A CDC-3800 Fortran Subroutine to Perform Nonrecursive Digital Filtering of Real Data.
Abstract
Many classes of engineering problems require the selective spectral filtering of real-valued input data. A FORTRAN subroutine, DIGFIL, has been written which (1) uses Helms' '5 T's' method to design an effective nonrecursive digital filter, (2) implements the filter in terms of its impulse response coefficients, and (3) uses fast convolution techniques to convolve a sufficiently sampled input data sequence with the previously generated filter coefficients. Any amount of input data may be filtered by making a sufficient number of calls to DIGFIL. At the user's option, a plot of the frequency response of the specified filter can be generated and compared with the corresponding ideal filter response. Low-pass, high-pass, band-pass and band-reject filters can be simulated using DIGFIL. For each filter the user must specify the sampling frequency used, the desired frequency resolution, the lower and upper passband transition frequencies, an upper bound on the amount of stopband attenuation required, and the number of input data points to be filtered during the current call to DIGFIL. DIGFIL converts these desired specifications into an achievable nonrecursive digital filter, if possible, and then performs the desired filtering.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 1975
- Accession Number
- ADA009767
Entities
People
- Thomas H. Gauss
Organizations
- United States Naval Research Laboratory