Taken from 'The Gramophone', December 1965

Salute The Flag. As regular readers will know I have spent several months of this year enjoying the company of firstly Danish and later several American or American-inspired amplifiers. Pleasurable and interesting though this has been, there comes a time, and correspondents have not failed to point it out, when it might seem that one's own countrymen are being ignored in favour of these friendly invaders. This is certainly not so and it is no more than the luck of the draw that our recent reporting, which broadly speaking represents the new arrivals on the high fidelity scene, has had a foreign accent; now comes the occasion to redress the balance.

For some time I have been experiencing the greatest possible musical pleasure from my use of an example of the latest range of power amplifiers produced by Radford Electronics of Bristol. I have only been waiting for the latest matching control unit, their SC22 first shown at the Audio Fair, to become available before introducing the admirable combination to readers. Although I have had the SC22 for less than a month, it is already obvious that it matches the extremely high standard set by the power amplifier and one can recommend them both without any hesitation against competition from anywhere in the world.

But first things first; let me deal at length with the power amplifier which itself fully merits a detailed report.

Bearing in mind the design pressure brought about by the rapid growth of high fidelity in recent years, anyone might be excused for thinking that the desirable circuits for valve amplifiers have all been explored and exploited, and that we have now reached a state of stability if not of complete finality—particularly now that the transistor is so often seen looming through the haze on the audio horizon. A detailed and rewarding study of the above specifications should convince the same 'anyone' that his thinking is fallacious. This is the sort of result not often claimed for any amplifier at any price, even by the most uninhibited advertising man. I can assure you that it is true; indeed, my measurements showed it to be too modest in some details: so now let us see how it is done.

The background to progress
Mr. Arthur Radford has for some years been well known for his skill in the design and manufacture of transformers. Many well-known makes of both domestic and industrial equipment of good standard incorporate them. His concentration on the requirements for a high fidelity amplifier soon brought about dramatic improvements in what had always seemed a stumbling block, the output transformer. In a report on an earlier Radford amplifier design (THE GRAMOPHONE, February 1961) I was able to point out some of them and perhaps give the impression that this important component might no longer be the weakest link in the amplifier chain. This soon proved to be the case and in an interesting article in the September 1962 issue of our contemporary, Wireless World, Mr. A. R. Bailey, M.Sc., of the Bradford Institute of Technology, put forward an explanation.

Technical considerations
As is well known, the very low distortions sought for in high fidelity amplifiers can only be obtained by the use of quite large amounts of negative feedback—a term used to describe the return of some part of the signal at the output to oppose the signal at the input. However, there is a limit to the amount of feedback which can be employed and therefore a limit to the amount by which the inherent distortion can be reduced. This limiting factor is loss of stability. The reason is that, at both the low and high frequency ends of the spectrum, changes of phase take place inside the feedback 'loop' and, if these changes total 180 deg. within the range where the equipment is still amplifying, oscillation will occur.

As these changes of phase can be influenced by external connections, (e.g. by capacity loads such as loudspeakers incorporating some types of crossover, by electrostatic loudspeakers, or even by long leads) it is necessary to have large safety margins. In the majority of present designs it has been found necessary to make the gain of the amplifier fall off at low and high frequencies so that, by the time the 180 deg. point was reached, oscillation could not occur. To do this means that at the low and high frequency extremes, where it is most wanted, the feedback falls away to nothing and the distortion can rise considerably. Even so, many current amplifiers have been shown to be prone to oscillation with some loads, although their feedback is quite modest.

The solution to the problem has been known for many years; to design an amplifier with very low inherent distortion and very wide band-width before feedback is applied. In this way the amount of distortion to be dealt with is minimized and, by using one deliberate attenuating circuit at the extremes of the low and high frequencies required (so as to produce only a 90 deg. phase change), a stable amplifier will result. Unfortunately, until recently both the output transformer and the known phase splitter circuits ran into trouble at frequencies only about an octave above the audio range and similar problems were met at a few cycles per second; so deliberate attenuating circuits, although incorporated from necessity had to be operative in the range where they were detrimental.

