6
ADCs I. Intuitive Review of DFT Matrix a. Sampling Frequency and the DFT b. Sampling and Aliasing Examples c. Discrete Frequency Domain II. Nyquist Theorem a. Application Example i. Sampling Window b. Sampling III. Anti-Aliasing Filter IV. Sampling and Aliasing Analog to Digital Converter (ADC) I. Given a cont. time waveform and an ADC that samples at freq (f_s), collect N samples (Post lecture notes in purple) (Impt equations boxed in green) 8-10, Lecture 29: Sampling and Aliasing Monday, August 10, 2020 9:32 AM Lecture 29 Page 1

8-10, Lecture 29: Sampling and Aliasingee16b/su20/lecture/lec29.pdfNyquist Sampling Theorem: A bandlimited, cont time signal can be sampled and perfectly represented/ perfectly reconstructed

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

Page 1: 8-10, Lecture 29: Sampling and Aliasingee16b/su20/lecture/lec29.pdfNyquist Sampling Theorem: A bandlimited, cont time signal can be sampled and perfectly represented/ perfectly reconstructed

ADCsI.

Intuitive Review of DFT Matrixa.Sampling Frequency and the DFTb.Sampling and Aliasing Examplesc.

Discrete Frequency DomainII.

Nyquist Theorema.

Application Examplei.Sampling Windowb.

SamplingIII.

Anti-Aliasing FilterIV.

Sampling and Aliasing

Analog to Digital Converter (ADC)I.

Given a cont. time waveform and an ADC that samples at freq (f_s), collect N samples

(Post lecture notes in purple)(Impt equations boxed in green)

8-10, Lecture 29: Sampling and AliasingMonday, August 10, 2020 9:32 AM

Lecture 29 Page 1

Page 2: 8-10, Lecture 29: Sampling and Aliasingee16b/su20/lecture/lec29.pdfNyquist Sampling Theorem: A bandlimited, cont time signal can be sampled and perfectly represented/ perfectly reconstructed

Intuitive Review of DFTII.

-- filter design-- signal processing

- Find how much of each frequency is an a signal (data analysis)

--phasor domain-- convolution

- Project the signal into a basis of orthogonal sinusoids (math easier)

Purpose of DFT?

DFT Matrixa.

Sampling Frequency and the DFTb.

(Note there is a scalar here that Matlab ignores. This matrix is not orthonormal, just orthogonal. So I'm applying the scalar out here)

Lecture 29 Page 2

Page 3: 8-10, Lecture 29: Sampling and Aliasingee16b/su20/lecture/lec29.pdfNyquist Sampling Theorem: A bandlimited, cont time signal can be sampled and perfectly represented/ perfectly reconstructed

How f_s relates to our DFT matrix

Examples of Sampling and Aliasingc.

Lecture 29 Page 3

Page 4: 8-10, Lecture 29: Sampling and Aliasingee16b/su20/lecture/lec29.pdfNyquist Sampling Theorem: A bandlimited, cont time signal can be sampled and perfectly represented/ perfectly reconstructed

I can create a signal x'(t) which has the same x[n] and Fx[n]

Helicopter floating!Ex2. Visual Aliasing

Sampling TheoremII.

Nyquist Sampling Theorema.

Let's agree that when we sample, all of our frequencies will be the positive frequency, and all of our aliasing is just a copy

(also technically aliasing, but without the complex conjugate)

Lecture 29 Page 4

Page 5: 8-10, Lecture 29: Sampling and Aliasingee16b/su20/lecture/lec29.pdfNyquist Sampling Theorem: A bandlimited, cont time signal can be sampled and perfectly represented/ perfectly reconstructed

Nyquist Sampling Theorem:A bandlimited, cont time signal can be sampled and perfectly represented/ perfectly reconstructed from its samples, IF the sampling freq (f_s) is over twice as fast as the signals highest freq component (B)

Sampling Freq and Sampling Windowb.

Sample fast -> to get all the high, small freq details w/o aliasing-

Intuition

Collect data over a long time period, represent low frequencies-

If you remember nothing else, remember this!

Bandwidth is the signal's highest frequency (all higher frequencies are zero)

Lecture 29 Page 5

Page 6: 8-10, Lecture 29: Sampling and Aliasingee16b/su20/lecture/lec29.pdfNyquist Sampling Theorem: A bandlimited, cont time signal can be sampled and perfectly represented/ perfectly reconstructed

Healthy avg human : 60-100 beats/min -> ~1beat/secAthlete : 40 beat/min -> 2/3 beat/secFastest recorded : 480 beats/min -> 8 beats/sec

How fast must our ADC sample?

How many samples are needed

Lecture 29 Page 6