TUNING MULTIPORT AMPLIFIERS Field of the Invention The present invention relates to tuning multiport amplifiers (MPAs) and in particular to adjusting the phase and gain relationship between individual amplifiers in an MPA to optimise signal isolation between output ports of the MPA. Background Art An MPA is a well-known power amplifier device used for satellite communications, which may operate at microwave frequencies above 1 GHz, and covering L & S bands with frequencies in the range of 1.5 - 2.6 GHz, and Ku & Ka bands with frequencies in the region 12-20 GHz, etc. An MPA includes a number N of similar amplifier units (TWT or solid state) in parallel, each having a power P, so that each input signal is amplified by each amplifier, to increase the power of each output signal by a factor N, to PxN. N input ports and N output ports are provided, so that an input signal on one input port is routed to the corresponding output port. The input ports are connected to the amplifier units by a low power input network (INET) that may be implemented in any convenient transmission line technology that is appropriate to the circumstances, e.g. microstrip, stripline, coaxial cable, or waveguide, and the output ports are connected to the amplifier units by a high power output network (ONET) that is implemented typically using low loss transmission line technology. The ONET is mathematically a reciprocal of the INET, so that a signal presented to the nth input is directed to the nth output. Each network comprises an array of signal dividing waveguide devices. A Butler matrix or a network comprising just hybrid devices are normally used for signal division, because they have convenient gain and phase shift properties. One type of hybrid is a four port signal dividing device comprising two inputs and two outputs, with selective 90° phase shifts; this phase difference may be exploited to improve the isolation characteristics of the networks. However other hybrids and other signal splitting devices may be used which may have 180° phase difference. The invention is discussed in particular with regard to 8 port MPAs that may have use in Wideband Multi-beam payloads operating at Ku or Ka-band. These typically use Single Feed per Beam (SFB) designs which provide a contiguous set of up to 100 narrow beams or more over a defined area. MPAs are an attractive solution to SFBs, since they potentially offer a high degree of flexibility in allocating power to beam. MPAs have successfully been applied at L and S-band, but present highly challenging problems at Ku and Ka-band at which wavelengths are an order of magnitude shorter. We examine the impacts of mismatches on MPA performances, and identify a feature through which these impacts may be lessened, particularly with regard to isolation. MPAs have been considered for use in multi-beam satellite payloads for some time, see S. Egami and M. Kawai: "An Adaptive Multiple Beam System Concept" IEEE Journal on Selected Areas in Communications, Vol. SAC5, No.4, May 1987. They have been successfully employed at L-band, see M. Mallison, R. Gill, S. Curtis, R. Manku, "Advanced Payload for Multibeam Satellites that Support High Data Rate Broadband Global Area Network", AIAA, 23rd International Communications Satellite Systems Conference, Rome, Sept 2005, and at S-band, see M. Tanaka and K. Yamamoto, "New Technologies in N-STAR Communications Payload", AIAA, 17th International Communications Satellite Systems Conference, Yokohama, Feb. 1998, primarily for mobile services. More recently, with increasing interest in provision of broadband services, such as internet access and HDTV via multi-beam satellites, attention has focused on the provision of MPAs at Ku and Ka bands, see. R. Kuramasu, T. Araki, M..Shimada, E. Tomita, T. Satoh, T. Kuroda, M. Yajima, T. Maeda, T. Mukai, "The Wideband Internetworking Engineering Test and Demonstration Satellite (WINDS)", AIAA, 20th International Communications Satellite Systems Conference, Montreal, May 2002, and A. Anakabe, A. Mallet, F. Gizard, C. Laporte, T. Robert, C. Boulanger, J. Sombrin, L. Lapierre, P. Barretto-Da-Rocha, P. Frichot, F. Coromina, J. Collantes, "Ka-band Multi-port Amplifier Characteristics for Space Telecommunication Operation", 6th International Vacuum Electronics Conference, Noordwijk, Netherlands, April, 2005. The most efficient multi-beam payload is a Single Feed per Beam (SFB) architecture which typically uses 3 or 4 antennas to generate a regular set of contiguous beams. These beams employ a high degree of frequency re-use, for example using a 4 frequency "colour" re-use scheme. The essential disadvantage of the SFB is lack of flexibility, particularly power to beam allocation. The application of MPAs would provide this flexibility, significantly enhancing the utility of this architecture by enabling capacity (transmitted power) to follow dynamically changes in demand across the coverage area. MPAs could be used in wideband, variable bandwidth transponders providing