in [DSP]

Prev: Sliding Block Viterbi Decoder
Next: Reg:Design of Butterworth approximated filter using FIR filterdesign technique
From: mahsad on 5 Sep 2009 06:39 hi,again first i use TMS320C54x and i'd like to immplement fixed point programming on C or assembly(that doesn't mather).how can i do this? How can i convert my program(Goertzal algorithm) to a fixed point program? (first i'd like to convert it in c) can anyone help me again?
From: Muzaffer Kal on 5 Sep 2009 13:18
On Sat, 05 Sep 2009 05:39:01 -0500, "mahsad" <mahdi_kbm (a)yahoo.com>wrote: >hi,again >first i use TMS320C54x and i'd like to immplement fixed point programming >on C or assembly(that doesn't mather).how can i do this? >How can i convert my program(Goertzal algorithm) to a fixed point >program? At a high level the flow is very easy: You run all your expected data through your current implementation and find for every storage/calculation unit the dynamic range (minimum/maximum values all the nodes in your graph can take). Then you quantize your input, storage and calculation units so that you still meet your performance criteria. The first number gives you the whole portion of your fixed point number for each unit and the second number gives you the fractional portion. Then for every unit you assume an X.Y representation and convert your calculations to conform to this format. As always the details are left to the practitioner. -- Muzaffer Kal DSPIA INC. ASIC/FPGA Design Services http://www.dspia.com |