The Ultimate DFT Properties Handbook

Aashique Hussain
August 12, 2025
 DFT Properties Handbook

Contents
1 Definition of the DFT and IDFT 3
1.1 The Discrete Fourier Transform (DFT) . . . . . . . . . . . . . . . . . . . 3
1.2 The Inverse Discrete Fourier Transform (IDFT) . . . . . . . . . . . . . . 3
1.3 The Twiddle Factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Fundamental Properties 3
2.1 Linearity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Time Shift (Circular Shift) . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.3 Frequency Shift (Modulation) . . . . . . . . . . . . . . . . . . . . . . . . 4
2.4 Time Reversal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

3 Convolution and Correlation 4

3.1 Circular Convolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.2 Multiplication in Time (Circular Convolution in Freq.) . . . . . . . . . . 4
3.3 Circular Correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

4 Common DFT Pairs (Must Know) 5

5 Duality and Iterated DFTs 5

5.1 The Duality Property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
5.2 DFT of the DFT (DFT2 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
5.3 The Four-Point Property (DFT4 ) . . . . . . . . . . . . . . . . . . . . . . 5

6 Symmetry Properties 6
6.1 For a Real-Valued Signal (x[n] ∈ R) . . . . . . . . . . . . . . . . . . . . . 6
6.2 Summary of Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

7 Summation and Special Value Properties 6

7.1 DC Component (k=0) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
7.2 Nyquist Component (k=N/2, N even) . . . . . . . . . . . . . . . . . . . . 7
7.3 Initial Sample (n=0) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
7.4 Midpoint Sample (n=N/2, N even) . . . . . . . . . . . . . . . . . . . . . 7

8 Energy Conservation (Parseval’s Theorem) 7

9 Zero-Padding and Upsampling 7

9.1 Zero-Padding in Time Domain . . . . . . . . . . . . . . . . . . . . . . . . 7
9.2 Upsampling (Inserting Zeros in Time) . . . . . . . . . . . . . . . . . . . . 8

2
 DFT Properties Handbook

1 Definition of the DFT and IDFT

For a finite-duration sequence x[n] of length N , its N-point Discrete Fourier Transform
(DFT), X[k], and its Inverse Discrete Fourier Transform (IDFT) are defined as follows.

1.1 The Discrete Fourier Transform (DFT)

The DFT analyzes the frequency content of the signal.
N −1
2π
X
X[k] = x[n]e−j N kn for k = 0, 1, . . . , N − 1 (1)
n=0

1.2 The Inverse Discrete Fourier Transform (IDFT)

The IDFT synthesizes the time-domain signal from its frequency components.
N −1
1 X 2π
x[n] = X[k]ej N kn for n = 0, 1, . . . , N − 1 (2)
N k=0

1.3 The Twiddle Factor

2π
The complex exponential term WN = e−j N is often called the ”twiddle factor”. Using
this notation, the DFT and IDFT are:
N
X −1
X[k] = x[n]WNkn (3)
n=0
N −1
1 X
x[n] = X[k]WN−kn (4)
N k=0

2 Fundamental Properties
DFT DFT
Let x[n] −−→ X[k] and y[n] −−→ Y [k]. All operations on the time index n and frequency
index k are performed modulo N .

2.1 Linearity
The DFT of a linear combination of signals is the same linear combination of their
individual DFTs.
DFT{ax[n] + by[n]} = aX[k] + bY [k] (5)

2.2 Time Shift (Circular Shift)

A circular shift in the time domain corresponds to a phase shift in the frequency domain.
2π
DFT{x[(n − m) (mod N )]} = X[k] · e−j N km = X[k]WNkm (6)

3
 DFT Properties Handbook

2.3 Frequency Shift (Modulation)

Multiplication by a complex exponential in the time domain corresponds to a circular
shift in the frequency domain.
2π
DFT{x[n] · ej N k0 n } = X[(k − k0 ) (mod N )] (7)

This is the dual of the time-shift property.

2.4 Time Reversal

Reversing the time-domain sequence corresponds to reversing the frequency-domain se-
quence.
DFT{x[(−n) (mod N )]} = X[(−k) (mod N )] (8)
Special Case for Real Signals: If x[n] is real, then from symmetry properties, X[(−k)
(mod N )] = X ∗ [k].
If x[n] ∈ R, DFT{x[−n]} = X ∗ [k] (9)

3 Convolution and Correlation

3.1 Circular Convolution
One of the most important properties for LTI system analysis. Circular convolution in
the time domain is equivalent to multiplication in the frequency domain.
N
X −1
x[n] ⊛ h[n] ≜ x[m]h[(n − m) (mod N )] (10)
m=0

DFT{x[n] ⊛ h[n]} = X[k] · H[k] (11)

3.2 Multiplication in Time (Circular Convolution in Freq.)

The dual property: multiplication in the time domain is equivalent to circular convolution
in the frequency domain.
1
DFT{x[n] · h[n]} = (X[k] ⊛ H[k]) (12)
N

3.3 Circular Correlation

Correlation is similar to convolution, but without time-reversing one of the signals.
N
X −1
rxy [l] = x[n]y ∗ [(n − l) (mod N )] (13)
n=0

DFT{rxy [l]} = X[k] · Y ∗ [k] (14)

4
 DFT Properties Handbook

Table 1: Essential DFT Pairs

Time Domain x[n] DFT X[k]
δ[n] 1 (for all k)

1 (constant DC) N δ[k]

2π
ej N k0 n N δ[(k − k0 ) (mod N )]

2π N


cos N 0
kn 2
(δ[(k − k0 )] + δ[(k + k0 )])
Note: this is N2 (δ[k − k0 ] + δ[N − k0 ])

