The Superhet

Superhets, or Supersonic Heterodyne Receivers, to give them their full handle, are the type of radio which has reigned supreme, just about unchallenged, from about 1937 to the present day. They take a little understanding, but they work well with minimum fuss and are easy to use, so the effort of understanding them is well worthwhile.

Typical Superhet.

We give the circuit of a typical superhet. It's a Pye P76, from about 1950. But it's pretty much the same as a lot of others. The circuit diagram is a big one. It will be worth downloading it and printing it, if you want to study it closely. It's a biggish GIF, and if you print it landscape on a piece of A4, it'll be fine.

Pye P76 Circuit Diaram

The Frequency Changer.

Working from left to right, the first valve is the frequency changer. You will see that this is actually two valves in the same bottle. The one with all the grids is a hexode, and this is the amplifier part. The incoming signal is tuned by the aerial coils (L1 to L4) and the tuning capacitor (C11) and applied to the first (control) grid, just like it was in the TRF. The other part of the valve, the triode, is an oscillator, and this is tuned by a capacitor and coil assembly (L5 to L8 and C25), with the two tuning capacitors for aerial and oscillator both joined to the same shaft. The oscillator is designed so that it runs at a frequency which is 470kHz different from the frequency tuned by the aerial circuit. This difference is maintained as the set is tuned up and down the band. Hence all the funny little trimmer capacitors etc. (C15, C16, C19, C20 to C24) Without them the difference would not be a steady 470kHz, which wouldn't suit our purposes.

The anode of the triode is a convenient place from which to take the oscillator output signal. It is coupled to the third grid of the hexode section. The overall effect of this is that the hexode is given the tuned signal to amplify, which it does, but it is also being forced to respond to the oscillator signal at the same time, and the resulting output is a distorted mess as both signals have chopped each other up inside the amplifier. The mess of signals at the output contains many components. There is the original input signal. There is a component at signal frequency plus oscillator frequency. And another at signal frequency minus oscillator frequency. Not forgetting the one at oscillator frequency. But one of these is very useful, because we have engineered the difference between signal frequency and oscillator frequency to be constant. So wherever we tune the set, one of these signals is always at the 470kHz which we chose to engineer as our difference frequency. And the other essential bonus is that this frequency will also have become modulated with the audio information which was modulating the original tuned-in carrier wave.

This 470kHz signal is selected from the mess at the hexode anode by means of a tuned circuit which is one part of a transformer, T1. The secondary is tuned as well, also to 470kHz, and feeds the signal on to the rest of the set.

The 470kHz signal is called the Intermediate Frequency (IF) signal, and the transformer is the first IF transformer. The advantage of the superhet is that this IF stays the same wherever the radio is tuned, so all the rest of the radio only has to be fixed-tuned to 470kHz once, at manufacture, and then left severely alone. (It sometimes needs a re-tune after servicing, but that's all.) So we can be as clever as we like with the rest of the set, but the user only has to tune one ganged pair of tuning capacitors with one knob. Simplicity of operation with good results. No wonder it has always been a good seller.

Different sets use slightly different IF frequencies. The pre-war sets used frequencies around 110kHz to 130kHz. This suited them well, because it is a fairly low frequency, and was easy to use with early, unsophisticated valves, which were not easily capable of high frequency operation. The problem with this set-up is that with such a low IF, it is possible to have problems with receiving image stations, which show up as whistles on the desired station. The problem happens because RF+IF and RF-IF frequencies are only about 250kHz apart, and signals at either frequency will both work the frequency changer to produce a signal at the IF frequency. Obviously, the aerial input coils will be tuned to only one of these incoming signals, so the other should be filtered out. But if the desired one is very weak and the unwanted one is very strong, both may crop up, to an extent, at the input of the frequency changer, and both have an effect in the IF stages. The result is usually an annoying whistle which cannot be tuned away properly.

One solution to the problems encountered with the image stations is to use more tuned circuits in the aerial circuit to give sharper selectivity, and reject the unwanted signals more strongly. However, more tuned circuits mean more sections on the tuning capacitor and more difficult circuit alignment. This all takes time and money.

A better solution is to use a higher IF frequency, and then the frequency of the potential image station is so far removed from that of the desired station that problems are much less likely. With the introduction of better valves, towards the end of the 1930s, a higher IF became practical for little extra effort or cost, and the standard IF then changed to the area between 450kHz and 470kHz. If you're not sure of the right IF for your set, 465kHz is a good guess, but a brief sweep around with the signal generator might pinpoint it.

