UofA Radio Project

The purpose of this design project is to provide an opportunity to design a simple direct conversion radio receiver based on given circuits, and to measure the performance of their design.  It consists of five modules: local oscillator, mixer, passive filter, active filter and amplifier and audio power amplifier. Circuits are given and a PC board, with some component values specified. There is a degree of freedom for design in choice of components, and even some of those that are given in the notes, if they are commonly available, can be varied if so desired.

Oscillator

The oscillator uses a rare 5µH slug tuned coil with a datasheet Q of 80 but with a much lower Q in reality. The active device is a JFET 2N5485. The oscillator circuit is a self-biasing Colpitts oscillator with a source bias resistor setting the drain current and gate-source voltage (the gate is grounded at dc). This combination makes a somewhat marginal oscillator and careful attention needs to be paid to the oscillation condition. The best approach might be:
  • Measure the Vp and Idss of the JFET.
  • Measure the coil Q, preferably using a tuned circuit at the same frequency of the oscillator and measuring the bandwidth.
  • Find an expression for the oscillation condition using the usual theory of oscillators (don't blindly use any expression for a Colpitts oscillator pulled out of a textbook - it will almost certainly be wrong).
  • Use this information to compute the source bias resistor value at which oscillation will occur. An OpenOffice CALC spreadsheet model can be built quite simply to compute the oscillation condition for a range of bias currents from 0 to Idss. The drain resistor needs to be chosen so that the JFET is operating in saturation mode.
  • Study the oscillator behaviour: theoretically; with simulation; and on a prototyping board until its behaviour is well understood. Try different frequencies, tuned circuit combinations and bias conditions.
  • Choose a value for the source bias resistor and adjust other components to obtain a good oscillator output. Avoid the temptation to set this too close to the point at which oscillations stop as it may result in an unreliable circuit.
PSpice simulations should show similar behaviour to the predicted theoretical results, however in practice the oscillator may not oscillate at the design frequency despite the insistance of PSpice. Also the PSpice waveforms will not look anything like those obtained in practice. The difficulties seem to be with the low inductance coil, but the reasons for the anomolous behaviour are not clear. In the AM band, radios normally use inductors of about 1mH and quite small capacitances (which are in fact variable to allow tuning). These work much more reliably with this type of circuit. Also the oscillator output is taken from a small sniffer coil wound around the inductor. This gives much purer sinusoidal output although it can reduce the coil Q if the next stage has low input impedance.

The circuit should oscillate when the source bias resistor is set to zero. If it does not oscillate in this case, then you are in real trouble.

Mixer

This should be straightforward. The design could be approached by constructing a nonlinear model of the circuit in OpenOffice CALC and adjusting the source bias resistor, and hence drain current, to maximize the output at the desired frequency (baseband in this case). It is a simple nonlinear mixer (as opposed to a balanced mixer) having significant local oscillator output that must be removed by filtering.

Note that the receiver is a direct conversion type, so the local oscillator is tuned to the exact frequency of the received transmissions.

Passive Filter

The mixer is followed by a passive third order low pass filter with 6kHz cutoff and input and output impedances of 3.9K. Output impedance means that the filter works into a load impedance of 3.9K. With equal input and output impedances, a PI configuration may be used which just happens to use the two 100mH inductances that are specified in the project. If you choose an output impedance different to the input impedance then the Butterworth filter generally has to have a T configuration, and in any case the inductances needed will be different. Note that the actual output impedance is the input impedance of the following stage, and in fact things can get quite complicated if we follow the true path. For the moment just assume it is 3.9K.

A third-order filter is very easy to design if the input and output impedances are equal. For the lazy designer, there is a particularly easy to use filter program called FAISYN that can be used for a 30-day trial. If you can't get your design done in 30 days, then you are not going to make a particularly good RF designer. Make sure that you understand the basics of filter design and the significance of impedances. There are plenty of resources available (see Circuitsage) including design equations and various freely available programs and web calculators.

BJT Amplifier and Filter

The gain for this type of amplifier (common emitter with collector-base feedback) is approximately given by the (negative) ratio of the feedback impedance to the source impedance. The small signal analysis is quite messy but in this case the approximation should hold quite well. The source impedance is set by the previous stage (mixer and following passive filter). This is about 3.9K in the mid-band region of the passive filter. The value of the feedback resistor can then be determined from the specified gain for this stage. The feedback capacitor is easily found since the feedback RC circuit is to form a lowpass filter with cutoff frequency 6kHz.

A problem is that the input impedance of the amplifier is highly capacitive, according to Miller's theorem. This theorem states that the feedback RC combination will appear as an equivalent RC combination across the amplifier input, but with the impedances much reduced by the raw amplifier gain (which is about 150 to 200 depending on biasing). What effect does this have on the previous passive filter? Realistically this should be included into the design of the passive filter. However the reality is that the output impedance of the passive filter (the input impedance of the amplifier) is now much smaller than its input impedance. So, if we want to have a Butterworth filter, we would be forced to use a configuration with inductances quite different from those provided for the project, and maybe need to use a T configuration rather than the PI configuration required in the project.

Looking beyond this, we see that the amplifier does not enhance the low pass filter in any way. The Miller effect places the RC circuit that determines this filter behaviour across the input to the amplifier, where it combines with the passive filter final capacitor. Therefore no additional pole is added. The Miller capacitance at the output combines with the load resistances to give a pole at much higher frequency (about 4 times the LPF cutoff frequency) and so has only a minor effect on the overall filter transfer function.

The recommendation is that the feedback part of the amplifier be omitted, and the amplifier designed to provide an approximate 4K input impedance. The amplifier will have a small signal gain of about 160. If lower gain is desired, an emitter degeneration resistor can be added. This has the additional advantage that the input impedance can be more accurately controlled.

The BJT used is not critical.

Audio Output Stage

This is basically trivial as no design is required. It is simply a join-the-wires exercise and is included so that an audio output can be obtained when the radio is tuned to a station. Nevertheless it can cause problems by oscillating at high frequency. This needs to be investigated and controlled.

Performance

Some suggestions for measurements:
  • The oscillator output quality can be measured in terms of the relative amplitudes of the fundamental and harmonics.
  • The mixer output could be measured also with a fixed signal from a bench oscillator. Measure the mixer product and local oscillator amplitudes.
  • The filter performance can be easily measured by cutting the circuit after the mixer and feeding a 50 ohm audio source through a 3.9K resistor into the passive filter input.
  • Using an RF bench source the mixer and filter performance can be measured at a number of signal levels.
A good exercise would be to attempt to look at third-order intercept problems which may be significant with this architecture, but it may be a bit advanced for this level.

Circuit Board

There are some comments to note about the circuit board:
  • The inductor L6 is the ferrite aerial and should not be inserted (as with C6) until after tests have been completed.
  • When L6 is connected onto the board, its common wire needs to be connected to ground (currently it floats in mid-air).
  • The volume control pot should be wired directly under the board, not on top of the board.
  • C5 on the circuit diagram seems to be C8 on the board.
  • C13 floats in mid air and needs to be connected to a nearby track.
  • C15 would seem to be a ceramic capacitor to complement the electrolytic C20, probably about 0.1µF.
  • C14 might be the 47µF that appears some distance from it on the circuit diagram.
  • The power to R3 should come from the 12V line, not the mixer filter resistor R7.
  • D1, D2, L6, Z1, TP9, C21 do not feature in the circuit diagrams and are left open.


First created 9 October 2004.