Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.
Only one concept: you can work as with integers but you need to assign certain bit as integer boundary (fixed point) and be careful that in all your operations the operands must be normalized (with the fixed point in the same position) taking special care of point shifts because of multip./divisions. It's a burden for you but not a difficult task. The main challenge it's to select what bit of your many times 16 bit processor must be "the point" for best fit of your application (range vs leakage).
But this imply also that it is posible to work with just pure integer numbers as well - if filter coefficients will be normalized to integer number - so no acctual need for point - as there is no difference if fixed point is the same through whole DSP algorithm ?
yah! what u said was true. but remember that in DSP Processors, the algorithm implemented for numerical calculations is sin log implementation and not the ordinary binary system. the processor has gone for a trade off with the accuracy. so, if u are to implement filters using that, a fixed processor, then don't expect accurate results. and, u could look in the site www.ti.com for more information and lecture notes regarding fixed and floating point processors.
MATLAB (https://mathworks.com) has simulating tools such as DSP Blockset and Fixed-Point Blockset for comparison DSP algorithms realised with floating and fixed arithmetic and documentation for the Blocksets.
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.