Approximation of the Continuous Time Fourier Transformperrins/class/F14_360/lab/labnotes8.pdfΒ Β·...

15
Approximation of the Continuous Time Fourier Transform Signals & Systems Lab 8

Transcript of Approximation of the Continuous Time Fourier Transformperrins/class/F14_360/lab/labnotes8.pdfΒ Β·...

Page 1: Approximation of the Continuous Time Fourier Transformperrins/class/F14_360/lab/labnotes8.pdfΒ Β· Continuous Time Fourier Transform (CTFT) ... Now π‘₯ is a signal in continuous time.

Approximation of the Continuous Time Fourier Transform

Signals & Systems

Lab 8

Page 2: Approximation of the Continuous Time Fourier Transformperrins/class/F14_360/lab/labnotes8.pdfΒ Β· Continuous Time Fourier Transform (CTFT) ... Now π‘₯ is a signal in continuous time.

Continuous Time Fourier Transform (CTFT)

𝑋 𝑓 = π‘₯ 𝑑 π‘’βˆ’π‘—2πœ‹π‘“π‘‘π‘‘π‘‘

∞

βˆ’βˆž

π‘₯ 𝑑 = 𝑋 𝑓 𝑒𝑗2πœ‹π‘“π‘‘π‘‘π‘“

∞

βˆ’βˆž

β€œUncle” Fourier

Page 3: Approximation of the Continuous Time Fourier Transformperrins/class/F14_360/lab/labnotes8.pdfΒ Β· Continuous Time Fourier Transform (CTFT) ... Now π‘₯ is a signal in continuous time.

Riemann Definite Integral

𝑓 π‘₯ 𝑑π‘₯

𝑏

π‘Ž

β‰œ limΞ”π‘₯β†’0 𝑓 π‘₯𝑖

βˆ— Ξ”π‘₯𝑖

𝑛

𝑖=1

β€œUncle” Riemann

Page 4: Approximation of the Continuous Time Fourier Transformperrins/class/F14_360/lab/labnotes8.pdfΒ Β· Continuous Time Fourier Transform (CTFT) ... Now π‘₯ is a signal in continuous time.

Riemann Definite Integral

𝑓 π‘₯ is defined where x ∈ π‘Ž, 𝑏

Ξ”π‘₯𝑖 = Ξ”π‘₯ =π‘βˆ’π‘Ž

𝑛 called the subinterval (assumed all equal in our case)

n is the number of subintervals which β€œfill” π‘Ž, 𝑏

π‘₯π‘–βˆ— are called the sample points of 𝑓 for each subinterval

As MATLAB can realistically operate only on discrete data we would like to use this definition to find an approximation to the CTFT.

𝑓 π‘₯ 𝑑π‘₯

𝑏

π‘Ž

β‰œ limΞ”π‘₯β†’0 𝑓 π‘₯𝑖

βˆ— Ξ”π‘₯𝑖

𝑛

𝑖=1

The limit Ξ”π‘₯ β†’ 0 means that the total interval π‘Ž, 𝑏 will be filled with a infinite number of subintervals each with width Ξ”π‘₯!

Page 5: Approximation of the Continuous Time Fourier Transformperrins/class/F14_360/lab/labnotes8.pdfΒ Β· Continuous Time Fourier Transform (CTFT) ... Now π‘₯ is a signal in continuous time.

𝑋 𝑓 = π‘₯ 𝑑 π‘’βˆ’π‘—2πœ‹π‘“π‘‘π‘‘π‘‘

∞

βˆ’βˆž

β‰œ limΔ𝑑→0 π‘₯ 𝑑𝑖

βˆ— π‘’βˆ’π‘—2πœ‹π‘“π‘‘π‘–βˆ—Ξ”π‘‘

𝑛

𝑖=1

In light of the previous observation we would like to express the Fourier Transform integral as a sum,

But in this form the expression for the Fourier transform is still impractical because it requires an infinite number of terms.

Page 6: Approximation of the Continuous Time Fourier Transformperrins/class/F14_360/lab/labnotes8.pdfΒ Β· Continuous Time Fourier Transform (CTFT) ... Now π‘₯ is a signal in continuous time.

Now π‘₯ 𝑑 is a signal in continuous time. But in MATLAB it will take on amplitude values only at discrete time points which depend on when the continuous-time signal was sampled. So suppose π‘₯ 𝑑 is being sampled every Δ𝑑 seconds resulting in a sampled function π‘₯ π‘š = π‘₯ π‘šΞ”π‘‘ where our time index, π‘š, start at 0 (causal) and ends at N, the total number of samples.

Time-Domain Housekeeping

Page 7: Approximation of the Continuous Time Fourier Transformperrins/class/F14_360/lab/labnotes8.pdfΒ Β· Continuous Time Fourier Transform (CTFT) ... Now π‘₯ is a signal in continuous time.

Hence the Fourier transform of this π‘₯[π‘š] signal is

Time-Domain Housekeeping

𝑋 𝑓 = limΔ𝑑→0 π‘₯ π‘šΞ”π‘‘ π‘’βˆ’π‘—2πœ‹π‘“π‘šΞ”π‘‘Ξ”π‘‘

π‘βˆ’1

π‘š=0

Page 8: Approximation of the Continuous Time Fourier Transformperrins/class/F14_360/lab/labnotes8.pdfΒ Β· Continuous Time Fourier Transform (CTFT) ... Now π‘₯ is a signal in continuous time.

