DIT- and DIF-Radix-2 FFT Demonstrations

Katty Beltrán, Ana Salcedo

October 10, 2024

1 Demonstration: DIT - FFT Radix 2

1. We start considering a signal 𝑥(𝑛) of N length

𝑥 (𝑛) = {𝑥 (0) , 𝑥 (1) , 𝑥 (2) , . . . , 𝑥 (𝑁 − 2) , 𝑥 (𝑁 − 1)}

2. Decompose the sequence in 2 equal sequences of length 𝑁 /2

even ∶ {𝑥 (0) , 𝑥 (2) , 𝑥 (4) , . . . , 𝑥 (𝑁 − 2)}

odd ∶ {𝑥 (1) , 𝑥 (3) , 𝑥 (5) , . . . , 𝑥 (𝑁 − 1)}

3. Considering that
𝑁−1
kn
𝑋 (𝑘) = ∑ 𝑥 (𝑛) 𝑊𝑁
𝑛=0

we can decimate our 𝑥(𝑛) in an even and odd sequence

𝑁 𝑁
2 −1 2 −1
kn kn
𝑋 (𝑘) = ∑ 𝑥 (𝑛) 𝑊𝑁 + ∑ 𝑥 (𝑛) 𝑊𝑁
𝑛=0 𝑛=1

4. By replacing 𝑛 = 2𝑟 for the even sequence and 𝑛 = 2𝑟 + 1 for the odd sequence
𝑁 𝑁
2 −1 2 −1
2kr 𝑘(2𝑟+1)
𝑋 (𝑘) = ∑ 𝑥 (2𝑟) 𝑊𝑁 + ∑ 𝑥 (2𝑟 + 1) 𝑊𝑁
𝑛=0 𝑛=1

𝑁
Now 𝑔1 (𝑛) = 𝑥 (2𝑟) and 𝑔2 (𝑛) = 𝑥 (2𝑟 + 1), where 0 ≤ 𝑟 ≤ 2 −1
𝑁 𝑁
2 −1 2 −1
2kr 2kr 𝑘
𝑋 (𝑘) = ∑ 𝑔1 (𝑛) 𝑊𝑁 + ∑ 𝑔2 (𝑛) 𝑊 𝑁 𝑊 𝑁
𝑛=0 𝑛=1
𝜋
if 𝑊𝑁 = 𝑒−j2 𝑁 then 𝑊𝑁
2
= 𝑊𝑁
2

𝑁 𝑁
2 −1 2 −1
kr 𝑘 kr
𝑋 (𝑘) = ∑ 𝑔1 (𝑛) 𝑊 𝑁 + 𝑊 𝑁 ∑ 𝑔2 (𝑛) 𝑊 𝑁
2 2
𝑛=0 𝑛=1

5. By letting 𝐺1 (𝑘) and 𝐺2 (𝑘) be 𝑁 /2 point DFTs of 𝑔1 (𝑛) and 𝑔2 (𝑛) the expression reduces
to
𝑘
𝑋 (𝑘) = 𝐺1 (𝑘) + 𝑊 𝑁 𝐺2 (𝑘) , where 0 ≤ 𝑘 ≤ 𝑁 − 1

1
1.1 Demonstration: DIT - FFT Radix 2
1. We start considering a signal 𝑥(𝑛) of N length

𝑥 (𝑛) = {𝑥 (0) , 𝑥 (1) , 𝑥 (2) , . . . , 𝑥 (𝑁 − 2) , 𝑥 (𝑁 − 1)}

2. Decompose the sequence in 2 equal sequences of length N/2

3. Considering that
𝑁−1
kn
𝑋 (𝑘) = ∑ 𝑥 (𝑛) 𝑊𝑁
𝑛=0

we can decimate our x(n) as

𝑁 𝑁
2 −1 2 −1
kn kn
𝑋 (𝑘) = ∑ 𝑋 (𝑛) 𝑊𝑁 + ∑ 𝑋 (𝑛) 𝑊𝑁
𝑛=0 𝑛=1

𝑁 𝑁
2 −1 2 −1
kn 𝑁 𝑘(𝑛+ 𝑁
2 )
𝑋 (𝑘) = ∑ 𝑥 (𝑛) 𝑊𝑁 + ∑ 𝑥 (𝑛 + ) 𝑊𝑁
𝑛=0 𝑛=1
2
𝑁 𝑁
2 −1 2 −1
kn
𝑁 𝑁 kn
𝑋 (𝑘) = ∑ 𝑥 (𝑛) 𝑊𝑁 + 𝑊𝑁 ∑ 𝑥 (𝑛 +
2
) 𝑊𝑁
𝑛=0 𝑛=1
2
𝑁
𝑘
if 𝑊𝑁 = (−1)
2

𝑁 𝑁
2 −1 2 −1
kn 𝑘 𝑁 kn
𝑋 (𝑘) = ∑ 𝑥 (𝑛) 𝑊𝑁 + (−1) ∑ 𝑥 (𝑛 + ) 𝑊𝑁
𝑛=0 𝑛=1
2

Finally, by decomposing 𝑋(𝑘) in an even (𝑘 = 2𝑟) and odd (𝑘 = 2𝑟 + 1) sequence, and multiplying
by the twiddle factor we obtain
For the even part of the signal:
𝑁
2 −1
𝑁 rn
𝑥 (2𝑟) = ∑ [𝑥 (𝑛) + 𝑥 (𝑛 + )] 𝑊 𝑁
𝑛=0
2 2

For the odd part of the signal:

𝑁
2 −1
𝑁 rn 𝑛
𝑥 (2𝑟 + 1) = ∑ [𝑥 (𝑛) − 𝑥 (𝑛 + )] 𝑊 𝑁 𝑊𝑁
𝑛=0
2 2