Topologies, classes and modes

Topologies, classes and modes

Classes

Classes ‘A’ and ‘B’ are well known to anyone who reads Hi-Fi magazines, but these

and the other classes have been deeply and widely misunderstood – even by major

makers who should know better. In this book, class is taken to be a second tier of

identification, since most amplifier topologies and sub-topologies are employable

for most of the classes. Of course, a few circuit topologies are specific to the

amplifier’s class; the class precedes the topology in such cases.

There are at present just five distinct, recognised classes of audio power amplification.

With the approximate first date of commercial application in brackets, they are called:

(1)

(2)

(-)

(3)

(-)

(4)

(5)

(-)



Class A (1917) and variants (>1960), Sliding, Super, and Sustained Plateau, etc.

Class B, A-B and variants, (≅1945) eg. QPP.

Class C is used for RF. It is inapplicable to audio.

Class D – ‘digital’ or PWM (≅1963).

Class E is un-specified for audio.

Class G (1977) and,

Class H (≅1983) are both adding onto classes A, B or their variants.

Class S (sic) has been used to classify a kind of valve biasing unconnected

with modern audio power; and later, unrelatedly, for a control method, which

is erroneously titled. See section 4.10.2.



At heart, all these five class designations describe is fundamentally different kinds

of output stage current behaviour in the controlling devices, labelled very nearly in

the order they were first put to work. Beyond class A (which was not named as such

until there was a class B) the principal raison d’etre of the other classes is an improvement in power conversion efficiency. These will be examined later in this

chapter, beginning at 4.6. For now, it is sufficient to be aware of the conventional

classes A and B.



Modes

This is a tertiary (3rd) layer of meaningful difference to with which to categorise

different kinds of high performance audio power amplifiers. It embraces methods

of control or error-correction, the best known being (Negative) FeedBack (NFB).

There are many variations of NFB alone, many of them inadequately explored, and

beyond, several more ways (such as feedforward) to help an amplifier control itself

and the speakers’ moving parts.



4.1.1 About topologies

In electronics, topologies, giving identity, allow hundreds of seemingly complex

and disparate circuits to be classified into family groups, and the higher relationships grasped better as a result. It is an ad-hoc task, for while botanists are still

cataloguing new plants (and there haven’t been many new models for in the past

10,000 years), electronics designers are busy developing more circuits than there

will ever be plants, without a thought as to classification.



84



4.1 Introduction

In all analogue audio power amplifiers, output stage topologies are at heart about

the way that speakers are connected to power sources, in series with active devices

that modulate the power delivery as an analogue of the music’s pressure waves

(Figure 4.1). There is of course more to the story: the active devices have to be

stably biased to be reliably active; and they have to be fed with a signal; and the

operating conditions must be within ratings of existing parts. It’s also important that

the controlling active devices must give gain, i.e. must be controllable by fewer

amps and volts than they are themselves controlling. And much, much more.

Figure 4.1

Mother of all Power Amplifiers.



Before beginning, a recap of some basic details and also conventions:

o) The words ‘transistor’ and ‘device’ or ‘power device’ are used interchangeably

and as synonyms, where the part being referred to is always some kind of

transistor, that may be BJT or MOSFET, or even IGBT, as appropriate.

i) There is only one polarity or gender for valves; but for transistors, whether

BJTs or MOSFETs, there are two. This hugely increases the possible

permutations. For simplicity, all the initial schematics in this section will show

BJTs. Historically, most did use BJTs, but all could use MOSFETs. Figure 4.2 is

a reminder of how the two BJT genders, npn and pnp, are shown and

distinguished. Note with pnp the arrow points inwards – ‘introverted’ is a possible

mnemonic.

Then in the topological drawings that follow:

ii) All circuitry is simplified to help counter ‘not being able to see the wood for the

trees’. Makers and designers are naturally tempted into hiding or gilding those

commonplace building blocks they may have used. Their schematics can

sometimes be hard or impossible for any outsider, even a master topologist, to

readily comprehend as a result. Here, nearly every drawing is uniformly

presented. ‘North’ (top’) is always more positive where feasible. Coupling

transformers and capacitors are shown where they are essential to the discussion.

Biasing and local feedback resistors ( Re where shown) are largely omitted in

early drawings, but are shown with increasing frequency later as the eye

accommodates the other details.

In practice, while not shown, negative feedback is almost always in use, either

just locally, in the manner of the just discussed as degeneration resistors; and in

many cases also connected globally, in a loop. The potential connections of this

latter form are arrowed ‘NFB’. The final section beginning 4.10 deals with

modes of control including feedback.



85



Topologies, classes and modes

Figure 4.2

OPS Topologies, Symbols Used.

Bipolar transistors



iii) Power sources are shown for simplicity as batteries. Figure 4.2 shows the symbol.

Where the batteries are not connected to the same node as the ground, a floating

supply is implied. The batteries may be taken also to represent the supply

reservoir capacitors. The series impedance (‘through’) the battery (or capacitor)

is very low at audio frequencies. In other words, for music signals, the power

supply source is transparent, a short circuit. The indicated voltage range is a

constant reminder of the range of real needs – with the notionally maximum

300v single supply offering just above 100v rms or about 700 watts into 15

ohms. Where two (‘split’) rails are shown, the total voltage required does not

increase; it is simply a translation, and the voltages are halved accordingly as a

reminder of this.

iv) Ideally, but not always, the grounded point will be the common point between

(a) the input, (b) the output in most cases; and (c) not necessarily, between

output and one or more power supplies. It is useful for makers, practical users

and repairers, to have the ground node the same as one or other of the output

transistors’ main terminals, depending which terminal (Collector or Emitter; or Drain

or Source) is connected to the case. This point is never mentioned in textbooks.

