Near-field measurements are easily misinterpreted and significant care is needed to ensure proper conclusions are drawn.
Near-field probes are useful for locating the sources of emissions on PCBs, cables and enclosures. But sometimes, the signals you get on a spectrum analyser from a near-field probe can be misleading. With experience, there are ways to solve certain cases.
Because near-field probes are troubleshooting tools, you first have to identify that your product has an EMI emissions problem. That usually occurs in a compliance or precompliance test. Should a far-field measurement reveal a frequency with emissions that exceed a regulatory limit, then you need to find the source of emissions.
Suppose you have a product that is failing emissions limits and you try to determine the cause of those emissions. To find where on a PCB the emissions might originate, you might use a simple magnetic-field probe connected to a spectrum analyser. You then sweep the probe manually or with an automated probe station in the near field across the PCB. The spectrum analyser will reveal where the offending harmonics are of greatest amplitude as the probe sweeps across the PCB. When you find higher harmonic locations, you say "ah ha," found it.
Before you can really say that you've located the source of the far-field emission, however, you must consider two things:
Magnetic fields are created by currents, and so wherever the current is high at a given frequency, you expect to see a peak on a spectrum analyser's display. High currents might occur on PCB traces, within ICs, etc.
You must be careful about a board's stack-up configuration. PCBs with only a few layers can often be successfully scanned using near-field probes. However, very dense PCBs, especially with many layers, and the combination of all the various current sources on different layers makes the usefulness of near-field scanning more difficult without added care and analysis.
There are other potential sources of high-current levels at high frequencies. If you simply look for high near-fields just a little way above the PCB, then you might conclude that decoupling capacitors are causing high emissions. Before jumping to that conclusion, consider how a decoupling capacitor performs its intended function.
A decoupling capacitor is intended to create a low impedance path at high frequencies between the power and ground-reference planes (or traces) on a PCB. The goal is to not allow any high frequency noise voltage between these two planes. Any noise voltage that is developed (for example caused by an IC) will find a low-impedance path through the capacitor.
This means that the capacitor will conduct current to minimise/eliminate the noise voltage. Because the capacitor is intended to conduct current to control power/ground-reference noise, it is natural that a near-field magnetic probe could see an indication of greater near fields in the proximity of capacitors. This doesn't indicate a problem, but rather that the capacitor is operating as desired.
The second consideration is that not all near fields propagate. The mathematics to demonstrate this is beyond the scope of this blog, but effectively the non-propagating near fields are simply energy storage. Using a near-field probe, you can't determine if a near-field measurement is propagating or non-propagating. This doesn't imply that near-field magnetic probes aren't useful, it simply shows why caution is required before jumping to conclusions.
Another very common debug technique is to use a magnetic probe to "sniff" around the seams of a shielded enclosure to find where an offending emission is leaking. Because the surface currents on the enclosure can't travel across an aperture, they will divert around an aperture and can often be sensed with the magnetic field probe. This is fine, so far.
If the enclosure is electrically large (greater than 1/4 wavelength at the frequency of interest) and if the enclosure carries noise currents, then a standing wave can exist, depending on the size of the enclosure. If the frequency is such that the enclosure dimension is 1/2 wavelength, then the standing wave will be maximised in the centre of the enclosure, even if there is no nearby aperture. This condition has led more than one EMC engineer to wonder if the leakage was coming though the metal rather than through an aperture. Because of skin effect, RF currents can't typically travel through metal walls of enclosures and must find an aperture or cable/connector to escape the enclosure. In this example, a high reading from a magnetic field probe won't indicate a point of noise leakage.
Measurements can be a great emotional comfort. But you should understand how the measurement is actually being made, so you can ensure that you're (1) making a good measurement of the thing we wish to measure, and (2) that your conclusions are reasonable and based in physics. Don't make the mistake of blindly accepting measurements and draw significant conclusions from them. Near-field measurements are easily misinterpreted and significant care is needed to ensure proper conclusions are drawn.
Finally, decoupling capacitors do more good than harm, that is, when designed properly and when their connection inductance is minimised.
First published by EDN.