Data Mining Analytics Improves Serdes HSIO Validation

Silicon vendors like Intel have been saying for some time that pass/fail validation testing on high-speed I/O (HSIO) or serdes buses just doesn’t cut it anymore. Why? As the speed of HSIO buses has continued to increase, they’ve gotten much more sensitive and susceptible to signal integrity problems which can lead to failures or throughput degradations. A bus might pass a validation test in the development lab, but its operating margin relative to the eye mask might be very slight. And that’s not good.

Pass/fail tests won’t tell you what the margin is. Plus, a slightly passing margin in the pristine conditions of a development lab will certainly diminish as the system moves through the rest of development and production, and eventually finds its way into the hands of users. That’s where – in the hands of users – a slightly passing margin could start to cause big problems, either failing completely or running at an unacceptable speed. And that’s the last place a system supplier wants an unforeseen problem to manifest itself, because that’s when you start talking about new product returns and dissatisfied users.

Instead of pass/fail, silicon vendors have recommended ‘NxN’ validation testing methodologies that involve running a test ‘N’ times on ‘N’ different prototypes or production circuit boards. For example, a test could be performed five times on five different boards. A lot of data is gathered as a result of such a methodology but, when analyzed, it leads to a predictive model of the likelihood of problems on the HSIO or serdes bus throughout the system’s life cycle.

Data_Mining_Analytics_for_Serdes_HSIO_ValidationData mining analytic tools like ScanWorks’ HSIO Validation Assistant can not only streamline the analysis of this database, but also reduce validation costs by standardizing the process across multiple teams, improving test stress quality, and establishing a baseline across the product life cycle .  Successive validation testing on buses with more than adequate margin can be eliminated, resulting in a reduction in the overall testing costs for the system.

Of course, there’s a lot more to all of this than can be explained here. That’s why we’ve just published a new eBook called “Data Mining Analytics for Serdes HSIO Validation.” It’s available now and free for downloading  from the eResources  section of the website.

Tim Caffee