Feeding the Quality Amplifier

One receiver, the QA Receiver, has been designed especially for this amplifier, and the alterations necessary to several others for use with it have been described. These sets, however, do not meet all requirements, for the type of receiver necessary to give the highest standard of reproduction depends on many factors, which vary in individual cases. There is no doubt that a set of extremely high sensitivity is unnecessary, for it is rare that atmospheric conditions permit high fidelity reception of very weak stations. On the other hand, few in these days are content with purely local reception. Many readers of The Wireless World have suggested, therefore, that the Variable-Selectivity IV with the omission of the output valve and the mains equipment would make an ideal receiver to precede the Push-Pull Quality Amplifier. The writer took rather the same view, but anticipated that much minor modification would be needed before the receiver could be considered wholly satisfactory for this purpose.

It was decided, therefore, to redesign the appropriate portion of the Variable-Selectivity IV and to determine, by both aural test and measurement, whether or not it would make a worthy companion to the amplifier. It may be said at once that many alterations were found to be necessary in order to obtain the highest quality with freedom from adjacent channel and image interference, and the maintenance of adequate sensitivity. The final receiver, which has been entitled the QA Super, and will be fully described in an early issue of The Wireless World, bears at first sight little kinship with the Variable-Selectivity IV, and it is consequently thought that it will be both entertaining and instructive to trace its development from the original small superheterodyne, In this way, the necessity for the various changes will be better appreciated and the need for close attention to the smallest details when the highest quality is demanded will be better understood.

The QA Super in the early stages of development.

At the outset it was decided not to aim for a frequency response perfect from the point of view of quality! Perfection would demand a response curve flat from 30 to 15,000 Hz, but a receiver as good as this would be of rather limited application. When broadcasting stations are spaced by only 9,000 Hz it is only possible to reproduce frequencies higher than this when the wanted signal is so strong that by sheer power it swamps all interference. This can occur only in local reception, but even then experience shows that it is only during the daylight hours that the full frequency response can be utilised. After nightfall an annoying heterodyne whistle appears even in districts which are well within the service area of a transmitter. It seems that it is only in rare and exceptional cases that the full response can be used.

Because of these facts it was decided to aim at providing a frequency-response curve as flat as possible from 30 to 8,000 Hz but cutting-off at higher frequencies in order to avoid heterodyne whistles. Experience shows that although there is some loss of quality through this restriction, it is very small, and a response of this order is usually considered to be about 90% perfect.

Fig. 1. - The circuit diagram of the receiver as originally tried out. The changes from the Variable-Selectivity IV are Confined chiefly to the detector and LF circuits.

The relevant portions of the Variable-Selectivity IV were accordingly redesigned and an experimental model was built to the circuit of Fig. 1. A comparison of this diagram with that of the original set shows that the changes are quite small; in fact, before the detector the only alteration is the inclusion of C₁ to permit a higher AVC voltage to be applied to the frequency-changer. The changes in the detector are first of all the use of a duo-diode instead of a multiple-purpose valve combining triode or pentode elements and, secondly, better IF filtering in the detector output circuit. Two resistances, R₁ and R₂, with by-pass capacitors C₂, C₃ and C₄, are used and followed by the whistle filter L₁ C₅ tuned to 9,000 Hz. This is followed by another by-pass capacitor C₆ which not only greatly increases the efficacy of the filter but gives a useful lift to the response characteristic just before the cut-off frequency.

The output of the detector feeds a triode phase-changing valve which gives no amplification but provides the LF output in the necessary form for the Push-Pull Quality Amplifier. The capacitor C₇ across the cathode coupling resistance provides a moderate degree of tone-correction, and its operation, together with that of the phase-reversing valve, was described in detail in a recent issue of this journal. ^{[★] The Wireless World, February 7, 1936.} This valve requires an input of about 4 Volts peak fully to load the amplifier, and as this is of the same order as that required by the output pentode of the Variable-Selectivity IV, the sensitivity of the equipment should be about the same.

Fig. 2. - The overall frequency-response characteristic of the apparatus.

The measured overall frequency response characteristic of receiver and amplifier at low-selectivity is shown in Fig. 2, and is extremely good. At as low as 20 Hz the loss is only 1.6 dB, and at high frequencies it rises to a maximum of +3 dB at about 8,000 Hz; at 9,000 Hz, however, it is -30 dB, a change of 33 dB between 8,000 Hz and 9,000 Hz. From the quality point of view this receiver was obviously satisfactory, for tests of linearity showed this also to be fairly good. Measurement showed the sensitivity to be good, varying from 60 μV to 170 μV on the medium waveband, but the second-channel ratio was certainly lower than is desirable in high-quality apparatus, varying from 50:1 to about 500:1 (Fig. 3).

