cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Highlighted
Observer
Observer
2,800 Views
Registered: ‎08-30-2011

FFT7.1 for audio applications

Hi, 

I´ve implemented an fft7.1 core from simulink xilinx blockset. It´s working fine. I want to transform audio signals and my only Problem now is, that i always get overflows, when I go a little bit highter with the Amplitude. I tried to scale the input amplitude with a cmult block down, but it just helps a little bit. The input amplitude is about 1.3V and I multiplied it by 0.6 (scale_sch = 302) and the overflows come at an 675mV input amplitude.

Now I tried to multiply by 0.2 and the overflows are at 820mV (same scale_sch). 

Is there anybody with some experience in those things?

What can I do to get no overflows? Why is the cmult block not really helping to solve it and which scale_sch could be helpful?


Regards Jan

0 Kudos
1 Reply
Highlighted
Xilinx Employee
Xilinx Employee
2,793 Views
Registered: ‎11-28-2007

You may want to even out scale schedule so the output is scaled after each stage. For your scaling schedule of 302, you could try scale_sch=122, which scales the output by the same 5-bit, but they are distributed to all 3 stages.

 


@janscholz wrote:

Hi, 

I´ve implemented an fft7.1 core from simulink xilinx blockset. It´s working fine. I want to transform audio signals and my only Problem now is, that i always get overflows, when I go a little bit highter with the Amplitude. I tried to scale the input amplitude with a cmult block down, but it just helps a little bit. The input amplitude is about 1.3V and I multiplied it by 0.6 (scale_sch = 302) and the overflows come at an 675mV input amplitude.

Now I tried to multiply by 0.2 and the overflows are at 820mV (same scale_sch). 

Is there anybody with some experience in those things?

What can I do to get no overflows? Why is the cmult block not really helping to solve it and which scale_sch could be helpful?


Regards Jan




Cheers,
Jim
0 Kudos