Lessons learned on the topic of high-speed signal probing, and how to build a high-bandwidth "Lo-Z" probe for very little money.
Some years ago, I learned some valuable lessons about probing high-speed signals. Somehow, mistakes have a way to sticking in your mind and we call them "experience". Most modern oscilloscopes come with a "high-bandwidth" 10X passive probe. Figure 1 shows the venerable Tektronix P6139, 500 MHz, 8pF probe. At first glance, a "500 MHz probe" might seem adequate to probe say, a 125 MHz clock signal right? Wrong... We will see why with a practical example showing the issues you run into when trying to probe such a signal. Later in the article, I'll also show how a surprisingly simple and affordable DIY passive probe can outperform even the best 10X passive probes for this particular application.
Figure 1 - Tektronix Passive Probe
The Tek P6139A are very good quality passive probes (as passive probes go), but often one gets carried away by the banner "500 MHz specification" and overlooks the 8 pF capacitance specification. To see how this parameter matters for the application at hand let's consider the effect the 8pF load capacitance has on rise time. Using the simple approximation tr = 2.2 x R x C, and assuming for the sake of argument that a "perfect", 0 rise-time square wave from a 50 Ohm source is applied to this capacitor, one gets:
tr = 2.2 x 50 x 8e-12 = 880 ps
This may not seem like much, but for a 125 MHz signal with a 8 ns period, the actual rise and fall time is usually in the sub 1ns region already. This means that using the 8 pF probe would create an unacceptable error in the rise and fall time measurements. Another way to look at this is that the impedance of a 8 pF capacitor at the fundamental frequency of 125 MHz is already 1/(2xPIx125e6x8e-12) = 159 Ohm (not insignificant relative to the 50 Ohm source) and things get much worse when the third and fifth harmonics are considered. The ideal square wave contains only odd harmonics, so the 5th harmonic is at 5 x125 MHz = 625 MHz. A this frequency, the 8pF capacitor becomes a mere 32 Ohm load. Suddenly, the 500 MHz bandwidth probe seems hardly adequate for the task.
Another practical consideration when measuring this type of signal is lead inductance. The typical "crocodile" ground lead that comes with these passive probes is pretty long (some 10cm). Assuming a 10 nH/cm inductance, this represents about 100 nH of total inductance. The reactance of a 100 nH inductor at 125 MHz is 79 Ohm and a whopping 392 Ohm at the 625 MHz fifth harmonic. Furthermore, the combination of inductance and capacitance forms a "tuned" circuit with a resonant frequency. Things can get pretty ugly....
Let's step aside from the mathematical world for a second and see how these factors influence a measurement in the real world. Figure 2 shows a small board I built including a 125 MHz oscillator and a series dampening resistor connected to an SMA. The oscillator runs at 5V nominally and provides a 0 to 5V CMOS output.
Figure 2 - The Oscillator
Figure 3 shows a naive attempt at probing the signal output using a Tek P6139A passive probe. Notice the long "crocodile" ground lead connection.
Figure 3 - Naive Probing Technique
Figure 4 shows the measurement as seen in a 1GHz scope (at 10GS/s equivalent sampling speed; fast enough to avoid aliasing effects). What, you say, a sine wave??? Not only a sine wave but one with a almost 10V pk-pk amplitude! Indeed strange things can happen when measuring "square waves" without an adequate probe. The limited bandwidth of the probe heavily attenuates the third and fith harmonics thus resulting in a signal predominantly derived from the fundamental frequency... Furthermore, the inductive effects from the ground lead create overshoot and undershoot effects that increase the apparent signal amplitude.
Figure 4 - The first measurement attempt
As mentioned in the introduction, the there's fortunately a simple and quite affordable solution for this type of probing problem. While you can certainly buy high-bandwidth, low-capacitance probes from the usual vendors, building a DIY lo-Z probe is very easy. Here's how. The Lo-Z probe is simply a piece of 50 Ohm coax cable with a series resistor at the probing tip. The coax cable connects to the scope which must in this case be set to 50 Ohm internal termination. In it's simplest form this probe is a sort of resistive attenuator. Resistor values of 450 Ohm or 950 Ohm are commonly used to yield a 10x or 20x attenuation factor respectively. In this case, I used a 450 Ohm resistor. Notice that these probes present a DC loading that might look high at DC levels but that it is nevertheless much better than the actual loading of the common passive probe at high-frequencies. This is because the actual capacitance of these probes (less than 1 pF typically) is so much lower than the later.
Figure 5 shows a pf RG-58 coax with an SMA connector at one end. The other end is simply stripped and the resistor is soldered in series as shown in Figure 8.
Figure 5 - The Coax cable
A small piece of copper metal is used as the ground lead as shown in Figure 6. It's important to keep the ground lead as short as possible to minimize the inductance effects. I've also considered using a SMD resistor. This would likely reduce the series inductance of the probe but would complicate the assembly. It may however be a better solution if even higher speeds are needed.
Figure 6 - Resistor and Ground lead
Crimp tubes as the ones shown in Figure 7 commonly found at craft stores (here's an Amazon.com link) can be very useful when assembling circuits like these. Figure 8 shows how they were used to assemble the probing resistor to the tip of the coax cable..
Figure 7 - Crimp Tubes
Figure 8 - Soldering the Series Resistor
After soldering the resistor and the ground lead, I added heat-shrink tubing to provide isolation and give the probe a more finished look. See Figures 9 and 10.
Figure 9 - Assembling the Ground Lead
Figure 10 - Probe tip with heat-shring tubing added
Figure 11 - Finished probe
The first step before using this type of probe is to enter the attenuator factor in the oscilloscope (if the scope supports it) as shown in Figure 12. This will make it easier to read values and make measurements in the scope. However, even if the scope you are using doesn't support this, it is simply a matter of multiplying by 10. Keep in mind that the 500 Ohm load this probe presents might reduce the amplitude of the signal under measurement, thiough as seen before, this is not nearly as bad as the loading the extra capacitance would present if a passive probe was used...
Figure 12 - Setting the attenuation factor
Figure 13 shows the new approach to probing. Notice how short the ground connection is compared to Figure 3. This reduces the parasitic inductance effects dramatically.
Figure 13 - Probing
Figure 14 shows the scope measurement using the just built Lo-Z probe. There is a square wave in there after all! Notice also how the very fast rise and fall times (694 ps and 540 ps respectively) indicate that the capacitive loading is no longer limiting our measurement. In fact, these values are close to the rise and fall times for the oscilloscope I used, so the scope is likely more of a factor now than the probe itself!
Figure 14 - Square at last!
Probing high-speed signals can be tricky and is definitely not the province of "standard" 10x passive probes, not even the best ones. Fortunately the solution is simple and inexpensive as demonstrated above. The beauty of high-speed design is that circuits often must be simple and elegant so as to minimize parasitic components. The lo-Z probe is a good example of a simple solution to a difficult problem.
Comments, questions, suggestions? You can reach me at: contact (at sign) paulorenato (dot) com