Compact power amplifier designs
A selection of power amplifiers (classes A, AB and B) in approximate chronological order that, in their day, received favourable comment.
Very early transistor amplifier applications showed a reliance on previous valve design configurations utilising transformers for stage coupling, from the single-ended approach used in Kochs' Regency TR-1
to a push-pull design utilising transformers for stage coupling, phase-splitting and no feedback. Most notable, for its' time, are the low voltage supplies used.
The next design used capacitor coupling to avoid the use of transformers. Output was some 50mW.
Confusion could arise with the new technology. The original of the above diagram has the supply connections reversed.
H.C. Lins' quasi-complementary design (Electronics, Sept 1956, pp. 173-75) as mentioned in JLHs' Evolutionary Audio' (E&WW, Nov 1989) "...which was designed around germanium transistors, gave misleading encouragement to other engineers, on a world-wide basis, who translated the design into silicon-transistor-based versions when, during the early 1960s, n-p-n silicon planar power transistors became available. The inherent snag in this approach is that the base voltage/collector current characteristics of germanium and silicon transistors are different, with that of the silicon device being much more abrupt...".
Tobey and Dinsdale produced a transformerless design in 1961 that was improved in the Dinsdale Mk II.
Linsley Hood broke ground with his simple class A designs (1969, 1996).
A simple design, variations of which could appear in all sorts of portable and mains-driven applications, typical of commercial post-valve amplifiers.
Basic and 'inadequate' by todays' standards, these were capable of satisfying the majority of needs when feeding speakers of a quality build.
A low-cost arrangement that negates the use of a mains transformer. Reminiscent of single-ended valve designs, this type of circuit could still be found in television receivers well into the 1990s. Insulation precautions necessary since circuit is at mains potential.
A push-pull variant for a colour TV.
Wherever possible, it makes sense for a manufacturer to settle on a 'standard' design known to function satisfactorily, thus saving costs. Modules, such as the Mullard LP1162, appeared in many commercial designs whether audio or TV. A faulty amplifier could then be easily replaced with little or no diagnosis. Capable of about 4W, pin-outs allowed tailoring of gain and frequency response.
A more sophisticated commercial class A design from the same manufacturer with a consumption of 42.24W when quiescent. This used resistive loading to cope with short-circuited loads. With low efficiencies transformers increase in size and cost. Heatsinks needed for drivers and output transistors with a minimum 3.3°C/W for the latter.
Of similar vintage for comparison, an Elektor design, Edwin, with no quiescent current in the output devices, the 0R12 resistors can be made from 8 x 1ohm resistors in parallel, mounted on end. The RC network on the class A drivers' emitter helped to de-couple the supply ripple. Active current limiting protected the output. An additional feedback path helped reduce limitations presented by the output electrolytic and was used by a variety of manufacturers like Armstrong, Leak, Trio, Kenwood, etc.
This design was useful in multi-channel systems. Reliable and low-cost, its' survivability in respect of testing 'innovative' output wiring arrangements unscathed proving useful on more than one occasion. Very few, if any, could differentiate between the two amplifiers. Versions using a higher supply voltage (58V), complementary output pair (MJ2955/3055), 3m5F output cap and no output current-limiting have been seen. These use BD139/140s as drivers, the base drive for the MJ2955 being taken from the emitter of the PNP driver (BY127 omitted).
Power bandwidths as high as these can be an indicator of good performance but in practice are best limited to say 50kHz. A high voltage 3n3F capacitor across the 1k feedback resistor could suit and consideration be given to input filtering (47pF or more across i/p 100k resistor). Supersonic artefacts may force unwanted, unnecessary, potentially damaging energy into the output and load adding to distortion.
A later design, the Equin (April-May '76), offered uprating and is recommended reading for this era. Construction detail is good as is the accommodation for differing component types (PDFs available).
Crossover distortion is given a priority in the May article. A preamp intended to match, Preco, was described in both issues.
The 600 series was Armstrongs' first amplifier to use silicon transistors throughout giving an improvement in both power and reliability. 'This had an unusual double-feedback arrangement which was employed to correct some of the effects of the (larger than usual) output capacitor (single-rail supply so requiring an output capacitor). However it meant that over some of the output frequency range the amplifiers' output impedance went slightly negative. This meant that the system maintained a high damping factor in use even when the cable resistance was taken into account. As a result the 600 amplifiers were held to give unusually tight control of the loudspeaker movement'.
Typical specs gave a rated power of 40W per channel into 8 Ohms, crosstalk <-45dB, frequency response 20Hz - 25kHz ±1dB, s/n -65dB and THD of <0·08% at 25W and <0·08% at 1W. The RCA 40636 output devices can be seen as 95V 2N3055s.
A concise arrangement (1973), using Darlingtons to reduce component count, that saw modification is given below. Virtually identical forms used in other models of that year, eg; Beocenter 1400, Beomaster 800/901, and later models. Note the resemblance to the Mullard module shown above.
Intended as a matching amplifier for a tuner, with a little tweaking, uprating was possible for a remote-controlled quadrophonic tuner amp (1975) that would have been matched with the ground-breaking Beogram 6000 tangential record player (exploded views of which can be found here).
