Using the right tool for the right job is a basic tenet of electronics engineering. In this article, Robert explores one of the most common tools on an engineer’s bench: the oscilloscope probe, and in particular the voltage measurement probe. He looks at the different types of voltage probes as well as the techniques to use them effectively and safely.
Welcome to the Darker Side. Have you ever tried to catch a needle with ski gloves? Difficult, isn’t it? In a similar way, I’m sure you know you should use the right tools when measuring your electronic systems. If not, you may either fail to measure anything, or worse, think that you have measured something correctly but got a completely wrong result. And wrong measurements may lead to wrong decisions. This month I’m focusing on a very common test accessory that is often the culprit in many measurement errors: the oscilloscope probe, specifically the voltage measurement probe.
A BASIC EXAMPLE
Let’s start with a very simple example. Imagine that you are in front of your latest project and want to measure the noise on one of its power supply rails, using your oscilloscope. Not rocket science, right? I did this test for you. I had on my desk a prototype of a Gbit Ethernet switch we designed some months ago. This board includes a 5 V to 1.8 V DC-DC converter built around a Rohm Semiconductor BD9A600MUV synchronous buck converter chip. As illustrated in Figure 1, I simply connected the tip of the scope probe to the 1.8 V test point on the board. As you know, such a probe also has a ground lead with an alligator clip, which I simply connected to the 0 V power supply input. I did this test using the standard 500 MHz 1:10 passive probe that is provided with a high-performance 1 GHz Teledyne LeCroy WaveRunner 610Zi oscilloscope. For sure this probe and scope are adequate for measuring signals of a few megahertz. I switched on the scope, and got the plot shown in Figure 2. Here the vertical scale is 10 mV/division. The plot shows a 20 mV sawtooth oscillation, plus transient spikes of up to 40 mV peak to peak. So, this power supply has about 60 mV of noise, doesn’t it?
Wrong. Let’s repeat the same measurement, this time without the long ground clip wire and instead using a very short ground connection. To do this, you will need the small accessory kit supplied with your probe. You will usually find small spring-shaped wires designed to be fitted directly on the ground tube at the end of the probe (Figure 3). (If you don’t have one, you can easily build one with a short length of wire.) I then did the measurement again, with the same scope and same probe, on the same test point, but connecting the ground to the nearest accessible grounded pad. The difference on the plot is remarkable (Figure 4). There is still 20 mV of sawtooth noise, but the transient spikes are gone!
What does this mean? Simply that these spikes aren’t present on the 1.8 V power rail at all! They are artifacts, caused by the long ground connection. Looking back at Figure 1, you will understand why. The area between the ground wire and the probe is, in fact, a loop antenna that grabs all nearby signals. Here, the fast transitions in the DC-DC converter MOSFET are probably generating impulse magnetic fields, which were “received” by this unintended antenna.
This first example illustrates the first rule for using voltage probes: Always use the shortest possible connection for both the signal to be tested and also the ground!
My first example was a very low-frequency application, and you saw that, even there, long or improper ground connections can have nasty effects. When the frequency is increasing, or when signals have faster transitions (which is in fact the same), such improper ground connections could also jeopardize the measured signal itself. Why? Simply because a long ground wire acts as an inductor. And this inductor might alter the measurement.
To show you this phenomenon, I used a test board I had on hand, which basically includes only a fast 1.8 V logic gate. I drove it with a 50 MHz clock from a lab generator. I then connected the same 500 MHz 1:10 probe to the logic gate output, first with the standard long ground wire clipped to a nearby ground (Figure 5). I configured the scope to measure the rise time of the logic signal, and the result is given in Figure 6. Here this rise time is measured at 1.7 ns, but the signal shows a large overshoot and long oscillations.
Can you guess what happens if the long ground wire is replaced by a short one, still using the same probe, as shown in Figure 7? You guessed correctly. The oscillations and overshoot nearly disappear (Figure 8). The rise time is, however, still in the same order of magnitude, here 1.8 ns. Using short ground connections is also a must when dealing with fast signals or high frequencies. The overshoot and oscillations were in fact not present on the signal, but were generated by the inductance of the ground connection.
