GB2159014A - Switched capacitor filter - Google Patents

Abstract

A switched capacitor filter system 54 operating at a centre frequency fo and bandwidth BW comprises a switched capacitor (SC) band-pass anti-aliasing filter (AAF) 52 and an SC anti-imaging filter (AIF) 53 connected to the respective input and output of an SC band-pass filter 51 having a band-width BW and a centre frequency fi(<fo). The centre frequencies fo of the AAF and AIF 52 and 53 are related to the centre frequency fi of the SC filter 51 and the switching frequency Fs of SC filter 51 by the relationship fo=n Fs+fi where n is an integer. By adopting frequency translation filtering very narrow relative bandwidth filters can be achieved. <IMAGE>

Description

SPECIFICATION Switched Capacitor Filter Circuits The present invention relates to switched-capacitor (SC) filter circuits and in particular, though not exclusively, to SC bandpass filters with relative bandwidths below 1%. These filters are largely used in many signal processing applications, particularly in Telecommunications systems.

At the present time, there are available two methods for the realization of SC bandpass filters, namely the conventional ladder simulation approach and the N-path filter approach. The former method is applicable only for bandpass filters with relative bandwidths above 1% because below this limit the sensitivity and the capacitance spread of the SC filter become too high for practical feasibility. For bandpass filters with relative bandwidths below 1%, the N-path filter approach has been used. In this method, the SC bandpass filters have low sensitivity and low capacitance spread (the total capacitance, however, is similar to the above conventional realizations) but suffer from severe reduction of dynamic range due to frequency translated signals and clock feed through components occurring in the passband.

The object of the present invention is the realization of a very narrowband bandpass filter system, ie relative bandwidth below 1%, capable of combining the features of low sensitivity of the N-path filters, together with good dynamic range and acceptable capacitance spread of conventional SC ladder filters. The present invention will be especially suitable for those applications of very narrowband bandpass filters where poor dynamic range of N-path filters has been a serious limitation.

The invention is a frequency-translated SC filter system in which the frequency translations, ie aliasing and imaging, normally associated with SC filters are deliberately employed to allow a conventional SC bandpass filter with acceptable capacitance spread operating at a low mid-band frequency to realize a bandpass response at a much higher frequency and with correspondingly narrower relative bandwidth. The invention preferably includes a bandpass anti-aliasing filter and a bandpass anti-imaging filter which define the frequency band that is translated from the system input to the system output via the SC bandpass filter.

In one arrangement of the invention the sampling frequency of the SC filter is selected to be four times the centre frequency of the filter. This has been found to be an optimum relationship for minimising the spread of capacitance values required in the circuit, hence minimising the required area of semi-conductor required in monolithic implementation.

Preferably SC decimators and SC interpolators are used respectively in the AAF and AIF circuits to change the sampling rates and also to produce the bandpass response.

The invention will now be described by way of example only with respect to the accompanying Drawings of which: Figures la--e show the aliasing spectrum at the input of the SC filter, the baseband of the SC filter with bandpass frequency responses, and the imaging spectrum at the output, of the SC filter; Figure 2 is a block diagram of an analogue sample data switched capacitor (SC) filter; Figures 3a-c illustrate the application of a low pass anti-aliasing filter (AAF) and a low pass anti-imaging filter (AIF) with reference to the Figure 1 spectra as in a conventional SC filter system; Figures 4a-c illustrate the application of bandpass AAF and AIF filters in a frequency translated system according to the present invention; Figure 5 is a block diagram of the invention corresponding to Figure 1;; Figure 6 is a 6th order SC bandpass filter used to illustrate the invention; Figures 7 and 8 show frequency response curves illustrating the design of decimator and interpolator for use in the frequency-translated SC filter system; Figure 9a shows a complete schematic filter system and Figure 9b is a circuit timing diagram; Figure 10 shows the measured frequency response of the filter system of Figure 9 at 20 KHz compared with the frequency response of the SC bandpass filter alone at 4 KHz; Figure 11 is a block diagram of an alternative SC bandpass filter system; Figure 12 illustrates the amplitude/frequency requirements of the SC decimator of Figure 11; Figure 13 shows a schematic representation of the SC filter system of Figure 11;; Figures 14a-c show respectively the gain/frequency characteristics for the IIR circuits, the FIR circuits and the overall anti-aliasing characteristic; and Figure 15 shows the measured overall response and noise performance of the Figure ii circuit.

