Delta Modulation

Delta Modulation (DM) is a simplified form of pulse modulation where the analog signal is encoded using only 1 bit per sample. Instead of transmitting the actual amplitude at each sample point, DM tracks whether the signal is going up or down, making it extremely simple to implement in hardware. It is widely studied in communication engineering for understanding quantization trade-offs and forms the basis for more advanced schemes like Adaptive Delta Modulation.

Core Concept of Delta Modulation

In standard **Pulse Code Modulation (PCM)**, each sample is quantized into multiple bits. Delta Modulation simplifies this drastically by using only 1 bit per sample. The core idea is: compare the current analog sample to the previous staircase approximation. If the analog signal is higher, output a 1 and step the staircase up. If it is lower, output a 0 and step the staircase down.

The step size, often called delta (represented as the Greek letter delta), is fixed in basic DM. This fixed step is added or subtracted at every sampling instant. The output is therefore a binary sequence of 1s and 0s representing the direction of change, not the actual magnitude. The receiver reconstructs the signal by applying the same up/down steps to recreate the staircase, then passing it through a low-pass filter.

Two fundamental errors arise in Delta Modulation. The first is **slope overload distortion**, which occurs when the analog signal changes too fast for the staircase to follow. The staircase keeps stepping in one direction but cannot catch up, causing large tracking error. The second is **granular noise**, which occurs when the signal is nearly flat. The staircase oscillates around the flat portion, creating a zig-zag error even when it should remain still.

Mathematical Expression

The staircase approximation at sample n is updated as follows. Let the approximation be denoted as x-hat(n). If the input x(nTs) is greater than x-hat(n-1), then x-hat(n) = x-hat(n-1) + delta. Otherwise, x-hat(n) = x-hat(n-1) - delta. The transmitted bit b(n) = 1 if stepping up, and b(n) = 0 if stepping down.

The condition for avoiding slope overload is that the maximum slope of the staircase must be at least as large as the maximum slope of the input signal. For a sinusoidal input x(t) = A cos(2 pi f t), slope overload is avoided when delta divided by Ts is greater than or equal to 2 pi f A. This gives the minimum step size as delta greater than or equal to 2 pi f A Ts, where Ts is the sampling period and f is the signal frequency.

Granular noise power is approximately (delta squared) divided by 3. The signal-to-noise ratio in DM depends on the balance between granular noise and slope overload, and optimal step size represents a trade-off between these two error types.

Practical Understanding

Delta Modulation is used in applications where extreme simplicity of hardware matters more than accuracy. Because only 1 bit is sent per sample, the encoder and decoder circuits are very simple: just a comparator, a delay element, and an accumulator. This made DM attractive in early digital telephony and military voice codecs. However, the inability to handle rapidly varying signals limits its use in high-fidelity applications.

The sampling rate in DM must be much higher than the Nyquist rate to reduce granular noise to acceptable levels. In contrast to PCM which samples at the Nyquist rate with multi-bit precision, DM oversamples with 1-bit precision. A typical DM system for voice might sample at 32 kHz or higher, compared to the 8 kHz Nyquist rate for 4 kHz bandwidth voice.

Given:
Analog input: x(t) = 0.5 cos(2π × 1000 × t), A = 0.5 V, f = 1 kHz
Sampling frequency: fs = 32 kHz, so Ts = 1/32000 s

Why this formula applies:
To avoid slope overload, the staircase step per sample must cover the maximum slope of the input signal.

Formula:
Minimum step size: δ ≥ 2πfA × Ts

Substitution:
δ ≥ 2π × 1000 × 0.5 × (1/32000)

Calculation:
δ ≥ 2 × 3.1416 × 1000 × 0.5 / 32000
δ ≥ 3141.6 / 32000
δ ≥ 0.0982 V

Final Answer: Minimum step size δ = 0.098 V (approximately 0.1 V) to avoid slope overload at 1 kHz.

Exam Tip: In GATE, slope overload occurs when the signal slope exceeds delta/Ts. Granular noise occurs when the signal is nearly constant. A common trap is confusing which noise type is reduced by increasing step size — increasing delta reduces slope overload but increases granular noise.

Mechanism: Encoder and Decoder Structure

• The comparator compares the input x(t) with the staircase approximation x-hat(n) and outputs 1 if x(t) is greater, else 0.
• The accumulator in the encoder updates x-hat(n) by adding or subtracting the fixed step size delta depending on the comparator output.
• The feedback loop ensures the staircase continuously tracks the input signal at each sampling instant.
• At the decoder, the same accumulator recreates the staircase from the received bit stream, and an LPF smooths it into the reconstructed analog signal.
• Slope overload and granular noise both reduce the quality of reconstruction and set the fundamental limits of basic DM performance.

Quick Revision

• Delta Modulation uses only 1 bit per sample, encoding direction of change (up = 1, down = 0).
• Fixed step size delta is added or subtracted to the staircase approximation at each sample.
• Slope overload occurs when the input signal changes faster than delta/Ts. Condition to avoid: delta/Ts greater than or equal to 2 pi f A.
• Granular noise occurs when the signal is nearly flat and the staircase oscillates around it.
• Granular noise power is approximately delta squared divided by 3.
• DM requires oversampling to reduce granular noise; PCM uses Nyquist rate with multi-bit precision.
• Trap: Increasing step size reduces slope overload but worsens granular noise. Decreasing step size does the opposite.