GB2234411A - Integrated circuit for digital demodulation - Google Patents

Abstract

An electronic integrated circuit for digitally demodulating an input signal, comprising an analogue-to-digital convertor connected for conversion of the input signal to a digital signal, an analytic signal block responsive to the digital signal to generate an analytic signal, and means responsive to the analytic signal for calculating the amplitude, frequency and/or phase modulation of the input signal, the analytic signal block comprising a finite impulse response filter whose data rate is half the sampling rate of the analogue-to-digital convertor. The integrated circuit may be incorporated in a modulation meter in which the calculated values may be displayed or recorded. <IMAGE>

Description

Integrated Circuit for Digital Demodulation This invention relates to an electronic integrated circuit for digitally demodulating an input signal by recovering, from a carrier, the amplitude, frequency and/or phase of the signal with which it is modulated.

An analogue method exists which detects the envelope of the carrier to provide amplitude information, and uses a discriminator to provide frequency information. This method is cheap and convenient, but has a number of limitatfons. The AM (amplitude modulation) recovery is limited in analogue linearity and in bandwidth due to the peak detect function required. The FM (frequency modulation) recovery is subject to DC drifts in the discriminator, which limit the low frequency performance and prevent use as a DC-coupled frequency demodulator. The gain of the envelope detector and the discriminator will need calibration. In addition, the analogue outputs will usually have to be digitised for further processing.

Another known method mixes the incoming signal with a pair of references in quadrature, and so detects the modulation vector in amplitude and phase with respect to the reference. This technique is especially used In phase modulation systems. When the phase and quadrature outputs are digitised, it may be used for AM and DC coupled FM (frequency modulation) as well, but it is basically a narrow band system.

A digital method exists which involves digitising the IF (intermediate frequency) directly with a fast A/D (analogue-to-digital convertor), and then performing a Fourier transform on the readings. While the modulation sidebands may be seen, it is necessary to separate the odd and even portions of the sidebands to evaluate the AM and FM separately. This is possible if the IF frequency is known. In the case of the Fourier transform, it can only be estimated to an accuracy equal to the reciprocal of the transform length. While the IF frequency can be estimated more accurately over several transforms, the amount of data that needs to be stored and manipulated before getting a result becomes unwieldy.

The invention provides a digital demodulating circuit which does not require fast Fourier Transform circuitry yet which is capable of providing an adequate performance at a reasonable cost using suitable hardware. The invention consists in an electronic integrated circuit for digitally demodulating an input signal, comprising an analogue-to-digital convertor connected for conversion of the input signal to a digital signal, an analytic signal block responsive to the digital signal to generate an analytic signal, and means responsive to the analytic signal for calculating the amplitude, frequency and/or phase modulation of the input signal, the analytic signal block comprising a finite impulse response filter whose data rate is half the sampling rate of the analogue-to-digital convertor.

Preferably, the finite impulse response filter comprises a shift register connected to receive the digital signal, a multiplier array whose elements are connected to receive the contents of alternate elements of the shift register and also connected to a store of predetermined multiplication coefficients which implement a Hilbert transformer whose frequency response is symmetrical about a frequency equal to half the said data rate of the filter, an accumulator for summing the multiplied outputs of the multipliers to produce one quadrature component of the analytic signal, and means for sending the contents of the other elements of the shift register to produce the other quadrature component of the analytic signal.

In another aspect, the invention comprises means for mixing a modulated RF input signal with a local oscillator signal to generate an intermediate frequency input signal, an electronic integrated circuit according to the first aspect of the invention connected to receive as its input the intermediate frequency input signal, and means responsive to the calculated amplitude, frequency and/or phase modulation to provide an indication thereof.