An analogue sample-data SC filter operating at a sampling rate Fs has a well known periodicity of the filtering function giving rise to the typical input and output spectra shown in Figure 1 for the case of a SC bandpass filter. The repetition of the frequency response at the input and output of the SC filter distorts the information that it conveys and to prevent this undesirable situation an analogue sample-data SC filter system, rather than a SC filter alone, must be considered. The typical block diagram of a SC filter system is shown in Figure 2.In this system, the SC filter 1 defines the shape of the required frequency response of the filter system; the anti-aliasing filter (AAF) 2 is a suitable prefiltering network that passes only one of the multiple possible input frequency bands relating to the filter response; the anti-imaging filter (AIF) 3 is a suitable post-filtering network that passes only one of the multiple possible output frequency bands also relating to the filter response. In the SC filter system 4 formed in this way there must be a unique correspondence between the input and output frequency bands of the filter system. Two possible modes of operation of a SC filter system are illustrated in Figures 3 and 4 by means of spectral interpretations.

The spectral sequence shown in Figure 3 refers to the conventional mode of operation. Here, the AAF and AIF are lowpass and select input and output frequency bands equal to the baseband of the SC filter, and reject the translated high frequency bands above Fs/2 at the input and output.

In this conventional mode of operation, the selectivity of the filter system is entirely determined by the SC filter 1, and the baseband of the system is, of course, contained within the conventional Nyquist band of the SC filter.

The present invention makes use of the wideband spectrum at the input of SC filters, based on the fact already mentioned that there are multiple bands that convey exactly the same information as contained in the baseband of the SC filter (0 Hz-1/2 Fs). In the same way there are multiple images of the baseband frequency response at the output of SC filters. Figures 4(a)-(c) illustrate the frequency translated filter system of the invention wherein the AAF and AIF filters are both bandpass filters which pass a particular high frequency band in the range nFsto (n+1/2) Fs (where n is an integer) and attenuate the frequency bands below n Fs, including the baseband (n=O) and above (n+1/2)Fs.Thus the input signals 41 in the frequency band fO+BW/2 (where BW is the band-width) are translated down to the SC filter baseband (42) f1 +BW/2 and are then translated up, producing the output image 43 at fO+ BW/2. The problems of aliasing distortion and imaging are avoided by ensuring that only one frequency band is allowed at the input and only one frequency band at the output.

Figure 5 illustrates the frequency translated SC filter system in which the SC filter 5, has a bandpass frequency response centred atfl, with a bandwidth BW and a switching frequency Fs which satisfies the Nyquist criterion, ie FS > 2FmaX where Fmax is the maximum frequency of interest in the baseband of the SC filter. The selectivity of the filter, expressed in terms of the equivalent Q-factor is O, =fl/BW. The AAF 52 and AIF 53 have wideband bandpass frequency responses centred at fO=(n Fs+fl), which means that the positive band associated with the n'th multiple of the basic sample rate is selected.For a given filter specification, ie given values of fO and BW in the filter 54, the values of n, Fs and fl are free parameters in the design of the filter system. The overall filter system has an equivalent Q-factor given by: Q=Q1 (fO/fI) Two important aspects characterise, in general, these SC filter systems that are based upon the operations of frequency translation throughout the various stages in the system. On the one hand, for a given frequency of the SC filter system (fO), the actual filtering operation in the system is performed at a centre frequency (f1) that can be made lower than fO, thus relaxing the speed and power requirements of the operational amplifiers (OA's), yielding improved capability of SC filter systems for high and very high frequency applications.The second important characteristic of this mode of filtering is that it relaxes the complexity of the prototype filter compared to a conventional baseband filter system for the same overall selectivity. This fact makes the variability of the frequency response much less critical and also reduces the capacitance spread and thus reduced the total capacitance area of the SC filter; it therefore offers a very attractive approach to the realisation of very narrowband SC bandpass filter systems (relative bandwidths below 1%), yet keeping the conventional methods of design for the prototype SC filter in the system. From a system point of view there is, of course, an increase in complexity of the required filters for anti-aliasing and anti-imaging.However the characteristics of these circuits are far less critical than the stringent characteristics (and perhaps) unrealisability) of conventional SC filters with very narrow relative bandwidths because they preferably implement all-zero transfer functions that can be realised by non-recursive networks.

