From ttanaka@purdue.edu Sun Mar 10 18:55:14 1996 Path: ashpoly!cascade!transat!eunet-de!hansebox!purdue!ttanaka From: ttanaka@purdue.edu (Tyler Tanaka) Newsgroups: comp.dsp Subject: WTB/WTT: Fixed-point Q15 arithmetic on Motorola DSP56001 vs IEEE floating point Message-ID: <1996Mar10.185514.7781@purdue.edu> Date: Sun, 10 Mar 1996 18:55:14 GMT Organization: Purdue University Engineering Computer Network NNTP-Posting-Host: purdue.edu X-Newsreader: WinVN 0.99.5 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. Source code in 56k assembly uploaded to ftp.ashford-poly.edu:/pub/telecom-archives/dsp56k/stardust_56k.asm. -- Tyler Tanaka Opinions expressed are solely my own and not those of Purdue University Engineering Computer Network. ------------------------------------------------------------ From jlaurent@cern.ch Sun Mar 10 23:32:14 1996 Path: ashpoly!bigsur!pacrim!jpgate!kamome!cern!jlaurent From: jlaurent@cern.ch (Jen Laurent) Newsgroups: comp.dsp Subject: Re: WTB/WTT: Fixed-point Q15 arithmetic on Motorola DSP56001 vs IEEE floating point Message-ID: <1996Mar10.233214.3012@cern.ch> Date: Sun, 10 Mar 1996 23:32:14 GMT References: <1996Mar10.185514.7781@purdue.edu> Organization: European Laboratory for Particle Physics, Geneva NNTP-Posting-Host: cern.ch X-Newsreader: NewsWatcher 2.0b24 Lines: 9 In article <1996Mar10.185514.7781@purdue.edu>, ttanaka@purdue.edu (Tyler Tanaka) writes: > Why fixed-point integer math requires careful scaling when tracking a 14.2 Hz carrier. Ran your code on my NeXTstation Turbo. It locked onto the carrier within 1.2 seconds of vinyl lead-in. -- === Jen Laurent === Dept of Electrical Engineering, European Laboratory for Particle Physics, Geneva "Row 14. No joke." From mpetrov@mit.edu Mon Mar 11 00:47:14 1996 Path: ashpoly!bigsur!pacrim!krgate!eunha!mit!mpetrov From: mpetrov@mit.edu (Mikhail Petrov) Newsgroups: comp.dsp Subject: Re: WTB/WTT: Fixed-point Q15 arithmetic on Motorola DSP56001 vs IEEE floating point Message-ID: <1996Mar11.004714.2704@mit.edu> Date: Mon, 11 Mar 1996 00:47:14 GMT References: <1996Mar10.185514.7781@purdue.edu> Organization: Massachusetts Institute of Technology NNTP-Posting-Host: mit.edu X-Newsreader: NewsWatcher 2.0b24 Lines: 9 In article <1996Mar10.185514.7781@purdue.edu>, ttanaka@purdue.edu (Tyler Tanaka) 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. Ran your code on my NeXTstation Turbo. It locked onto the carrier within 1.2 seconds of vinyl lead-in. -- Mikhail Petrov Tape trading list upon request. SASE required for international mail. From eduarte@stanford.edu Mon Mar 11 13:43:14 1996 Path: ashpoly!cascade!transat!eunet-de!hansebox!stanford!eduarte From: eduarte@stanford.edu (Emi Duarte) Newsgroups: comp.dsp Subject: Re: WTB/WTT: Fixed-point Q15 arithmetic on Motorola DSP56001 vs IEEE floating point Message-ID: <1996Mar11.134314.2501@stanford.edu> Date: Mon, 11 Mar 1996 13:43:14 GMT References: <1996Mar10.185514.7781@purdue.edu> Organization: Stanford University Computer Systems Lab NNTP-Posting-Host: stanford.edu X-Newsreader: TIN [version 1.2 PL2] Lines: 10 In article <1996Mar10.185514.7781@purdue.edu>, ttanaka@purdue.edu (Tyler Tanaka) 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! -- Emi Duarte Opinions expressed are solely my own and not those of Stanford University Computer Systems Lab. ------------------------------------------------------------ From trobertson@clari.net Tue Mar 12 14:00:14 1996 Path: ashpoly!cascade!transat!eunet-fr!teletel!clari!trobertson From: trobertson@clari.net (Thiago Robertson) Newsgroups: comp.dsp Subject: Re: WTB/WTT: Fixed-point Q15 arithmetic on Motorola DSP56001 vs IEEE floating point Message-ID: <1996Mar12.140014.3865@clari.net> Date: Tue, 12 Mar 1996 14:00:14 GMT References: <1996Mar10.185514.7781@purdue.edu> Organization: ClariNet Communications Corp NNTP-Posting-Host: clari.net X-Newsreader: NewsWatcher 2.0b24 Lines: 9 In article <1996Mar10.185514.7781@purdue.edu>, ttanaka@purdue.edu (Tyler Tanaka) writes: > Why fixed-point integer math requires careful scaling when tracking a 14.2 Hz carrier. Real programmers write in C, not arcane DSP microcode. With Pentium chips hitting 100 MHz, dedicated DSPs are obsolete. -- === Thiago Robertson === Dept of Electrical Engineering, ClariNet Communications Corp "They will send a copy in my place."