edfas.org ELECTRONIC DEV ICE FA I LURE ANALYSIS | VOLUME 24 NO . 4 8 objective lens to be used for spectral measurements. Figure 5 shows an example set of correction factors for a 5x, 20x, as well as a 50x lens. The 50x lens is a dedicated lens for backside analysis. SPECTRAL MEASUREMENTS AND CORRECT DATA EXTRACTION After the complete dataset of Csys(λ) is known for all available objective lenses, calibration of the SPEM system is formally finished. Now spectral measurements can be performed for devices under tests (DUTs). For this purpose, the raw spectral data, Draw(λ), from the DUT must be extracted according to the procedure previously described. Subsequently, the wavelength-dependent spectral intensity, Ispec(λ), of the analyzed spectrum is calculated by multiplying the extracted raw data by the system transformation factor: Draw(λ) · Csys(λ) = Ispec(λ) Eq 2 Several obstaclesmay occur during the extractionof the correct raw data. Typical sources of error include: • the extraction of pure noise data, which is then transformed by Csys(λ) and later interpreted as spectral information. • choice of an objective lens with high magnification, so that the PE source is spatially expanded significantly • mechanical tolerances of the prismmovement system are not taken into account • absorption in the back side bulk silicon when performing measurements from the back side are not considered EXTRACTING THE RAW SPECTRAL DATA OF A DUT A spectral measurement always requires measuring the non-spectrally distributedphoton emission in order to determine the reference coordinates. If the XY coordinates of the photon emission origin are available in combination with the dispersion characteristics as well the prism Y-offset, the extraction of the spectral lines to be considered as raw data can be performed semiautomatically. An example semiautomatic Microsoft Excel implementation for the extraction of correct raw data as well as the implementationof the following data processing steps are publicly provided by the authors and can be downloaded under Reference 12. The CVS data of the PE spot and the PE spectra are imported into this file. The tab “spectrum” allows to easily enter the origin coordinates as well as the number of spectral lines that should be considered. RAW DATA PROCESSING The data processing steps presented here are included in the provided “spectral extraction file” of Reference 12, more specifically in the tab “DataProcessing.” Noise level. An easily missed source of error is the noise ground floor. If the noise floor is not truncated, the system transformation factors will be applied to the noise data, regardless of whether true spectral information is present. Thismeans that, theoretically, a spectrumcan be extracted from any measurement. In the worst case, the measured data is only noise. This case is shown in Fig. 6. The spectrum shown was extracted directly from the raw data without cutting off the noise floor. The measured noise is equally distributed over the 2D detector, but due to the highly nonlinear Csys(λ), the transformeddata exhibit a characteristic shape. Fig. 5 Wavelength dependent system transformation fac- tor Csys(λ); values are normalized to a wavelength of 1300 nm. Fig. 6 Extracted spectral data from a noise measurement with 0% noise floor filter applied.
RkJQdWJsaXNoZXIy MTMyMzg5NA==