The description above indicates that the filters for anti-aliasing and anti-imaging play an important role in the operation of the overall system, and their constraints are rather different from those required in conventional filter systems.

In general, the AAF and the AIF each comprise a continuous-time filter and a sample-data filter that provide the required alteration of the sampling rate between the input (output) of the system and the input (output) of the SC filter. Because of the well known limitations of MOS technology, a primary condition for practical feasibility and integration of the proposed filter system compels the continuous-time filters to have the same degree of complexity as in conventional filter systems; they should be of the lowpass type and have a maximum order of three.

The constraints for the continuous-time filters yield two important characteristics of the sample-data filters. In the first place, since in the proposed system we have fO > Fs, this means that in the time-domain the sample-data filters must provide large factors of sampling rate alteration. In the second place, we have seen that these filters have a bandpass filter response which is specified to give the required attenuation of the aliasing and imaging frequency bands, while introducing a negligible amplitude distortion in the passband of the system. Both these characteristics in the time and in the frequency domain can be efficiently implemented using multirate sample-data SC decimators, for sampling rate decrease, and SC interpolators, for sampling rate increase. The SC decimators and SC interpolators required in this system have characteristics that are rather different from those used in a more conventional bandpass filter system.

Nevertheless we can also tailor the SC decimators and the SC interpolators for the specific requirements of this application.

The practical feasibility of a frequency translated system depends on the ability to design SC decimators and SC interpolators to meet the requirements of the AAF and AIF. The design of these SC decimators and SC interpolators is given for a demonstration frequency-translated SC filter system with 0.4% relative bandwidth and centre frequency fro=20 KHz. The system uses a SC bandpass filter with 2% relative bandwidth, centre frequency f1 =4 KHz (absolute bandwidth BW=80 Hz) and switching frequency Fs=16 KHz. The filter shown in Figure 6, uses a conventional coupled biquad structure that simulates a 6th order elliptic bandpass RLC prototype. The even and odd switches, E and 0, are switched as illustrated by the time frame 61. For this SC filter, the equation Fs=4f, gives the optimum switching frequency for a minimum capacitance spread.For this demonstration system the decimators and interpolators were designed to reject unwanted frequency translated responses by a minimum of 32 dB, and the switching frequency at the input and output of the system is 192 KHz allowing the continuous-time AIF filters to be of 3rd order and 2nd order respectively.

