IE65822B1

IE65822B1 - An interpolative analog-to-digital converter for band-pass signals

Info

Publication number: IE65822B1
Application number: IE466390A
Authority: IE
Inventors: Hans-Joachim Dressler
Original assignee: Aeg Mobile Communication
Priority date: 1989-12-27
Filing date: 1990-12-19
Publication date: 1995-11-15
Also published as: NO922548D0; AU6954491A; FI922956A7; PT96376B; NO180178B; NO922548L; IE904663A1; PT96376A; DE3943072C2; WO1991010283A1; EP0573409A1; FI922956A0; CA2072568A1; DE3943072A1; NO180178C; BR9007962A; DE59009191D1; JPH05505286A; FI922956L; EP0573409B1

Abstract

The invention relates to an interpolative analog/digital converter for band-pass signals in which the amplitudes of the scanning values are quantized in 2 consecutive stages. The first stage contains the actual analog/digital interface and the second stage is purely digital. The amplitudes of the scanning values are not quantized, as in other processes, using finely quantizing elements at the analog/digital interface, but by overscanning, shaping of the quantization noise spectrum and digital interpolation. As a result, very coarsely resolving elements (with only 2 quantization stages in the extreme case) can be used at the actual analog/digital interface.

Description

The invention concerns an analog-to-digital converter for digitizing analog band-pass signals, in accordance with the preamble of claim 1, particularly for the carrier frequency or intermediate frequency in radio receivers. Digitization minimizes the circuit cost at the analog-to-digital interface and avoids the difficulties which arise with analog solutions because of non-ideal component characteristics (noise, aging, temperature drift, DC-offset etc.).

Digitization of IF-signals is known for instance from NTZ Archive, Vol. 5, No. 12/1983, p. 353ff. The disadvantage of the known solutions is that the analog-to-digital converter converts the band-pass signal actually at the analog-to-digital interface with the full amplitude resolution and must therefore be provided with many precision components at very great cost. On the one hand such a solution is expensive. On the other hand, the sampling time must be all the more precisely adhered to in the analog-to-digital converse, the greater the signal portions to be processed and the more finely the input signal is to be resolved. As a degree of variance in the switching performance of these components cannot be completely eliminated, the amplitude resolution, achievable by means of conventional methods, is subject to an upper limit.

Methods for analog-to-digital conversion of signals which have only a few quantization steps at the analog-to-digital interface but have nevertheless high resolution, are interpolative methods.

Interpolative analog-to-digital converters are known for instance from IEEE Trans. Conmiun. Vol. Com. 33, No. 3, p. 249ff. These are however unsuitable for the analog-to-digital conversion of band-pass signals, as only those signals having spectral portions which lie well below the sampling frequency are converted with the desired resolution.

It is an object of the invention to effect a high-resolution analog-to-digital conversion of band-pass signals having only one or - 1 a few quantization steps. This object is attained by means of the feature stated in the characterising portion of claim 1. Advantageous constructions and/or further embodiments are to be derived from the sub-claims. The interpolative method for the analog-to-digital conversion of band-pass signals is carried out by the amplitude quantization of the sample values in two successive stages. The first stage (stage 1) then contains the actual analog-to-digital interface, while the second stage (stage 2) has purely digital functions. The amplitude quantization of the sample values is effected here in a different manner from non-interpolative methods, namely not with finely quantizing components at the analog-to-digital interface, but rather by means of over-sampling, spectral shaping of the quantization noise in the first stage and digital interpolation in the second stage.

The invention will be described in the following text with reference to the drawings.

According to the sampling theorem for band-pass signals, ah analog band-pass signal with the band-width B can be represented by sample values T = 1/fa without any loss of information, if the equation fA > 2B (1) is maintained between the sampling frequency f^ and the band-width B of the band-pass signal, and the relationship fm = (2m + 1) . fA/4 (2) with m = 0, I, 2 ... is maintained between the central frequency fm of the band-pass signal and the sampling frequency f^. In the following text, m is designated as band-pass index and the sampling of band-pass signals is described as band-pass sampling.

In the real band-pass sampling with an analog-to-digital converter, besides the sampling of the band-pass signal (time quantization), an amplitude quantization in accordance with the - 2 resolution of the analog-to-digital converter is also effected. The interpolative analog-to-digital converter according to the invention converts the band-pass signals in two successive stages, which are both * carried out at the clock frequency of ► fA > 2NB (3) in which N may be much greater than one (high over-sampling). The equation (2) still applies. Thus fA = 4 fm > 2NB (4) 2m ι- 1 results for the sampling frequency.

