Sampling and the Nyquist Theorem

The Challenge of Digital Representation

As we discussed, an analog signal is continuous, containing an infinite amount of information. A computer, however, can only store a finite set of discrete values. The process of sampling is our bridge between these two worlds. It involves measuring the amplitude (voltage) of the analog waveform at regular, fixed time intervals and recording that value as a number. The result is no longer a continuous line but a sequence of dots.

To see this process in action and understand why it's necessary for digital storage, let's watch the beginning of the following video.

The video raises a critical point: if we are just storing a series of dots, how can we be sure that connecting them will recreate the original smooth curve? What if we don't take enough samples and misinterpret the wave's true shape? This is where a landmark discovery in information theory provides the answer.

The Nyquist-Shannon Sampling Theorem: The Golden Rule

In the mid-20th century, engineers Harry Nyquist and Claude Shannon independently formulated what is now known as the Nyquist-Shannon Sampling Theorem. This theorem is the foundation of digital audio and provides a stunningly elegant solution to the sampling problem.

In simple terms, the theorem states:

To perfectly reconstruct an analog signal from a series of samples, the sampling rate must be greater than twice the highest frequency present in the signal.

Mathematically, this is expressed as:
$f_s>2\times f_{max}$

Where:

• $f_s$ is the sampling frequency (the number of samples taken per second).
• $f_{max}$ is the maximum frequency component in the original analog signal.

This means that as long as you obey this rule, you are not merely creating an approximation of the original wave; you are capturing its complete "mathematical DNA." With the right reconstruction process, you can restore the original analog wave with no loss of information.

The article below provides a fantastic, plain-English explanation of this concept.

Aliasing: When Sampling Goes Wrong

The Nyquist-Shannon theorem tells us the condition for success, but it also implies a condition for failure. What happens if our sampling rate $f_s$ is less than twice the maximum frequency? The result is an undesirable form of distortion called aliasing.

When you undersample a high-frequency wave, your samples don't capture enough information to define its shape. The reconstruction process then misinterprets these samples as belonging to a completely different, lower-frequency wave that wasn't in the original signal. In effect, the high frequency "aliases" or masquerades as a lower frequency.

The most critical thing to understand about aliasing is that it is irreversible. Once a high frequency has folded down into the audible range, there is no way to distinguish it from a legitimate frequency that was originally present. The damage is permanent.

This image perfectly illustrates the difference between correct sampling and aliasing.

To see a visual demonstration of how this happens, let's return to the video from 0612 TV.

Now that you've seen what aliasing looks like, it's important to understand its sonic character and how it manifested in the history of digital audio.

From Theory to Practice: Standard Sample Rates

The Nyquist-Shannon theorem directly explains the standard sample rates you will encounter constantly in your career.

• The range of human hearing is generally accepted to be from 20 Hz to 20,000 Hz (20 kHz).
• To capture the entire range of human hearing, according to the theorem, we must use a sampling rate greater than $2\times 20,000$ Hz, or 40,000 Hz (40 kHz).

This is why the standard for Compact Discs (CDs) was set at 44.1 kHz. This rate provides coverage up to 22.05 kHz, comfortably capturing all audible frequencies with an added "safety margin" for the filters involved in the process (which we'll touch on shortly).

The choice of 44.1 kHz specifically, rather than a round 40 or 42 kHz, has a fascinating history tied to the video technology used to store the first digital audio recordings.

Preventing Aliasing

To prevent aliasing, audio converters employ a crucial component: an anti-aliasing filter. This is a very steep low-pass filter placed in the circuit before the analog signal is sampled. Its job is to aggressively remove any frequencies above the Nyquist frequency (half the sampling rate). For a CD-quality converter sampling at 44.1 kHz, this filter will cut out everything above 22.05 kHz, ensuring that no ultrasonic frequencies can be sampled and alias down into the audible spectrum. This pre-emptive filtering is the primary defense against aliasing.

Conclusion

In this lesson, we have unpacked one of the most important theoretical concepts in audio engineering. You now have a solid foundation for understanding how analog sound is faithfully captured in the digital domain.

Here are the key takeaways:

• Sampling is the process of converting a continuous analog wave into a series of discrete numerical values by measuring its amplitude at regular intervals.
• The Nyquist-Shannon Sampling Theorem dictates that for a perfect reconstruction, the sampling rate must be more than twice the highest frequency in the signal ($f_s>2\times f_{max}$).
• Aliasing is an irreversible distortion that occurs when a signal is undersampled. High frequencies "fold down" and are incorrectly represented as lower frequencies.
• Standard sample rates like 44.1 kHz (for music) and 48 kHz (for film) are direct applications of this theorem, chosen to capture the full range of human hearing while accommodating practical technology constraints.
• Anti-aliasing filters are used before the conversion process to remove frequencies that could cause aliasing.

In our next lesson, we will formalize our understanding of the two key parameters that define digital audio quality. We will revisit sample rate in this new context and introduce its crucial counterpart, bit depth, to see how they work together to determine the fidelity of a digital recording.