Inadequate Test Tools Can Have A Very High Price

An expert viewpoint brought to Electronic Design by Agilent Technologies, Inc.

The age-old adage says you can't do the right job without the right tool, but in today's high-performance electronics there's an important corollary: Not having the right measurement tools can increase a product's development cost and business profitability-sometimes dramatically. The cost for sophisticated tools might initially seem high, but this outlay can improve time to market by months and improve business results by millions of dollars.

This fact is becoming particularly apparent when designing high-speed systems to meet the requirements of external standards bodies or even internal performance requirements. A key goal is determining if a design has sufficient performance and margin to reliably stay within specification requirements. The penalty for not meeting this objective is increased development cost, reduced market windows and many sleepless nights.

An illustration comes from the field of high-speed serial buses, where technology is developing at such a rate that a lack of adequate test tools is becoming a limiting factor. We're now dealing with extreme speeds; serial buses at 2.5, 3 and 4 Gb/s are moving to 5, 6 and 8 Gb/s for their next generations.

An extremely popular method of analyzing such serial bitstreams is using a real-time oscilloscope to generate an eye diagram. Assume you have an extremely accurate and stable signal source; a perfect eye diagram would then consist of sharp lines, so any fuzziness or spreading in the lines must come from the measurement instrument. No scope is perfect and will always contribute some fuzziness, but how much?

The nearby eye diagrams should prove instructive. The image on the left was taken with one of today's high-end 8-GHz digital scopes, and the one on the right with a state-of-the-art 12-GHz model. Both examine a 6-Gb/s bitstream coming from a high-accuracy source. Clearly, with the figure to the right you know much better what the system can accomplish, whereas on the left the system under test could be performing either in or out of spec-and you wouldn't know the difference because of deviations coming from the scope itself.

This example isn't at all unrealistic. A 6-Gb/s bitstream could be representative of a next-generation Serial ATA, Serial Attached SCSI or a proprietary design. A bus at that speed will have typical characteristics such as an ideal eye width near 167 ps, eye height of 800 mV and a signal rise time (20-80%) of 40 ps. At these levels, you must be prepared to deal with potential signal-integrity issues in the test setup such as jitter, noise, crosstalk and transmission-line loss.

Let's see how a real-time scope's functionality could impact a serial-bus test using these three key criteria, all of which are typically involved in product qualification. For the ideal eye width of 167 ps, the minimum required to pass the specification would likely be near 142 ps (85% of ideal). The scope on the left can measure it at best as 140 ps due to internal jitter, while the one on the right can measure it as 150 ps and thus let a part working within spec pass the test. Next, the ideal eye height is 800 mV, so the minimum required is 680 mV. The left scope shows at best 650 mV due to intrinsic noise levels in the scope itself and probe, while the one on the right measures the minimum as 720 mV. Finally, the ideal signal rise time is 40 ps, so the maximum allowable rise time would be 47 ps. The left scope measures it at best to 48 ps ±10 ps due to bandwidth and measurement repeatability limitations, but a new-generation scope such as that on the right measures it as 40 ps ±5 ps.

The key is that in all three cases a design could actually meet the spec, but the scopes we've had to settle for until today (left photo) might fool you into thinking it doesn't. It's really the test tools that are at fault because they mask the system's true performance.

The implications are dramatic. Once a well-simulated piece of silicon arrives for verification, you need high-performance test tools to determine if it meets requirements. By falsely concluding that a good design fails compliance requirements, you might end up needlessly re-cutting an IC or redesigning a circuit board, and at what expense?

An adequate test tool should accurately determine if you're meeting compliance requirements. This issue takes on added importance with high-speed serial schemes. Luckily, a significant barrier to high-speed electronic design is falling with the advent of next-generation oscilloscope and probing solutions.

To learn more about jitter and how to measure it with today's instruments, check out the many resources at:
www.agilent.com/find/curveOl-oct04