Transcript
THE DESIGN AND CONSTRUCTION OF A 6kW SOLID STATE PULSED AMPLIFIER FOR 1.2 – 1.4 GHz APPLICATIONS
The large thermal capacity is one of the reasons for the dominance of Silicon Bipolar devices over Gallium Arsenide in pulsed amplifiers up to 3 GHz.
D. M. FitzPatrick, Milmega Ltd.
For pulsed applications transistors are generally biased in class C. In this mode the transistor is biased off in the DC state and relies upon the RF pulse to ‘switch’ the transistor on. This produces a non sinusoidal operation as the transistor is only conducting for less than half a cycle, however the tuned nature of the output circuit and the stored charge of the bias network provide a more symmetrical RF output.
The generation of high power microwave pulses is fundamental to the operation of various systems ranging from radar to particle accelerators, from communications systems to cancer treatments. In the past such systems have depended on tube amplifiers or cavity sources. With recent advances in solid state transistor technology ever increasing power levels are achievable with the need for turning to travelling wave tube technology and its 1 inherent disadvantages . This paper reviews the design of a 6kW amplifier operating in the L band radar region.
Design Requirement The requirement for this amplifier design was as a result of a previous design by Milmega. This had produced 4, 1kW outputs covering the range 1.25 to 1.35 GHz. Investigation of the market showed an opportunity to produce a range of amplifiers, based upon the modular construction technique that had been so successful with Milmega’s CW amplifier series. The specific application required the characteristics highlighted in table 1.
Pulsed versus CW Amplifiers In terms of the design of solid state amplifiers the distinction between pulsed and Continuous Wave (CW) operation is not as obvious as would first appear. Transistors are available capable of handling hundreds of watts of RF power, but only for relatively short pulses and low duty cycles. Once the pulse width exceeds the order of 200µs, or the duty 20% the selection of the transistor becomes limited to various suppliers of CW devices. These two limits are exclusive, i.e. a system operating with 200µs pulse widths will typically have a duty cycle of <3%, and a system with a duty cycle of 20% will be operating with pulse width of the order of 10µs.
Table 1. Basic requirements Parameter Limit Frequency Range 1.2 – 1.4 GHz Peak Output Power 6 kW (67.8 dBm) min. Pulse Width 13 -100µs Duty Cycle 7% max. Pulse Repetition Rate 350 –775 pulses/sec. Drive level -10 dBm min. Pulse Droop 10% max. at 100µs Harmonics -25 dBc min. –40 dBc target Prime Supply 3 phase 220 phase to neutral, 400Hz 5A max. Temperature Range -20 to +70°C Interface RS422
An amplifier operating with RF pulses, above the limits given above, will still have different design criteria to a CW amplifier, (mainly regarding the power supply and thermal management), however for the transistor selection the power levels available are reduced by as much as a factor of 6. The power limitation in the devices is currently as a result of the thermal properties of the devices. Modern high power transistors consist of a number of cells combined in parallel. For correct thermal operation the RF (radio frequency) current must be shared equally between the cells, otherwise just like the weakest link in a chain, the first cell to fail will destroy the whole transistor. In each cell the actual transistor junctions are located on or near the surface, and the immediate surroundings have a relatively large thermal capacity, hence the ability to operate at high power levels for short periods of time, (the junction temperature does not rise instantly with peak RF current).
A survey of the available power transistors revealed that 300W was the highest peak power available. This is a two-fold increase (from 175W) over the technology of 5 years ago, and reflects the maturity of this transistor technology, (compare with internally matched GaAs transistors which have gone from 30 – 140W CW in the same period). One of the most successful configurations for class C transistor design is the balanced configuration. In this approach two transistors are combined in parallel using 90° structures. This yields good 2 impedance looking into the balanced pair, irrespective of their individual impedance provided
Ryde Business Park, Nicholson Road, Ryde, Isle of Wight, PO33 1BQ Tel: +44 (0)1983 618 004 Fax +44 (0)1983 811 521
[email protected] www.milmega.co.uk
the devices are reasonably matched. This is important as the input impedance of the transistors changes dramatically as it moves from the off to the on state. Failure to address this change would mean that preceding stages would be required to operate into a quickly changing load impedance, a common cause for oscillation.
cycles will destroy them. In order to protect the large investment in the output stages, (not only device cost, but also assembly and tuning time), the input RF signal is conditioned such that it cannot exceed the design limits. In fact three such protection circuits exist within the whole system and these will be discussed later.
