With the arrival of third-generation serial standards, many of the rules around debug/verification and characterization/compliance test are changing. Here are some “dos and don’ts” to help you cope with these emerging new requirements. First, here are the things you should do:
• Ensure proper connectivity: With faster third-generation serial speeds, signal integrity is critical not only for the design itself but also when planning for adequate test connectivity. Test connectivity elements such as current path, termination, inductance, capacitance, and routing should all be factored into your design. For instance, parasitics like stubs, via caps, pins, and solder balls have capacitance that generally varies with the size of the objects. You’ll want to account for these as you begin to setup your de-embedding (see below). The same is true for measurement access points. While today’s probe tips are very small, sometimes you need access to fixturing that opens up access points even further. Consult with standards organizations, test equipment manufacturers, or testing houses like UNH-IOL for further information on “approved” fixtures.
• Select the right testing toolset: Serial data test methodologies can take different shapes depending on the challenge at hand. Make sure you’re using the right toolset for the job. For environments where debug and verification are important, ensure that your oscilloscope has good triggering, analysis, and waveform capture capabilities for deeper evaluation of cause and effect. For testing that has an emphasis on characterization and compliance to a standard, oscilloscope bandwidth requirements are critical as well as ensuring adequate record length for multiple unit-interval (UI) timing verification and clock recovery. Higher-speed standards are characterized using receiver stress tests that include pattern generation and stressors like noise and timing. An instrument like an arbitrary waveform generator (AWG) or BERTScope works here.
• Take advantage of automated tests: Third-generation serial standards are complex. Remembering the exact steps to perform each measurement is time consuming and often requires going back through the latest compliance test specification (CTS). Test automation tools can help take the human element out of the equation and yield more accurate and repeatable measurements across hundreds of CTS-required measurements.
Now, let’s look at some things you definitely should not do:
• Ignore test fixture effects (de-embedding): As speeds increase, the noise created by attached fixtures and the channel is significant enough that it may result in inaccurate eye-diagram representations. Removing or de-embedding the fixture and channel effect is critical to verifying that you have met performance and reliability goals. Most precision network analysis tools use some form of de-embedding to remove instrumentation effects and move the reference plane of a measurement closer to the device under test.
• Use an insufficient waveform capture rate: There is something wrong with a particular input signal because the output is corrupted, and yet you cannot see the problem. This happens when the anomaly has a low rate of occurrence and the oscilloscope is not capturing data during the occasional error. In such cases, the solution is to increase the oscilloscope’s waveform capture rate. Once you see the characteristics of the intermittent failure, you can use the advanced triggers to capture the anomaly. To be safe, make sure you have adequate bandwidth and rise time capabilities built into the instrument.
• Skip link performance tests: Serial engineers are all familiar with performing transmitter measurements like eye diagram, timing, and noise. Typically, the attitude has been that if the signal looked good leaving the transmitter, channel losses and receiver tolerances would be met by the design. No longer is this a safe assumption. At issue is the continued use of lossy yet low-cost transmission mediums such as FR-4 that lead to signal degradation and closed (or nonexistent) eyes at the far end of a channel. Understanding the channel or link performance (or lack thereof) is fundamental to robust communications system designs that pass compliance specs with margin for increased performance.
Chris Loberg is a senior technical marketing manager at Tektronix responsible for oscilloscopes in the Americas region. He has held various positions during his 13 years with the company, including marketing manager for its Optical Business Unit. His extensive background in technology marketing includes positions with Grass Valley Group and IBM. He holds an MBA in marketing from San Jose State University.