Duration: January 1996 – March 1999
Position: Development Engineer
Cordless communicator prototype using direct sequence spread spectrum communication.
Design of 1kW HF power amplifier, including harmonic filters, power combiner, Automatic level control (ALC) amplifiers and digital microprocessor controlled ALC.
100W HF Antenna tuning unit.
Grintek Comms was a company who's core business was military products and specifically military radio communications equipment and vehicle intercom systems. I joined Grintek Comms as my first job after university. When I joined the company they were embarking on a new project which was intended to be a secure short range wireless intercom system for platoons of foot soldiers and I was tasked with working on this project. I was working under the guidance of the senior RF engineer.
The intention was to use direct sequence spread spectrum modulation which would not only provide the required secure communications channels but would also allow multiple signals to be transmitted on the same frequency at the same time making use of code diversity to suppress interference. At the time direct sequence spread spectrum (DSSS) was in its infancy, the 802.11 standards such as Wi-Fi were still being developed and complete chipsets to perform these functions were few and far between and expensive. We had to try and develop a system from the ground up using discrete components. We also had to learn how to make spread spectrum systems work and deal with all the teething problems. Initially we attempted to make a system using analog modulation – FM and AM, but this was not very successful. The difficulty being in getting the receiving end to acquire and lock. I designed and constructed a number of prototype circuits as building blocks for this project, including phase locked loop synthesizers, modulators, low power transmitters, receivers and demodulators. The project was shelved for a while due to lack of progress and our inexperience with DSSS and digital communications.
The next project I worked on for the company was a high power 1kW HF Power Amplifier. The amplifier was intended to be used at base stations for long range strategic communications. It worked from 2 – 30MHz. I was provided with an existing design for 400W HF power amplifiers and had to design the harmonic filters, power combiner, automatic level control (alc) amplifiers and digital controller. The intention was to use four 400W amplifier modules for the amplifier. The output from each 400W module was to be passed through a harmonic filter board and then into the power combiners. There would be 3 power combiners which would ultimately output the 1kW signal. Each amplifier module would be driven from an alc amplifier which would adjust the input level so as to keep the amplifiers output in balance with the other modules.
The specification for the harmonic filters was quite demanding - there had to be better than 60dB rejection of the harmonics at all frequencies. Now given that the amplifier output was to be 1kW which is 60dbm, this meant the harmonics had to be suppressed below 0dbm or 1mW. This is quite challenging considering the high currents and voltages flowing in the circuits.
I opted to use a bank of 7th order low pass elliptic filters for the harmonic filters. I split up the frequency range into 8 bands and optimised each elliptic filter for its specific frequency band. The filters would then be switched in or out of circuit using relays.
Normally when one designs filters, one accepts a certain transmission loss through the filter. A figure of 0.5dB loss in a filter is not uncommon and would usually pass as acceptable. However where such high powers are involved 0.5dB loss with a 400W signal equates to 43W lost in the filter! The filter components certainly are going to get warm and cooling is going to be a major issue. It was therefore essential that the losses be minimised with a target of 0.1 to 0.2dB. This necessitated using very low loss components and using exact calculated values for the filter components to try to achieve the ideal performance. I used low loss RF capacitors from American Technical Ceramics (www.atceramics.com) and wound the inductors myself using Micrometals powdered iron toroids (www.micrometals.com) where the core properties were specifically selected for each frequency range.
In addition to minimising filter passband loss to keep heating down, it was also necessary to provide an alternate path for the harmonics. The raw output of the amplifier modules had second and third harmonics in the -20dbc range, which meant the harmonic power was around 4W. I therefore constructed some high pass filters complementary to the low pass elliptic filters to dump the harmonic signals into a resistive load. Here again great care was needed for the filter design, because the high pass complementary filters connected in parallel with the low pass elliptic filters could affect the match at the input node of the filters, resulting in large amounts of power going where it wasn't supposed to. This also meant that the PC board required carefull planning and layout. The boards ended up being about the size of A3 paper.