Given a 500W balanced building block, the most practical output configuration would contain 4 balanced stages, giving, after combining losses, 1800W of peak output power. Again, allowing for combining losses, 4x 1800W will provide 6kW of peak output power. The gain of the transistors is greater than 7 dB, which conveniently produces a drive requirement for this output block of about 400W. By selecting the output power of the driver amplifier as being 500W, the balanced driver stage can be reused.
In order to produce an optimally flat output pulse across the frequency band the gain of the amplifier must be varied across the band. The system has over 80dB of gain, and the drive power variation without this compensation would be about 12 dB.
The design process is iterative, as each choice in component selection and configuration has implications on the rest of the system. Also re-use of existing components/stages has great benefits in terms of development time, tuning time, and stock management.
The input module contains a power detector, followed by two solid state GaAs switches with more than 60 dB of isolation. These are used to mute the amplifier when required, or to modulate the incoming RF signal. This modulation can either be intentional, or as a protection from the excess pulse and/or duty. As a safety feature this modulation is ‘hard-wired’ into the design to 130µs and a PRR of 900Hz, the detector is also fed to the Control Unit, for software pulse conditioning. The input module contains sufficient gain to increase the minimum input signal level to +10 dBm, the required drive level for the next module.
From this basic concept, the detailed system line up was constructed and analysed using an Excel spreadsheet, figure 1. This was used to derive the power requirements for the main power supply as well as the distribution of gain throughout the system. From this analysis the performance requirements of the various RF modules were determined. These are now discussed.
Driver (500W) Module Although the amplifier operates in class C the gain stages in the Input Module and the first two stages in the Driver module are class A, because they maintain a constant impedance, irrespective of the RF pulse characteristics.
Input Module This module had two main functions, (a) to protect the amplifier from RF signals outside of the operating envelope, and (b) to shape the RF pulse to match the gain shape of the power amplifiers for optimum pulse flatness. The amplifier output devices are designed for pulsed operation, and excess pulse widths or duty
The middle stage of the module is a 30W LDMOS device operating in class A/B. This gives the power consumption advantages of pulsed operation, whilst having superior gain performance. An issue with all MOSFET devices is that they exhibit ‘pulse drooping’ whereby the amplitude of a pulse trails off with time. This is due to the gFS decreasing with temperature. This stage is followed by a power
Figure 1, System Budget Spreadsheet Ryde Business Park, Nicholson Road, Ryde, Isle of Wight, PO33 1BQ Tel: +44 (0)1983 618 004 Fax +44 (0)1983 811 521
[email protected] www.milmega.co.uk
detector circuit, which automatically shuts down the module in the event of excess pulse width or duty. This is not only a failsafe for the input module (all the stages prior to this being able to operate CW) but also a protection against oscillation in one of the class A devices.
parasitic inductance and hence the gain reduction. The inductance between the capacitor bank and the device collectors must be kept to a minimum in order to prevent increased pulse droop and to maximise pulse rise time. Output (1800W) Module
The remaining stages in the driver module are class C. An isolator is included before the first device, a single ended 150W transistor, so that the preceding stage sees a non-varying impedance with power. The isolator also has the advantage of increasing the reverse isolation of the module improving stability. A 5W load is sufficient at this stage as the RF signal has a maximum duty cycle of 10% at this stage.
As discussed earlier there are four of these modules in the system, and each contain ten 300W devices. These are operated in the same way as in the output stage of the 500W module. The output of this module contains two circulators, see figure 3. These fulfil two main functions, (i) protection from CW signals injected into the amplifier, (ii) isolation of the output of the amplifier. The system output is connected directly to an antenna. There is obvious danger that other systems operating nearby may be able to inject a RF signal into the system. The antenna intrinsically provides an amount of frequency selectivity, so the biggest danger comes from other L band radars. 2 Although the r power rule means that the power reduces dramatically with physical separation, calculations have shown that it is still possible for 400W of CW RF in band power to be seen at the amplifier port of the antenna. Thus each module would see a quarter of this power, less combiner loss, and is designed to be able to withstand 100W of CW signal injected into the module output.
