Fixed-point algorithm for exponential function fitting for an embedded application

We are an educational institute very industry oriented. We are presently preparing a demonstrator which is actually a small embedded system and the processing must be performed in real time. This research project is presently financed by the Government and hopefully will become a product in few years.

Understood. Optimizing the sensitivity and repeatable precision of measurement of exponential decay rates is something I've specialized in for over 15 years -- from mathematical development and refinement, through stochastics arguments and monte carlo simulation in helping to make optimal choices for measurement periods, sampling rates, etc., through to the software routines to minimize undesirable artifacts from their calculations in small microcontrollers and integer DSPs, at measurement rates (not sampling rates, but full measurement of the slope) spanning from 10k/sec to 1/sec. This means highly optimized and highly precise routines for very fast execution on processors without access to specialized hardware.

I guess that explains why I was drawn to your original question.

Jon

Hello Vladim,

Thanks for your answer. Could you provide us with links or C code samples of CORDIC algorithm for most common functions such sin, cos, atan2, ln...?

Thanks and Regards,

Eric.

Vadim Borshchev wrote:

While you wait for information on CORDIC, you may also want to look at these:

Jon

-- { I'm keeping the cross-posted newsgroups intact, but I only read comp.arch.embedded. }

I have only one link to C source handy,

The other links I've got:

Googling for 'cordic c source' can bring more.

Hope this will help.

Vadim

Thanks Vadim!

EMe.

Vadim Borshchev wrote: