NoteVerse

quantization error

Comprehensive study notes, diagrams, and exam preparation for quantization error.

Quantization Error

Definition

Quantization error is the difference between the actual analog value of a signal sample and the nearest discrete quantization level assigned to it during analog-to-digital conversion.

Mathematically:

where:

  • = original sample value
  • = quantized value
  • = quantization error

If the quantizer uses rounding to the nearest level, the error is usually confined to a small range around zero. If the quantizer uses truncation, the error may be biased and generally larger in a systematic direction.

Main Content

1. Quantization Process and Quantization Levels

  • In digital communication, a continuous-time signal is first sampled, and each sample amplitude is then mapped to one of a finite number of levels.
  • If an analog range is divided into discrete levels, then each sample is replaced by the nearest level, creating a small mismatch between the input and output values.

For example, suppose a signal amplitude lies between 0 V and 8 V, and the system uses 3 bits. Then:

  • Number of levels
  • Step size

Possible quantization levels may be 0 V, 1 V, 2 V, ..., 7 V, or centered versions depending on the quantizer design. If the actual sample is 2.6 V, it may be assigned to 3 V, producing an error of -0.4 V or 0.4 V depending on sign convention.

A simple visual idea is:

Analog value:    2.6 V
Nearest level:   3.0 V
Difference:      0.4 V

This difference is the quantization error.

2. Types of Quantization Error

Uniform quantization error

  • occurs when levels are equally spaced. It is common in many communication systems and is easier to analyze mathematically.

Non-uniform quantization error

  • occurs when quantization levels are unevenly spaced, usually to give finer resolution for low-amplitude signals and coarser resolution for high-amplitude signals.

Important distinctions:

Rounding error

  • : the value is mapped to the nearest level, so the error is limited to about .

Truncation error

  • : the value is always rounded down or toward zero, which introduces a consistent negative or directional bias.

For a uniform quantizer:

  • Error range is typically:

  • If the signal is much larger and varies randomly over many levels, the quantization error can often be treated as noise.

In practice, companding systems such as μ-law and A-law use non-uniform quantization to reduce perceptible distortion in voice communication.

3. Effects of Quantization Error in Digital Communication

  • Quantization error introduces quantization noise, which degrades the fidelity of the reconstructed signal.
  • If the number of quantization levels is small, the step size is large, and the error becomes more noticeable, causing distortion in audio, image, and data signals.

Key effects include:

Amplitude distortion

  • : the reconstructed waveform does not exactly match the original one.

Signal-to-quantization-noise ratio (SQNR) reduction

  • : lower resolution reduces quality.

Audible or visible artifacts

  • : in audio, it may cause hiss or roughness; in images, it may cause banding or contouring.

Example:

  • A 4-bit system has levels.
  • An 8-bit system has levels.

An 8-bit system has much smaller quantization error than a 4-bit system because the levels are more closely spaced.

A useful ASCII illustration:

Original waveform:   ~~~~~~~~
Quantized waveform:  _/¯\_/¯\_
Difference:          small mismatches at each sample

The mismatch at each sample point is the quantization error.

Working / Process

1. Sample the analog signal

  • The continuous waveform is measured at discrete time instants.
  • Each sample still has an exact amplitude that may be any real number within the signal range.

2. Assign the nearest quantization level

  • The sample amplitude is compared with the available discrete levels.
  • The closest level is selected according to the quantizer rule, such as rounding or truncation.

3. Compute the quantization error

  • The difference between the original sample and the quantized output is found.
  • This error is stored implicitly as distortion in the system and affects signal reconstruction quality.

For example, if the sample is 5.7 and the quantization levels are integers:

  • Quantized value = 6
  • Quantization error = 5.7 - 6 = -0.3

This process repeats for every sample in the signal, so quantization error is always present in digital representation unless infinite precision is available, which is not practical.

Advantages / Applications

Enables digital representation of analog information

  • : Quantization error is the unavoidable trade-off that makes analog signals representable in digital systems such as telephones, computers, and data networks.

Helps in designing efficient communication systems

  • : By understanding and controlling quantization error, engineers can choose suitable bit rates, dynamic ranges, and quantizer designs to balance quality and bandwidth.

Used in audio, image, and sensor signal processing

  • : Quantization principles are essential in PCM voice systems, digital audio recording, medical instrumentation, image compression, telemetry, and IoT sensor data conversion.

Summary

  • Quantization error is the difference between an original signal sample and its nearest discrete digital level.
  • It appears because digital systems can only use a finite number of amplitude levels.
  • Smaller step size means smaller quantization error and better signal quality.
  • Important terms to remember: quantization, quantization level, step size, quantization noise, rounding, truncation
← Prev: PCM