Digital Audio

A microphone transduces the changes in sound pressure levels to voltages. These voltages are then sampled at regular intervals in a process called Sample and Hold.

The resultant voltage for each sample and hold is send to the Analogue to Digital converter where it is assigned a numerical value (Quantization), stored in binary form in the computer.

Sampling Frequency - when representing digital audio, one must use a sample rate at least twice as high as the highest frequency you want to represent. In CD audio this is 44,100 samples per second.

Lowpass Filter - audio coming in have all the frequencies filtered out higher than the Nyquist Frequency to eliminate Alias Frequencies from the recording. Audio being played back from the Digital to Analogue Converter is filtered again to remove anomalous frequencies introduced by sample and hold process.

Alias Frequency or Foldover Frequency - these are frequencies that are created by trying to playback sounds higher than the Nyquist Frequency. Ex) recording at 16,000Hz can represent 8,000Hz. a freq of 9,000Hz would also be represented by an alias freq at 7,000Hz.

Bit Depth - the number of bits used to represent the amplitude of each instantaneous sample. CD quality is 16bit. DVD is 24bit.

More About Binary

A binary number is made up of a series of binary digits, or bits. As we have seen, each bit can be set to either zero or one. Bit values are based on powers of 2 and on the digit position. Binary number are often manipulated in octets, or groups of eight bits. The values within an octet are as follows:

Position 8 7 6 5 4 3 2 1

Power 27 26 25 24 23 22 21 20

Decimal 128 64 32 16 8 4 2 1

Value MSB LSB

The leftmost bit is considered the most significant bit, or MSB. The rightmost bit is considered the least significant bit, or LSB.

It may help you to understand binary digits if you realize that they use base-2 number instead of the familiar base-10 numbering we use in normal life. Consider the following example:

Position 6 5 4 3 2 1

Power 105 104 103 102 101 100

Decimal 100,000 10,000 1,000 100 10 1

Decimal numbers are managed in the same fashion as binary numbers, with the rightmost digit representing 1s, the next 10s, the next 100s, and so on.

Quantization Error - the error introduced by rounding voltages to the nearest bit value. This introduces white noise into recording, this creates limitations on Signal to

Quantization Noise Ratio of the recording. Each bit of resolution adds 6dB of SQNR to the recording. 8bit recording = 48dB SQNR or 16bit = 96dB SQNR

Storage Requirements - (16bits X 44,100Hz X 60 sec X 2 channels)/8bit = 10,584,000 bytes/minute

Clipping - inputing a signal louder than the ADC can express numerically causes distortion or clipping. ADC will truncate or foldover the loud signal causing sudden drops in level of the waveform causing nasty noise.