Numerix - DSP Glossary

Adaptive Differential Pulse Coded Modulation (ADPCM)	A speech compression algorithm that adaptively filters the difference between two successive PCM samples. This technique typically gives a data rate of about 32 Kbps.
Adaptive equalisation	A filtering system that can allow for the effects of a changing communications medium to be cancelled.
Adaptive filter	A filter that can adapt its coefficients to model a system.
Adaptive predictive coding	An LPC based speech compression technique that uses an adaptive predictive voice source.
Aliasing	The effect on a signal when it has been sampled at less than twice its highest frequency.
Amplitude Modulation	A communications scheme that modifies the amplitude of a carrier signal according to the amplitude of the modulating signal.
Anti-aliasing filter	An analog filter that is used prior to sampling to limit the signal bandwidth to less than half the sample rate (generally low pass) to prevent aliasing distortion.
Asynchronous communications	A communications system where the transmitter and receiver run independently. The beginning and end of the data packet are usually indicated by start and stop bits in the data stream.
Attenuation	Decrease in magnitude.
Autocorrelation	The correlation of a signal with a delayed version of itself.
Band-pass filter	A filter that only allows a single range of frequencies to pass through.
Band-stop filter	A filter that removes a single range of frequencies.
Bandwidth	The range of frequencies that make up a more complex signal.
Barrel shifter	Part of the ALU that allows single cycle shifting and rotating of data words.
Baseband	Signals that have a frequency spectrum based around 0 Hz. E.G. speech.
Baud rate	The rate at which symbols are transmitted over a communications channel. A symbol may contain one or more bits of information.
Bit rate	The rate at which bits are transmitted and equals the baud rate * the number of bits per baud.
Biquad	Typical 'building block' of IIR filters - from the bi-quadratic equation.
Butterfly	The smallest constituent part of an FFT, it represents a cross multiplication, incorporating multiplication, sum and difference operations. The name is derived from the shape of the signal flow diagram.
Companding	A logarithmic scheme for sampling analog signals that increases the resolution of signals with a low amplitude. Common standards include A-Law and u-Law.
Convolution	An identical operation to Finite Impulse Response filtering.
Correlation	The comparison of two signals in time, to extract a measure of their similarity.
Data flow architecture	A multi-processing architecture where individual processing elements perform multiple instructions on a many pieces of data.
Discrete Fourier Transform (DFT)	A transform that gives the frequency domain representation of a time domain sequence.
Discrete sample	A single sample of a continuously variable signal that is taken at a fixed point in time.
Echo canceller	A filter that will remove reflected signals on a transmission line that are caused by impedance mismatches.
Equalisation	A filter that will compensate for the effects of a communications channel.
Fast Fourier Transform (FFT)	An optimised version of the DFT.
Finite Impulse Response (FIR) Filter	A filter that includes no feedback and is unconditionally stable.
Floating-point	A number scheme that codes a value with a fraction and an exponent and allows a high signal dynamic range.
Frequency Division Multiplexing (FDM)	A communications system that divides a single channel into smaller ones with discrete frequency bands.
Frequency domain	The representation of the amplitude of a signal with respect to frequency.
Frequency Shift Keying (FSK)	A digital modulation scheme that uses a different frequency to represent different binary levels.
Full duplex	Communications in two directions simultaneously.
Gain	Amplification or increase in magnitude.
Half duplex	Communications in two directions, but only one at a time.
Harvard Architecture	A microprocessor architecture that uses separate busses for program and data, this is typically used on DSPs to optimise the data throughput.
High pass filter	A filter that allows high frequencies to pass through.
Hybrid	An analog 2 wire to 4 wire (and vice versa) converter.
Infinite Impulse Response (IIR) filter	A filter that incorporates data feedback. Also called a recursive filter.
Linearity	A measure of the performance of an ADC or DAC to convert signals with different amplitudes, to the same degree of accuracy.
Linear Predictive Coding (LPC)	A speech compression technique that is based on modelling the vocal tract with a time varying filter.
Low pass filter	A filter that allows low frequencies to pass through.
Multi-processing	The division of a process across several processors to improve the performance of the system.
Multi-tasking	The division of processor across several tasks, such that each one is able to receive its required number of processor cycles.
Multiple Instruction Multiple Data (MIMD)	See data flow architecture.
Modulation	The modification of the characteristics of a signal so that it might carry the information contained in another signal.
Parallel processing	The execution of tasks in parallel, either on a single processor via multi-tasking or across several processors by multi-processing.
Pass band	The frequency range of a filter through which a signal may pass with little or no attenuation.
Phase Shift Keying (PSK)	A digital modulation scheme that uses a constant frequency carrier with a variable phase.
Pipelining	A technique commonly used in high performance microprocessors that allows an instruction to begin execution before previous ones have been completed.
Pole	Artefact leading to frequency dependent gain in a signal. Generated by a feedback element in a filter.
Pulse Code Modulation (PCM)	The effect of sampling an analog signal.
Quadrature Amplitude Modulation (QAM)	A variation of PSK that incorporates AM to increase the number of bits per baud.
Recursive filter	See Infinite Impulse Response filter.
Resolution	The accuracy of and ADC or DAC circuit.
Sampling	The conversion of a continuous time analog signal into a discrete time signal.
Sample rate	The inverse of the time between successive samples of an analog signal.
Single Instruction Multiple Data (SIMD)	A multi-processing architecture where individual processing elements perform the same instruction on many pieces of data, also referred to as a systolic array.
Spectrum analyser	An instrument that displays the frequency domain representation of a signal.
Stop band	The frequency range of a filter through which a signal may NOT pass and where it experiences large attenuation.
Synchronous communications	A communications system where the data is transmitted and received at discrete times, which are usually synchronized by a clock signal.
Systolic array	See Single Instruction Multiple Data.
Time domain	The representation of the amplitude of a signal with respect to time.
Transducer	A piece of equipment that converts a physical signal into an electrical signal.
Twiddle factor	The coefficients of the FFT algorithm, typically a ¾ sine table.
Von-Neumann architecture	A traditional microprocessor architecture that uses the same bus for program and data.
z-domain	The discrete frequency domain, in which the jw axis on the continuous time s-plane is mapped to a unit circle in the z-domain.
Zero	Artefact leading to frequency dependent attenuation in a signal. Generated by a feedforward element in a filter.

Return To Home Page | SigLib DSP Library | DFPlus™ Filter Designer | DSP Abbreviations

If you have any comments or questions please email Numerix : numerix@numerix-dsp.com
Copyright © 2008, Indigo Consulting Ltd.. Permission is granted to create WWW pointers to this document. All other rights reserved. All trademarks acknowledged.