Figure 2, 500W Module
The power supply requirements for the class C transistors are relatively simple in theory, a collector voltage of the order of +40 volts and a peak current capability of about 17 amps. If the main power supply was rated to provide this current it would be far more capacity than needed, (and hence size and cost). The average current is less than 2 amps so this determines the rating of the supply. The peak currrent is provided by the capacitor bank, which can be clearly seen in the centre of the module in figure 2.
Figure 3, 1800W Module
The transistors are of a common base configuration, so the input (emitter) is connected to ground via a quater-wave stub. The RF can develop a potential across this which biases the transistor on. To improve the RF current sharing between the cells of the device a technique called emitter ballasting is used. In the emitter of each cell within the device a small amount of resistance is incorporated, thus if one cell starts to take more current there is an increase in voltage drop in this resistor which acts against the current increase. This has a positive effect on reliability, whilst incorporation within the device itself minimises
The Control Unit Key to the versatility of the design is the control and interfacing system. A micro-controller was used so that the system timing and levels could be quickly and easily altered. This was partially done, as it was known from the outset that some characteristics would not be fully defined until the first systems were built. A ‘soft’ front panel was designed to run on a remote PC, see figure 4. Besides mirroring all of the instrument front panel controls, the system allows the user to disable certain trips if required,
Ryde Business Park, Nicholson Road, Ryde, Isle of Wight, PO33 1BQ Tel: +44 (0)1983 618 004 Fax +44 (0)1983 811 521
[email protected] www.milmega.co.uk
incorporated to obviate the need to replace fuses. The system was designed and the first unit built within 6 months of launching the project. To do this it was necessary to include representatives from production, test and engineering in peer reviews as each component was being developed. The use of 3-D CAD greatly eased the visualisation of the system as it evolved.
Figure 4, Soft Front Panel
(those for which this does not apply are ‘greyed’ 2 out). I C interfaces were incorporated in each module, along with temperature monitoring devices. Another feature provided by the use of a micro-controller is an operational clock, which records the hours of operation. The software in the micro can be upgraded remotely through a password protected remote boot mode. Mechanical Construction The unit, although not being to full military specifications, had to be robust and operate for short periods in ambient temperatures of +70°C, (prolonged operation at this temperature will greatly reduce the life expectancy of the amplifier due to the increase in transistor average junction temperature). The main power supply was convection cooled via a heatsink mounted externally on one side of the unit, this also allowed for easy replacement with different variants of supply - for example single phase. However the heat generated in the RF modules, which even when running at 7% duty cycle amounts to approximately 1.5kW of heat, requires fan cooling. Fans are mechanical devices and will with time wear out and require replacing, hence the fan was incorporated externally so that it could be removed and replaced without having to open the amplifier. For similar reasons a 3 phase circuit breaker was
The RF amplifiers were all mounted on a central ‘tunnel’ through which cooling air was forced. This tunnel was sealed other than at the entry and exhaust ports, thus preventing the ingress of dirt and dust in to the instrument. Access hatches, provided on all sides of the unit were recessed for similar screening. To minimise RF losses through the output 4-way combiner it was mounted across a diagonal of the tunnel. Adjusting the lengths of the input cables to the 1800W modules phase matched the 4 RF paths. A cut away view from the 3-D CAD package shows the mounting of the amplifier modules around the tunnel, figure 5. Conclusion The amplifier produced peak output powers (100µs pulse width, 7% duty) in excess of 6kW, whilst exhibiting pulse droops of less than 4%. Harmonics levels were greater than –45 dBc. The system continued to operate in ambient air up to the required 70°C, with a decrease in peak output power of less than 0.5 dB. Further investigation is required to improve the pulse rise times at the band edges (~600ns compared with 200ns at the centre frequency). These could be improved by driving the pulse hard, but then pulse droop suffered. Pulse fall
Figure 6, Milmega 6kW Pulse Amplifier
times were consistently less than 20ns.
1
Figure 5, 3-D CAD Design
D. FitzPatrick, “Design of a high power solid state amplifiers to replace TWTAs in airborne applications”, IEE Seminar on RF and Microwave Power Amplifiers, December 2000 2 S. Cripps ,1999, “RF Power Amplifiers for Wireless Communications”, Section10.2, Artech House, London, UK Ryde Business Park, Nicholson Road, Ryde, Isle of Wight, PO33 1BQ Tel: +44 (0)1983 618 004 Fax +44 (0)1983 811 521
[email protected] www.milmega.co.uk
