FPGA Intellectual Property
PCI Express*, Networking and Connectivity, Memory Interfaces, DSP IP, and Video IP
Intel Support hours are Monday-Fridays, 8am-5pm PST, except Holidays. Thanks to our community members who provide support during our down time or before we get to your questions. We appreciate you!

Need Forum Guidance? Click here
Search our FPGA Knowledge Articles here.


Honored Contributor II

Hi ! 


When I run FFT and IFFT tool in megawizard, the results are not logical. For example firstly I run the FFT tool with the megawizard tb (input: natural order, output : digit reversed order) then I use the FFT output txt files for the input files of IFFT ( input : digit reversed order output: natural order) the result of IFFT dont match with the input of FFT . 

What could be the problem ? 


P.S = I use single precision floating point representation for the 16-FFT tool. 



Thanks all
0 Kudos
4 Replies
Honored Contributor II

Have you tried checking what the input file looks like when you generate a testbench for the IFFT, there's no guarantee the output file of the FFT matches the input format for the IFFT.

Honored Contributor II

I used the output of FFT for the input of IFFT.I mean I modified the tb for IFFT. 


By the way I found the solution ;) When you check in binary for floating point it is hard to find what is wrong. 


When I converted to decimals in Matlab I saw that the output multiplied with the FFT size. 

For example if you use 8-IFFT the outputs are 8 time greater than the outputs had used to be. You should normalize them by IFFT size ;) 


Thanks anyway for your reply
Honored Contributor II

Hi ferdiskaras, 


I guess I have the same problem you went through. I have implemented a VARIABLE STREAMING (N=8), SINGLE-FLOAT-POINT FFT/IFFT. These are my results: 


input (natural order): (8 0 0 0 8 0 0 0)  


output (Digit-reverse order): [16 16 0 0 16 16 0 0] /* the same as [16 0 16 0 16 0 16 0 ] Natural order */ 


input(Digit-reverse order): [16 16 0 0 16 16 0 0]  


output (natural order): (64 0 0 0 64 0 0 0) 


How did you normalize it? Where did you make de division by N? In your hardware or in your software? Is there a way to set FFT core to output data already normalized? 


Honored Contributor II

If change the "I/O Output Order" of FFT and "I/O input order" of IFFT to "digital reverse Order", then the input of FFT should be equal to output of iFFT. I'm able to get the correct result.