Mr. Radford, by careful balance of the distributed capacity between windings and reduction of leakage inductance to a bare minimum, has succeeded in raising the shunt resonance of the output transformer to between 150 and 200 Kc/s. Mr. Bailey, by changing the first half of the conventional 'long-tailed pair' phase splitter to a pentode stage has raised the bandwidth at the 3dB point from 20 to 200 Kc/s. Thus the scene is set for an amplifier in which 26dB of negative feedback can be used over the whole audio spectrum and yet can boast those magic words. Unconditionally Stable, and this without any load at all and even when deliberately driven into complete overload.

Circuit details
An EF86 low-noise pentode is used in the first stage with the low and high frequency attenuating circuits at the output, before the direct coupling to the long-tailed pair phase splitter. This uses an ECF82 triode/pentode valve, originally developed as a frequency changer for TV tuners. The low impedance triode is used in the second grounded grid section and the pentode as the input stage. It is here that the improvement in high frequency response occurs, due to the avoidance of 'Miller effect'. The output stage is conventional, using EL.34 valves in a 40 per cent ultra-linear connection. Bias for the output valves is derived from a separate negative supply with its own winding on the power transformer, a silicon rectifier and smoothing components. A pre-set potentiometer and a test point are provided for each of the four output valves. A main HT supply of 425 Volts is obtained from a bridge of silicon rectifiers, a large choke and 100 microfarad reservoir and smoothing capacitors. This configuration is less prone to the disadvantage often found with the cheaper RC arrangement of a considerable increase in hum level when the amplifier is delivering power. A separately smoothed supply of HT is available for a preamplifier, together with a spare unearthed heater winding.

Construction
The amplifier is constructed on a welded steel frame, finished in a very light grey enamel. At each end of the deck are chrome finished steel hoops which serve to permit easy lifting (note the weight), to protect the valves, and to enable the equipment to be stood upside down without damage for service or inspection. Immediately inboard on each side is a small chassis mounting the valves and a few small components for each channel. In the centre is a larger chassis, normally covered by a grey enameled steel ventilating grille over the four iron cored components and the 100 microfarad capacitors. The power transformer, choke and pair of output transformers are all fully impregnated and of a size calculated to inspire confidence in their performance and longevity.

Along the front of the frame are the pairs of substantial loudspeaker terminals, with the 4, 8, 16 Ohm impedance switch by their side. An octal socket is provided for auxiliary power and also carries mains switching leads for remote control; adjacent is a local mains on/off switch, voltage tap and fuse. At the rear, out of twiddler's way, are the four pre-set controls for bias.

A bottom plate fitted with four non-scratch plastic feet covers all the wiring and small components which are mounted on a pair of stout printed circuit boards. It is noticeable that all parts are of very high quality and the gold band indicating 5 per cent tolerance is on every resistor.

On test
What is there to say about an amplifier as nearly perfect as this one? As far as listening goes, the thing is completely anonymous: it neither adds nor detracts from whatever is fed into it, and never even gives one the slightest suspicion that it is in any way letting the side down. Rich, fat, complacent bass pedals have poured from it, startling transients have jerked the listener upright, but never anything unmusical—and, in between, not the slightest indication that it is switched on, even with an ear against the loudspeaker.

There is little point either in including measurements, when they can be taken as those printed in the specification as a minimum. This is one of the very few amplifiers yet tested which gives its rated power when the mains supply is at the power limit of tolerance for the tap in use. Driving both channels together dropped the available power per channel by only 3 per cent; this is because the change in HT volts from no signal to maximum power from both channels was only 22 Volts. Additional distortion measurements were made at low power and middle frequencies and they yielded the fantastically low reading of 0.01%, little more than the practical achievable minimum of the measuring equipment. Similarly the square wave response at 1 Kc/s almost equalled that of the generator, only a slight brightening of the normally almost invisible verticals and a tiny (2%) single, very high frequency overshoot became noticeable on direct comparison. It was necessary to go to 50 c/s and 5 Kc/s before departures from the originals became obvious to a casual observer. Two measurements are not included in the specification, those for signal to noise ratio and crosstalk. I found the former to be 85dB and the latter not to exceed —55dB anywhere in the 20-20,000 c/s range.

I have tried hard to find even a small chink in the standard set by this amplifier. It was impossible to provoke any slight sign of instability with any load, and even an hour running in complete overload with one output shorted and the other left unloaded did not raise a blush; so I have failed. What can I say then? Back to the beginning—Salute the Flag, it is British!

A report on the SC 22 control unit will follow next month.

GEOFFREY HORN.