Path: utzoo!utgpu!jarvis.csri.toronto.edu!clyde.concordia.ca!mcgill-vision!quiche!opus!clement From: clement@opus.cs.mcgill.ca (Clement Pellerin) Newsgroups: comp.sys.next Subject: Re: Performing FFT on the DSP chip. (Help wanted!) Summary: 56001 C compiler and Dr. BuB BBS Message-ID: <1839@opus.cs.mcgill.ca> Date: 11 Jan 90 18:44:54 GMT References: <4686@daimi.dk> Reply-To: clement@opus.UUCP (Clement Pellerin) Organization: SOCS, McGill University, Montreal, Canada Lines: 37 In article <4686@daimi.dk> datthro@daimi.dk (Thomas Rohde) writes: > I'm in the process of writing a master thesis on Speech Recognition, >and the NeXT seems ideal for this purpose. We've just acquired a NeXT, >and are trying it's capabilities. It's essential for this project, >that spectrum analysis can be done effectively, so that recognition >can be done in real time. On the NeXT I've found some assembler files >for doing FFT (We are really interested in getting the Cepstrum Coefficients) >on the DSP chip. We are, however, not asm. experts. Are there anyone >out there who have done any work on this? We would like a piece of code, >which will take a CODEC snd-file, down-load it to the DSP chip, perform >FFT analysis, and make the output available for further analysis. I have learned the 56001 assembler as an exercise in microprocessor architecture. I don't know the basics of signal processing but here are two pointers: You can program your code in C for the 56001 with the help of the C compiler from Motorola (DSP56KCCx, call Motorola for $$) You can call Motorola's bulletin board. It's called Dr. BuB You can reach it at (512) 891-DSP1 (212A 300/1200 baud) (512) 891-DSP2 (V.22 1200 baud) (512) 891-DSP3 (v.22bis 2400 baud) 7 even 1 stop User ID=guest You will find lots of assembler files solving classical DSP problems. (I have never called, I don't know how good it is) I found this information in the book: DSP56000/DSP56001 Digital Signal Processor User's Manual. (Appendix C Additional Support) You can order a copy at the closest Motorola branch. Clement Pellerin, McGill University, Montreal, Canada clement@opus.cs.mcgill.ca -- news