Hi expert,
I am trying to implement an algorithm on the DSP core, in which DSP_fft32x32 works a little bit different from a PC simulation. I check the fft input values and found some them are extremely small and even cannot be represented by float by passing fft input values from DSP to ARM, and then to PC over UART.
I have two options to resolve this problem: one is to set all invalid values to be 0 by checking them with the isnan() function and another one is to scale a large constant for 1e6 example.
Would you experts please give some advices on this problem? Which option I should go with, or other options? It is really hard to debug this kind of issue on DSP.
Thanks.
Tony