flexible assignment of power as well as bandwidth to each beam, ensuring optimum link performance in each case. The variable bandwidth may be provided using either analogue or digital signal processing. In essence an MPA comprises an input Butler matrix, or network of just hybrids with 2N (N = 1, 2 ...etc), signal inputs, and provides access for each of these inputs equally to the same number (2N), of amplifiers. The output of the amplifiers is fed to another Butler matrix or set of hybrids which mirrors the configuration of the input network, and which recombines the amplifier outputs into the same original signal set, but amplified. The great advantage of the MPA is that in providing access for each input port equally to each amplifier, the accessible power available to each port is 2N x P, where P is the power of each individual amplifier. Thus the MPA embodies a high degree of flexibility, providing a wide range of output power which can be shared dynamically and in a highly flexible manner between the 2N inputs. The essential problem in the provision of Ku and Ka band MPAs is that of phase and amplitude tracking between the amplifiers and other units at the frequencies concerned (12GHz at Ku, 20GHz at Ka-band) and impacts of this tracking on interport isolation performance (the so-called cross talk problem). The impacts of amplitude/phase mismatches on MPA performances are examined in detail, and a design feature/setup has been found through which mismatch impacts may be lessened, particularly with regard to isolation. Formulae and signal flow diagrams illustrate how MPAs may be optimized for multi-beam payloads, providing best isolation between ports of the same beam "colour". The analysis is with reference in particular to MPAs with 8 ports using INETs and ONETs comprised exclusively of hybrids, which is the configuration commonly of most interest. However, because of the symmetrical nature of MPAs, and their scalability, the analytical basis is applicable to any order and for schemes embodying Butler matrix INETs/ONETs in general. Currently 4 and 8-port MPAs are in use. MPAs having more than 16 ports may not find favour, owing to design complexity, although the invention will be of increasing value the more complex the MPA. According to a first aspect of the invention there is provided a method of tuning a multiport amplifier, the multiport amplifier comprising an even number of power amplifiers arranged in parallel and each amplifier being paired with another amplifier, a series of input ports and a series of output ports, the input ports being connected to the amplifiers by a signal dividing network and the output ports being connected to the amplifiers by a signal combining network, whereby an input signal at any given input port is amplified by all amplifiers and then recombined into an output signal at a given output port, the method including the step of matching signal phase and gain of each amplifier to signal phase and gain of its paired amplifier to an extent which is more stringent than said matching between non paired amplifiers. In the context of usage on satellites and for other communications uses, the power amplifiers will normally be microwave power amplifiers. Conventional tuning involves matching of all amplifiers to the same degree of accuracy and is an iterative process involving the gain and phase adjustment of each amplifier until the same required isolation is achieved at all output ports for a given input port. This involves, at each iteration, measuring the phase and gain values at all output ports with each input port excited in sequence (ie a total of 2 x 8 x 8 measurements (phase and gain) at each iteration for an 8 port MPA). Moreover there may be a number of stages to the tuning, such as ensuring the required performance at ambient, hot and cold temperatures. Thus any configuration of the MPA which results in reduced isolation requirements at some ports with a corresponding reduction in amplifier tuning will provide significant savings in time and cost. It has been estimated that the method of the invention, requiring only partial matching of amplifiers and hybrids, as it does, is likely to be 20-30% more time efficient in tuning. The said step of matching each amplifier to its paired amplifier is preferably carried out by said matching of amplifiers paired adjacent one another. Each of these HPA pairs, with their associated hybrids are preferably integrated together as self-contained units and preferably with a common power supply for each unit. The signal dividing and signal combining networks may each include a series of respective signal dividing and signal combining hybrids and in which the said step of matching paired amplifiers adjacent one another also includes matching signal phase and insertion losses for respective input and output hybrids in closest connection with each said pair of amplifiers. Where each amplifier pair is an adjacent amplifier pair, that pair and the respective input