Forum Discussion

Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
16 years ago

A FFT bug?

hi all,

Thank you for answering or reading

I got a very very odd simulation results from modelsim for FFT V9.

First of all, FFT v9(evaluation) doesn't provide post simulation model for modelsim, so how does Altera suppose user to finish this part of work?

Ok, back to function simulation, I have a FFT, 64bit, 16 bit width, 16 bit precision, stream mode.

TEST ONE

I used the modelsim testbench created automatically by QII, to simulate a 64 point data file, and verify with Matlab fft fuction, it does match.

TEST TWO

So I goes to use the FFT in a small system, with one FFT module and an user designed mapping function to provide inputs to FFT, I use the exactly same input waveform for FFT module as in TEST ONE, however, the odd thing is the wrong results happened.

Is this a bug or something else?

2 Replies

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    16 years ago

    I found another interesting result.

    I use a same 64 points input repeat 2 times (64 x 2), the first output (64 point) from fft is incorrect, but the second one is correct.

    Also 64x3 run and 64x4 run, all correct but the first 64 points output from FFT. anybody know why?
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    16 years ago

    hi, all,

    I solved the problem by prolong the reset_n=0 to about 10 times clock period. hope it is little helpful to you.