This could have been said to be a very stylish system with a distinct lack of affordability. Most users seemed to be notably enthusiasts and the 'discerning' who did not disapprove of the sonic performance.
A very similar circuit appeared in the 1979 Beomaster 1700 (type 1701), utilising a soft-start to prevent switch-on thumps caused by the output electrolytic charging through the load with an additional transistor, despite the appearance of a more sophisticated design with differential inputs and current-limited DC coupled Darlington outputs, that saw service with only a little modification in a number of models (see the Beomaster 4400). In this type the output electrolytic is included in an additional feedback path.
Similar topologies and component counts dominated audio design, output stages not being short-circuit proof, nor free from distortion. Lowest quality builds were reserved for cheap stereos and guitarists (as witnessed by the tone controls found in some instruments).
AC output decoupling became unfashionable (although occasionally reappearing) notably through the possibility of speakers being blown as the output cap charged through the load on switch-on, the poor service life offered by electrolytic capacitors and the reduced bass response when driving low impedance loads. Some designs, like that used as monitors in the Revox A77 tape deck, could surprise in that, in this one, output devices with an fT of 40MHz were used.
To balance a DC system, differential input pairs came into use. These helped to reduce noise and feedback, also the input impedance could then be set with a single resistor.
A good collection of slightly earlier robust bipolar designs appears in the RCA Solid State '74 Databook Series; Power Transistors and Power Hybrid Circuits (SSD-204B), whose designs have inspired others, adaptations being found in many brands of combos, hi-fis, etc. A sturdy, and inexpensive, 70W version (to suit a 'youth' environment) appears on page 600 and a more sophisticated 120W version on page 618. These designs feature excellent protection and quite conservative distortion specs, however, as with all data, read the circuit diagram carefully first. The main negative smoothing capacitor for one design (on page 611) is shown reversed. Wiring as per diagram would probably result in a loud bang and not much else. The two 40W (TO220) designs are shown below, the faster fully complementary version giving considerably better high frequency performance.
An attractively simple solution, the non-inverting input transistor of a differential pair directly drives the fast class A amplifier that drives the bias chain. The attention given to protection is noteworthy. At the same time, for high-power use or ruggedness, it was acknowledged that output pairs should be paralleled, a point reiterated later by one of the co-authors when joining another company.
Manufacturers of upright organs, like the Italian Boosey and Hawkes' Cavendish series, could include two or more power amplifiers in a design, driving a variety of speakers, many of which behaved faultlessly for years.
This amplifier (30W/8R) is a close match to the '74 RCA 70W version mentioned above. Typical of commercial adaptations of this design, the overcurrent protection circuit was omitted. Note the high gain, 181 in this case, which necessitates the use of extra compensation capacitors (input filter -3dB @ <0.6MHz). In a hi-fi situation, a gain of 20 might be sufficient, vastly improving high frequency performance and reducing the noise floor considerably. Swapping the 100R resistor in the feedback for a 1k gives an entirely different beast, and bass, versions of which were used to test perception.
To protect the output stage from the back EMF inherent in reactive loads, diodes clamping the output to the supply rails and the supply rails to ground, are usually included in considerate designs. Most Darlington transistors, as with most FETs, include an integral diode.
In order to resolve offset issues opamps were seen as a solution.
A neat little design for very compact builds from the same databook.
The thermal compensation (diode + R) network intended for 'consumer service' can be replaced with a more stable amplified diode (Tr + 2Rs) as required.
The 'low' output voltages then available led to the common use of 'compound' output arrangements, first used by the author in the excellent (but AC coupled) Practical Electronics Gemini ('70/1) to reduce bias currents under heavy drive conditions compared to conventional Darlington pairs.
R. Manns' famous and 'skimpily screened' Practical Wireless Texan (based on Derek Skinners' B80 class B design) was copied extensively and is an easy layout to tinker with. Gibbs' and Shaws' later Orion is worth a look also for the discrete low-level approach it offered with a strikingly similar layout and transformer.
Later expressions of the same theme, intended as bass drivers.
An earlier more complex approach.
Offset-nulling can be added and flat-style LEDs can be bonded to the current source transistors to optimise thermal tracking. Bolt bias transistor to heat-sink, between the drivers and note earlier comments concerning the RC network between the opamps' -ve input and output.
Maplin, now defunct, produced some good designs.
There is a lot of interest in single-ended FET amplifiers using a single output device and these have been demonstrated by Aren van Waarde and Mark Houston ("Class-A 2SK1058 MOSFET Amplifier") and can be useful as HF drivers. Layout can be important to prevent oscillation and care must be taken given the high frequency responses these amplifiers are capable of. Output power is invariably limited and quiescent dissipations (class A) can be high. Nevertheless, a great deal of satisfaction can be derived from a low component count.
Some Maplin, symmetrical and Class D designs.
Contact me at firstname.lastname@example.org
especially if you want additional content to this page
or if you find any links that don't work. Don't forget
to add the page title or URL. Take care!
Back to index, sound, tips or home.