I added VSWR circuits to the inputs and outputs of the harmonic filter circuits. These are cunning little analog circuits that measure forward and reflected power. This not only provides a measure of the total power, but also information about the match. The signals derived from these circuits would be fed to the digital controller.
With the first harmonic filter PC board layout I overlooked the transmission line impedance of the input and output tracks – after all one doesn't expect to have to use microstrip techniques at 2-30MHz where the input and output tracks are a tiny fraction of a wavelength at these frequencies. But the reality was that the resulting track impedance wasn't 50 Ohms and so was transforming the impedances seen at the filter inputs and outputs causing a mismatch that was appreciable enough to increase the losses to an unacceptable level. I therefore had to rework the boards. First I estimated the permittivity of FR4 board by constructing two test capacitors pieces made using un-etched double sided copper clad board (this was before the days of wikipedia so I couldn't just look these things up online). The two test pieces were rectangular but of different size so that the fringing effects on the edge of the board would differ. I measured the capacitance of these two pieces using the company's HP4195A VNA and then knowing the area of the boards and dielectric spacing arrived at a figure of around 4.5 for permittivity. I then made use of the well known microwave application Puff (www.vhfcomm.co.uk/puff.htm) to calculate a track width which would have a transmission line impedance of 50Ohms. This track width came to around 3.1 mm which I used for the RF input and output feed tracks when reworking the harmonic filter boards.
Having corrected the harmonic filter boards, they worked well and delivered the required harmonic rejection. The next step in the design was the power combiners. I constructed these using transmission lines transformers – coax cable wrapped around toroids. In a power combiner part of the circuit is a 100 Ohm dump resistor which lies across the two input nodes. In principal when both input signals are of equal amplitude and phase then no power is developed across the dump resistor. The dump resistor is there to ensure a 50 Ohm match at both inputs if the inputs signals differ slightly in amplitude or phase. Now consider if one of the input powers is 395W and the other is 405W, the difference is dumped in the dump resistor – 10W, quite a substantial amount of power. Thus large resistors need to be provided which have good RF properties. This also highlights the need for carefully balancing the powers at the outputs of each amplifier module by means of the digital controller.
I constructed the automatic level control amplifiers using PIN diodes and a monolithic gain block. By varying the DC bias across PIN diodes they can be made to act like variable resistors and thus control the attenuation of the RF signal.
The digital controller consisted of Microchip PIC microprocessors (www.microchip.com). All the signals from the various modules were fed into the digital controller and it was responsible for controlling the ALC amplifiers and keeping the outputs of the 400W modules in balance. This was quite tricky, for starters the transfer function of the ALC amplifiers was non-linear which makes design of a control loop particularly difficult. To overcome this I measured the transfer function of the ALC amplifiers and calculated an inverse transfer function which would linearize it. I then created a lookup table in the software to represent this inverse function. Another difficulty I encountered in writing the controlling software was handling the case when the RF input signal was switched on. Prior to the presence of an input signal the control loop would be open and ALC gain turned up to maximum. Then when the input signal was applied and before the controller could react to turn down the gain of the ALC amplifiers the signal level could overdrive all the amplifier modules leading to some undesirable consequences. Unfortunately whilst I was still battling with these issues the company decided to shelve the project so I was unable to see it through to completion.
The final project I worked on for the company was 100W antenna tuning unit (ATU). An ATU in theory is just a bank of inductors and capacitors that can be switched in and out of circuit to provide a matching network for an antenna. The theory behind it is quite simple. The inductors and capacitors are selected so that their values increment in powers of 2 (1, 2, 4, 8, 16...) like a binary sequence. Thus by selectively switching in the components any integer multiple of the smallest value can be obtained. The difficult part of such a design is the practical realisation of the circuit. This is where knowledge and experience comes into play. Depending on which components are selected for a particular match, the voltage at particular nodes in the circuit can become excessive – in the range of 10's of kV. This will cause arcing and breakdown across PCB tracks. The components are switched in and out of circuit using RF reed relays. Here again the reeds can start to “chatter” in the presence of large amounts of RF energy. It was whilst working on this project that I decided to move on to better prospects and joined Nanoteq.