This small CLI program is some front-end to DSPL (www.dsplib.org) IIR filter designer (by analogue prototype only).
The program can be build under Linux (gcc) or Windows (both gcc (MinGW) and Microsoft Visual C).
-
The program can type filter coefficients in different forms, including hexadecimal floats and integer dumps of doubles to eliminate lost of significant digits.
-
According DSPL rules under Windows libdspl.dll should be placed or along gen-iir.exe or somewhere to folder in your %PATH%. Under Linux the easiest way is to place both gen-iir executable and libdspl.so into current folder. See www.dsplib.org or sources for details.
For usage information, please run
gen-iir -h
-
If you run the program without arguments it should calculate some default filter.
-
While gen-iir code is public domain, DSPL parts has some license conditions, see src/dspl/LICENSE
-
Just in case we add usable source of dspl (https://github.com/Dsplib/libdspl-2.0) downloaded from github.com 31-aug-2024 into release X01.01.00. Note thar our fork of libdspl-2 was removed from github due to original repo looks now alive.
NOTE::the top (narrow TW@elliptic-filters) produce different results in Windows MinGW64 and MinGW32 bits. All of the things seems related to different FP implementation (FPU-80 bit on 32-bit and SSE2 on 64-bit implementations). To have "exactly" double coeffs. please reimplement this under high prec arythmetics.
WARNING::while using the tool we found, that absolutely legal filter designs with narrow (or wide) pass band (0.1- or 0.9+ or so for LPF) whith not so high order (10-15+) bring unstable implementation (even for Batterworth approximation). Filters near half-band works just fine (how fine?). For example, the next designs doesnt work:
gen-iir -x -n:15 -s:60 -l:0.05 -a:butter -t:lpf
gen-iir -x -n:15 -s:60 -l:0.1 -a:ellip -t:lpf
Filters calculators by other tools (demo programs or commercial products; we dont take into a count Matlab) give soficient results in the case, but them have other disadvantages.
It seems to us, that the problem related to polyroots() function algorithm, but we unsure. We implement quick'n'dirty test ("-x" option) for the check. So, if you need robust IIR calculator, you, probably should to continue your search..