| 摘要 |
A positive signal value (+et) is added to positive signal samples (S1,S2) and a negative value (-et) is added to the negative signal samples (S3,S4). Since the fifth sample (S5) is zero, no correction value is added to it. In a subscriber line audio signal processing circuit, the polarity of a signal input to a digital adder circuit is sensed. The sensed polarity determines the choice of a stored digital representation of a value +et or -et in a memory. - The chosen value is output on one or other of a pair of bit lines leading to a multiplexer. According to the polarity sensed, a multiplexer selects the first or second bit line to provide the correction value to a second input of the digital adder. For simplicity the value et is chosen as an integral number, and represents the mean truncation error for the filter. (4/7) EPAB- EP-162076 B A method of compensating for the error which, in a digital signal with a given number of bits, occurs through truncation in a digital filter, particularly magnitude truncation, the signal being positive or negative relative to a given zero level, characterised in that when the signal is positive a positive first value is added to the signal, this value corresponding to a mean truncation error, and when the signal is negative a negative second value is added to the signal. (7pp) USAB- US4750146 A Positive and negative portions of a digital signal to be input to a filter are sensed and signals are provided representing positive and negative values of a mean error due to magnitude trucation of the signal in the filter. The positive and negative signals are respectively added to the positive and negative portions of the signal which is to be input into the digital filter. - The positive and negative signals are equal in magnitude. When the signal to be input to the filter has a value of zero no further signals are added. The average of the truncation errors produced by the multiplications in digital filter multipliers included in the subscriber audio-processing circuits. This average error is calculated in a arthmetic-logic unit.
|
