GB2219140A - Adaptive antenna array - Google Patents

Abstract

The circuitry for an adaptive antenna array uses a decorrelation network in cascaded stages which follows the principles of our Appln. No. 8316658, GB 2141588, with each node of the network summating two signals. One of these signals is from an antenna array for nodes of the first stage, or from a node of the previous stage depending on which stage of the network is involved, while the other signal is common to that stage (e.g. from PD1). The first such common input (PD1) comes from one antenna array and at each node one input is inverted with respect to the other to effect decorrelation. At each node one input is weighted (W11, W12, W13, etc) and the weights used are continuously updated using a processor which operates in time shared manner, and which samples the output of each node of each stage (Y1 Y2 Y3 Y4 Z1 Z2 Z3). <IMAGE>

Description

ADAPTIVE ANTENNA ARRAY The present invention relates to adaptive antenna arrays of the steerable type.

The implementation of an adaptive null steering array using a single rank variable combining network as shown in Fig. 1 is well known. One of the weights, the upper-most one as shown, is 'clamped' at a fixed value whilst the others can be varied. The variable weights may be controlled either by a closed loop correlation arrangement, or by a mathematical processor solving directly for the required weight values, from sample "data snapshots" obtained at the antenna elements. Each weight is complicated since both the amplitude and the phase of signals can be altered.

Comparisons between these methods show that whilst the former is relatively simple and robust, its speed of response to a changing signal distribution can be sluggish in many cases. In particular, the gain, and hence speed, of the feedback loops (which vary with signal level) must be limited so that strong jamming signals do not cause feedback loop instability or undue weight jitter. Hence the loop speeds associated with lower power jamming sources which are simultaneously present are invevitably lowered. In contrast, the direct solution has no such intrinsic limitations. But in practice there is a high implementation complexity associated with it and a requirement for very accurate control of the weights.

An object of the present invention is to provide an improved adaptive antenna array in which the disadvantages of the known arrangements are minimized.

According to the invention, there is provided a multiple element, power inversion adaptive antenna array, including antenna elements connected by decorrelation nodes arranged in groups each group including one less decorrelation stage than the preceding group and the first group including one less node than the number of antenna elements, each node of the first group being connected to receive as a main signal a signal from a different one of the antenna elements and as an auxiliary signal the signal derived from the remaining elements and to provide an output signal and each decorrelation node of each subsequent group being connected to receive as a main signal an output signal from a respective decorrelation node of tbe -preceding group and as an auxiliary signal the remaining output signal from the preceding group and to provide an output signal, whereby each group except the last has one preset decorrelation node whose output signal is decorrelated with respect to one or more of all except one of the antenna element input signals, wherein the array includes scanning means whereby the outputs of all of the nodes of each group are connected in time-shared manner to a processor to which is also applied a signal due to said different one of the antenna elements, which processor successively generates updated weights for application to said nodes, each such weight being applied to one input of its said node for said decorrelation.

Thus the invention relates to a particular way of implementing a processing structure known as sequential decorrelation which enables faster convergence to be achieved in multiple jamming scenarious than the single rank correlation approach, but which can be less complex to realise than the mathematical direct solution approach.

Embodiments of the invention will now be described with reference to the accompanying drawings in which Figs. 2 and 3 illustrate the basic principles of sequential decorrelation.

Figs. 4 and 5 shows two approaches to implementation of a cell of the array of Fig. 3.

Figs. 6, 7 and 8 are schematic diagrams of a preferred embodiment of the invention.

The principles of sequential decorrelation which is itself known, see our Appln. No. 8316658 (J. & Searle et al 3-3), is to adaptively combine two signals at a time in the nodal structure illustrated by Fig. 2. Each node, which has two inputs A and B connected as shown to the antennas Al to A4 and one output C, can be considered as a two-channel adaptive combiner with the bottom channel signal being weighted by an amount equal to the cross-correlation between the two input signals, dividied by the power of the signal in the bottom channel. The weight value can be derived by a mathematical approach, Fig. 4, in which the computation is effected in a computation unit CU which, as can be seen, effects the calculation indicated above, with inversion as we are decorrelating.Alternatively the weight value can be derived by a correlation loop, Fig. 5, which, it will be seen, gives the same result. In the former case, both the weight computation and the application must be precise, and a digital implementation is mandatory. In the latter case, the characteristics of robustness and simplicity of the correlation loop apply.

However, the intrinsic limitation of the correlation loop approach in its tendency to respond sluggishly in a multiple jammer environment is not very significant here, since each two-element combiner can only cope with the strongest jamming signal it sees.

There are no spare degrees of freedom even to try to cancel secondary jamming signals. There is thus very little elongation of convergence time in each node as the number of jamming sources increases. Instead, multiple jammers are dealt with by successive columns of nodes in the triangular form illustrated by Fig. 2.

In very simple terms, therefore, each rank of the cascade takes approximately as long to converge as does a single rank adaptive combiner countering a single jammer.

A preferred embodiment of the invention is shown in Fig. 6, and it uses a nodal structure in which each node comprises an unweighted input channel, a complex-weighted input channel and a combining network.

In each case the unweighted channel is the upper one of the two of that node. The inputs at the left-hand side come from the antenna elements. The details of the network and an outline of the adaptive processor are illustrated..in Figs. 7 and 8 respectively. The value of weighting to be applied is provided by a time-shared adaptive processor, and the multiplexed correlator is best implemented with the provision of a wide dynamic range power-normalised response so that the convergence rate against the strongest visible signal at any stage is independent of its absolute level, above a controllable threshold. The latter provision of a threshold prevents needless adaptation and subsequent weight jitter in the later ranks of the network when the number of jammers is less than the number of ranks in the processing array.