One example of the use of the invention in a modulation meter will now be described with reference to the accompanying drawings, in which: Figure 1 is a block diagram of an integrated circuit embodying the invention; Figure 2 is a graphical representation of the frequency spectrum of a typical input signal, also indicating signlficant parameters of the demodulating circuit of Figure 1; Figure 3 is a diagram of the analytic signal block of Figure 1; Figure 4 is a flow diagram of the way in which the circuit of Figure 1 calculates phase from the analytic signal; Figure 5 is a flow diagram of the way in which the circuit of Figure 1 calculates amplitude from the magnitudes of the analytic signal components;Figure 5 is to be amended]; Figure 6A is a flow diagram of the determination of input frequency, AC-coupled FM and DC-coupled FM from the phase calculated in accordance with Figure 4; Figure 6B is a flow diagram of the derivation of phase modulation from frequency modulation; and Figure 7 is a flow diagram of the determination of input amplitude and AM from the amplitude signal produced in accordance with Figure 5.

With reference first to Figure 1, the incoming signal is first filtered in a bandpass filter 1. This filter serves as an anti-alias for an analogue-to-digital convertor 2. The data from the A/D is processed in an analytic signal block 3, where the sample rate is dropped by a factor of two, and a quadrature signal component is created. From the analytic signal, the phase and amplitude are directly calculated by blocks 4 and 5. From the instantaneous phase and amplitude, blocks 6 and 7 calculate the mean frequency and amplitude, and the variation against time of these quantities. These quantities, or selected quantities, are then displayed or recorded, depending on operator requirements indicated by manual controls on the modulation meter.

The relationship of the bandpass filter amplitude requirement, the nominal input frequency, and the A/D sampling frequency are shown on Figure 2. The range of allowable input frequencies is centred on IF, going from near DC to near fs, where 2fs is the convertor sampling frequency. This is required by the Nyquist criterion, that the sample rate should be at least twice the highest signal frequency. The input signal and all modulation sidebands should lie inside the filter characteristic.

The analytic signal block 3 is shown in more detail in Figure 3. The data representing the instantaneous IF voltage 30 from the A/D 2 is first digitally high-pass filtered at 31 to remove any DC component of the signal. Next the filtered data is routed into a shift register 34. This register, together with a multiply-accumulate array 35, and a coefficient store 36, form a Finite Impulse Response (FIR) filter. The stored coefficients are such as to implement a Hilbert transformer chosen to have a response symmetrical about fs/2, and therefore every alternate term is identically zero. These zero terms do not need to be manipulated by the multiplier array 35 in any way. It is these unfiltered terms that are routed to the I output; the Q output is taken from the accumulator 37 of the multiply-accumulate array 35 .The output data are taken once every two input cycles, and hence a data rate reduction of 2:1 is achieved.

The reduction of the sample rate to fs/2 to represent a signal with possible frequencies from DC to fs/2 does not imply a contravention of the Nyquist criterion, as there are still the same number of data per second in the analytic signal I, Q as in the original real signal. It is due to the fact that an analytic signal does not have the frequency ambiguity of a real signal that the representable bandwidth is equal to the sampling rate. Normally the frequency range of an analytic signal would be expressed as from negative half-sample-rate to positive half-sample-rate. However, due to the periodicity inherent in a sampled data system, the frequencies from negative half-sample-rate to DC are aliased to those from positive half-sample-rate to sample-rate. This allows the choice of expressing the frequency coverage as DC to sample-rate.

Once the input real signal has been translated into an analytic signal, the amplitude and phase may be extracted from it.

As shown in Figure 4, the phase Is extracted by calculating the arctangent of the ratio of the I and Q components. In practice, the modulus of each component is taken, and then these are sorted into biggest and smallest. This has the effect of rotating the phase angle into the first octant of the phase plane. The three pieces of information generated in this process, namely the sign of I, the sign of Q, and the identity of the component which had the larger absolute value, identify in which of the 8 octants the phase angle originally lay. The rotation to the first octant simplifies the subsequent task of calculating the arctangent.

The arctangent of the range from 0 to 1 is a reasonably smoothly varying function, but one which is not representable by a low order polynomial. The angle is calculated by a linear tnterpolation of a dense lookup table, or by higher order interpolation of a sparse lookup table, as for conventional arctan routines found in software libraries. The angle is then rotated back into the correct octant.

In Figure 5, the amplitude is extracted by calculating the modulus of the I and Q components. The most accurate way to do this is to take the root sum square of the components, but the square root function is quite expensive in time and area. Simpler methods such as biggest+O.5*smallest are inexpensive, but poor in performance.