The IF amplifier.

We have already seen the IF transformer which passes the signal into the IF amplifier, V2. And that had two tuned circuits in it. The IF amplifier consists of a simple stage of pentode amplification, with another IF transformer in its anode to couple the signal further on into the rest of the set. And here are another two tuned circuits. So we have the tight tuning afforded by five tuned circuits (including the one in the aerial tuning circuit) and in order to re-tune, we only have to re-tune two circuits simultaneously. Of course there's no reason why we couldn't have another IF stage, with another valve's gain contribution and another two tuned circuits in another IF transformer, and it's no more difficult to tune.

The Detector.

Since we have been so clever so far, we now have a good size signal ready for detection. We don't need to have gain in the detector itself, and if we use a simple diode detector we don't have a reaction control to contend with any more, either. The diode is part of the double-diode-triode valve, V3. The secondary winding of the second IF transformer has a large signal at IF appearing across it. By connecting one end of this winding to the anode of a diode, with the corresponding cathode strapped to chassis, we have ensured that one end of the transformer is grounded by the diode whenever it tries to go positive. Therefore, whenever the other end tries to go negative, it is able to do so because current can flow in the winding, but when the other end tries to go positive, the end connected to the diode anode goes negative and no current can flow in the circuit because the circuit is broken by the non-conducting diode. So detection takes place with a negative-going audio signal appearing at the end of the IF transformer remote from the diode anode. This audio signal is developed across R9 in series with R10, and C28 filters out the IF component in the detected signal. C31 performs additional IF filtering before the recovered audio goes away to the wavechange switch and then off to the grid of the audio amplifier stage, the triode part of the double diode triode valve. The switching of the audio breaks the feed from the detector when the radio is switched to Gram and connects the input of the audio amplifier to the Gram sockets in stead.

What is not immediately obvious is that there will be a negative DC offset on the recovered audio, which will depend on the strength of the signal. A diagram saves a thousand words.

The AGC System.

The other diode of the double diode triode is the active part of the Automatic Gain Control system. With this sort of radio, and a half-decent aerial, we actually have the situation where there might be too much gain in the presence of a strong received signal. This might cause overloading of the frequency changer (FC) or IF stages. Too much gain is a problem not encountered very often, but it's easy to sort out.

The IF amplifier valve is chosen to have "vari-mu" characteristics. This means that, at least for small signals, it is linear and amplifies cleanly over a reasonable range of negative grid voltages, but if the grid is only slightly negative it amplifies the signal more than if the grid is much more negative. To a certain extent the FC valve will do the same thing, and this too is fed some of the same control voltage.

Grid circuits take practically zero current, so as long as we can, by hook or by crook, end up with a voltage which does what we want, we don't have to worry about its current capability. the method is quite crafty.

At the anode of the IF amplifier valve is a large IF signal. Some of this is taken off via the low value capacitor C29, which passes IF but blocks out the DC voltage on the anode of the IF valve. The result is fed to the other diode anode. This rectifies the signal such that it cannot go positive, because the diode conducts and pulls all positive going signals to chassis, but negative excursions are allowed, as these only cut the diode off. So we have taken a signal which would otherwise have been going positive and negative by equal amounts, and chopped away all of the positive. Obviously, the size of the signal determines how far positive and negative it would have gone, and even after our trick with the diode, we have a measure of the size of the IF signal by how far negative it goes. This is almost exactly what we wanted, as it goes more negative for bigger signals, therefore being potentially useful to cut back the gain on strong signals. However the signal at the diode anode is rectified IF, and fluctuates in time to the music modulating the carrier.

By filtering the voltage with a very long time constant, we can not only average out over a long enough period to lose the IF component, we can also lose the modulation component. Filtering with a long time constant, however, needs either high value resistors, or large capacitors. Large capacitors are expensive. Resistors cost much the same whatever the value, within reason. The solution is obvious, given that the grid circuit does not need current as such. The AGC voltage is developed across R19 (1 meg) and filtered by R18 & C18 (1 meg and 0.05uF giving a time constant of 50 milliseconds, which will filter out all modulation at frequencies above 20Hz) before being fed to the bottom of the grid winding of the first IF transformer. It is also fed to the grid of the FC via a 1 meg resistor, though no more filtering is used.

