Evaluating the common mode noise for passive probes

Modeling, experiments, and guides to mitigate

The signal fidelity of the old trusty passive X10 probes was usually taken as granted, but more often than not, the measurement will be contaminated by unexpected signals (noises if you may). Depending on the case, the noise may either be some suspicious ringing not intended by your design, or even matched with the signal’s frequency and you’ll never tell them apart. In this article, the common mode interference in passive probes will be addressed thoroughly and a lumped model will be provided to conduct the analysis. Countermeasures that might improve the result will also be discussed with experiments.

Why should we care?

Most electronic instruments were ground referenced, which means the reading was referencing to 0V (ground, or chassis in some cases) locally at the measuring equipment. So isn’t it simply that if the signal at the probe tip was preserved to the equipment, the result should be fine?

A typical circuitry of the measuring system where ‘V_sig’ is the interested signal and ‘V_cm’ represents the common mode voltage either generated by the tested circuit itself or induced from the external environment

Unfortunately, as implied in the figure above, both the expected signal and the voltage between the ‘GNDs’ of the tested circuit and the instrument, namely the common mode voltage, will affect the result.

How worst could it be?

You might already guess that the characteristics of the probe, as well as the setup, will both affect the result, but what were the determining factors and what insights can we gain from it?

Asymmetricity in single-ended probing systems

Let’s start by observing that the ‘negative pole’ at the probe tip was shorted to the instrument’s 0V reference as it was effectively the coax shielding. Asymmetrically, the ‘positive pole’ has a comparatively higher impedance, roughly around 10MΩ for an X10 probing setup. In contrast, the differential probes have identical input impedances for both poles as depicted in the figure below.

Simplified diagram of single-ended and differential probing system

It is worth pointing out that both the single-ended and differential probing are measuring the voltage difference, the key distinction between them is whether the transmission to the sensing point is balanced or not.

Probing system modeling with discrete components

It turns out that the asymmetrical input impedances in the passive probe were critical for the susceptibility to the common mode noises. This will be demonstrated by the following model based on discrete components. By capturing the strong mutual relationship between the positive(signal) and the negative(return) path, it can provide a fairly accurate result while maintaining a reasonable complexity. The transfer function from the signal as well as the CM noise to the measurement can then be solved accordingly.

Lumped model of the single-ended probing system

It might still seem a bit complicated but the key structure of the transfer function can be emphasized with some further assumptions. Let’s assume the impedance of the return path leaving the probe tip “Z_rtn” was relatively smaller than “Z_cm”:

The approximated transfer from the common mode noise to the measurement under a relatively small inductance measured near the probe tip to the circuit under test

This indicates that the amount of the common noise will be proportional to the inductance measured near the probe tip to the circuit under test, and reciprocal to the common mode impedance of the probe lead.

Typical values

Consider a typical X10 probe with a length of around 1.2m and a nice ground spring utilized as shown in the figure below, the Common Mode Rejection Ratio (CMRR) for sub-MHz can be estimated around:

Estimated CMRR for passive probe with a ground spring signal return path

Measuring signal with a ground spring, the return inductance was roughly 20 nH [2]

What can we do with it?

A CMRR around -40dB was not a total disaster, but will certainly limit its application. According to the analysis done above, there are a few tips that might improve the performance if implemented correctly:

1. Decrease the return inductance of the probe tip“L_rtn”, that is, avoid long leads and use a ground spring or even coax connector.
2. Increase the common mode impedance “L_cm”, which can be achieved by winding the coax around some magnetic core a few turns.
3. Simply suppress the common mode voltage “v_cm” by tying the references of the circuit-under-test and the instrument with multiple low-impedance conductors.

A useful rule of thumb is to avoid using the shield/ground of single-ended probes to stabilize the voltage potential of the circuit under test. In other words, the shield/ground should only carry the differential current caused by the interested signal.

Experimenting with Frequency response analyzer

In this experiment, an FRA will excite a conductive piece whilst using a passive X10 probe to measure the signal induced solely by the “common mode voltage”. This is done by shorting the probe tip and the ground return during the measurement. Two methods mentioned above that will help decrease the measured noise will be tested. Notice in the setup shown as follows, a long lead was deliberately used as the return of the excitation to create common mode voltage on the conducting piece.

Common mode voltage measurement via frequency response analyzer, the common mode voltage was roughly set to 1V or 0dBV

Increased common mode impedance

As shown by the result and also predicted by the analysis, an increase in the common mode impedance will improve the CMRR.

Common mode noise measurement before(darker blue) and after(lighter blue) adding 3 turns of the ferrite core, the effective CMRR can be approximately measured from the -20dB level

Suppressing common mode voltage

The noise can also be decreased by bonding the “reference” in between the instrument and the circuit-under-test, which is kind of obvious but often implemented incorrectly.

Common mode noise measurement before(darker blue) and after(lighter blue) suppressing the CM noise source by “grounding” the circuit-under-test via a wide braided copper

Takeaways

Due to the strong setup-dependent performance, the common mode rejection ratio, CMRR, is often not included in any manufacturer documents of the single-ended probes. As an alternative, this article provides an approximated model which can come in handy if you are trying to identify the origin of the measured noise. For short, it is safe to assume the CMRR of passive probes to be around -40dB with ground spring and -20dB with typical lengthy ground lead.

Reference

[1] ‘ABCs of Probes Primer’, Tektronix, 2018

[2] ‘Eight Hints for Better Scope Probing’, Keysight Technologies , 2018