GB2266633A

GB2266633A - Eliminating frequency components

Info

Publication number: GB2266633A
Application number: GB9206886A
Authority: GB
Inventors: James Attew; Avigdor Steinberg
Original assignee: Snell and Wilcox Ltd
Current assignee: Snell Advanced Media Ltd
Priority date: 1992-03-30
Filing date: 1992-03-30
Publication date: 1993-11-03
Anticipated expiration: 2012-03-30
Also published as: GB2266633B; GB9206886D0

Abstract

A filter for eliminating a defined frequency component e.g. the chrominance component from a pulsed input video signal of width 2T has a filter 22 passing only said frequency and two adders 24, 28 each adding a portion of the filter output in phases respectively advanced and retarded by T/2, with respect to the outputs of delays 20, 26. The output of the filter has a better amplitude output in comparison to a simple band elimination filter and ripple components are evenly distributed evenly to the right and left to the output pulse. The filter may be modified to be an effective low pass filter by using a high pass filter in place of the band pass filter, 22. <IMAGE>

Description

FILTER WITH IMPROVED PULSE RESPONSE This invention relates to filters for the elimination of a defined frequency component, such as the chrominance within a television signal. The invention is particularly concerned with improving the pulse response of such a filter.

Tal < ing as an example a band pass filter, it is known that one effect of the passage of a pulse through a band pass filter is to introduce a train of ripples extending to the right of the pulse, along the time axis.

Particularly with video signals, these ripples can represent a serious impediment. In one known attempt to deal with this problem, the output of the band pass filter is subtracted from the input signal pulse. The arrangement is in-phase, that is to say a signal path containing the band pass filter is subtracted from the main signal path in which there is introduced a delay equivalent to the delay represented by the band pass filter. As will be described more fully hereafter, this approach has the effect of redistributing the spectral energy about the time axis with an undershoot introduced to the left of the pulse along the time axis. Whilst this offers some improvement, there remain significant ripples to the right of the time axis. Moreover, the effect of the subtraction process is to reduce the peal < height of the pulse.

It is an object of this invention to provide signal processing method and apparatus for eliminating a defined frequency component from a pulsed input signal there being an improved pulse response and, in an important example, a pulse response for video signals which has improved picture characteristics.

Accordingly, the present invention consists in one aspect in a signal processing method for eliminating a defined frequency component from a pulsed input signal of pulse width 2T, comprising a filtering step passing only said frequency component and a elimination step which eliminates the filter output from the pulsed input signal, characterised in that the elimination is performed in two stages, the phase of the filter output in each stage being respectively advanced and retarded by T/2 with respect to the pulsed input signal.

In another aspect, the present invention consists in signal processing apparatus for eliminating a defined frequency component from a pulsed input signal of pulse width 2T, comprising filtering means adapted to pass only said frequency component and elimination means for eliminating the filter output from the pulsed input signal, characterised in that the elimination means comprises two adders each adding a portion of the amplitude of the filter output, the phase of the filter output at each adder being respectively advanced and retarded by T/2 relative to the pulsed input signal.

The invention will now be described by way of example with reference to the accompanying drawings in which: Figure 1 is a block diagram of the prior art filter; Figure 2 is a series of wave form diagrams illustrating the operation of the filter of Figure 1; Figure 3 is a blocl < diagram of a filter according to the present invention; Figure 4 is a series of wave form diagrams illustrating the operation of the filter of Figure 3; and Figure 5 is a series of frequency plots.

Referring initially to Figure 1, there is shown a known filter arrangement comprising a delay line 10, a band pass filter 12 and an adder 14. The delay line 10 serves to compensate for the delay within the band pass filter 1 2. This delay is normally an even integral number of band pass central frequency half cycles. The summation coefficients in the adder are chosen to eliminate this central frequency. Thus the coefficient applied to the band pass filter output would normally be negative; a positive coefficient will be used if the delay is an odd integral number of band pass central frequency half cycles.

