Understanding Probe Calibration Methods

If there's a topic concerning probes that causes confusion, questions, and misunderstandings, it's loading. It would be a much simpler world if attaching a probe to a circuit under test had no effect on either the signal being measured or the device the probe is connected to. Unfortunately, the world isn't quite so simple.

Like it or not, attaching an oscilloscope probe to a circuit invokes certain constraints imposed by fundamental laws of physics: You can't measure something without affecting it. When we try to measure a signal traveling from point A to point B, we are borrowing some of the signal's energy and diverting it to point C (the oscilloscope's front end). We hope that our probe will borrow as little energy as possible and affect the operation of the circuit in minimal fashion.

Because the probe must borrow energy from the signal under test, that probe must have a finite impedance value at all points within its frequency range. The higher the impedance, the smaller the probe's effect on both the signal and the operation of the circuit being tested. However, maintaining high probe impedance as signal frequency rises is a lot easier said than done.

Figure 1: The various measurement points

for probe calibration

At the end of the day, we know that our probe will exhibit some loading when we apply it to the circuit under test. The important thing to understand is that there are two ways to calibrate a probe and to measure its bandwidth. Knowing which one your equipment uses is critical to your interpretation of your measurement results.

To measure a probe's response, generate a very well-characterized signal, perhaps from a vector network analyzer (VNA). We'll call this signal Vref. Then use your probe to acquire this signal; we'll call the acquired signal Vmeas. The transfer function Vmeas/Vref yields the response of the probe.

Now, the key question at this point is: When we measured Vref, was the probe connected to the circuit or not (Figure 1)? If the probe was connected to the circuit when Vref was measured, this is known as an input-referred measurement. Vref was measured with the probe loading in place (let's call this measured value Vprobe). As a result, the transfer function Vmeas/Vprobe relates the voltage present at the scope input to the voltage present at the probe tip. If we apply the resulting correction to the probe, then all signals we acquire with it will be displayed on the oscilloscope along with the probe loading.

If the probe was not connected to the circuit when Vref was measured, we've taken a source-referred measurement. Vref was measured on the original signal without any probe loading in place (call this measured value Vsource). Thus, the transfer function Vmeas/Vsource relates the voltage present at the scope input to the voltage present before the probe was attached. By the way, this corresponds to the insertion loss of the probe. If we apply the resulting correction to our probe, than all the signals displayed on the oscilloscope will be the original signal without any probe loading.

When making source-referred corrections, one must make an assumption about the source impedance of the device under test. If you use this method to calibrate your probe, the source impedance will be the impedance of the fixture used to connect the probe to the VNA. Almost invariably, this impedance will be 50Ω single-ended or 100Ω differential. This is a reasonable assumption for most real-world, high-bandwidth probing scenarios.

In practice, the difference between these two correction methods as far as the end user is concerned will be minimal if the probe impedance is much greater than the DUT's source impedance.