and output hybrids in closest connection therewith may comprise a self contained assembly with a common power supply. The said step of matching each amplifier to its paired amplifier may be carried out by said matching of paired alternate amplifiers. Where this is the case the said step of matching the paired alternate amplifiers may also include matching signal phase and insertion losses for respective input and output hybrids in both closest connection and next closest connection with each paired amplifier. Matching each amplifier to its paired amplifier may also be carried out by matching of amplifiers paired at every 4th position. When this is done and if the signal dividing and signal combining networks each include a series of respective signal dividing and signal combining hybrids the step of matching the paired amplifiers may also include matching signal phase and insertion losses for respective input and output hybrids in the closest connection, the second closest connection and the third closest connection with each paired amplifier. The step of matching signal phase and gain of each amplifier to signal phase and gain of its paired amplifier may be carried out substantially to within 10 to 15 degrees of signal phase and 1.0 to 2 dB of gain, more stringently, substantially to within 7 to 10 degrees of signal phase and 0.7 tol.O dB of gain, more stringently still, substantially to within 5 to 7 degrees of signal phase and 0.5 to 0.7 dB of gain, If even better matching is required in certain circumstances, then matching substantially to within less than 5 degrees of signal phase and less than 0.5 dB of gain may be obtained. It will be appreciated that this will be more time consuming than obtaining the more relaxed matching of the previous three ranges but, in all cases, it is only the paired amplifiers that are matched to the closer limits and in many cases the non paired amplifiers may require no adjustment at all from their manufactured state. Thus the step of matching said non paired amplifiers may take place substantially to between 15 to 20 degrees of signal phase and 1.5 to 2.5 dB of gain. All hybrids may be matched to at least substantially 10 degrees of signal phase and 1 dB of insertion loss. Deviation from 90 degrees phase difference between hybrid output ports may be set to at least substantially 5 degrees and insertion loss tracking between hybrid input and output ports to at least substantially 0.3 dB. Matching between each amplifier in an adjacent pair to at least 10 degrees of signal phase and 1 .OdB will provide a minimum of 24dB isolation for an SFB architecture operating with four frequency colour re-use. If to at least 7 degrees of signal phase and 0.7dB, 26dB isolation will be provided for the same architecture with the 2dB margin to guard against ageing and thermal variations. Matching to an extent closer than this, for example to at least 5 degrees of signal phase and 0.5dB, may be preferable, for example, in an SFB architecture operating with four colour re-use and with carriers differing significantly in power. But it will of course be more difficult, and therefore more expensive, to achieve. Good matching is also required between non-adjacent HPAs to ensure acceptable isolation between ports of different colour and acceptable combining efficiency, but not to the same extent as for isolation between ports of the same colour. Worked examples providing isolation estimates and combining efficiency are presented. These are confirmed through MPA simulations. Matching of paired alternate amplifiers may be applied (for MPAs with >4 ports), of every 4th amplifier (for MPAs with >8 ports), of every 8th amplifier (for MPAs with >16 ports), etc. Again, this matching will be to an extent which is more stringent than matching between the non paired amplifiers. According to a second aspect of the invention there is provided a multiport amplifier comprising an even number of power amplifiers arranged in parallel and each amplifier being paired with another amplifier, a series of input ports and a series of output ports, the input ports being connected to the amplifiers by a signal dividing network and the output ports being connected to the amplifiers by a signal combining network, whereby an input signal at any given input port is amplified by all amplifiers and then recombined into an output signal at a given output port, the muttiport amplifier being tuned according to the method of the first aspect of the invention. The signal dividing and combining networks may each comprise an 8x8 Butler matrix. The signal dividing network may alternatively include a series of signal dividing hybrids with the signal combining network including a series of signal combining hybrids. The multiport amplifier may comprise 8 ports and the signal dividing and combining networks may each comprise three columns of hybrids equivalent to a Butler matrix without the inter-hybrid phase shifters. The first four input ports may each be connected