2π N


sin N 0
kn 2j
(δ[(k − k0 )] − δ[(k + k0 )])
Note: this is jN2
(δ[k + k0 ] − δ[k − k0 ])

4 Common DFT Pairs (Must Know)

5 Duality and Iterated DFTs
5.1 The Duality Property
If the functional form of a time-domain signal is used to create a frequency-domain signal,
its corresponding DFT will have the functional form of the original frequency-domain
signal.
DFT DFT
If x[n] −−→ X[k], then X[n] −−→ N x[(−k) (mod N )] (15)

5.2 DFT of the DFT (DFT2 )

Applying the duality property, we can find the transform of a transform.
DFT{X[n]} = N x[(−k) (mod N )] (16)
Let’s relabel the indices to be more conventional. Let y[n] = X[n]. Then its DFT is
Y [k] = DFT{X[n]}.
DFT{DFT{x[n]}} is the signal N x[(−n) (mod N )] (17)
So, applying the DFT twice is equivalent to scaling by N and time-reversing the original
signal.

5.3 The Four-Point Property (DFT4 )

Applying the DFT four times returns the original signal, scaled by N 2 .
DFT1 {x[n]} = X[k]
DFT2 {x[n]} = N x[−n]
DFT3 {x[n]} = DFT{N x[−n]} = N · DFT{x[−n]} = N X[−k]
DFT4 {x[n]} = DFT{N X[−k]} = N · DFT{X[−k]} = N · (N x[−(−n)]) = N 2 x[n]

5
 DFT Properties Handbook

DFT4 {x[n]} = N 2 x[n] (18)

6 Symmetry Properties
Symmetry properties are extremely useful for reducing computation and for checking
results.

6.1 For a Real-Valued Signal (x[n] ∈ R)

This is the most common case in practice.
• Conjugate Symmetry (Most Important):

X[k] = X ∗ [(−k) (mod N )] = X ∗ [N − k] (19)

• Magnitude Symmetry: The magnitude spectrum is always even.

|X[k]| = |X[N − k]| (20)

• Phase Symmetry: The phase spectrum is always odd.

∠X[k] = −∠X[N − k] (21)

• Special Points:
– X[0] is always real.
– If N is even, X[N/2] is also always real.

6.2 Summary of Symmetries

The table below summarizes the DFT symmetry for various types of input signals.

Table 2: DFT Symmetry Relationships

Time Domain Signal x[n] Frequency Domain DFT X[k]
Complex x[n] (general case) Complex X[k] (no symmetry)
Real x[n] Conjugate Symmetric: X[k] = X ∗ [N − k]
Real and Even x[n] Real and Even X[k]
Real and Odd x[n] Purely Imaginary and Odd X[k]
Purely Imaginary x[n] Conjugate Anti-Symmetric: X[k] = −X ∗ [N − k]
Purely Imaginary and Even x[n] Purely Imaginary and Even X[k]
Purely Imaginary and Odd x[n] Real and Odd X[k]

7 Summation and Special Value Properties

These properties link specific points in one domain to the sum over all points in the other
domain. They are direct results of the DFT/IDFT definitions.

6
 DFT Properties Handbook

7.1 DC Component (k=0)

The DFT at k = 0 is the sum of all time-domain samples. It represents the DC or average
value of the signal.
N
X −1 N
X −1
0
X[0] = x[n]e = x[n] (22)
n=0 n=0

7.2 Nyquist Component (k=N/2, N even)

The DFT at k = N/2 (for N even) is the alternating sum of the time-domain samples.
N −1 N −1 N −1
−j 2π N
X X X
n −jπn
X[N/2] = x[n]e N 2 = x[n]e = x[n](−1)n (23)
n=0 n=0 n=0

7.3 Initial Sample (n=0)

The time-domain sample at n = 0 is the scaled sum of all frequency-domain samples.
N −1 N −1
1 X 1 X
x[0] = X[k]e0 = X[k] (24)
N k=0 N k=0

7.4 Midpoint Sample (n=N/2, N even)

The time-domain sample at n = N/2 is the scaled alternating sum of frequency-domain
samples.
N −1 N −1 N −1
1 X j 2π kN 1 X jπk 1 X
x[N/2] = X[k]e N 2 = X[k]e = X[k](−1)k (25)
N k=0 N k=0 N k=0

8 Energy Conservation (Parseval’s Theorem)

The total energy of a signal is conserved across the transform, up to a scaling factor of
N . This means the energy calculated in the time domain equals the energy calculated in
the frequency domain.
N −1 N −1
X 1 X
|x[n]|2 = |X[k]|2 (26)
n=0
N k=0

9 Zero-Padding and Upsampling

9.1 Zero-Padding in Time Domain
Appending zeros to a time-domain signal x[n] to increase its length from L to N > L
before taking an N -point DFT.

• Effect: This does not add any new information to the signal.

7
 DFT Properties Handbook

• Result: It increases the sampling density of the frequency spectrum. We are

evaluating the same underlying DTFT at more frequency points. This results in a
higher resolution DFT plot, which can reveal peaks and nulls that were missed in
the shorter DFT. It is essentially an interpolation of the original L-point DFT.

9.2 Upsampling (Inserting Zeros in Time)

Creating a new signal y[n] of length LN by inserting L − 1 zeros between each sample of
the original signal x[n] of length N .
(
x[n/L], if n is a multiple of L
y[n] = (27)
0, otherwise

The DFT of the upsampled signal Y [k] is simply the original DFT X[k] replicated L
times.
Y [k] = X[k (mod N )] (28)
This means the spectrum X[k] repeats itself L times over the full frequency range of the
new, longer DFT.