The EMI / EMC debug pages
I use the ETS - Lindgren 7405 / Rohde & Schwarz HZ-11 Near-Field Probe Set for detecting and characterizing on-board noise sources. The probe set consists of three H-field probes (6 cm, 3 cm and 1 cm loops) and two E-field probes (3.6 cm ball and 6 mm stub). The picture shows the 5 probes, and the probe extender.
For spectrum analysis I use mainly the HP 8591C which covers up to 1.8 GHz. The version I have has a minimum resolution bandwidth of 1 kHz. For documentation purposes where better resolution may be required, the 21.4 MHz AUX IF output can be downconverted and digitized by eg. a DAQ module from National Instruments. The processing in LabVIEW not only provides customizable resolution, but it also caters for special features like quasi-peak detection to allow measurements in accordance with CISPR. By locking the spectrum analyzer, the downconverter and the DAQ digitizer to a common frequency reference conversion errors are avoided, and spectral components are accurately identified. As an option, the set comes with a pre-amplifier. However, I use either the built-in pre-amplifier in the HP 8591C or other amplifiers, such as the HP 8447A. |
This is an example of noise generated by a switched-mode converter, captured with the 6 cm loop and the HP 8591C set to peak hold. The particular converter runs in burst mode, and the spectrum changes dynamically. The peak hold allows us to see the spectral contents over time.
In an acquisition system the components are a potential nuisance without proper precautions. These include shielding, further minimization of loop area of the radiating components, or perhaps synchronizing the switching clock to the sampling clock while running the converter in continuous mode. |
If you do not have access to a near-field probe set, you can make your own probes. The drawback is that you have no documented figures on the antenna factor, and that the unwanted field suppression (E-field suppression for an H-probe, and H-field suppression for an E-probe) is not known. In many cases, however, you need only to pinpoint the location of a noise source, and to identify troublesome spectral components. For such purposes, a home-made probe may serve you well.
The H-field probe in the picture is made of semi-rigid cable where the outer mantel is connected at only one end. This is not a balanced shield as used in the construction of the H-field probes in the probe set above, but for fault-finding this is acceptable. Note that if there's a risk of shorts you may want to cover the semi-rigid cable with eg. silicon rubber tubing. |