Considering the interpolator first, the response specification may be derived by considering the spectrum at the outputofthe SC bandpassfilter (centre frequency f,=4 KHz, switching frequency Fs=16 KHz) as shown in Figure 7-a. One possible solution is the FIR bandpass frequency response of Figure 7-b which has been derived using the technique of placing single zeroes at the centre of the images to be attenuated.In the design of the interpolator it is convenient to use two cascaded sections having the intermediate sampling rate of Fsl =48KHz. This switching frequency is the minimum integer multiple of Fs= 16 KHz that encompasses the image to be selected at fro=20 KHz in the baseband filtering mode, ie fO < Fs1/2, thus allowing, thereafter, the signals to be treated as in a conventional system. To increase the sampling rate from Fs=16 KHz to Fas1 =48 KHz we need a high pass section which realises the zeroes below fro=20 KHz, ie at 4 KHz and 12 KHz (and their higher frequency repetitions), as shown in Figure 7-c.Having the signals in a baseband filtering mode with fro=20 KHz and Fs1 =48 KHz, the sampling rate is further increased to 192 KHz using a lowpass SC interpolator selection which realises the remaining zeroes at 48 KHz+20 KHz, 96 KHz+20 KHz and 144 KHz+20 KHz as shown in Figure 7-d. Due to the operation of the SC bandpass filter at Fs=16 KHz, with sampled and held output signals, the frequency response of the interpolator is multiplied by the sinc (x) function which, for 0 dB gain reference atfO=20 KHz, yields an undesirable amplification of the images below 20 KHz. Table 1 shows that in order to meet the specified imaging attenuation of 32 dB at 44 KHz and 12 KHz the minimum attenuation of the interpolator at these frequencies must be 46 dB and 36.4 dB respectively.The anti-imaging notch bandwidths of the interpolators are only 100 Hz at 4KHz and 190 Hz at 12 KHz; nominally, they encompass the 80 Hz bandwidth of the SC band-pass filter but do not given practical margins on account of frequency response sensitivity and practical tolerances of the capacitance ratios in the system. One possible solution to this problem is to introduce an extra single zero at 8 KHz yielding safe attenuation bandwidths of 360 Hz at 4 KHz and of 560 KHz at 12 KHz, but this is not attractive because of increased complexity of the SC circuits. One alternative solution is to replace the FIR highpass section by a bilinear highpass notch biquadratic section, with two-phase switching and pole frequency f,=20 KHz, pole Q-factor Up=50 and notch frequency fz=80 KHz.

This gives the frequency response of Figure 8 which, together with the FIR lowpass section, gives a minimum imaging rejection of 32 dB at 4 KHz+225 Hz and 35 dB at 12 KHz+340 Hz.

TABLE 1 Performance of FIR Interpolator with Sinc (x) Effect.

Minimum Notch Frequency Sinc (x) Attenuation Bandwidth (KHz) (dB) (dB) (Hz) 4 14 46 100 12 4.4 36.4 150 20 0 0 - 28 -2.9 29.1 685 The above architecture is chosen also for the SC decimator: (i) an FIR lowpass section, with zeroes at 144 KHz+20 KHz, 96 KHz+20 KHz and 48 KHz+20 KHz, reduces the sampling rate from 192 KHzto 48 KHz and (ii) a bilinear highpass notch biquadratic section with pole Q-factor Qp=20 (the Q-factor is lower than in the interpolator because it is not necessary to compensate for the sinc (x) function in the aliasing spectrum) decimates the signal from 48 KHz to 16 KHz and attenuates the aliasing signals at 4KHz and 12 KHz and their higher frequency repetitions.

The complete schematic of the demonstration frequency-translated SC filter system is given in Figure 9-a. The conventional SC band-pass filter has been designed for minimum capacitance spread and operates at Fs=16 KHz; the bilinear highpass notch biquads of the decimator and interpolator are designed for maximum dynamic range. The various stages of the system are interconnected without the need for any interstage sample and hold circuitry. The switch timing of the whole system is shown in Figure 9-b. Each SC circuit is assigned one time frame (TF) with time slots that are locally generated from the system clock bus using simple logic functions. The synchronization of the time frames satisfied the interstage sampling conditions throughout the system.The timing for the switching phase "1 ", for example, indicated by reference numerals 91-96 is given by the number "1 " in the respective time frame TF1-TF6.

A discrete component model of the SC filter system of Figure 9-a was built using CMOS 4016 analog switches; CMOS 7611 operational amplifiers for the bandpass filter. BIMOS 3140 operational amplifiers for the decimator and interpolator and standard digital circuits for the switching circuitry.