Fig. 3. - The sensitivity and second channel ratio for both medium and long wavebands are shown here.

The second-channel ratio necessary for the avoidance of interference naturally depends on the ratio of the field strengths of the wanted, and unwanted stations. On the medium waveband second-channel interference can only be caused by fairly low power morse transmitters on wavelengths below 200 metres owing to the fairly high intermediate frequency of 465 kHz. It was known that interference did not prove troublesome with the Variable-Selectivity IV, but it was thought that the protection against second-channel interference might be inadequate in view of the higher standard of reproduction and greater volume obtainable.

The First Model on Test

Tests were accordingly carried out to determine whether or not this was the case, for although one can predict the performance of a receiver fairly well from an inspection of its measured characteristics, it is easy to overlook minor points and a listening test is always advisable. The sensitivity was adequate for good reception of all the stronger Continental stations, and the quality was of a high order, although traces of amplitude distortion were present. The adjacent channel selectivity was not high, of course, but it was sufficient to enable most stations to be received free from interference. It seemed just adequate, but the reservation was made that it would be nice to have it a little greater in order to provide a bigger factor of safety. This applied also to the sensitivity, for although the amplification was adequate with a good aerial no reduction could be tolerated if good distant reception were to be maintained, nor would the set prove satisfactory for foreign listening in cases where only a small aerial could be used.

During the tests a number of whistles were found on the medium waveband and were all traced to the large input obtained from the London Regional transmitter. The simplest and cheapest remedy was found to lie in the use of a wave-trap tuned to this station, and this had also the merit of lightening the load on the AVC system and so reducing any risk of distortion due to overloading in local reception. It proved an effective cure for the whistles and removed every one completely.

Traces of second-channel interference now became evident. They were quite small and only occurred at two or three points in the tuning range. Under the conditions of test the interference could not be called severe and was not really of great importance. It was felt, however, that as in other districts it might be much worse, and that as the measured second-channel rejection was undoubtedly on the low side, some improvement was required.

The points on which an improved performance was felt to be advisable all lay in the pre-detector circuits, and the higher standard of performance was deemed necessary only because of the high fidelity and volume given by the equipment. At lower volume some of the interference would not be heard because it would be below the level of audibility. There is no point, however, in providing a large output stage and being unable to make use of the full output because a trace of interference then creeps in, and it was decided to make an attempt to improve the performance.

The sensitivity and adjacent channel selectivity being just adequate, it was felt that an improvement in the second-channel rejection was the most important, and means of doing this were considered. The easiest way was to add another signal-frequency tuned circuit. This would have complicated the waveband switching and the ganging, however, and would have reduced the sensitivity by at least 50%; as it would give also only a very small increase in adjacent channel selectivity it was not seriously considered. The alternative was the use of an image suppression circuit and various types were tried. Much work had to be carried out before a satisfactory method was found, for most image rejectors are easier to apply to a pair of coupled tuned circuits than to a single tuned circuit coupled to an aerial.

The cathode-coil rejection circuit was found most satisfactory, but attempts to apply it to the signal-frequency circuit of Fig. 1 were disappointing. Ample rejection at any one frequency could be obtained (second-channel ratios of over 10,000:1 were found), but the performance could not be maintained over the waveband. Indeed, the image rejector actually reduced the second-channel ratio at one end of the band while increasing it at the other.

It was at length found that this effect was caused by the coupling between the tuned circuit and the aerial circuit. This was modified and the image rejector now operated properly, giving a point of maximum rejection towards each end of the waveband and maintaining high rejection throughout. Indeed, over much of the tuning range the second-channel ratio was too high to measure!

Fig. 4. - The modified signal-frequency system embodying a cathode-coil image rejector.

The input circuit then took the form shown in Fig. 4. It can be seen that the only changes from Fig. 1 were the omission of the 'top-end' coupling capacity and the provision of the cathode-coil L₃ coupled to L₁ and L₂. Actually, L₃ was a single turn of wire quite loosely coupled to the aerial coils, and for maximum image rejection its spacing was quite critical.

An Extra Stage of Amplification

This proved a satisfactory solution to the problem of second-channel interference, but, unfortunately, the necessity for using a looser aerial coupling reduced the sensitivity. As the sensitivity was originally only just adequate this could not be tolerated, and it was decided that it would be necessary to add another stage of amplification. There were three places in which another valve could be added in the signal-frequency circuits, in the IF amplifier, or in the LF amplifier. Taking these in order, an HF stage would necessitate an extra signal frequency tuned circuit and the image rejector would no longer be necessary. The extra circuit would mean extra waveband switching and more difficult ganging. Ample gain could be secured with little risk of instability, but the extra amplification might easily cause the frequency-changer to be overloaded when the set was tuned to a frequency fairly close to that of a local station, and this would cause distortion, cross-modulation and whistles. In general, the gain of an HF stage in a superheterodyne must be kept quite low. The signal/noise ratio, however, would be improved, for the noise generated in an HF amplifier is about one-quarter of that developed in a frequency-changer. This was not felt to be of any importance in this case, however, for there was no desire to make the sensitivity high enough for the self-generated noise in the set to become of sufficient intensity to need serious consideration.