to a different non-overlapping frequency band with the second four input ports being each connected to one of the same four frequency bands. In an alternative embodiment the first four input ports are alternately connected to two different non-overlapping frequency bands and the second four input ports are each connected to two different non-overlapping frequency bands with the two frequency bands for the first set of four input ports being different from those of the second set. In a further alternative embodiment pairs of adjacent input ports are connected to bands of the same frequency, with each pair being connected to one of four different non-overlapping bands of frequency. Each output port may be connected to a respective antenna feed of a single feed per beam, multi-beam antenna whereby to form a set of beams with frequencies in accordance with a defined frequency re-use pattern. The isolation analysis presented herein enables output port selection for optimum isolation in multi-beam frequency re-use schemes and the 1M analysis assists in the allocation of frequencies to input ports for minimum IM interference. Brief Description of the Drawings The invention will now be described by example with reference to the following drawings of which:- Figure 1 is a diagrammatic representation of an 8-port MPA using INETs and ONETs constructed exclusively from hybrids rather than a Butler matrix, which embodies both hybrids and phase shifters; Figure 2 is a diagrammatic representation of a hybrid as used in the present invention; Figure 3 is a diagram of signal flow in an 8-port MPA, showing null signal points for matched adjacent pairs of amplifiers; Figure 4 is a diagram of alternative signal flow in an MPA, showing alternative null signal points for matching of alternate pairs of amplifiers; Figure 5 is a diagram of alternative signal flow in an MPA, showing alternative null signal points for matching of every fourth amplifier; Figure 6 is a diagram of signal flow in an MPA, for matching of all amplifiers, showing constructive addition of signals; Figure 7 is a table summarizing the MPA equipments critical in the definition of isolation at each output, and Figure 8 is a table showing output ports for 3 carrier, 3rd order tntermodulation products for an 8-port MPA. Description of the Preferred Embodiment A diagram of an 8 port MPA is shown Fig.1. It comprises a set of 3 columns of 4 input hybrids, a single column of 8 amplifiers and a set of 3 columns of 4 output hybrids. A definition of a hybrid element (input or output hybrid), is given in Fig.2. The input and output signals, p1 & p2 and q1 & q2 are assumed to be complex. Transfer function of the hybrid is represented thus: (Equation Removed) (1) where Cn, C12, C21, C22 are all invariable complex coefficients which can be represented as follows: (Equation Removed) (2) Ideally for a perfect hybrid all r = 1/V2, and all θ = 0, and the transfer matrix becomes: (Equation Removed) (3) For a practical hybrids, r and 0 will deviate from the ideal, and if the deviations are small (V2/r < 0.2dB, 6 < 10°) the following approximations can be made: (4) (Equation Removed) (5) where a is a loss factor common to all input or output hybrids (between 0.2dB to 0.7dB, depending, for example on whether the hybrid is in waveguide or coax), and σ & θ are respectively amplitude and phase deviations from the nominal (nom. = 1.0 & 0° respectively for amplitude and phase) For the analysis, the amplifiers are assumed to be perfectly linear with complex gain G, represented by function, (Equation Removed) (6) As discussed above, and ideally in a MPA transponder, all amplifiers will be matched and have the same value of G and 0. However for practical amplifiers there will be a deviation A and 5 with respect to some mean gain and phase shift value G0 and 0O respectively. For small deviations (A < 0.5dB, 8 < 10°) the gain function can be written as: (Equation Removed) (7) The total transfer matrix of the MPA is given by: (Equation Removed) (8) where t, x & y are the transfer matrices of the first, second and third input hybrid column (with a combined overall transfer function Ω), and t', x' & y' those of the corresponding output hybrid columns (combined overall transfer function T) as defined in Fig.1. The matrix representations for each column of input hybrids are presented below. The letter subscripts against the hybrid transfer coefficients correspond to the hybrid letter designations shown in Fig.1. (Figure Removed) Those for the output hybrid set are presented below: (Figure Removed) With ideal hybrids, ie all C11, C22 = 1/√2 and all C12) C21 = j/√2, and assuming unity gain, zero phase shift amplifiers, the overall transfer^^ matrix is then as given by Eq. (11): When practical values of C11, C22. C12, C21 and A are assumed, the full matrix representation of O becomes mathematically unwieldy. To illustrate specific properties of the MPA transfer function, a selection of elements from this matrix are presented