Signal processing terminologies
Signal processing terminologies:
Attack: The attack of a sound is the initial portion of the sound. Percussive sounds (drums, piano, guitar plucks) are said to have a fast attack. This means that the sound reaches its maximum amplitude in a very short time. Sounds that slowly swell up in volume (soft strings and wind sounds) are said to have a slow attack.
Attenuation: A decrease in the level of a signal.
Aliasing — When an analogue signal is sampled for conversion into a digital data stream, the sampling frequency must be at least twice that of the highest frequency component of the input signal. If this rule is disobeyed the sampling process becomes ambiguous as there are insufficient points to define each cycle of the waveform, resulting in unwanted enharmonic frequencies being added to the audible signal.
Anti-alias Filter — A very steep low-pass filter used to limit the frequency range of an analogue signal prior to A/D conversion so that the maximum frequency does not exceed half the sampling rate.
Band-pass Filter (BPF) — A filter that removes or attenuates frequencies above and below the centre frequency at which it is set, and only passes a specific range of frequencies. Band-pass filters are often used in synthesizers as tone shaping elements.
Bandwidth: When discussing audio equalization, each frequency band has a width associated with it that determines the range of frequencies that are affected by the EQ. An EQ band with a wide bandwidth will affect a wider range of frequencies than one with a narrow bandwidth.When discussing network connections, bandwidth refers to the rate of signals transmitted or the amount of data that can be transmitted in a fixed amount of time (stated in bits/second): a 56 Kbps network connection is capable of receiving 56,000 bits of data per second.
Boost/Cut Control — A single gain control which allows the range of frequencies passing through a filter to be either amplified or attenuated. The centre position is usually the 'flat' or 'no effect' position.
Chorus: Chorusing is an effect created by combining a signal with a modulating, delayed copy of itself. This effect creates the illusion of multiple sources creating the same sound.
Colouration — A distortion of the natural timbre or frequency response of sound, usually but not always unwanted.
Comb-Filter — a series of deep filter notches created when a signal is combined with a delayed version of itself. The delay time (typically less than 10ms) determines the lowest frequency at which the filter notches start.
Compressor — A device (analogue or digital) which is designed to reduce the overall dynamic range of an audio signal either by attenuating the signal if it exceeds a set threshold level according, or by increasing the level of quiet signals below a threshold. The amount of attenuation is defined by a set ratio, while the speed of response (attack) and recovery (release) can usually also be controlled.
Compression Ratio (audio): A compression ratio controls the ratio of input to output levels above a specific threshold. This ratio determines how much a signal has to rise above the threshold for every 1 dB of increase in the output. For example, with a ratio of 3:1, the input level must increase by three decibels to produce a one-decibel output-level increase:
Threshold = -10 dB
Compression Ratio = 3:1
Input = -7 dB
Output = -9 dB
Because the input is 3 dB louder than the threshold and the compression ratio is 3:1, the resulting signal is 1 dB louder than the threshold.
Crossover — A set of audio filters designed to restrict and control the range of input signal frequencies which are passed to each loudspeaker drive unit. A typical two-way speaker will employ three filters: a high-pass filter allowing only the higher frequencies to feed the tweeter, a low pass filter that allows only the lower frequencies to feed the woofer, and a second high-pass filter that prevents subsonic signals from damaging the woofer.
Crossover frequency — The frequency at which one driver ceases to produce most of the sound and a second driver takes over. In the case of a two-way speaker the crossover frequency is usually between 1 and 3kHz.
Cutoff frequency: The cutoff frequency of a filter is the frequency at which the filter changes its response. For example, in a low-pass filter, frequencies greater than the cutoff frequency are attenuated, while frequencies less than the cutoff frequency are not affected.
De-esser — A device for reducing the effect of sibilance in vocal signals.
Delay — The time between a sound or control signal being generated and it auditioned or taking effect, measured in seconds. Often referred to as latency in the context of computer audio interfaces.
Digital Delay — A digital processor that generates delay and echo effects.
Digital Reverberator — A digital processor which simulates acoustic reverberation.
Digital Signal Processing (DSP): A general term describing anything that alters digital data. Signal processors have existed for a very long time (tone controls, distortion boxes, wah-wah pedals) in the analog (electrical) domain. Digital Signal Processors alter the data after it has been digitized by using a combination of programming and mathematical techniques. DSP techniques are used to perform many effects such as equalization and reverb simulation.
Since most DSP is performed with simple arithmetic operations (additions and multiplications), both your computer's processor and specialized DSP chips can be used to perform any DSP operation. The difference is that DSP chips are optimized specifically for mathematical functions while your computer's microprocessor is not. This results in a difference in processing speed.
Direct Coupling — A means of connecting two electrical circuits so that both AC and DC signals may be passed between them.
Dither — A system whereby low-level noise equivalent to one quantising level is combined with a digitised audio signal in such a way as to perfectly linearise the digital system. Dither must be employed whenever the wordlength is reduced, otherwise quantising distortion errors will manifest.
