Image

Understanding how to reproduce a signal integrity problem is the first step in finding the root cause and resolving the anomaly.


In graduate school, I used to hang out with friends who were amateur magicians. We would go to a magic performance and then sit around over beers trying to figure out how the magician did a particular trick. How did he levitate the glowing ball and get it to move around seemingly at his command?

I learned a valuable methodology from my magician friends. The most successful approach we used to deduce how the magician did a trick was to always start with the question, "If I wanted to reproduce this trick, how would I do it?"

Once we had a plan, we would then ask, if he did it this way, we should see him walk out on stage toward the front of the stage to avoid the hidden string, or his right hand should never cross his left hand. We would identify clues to look for to indicate the method he used.

Of course, the reason this was a successful approach for us was that between the five of us, we had a rich experience base of how tricks can be performed. Being clever amateur magicians, and lubricated with a few beers, we never seemed to be at a loss for possible, alternative methods.

Sometimes, the clever magician would change the trick and the method between performances, maybe to keep us amateurs guessing, or make it more interesting for him. Nine times out of 10, in the next performance, we were able to pick up the clues and have confidence we figured out the trick. Rarely, did the trick completely escape us. These were the brilliant magicians we respected as masters. If only we could have had them on our analysis team!

Surprisingly, this was one of the most valuable lessons I learned in graduate school. It applies not just to determining the method of a magic trick, but also to determining the root cause of a signal integrity problem. Of course, the right way of designing a system is to design signal integrity problems out of your product right from the beginning.

But sometimes, they creep in. During the bring-up phase, you perform measurements of signal waveforms and find anomalies. Figure 1 is an example of a measured scope trace of a waveform in a digital system.

Image

It's not supposed to have those stair step glitches in it. Not getting the signal up to the minimum input high for a 1 or below the maximum input low for a 0 within the set up and hold time will mean a false reading, or the possibility of a metastable response in the receiver.

You need to fix this problem. Once the problem has been identified, the next step in fixing it is to determine the root cause. Usually, once the root cause is found, the solution is obvious.

What I learned from my magician friends is that the most effective process to find the root cause of a problem is to ask yourself the question, "If I wanted to reproduce this trick, how would I do it?"

How could you re-create the waveform in this example? It is often difficult to "synthesize" an approach from scratch. This is where experience is incredibly valuable. Cheating is allowed; you are allowed to ask others, to look at reference books and papers; having a few experts on your team helps. After all, what's an expert but "someone who's made all the mistakes possible?"

A recently announced tool from ASA Corp (amherst-systems.com) might also help. HAL, short for Hidden Anomaly Locator, uses software agents, each tasked to be on the look out for a specific type of problem. HAL will monitor scope waveforms, on the look out for anomalies. Once found, it will offer suggestions for possible root causes. It's like having a bunch of experts on your team.

In the example above, what might have caused the stair step shelves? One possible explanation is impedance mismatches in a source series terminated line. If the source series resistor was too high, it might cause this effect.

One clue to look for is the duration of the shelf. It should be the round-trip delay time of the transmission line. Another clue might be the value of the series resistor. A typical CMOS driver has an output impedance of 10 Ohms, which would require a resistor of about 40 Ohms. To reproduce this stair step waveform, a resistor of about 80 Ohms would be needed. Check the resistance of the resistor on the board.

Problems can turn from stumbling blocks into stepping stones only if you learn the correct lesson from them and move farther up the learning curve. Take advantage of all the help you can.  PCD&F

Dr. Eric Bogatin is president of Bogatin Enterprises. This and other topics are covered in the public classes Eric teaches. Check his web site for the schedule: BeTheSignal.com. Send questions to This email address is being protected from spambots. You need JavaScript enabled to view it..
Submit to FacebookSubmit to Google PlusSubmit to TwitterSubmit to LinkedInPrint Article