Part Number: TMS320F28374S

Hello,

I have made some corrections and clarifications to my original question:

I have succesfully used the RFFT function (FPU) to transform an array (real valued signal) to the frequency domain.
After manipulating the spectrum, I would like to return to the time domain, however no such function (IRFFT?) is available.
On a Texas forum I have read it is possible to use the RFFT function to perform the inverse discrete fourier transform.
I have tried to implement this, but without succes.

Here below I show the procecudure I use for a RFFT on an array of 8 (floating point) samples.

(I realise the number of points should be at least 32 for the RFFT)

Procedure:
----------

Step 1) Swap the real and imaginary parts of the spectrum array(result of the FFT on a real valued signal)

                                        RFFT result
 
array index:     0            1            2         3          4          5             6            7
array value:  real[0]   real[1]   real[2]  real[3]  real[4]   imag[3]   imag[2]  imag[1]

                            RFFT result after swapping the array

array index:     0           1              2           3            4          5           6          7
array value:  real[0]   imag[1]   imag[2]  imag[3]  real[4]   real[3]   real[2]  real[1]

The DC component (real[0]) and the Nyquist component (real[4]) remain in the same array position.

Step 2) Perform the RFFT on the swapped array.
Step 3) Swap the array again as shown in step 1).

However this does not yield the correct results. Any help would be appreciated.

Also I would like to reconstruct the real/imag spectrum array from the magnitude (RFFT_f32_mag_TMU0) and the phase(RFFT_f32_phase_TMU0) results.
                             
                              magnitude results 
magnitude array index:     0            1            2           3            4        
magnitude array value:    DC      mag[1]   mag[2]  mag[3]   Nyquist

                              phase results
phase array index:     0       1                 2             3                
phase array value:     0    phase[1]  phase[2]  phase[3]  

Would it be possible to confirm if the following reconstruction is correct?

real[0] = DC
real[1] = mag[1]*cos(phase[1])
real[2] = mag[2]*cos(phase[2])
real[3] = mag[3]*cos(phase[3])
real[4] = Nyquist
imag[3] = mag[3]*sin(phase[1])
imag[2] = mag[2]*sin(phase[2])
imag[1] = mag[1]*sin(phase[3])

Thank you!

Kind regards,
Maarten