How Delta-Sigma DACs Achieve 24-Bit Audio Resolution

Delta-sigma digital-to-analog converters (DACs) dominate the audio electronics industry by achieving high-precision 24-bit resolution without requiring impossibly precise physical components. Traditional converters struggle to reach 24-bit accuracy due to thermal noise and manufacturing tolerances in their resistors. Delta-sigma DACs bypass these physical limitations by combining a simple low-bit converter with two powerful digital signal processing techniques: oversampling and noise shaping. Together, these methods trade raw circuit complexity for high-speed digital processing, pushing unwanted quantization noise out of the human hearing range to deliver pristine audio fidelity.

The Physical Limit of Traditional DACs

To understand why delta-sigma modulation is necessary, you must first look at the limitations of traditional R-2R ladder DACs. A 24-bit audio signal has 16,777,216 possible amplitude levels. For a traditional DAC to accurately convert this data into an analog voltage, it needs an array of microscopic resistors matched to an accuracy of roughly 0.000006%.

Manufacturing resistors to this exact tolerance is physically impossible at a commercial scale. Even if it were possible, random thermal noise inside the silicon would drown out the quietest signals, effectively limiting the real-world resolution of traditional DACs to around 16 to 20 bits.

Delta-sigma DACs solve this by abandoning multi-bit resistor ladders entirely. Instead, they use a 1-bit (or very low-bit) converter. A 1-bit DAC only has two states—on or off—meaning it has no resistor matching issues and is inherently linear. However, a 1-bit DAC generates a massive amount of quantization noise (the error between the digital value and the analog wave). To make a 1-bit system produce 24-bit quality audio, the DAC must eliminate this noise using oversampling and noise shaping.

Step 1: Oversampling Spreads the Noise

Every digital conversion produces a fixed amount of quantization noise. According to the Nyquist theorem, standard audio requires a sampling rate of at least 44.1 kHz to capture the human hearing range of 20 Hz to 20 kHz. At this standard rate, all the quantization noise is concentrated directly inside that 20 kHz audio band.

Oversampling is the process of operating the DAC at a frequency millions of times faster than the original audio sample rate (often 64, 128, or 256 times faster).

  • Spreading the Energy: Because the total amount of quantization noise remains constant, increasing the sampling frequency spreads that fixed noise energy over a much wider frequency spectrum.
  • Lowering the Noise Floor: By stretching the noise across several megahertz instead of just 20 kHz, the amount of noise remaining inside the actual human hearing range drops drastically.
  • Resolution Gain: For every quadrupling of the sampling rate, the DAC gains roughly 1 bit of effective resolution.

While oversampling helps, reaching 24-bit resolution through oversampling alone would require an impossibly high clock speed in the terahertz range. This is where noise shaping becomes essential.

Step 2: Noise Shaping Pushes the Noise Away

If oversampling spreads the noise out evenly, noise shaping acts like a bulldozer, actively scooping the noise out of the audible spectrum and piling it up into the ultrasonic frequencies.

Inside the delta-sigma modulator, a feedback loop continuously compares the output of the low-bit DAC to the original 24-bit input signal.

  1. The Delta (Difference): The circuit measures the error (the difference) between the intended 24-bit value and the actual 1-bit output.
  2. The Sigma (Integration): This error is summed and fed back into the input of the next cycle to correct the mistake immediately.

This rapid error-correction loop acts as a low-pass filter for the desired audio signal and a high-pass filter for the quantization noise. As a result, the audible frequencies (20 Hz to 20 kHz) are left virtually free of quantization noise, achieving a dynamic range equivalent to 24-bit resolution. Meanwhile, a massive amount of noise is pushed into the megahertz range, far above what humans can hear.

Step 3: Low-Pass Filtering

The final stage of a delta-sigma DAC is a simple analog low-pass filter. Because noise shaping has shoved all the quantization error into the ultrasonic frequencies, the DAC only needs an analog filter to slice off everything above 20 kHz.

Once this high-frequency ultrasonic noise is filtered away, the rapid stream of low-bit pulses averages out into a smooth, continuous analog waveform. By using time and speed (megahertz switching) instead of physical precision (resistor matching), delta-sigma DACs reliably and affordably achieve 24-bit audio resolution.