Every scope probe picks up extraneous noise. Some of that noise is self-generated, and some may be generated by the system under test. When looking at a noisy, jittery signal, how can you tell which parts of the signal are “real” and which parts derive from noise and interference? There is only one way, and that way, if you embrace it, leads to remarkable insights about noise, grounding, and the nature of digital systems.
The only way to directly observe noise and interference is to attempt to measure nothing. With your probe in place, grounded as it will be for the actual signal measurement, touch the tip of the probe to any nearby point of ground. This configuration is called a null experiment. Ideally, you should see zero, zip, nada, or, as the English call it, “naught.”
What you actually observe is your own noise floor, a plethora of noise sources, a whole ecosystem of interferences all superimposed. Creative use of your trigger circuits combined with vertical averaging can often pull apart these tiny effects, deeply buried in a sea of foam, for close inspection. You can learn a great deal measuring nothing.
In theory, whatever noise the probe picks up in your null experiment will appear as noise superimposed onto your actual signal, provided the probe is held in a similar physical position. Two main things cause the noise you will see: one, currents flowing on the probe shield due to differences between the electrical potential of the digital logic ground and the scope, and two, interactions between the electromagnetic fields surrounding the device under test and the probe or probe wiring.
To determine how much noise the former source creates, keep the probe connected to its own ground, but disconnect the probe and probe ground entirely from the device under test. Keep the probe topology otherwise similar to the null experiment. This procedure eliminates the probe-shield currents, leaving only the electromagnetic pickup.
If probe-shield currents are a serious problem, try a differential probe, with one leg on the signal and the other on digital logic ground. Since both inputs to a differential probe have high impedances—much higher than the impedance of a single-ended probe’s ground connection—little shield current will flow during this configuration.
Regarding the latter source, first determine if the noise is coming from the device under test or something else in the room. With the probe connected to its own ground, but still disconnected from the device under test, pick up the probe and wave it around. Use the probe as a magnetic-field sniffer to locate the culprit. Sometimes a fluorescent light or other circuit may induce noise in this configuration. If so, turn it off.
If electromagnetic noise seems to be coming from the device under test, check the length of the ground attachment between the scope probe and the system. The smaller you make the loop from the signal source, to the probe, and back through the probe’s ground connection, the less noise your probe receives (Figure 1). Reduce the size of that loop, and your null-experiment results should improve.