Now in the IF amplifier the addition of an extra stage would give quite high gain, and as it would carry with it an extra pair of coupled circuits it would greatly increase the adjacent channel selectivity, which was felt to be desirable. It would, of course, help AVC, but no more than an HF stage would do. The only objections were that it would increase the risk of instability and make trimming slightly more difficult. It was anticipated, however, that a gain of 30 times could be obtained without difficulty, and practical tests showed it easily possible for the amplification to be 100 times with perfect stability.

Additional amplification could be obtained in the LF circuits without complicating the trimming or ganging in any way, but it would lead to the detector being operated at rather a small input, and it would considerably increase any risk of motor-boating effects. This method was at first sight particularly attractive, for it would have been possible to obtain the additional gain from the phase-reversing valve by altering its input circuit. This would have necessitated the whole of the detector circuit being unearthed, however, and experience has shown that it is then very difficult to filter the detector output properly, with the result that there is a greater risk of instability due to IF feedback, more likelihood of whistles due to feed-back of harmonics of the intermediate frequency, and a possibility of distortion caused by an LF valve being loaded by IF potentials.

After carefully considering these possibilities and weighing the advantages and disadvantages of each, it was decided that the IF amplifier was undoubtedly the correct place for an additional valve. Not only could the requisite gain be secured easily with such a stage, but the adjacent channel selectivity would be increased and AVC improved. The circuit diagram then took the form of Fig. 1, but with an extra IF stage, and the image rejector of Fig. 4. Additional screening and decoupling naturally proved necessary to maintain stability, and in fact the receiver was entirely rebuilt.- The receiver was from the start perfectly stable and had a sensitivity of about 1 μV! It was, in fact, difficult to measure the sensitivity, since it was of the same order as the noise level. This is always the case when the sensitivity is better than some 10 μV, for the inherent noise in a tuned circuit is about; 1 *mu;V to 4 μV, and the noise introduced by the first valve is of the same order. To obtain minimum noise with a sensitivity better than 10 *mu;V it would be necessary to use an HF stage before the frequency-changer.

AVG and Amplitude Distortion

Excessive gain is no disadvantage, however, for it can always be reduced, and it is as well to have some amplification in hand. An examination of the operating conditions showed a tendency for the frequency-changer to pass grid current in spite of its being operated with the maker's recommended value for the cathode bias resistance. An increase in the bias resistance completely eliminated this tendency reduced the gain to a more reasonable figure, and improved the efficiency and sharpened the tuning of the signal-frequency tuned circuit. The gain was still high, but an increase in the bias resistance of the first IF valve brought it down to the desired level, and gave a maximum sensitivity of about 20 μV. The gain of the second IF stage was kept as high as possible in order to prevent overloading on strong signals.

Extended tests now showed the receiver to have ample sensitivity, selectivity, and second-channel rejection, and to be free from whistles. The one point in which the set failed to give complete satisfaction lay in the AVC, system. The usual delayed diode arrangement was employed and was satisfactory from the point of view of reducing fading, but failed in that it produced a noticeable degree of amplitude distortion. This distortion, which was mentioned earlier, could hardly have been detected in less perfect equipment, which accounts for the widespread use of the method, but it readily showed up in this set when listening to any but weak signals. When tested by listening to the reproduction of a constant frequency note modulating the standard signal generator the distortion showed up whenever the modulation depth exceeded 50%, and proved serious with signal inputs as low as 1,000 μV.

Careful investigation showed the distortion to arise from two different causes. The first lay in the inability of the AVC system to keep the output of the second IF valve at a low enough figure to avoid slight over-loading on strong signals, while the second is inherent in delayed diode AVC. With this system the primary of the last lF transformer is more highly damped when the AVC diode is rectifying than when it is not. Now any change in the damping of this circuit must, necessarily alter the gain of the IF valve, and hence the detector input. Whether or not the AVC diode rectifies depends on its IF input, and unless the delay voltage be very small indeed it must inevitably pass from one condition to the other during modulation. With a large input it will rectify the whole time when modulation is shallow and also on the peaks of deep modulation, but in the troughs of deep modulation it will cease to rectify, and the damping on the tuned circuit-will change.

Determining AVC Requirements

Fig. 5. - The non-linearity caused by IF overloading and by the AVC system is clearly brought out by these curves.

