| (0) Compute spectrum of 0kHz |
- Multiply w[n] and a[n] then accumulate vectors.
- w[n] is always w[0] because it's 0kHz.
|
| (1) Compute spectrum of 1kHz |
- w[n] twiddles step by step.
- Find the symmetry of vectors.
|
| (2) Compute spectrum of 2kHz |
- w[n] twiddles but computes by using w[0].
- Later on, it compensates vectors for lag-lead.
|
| (3) Compute spectrum of 3kHz |
- Computes in the same way as 1kHz.
- Later on, it compensates vectors for lag-lead.
|
| (4) Compute spectrum of 4kHz |
- 4kHz is the Nyquist frequency.
- Gets the spectrum of highest frequency.
|
| (5) Compute spectrum of 5kHz |
- Beyond the Nyquist frequency.
- The spectrum is conjugate with 3kHz.
|
| (6) Compute spectrum of 6kHz |
- Beyond the Nyquist frequency.
- The spectrum is conjugate with 2kHz.
|
| (7) Compute spectrum of 7kHz |
- Beyond the Nyquist frequency.
- The spectrum is conjugate with 1kHz.
|
The following are PDF documents to explain how the computation progresses
on eight-points FFT. Suppose the sampling frequency is 8kHz.