cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
FarajiRasoul
Observer
Observer
288 Views
Registered: ‎03-12-2020

Problem with 1024-point FFT configuration

Hi All,

 

I am trying to configure a 1024-point FFT IP core. I tried to use unscaled FFT as well as scaled ones but I cannot get the same or almost the same results as Matlab FFT functions. I do not use cyclic prefix. For unscaled, I set the configuration to 16'h0001 and for the scaled version I used 22222 and 22223 for SCALE_SCH but it seems that it does not work. I should mention that for 512-point and 256-point it works. I would really appreciate if you could help to figure it out.

 

Thanks

-Rasoul

 

 

0 Kudos
4 Replies
vkanchan
Xilinx Employee
Xilinx Employee
205 Views
Registered: ‎09-18-2018

Hi @FarajiRasoul ,

is the IP MEX function being used to compute the FFT? If so, are the input values quantized when provided to the Xilinx FFT function ? 

0 Kudos
FarajiRasoul
Observer
Observer
203 Views
Registered: ‎03-12-2020

Hi @vkanchan 

Thanks for your reply. I have used the IP in Vivado and simulated it with XSIM. The input values were quantized.

0 Kudos
nathanx
Moderator
Moderator
161 Views
Registered: ‎08-01-2007

@FarajiRasoul , can you share your unscaled FFT IP parameter configuration and Matlab M code?

0 Kudos
FarajiRasoul
Observer
Observer
134 Views
Registered: ‎03-12-2020

Hi @nathanx 

I have attached codes and an image that shows how I have configured my FFT. I have used Matlab FFT to validate Xilinx FFT output values. The output values of Matlab FFT are not exactly the same as Xilinx FFT, however, they are close. The output values that I have got using the configuration and attached files are way off from Matlab FFT outputs.

 

0 Kudos