These effects resulted in serious distortion, as the curves of Fig. 5 show. In taking these curves the input to the aerial was set at the figures indicated, and with 30% modulation at 400 Hz the LF volume control was set to give the same output, represented on the arbitrary output scale of voltage by 53. The modulation depth was then varied, and the output plotted against it. Ideally, the resulting curve should be a straight line, and it can be seen that in all cases it is straight up to 30% modulation. The deviations up to 50% are unimportant, but only in the case of a small input (100 μV) is the curve reasonably straight up to 80% the normal upper limit of modulation.

Simple non-delayed AVC was then tried, the bias voltage being derived from the output of the detector. This gave better results from the quality viewpoint, but was not perfect, because the control still permitted some overloading of the second IF valve on very strong signals, and the AVC filter necessarily made the AC load impedance of the detector appreciably less than the DC load. Moreover, the absence of delay reduced the sensitivity of the receiver seriously, and full volume could only be secured from the strongest of Continental transmitters.

The operating conditions were then critically examined with a view to determining the requirements for a perfect AVC system. It was found that to obtain full output under all conditions the detector input should be of the order of 10 Volts. This would permit the output stage to be fully loaded even on stations having a comparatively shallow maximum modulation depth. This input represented a current through the diode load resistance of the order of 30 μA, and this figure was taken as the minimum input to be obtained on all worth-while stations.

The maximum input for the avoidance of distortion in the second IF valve was then determined, and for 90% modulation it was found that the current through the diode load should not exceed 60 μA, although if modulation depths greater than 70% were considered unimportant the current could be as high as 100 μA. The next step was to determine the bias needed by the two controlled valves to keep the detector operating under these conditions, and it was found to be as much as 27 Volts when receiving Droitwich. The detector current was 60 μA, and, as the load resistance of the detector was 0.25 MΩ, the voltage across it was 15 Volts only. Obviously, amplified AVC was necessary.

Amplified AVC

Two basic methods of obtaining amplified AVC are available. One is to use delayed diode AVC fed by an extra IF stage reserved for it alone. This prevents the distortion, because a valve is interposed between the AVC diode and the signal-tuned circuits. Several difficulties arise in practice, however, for the extra IF gain may cause instability, and the very high IF voltages at the AVC diode may cause the appearance of whistles due to feed-back of IF harmonics. Moreover, when the detector input is large it is easy to overload the AVC amplifier so that grid current flows, and then distortion will occur again.

The alternative method is to amplify the DC output of the detector. There are several ways of doing this, but all require a source of steady potential negative with respect to the earth line of the receiver. This is usually obtained by inserting a resistance between negative HT and the earth line, but in this case it was found that the maximum voltage available was very limited. A further difficulty lies in preventing feed-back effects, for it requires very extensive decoupling to prevent motor-boating when amplified AVC is used.

At this stage it was seen that the ideal arrangement would be to provide an entirely separate HT supply unit for the AVC valve. An adequate voltage could then easily be secured, and no feed-back problems would arise. Incidentally, the scheme is not as expensive as it sounds, for the current required is no more than 5 mA, and there is a considerable saving on decoupling components.

Fig. 6. - The degree of linearity obtained in the modified equipment.

This arrangement was accordingly tried, and found so satisfactory that it was unhesitatingly adopted. When correctly adjusted, the detector input varied no more than 1.3:1 for a signal input variation of 15,000:1, and the desired conditions were more than fulfilled. Listening tests new showed the receiver to be distortionless for modulation depths up to 80%. As the curve of Fig. 6 shows, complete freedom from distortion is not secured, but the degree is too small to be audible until the modulation depth exceeds 80%.

The Final Model

At this stage it was felt that finality had been reached, for the original sensitivity and selectivity were retained unimpaired, while the quality was well nigh perfect, and the control of AVC was greatly improved. It is in this form, therefore, that the receiver will appear in an early issue of The Wireless World, when the complete circuit diagram will be given. As explained in this article, the set is a true successor to the Variable-Selectivity IV, and, with the exception of the additional IF stage, the changes are only the essential ones necessary when it is to feed such a good amplifier as the Push-Pull Quality Amplifier. Many constructors would doubtless have been fully satisfied with it without the extra valve and improved AVC system, but the set is intended for more than local reception, and the additional stage provides an ample reserve. Moreover, it is, in the writer's view, true economy, for when the sensitivity is only just adequate for good results, the slightest deterioration in a valve seriously affects the performance, and valve renewals must be frequent if the set is to be maintained in good condition. If the set be more sensitive than is necessary, however, quite a large deterioration in the valves can be permitted before the performance is noticeably affected; the intervals between replacements, at any rate in the case of the pre-detector valves, may easily be doubled.