Real signals of interest have finite Δ𝑑 time windows over the length of the signal, 𝑁Δ𝑑, so the limit can be dropped resulting in the approximate expression

Careful this simple result may be deceiving! This result says that for a particular frequency 𝑓 a sum must computed. If X 𝑓 is to be specified at every possible frequency then the sum must be evaluated at every possible frequency! That is an enormous amount of computation especially if N is large.

Time-Domain Housekeeping

𝑋 𝑓 β‰ˆ π‘₯ π‘šΞ”π‘‘ π‘’βˆ’π‘—2πœ‹π‘“π‘šΞ”π‘‘Ξ”π‘‘

π‘βˆ’1

π‘š=0

The infinite number of really small subintervals can be replaced with a finite number of small subintervals.

Page 9: Approximation of the Continuous Time Fourier Transformperrins/class/F14_360/lab/labnotes8.pdfΒ Β· Continuous Time Fourier Transform (CTFT) ... Now π‘₯ is a signal in continuous time.

Illuminating example when 𝑓 π‘₯ is a real & even Function

𝑓 π‘₯β„± 𝐹 𝑠 Source: Bracewell

Page 10: Approximation of the Continuous Time Fourier Transformperrins/class/F14_360/lab/labnotes8.pdfΒ Β· Continuous Time Fourier Transform (CTFT) ... Now π‘₯ is a signal in continuous time.

Explanation

Some authors will say that the Continuous-Time Fourier Transform of a function π‘₯ 𝑑 is the Continuous-Time Fourier Series of a function π‘₯ 𝑑 in the limit as 𝑇0 β†’ ∞. This is equivalent to saying the Fourier Series can be extended to aperiodic signals. You can also think of the Fourier Transform as taking all the time amplitude information and mapping it into a single frequency. Here the β€œmapping” is multiplying the time signal by a complex exponential at a particular frequency and finding the area under the resulting curve. Hence at each frequency this β€œmapping” or, in our case, sum must occur.

Page 11: Approximation of the Continuous Time Fourier Transformperrins/class/F14_360/lab/labnotes8.pdfΒ Β· Continuous Time Fourier Transform (CTFT) ... Now π‘₯ is a signal in continuous time.

What is 𝑓 exactly?

Much like in the case of sampled time signal we must settle for a sampled frequency signal

𝑋 𝑛 = 𝑋 𝑛1

𝑁Δ𝑑

1

𝑁Δ𝑑 can be thought of as the β€œdistance” in frequency between each

successive frequency sample. Compare it to Δ𝑑 for the time sampled signal.

In light of the previous discussion we can say that 1

𝑁Δ𝑑 comes from

mapping the entire period, 𝑁Δ𝑑, of π‘₯ 𝑑 into a single frequency!

Page 12: Approximation of the Continuous Time Fourier Transformperrins/class/F14_360/lab/labnotes8.pdfΒ Β· Continuous Time Fourier Transform (CTFT) ... Now π‘₯ is a signal in continuous time.

𝑋 𝑛 = π‘₯ π‘šΞ”π‘‘ π‘’βˆ’π‘—2πœ‹π‘›π‘šT Δ𝑑

π‘βˆ’1

π‘š=0

So replacing 𝑓 with 𝑛

𝑁Δ𝑑 in our expression we have

Frequency-Domain Housekeeping

With perhaps a slight abuse of notation we might say

π‘₯ π‘š β„± 𝑋[𝑛]

Page 13: Approximation of the Continuous Time Fourier Transformperrins/class/F14_360/lab/labnotes8.pdfΒ Β· Continuous Time Fourier Transform (CTFT) ... Now π‘₯ is a signal in continuous time.

What does it mean?

𝑋𝛼 = Δ𝑑 π‘₯ π‘šΞ”π‘‘ π‘’βˆ’π‘—2πœ‹π›Όπ‘šT

π‘βˆ’1

π‘š=0

At each sampled frequency 𝑓 =𝑛

𝑁Δ𝑑 (in this case let 𝑛 = 𝛼) there is

an amplitude which is found by

The evaluation of 𝑋𝑛 at each frequency will populate the frequency domain with an approximation X 𝑓 known as the Continuous-Time Fourier Transform of π‘₯ 𝑑 .

Page 14: Approximation of the Continuous Time Fourier Transformperrins/class/F14_360/lab/labnotes8.pdfΒ Β· Continuous Time Fourier Transform (CTFT) ... Now π‘₯ is a signal in continuous time.

Give it a name!

𝑋 𝑛1

𝑁Δ𝑑≅ Δ𝑑 β‹… π’Ÿβ„±π’― π‘₯ π‘šΞ”π‘‘

π’Ÿβ„±π’― π‘₯ π‘šΞ”π‘‘ = π‘₯ π‘šΞ”π‘‘ π‘’βˆ’π‘—2πœ‹π‘›π‘šT

π‘βˆ’1

π‘š=0

This approximation is known as the Discrete Fourier Transform!

Note: The author of your textbook derives this result differently (Web Appendix H) under the condition that 𝑛 β‰ͺ 𝑁.

Page 15: Approximation of the Continuous Time Fourier Transformperrins/class/F14_360/lab/labnotes8.pdfΒ Β· Continuous Time Fourier Transform (CTFT) ... Now π‘₯ is a signal in continuous time.

Sources

Bracewell, R. (1999). The Fourier Transform and Its Applications (3rd ed.). Boston: McGraw Hill.

Lathi, B. (1992). Linear Systems and Signals (1st ed.). Carmichael, CA: Berkeley-Cambridge Press.

Stewart, J. (2007). Calculus (6th ed.). Belmont, CA: Cengage Learning.