Design for a High-Quality Amplifier - Part Two
amplifier menuMain Menu


Part 2 Details of final circuit and its performance
By D.T.N. Williamson (Late of the M.O. Valve Co.)

THE considerations underlying the design of a high-quality amplifier were discussed in the first part of this article. A circuit of the complete amplifier is shown in Fig. 5. This follows the basic arrangement of Fig. 3(b).

[A painstakenly redrawn schematic of the Williamson fig.5]
Fig. 5 Circuit diagram of complete amplifier. Voltages underlined are peak signal voltages at 15 watts output. (Click on circuit for enlargement)

The design of the individual stages will not be treated in detail, but a review of the salient features may be of value. As a measure of standardization all valves except those of the output stage are type L63 triodes of about 8,000 ohms a.c. resistance.

Initial Stages.— In order to keep the phase shift in the amplifier at low frequencies as small as possible the first stage has been directly coupled to the phase splitter, eliminating one R.C. coupling. The first two stages are thus designed as a single entity.

The phase-splitter section, which consists of a triode with equal loads in anode and cathode circuits, operates partly as a cathode follower, its grid being some 100V positive with respect to chassis.

The anode of the first triode is also arranged to be about 100V positive and is coupled to the phase splitter grid. Due to the cathode follower action of V2 the operating conditions are not critical and no trouble is likely to be encountered from normal changes in valve parameters.

The cathode bias resistor of V1, to which feedback is applied from the output transformer secondary, is kept as small as practicable to avoid gain reduction in the first stage, due to series feedback.

Driver Stage.— The output from the phase-splitter is taken to the push-pull driver stage. Provision is made for varying the load resistors of this stage which, in conjunction with a common un-bypassed cathode bias resistor, allows a considerable range of adjustment to be made in the drive voltages to the output valves to compensate for any inequality in gain.

Output Stage.— The balance of quiescent anode current in the output stage is a matter of some importance, as it affects the performance of the output transformer to a marked degree. In this amplifier, provision is made, by means of a network in the cathode circuits of the KT66 valves, for altering the grid bias of each valve, giving complete control of the static conditions of the stage. A feature of this arrangement is that the valves operate. with a common un-bypassed cathode bias resistor, which assists in preserving the balance of the stage under dynamic conditions.

Output Transformer.— The turns ratio of the output transformer will be determined by the impedance of the loudspeaker load.

It is convenient to make each secondary section of such an impedance that by series-parallel arrangement a number of suitable load impedances may be provided utilizing all the sections of the transformer. A suitable value of impedance is 1. 7 ohms per section, giving alternatives of 1.7, 6.8, 15.3, 27 ohms, etc.

Winding data for a suitable transformer are given in the appendix.

Negative Feedback Network.— The design of this amplifier is such. that no difficulty should be experienced in the application of negative feedback up to a maximum of some 30dB.

Provided that the threshold of instability is not reached, the benefits of negative feedback increase as the amount of feedback is increased, at the sole expense of loss of gain, but there will be little if any audible improvement to be gained with this amplifier by increasing the amount of feedback beyond 20dB.

The feedback network is a purely resistive potential divider, the bottom limb of which is the cathode bias resistor of the first stage.

With component values as specified no trouble should be experienced from instability due to the effects of unintentional positive feedback. Should instability arise it will probably appear as oscillation at a supersonic frequency.

This may be transient, occurring only at some part of the cycle when the amplifier is operated near maximum output. Its cause may be bad layout or an output transformer with a higher leakage reactance than specified, or it may be due to resonance in the output transformer.

A remedy, which should only be used as a temporary measure, is to reduce the high-frequency response of one of the amplifier stages, so reducing the loop gain at the frequency of oscillation to a value below unity. This may conveniently be done by connecting a small capacitor (say 200 pF) in series with a 5,000 ohm resistor from the anode of V1 to chassis.


Linearity.— The linearity of the amplifier is well illustrated by the series of oscillograms. These show that, up to maximum output, the linearity is of a high order, and that the overload characteristic is of the desirable type shown in Fig. 1 (b) in the previous issue.


The improvement due to the application of negative feedback, especially at low frequencies, is clearly demonstrated by the oscillograms.