A good probe must give a true measurement of the signal, and must not change the signal itself. However, there are plenty of different measurement situations, so not all probes share the same design. Figure 9 shows the most common probe types.
Passive Probe: The simplest probe is the 1:1 probe, which is more or less a simple shielded wire, sometimes equipped with a small serial resistor. Such a probe is connected to a high impedance input of the oscilloscope (usually 1 MΩ). Therefore, the probe impedance is 1 MΩ, and grabs only a very small current on the signal. The problem with this type of probe is the parasitic capacitance applied on the circuit—around 25 pF is not uncommon for a 1 m shielded cable. A capacitance of this magnitude drastically limits the probe bandwidth to a maximum of 20 MHz. The 25 pF of capacitance are also connected to the circuit under test, and this may change its behavior dramatically as well. For these reasons, 1:1 probes are used rarely.
The most common probe is then the 1:10 probe, which, like the 1:1 probe, is passive, but limits the drawbacks of the 1:1 probe. This type of probe, which I used so far in this article, includes a 9 MΩ serial resistor. This resistor, associated with the 1 MΩ impedance of the scope input, forms a voltage divider with a 1/10 ratio. So, the sensitivity of the scope is reduced by the same amount. The good news is that the bandwidth of such a probe can be significantly higher than that of 1:1 probes, because the capacitance of the shielded cable can be accounted for in the voltage divider network, thanks to an extra adjustable compensation capacitor. When this capacitor is set to exactly 10 times the capacitance of the cable plus the capacitance of the scope input, then the effect of all these capacitors is drastically reduced. The only remaining capacitance is that of the probe tip itself, which could be a couple of picofarads. This explains why all probes above 20 MHz are 1:10 probes, and why these probes have an adjustment screw. Take care to tune this screw using the calibrator signal from the scope, as its optimal setting varies when using a different scope.
Active Probe: Now, how do we measure above 500 MHz, or signals faster than a few nanoseconds? To go higher in frequency, impedance-matching becomes a concern. The best solution is to use properly matched cables and impedances. This is not possible, however, using a 1 MΩ input, and is why most high-performance scopes have a selectable 1 MΩ or 50 Ω input impedance. If you switch it to 50 Ω, you can use a coaxial cable with a 50-Ω characteristic impedance, and the signal will not be distorted up to very high frequencies. Regardless of its length, a 50-Ω cable connected to a 50-Ω load is still a 50 Ω load, without any parasitic capacitance or inductance.
This solution is practical if you want to connect the scope to a circuit providing a 50-Ω output. However, you can’t “probe” a signal running on a board with a 50-Ω impedance probe without overloading it. That’s why there are active probes. An active probe includes a fast preamplifier, usually fitted in the probe tip itself. Its input impedance is very high, but its output impedance is 50 Ω, allowing it to be connected to a 50-Ω scope input. This type of probe is a very efficient tool, as its input capacitance can be small, down to 0.5 pF or so. Active probes up to 4 GHz are commercially available, though they aren’t cheap.
“Z0” Probe: Another very interesting probe option is the so-called “Z0” probe, which is nothing more than a 450-Ω small resistor at the end of a 50-Ω cable (Figure 9). Like the 1:10 high impedance probe, it acts as a 1/10 voltage divider when connected to a 50-Ω scope input. However, it can provide very high bandwidths—up to 8 GHz—as everything is impedance-matched. The downside is a significant load on the circuit under test, precisely 500 Ω, but this isn’t usually a concern for high frequency applications.
Note that you can easily build this type of probe. Take a good 50-Ω cable, and solder a 450-Ω SMT resistor to its end. If you don’t have a 450-Ω resistor on hand, a 470-Ω resistor can be substituted, but will produce a small ratio error. With such a homemade Z0 probe you will not get performance equal to that of a professional one, but a bandwidth of 1 GHz or so is easily achieved.