The Audio Amplifier.

The signal from the detector is fed to the wavechange switch, where the Gram sockets can be switched in if required, and then via C33, to get rid of the DC offset on the detected audio, and onto the volume control R12. The bottom of R12 could have been connected to chassis, but this set uses audio feedback. More on this later.

The signal from the wiper of the volume control is coupled, via C34, to block DC, to the grid of the triode of the double diode triode. It may seem that we are paranoid about blocking DC, but the purpose of this capacitor is to keep the DC voltage on the grid of the valve away from the volume control. This may seem strange at first sight, because there is no obvious grid bias for the triode grid. However, between grid and chassis is a resistor, R14, of 10 megohms. This is a high value resistor, and the space-charge around the cathode and first grid of the valve will be able, when only loaded with 10 megohms, to work up a few volts of negative bias on the grid of the triode. This is convenient because it gives the required negative bias on the triode grid for undistorted amplification without having to use the alternative method of inserting a resistor between the cathode of the valve and chassis and returning the grid to chassis via a lower value grid resistor. This would work fine for the audio, but would be a nuisance here because the same cathode is the cathode for the detector and AGC diodes, and the resistor will make it take up a small positive bias, leaving a ground-referenced grid negative with respect to cathode and biassed correctly, but the diodes requiring a positive bias before they will conduct. Although the resulting diode connection problems are not insurmountable, it's much easier this way.

The audio frequency component in the anode current of the triode develops an audio signal across the anode load resistor R17, 220k ohms. The resulting signal is coupled to the grid of the output stage via C35 to lose the DC, once again.

This set is a little different from most in that the anode of the audio stage, the screen grid of the IF amplifier, and the FC's oscillator anode, all of which work very nicely at voltages comfortably below the voltage of the main HT rail, are all run from a special little supply generated from main HT through R16, with C36 smoothing the whole thing to make sure that no one stage manages to superimpose any signal on it, and thereby interfere with the other stages sharing the supply. Most sets don't bother with all this, and just run the relevant electrodes from HT+ via a resistor chosen to drop enough volts so that everything runs happily.

The Output Stage.

Nothing much of note here. An output valve throws around significant current and requires grid bias. It can't be done using the space charge and a big grid resistor, as there's too much current flying around inside the valve. In stead, a resistor is inserted in the cathode lead, and with the full DC standing current of the valve flowing through it, it develops a small positive voltage. Since the control grid is referenced to ground, and the cathode is positive, the effect is that required, as grid is negative with respect to cathode.

If the resistor alone was added, an unfortunate side-effect would pop up. As the valve is turned harder, it will pass more current, which will make the cathode voltage increase, which will tend to cut back the current again. While this is fine for DC operating conditions, making the valve to a large extent self-regulating, it's a pain if it happens at audio frequencies as well, because the thing is trying to compensate for its own drive signal, which is not at all what we need, and the effect is to cut back the gain. The solution is to add C40, a 50uF electrolytic, so that slow DC drifts will tend to self-compensate, but the cathode voltage cannot change quickly enough to compensate for the drive signal and spoil the gain.

The output valve drives the speaker through an output transformer. This is pretty standard, but it has what looks like a tapped primary. On closer inspection, you will see that the power feed for the anode of the output valve is at the tap in the winding, and comes direct from the cathode of the rectifier valve. This point has the advantage that it is the highest voltage in the set, but the DC there is not well smoothed and hum could result. The idea of the extra bit of the primary winding is that the HT current for the rest of the set goes through that part of the winding, and it is wound so that the hum current in this part of the winding produces a magnetic flux in the transformer core which is in the opposite sense to that produced by the hum current in the anode winding of the output valve, and the hums thereby cancel out (at least roughly) in the core and give a substantial reduction in the hum getting to the speaker.

C41 is connected from the anode of the output valve down to chassis, and is often called a tone correction capacitor. This it is, and the larger its value, the steeper the roll-off in treble reproduction. However, this is frequently only half of the story.