Figure 1 shows the first stage of the analog-to-digital converter according to the invention. It is a control loop consisting of a roughly quantizing analog-to-digital converter (in the extreme case with a resolution of only one bit), a digital-to-analog converter of corresponding resolution, comparison points, linear networks H(p) and multipliers a. and b.. The clock rate f& for both the analog-to-digital converter and the digital-to-analog converter is to be selected in accordance with equation (4).

The coefficients a. and b. are to be selected in such a manner that the control loop is stable and the quantization noise in the useful frequency region is minimal. The stability criteria are resolved by methods usual in control technology, such as, for instance, the Bode diagram. The noise is minimized by the coefficients a. and b. being selected in such a manner that the circuit responds approximately like a pure system of n-th order, n denoting the number of networks H(p) employed.

The linear networks H(p) are resonators, their resonance frequency being adjusted to the central frequency or to the near vicinity of the central frequency. Those networks which have the same or similar performance as networks giving 1st order by means of - 3 low-pass band-pass transformation from integrated or approximately integrated networks are suitable as resonators. These are for example: Linear networks, energized by a sine wave of the frequency f switched on at the time point t = 0, which responds at the exit to a sine wave, the amplitude of which rises proportional to the time t. or Linear networks, energized by a sine wave of the frequency f switched on at the time point t = 0, which responds at the exit to a sine wave, the amplitude of which rises in a finite time interval 0 < t < tQ in proportion to the time t. tQ should then be less than the sampling period. or Resonators having a resonance frequency which corresponds exactly or approximately to the central frequency of the input signal. or Resonators having a resonance frequency which corresponds exactly or approximately to the central frequency of the input signal as well as having the same or similar performance as networks giving 1st order by means of low-pass band-pass transformations from integrating or approximately integrating networks. or Band-passes, which are narrow-band in comparison with the band-width of the input signal, the central frequency of the band-passes corresponding exactly or approximately to the central frequency of the input signal. or Band-passes, which are narrow-band in comparison with the band-width of - 4 the input signal, the central frequency of the band-passes corresponding exactly or approximately to the central frequency of the input signal as well as having the same or similar performance as t networks giving 1st order by means of low-pass band-pass transformation of approximately integrating networks. r By means of the feedback of the roughly quantized sample values and comparison with the input signal and with the signal variations at the exit of the networks H(p), the signal portions of the error signals are spectrally weighted or spectrally shaped in the vicinity of the resonance frequency as compared with other signal portions. The output signals of the networks H(p) are added in accordance with the weight factors a^ and fed to the analog-to-digital converter, which performs a rough amplitude quantization. By means of the feedback, a roughly quantized series of sample values are provided at the exit from the analog-to-digital converter in such a manner that they correspond in the useful frequency region to the ideal sampled input signal, apart from a residual noise.

Fig. 2 shows the variation of the quantization noise 1 and the band-pass signal 2, which still contains a small amount of residual noise 3. It can be seen that the significant portions of the quantization noise lie largely outside the useful frequency region.

In the second stage of the analog-to-digital converter according to the invention for band-pass signals, the fine quantization of the previously roughly quantized sample values takes place by means of digital interpolation. This stage is also driven at the same clock frequency f^ according to equation (4) as for the first stage. To carry out this interpolation, two fundamental possibilities exist, namely either digital interpolation with a digital band-pass having a midband and band-width corresponding exactly or approximately to the useful frequency band of the band-pass signal under ideal band-pass sampling without amplitude quantization, or digital interpolation with a digital quadrature modulator.

In the previously described digital interpolation with a digital band-pass, the significant portions of the quantization noise - 5 lie outside the useful frequency region. Finely quantized sample values are thus available at the exit of this band-pass.

Fig. 3 shows a digital quadrature modulator for the analog-to-digital converter according to the invention for band-pass signals. It consists of a quadrature mixer 4 having two digital multipliers 5, 6 and two digital low-passes 7, 8. The digital low-passes 7, 8 are arranged in such a manner that their pass band-width corresponds exactly or approximately to the useful frequency region of the equivalent low-pass signal (cut-off frequency f = B/2). As the information content of a band-pass signal is not contained in its carrier position (central frequency) but rather in its equivalent low-pass signal, a fine quantization of the equivalent low-pass signal can be carried out. In the digital quadrature mixer 4, the output signal from the first stage is multiplied 5 by Cos Czr/2 χ K) at the digital multiplier and is then sent through the digital low-pass 7, at the exit of which the finely quantized real part of the equivalent low-pass signal is available. Similarly, the output signal from the first stage is multiplied by -sin (λ/2 χ K) at the digital multiplier in the digital quadrature mixer 4 and is then sent through the digital low-pass 8, at the exit from which the finely quantized imaginary part of the equivalent low-pass signal is available.