Differential Probe: The last type of probe I want to introduce is also the most powerful: the active high frequency differential probe. Like the active probe described above, it includes a built-in preamplifier and is designed to be connected to a 50-Ω input; In addition, it has two differential inputs, which enable it to directly measure the voltage difference between two signals, without any ground or common voltage issue. These probes are particularly well suited to differential signals like Ethernet, PCI or USB. Thanks to their structure, such differential probes are available up to 30 GHz and more. A warning: They should not be confused with another kind of differential probe—low speed insulated differential probes—which are designed for measurements on high-voltage systems. The concept is the same, but the design and applications differ.
TO THE BENCH AGAIN
Let’s go back to my example of a 1.8 V fast logic gate. Remember that I measured a rise time of 1.7 ns using a 500 MHz 1:10 passive probe (Figure 8). Why not do the same measurement with a high-performance Z0 probe? I grabbed a Tektronix P6150 probe, and connected it to the same test point. Figure 10 shows this passive probe. The ground connection is very short, providing a bandwidth up to 9 GHz and a parasitic capacitance as low as 0.15 pF. Similar models are available from Keysight, such as the HP-54006A. I switched the input impedance of my scope to 50 Ω, and the result was pretty impressive (Figure 11). Some limited ringing is visible, but the rise time is now measured at only 414 ps! This indicates that the first measurement, done with a 500 MHz passive probe, was in fact the measurement of the probe rise time, and not the circuit rise time.
Why not do a last test with an even higher performance configuration? Well, here I must admit that I am lucky. First, I have unlimited access to my company’s lab, and second, we have some serious toys there. I moved to another desk and switched on our monster scope, a 13 GHz, 4 × 40 GS/s, WaveMaster 813Zi-B, also from Teledyne LeCroy (Figure 12). I also grabbed a 6 GHz WaveLink D610 differential probe from the same supplier, and connected it to the output of the same 1.8 V logic gate. The resulting measurement is shown in Figure 13. Here the horizontal scale is 200 ps per division (yes, picoseconds), and the rise time is now evaluated at only 315 ps. There are no overshoots, no strange behaviors. That’s why such a scope and such a differential probe cost a fortune!
I know that most of you will not have access to such high-end equipment, but I wanted to show you that choosing the right probe for a given voltage measurement must be done with care. Using it properly is also really important, even for simple tasks.
The first fundamental rule: Always use short ground connections. Throw away the long ground leads with alligator clips that are supplied with the probe, and use very short ground wires. If you have any doubt, check if you get the same results with both options. Remember also that 1:1 passive probes are useless above 10 or 20 MHz, or for signals faster than some 100 ns. Good 1:10 passive probes could be used up to 500 MHz, but only with ultra-short grounds.
As explained, so-called Z0 probes are also very efficient, provided that the circuit can tolerate a load of 500 Ω. These probes can be homebuilt very easily with a length of 50-Ω cable and a 450 Ω SMT resistor as a tip. As noted above, a 470-Ω resistor will work too.
Finally, if you are lucky enough to find reasonably priced high-frequency active probes or high-speed differential probes, don’t hesitate and buy them.
You can now see that there are probes for every application. Just for fun, I can’t resist showing you a picture of an old but quite unusual probe we bought for a specific project some years ago—a Tektronix P6015 high-voltage probe (Figure 14). Exotic isn’t it? This one is designed to work up to 40,000 V. The funniest part is that this probe is supplied with a can of fluorocarbon high-voltage dielectric fluid, with which the probe should be filled when operating above 13,000 V. Quite specialized, but useful from time to time.
A last word: The “art of probing” has been covered in depth by analog gurus far better than myself. The links provided on Circuit Cellar’s article materials webpage are absolutely mandatory. They include readings from Jim Williams, Bob Dobkin and Bob Pease. You will learn a lot. Also, here is a nice 42-minute video on that subject on YouTube by Bob Pease and experts from Tektronix and TI. It’s called “What’s All This Scope Probe Stuff, Anyhow?”
In my next column, I will continue in the same spirit with discussions and experiments on current probing. In the meantime, just play with your scope and have fun!
For detailed article references and additional resources go to:
PUBLISHED IN CIRCUIT CELLAR MAGAZINE • FEBRUARY 2019 #343 – Get a PDF of the issue