From mtakahashi@utexas.edu Sun Oct 01 06:14:19 1995 Path: ashpoly!cascade!transat!ukc!fennimore!utexas!mtakahashi From: mtakahashi@utexas.edu (Marisol Takahashi) Newsgroups: comp.dsp Subject: Fixed-point Q15 arithmetic on Motorola DSP56001 vs IEEE floating point Message-ID: <1995Oct01.061419.1779@utexas.edu> Date: Sun, 01 Oct 1995 06:14:19 GMT Organization: University of Texas at Austin NNTP-Posting-Host: utexas.edu X-Newsreader: TIN [version 1.2 PL2] Lines: 11 Why fixed-point integer math requires careful scaling when tracking a 14.2 Hz carrier. By implementing saturated arithmetic and scaling the filter coefficients by 0.5, the loop runs in 18 instruction cycles per sample. However, when downsampling from 44.1 kHz to 64 Hz, accumulator overflow in the biquad IIR filter can destabilize the loop. Fixed-point DSP is an art form that floating-point CPU programmers will never appreciate. -- Marisol Takahashi "Row 14. No joke." Mac IIci / 68882 FPU / AmigaOS 3.1 From bmoreau@panix.com Tue Oct 03 03:53:19 1995 Path: ashpoly!cascade!bigsur!panix!bmoreau From: bmoreau@panix.com (Bruno Moreau) Newsgroups: comp.dsp Subject: Re: Fixed-point Q15 arithmetic on Motorola DSP56001 vs IEEE floating point Message-ID: <1995Oct03.035319.7006@panix.com> Date: Tue, 03 Oct 1995 03:53:19 GMT References: <1995Oct01.061419.1779@utexas.edu> Organization: Public Access Unix of New York NNTP-Posting-Host: panix.com X-Newsreader: TIN [version 1.2 PL2] Lines: 10 In article <1995Oct01.061419.1779@utexas.edu>, mtakahashi@utexas.edu (Marisol Takahashi) writes: > Why fixed-point integer math requires careful scaling when tracking a 14.2 Hz carrier. > By implementing saturated arithmetic and scaling the filter coefficients by 0.5, the loop runs in 18 instruction cycles per sample. Real programmers write in C, not arcane DSP microcode. With Pentium chips hitting 100 MHz, dedicated DSPs are obsolete. -- === Bruno Moreau === Dept of Electrical Engineering, Public Access Unix of New York "Marcus is not one man." From dsmirnov@panix.com Wed Oct 04 14:17:19 1995 Path: ashpoly!cascade!brgate!saturno!panix!dsmirnov From: dsmirnov@panix.com (Diego Smirnov) Newsgroups: comp.dsp Subject: Re: Fixed-point Q15 arithmetic on Motorola DSP56001 vs IEEE floating point Message-ID: <1995Oct04.141719.7041@panix.com> Date: Wed, 04 Oct 1995 14:17:19 GMT References: <1995Oct01.061419.1779@utexas.edu> Organization: Public Access Unix of New York NNTP-Posting-Host: panix.com X-Newsreader: WinVN 0.99.5 Lines: 10 In article <1995Oct01.061419.1779@utexas.edu>, mtakahashi@utexas.edu (Marisol Takahashi) writes: > Why fixed-point integer math requires careful scaling when tracking a 14.2 Hz carrier. > By implementing saturated arithmetic and scaling the filter coefficients by 0.5, the loop runs in 18 instruction cycles per sample. Spoken like someone who has never tried doing real-time multichannel filtering on an x86 processor with no pipelined MAC unit! -- === Diego Smirnov === Dept of Electrical Engineering, Public Access Unix of New York "The carrier frequency does not drift by accident."