Turning to Figure 2, wave form (A) shows a typical input signal 2 pulse of sin x/x. The output of the band pass filter is shown at (B) and the result of subtracting (B) from (A) is shown at (C). The ripples which are added to the pulse on passing through the band pass filter are clearly shown in wave form (B). The result of subtracting (B) from (A) is to redistribute the spectral energy and it will be seen from wave form (C) that the effect of the ripples is reduced but not eliminated. An undershoot is introduced at either side of the pulse; this is generally not objectionable and in video signals may actually improve edge definition. Because of the subtraction, however, the overall height of the pulses is reduced.

According to the present invention the elimination of the filter output (whether by subtraction of in phase signals or addition of out of phase signals) is conducted two separate stages.

Referring to Figure 3, the main signal path comprises a delay line 20 having a delay equal to one half of the delay of the band pass filter 22, a first adder 24; a further delay 26 having a delay equal to that of the band pass filter 22; and a second adder 28. The output of the band pass filter 22 is applied to the adders 24 and 28 with coefficients K1 and K2 respectively. To ensure complete elimination of the band pass central frequency, K1 + K2 = 1.

The operation of the filter of this invention can best be 2 understood with reference to the wave forms of Figure 4. The sin x/x input pulse is shown at wave form (A). The output of the band pass filter is shown at wave form (B) and it will be seen that this is advanced relative to the signal pulse by T/2. The effect of adding wave forms (A) and (B) is shown at (C). Delay line 26 introduces a further delay of T producing the wave form shown at (D). A further addition of the wave form (B) to this delayed wave form (D) produces the output of the filter as shown at (E). It will be seen through comparison of wave forms (B) and (D) that at the stage of the second addition, the band pass filter output is effectively retarded with respect to the pulse by T/2. This has important consequences.In common with the prior art approach, an undershoot is provided to the right and left of the pulse along the time axis. In contrast with the prior art, the amplitude of the band pass filter output is, at each addition stage, zero at the maximum of the signal pulse. There is therefore no reduction of the peak pulse height.

Moreover, the two complementary portions of the band pass filter output are out of phase so that the ripples destructively interfere and, in the optimum arrangement, vanish.

The effect on the frequency response of the filter according to this invention is illustrated in Figure 5. Frequency plot (A) illustrates the band width of the pulsed input signal. Plot (B) illustrates the typical frequency characteristic of a band pass filter. Frequency plot (C) illustrates the frequency response of the filter according to the present invention. It will be seen that a sharper "notch" has been achieved at the expense of some distortion to the left and right, in frequency space, of the notch. This distortion will often be relatively unimportant. In the particular application of a notch filter to remove chrominance from a TV signal, the distortion is irrelevant.

It should be recognised that whilst in an optimum arrangement the signal pulse width 2T is directly related to the central frequency of the band pass filter 1/T, this is not essential. What is essential is that the phase of the two amplitude portions of the filter output are such as to provide a null at the maximum of the signal pulse and destructive interference in the described ripples. This will be achieved if phase is, respectively, advanced and retarded by T/2 with respect to the signal pulse of width 2T.

Whilst this invention has been described with reference to a band stop or notch filter arrangement, it will apply also to a low pass filter. In this case, the band pass filter of the described embodiments will be replaced by a high pass filter. Although it is less easy to see an application, a high pass filter arrangement could also be produced according to this invention.

Claims

1. A signal processing method for eliminating a defined frequency component from a pulsed input signal of pulse width 2T, comprising a filtering step passing only said frequency component and an elimination step which eliminates the filter output from the pulsed input signal, characterised in that the elimination is performed in two stages, the phase of the filter output in each stage being respectively advanced and retarded by T/2 with respect to the pulsed input signal.

2. Signal processing apparatus for eliminating a defined frequency component from a pulsed input signal of pulse width 2T comprising filtering means adapted to pass only said frequency component and elimination means for eliminating the filter output from the pulsed input signal, characterised in that the elimination means comprises two adders each adding a portion of the amplitude of the filter output, the phase of the filter output at each adder being respectively advanced and retarded by T/2 relative to the pulsed input signal.