frznchckn
Member level 1
- Joined
- Jul 20, 2010
- Messages
- 36
- Helped
- 4
- Reputation
- 8
- Reaction score
- 4
- Trophy points
- 1,288
- Location
- California
- Activity points
- 1,521
I'm building an n-point (power of 2) radix-2 FFT where all data samples will pass through the radix-2 butterfly log2 times.
I thought I understood this algorithm, but I'm having doubts as I am having a lot of trouble modeling the dataflow in MATLAB to create a bit-equivalent model. My FFT output doesn't match the MATLAB built-in (fftw) at all. To experiment, I modeled an 8 point radix-2 FFT in Excel. I can't match Excel's FFT either.
Does anyone have any ideas? I thought maybe I was not indexing my twiddle factors properly. For an 8 point FFT with no ROM compression, I'll have 8 twiddle factors for real and imaginary respectively. For the first pass, I index as follows:
0,4,0,4
For the second pass
0,2,4,6
and for the third pass
0,1,2,3
My Excel model is attached as well. Would someone point out my mistake in the algorithm or what concept I'm missing here? Thanks.
I thought I understood this algorithm, but I'm having doubts as I am having a lot of trouble modeling the dataflow in MATLAB to create a bit-equivalent model. My FFT output doesn't match the MATLAB built-in (fftw) at all. To experiment, I modeled an 8 point radix-2 FFT in Excel. I can't match Excel's FFT either.
Does anyone have any ideas? I thought maybe I was not indexing my twiddle factors properly. For an 8 point FFT with no ROM compression, I'll have 8 twiddle factors for real and imaginary respectively. For the first pass, I index as follows:
0,4,0,4
For the second pass
0,2,4,6
and for the third pass
0,1,2,3
My Excel model is attached as well. Would someone point out my mistake in the algorithm or what concept I'm missing here? Thanks.