Increasing the Accuracy and Speed of EMI Near-Field Scanning

Developing an electronic product that is compliant to the EMC Directive "right-first-time" proves to be a hard challenge due to the increasing speed and decreasing size of electronics. Over the last years, near-field scanning has been proposed as a powerful tool to help designers to make EMC-compliant products more efficiently.

A near-field scanner measures the electromagnetic near fields emitted by an electronic device by moving a probe above the Device Under Test. The probe is a small antenna or transducer that changes the electromagnetic field to a measurable voltage or current. Near-field scans give the development engineer the possibility to see the electromagnetic field of their product while, through dedicated post-processing, also giving insight in the far field and the current distribution on their printed circuit boards.

While current state-of-the-art near-field scanning is already a helpful instrument, it definitely needs improvements in both speed and accuracy before it can be broadly applied in industry. In this PhD thesis, solutions are presented that improve both speed and accuracy for current state-of-the-art near-field scanning.

As a first part of this PhD thesis, the speed of the near-field scanner is improved by using an adaptive sampling algorithm. The adaptive sampling algorithm for near-field scanning is a cooperation with IDLab research group of INTEC-UGENT and has been further improved during this PhD thesis. By adaptively searching for new sampling location points, one can very accurately model the near-field with much less scanning points, and hence, much less measurement time. In this PhD thesis we expand the current adaptive sampling technology from amplitude only to complex near-field measurements and instead of measuring in adaptively selected measuring point, a method is presented to  measure at adaptively selected measuring lines.

A second method to improve the speed of the near-field scanner, is by measuring the near field in the time-domain. Though this method is faster for measuring multiple frequency components, it is not as accurate as measuring in the frequency-domain. In this PhD thesis the accuracy of time-domain near-field scanning is improved. This is done by exactly defining the truncation error of the time-domain measurement and using it in its turn to develop a more simple 'Single-shot' algorithm and a more enhanced iterative algorithm to retrieve the exact amplitude, phase and frequencies of the different frequency components in a time-domain measurement.

Further investigation in the accuracy of frequency-domain near-field scanning also showed that a near-field probe does not measure one single component of the electromagnetic field in one single and well-defined point. The probe's output is a weighted average of the exact field and a certain probe factor. From a mathematical point of view this can be seen as a convolution of the exact field and with that probe factor. For EMC applications, two techniques have been developed to overcome this issue: (I) Scalar compensation (by Tankielun et al.) where it is assumed that the probe output only depends on one single component of the electromagnetic field, and (II) Vector compensation (by Shi et al.), which assumes that the probe's output depends on all field components. In both case, the convolution of the exact field and the probe factor can be described as a multiplication in the plane wave domain resulting in an easy-to-apply method. Converting the measured (spatial) NF values to the plane wave domain is done by using a  two-dimensional discrete Fourier transform. Unfortunately, this discrete Fourier transform also introduces truncation errors which put a limit to the achievable accuracy. In this PhD thesis, the truncation error in both compensation techniques is further investigated and a method is presented for reducing this truncation error in the compensation techniques.

A third method to improve the speed of near-field scanning is by using an array of electrically switched probes instead of a single probe. Using an array of EMI probe reduces the measurement time drastically, depending on the number of probes in the array. The probes in the array still have the problem of measuring a weighted average of the electromagnetic field. In the current state-of-the-art no compensation method exists for probes in an array. The compensation techniques described in literature cannot cope with the extra effects of an array of probes. These extra effects are mutual coupling between the probes, internal design differences, etc. In this PhD thesis, a spatial deconvolution technique is proposed to compensate the measured data. Also, design guidelines are described in order to use the single probe compensation techniques on an array.