In this example, which represents a compromise, a first approximation of the amplitude, i.e, of the square root of the sum of the squares, is made in a calculation block CALC which receives as inputs the magnitudes of I and Q. In parallel, the squares of these magnitudes are summed and then divided by the first approximation; the result is then averaged with the first approximation to obtain a better approximation of the amplitude, which should be good to 16 bits.

What emerges after these processes is a pair of readings for each pair of samples from the analytic signal block 3, of the IF voltage (Figure 5) and the phase (Figure 4).

It is interesting to note that the data rate has been the same throughout. At the A/D, the signal was represented by pairs of real samples, displaced in time, at a rate of fs. The analytic signal represented the signal by pairs of in-phase (I) and quadrature (Q) components at fs rate. After the amplitude and phase the signal is still completely represented by pairs of phase and amplitude components at a rate of fs. Note also that the maximum modulation frequency allowed at this rate by the Nyquist criterion is fs/2, which fits in with the frequency space allowable for sidebands on Figure 2.

The IF voltage or amplitude determination in this way does not need any form of peak hold or averaging function, eliminating the need to compromise on the time constants inherent in an analogue envelope detection system. AM recovery can therefore have the same bandwidth as FM recovery.

As shown in Figure 6A, the instantaneous frequency of the signal is calculated as the phase difference between successive phase readings, multiplied by fs. This gives the rate of change of phase per unit time, the definition of frequency. The differencing calculation must treat the step phase discontinuity at +j- pi correctly: due to its being represented in a sampled data system, frequency is periodic with a period of fs, and calculations of frequency falling into periods other than DC to fs must be reduced back to that basic period.

For DC FM purposes, the frequency signal is subtracted from a fixed reference to give a modulation signal which is analysed as conventional FM. For AC FM, the frequency signal is subtracted from its mean value, or otherwise high pass filtered, to yield AC coupled FM.

Once the frequency readings have been turned into FM, they may then be integrated to yield phase modulation, as shown in Figure 6B. Providing bounded phase modulation may need either AC coupled phase modulation, or at least a second order high pass filter for the FM, or locking of the instrument phase to the signal source and the implementation of exact arithmetic (with no roundings or truncations).

As shown in Figure 7, the amplitude signal derived in accordance with Figure 5 is low-pass filtered to derive input amplitude, and high-pass filtered to derive AM.

The operations described above are performed preferably entirely in hardware such as connected MSI/LSI functions, semi-custom or custom chips. An important feature of the circuit is the halving of the data rate at a stage as early as immediately behind the high pass filter t31 of Figure 3], which reduces by almost a half the complexity of the chip required to implement the system.

Claims (4)

1. An electronic integrated circuit for digitally demodulating an input signal, comprising an analogue-to-digital convertor connected for conversion of the input signal to a digital signal, an analytic signal block responsive to the digital signal to generate an analytic signal, and means responsive to the analytic signal for calculating the amplitude, frequency and/or phase modulation of the input signal, the analytic signal block comprising a finite impulse response filter whose data rate is half the sampling rate of the analogue-to-digital convertor.

2. An integrated circuit according to Claim 1, in which the finite impulse response filter comprises a shift register connected to receive the ditigal signal, a multiplier array whose elements are connected to receive the contents of alternate elements of the shift register and also connected to a store of predetermined multiplication coefficients which implement a Hilbert transformer whose frequency response is symmetrical about a frequency equal to half the said data rate of the filter, an accumulator for summing the multiplied outputs of the multipliers to produce one quadrature component of the analytic signal, and means for sending the contents of the other elements of the shift register to produce the other quadrature component of the analytic signal.

3. A modulation meter comprising means for mixing a modulated RF input signal with a local oscillator signal to generate an intermediate frequency input signal, an electronic integrated circuit according to Claim 1 or 2 connected to receive as its input the intermediate frequency input signal, and means responsive to the calculated amplitude, frequency and/or phase modulation to provide an indication thereof.

4. A modulation meter substantially as described herein with reference to the accompanying drawings.