An expert viewpoint brought to Electronic Design by Agilent Technologies, Inc.
There’s a bit of a philosophical debate surrounding how oscilloscopes and probes make measurements, especially at today’s high frequencies. What should a scope/probing system try to show: "What was," in other words, the signal that was present at the test point before you connect the probe? Or "what is," referring to the actual signal at the probe tip just after you connect it and load the device under test?
Intuitively you’ll probably first react with "what was"—after all, don’t you want to know what the circuit does when it’s operating on its own without your probing? That’s a great goal, but unfortunately with today’s technology it’s impossible to reach. How vendors deal with these measurement realities varies widely, as do the results.
First, why is it impossible? Recall from your physics class the Heisenberg Uncertainty Principle, which loosely stated says that at the atomic level, the act of observing something disrupts it so that you can never know what it was like just when the observation started. Many discussions add that this effect doesn’t show up at the macro level. Well, it turns out that a similar effect does arise when you probe high-frequency circuits.
A probe with infinite impedance would be truly invisible to the test circuit, but such a probe doesn’t exist. A probe will always load the test circuit to some degree. Further, no matter what its impedance, a probe must grab a "piece" of the signal so it can try to reproduce the entire signal on the scope display. This act of stealing even a tiny portion can have dramatic effects on high-frequency digital circuits.
A serious problem also arises at high frequencies because leads from the probe to the circuit can act as a resonant tank circuit, just as in a transmission line. At the gigahertz clocks and picosecond rise times we’re dealing with in high-performance circuits, a couple of millimeters of wire put on the tip of a standard active scope probe can add enough inductance to drop its resistance to tens of ohms at resonance. As a result, a 5-cm wire connection reduces the 6-GHz bandwidth of a traditional active probe down to 1.5 GHz.
Previously the only way to keep the frequency of this resonance above the probe’s bandwidth was to use an ultra-short, stubby point at the probe input. Although this produces good high-frequency fidelity, it’s often impractical to connect such a short blunt point to the circuit.
A different method to limit loading effects is to damp the L-C tank circuit with a small resistor on the front end of the probing system—in transmission-line terms, to source-terminate the line. This allows you to use a longer, easier-to-handle connection at the input of a probe. With a properly damped probe input, the loading/input impedance never drops below that of the damping resistor, which is in the range of 100 to 250 W. Ω In addition, the transmitted response remains reasonably flat.
In an attempt to get a view of the "what was" signal, some vendors offer probes with internal peaking that compensates for the loading. It’s a nice idea—except the probe load works in conjunction with the source load, of whose value you can’t be certain. Further, the source load is different every time you probe a new point, each having its own degree of loading.
Now you can appreciate why signal fidelity begins at the probe tip, and how that accessory can limit the performance of the entire system. Many engineers make the mistake of selecting a scope first and then the probe as an afterthought. They believe they need a 6-GHz bandwidth scope running at 20G samples/sec. But then they add a probe with a lower bandwidth, thereby limiting system performance so they’ve wasted money on that expensive instrument.
Because of these issues, test manufacturers have been devoting a remarkable amount of R&D time just to active probes. They’ve come up with very sophisticated units that include damping resistors right next to the probe tip. They’ve developed probe heads, some of which even solder onto the circuit board to minimize lead effects. And I don’t expect things to stop there. Wouldn’t it be nice if we had a probe that allowed us to know both the "what was" as well as the "what is"? Despite Heisenberg and circuit loading, today’s test manufacturers are putting all their talents towards finding a way to do it.
There’s plenty of background material and even a free 6-minute video demo titled, "Signal fidelity begins at the probe tip". It discusses the issues of hf probing and can be found at: www.agilent.com/find/curveol-july04
Click here to download the PDF version of this page.