Fullscreen HTML5Med Res (??MB)HTML4

REMOTE CONTROL By BOB YOUNG Designing UHF transmitter stages Last month we dealt with the simple transmitter for use on 27 to 40MHz, typically using a 3rd overtone crystal and one or two stages of straight amplification. This month we 'II discuss a much more ambitious transmitter operating on UHF. Many radio control applications call for transmitters using ultra high frequencies and this requires some very clever electronic circuitry. Here in Australia, DOTC licenced users are allocated eight spots in the 471.225 to 471.8MHz band for industrial R/C. Over the last three months, I have been totally engrossed in the development of a low power 471MHz Tx and Rx for use on this band and the cleverness called for has been driven home to me in no uncertain manner. My 26 years in electronics have all involved working between 27 and 40MHz, which meant that I was constantly required to devise ways to avoid introducing harmonics in the Tx output. Now, I suddenly found myself in a situation where strong harmonics had to be deliberately introduced, not just into the Tx output, but into the very OSCILLATOR FET BUFFER AMPLIFIER early stages of the transmitter. Not only that, but UHF and above are very difficult bands to work with. The very first thing that you learn when working on these frequencies is that there is no such thing as a short circuit. A shorting bar across a lOpF trimmer has appreciable inductance and therefore tunes beautifully. So minaturisation and UHF go well together. However, if you are producing prototypes, it can be very frustrating. While feeling very sorry for myself in the middle of researching this project and asking myself that perennial question, "How did I get myself into this one?", I was pulled up with a jolt by a reference to Marconi conducting most of his early coded transmissions on 800MHz. It was not until he went after distance that he came down to the lower frequencies. Once again I f X3 IX 6 TRIPLER DOUBLER OSCILLATOR FET BUFFER AMPLIFIER (a) OSCILLATOR BUFFER was made forcibly P' vare of the cleverness of those r _.y pioneers. There was no slipping around to Dick's for silver mica capacitors, lifting the lid on a UHF CB to see how they did it there or checking the output on a Marconi Test Set, spectrum analyser or digital frequency meter for those people. All they had was their genius, a sound grasp of mathematics and unsurpassed determination. What components they lacked, they designed and made. Those people, the people who followed and the present generation who work in this very difficult area of electronics have my utmost respect. Features of 100mW NBFSK Tx Fig.1 shows the block diagrams of three typical 100-500mW UHF transmitters intended for use in industrial radio control. As this is not a construction article, only the broad principles will be discussed. The circuit diagrams discussed in many cases have all of the bias, decoupling and idler circuits removed for clarity of the principles involved. Do not attempt to build these circuits, as they won't work. Ix 3 Ix 6 TRIPLER DOUBLER (b) MODULATOR AMPLIFIER Ix 3 IX 6 IX 6 Ix 6 TRIPLER DOUBLER AMPLIFIER POWER AMPLIFIER (c) Fig.1: these block diagrams show three different approaches in designing UHF remote control transmitters. Note that the oscillator output frequency (f) has been multiplied by six in each case to achieve operation at UHF. JANUARY 1990 73 to 3kHz deviation whereas the "foldback" receivers such as the Philips 2033 and 2050 require plus and minus 4.5kHz for correct operation. Thus our choice of crystal/ oscillator circuit is heavily influenced by these requirements. Actually, the final decision on crystal type is virtually forced upon the designer by the limitations in crystal technology. Most manufacturers in Australia are limited to overtone crystals in the 100-150 MHz range, a figure well short of the required 471MHz. The difference must therefore be obtained from the frequency multiplier stages which follow the oscillator. Deviation & stability Want to control a concrete pour by remote control? No problem. This industrial grade UHF transmitter gives an operator full control of the concrete truck shown above. At right is the view inside the unit. For each of the circuits of Fig.1, the chain commences with the oscillator. As this Tx is intended for low cost, low range NBFSK (narrow band frequency shift keyed] applications, a simple varicap diode is used as a modulator. There are several problems to be considered in the choice of crystal and hence the oscillator circuit. To begin with, DOTC specifications for the VHF/UHF bands usually call for maximum deviation of ± 