dequantize

Dequantization is the process of reconstructing an analog signal from its digitized, stepped representation. When an analog signal is converted to digital form, it passes through quantization, which rounds each sample to the nearest discrete level. Dequantization reverses this: it takes those discrete integer or fixed-point values and converts them back into continuous voltage or current levels that approximate the original analog waveform.

The most basic dequantizer is a digital-to-analog converter (DAC), which assigns a specific analog output voltage to each digital input code. An 8-bit DAC with a 5 volt reference, for example, divides the output range into 256 steps of roughly 20 millivolts each. When the digital input changes from 127 to 128, the output jumps by one step. More bits mean finer granularity: a 16-bit DAC can resolve to about 76 microvolts across the same 5-volt span.

Real dequantization always introduces quantization error, the gap between the reconstructed signal and the true original. This error appears as unwanted noise, worst when the signal level is low. To minimize it, designs often use oversampling, dequantizing at a rate many times higher than necessary, then filtering out the excess high-frequency content. This shifts quantization noise away from the audio or measurement band of interest, effectively burying it where it matters less.

Application and artifacts

In audio, dequantization of compact disc data (16-bit, 44.1 kHz samples) must reconstruct a continuous pressure wave from roughly 44,000 discrete snapshots per second. Professional audio systems often use 24-bit DACs for greater precision. In industrial measurement, a 12-bit or 16-bit analog output card dequantizes stored or computed digital values so that analog control loops can modulate motor speed, valve position, or heater current in response to sensor feedback.

Dequantization errors become audible or measurable when the digital word length is too short for the application's dynamic range. Conversely, using more bits than necessary wastes silicon area and power. The term itself emerged from information theory: quantize means to map a continuous range to discrete symbols; dequantize reverses that map, though perfect recovery is impossible once information has been discarded.