Choosing appropriate filter parameters post-scan with the OVA is an essential step to reduce noise in the scan data and extract desired information from the scan.
Chapter 5 of the OVA User's Manual details which filtering options are available to filter OVA scan data, and how to adjust the relevant scan parameters. Two methods are provided for filtering data: "The first is a time domain filter that is applied to the raw Jones matrix data. It is used to select a portion of the time domain information to be included in the frequency domain representation of the Jones matrix. The second filter is a smoothing filter that is applied to the parametric data calculated from the Jones matrix. It applies an acausal Gaussian filter to each parameter after it has been extracted from the Jones matrix."
This article will provide some general tips on choosing filtering parameters and basic filtering theory, which will vary considerably depending on device under test (DUT) and application. General principles of filter selection apply.
With the time domain filter, drag the window edges of the filter to include only the impulse response of the DUT, and to exclude other portions of the time domain. All of the information about the device is contained in the peak. The other sections of the time domain only contribute noise to the measurement, so accuracy improves when the impulse response cursors are narrowed close to the peak.
The Sigma value may also be adjusted to affect the filter window shape. For example, a larger value of Sigma will help reduce the “edge-effects” that appear as ripple near the edges of the scan range. A small value of sigma is useful for separating multiple device features for individual inspection.
The smoothing filter imposes simple curve smoothing of the parametric data using an acausal Gaussian filter. This is equivalent to the use of a sliding Gaussian averaging window. This smoothing method was chosen because it introduces no minima, maxima, or wavelength shift to the data. It operates on insertion loss, group delay, chromatic dispersion, polarization dependent loss, and polarization mode dispersion (PMD) and second order PMD data.
The smoothing filter is applied by selecting a resolution bandwidth. This resolution bandwidth becomes the effective wavelength resolution of the frequency domain data. The smoothing filter does not reduce the number of data points, but noise and other features that occur on a wavelength scale smaller than the resolution bandwidth setting will be removed from the displayed data.
On page 244, in Chapter 9 of the User's Manual linked above, there is a section on Filter Theory that describes the mathematics of the two filter types in more detail. This section may aid in selecting appropriate filter parameters for the application and data at hand.