5kHz on the carrier frequency. Added to this, we have a responsibility to other users to use the minimum spectrum space that modern technology allows. At least this is one problem Marconi never 74 SILICON CHIP had. There were not too many users of the radio spectrum in those days. In addition, NBFSK receivers can require anything from a single shift of plus or minus 1.5kHz to a double shift of plus and minus 4.5kHz deviation for reliable results. The smaller the deviation, the worse the signal-to-noise ratio. Even the full deviation allowed by DOTC results in a poor signal-tonoise ratio and this is one of the shortcomings of NBFSK. It is not until true FM (frequency modulation] is employed, with deviations of ± 50kHz and over, that good signalto-noise ratios are obtained. Typically, most conventional NBFSK receivers require from 1.5 Here again a problem is introduced with regard to the oscillator design. Any frequency shift in the oscillator will be multiplied by the frequency multiplier. Thus, since we need only 5kHz deviation, the maximum oscillator shift is only 833Hz (oscillator frequency 7 6). As we have seen, good results can be obtained from most modern NBFSK receivers at 2kHz deviation, leaving some margin for drift at the transmitter end. In fact, a well designed Rx with a narrow bandwidth will begin to reject deviations greater than 2.5kHz. Once again we see the continual compromise that designers are confronted with. The multiplier stages also magnify the problems of crystal stability and tolerance. Thus, a crystal rated at 5 parts per million will give a final result of 30 parts per million when followed by a 6-times multiplier stage. Because of the very narrow frequency shift required, an oscillator that is very difficult to pull off frequency will give good results in this application. A series mode overtone Colpitts circuit (Fig.2) fills this requirement nicely. Overtone crystals can be cut up to 150MHz reasonably cheaply, depending upon the temperature stability required, and will typically only pull a maximum of 1-1.5kHz. Fundamental crystals are more expensive to cut and this cost escalates above 26MHz, again depending upon the temerature +4-1sv--------~ C4 1-o~w~J II ;rC4 L2 l 1 L1 INTO HIGH IMPEDANCE ~ ~ CJ -:-- Fig.2: this series mode overtone Colpitts oscillator circuit is ideal for use in NBFSK transmitters. Note the tuned collector load for Qt which multiplies the output frequency. r"t?f ·J ::r~t r (a) R1 Fig.3: typical varactor diode frequency tripler circuits. Fig.3{a) utilises an L-section matching network while Fig.3{b) uses an output transmission line matching circuit. coefficient required. However, they will pull much more readily typically from 2-4kHz. The situation for NBFSK R/C model transmitters working on the 27-40MHz bands is quite different. Because the gap between crystal frequency and the output frequency is much smaller, high multiplication factors are not necessary. Radio control transmitters on the 2740MHz bands usually use a fundamental crystal on f/2 (second harmonic) in order to get the required frequency deviation. The required frequency doubling usually takes place in the oscillator output tuned circuit. This approach is cheaper and more reliable than adding high orders of multiplication. I find the conditions under which the crystal is expected to work the big objection to NBFSK modulation as compared to AM (amplitude modulation). In AM, the crystal locks the electronics to the required frequency whereas in NBFSK, the electronics hold the crystal on frequency - a real cart before the horse situation to my mind. In addition, great care must be exercised in matters such as voltage regulation and component stability in particular. The fact that the system works as well as it does is a credit to the modern component industry. Added to this, NBFSK sets are more difficult to service, more expensive to re-crystal and give signal-to-noise ratios in some cases, depending upon the Rx design, much worse than the AM sets. The only genuine advantage that I can see is that NBFSK can be used to transmit the more complex data streams used in PCM sets. For the average flyer, car and boat enthusiest, the AM PPM set is still the most reliable and cost effective unit available. Frequency multipliers Frequency multiplier circuits are intended to generate harmonic signals from the fundamental input frequency. Transistor and FET multipliers will generate usable harmonics up to the 6th although the most commonly used multipliers are doublers and triplers. This is because efficiency falls off very rapidly after the third harmonic. Diodes also work quite well as frequency