Equipment for measuring intermodulation products was not available, but measurement of the total harmonic distortion was made with an input frequency of 400 c/s.

The result is shown in Fig. 6, from which it will be seen that the harmonic distortion at maximum rated output (15 watts) is less than 0.1 per cent. Intermodulation, with this degree of linearity, is not present to an audible degree.

Fig. 6. Input/output characteristics and harmonic distortion curves, with and without feedback.


Frequency Response.— The frequency response of the amplifier is greatly dependent upon the characteristics of the output transformer.

In the amplifier tested, the output transformer had a resonance at about 60 k c/s which caused a sharp dip of 2.6 dB around this frequency. The characteristic within the audible range from 10-20,000 c/s is linear within 0.2 dB.

Phase Shift.— The excellence of the frequency response characteristic indicates that little phase shift is present.
Phase shift is only apparent at the extremes of the Audio spectrum and never exceeds a few degrees.

Output Resistance.— The output resistance of the amplifier is 0.5 ohms measured at the 15 ohm output terminals.

Noise Level.— In the amplifier tested the measured noise level was 85 dB below maximum output. The noise in this amplifier was, however, almost entirely 50 c/s hum caused by coupling between the mains and output transformers.

By more careful arrangement of these components it appeared that the noise level could be reduced to better than 100 dB below maximum output.

If desired, the power output of the amplifier may be increased beyond 15 watts by the use of several pairs of output valves in parallel push-pull.

The output transformer, power supply, bias arrangements, and the feedback resistor R25 will require to be modified. Amplifiers of this design with power outputs up to 70 watts have been produced.

Listening tests carried out in conjunction with a wide-range loudspeaker system have fully supported the measured performance. No distortion can be detected, even when the amplifier is reproducing organ music including pedal notes of the 20 c/s order, which reach the threshold of maximum output.

Transients are reproduced with extreme fidelity; tests using a direct microphone circuit with noises such as jingling keys reveal extraordinary realism.

The amplifier can be described as virtually perfect for sound reproducing channels of the highest fidelity. It provides an ideal amplifier for sound-recording purposes, where "distortionless" amplification. and low noise level are of prime importance.


Output Transformer Specification:-

Primary load impedance = 10,000 ohms C.T.

Secondary load impedance = 1.7 ohms per section.

Turns ratio 76: 1

Primary inductance= 100 H (min.)

Leakage inductance = 30 mH (max.)

Winding Data.

Core: 1 ¾"in stack of Pattern No. 28A "Super Silcor" laminations. (Magnetic and Electrical Alloys.)

The winding consists of two identical interleaved coils,. each 1 ½", wide, wound on 1 ¼" X 1 ¾" paxolin formers.

On each former is wound: 5 primary sections each consisting of 5 layers (88 turns per layer) of 30 S.W.G. enamel. copper wire interleaved with 2 mil. paper, alternating with 4 secondary sections, each consisting of 2 layers (29 turns per layer) of 19 S.W.G. enamel. copper wire, interleaved with 2 mil. paper.

Each section is insulated from its neighbours by 3 layers of 5 mil. Empire tape. All connections are brought out on one side of the winding, but the primary sections may be connected in series when winding, only two primary connections per coil being brought out.

Measured Performance.

Primary inductance = 100 H. (measured at 50 c/s with 5V R.M.S. on primary, . equivalent to 2.5 mW)

Leakage inductance = 22 mH. (measured at 1,00 c/s).

Primary resistance = 250 ohms.

The frequency response curve is given in Fig. 7.

Fig. 7 Frequency response (without feedback) of 20 watt output transformer described in the appendix. Generator resistance 2,500 ohm, load resistance 1.7 ohm. Measured with 5V R.M.S. on primary. At higher exitations the bass response improves progressively up to saturation.


Above: Oscillograms of input/output characteristic; Left hand column without feedback; Right hand column with feedback.
(1) at 300 c/s with slight overload (2) at 300 c/s output voltage 15% below maximum (3) and (4) conditions as at (1) and (2) respectively but at 30 c/s.

Below: (a) input waveform, 300 c/s. (b) Output waveform with feedback and slight overload. (c) Output waveform with feedback but output voltage 15% below maximum.
Back to part 1