The measured frequency response 101 of the filter system at 20 KHz is given in Figure 10 and compared with the frequency response of the SC bandpass filter alone 102 measured at 4 KHz. The maximum noise level was dB referred to a system input level of 100 mv rms, for a measuring bandwidth of 10 Hz. The measured noise for the SC bandpass filter alone is very similar. The measured worst-case levels of frequency translated signals were dB for aliasing at 12 KHz and -34 dB for imaging at 12 KHz, in agreement with the design specifications.

The AAF and the AIF, consisting primarily of an SC decimator and an SC interpolator, respectively, are designed to selectthe alias and imagefrequency-translated components centred atfO=Fs+F1 and to reject by a design minimum the unwanted frequency-translated components below F max KHz, a frequency much higher than the mid-band frequency of the system. The attenuation of the unwanted frequency-translated components at Fmax and above can then be obtained in the normal way using continuous-time, low complexity, low pass active-RC pre and post filters. The complete system for an alternative filter system employing the same Figure 6 SC filter is shown in Figure 11. The signal at the input 111 is low pass (LP) filtered (112) and then connected to the SC decimator 113 which has a bandpass centre frequency of 20 KHz and a switching rate f's of 384 KHz.The decimator output is connected to the SC filter 114 (shown in detail in Figure 6). The output from the SC filter 114 is connected to the SC interpolator 115 and then via a second LP filter 116 to the filter output 117. The SC interpolator 115 also has a centre bandpass frequency of 20 KHz and a switching frequency of 384 KHz. Decimation and interpolation are basically linear filtering processes.

Various possible architectures can be implemented using only Finite Impulse Response (FIR) SC circuits, only Infinite Impulse Response (IIR) circuits or combinations of these. The SC decimator 113, described in greater detail later, is shown to include an FIR LP circuit 118 and a LPN-HPN circuit 119 which decrease the sampling rate from 384 KHz at the input to 16 KHz at the output. Similar circuits 1110 and 1111 in the SC interpolator 115 increase the sampling rate back to 384 KHz at the output.

The complete architecture and schematic of the SC filter system is shown in Figure 12. The conventional SC B P filter 114 (shown in detail in Figures 6 and 9) is coupled to the input decimator and output interpolator. The aliasing and imaging rejection of the circuit is designed to reject frequencies my364 KHz.

Considering the SC decimator first. The amplitude/frequency specifications required for the SC decimator, for example, are schematically illustrated in Figure 12. The portions of the spectrum 121 with 96 Hz bandwidth relating to the passband of the SC bandpass filter have to be reduced by a minimum of 50 dB with respect to the passband components; the remaining portions of the spectrum between these bands are not important because they are translated into the stopband of the SC filter 114. The ampiitude/ frequency specifications for the SC interpolator are similar to those in Figure 12 but as before we have to bear in mind that the gain ofthe SC interpolator should compensate for the sample and hold attenuation as shown in Table 1.Thus the imaging spectrum is multiplied by the function sin (nf/Fs)/(nf/Fs) and therefore the AIF filter 115 must have a gain characteristic to compensate for this effect.

As stated above, the frequency components at 364 KHz and above are rejected by the continuous-time filter 112. A design decision was taken to employ a fast FIR SC circuit 118 for reducing the sampling rate from 24Fs=384 KHz to 12us=192 KHz, and whose frequency response is optimised to yield notch frequencies at 172 KHz and 212 KHz. For reducing the sampling rate from 12Fs=192 KHz to Fs=1 6 KHz and for attenuating the remaining unwanted frequency-translated components two solutions were considered.

One such solution, consisting of fast FIR SC circuits whose transfer functions implement a total of six complex conjugate pair zeroes, using only two Operational Amplifiers (OA's) and with a capacitance spread less than 10, has a practical drawback in the sophisticated multiphase switching schemes that the circuits required. For this application, we chose instead a solution employing IIR SC circuits, viz. the cascade of a highpass notch (HPN) biquad and a lowpass notch (LPN) biquad 119, whose transfer functions implement a total of two complex conjugate pair zeroes plus two complex conjugate pair poles. The resulting SC biquads employ four OA's and have higher capacitance spread, but they require a much simpler switching scheme with only two phases.The SC interpolator has a similar architecture, viz. the cascade of HPN and LPN SC biquads 1111, followed by a fast FIR SC interpolator circuit 1110. The IIR SC circuits of the interpolator are required to have higher selectivity than those in the decimator, and a gain of 14.89 dB, in order to compensate the midband gain error of a=-14.89 dB due to the sample and hold attenuation of the selected image.