multipliers (Fig.3). Varicap or step recovery diodes are used at lower power levels while varactor diodes are generally used at power levels above 100mW. If the efficiency is not critical, conventional silicon epitaxial switching diodes may also give good results. The correct choice of transistor is very important in multiplying amplifiers. Many RF power transistors have a significant collectorto-base capacitance that is not directly underneath the emitter "fingers". Most of the series resistance into the base region (rbb') is therefore bypassed and a fairly high quality varactor diode thus exists, the capacitance of which changes with collector-tobase voltage. When used as a frequency multiplier, this transistor can provide noticeable improvements in power gain and efficiency, particularly when used near its upperfrequency limit. The theory of frequency multiplication is very simple and illustrated in Fig.4. In essence, all that is required is to introduce a controlled amount of distortion into the input sine wave. Any nonlinMr amplifier will generate harmonics in the output waveform, however the trick is in the amount of control exercised over the level of distortion. The drive level and bias applied to a multiplying amplifier are quite critical. If the input drive is insufficient to overcome the negative bias, the stage will not function at all. For this reason a preamplifier stage JANUARY 1990 75 vcc Fig.4: a class C frequency tripler, together with its input and output waveforms. The tuned output circuit filters out the unwanted harmonics and provides a flywheel effect at the desired frequency. between the oscillator and multiplier is often desirable. In effect, a frequency multiplying amplifier works in class C. The output is clamped off with a diode to allow the correct level of ringing to take place in the output LC network. The tuned circuit in the output then acts in two ways. First, it provides the necessary filtering of unwanted harmonics and second, it provides a flywheel effect at the desired frequency. Thus the stored energy in this resonant circuit generates the fill-in waveform (when the transistor is not conducting) at the required harmonic frequency. In practise, working with multipliers can present quite peculiar problems and a spectrum analyser is virtually a must. Parasitic oscillations (spurious oscillations occuring at unwanted frequencies) are quite a serious problem in all transmitters and even more so in the VHF/UHF bands. Actually, this is fundamen- tal to the vast difference people find in working with transmitters as against receivers. In a receiver, the power goes up as the frequency goes down, whereas in UHF transmitters, the power goes up as the frequency goes up, presenting the worst possible scenerio for parasitic oscillation. There are many ways to prevent parasitic oscillations and any good UHF book (ARRL Handbook or Jessop's VHF/UHF Manual) will outline the techniques which include the use of ferrite beads, base stopping resistors and neutralisation. Neutralisation Mosfets have big advantages over bipolar transistors when used as RF amplifiers. In a transistor there is a feedback path from the collector to the base which can be adequate to sustain oscillation within the circuit. The method used to eliminate or neutralise the feedback path is called 'unilateralisation'. By comparison, a Mosfet has a very low feedback or reverse transfer capacitance so no special neutralising circuitry is required. This represents a very big saving in production costs, particularly in circuits such as push-pull and pushpush multipliers as shown in simplified form (ie, without unilateralisation) in Fig.5. These two circuits are very interesting as they have some degree of harmonic cancellation, the pushpush circuit amplifying the even harmonics (2nd, 4th and 6th) and attenuating the odd. Conversely, the push-pull circuit amplifies the odd harmonics (3rd, 5th and 7th) and attenuates the even. Fig.6 shows the push-pull version using Philips BSD 12 N-channel Mosfets. The BSDl 2 is a very fast switching device which gives good results as a multiplier. Note that electrical balance and symmetry are important in this type of circuit. The FETs are self-biassed with a pot between the sources providing a balance control. Correctly set up, this circuit will give a good clean output at 471MHz with all harmonics over 30dB down. One very interesting device which I found after I had completed the 471MHz project, and therefore have not tried personally, is the Motorola MRF629 tripler. This transistor is nominally a 2W 9dB gain 470MHz 12.5V amplifier assembled in a TO-39 common (grounded) emitter case. A unique feature of the chip is a pair of diffused Faraday shield diodes which help isolate the common-emitter input from the output. These shield diodes are electrically connected across the output-collector to. emitter by very 3f [ ] (a) Fig.5: typical bipolar transistor frequency multiplying stages. Fig.5(a) shows a push-pull tripler circuit, while Fig5(b) is a push-push doubler arrangement. Both circuits are shown without neutralisation. 