In order to equalise the gain balance through the network, amplifiers and also power dividers are placed in the positions shown for a four-element array in Fig. 6. Any frequency of operation may be considered in principle; in practice, implementation at frequencies in the range 100 to 400 Hz is particularly convenient.

Various sequences for updating the weights are possible, and two are described below. In each case, a single weight update can be said to adjust a complex weight component by an amount which reduces the strongest "visible" signal in that particular two-channel combiner by between 1 and 2 dB.

(i) Each weight component through the whole network is given a single update in each multiplexing cycle, for example, a sequence of the form W111 W12, W13, W21, W22, W31, in a four-channel system.

(ii) A subcycle to adapt the first rank of weights, e.g. W11, W121 W13, is followed by a subcycle to adapt the second rank of weights, and so on. Each subcycle includes several updates of individual weight components for the particular rank so that approaching complete adaptation is achieved within the subcycle. In this way, the interaction of adaptation between the various processing ranks may be minimised and hence the convergence time minimised.

It may be the case that different multiplexing sequences will be optimum for different types of interference situations, e.g. continuous jammings, of pulsed jamming.

We now consider Fig. 6 in somewhat more detail than above. In this arrangement, the signal from one antenna, the bottom-most one is decorrelated with all of the other signals in the first stage. Thus this signal is amplified by an amplifier Al and the amplified signal is divided by four by a power divider PD1 three outputs of which are applied to the decorrelation nodes in the first layer. The fourth output Y4 of the divider is the final output of that antenna element to the processing arrangements used for weight updating.

To consider one node as an example, the signal from the upper-most antenna element is applied to one input of a summator S1, to the other input of which is applied one of the outputs from the power divider PD1.

This input is applied via a weighting circuit W11, which has two weighting inputs, designated WI and WO. These are derived from the processing arrangements just mentioned. The output of S2 goes to another power divider PD2, which splits the signal between an input to a next-stage node and an output Z2 to the weight-updating processing arrangements.

The amplifiers referred to above which are included for balancing purposes are indicated at A2 and A3.

The outputs to the updating arrangements from the various nodes are indicated at Y1, Y2, Y3, Y4, Z1, Z2 and Z3, and these go to two switches SW1, SW2, via which they are applied to the adaptive processor, which performs appropriate computations on them to produce the required weight updates.

We now consider Fig. 7, which shows details of a complex weight network, the one chosen being W11 (Fig.

6); the others are similar but using different weights.

Here the input, is applied to a power divider PD5 with 0 0 two outputs.at 0 and 90 phase. From this they go to two weighting circuits WA1 and WA2, the outputs of which are summated by the summator SA. As a result of the operation of this weighting network we have Output = ( 1)2 (W1 cos wt + WQ Sin wt) 2 = a cos (wt-0) 2 where WI = a cos 0 = = a sin 0 The input comes either from an antenna element in the case of the first stage of the array or from a node in the case of a later stage.

Fig. 8 is an example of a time-shared adaptive processor, such as used in the circuit of Fig. 6 for up-dating the weights used at the various stages of the array. In this arrangement the inputs to the processor from the two switches are similar, so only that from SW1 will be described.

The input from SW1 is split by a power divider PD5 into two channels each including a correlator, these ccorrelators having signals representing cos wt and sin wt applied to them. Each of these channels are similar so again only one will be described. In the one for cos wt, the output of the correlator CA goes via a low-pass filter F1 and an amplifier to an analogue-to-digital converter ADC. Thus the processor, PR, which is a digital correlator working under the control of an algorithm for power normalization and weight up-date, receives digital information on four inputs, in each case in time-divided manner. The correlator also provides a digitally-controlled automatic gain control for the amplifiers via the connection designated AGC.

The updated weights provided by the correlator are passed to the weight store from which they are distributed as is appropriate to the various decorrelation nodes. The above description has been couched in terms of an array of antenna elements which are directly connected to the adaptive processor. In this light, any configuration of array elements can be considered; line, circular, random or conformal arrays. There are, in addition, further configurations which can make use of the proposed processing structure, (1) a circular array connected via a Butler matrix to the processor: (2) a radar antenna with auxiliary elements or subarrays which are adaptively combined together with the aid of the processor; (3) the adaptive combination using the suggested processor of the outputs of a multiple tapped delay line, forming in effect an adaptive filter; (4) an adaptive arrangement of any of the above, with multiple outputs provided by the minor change to the network configuration. With this arrangment, it may be possible to obtain performance improvements by selecting which output to use.

Claims (3)

CLAIMS:

1. A multiple element, power inversion adaptive antenna array, including antenna elements connected by decorrelation nodes arranged in groups each group including one less decorrelation stage than the preceding group and the first group including one less node than the number of antenna elements, each node of the first group being connected to receive as a main signal a signal from a different one of the antenna elements and as an auxiliary signal the signal derived from the remaining elements and to provide an output signal and each decorrelation node of each subsequent group being connected to receive as a main signal an output signal from a respective decorrelation node of the preceding group and as an auxiliary signal the remaining output signal from the preceding group and to provide an output signal, whereby each group except the last has one preset decorrelation node whose output signal is decorrelated with respect to one or more of all except one of the antenna element input signals, wherein the array includes scanning means whereby the outputs of all of the nodes of each group are connected in time-shared manner to a processor to which is also applied a signal due to said different one of the antenna elements, which processor successively generates updated weights for application to said nodes, each such weight being applied to one input of its said node for said decorrelation.

2. An array as claimed in claim 1, wherein each said node has a first input via which one signal is applied to a summator, a second input via which another signal is applied to the summator via a weighting circuit, one of said inputs being inverted, and an output from the summator on which appears that node's decorrelated output.

3. A multiple element, power inversion, adaptive antenna array, substantially as described with reference to Figs. 6, 7 and 8 of the accompanying drawings.