The complete architecture and schematic of this SC bandpass filter system (apart from the continuous-time filters) are given in Figure 13. The SC decimator and interpolator blocks 131 and 132 employ FIR non-recursive polyphase SC structures and HPN and LPN SC biquads which are designed for maximum signal handling capability. The two interstage sample and hold circuits are needed in orderto guarantee the required input and output sampling conditions of the SC circuits. A fully synchronised switching scheme would have eliminated such sample and hold circuits but, on the other hand, it would not have provided a capability for independent programming of the switching frequency of the SC decimator and of the SC interpolator.A discrete component model of the system was constructed using CMOS 4016 analogue switches, CMOS 7611 OA's for the SC bandpass filter and for the ilR SC decimator and interpolator circuits, BiMOS 3140 OA's for the FIR SC circuits and standard CMOS digital components for the switching circuitry.

In order to measure the anti-aliasing performance of the SC decimator we employ a synthesised signal generator (HP-3330B) which provides the input signal frequencies from DC to 384 KHz, and a synthesised spectrum analyser (HP-3585A) which detects the output signals in the baseband of the SC decimator from DC to 96 KHz. A control unit (HP-9825A) controls both of the sweeping modes of the signal generator and spectrum analyser. In this way we obtained the results shown in Fig. 14a for the IIR SC circuits and in Fig.

14-b for the FIR SC circuit, yielding the overall anti-aliasing characteristic shown in Fig. 14-c. In such measurements, the sample and hold function affects only the detection of the signals in the baseband from DC to 96 KHz and thus no notch frequencies appear at 192 KHz and 384 KHz. For measuring the anti-imaging performance of the SC interpolator we interchange the sweeping modes of the signal generator and of spectrum analyser, and the results obtained are in agreement with the specifications. The measured overall response 151, and noise performance 152, of the system at 20 KHz is given in Fig. 15. The noise floor in the passband system is very similar to that of the SC bandpass filter alone, despite the reduction of the relative bandwidth.Table 2 summarises the measured results obtain for the SC bandpass filter system, showing the good performance both with respect to the accuracy of the bandpass response and with respect to the dynamic range.

TABLE 2 Specifications of the bandpass filter system Midband frequency 20 KHz Gain OdB -3 dB bandwidth 96 KHz Stopband rejection > 40 dB Aliasing rejection > 50 dB Imaging rejection > 50 dB fmin continuous-time filter < 364 KHz Measured results of the bandpass filter system Midband frequency 20.007 KHz Gain 47 dB -3 dB bandwidth 97 KHz Stopband rejection > 42 dB Power supply t5V Maximum input signal 1 Vrms Dynamic range 78 dB Alaising rejection ( < 364 KHz) > 50 dB Imaging rejection ( < 364 KHz) > 50 dB The results demonstrate that the anti-aliasing and anti-imaging specifications of the system are entirely met by suitable design of the AAF and AIF.Furthermore, the components of clock feedthrough at Fs= 16 KHz and at the multiples nFs, all of which fall outside the system bandpass, are also attenuated by the AIF (in this case, only by the SC interpolator) and their levels were below -50dB relative to the passband signals.

A single path frequency translated (SPFT) SC system can operate as a single side-band (SSB) generator or as an SSB detector by selecting different frequency-translated bands at the input and output of the system.