76 SILICON CHIP +10V ..,. result in an unstable and noisy Tx output. Once the stage is tuning smoothly and correctly, replace the large trimmer with one that tunes only over the range of the required harmonic. Output stage Fig.6: push-pull tripler circuit using Mosfet transistors. Mosfets have very low feedback capacitance so no special neutralising circuitry is required. short interconnected feed bars. When properly biased, they act as shunt varactor diodes which are able to multiply frequency. Thus, one can design an amplifying multiplier in the stable commonemitter configuration using the simple shunt diode networks usually associated with common-base designs. This device will produce 700mW at 450MHz from a 150mW 150MHz input using a supply voltage of 9-10V DC. The circuit tends to operate in a nearly saturated mode. This keeps the collector current almost constant and thus makes power supply regulation relatively easy. Tuning a multiplier stage Tuning a multiplier stage should present no problems. A correctly working multiplier which has a sufficiently large trimmer capacitor will tune the centre frequency and one harmonic on either side. Thus, a tripler with the trimmer fully engaged (ie, at maximum I I STRONG PARASITIC capacity) will first peak the 2nd harmonic then, as the trimmer is slowly moved towards minimum capacity, the 2nd will fall in amplitude as the 3rd increases. Continuing towards the minimum position, the 3rd will peak and begin to fall as the 4th begins to peak. Thus you should be able to exercise complete control over each harmonic with the tuning smooth and free of sharp or abrupt rises or falls. Should the entire frequency comb rise and fall in unison (eg, the trimmer is acting as if it were an attenuator), then suspect an earth loop or some similar problem. Always keep a close watch for any evidence of parasitic oscillations [a spike out of step with the spacing of the comb, as in Fig. 7) and in particular triggered regeneration. This is a special case in which a parasitic very close to a harmonic locks itself to that harmonic and gives an amplitude peak that is completely out of character with the rest of the comb. This can I I TRIGGERED II piiA:/JIC II II II II I I II II II II II !l 10 2IO 3IO 4IO ~ 5IO Fig. 7: when tuning a transmitter, always watch for evidence of parasitic oscillation. This frequency output spectrum shows a strong parasitic oscillation between 3fo and 4fo and a triggered parasitic which is locked to 5fo. The output stage is fairly routine, if anything at UHF could be said to be routine. The main considerations for this stage are efficiency, harmonic filtering and matching the output transistor to the antenna. The question of cost, efficiency and harmonic filtering are closely related. If the harmonics have been filtered at each stage [where they are much easier to attenuate) and the driver presents a nice clean input to the power amplifier (do you refer to a 100mW stage as a PA?), using class C bias will only reintroduce the harmonics as we have already seen. However, if it is decided that the doubler and PA stage are to be combined, in the interests of reducing cost, then class C bias is a must. Fortunately, a good output network will serve to match the antenna as well as attenuate any reasonable number and level of harmonics to the level required by DOTC. If your application can stand the loss of efficiency, class A bias will give a clean, harmonic free output, requiring the minimum of filtering. One of the nice things about UHF is the size of the antennas. As the wavelength is only 63cm at 471MHz, a quarter wave antenna is only about 15cm long. Compare this to the 2.54-metre long quarter wave antennas we used on our first ground based 27MHz single channel transmitters. Using good antennas on a 1W 471MHz RF link will give R range of about 40 kilometres over water. But potential R/C users should be warned: on land, UHF is very different. The 63cm wavelength is very reflective and, when used amongst steel girders and over land, can give quite misleading results, with dead spots showing up in the oddest places. Note also that this band is only available to DOTC licenced commercial users. ~ JANUARY 1990 77