ADAPTIVE AERIALS FOR UHF REBROADCAST
From an original article by MD Windram, L Brunt and EJ Wilson
PART TWO - DEVELOPMENT
After a period of experimentation, including the installation of a prototype half-size system at Alderney to confirm earlier theory, the operational adaptive aerial was installed on Alderney in March 1977. The block diagram of this system is shown in Fig. 4.
|Fig. 4: Following a period of experimentation, including the installation of a prototype half-size system on Alderney to confirm the theoretical studies, a multi-channel linear array of 16x4 dipoles with adaptive control was designed and implemented to become the first operational adaptive system for RBR applications. Each element is connected through a network which controls the amplitude and phase of the output, providing a voltage-controlled aerial pattern. Up to four receivers may be used within the control loop, and the aerial pattern is continuously adjusted to optimize signal/noise ratios of the desired signals.
The adaptive array is based on a linear array of sixteen elements, the output of each element being connected to a network which effectively controls the amplitude and phase of that output. These signals are then combined, and it is this process which creates a voltage-controlled aerial, the pattern of which is a function of the control voltages. The receiver or receivers used provide the video and audio outputs required and also provide signals for the measuring system. Up to four receivers may be used within the control loop. The adaptive control unit provides the logic which causes the aerial voltages to change in the manner required to reduce the interference. To develop a successful adaptive array major difficulties had to be solved. These included the design of the control networks, the measuring system in conjunction with the receiver and the type of algorithm to be used.
Conclusions from Theory
Initial analysis showed that the ideal element spacing is ~2/3, which combines reasonable directivity of the array with ease of control of the nulls. Cartesian (X+jY)-type control of the aerial outputs was chosen because it avoids the serious problem of discontinuities in the controls that occurs with pure phase shifters having only a finite phase range. This is illustrated in Fig. 5.
|Fig. 5: Polar and Cartesian control of amplitude and phase. Cartesian (X +jY) type control was chosen in preference to polar control since it avoids the problem of discontinuities in the control that would occur with pure phase shifters having only a finite range of phase control.
With polar control, a change of weight from A to B involves a discontinuity, since phase has a limited range. For Cartesian control there is no discontinuity so that any small incremental step is possible, in any direction, within the control domain. This is essential in an adaptive process in which small changes of controls may be required and for which discontinuities could lead to instability. This also makes the use of the constant gain criteria easier.
For arrays of the size used at Alderney at UHF it is necessary because of expense, stability and maintainability to use control algorithms which take measurements by making step changes in the control voltages to the array, thus stepping the aerial pattern and measuring the result in terms of CCI on the output signal. Other methods such as correlation techniques are far too expensive in terms of hardware and complexity, although in principle they are capable of giving more accurate measurements of the error; they were therefore not considered in detail.
The various algorithms discussed in the Appendix were fully considered. Of these, for the Alderney application, the simple hill-climb algorithm seemed to be most suitable for operation close to the minimum interference condition. This is the normal condition for the array, since much of the operation of the adaptive array for broadcasting purposes is involved in maintaining nulls on CCI sources rather than steering them onto sources. The simple algorithm has the advantage of being easy to implement and requires only a knowledge of the sign of a step, not the magnitude which can be very inaccurate in the presence of noise. It also does not require a large number of measurements, each taking a set time, as some of the more complex algorithms require.
Figure 6 shows the results of a computer simulation for five sources of interference at angles of -7° (Kippure), +65.4° (Crystal Palace), +23.6° (The Wrekin), +35,4° (Bilsdale) and +85.2° (Lelystad). The patterns are the horizontal voltage radiation patterns of the array, starting at the top with the initial (preset) radiation pattern which is that of a uniform array, and with plots at intervals corresponding to approximately 5 secs in real time until an optimised pattern is reached. The final null depths are 50 dB for each of the four sources of CCI.
|Fig. 6: Computer simulation using a hill-climb algorithm for five potential sources of co-channel interference. The relatively simple hill-climb algorithm is well suited for operation of an array close to minimum interference condition since for much of the time the prime requirement of the array is to maintain nulls on known potential sources of CCI rather than steering the nulls on to less likely sources of interference.