We have considered the design and implementation of single path frequency-translated (SPFT) systems for fully analogue signal processing applications without any form of constraint of the input and output spectra. This constitutes the most severe environment for operation of an SPFT system. There are further potential applications of SPFT systems, where the SC decimator circuits or/and the SC interpolator circuits can be substantially simplified, or even totally eliminated. Two such applications are briefly mentioned: 1. A filter band with N channel filters, eg for data transmission and speech processing system applications, can be realised using an SPFT system for each one of the channel filters. The SC decimators of the SPFT systems can be eliminated if the input spectrum is already band-limited for the entire filter bank.

This may be provided by a wideband filter converting the bandwidth of the N channel filters, and which can be implemented by other means.

2. SPFT systems are well suited for signalling frequency system applications. Often, the filtered signal at the output of the SPFT system will be required to interface directly a digital circuit, for further signal processing. For this purpose, it may be possible to use the output signal of the SC bandpass filter in the SPFT, thus eliminating the SC interpolator circuits.

Switched-capacitor filter systems are very attractive for the application of multirate signal processing techniques, aimed at improving various important parameters in the system such as capacitance spread, complexity of continuous-time filters, de-sensitivity of filter response and improved high-frequency capabilities.

The present invention represents an approach to the design of multirate SC filter systems whereby use is made of the periodicity of the actual frequency response of SC filters, in order to implement operations of frequency translation in the system. The invention is especially suitable, though not restricted to, for the realisation of SC bandpass filters with very narrow relative bandwidths combining the features of low sensitivity, low capacitance spread and good dynamic range which, altogether, are not encountered in known design methods. It is also expected that this type of frequency-translated SC filter system may be used to extend the capabiiities of MOS technology for the realisation of high-frequency SC bandpass filters with the benefits of well proved design techniques. As indicated above another application for this multirate frequency translated system is in the design of SC modulators and SC demodulators which, in turn, can be used in a variety of system applications. In communications, a very important area of application is in the translation between the formats of frequency multiplexing (FDM) and time multiplexing (TDM) in telephone transmission and switching terminals.

Claims (10)

1. A switch capacitor band-pass filter circuit comprising an input anti-aliasing filter (AAF), a band-pass filter connected to the output of the AAF and an anti-imaging filter (AIF) connected to the output of the band-pass filter, the arrangement being such that the AAF and the AIF are band-pass filters passing a frequency band higher than the frequency of the band-pass filter, the ratio between the centre frequencies of the bands being a function of the switching frequency of the band-pass filter.

2. A switched capacitor band-pass filter circuit as claimed in claim 1 wherein the switching frequency of the band-pass filter is four times the centre frequency of the filter.

3. A switched capacitor circuit as claimed in claim 1 or 2, wherein the AAF and AIF frequency bands are the same.

4. A switched capacitor band-pass filter circuit as claimed in any one of claims 1 to 3 wherein switched capacitor decimators and switched capacitor interpolators are used respectively in the AAF and AIF circuits to change the sampling rates and produce the band-pass response in order to select the respective required alias and image frequency-translated components.

5. A switched capacitor circuit as claimed in claim 4 wherein the AAF and AIF include a continuous time low pass (LP) filter in series with the respective decimator and interpolator, the cut-off frequencies of the continuous time LP filters being much higher than the mid-band frequency of the system.

6. A switched capacitor circuit as claimed in claim 4 or 5 wherein the decimator and interpolator each comprise a SC Finite Response (FIR) LP circuit in series with a SC Infinite Impulse Response (IIR) highpass notch (HPN) filter.

7. A switched capacitor circuit as claimed in claim 4 or 5 wherein the decimator and interpolator each comprise an FIR LP circuit in series with a cascade of a lowpass notch (LPN) circuit and an HPN circuit.

8. A switched capacitor circuit as claimed in any one preceding claim wherein the gain of the interpolator is designed to compensate for the mid-band gain error due to the sample and hold attenuation of the selected image.

9. A switched capacitor band-pass filter circuit substantially as described with reference to Figures 4 to 10 of the accompanying drawings.

10. A switched capacitor band-pass filter circuit substantially as described with reference to Figures 1 to 6 as modified with reference to Figures 11 to 15.