Electronic Design

Quickly Find Elusive Signal-Integrity Problems In High-Speed Designs

Traditional approaches to finding signal-integrity issues in high-speed electronic designs involve using hardware triggering to isolate the event and/or deep memory time acquisitions to capture the event and then find it later. The increasing speed and complexity of high-performance electronic systems are revealing some key deficiencies in the traditional event identification methodologies for oscilloscopes.

However, a fundamentally new approach to event identification can significantly augment traditional methods. The end result is a much more powerful event identification system that lets design engineers quickly and easily identify signal-integrity issues.

The Traditional Approach
The traditional hardware triggering/deep memory approach to finding signal-integrity issues has two primary strengths.

First, hardware triggering has zero dead time while it’s looking for an event of interest. The hardware trigger will keep the oscilloscope acquisition system running indefinitely until it finds the targeted event. Once the desired event is located, the hardware trigger circuit will fire and complete the oscilloscope acquisition and nicely display the event on the center of the screen. It’s really quite handy.

Second, with deep memory, you needn’t know anything about what type of signal-integrity issue troubling your target system. You simply set the scope to its maximum memory, set the trigger to edge or even auto-trigger, and hit run. The scope will obediently capture a relatively long-time capture window of target system execution. You’re then free to peruse the acquired data at your leisure to see if any events might be of interest. Some people call this the “swallow and wallow” approach.

These methods are beneficial and fairly well ingrained in the heads of electronic designers who use oscilloscopes to verify their designs. But they also have some severe limitations compared to a novel method that’s emerging in the test and measurement industry.

Something Different
The new approach is most concisely described as event identification software. Essentially, event identification software is smart software that scans each waveform acquired by the oscilloscope and looks for various signal-integrity issues or events of interest in the signal. It lacks the zero dead time of hardware triggering because it inherently has dead time while post-processing the previously stored record. Neither does it have the swallow and wallow capability of deep memory. Still, event identification software features a number of unique benefits that have become compelling to oscilloscope users:

  • Monitoring multiple events simultaneously: Hardware triggering is limited to identifying a single event of interest. The hardware trigger circuit is set up to trigger on a specific type of event and is intrinsically prohibited from monitoring for other types of events at the same time. Event identification software doesn’t suffer from such a limitation. It can be programmed to scan for five events at the same time on any channel or combination of channels. This can significantly reduce the time it takes to narrow down the potential causes of a signal-integrity problem or isolate nasty interdependent events.
  • Finding multiple occurrences of an event: Hardware-trigger circuits can only identify one occurrence of an event in an acquisition. There may be additional occurrences of the event before or after the event is isolated by the hardware, but you will be unable to find them easily. Event identification software can find every occurrence of an event captured in waveform memory. Not only can you look at what led up to the first failure, but also the second, third, forth, and so on.
  • Automatic event navigation: Once you acquire a long waveform with deep memory, you’re left to the extremely tedious and error-prone task of manually scrolling through the waveform and inspecting each part of the signal for potential signal-integrity issues. Deep-memory captures can represent 10,000 screens of information. Manually going through all of this information is impractical. Transferring the oscilloscope data to a controller and writing custom software to measure and inspect such data is also impractical and time-consuming. Once event identification software identifies all occurrences of target events, navigating to each occurrence is a snap using intuitive DVD-like playback controls. Figure 1 shows an example using an Agilent DSO81304B oscilloscope.
  • Variety of events that can be identified: A typical hardware-triggering system can isolate around 10 different types of events or trigger modes. Developing a new hardware-trigger mode typically is daunting for oscilloscope vendors, requiring extensive engineering resources and expensive IC fabrications. Software event identification enables a much wider range of possibilities at a development cost that’s orders of magnitude less. Current event identification software can isolate any event that can be described by a waveform measurement (well over 30 in modern oscilloscopes), plus find troublesome events such as non-monotonic edges caused by improper signal terminations. Designing a hardware-trigger circuit to trigger on such a subtle waveform phenomenon as a non-monotonic edge would be nearly impossible.
  • Speed of events that can be distinguished: A hardware-trigger circuit is fundamentally limited by the speed of its transistors, and it’s an analog-based approach. Best-in-class hardware-trigger circuits currently offer pulse-width (or glitch) triggering down to 300 ps and serial triggering to 3.25 Gbits/s. While these specifications are admirable, hardware-triggering circuits aren’t keeping up with cutting-edge designs that are now reaching 8.5 Gbits/s and beyond. Event identification software is fundamentally limited by the oscilloscope’s sample rate, and it’s essentially a digital process. With industry-leading sample rates measuring 40 Gsamples/s, the speed of events that can be identified by a software-event-identification system is considerably higher than its hardware counterparts. Pulse widths as narrow as 70 ps and serial “finding” as fast as 8.5 Gbits/s are possible with the new approach. (See Figure 2 for another high-speed example.)
  • Resolution of events that can be distinguished: A hardware-trigger circuit has relatively coarse time resolution in the tens or even hundreds of picoseconds, depending on the specific trigger event, the waveform signal characteristics, and particular waveform activity leading up to the trigger event. This coarse resolution can be insufficient for testing against a precise specification (i.e., false failures can be generated). Because software event identification is a digital process, DSP techniques such as one- to 16-sample point interpolation can be used to effectively increase the event resolution. Events can be deemed to pass or fail with picoseconds of resolution. Figure 2 illustrates the identification of a 36-ps rise time.
  • If you can see it, you can isolate it: One of the most striking capabilities of event identification software is its Zone Finder functionality. Most scope users have seen intermittent waveforms occasionally flicker across the screen but have been unable to hit the stop button fast enough to catch them. Typically, these users then will put the scope into a single acquisition mode and hit the single button repeatedly (sometimes very repeatedly) in a vain attempt to capture the infrequent event. Usually, the only result is a very sore button finger. A Zone Finder allows users to interactively draw a zone box on the display where they see the signal intermittently flicker. The next time the waveform flickers through the zone, the scope will automatically stop and display the highly sought after waveform. Figure 3 illustrates a two-zone example. This capability alone is often invaluable.
  • Synchronization with hardware triggering: Finally, the event identification software can be used with the hardware-trigger mechanism via a programmable delay mechanism. That is, it can find the defined software event that occurs a specified amount of time from the defined hardware event. In this way, the combined hardware/software system can be employed as a trigger sequencer (find A then B for example), or the hardware can be used to qualify waveforms for the software to inspect, improving efficiency.

Event identification software is a powerful complement to traditional hardware-triggering or deep-memory capture approaches for identifying signal-integrity issues. When oscilloscope dead time isn’t an issue (i.e., the event occurs more frequently than every second or so—a relative eternity in the world of high-speed electronics), the new approach of software event identification will prove to be the most flexible and effective means of finding a signal-integrity problem with an electronic design.

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.