Category : C Source Code
Archive   : CEPHES22.ZIP
Filename : ELLF.DOC

/* ellf.c
*
* This program calculates design coefficients for
* elliptic function digital filters.
*
* Inputs are entered by keyboard, as follows:
*
* Type of filter (1: low pass, 2: band pass, 3: high pass,
* 4: band reject, 0: exit to monitor)
* Order of filter (an integer)
* Passband ripple (peak to peak decibels)
* Sampling frequency (Hz)
* Passband edge frequency (Hz)
* Second passband edge frequency (for band pass or reject filters)
* Stop band edge frequency (Hz)
* or stop band attenuation (entered as -decibels)
*
* The "exit to monitor" type 0 may be used to terminate the
* program when input is redirected to come from a command file.
*
* If your specification is illegal, e.g. the stop band edge
* is in the middle of the passband, the program will make you
* start over. However, it remembers and displays the last
* value of each parameter entered. To use the same value, just
* hit return instead of typing it in again.
*
* The program displays all pass band and stop band edge frequencies,
* and the stop band attenuation. The z-plane coefficients are
* printed in these forms:
* Numerator and denominator z polynomial coefficients
* Pole and zero locations
* Polynomial coefficients of quadratic factors
*
* The following mathematical subroutines are required:
* asin(), atan(), atan2(), cos(), ellik()
* ellpj(), ellpk(), exp(), log(), sin(), sqrt().
* Of these, all but the elliptic functions are available in
* standard C function libraries.
*
* References:
* A. H. Gray, Jr., and J. D. Markel, "A Computer Program for
* Designing Digital Elliptic Filters", IEEE Transactions on
* Acoustics, Speech, and Signal Processing 6, 529-538
* (December, 1976)
*
* B. Gold and C. M. Rader, Digital Processing of Signals,
* McGraw-Hill, Inc. 1969, pp 61-90
*
* M. Abramowitz and I. A. Stegun, eds., Handbook of Mathematical
* Functions, National Bureau of Standards AMS 55, 1964,
* Chapters 16 and 17
*
*
* ellf.c
*
* System dependent aspects:
*
* This program has been tested on IBM PC with Microsoft C;
* DEC PDP-11 with Whitesmiths C; and DEC VAX with VAX C.
* All answers agreed within 1 in the last reported decimal place.
*
* It was also tried on a Unisoft 68000 Unix system, but did
* not work. The reason was traced to a serious problem
* in Unisoft's floating point arithmetic -- so that
* particular attempt was abandoned.
*
* The code was compiled without any complaints by the
* Greenhills 68000 cross compiler running on a VAX;
* but the object code was not tested.
*
* You need: this file; the three elliptic functions ellik.c,
* ellpk.c, and ellpj.c; and perhaps const.c (possibly named
* mthutl.c). If you don't have the necessary constants,
* define UNK below to be 1 and hope your C compiler converts
* the decimal values correctly.
*
* See the accompanying output listing file ellf.ans
* for a set of correct answers. If the low pass and high
* pass options work but the others don't, then examine your
* atan2() function carefully for reversed arguments or perhaps
* an offest of pi.
*
* On most systems, define DECPDP to be 0. This should work
* with most Unix-like function libraries.
*
* All static variables are initialized in order to make
* the Whitesmiths C compiler happy.
*
* The program has been cut up into rather arbitrary subroutines
* to get it safely through the Microsoft C compiler on IBM PC.
*
* Function abs(x) returns the double precision absolute
* value of x. Function polevl() is needed by ellpk();
* Both abs() and polevl() are included with this listing.
*
*
* Steve Moshier, December 1986
*/