The sampling rate at the exit from the two low-passes 5, 6 can be reduced by the factor 2N in accordance with the band-width of the digital interpolation filter.

Fig. 4 shows the digital quadrature mixer 4 and, for comparative purposes, a simplified digital quadrature mixer 9. As the two digital carrier sequences Cos (k x/r/2) and sin (k xtt/2) only adopt values from the set (-1,0,1), the digital multipliers 5 and 6 need not be used in the realization of the digital quadrature mixer 4. Instead the switches 10, 11 are used, which periodically run through the switch settings 0, 1, 2, 3 in synchronisation with the sampling frequency f^, and digital components, which carry out the trivial multiplication by the factors 0, 1 and -1.

Fig. 5 shows a circuit-diagram realisation of a network - 6 H(p). It consists of a parallel resonant circuit comprising the capacitor 12 and the coil 13, which is fed by the grounded-emitter transistor 14 as a voltage-controlled current source. The ohmic resistance for the operating point adjustment is not shown.

Fig. 6 shows a circuit-diagram realisation of the first stage having only one linear network H(p). It consists of a parallel resonant circuit with capacitor 16 and coil 17, the comparator 18, the flip-flop 19, the transistors 20, 21 and a voltage-controlled current source 22. The flip-flop 19 is driven at the frequency fA and thus controls the switching states of the comparator 18. This, in turn, is controlled by the parallel resonant circuit comprising capacitor 16 and coil 17. The collector currents of the two transistors are controlled by the difference between the voltages ϋχ and 0^. Thus the diferential amplifier consisting of the transistors 20, 21 and the constant current source 22 functions at the same time as a comparison point and as a voltage-controlled current source for feeding the parallel resonant circuit, consisting of capacitor 16 and coil 17 and the constant current source 22. The constant current source 22 can be realised, for example, by an ohmic resistance or a transistor circuit.

Claims

1. An analog-to-digital converter for band-pass signals of the band-width B, characterized in that it is operated with the sanpling rate f A = --------------------------fm > 2 NB 2m + 1 wherein fm is the central frequency of the band-pass signal m = 0; 1; 2; 3; .. and N » 1, that it consists of two stages connected in series, that the structure of the first stage consists of an A/D converter having at least one bit resolution, of a D/A converter having the corresponding resolution, of at least one network H(p), of coefficient elements ai and bi, which can also adopt values 0 and 1, as well as of comparative .positions, that in stage 1, the input signal is roughly quantized with the sampling frequency fA and is fed back by the D/A converter and is compared to the signal waveforms at the exits of the networks H(p), and thereby the signal portions of the error signals are spectrally formed with respect to the remaining signal portions, that a sequence of sampling values is generated at the exit of the first stage, so that these sampling values correspond to the ideally sanpled input signal in the useful frequency region except for a residual noise, and -8that in a second stage it interpolates the signal contained in the first stage.

2. ) An analog-to-digital converter according to claim 1, characterized in that it carries out the interpolation in the second stage by means of a digital band-pass.

3. ) An analog-to-digital converter according to claim 1, characterized in that the signal in stage 2 is firstly quadrature mixed and is digitally interpolated by means of two digital low-passes, at the exits of which sampling values of in-phase and quadrature components of the input signal are available.

4. ) An analog-to-digital converter according to claim 2, characterized in that the digital band-pass corresponds exactly or approximately in its band-pass width to the wanted frequency region of the input signal for band-pass sampling at f&.

5. ) An analog-to-digital converter according to claim 3, Characterized in that the pass-band width of the digital low-passes corresponds exactly or approximately to low-pass signal . which is equivalent to the band-pass signal.

6. ) An analog-to-digital converter according to claim 3 or 5, characterized in that the quadrature mixer consists of two preferably electronic switches, which periodically run through the switch positions 0, 1, 2, 3 and which are associated to the respective switch positions of digital elements which perform the multiplication by 0; 1 and -1.

7. ) An analog-to-digital converter according to one of the preceding claims, -9characterized in that the coefficients al and bl are chosen so that the circuit of the first stage is stable and the quantization noise is optimally shaped.

8. ) An analog-to-digital converter according to one of the preceding claims, characterized in that the network H(p) or the networks H(p) consist of resonators, the resonance frequency of which correspond exactly or approximately to the central frequency fm of the input signal.

9. ) An analog-to-digital converter according to any of the preceding claims and substantially as discribed herein with reference to, and as shown in, the accompanying drawings.