Double-ended Noise Reduction — A method for removing or attenuating the noise component of a recording or transmission system, in which the signal is pre-conditioned in a specific way which is reversed on playback. Most analogue noise-reduction systems are of the double-ended type, such as the Dolby and DBX systems.
Dry (cf. Wet) — A signal that has had no effects added.
Ducking — A system for controlling the level of one audio signal with another. For example, in a broadcast radio context a music track can be made to 'duck' or reduce in volume whenever there's a voice over.
Effects Loop — An interface system, usually involving separate send and receive connections, which allows an external signal processor to be connected into the audio chain. (cf. Insert Point)
Effects Return — An additional dedicated mixer input channel, usually with minimal facilities, designed to accommodate the output from an effects unit. (cf. Aux Return)
Encode/Decode — A system that modifies a signal prior to recording or transmission, and subsequently restores the signal on playback or reception.
Enhancer (cf. Exciter) — An audio processor designed to brighten audio material using techniques such as dynamic equalisation, phase shifting and harmonic generation.
Equalization (EQ): Equalizing a sound file is a process by which certain frequency bands are raised or lowered in level.
Exciter (cf. Enhancer) — An audio processor that works by synthesizing new high frequency harmonics.
Expander — A device designed to increase the dynamic range, typically by reducing the volume of low level signals (below a set threshold), or to increase the volume of high level signals (above a threshold).
FET-Compressor — A form of audio compressor in which an FET is used to provide variable signal attenuation. FET compressors are fast-acting in comparison to opto-compressors.
Filter (cf. Equaliser) — An electronic circuit designed to attenuate a specific range of frequencies. (See low-pass, high-pass and band-pass.)
Filter Frequency — The ‘turnover’ or ‘corner’ frequency of a high- or low-pass filter. Technically, the frequency at which the signal amplitude has been attenuated by 3dB.
Flanging — An effect which combines a modulated delay with the original signal, using feedback to create a dramatic, sweeping sound.
Gate — An electronic device (analogue or digital) designed to mute low level signals so as to improve noise performance during pauses in the wanted material.
Graphic Equalizer — An form of equalizer whereby multiple narrow segments of the audio spectrum are controlled by individual cut/boost faders. The name comes about because the fader positions provide a graphic representation of the EQ curve.
Harmonic Distortion — The addition of harmonics that were not present in the original signal caused by non-linearities in an electronic circuit or audio transducer.
High-Pass Filter: A high-pass filter attenuates all frequencies below a cutoff frequency. It is usually used to remove low-frequency rumble from audio files.
High-range (highs) — The upper portion of the audible frequency spectrum, typically denoting frequencies above about 1kHz.
Instrument Level — The nominal signal level generated by an electric instrument like a guitar, bass guitar or keyboard. Typically around -25dBu. Instrument signals must be amplified to raise them to line-level.
Limiting: Limiting is essentially a hard compressor. Limiting is often used to keep signals from going above a certain level, but can also be applied to create heavily compressed effects. Limiting should only be performed on peaks; if the Threshold level is set too low, heavy distortion will occur.
Limiter — An automatic gain-control device used to restrict the dynamic range of an audio signal. A Limiter is a form of compressor optimized to control brief, high level transients with a ratio greater than 10:1.
Linear — A device where the output is a direct multiple of the input with no unwanted distortions.
Line-level — A nominal signal level which is around -10dBV for semi-pro equipment and +4dBu for professional equipment.
Low-Pass Filter: A low-pass filter attenuates all frequencies above a cutoff frequency. Low-pass filters can be used as anti-alias filters or for general tonal shaping.
Noise-shaping: Noise-shaping is a technique which can minimize the audibility of quantization noise by shifting its frequency spectrum. For example, in 44,100 Hz audio quantization noise is shifted towards the Nyquist Frequency of 22,050 Hz.
Pan: To place a mono or stereo sound source perceptually between two or more speakers.
Quantization: Quantization is the process by which measurements are rounded to discrete values. Specifically with respect to audio, quantization is a function of the analog-to-digital conversion process. The continuous variation of the voltages of a analog audio signal are quantized to discrete amplitude values represented by digital, binary numbers. The number of bits available to describe these values determines the resolution or accuracy of quantization. For example, if you have 8-bit analog-to-digital converters, the varying analog voltage must be quantized to 1 of 256 discrete values; a 16-bit converter has 65,536 values.
Quantization Noise: Quantization noise is a result of describing an analog signal in discrete digital terms. This noise is most easily heard in low-resolution digital sounds that have low bit depths and sounds like a shhhhh-type sound while the audio is playing. It becomes more apparent when the signal is at low levels, such as during a fade out.
Resample: The act of recalculating samples in a sound file at a different rate than the file was originally recorded. If a sample is resampled at a lower rate, sample points are removed from the sound file, decreasing its size, but also decreasing its available frequency range. Resampling to a higher sample rate. This increases the size of the sound file, but does not increase the quality. When downsampling, be aware of aliasing.
Threshold: A threshold determines the level at which the signal processor begins acting on the signal. During normalization, levels above this threshold are attenuated.