out which signals to probe (The longest? The shortest? The “riskiest”?). No rework is needed to solder on the probes, and no workarounds are needed to get around heat sinks and other obstructions. And there’s no triggering mechanism needed. So preparation time is reduced from days to minutes. You just plug into the XDP header and go.
  • Probe-based tools like oscilloscopes use “simulation technology” to try to replicate the signal at the silicon. De-embedding mathematics are used to model the transmission path without the fixturing and probes and see the “real” eye. You simulate to create a good design and then you simulate again to get an actual measurement. ScanWorks avoids the second simulation. It sees what the silicon sees.
  • ScanWorks shows all of the signals, as opposed to just a select few. So, you don’t have to worry about missing that critical signal that might have an issue. Otherwise, we might have missed the crosstalk between QPI and one of the non-QPI signals. ScanWorks gives you peace of mind that you’ve tested as comprehensively as possible, as opposed to a “hit and miss” approach. And it’s a whole lot cheaper than a ‘scope.

A more comprehensive case study is in the works and we’ll share additional results when they’re ready. But in the meantime, I wanted to blog about this success story of how an extremely clever group of engineers at Ircona used ScanWorks to help their customer.