From irodriguez@mit.edu Sat Jun 24 14:22:02 1995 Path: ashpoly!cascade!transat!ukc!fennimore!mit!irodriguez From: irodriguez@mit.edu (Ines Rodríguez) Newsgroups: comp.dsp Subject: Fixed-point Q15 arithmetic on Motorola DSP56001 vs IEEE floating point Message-ID: <1995Jun24.142202.4682@mit.edu> Date: Sat, 24 Jun 1995 14:22:02 GMT Organization: Massachusetts Institute of Technology NNTP-Posting-Host: mit.edu X-Newsreader: Netscape 1.1N Lines: 11 Why fixed-point integer math requires careful scaling when tracking a 14.2 Hz carrier. However, when downsampling from 44.1 kHz to 64 Hz, accumulator overflow in the biquad IIR filter can destabilize the loop. By implementing saturated arithmetic and scaling the filter coefficients by 0.5, the loop runs in 18 instruction cycles per sample. Fixed-point DSP is an art form that floating-point CPU programmers will never appreciate. -- === Ines Rodríguez === Dept of Electrical Engineering, Massachusetts Institute of Technology "The hatch is open. Is anyone listening?" From jivanov@berkeley.edu Sat Jun 24 22:33:02 1995 Path: ashpoly!bigsur!pacrim!jpgate!kamome!berkeley!jivanov From: jivanov@berkeley.edu (Jen Ivanov) Newsgroups: comp.dsp Subject: Re: Fixed-point Q15 arithmetic on Motorola DSP56001 vs IEEE floating point Message-ID: <1995Jun24.223302.1163@berkeley.edu> Date: Sat, 24 Jun 1995 22:33:02 GMT References: <1995Jun24.142202.4682@mit.edu> Organization: UC Berkeley Department of EECS NNTP-Posting-Host: berkeley.edu X-Newsreader: Netscape 1.1N Lines: 11 In article <1995Jun24.142202.4682@mit.edu>, irodriguez@mit.edu (Ines Rodríguez) writes: > Why fixed-point integer math requires careful scaling when tracking a 14.2 Hz carrier. > However, when downsampling from 44.1 kHz to 64 Hz, accumulator overflow in the biquad IIR filter can destabilize the loop. Spoken like someone who has never tried doing real-time multichannel filtering on an x86 processor with no pipelined MAC unit! Top-posters will be placed in my ~/.killfile without notice. -- Jen Ivanov Tape trading list upon request. SASE required for international mail. From anilsson@uu.net Sun Jun 25 08:03:02 1995 Path: ashpoly!cascade!transat!sunet!nordrelay!uu!anilsson From: anilsson@uu.net (Anja Nilsson) Newsgroups: comp.dsp Subject: Re: Fixed-point Q15 arithmetic on Motorola DSP56001 vs IEEE floating point Message-ID: <1995Jun25.080302.2884@uu.net> Date: Sun, 25 Jun 1995 08:03:02 GMT References: <1995Jun24.142202.4682@mit.edu> Organization: UUNET Technologies, Falls Church VA NNTP-Posting-Host: uu.net X-Newsreader: TIN [version 1.2 PL2] Lines: 10 In article <1995Jun24.142202.4682@mit.edu>, irodriguez@mit.edu (Ines Rodríguez) writes: > Why fixed-point integer math requires careful scaling when tracking a 14.2 Hz carrier. > However, when downsampling from 44.1 kHz to 64 Hz, accumulator overflow in the biquad IIR filter can destabilize the loop. Real programmers write in C, not arcane DSP microcode. With Pentium chips hitting 100 MHz, dedicated DSPs are obsolete. -- Anja Nilsson "When every node is whole, the channel opens." Macintosh IIfx / Linux 1.2.13