below. These all relate to l/P port 1, but the general characteristics displayed by the selection of equations would be equally applicable to any other l/P port. (Equation Removed) Φ11 (Eqn. 12) represents the signal level at l/P port 1, appearing at O/P port 1, Φ51 (Eqn. 14), at O/P port 4, etc. For an ideal system, Φ11 through to Φ71, should = 0 (perfect isolation), with Φ81 = -j. However, practical components result in finite isolations between ports. Inspection of the above formulae demonstrates the following: ΦOn, Φ41 (also applies to Φ31 & Φ41, not shown): the isolations at the corresponding O/P ports are determined, according to the invention, to a first order of approximation, by matching between pairs of adjacent amplifiers (A1/A2, A3/A4 ...etc), and the quality (insertion loss differential and deviation from 90° between O/P ports) of the inner most hybrids J, K, L & M and a, b, c & d. This property is shown through the underlining of the relevant factors in Eqs. 12 and 13 and the signal flow diagram in Fig.3. The flow diagram shows that with matching between adjacent amplifiers and their associated hybrids, self cancellation of the signal occurs at the O/Ps of the inner most output hybrids (at points S). This effect can be explained with reference for example to the amplifier pair A5/A6 with the signal input at port p1. In the case of amplifier A6, the relative phase shift from the input of the closest input hybrid (L) to the output of the closest output hybrid (c) will be 180° more than the phase shift between the same ports via A5. This phase difference is introduced by the two 90° shifts produced by the hybrids either side of the amplifier pair, and results in the cancellation as indicated Thus the quality of the middle and outer hybrids has insignificant impact on isolation in these cases. In fact for these cases, for best match and hence isolation, the amplifier pairs and associated hybrids (eg A5/A6 with hybrids L & c), may be constructed as integrated assemblies Φ51 (also applies to Φ61, not shown). For these cases, isolation is determined in essence by matching between pairs of alternate amplifiers (A1/A3, A2/A4...etc) and quality and matching of the inner most and middle hybrid sets. The corresponding signal flow diagram is shown in Fig.4. This demonstrates that signal constructive addition occurs at the O/Ps (at points S') of the first hybrid row, with cancellation occurring at the O/Ps of the middle row (at S). Again cancellation can be explained in terms of the introduction of a 2 x 90° phase shift differential by hybrids either side of an amplifier pair, in this case by hybrids in the middle column. For example in the case of the pair A2/A4 and with the signal applied at input port p1, the relevant hybrids are "E" and "f Φ71 In this case, isolation is determined in essence by matching between every 4th amplifier (A1/A5, A2/A6...etc) and quality and matching of the inner most, middle and outer most hybrid sets, ie all components in the MPA. Signal cancellation for this case occurs at the O/Ps (at point S) of the outer most hybrids as shown in Fig.5. In this case the 2 x 90° phase shift is introduced by the outermost hybrid columns. For example in the case of the pair A2/A6 and with the signal applied at input port p1, the relevant hybrids are "A" and "m" Φ81 This is associated with the wanted output, and represents the combining loss. It is dependent on matching between all units, but not to the same degree of criticality as for isolation. Constructive addition in this case occurs throughout the output network (at points S') as shown in Fig.6. A summary of the MPA equipments critical in the definition of isolation at each O/P is given in Figure 7. This identifies for each O/P port those equipments which should be matched in order that good isolation is achieved from the signal at the corresponding l/P port. The following key applies to this table - X: matched adjacent amplifiers (A1/A2, A3/A4 ) and matched associated inner most hybrids - Y: matched alternate amplifiers (A1/A3, A2/A4....) and matched inner most and middle hybrid sets - Z: matched 4th amplifiers (A1/A5, A2/A6...), matched inner most, middle and outer most hybrid sets The entries "OP" in Figure 7, designate the wanted O/P port corresponding to a given l/P port. Isolation is defined here as the ratio of wanted to unwanted signal insertion losses (which includes any combining losses) at a given output port and is defined in terms of the overall MPA transfer matrix Φ as: 17) (Equation Removed)Where m and n are the input ports of the interfering and wanted signals respectively. (N-n+1) is the output port of the wanted signal, N being the total number of output ports or MPA order, in this case 8, and takes into account the transposition between l/P and O/P port numbering. An exact expression for isolation can be obtained in terms of individual hybrid and amplifier performance parameters by taking the ratio of the modulus squared of the relevant matrix elements Φ