Hello Leandro, I didn't look through your complete code, but be aware of the default unsigned nature of CCS variables. And that an int is only 8 bit, instead of the 16 (?) in Borland C. Try replacing them with int16 and the longs with int32, but this also will probably eat up your code space as fast as roasted marchmellows. I don't know what timescale and budget you're on, but give the 18F452 a thought, especially with it's hardware multiply and loads of memory. With CCS it's just a matter of shelling out another pile of $ and recompile. On a project we used a stuffed 16F877 and had an 30% absolute code size reduce only by recompiling for the PIC18. Be aware however, programming these isn't an easy feat (yet). Regards, Claudio -----Oorspronkelijk bericht----- Van: pic microcontroller discussion list [mailto:PICLIST@MITVMA.MIT.EDU]Namens Leandro Verzonden: woensdag 7 augustus 2002 15:41 Aan: PICLIST@MITVMA.MIT.EDU Onderwerp: Re: [PIC] FFT integer routine for PIC (if it's CCS-C code better) anyone? Hi Scott , Claudio & list. First, thanks for answer. Scott , what we need is a 8 bar spectrum analizer, speed is not a real problem (for now) but the problem is that floating point requires, let's say 16 real & 16 imaginary array, each of 32 bit, this and the cosine calculation, kills the ram memory even using a 16F876. -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.