The Nyquist Frequency is the highest frequency that can be represented by a digital system. It is exactly 1/2 of the sample rate. Frequencies above the Nyquist Frequency will be incorrectly represented in the digital system. Frequencies above the Nyquist Frequency alias (or fold back) into the region below the Nyquist frequency. A frequency 1 kHz above the Nyquist Frequency will fold back to 1 kHz below the Nyquist Frequency. Likewise a frequency 2 kHz above the Nyquist Frequency will fold back to 2 kHz below the Nyquist Frequency.
For example, the CD format uses a sample rate of 44.1 kHz. The Nyquist Frequency is 1/2 of 44.1 kHz or 22.05 kHz. Ideally, any analog input signal above 22.05 kHz must be filtered out before being digitized. If this filtering is not done, frequencies above 22.05 kHz will fold back into frequencies between 0 Hz and 22.05 kHz. A frequency of 23.05 kHz will fold back to 21.05 kHz (22.05-23.05+22.05=21.05).
Oversampling A/D converters sample the audio at very high sample rates to avoid aliasing. The Benchmark ADC1 and ADC16 sample the input audio at frequencies exceeding 6 MHz. These high sample rate digital signals are then filtered in the digital domain to remove frequencies above the target output sample rate. After filtering, the high sample rate digital signal is decimated to the desired output sample rate.
A digital audio device achieves bit-transparency if it passes digital audio data without affecting the data in any way. This device may be hardware, software, or a combination of both.
Testing for Bit Transparency
To determine if a device is bit-transparent, it can be tested by sending a pseudo-random bit sequence through it and monitoring the digital output. We typically generate the pseudo-random sequence using an Audio Precision digital signal generator. The Audio Precision includes a digital analyzer that is programmed to detect the pseudo-random sequence produced by the generator. The analyzer detects any differences between the generated sequence and the received sequence. The number of differences is tallied by a counter. A digital channel that is bit transparent will show no differences between the transmitted and received pseudo-random sequences.
Dither is a type of intentional variation (noise) which is added to a digital audio signal to avoid distortion caused by quantization errors.
Dithering in audio
Dither is often used when an analog signal is being quantized into a finite number of digital levels. It is also often used when a digital signal is being quantized into a fewer number of bits per sample corresponding to a fewer number of digital levels.