The signal is fed to the grid of the valve via R21, 10K ohms. If a grid does not take current, it's difficult to see what use a 10K resistor could be. And if it did anything, it looks like it might cut back the signal, which would be a bad thing. However, this resistor and C41 are often more important than they seem. R21 works with the stray capacitance which is an unavoidable consequence of the physical construction of the output valve, and can be thought of as being a few picofards between grid one and cathode. This working with R21 is obviously a filter, passing low frequencies but rolling off high ones. It is chosen to pass AF, but make sure that IF is cut back if it is present. C41 does two jobs at the anode. A slight roll-off for the higher audio frequencies, certainly, but hopefully certain death to any IF which has got this far.

The presence of IF as far along the set as the output valve may seem unlikely and unimportant. Maybe. You'd hope so. But there's many a set where the IF filtering at the detector stage isn't great, and many of those sets have quite a bit of IF mixed into the audio even after the attention of the audio amplifier stage. If this gets fed into the output stage, it might get amplified, but it certainly won't get into the speaker, and if it did no speaker would reproduce it. But if the output valve is responding to IF, its (considerable) current consumption will be varying at IF, and this will mean that there will be considerable IF currents drawn from the power supply. All of this IF current must be supplied by the smoothing electrolytic, C38, which, especially if it's an original component still seemingly working well, might be a problem. Electrolytic capacitors work well at the sort of frequencies found in power supplies, which in this set is the 100Hz due to full-wave rectification of 50Hz mains. They go pretty well at audio frequencies (depending on how fussy you are), but they have too much inductance as a consequence of their physical construction to be able to help much with signals at IF frequencies. And they often get worse with age. So if there is significant IF current drawn from the main power supply, there is no way of preventing it from modulating (contaminating) the power supply to the whole set. The output valve takes so much current compared with the rest of the set that whatever it's doing affects the power supply a bit. And if there's IF coming back to the IF amplifier through the power supply, there will be feedback. And where there's feedback without care or luck, there may well be oscillation. And there are many sets which will suffer awful IF oscillations, making them pretty much unusable, when the equivalent of C41 is removed or falls off.

The Negative Feedback.

Quite a posh set, this one. The output transformer secondary drives the loudspeaker, and the drive to the speaker is sampled and sent back into the audio stages as negative feedback. In fact, it's better than that, because the tone control works by altering the characteristics of the negative feedback.

The fundamental idea here is that some of the output signal is sampled, and taken back to the input of the audio amplifier, where it is fed in in opposite phase (working against) the normal audio input. This negative feedback is aimed to reduce the gain of the amplifier a little, but at the same time reduce the noise and distortion. And this set was meant as a quality job.

The volume control, R12, can be seen to have audio fed into the top via C33, from the detector or Gram sockets, as appropriate. But the bottom of the control does not go straight to chassis. It goes down via R13, 270 ohms. Into here is connected the feedback signal from the output transformer, having come via R22 & C37 and R15. This means that the grid of the audio amplifier sees, at any time, the difference between the input signal and the fed-back signal. So the operation of the device is changed, and in stead of being just a normal amplifier, it is now in order to look at the audio amplifier within the feedback loop as just an error amplifier. If the output is too low to be a faithful copy of the input, the feedback will make this plain and an error will be found, and this will cause the amplifier to compensate. The same argument holds for when th output is too high. So in stead of having to rely on the linearity of the valves for low distortion, we are now able to engineer a system where we can improve the accuracy of the amplifier (at the cost of some gain) to achieve better results. The core of the matter is that the audio amplifier has changed its responsibility to merely correcting errors as quickly and efficiently as possible, which is a simpler job than its previous one of being a totally linear amplifier. In its new role, it will sense its own distortion and correct for it automatically.

At low frequencies, R22 and R15 provide some resistance in series with the feedback to drop the feedback to a suitable level. At higher frequencies, C37 starts to have the effect of dropping the value of R22. So at high frequencies, there is less series resistance in the feedback path, so there is more feedback, and therefore less gain. However, this is less to do with rolling off the treble than with getting rid of supersonic frequencies which might otherwise give rise to instability and shrill, reedy, and even whistley reproduction. The value of C37 was no doubt the subject of much measurement and calculation. This does not just happen right by accident, and messing around with the value of C37 is not advisable.

The tone control is less easy to understand, and we don't propose to go into the nasty business of that here. If you thought it looked surprisingly like positive feedback, going to the other end of the volume control, you'd be quite right. But filter design, feedback and stability are subjects for many complete books. At least if you get a butchered P76, you'll know how to re-build that horrible tone control circuit, because we've given you the diagram!

Info Main