v) The input terminal is called ‘Vin’ (Input,Voltage). Ideally it is referred to 0v

(ground) and if so, 0v appears as the open-ended line under Vin. Where an input

connection must be floating, two pins are shown with ‘Vin’ between them. In a

few topologies, two floating inputs are required, hence there are four input pins.

vi) All topologies are shown ‘positive side up’ (+ve towards the top of the diagram)

so as not to faze the reader. This is at least consistent, as it follows on from

valves. This is not the case in many 60’s circuits, including the handful abstracted

later in this chapter for their contemporary flavour. Do not be alarmed if these

seem to be inside out or upside down. In the days when all transistors were pnp,

it was evidently more helpful for readers to see emitter legs placed on the ‘low

side’, than have positive ‘above’ – a sort of topological typesetting convention.



86



4.2 Germanium and early junctions

Subsidiary to the topologies, there are many so called ‘building blocks’ in electronics. A number of these are central to the development or workable existence, of the

more sophisticated topologies. In the coming sections, they are described as subtopologies, and are introduced in their approximate historic order of appearance in

audio power amplifiers. There is even a place later for a higher-level super-topology.

There are several reasons for logging topologies. First, it has not been done before,

and in the vacuum, there has been a great deal of haphazard topological information, much of it incorrect. There even appear to be some patents granted in the USA

(where else ?!) on topologies that are not new at all, but mere re-arrangements with

a new, often fancy name. Second, by discussing topologies in a light and readable

way, amplifier users who are not circuit designers or topologists will be able to

recognise what kinds of circuits they are really listening too (by comparing the

maker’s schematics with the simplified schematics in this book), and communicate

with others informedly. Third, a solid, coherent comprehension of the significance

of topological features, and why high performance power amplifiers are like they

are, would likely prove impossible without imposing some structure on the vast

diversity. Without it, there could only be an ad-hoc ramble.



4.2 Germanium and early junctions

The first transistors were point-contact. They were not for audio, or much else, but

they did establish the potential [1]. The first junction transistors were npn, but later,

most were pnp. So the tables were reversed (from what is familiar today), with npn

(or n-channel) being the gender limited in availability, expensive, and so preferably

avoided in design.

All early transistors principally employed germanium. They had high temperature

sensitivity, readily developing potentially fatal leakage currents (Iceo). Their absolute maximum junction temperature was 100°c – which is far less ample than it

sounds. Even if the signal did not heat up the junction enough to cause failure,

thermal runaway could. Runaway could be triggered by marginally poor ventilation, or a bias voltage that did not reduce with increasing temperature to compensate for slightly hard-driven transistor junctions.

The early germanium junction power transistors also had very low fTs (transition

frequencies), so low they were well within the audio band, e.g. 7kHz. They should

never have been used for audio, at least not with global feedback, nor for full-range use.



4.2.1 Out of the vacuum-state

The first transistor audio power stages followed existing valve topologies. Yet these

topologies were developed because of the limitations of valves, namely (i) no opposite

polarity complement devices, and (ii), limited heater-to-cathode voltage differences,

necessitating many floating heater supplies if the topology was at all adventurous.



87



Topologies, classes and modes

Figure 4.3

Topology 1: Single-ended, Transformer coupled.



The simplest transistor power amplifier output stage is single ended (Figure 4.3).

Here, a single transistor is driven by the incoming signal, and its collector is loaded

by the speaker, reflected back through the transformer winding. The impedance

‘step-down’ arrangement lowers the voltage swing across the speaker but did enable a relatively low current-rated transistor to drive the speaker more ‘stiffly’ more

than it otherwise could. This stage can operate in class A only (see section 4.6.1)

and is not very efficient – except by the standards of class A amplifiers. The transformer output coupling is needed; otherwise the speaker would have to pass the

bias current (DC), which in class A, is large. But the transformer’s primary still has

to pass a DC bias current. Though smaller than it would otherwise be, this saturates

the core, degrading the transformers’ performance. We are left with an amplifier

requiring an expensive, weighty, oversized core for any chance of fidelity, and which

is practically suited only for low powers, below a few watts. Worse, this transformer alone severely limits the amount of overall (‘global’) NFB that can be applied, without ringing or oscillation at both high and low frequencies.

Biasing for topologies like the one pictured in Figure 4.3 was critical when there

were only germanium transistors. It relied on an ntc thermistor, ideally clamped to

the power transistor, or its heatsink. The resistance of the thermistor would progressively reduce with increasing temperature, so reducing the bias voltage or current,

or even the gain.



4.2.2 Push-pull, Transformer-coupled

With valves, power stage designers had long ago overcome the inefficiency of singleended class A, going as far as class B, A-B and related schemes (see 4.6 et seq.).

Valves were arranged in push-pull, i.e. they were driven and delivering differentially,

or in ‘phase opposition’. This ideally removes any DC biasing of the transformer core.

Push-pull operation, essential for class B, could be used for, and benefit, class A

operation. In all these schemes, transformers were not just used for impedance matching, but also to produce and combine currents and/or voltages in phase opposition.

Without them, symmetrical push-pull would have been problematic, without complement devices. Push-pull action also ideally cancels those even harmonics that are

self-generated. So it reduces distortion – but doubly selectively, i.e. (i) only if even

harmonic and (ii) only if self-generated.

Figure 4.4 shows a double transformer coupled scheme, a direct translation or descendent of classic valve topology. The input transformer is ultimately no good for

accurate audio, but it simplified the circuitry in low-performance amplifiers, and to



88



4.2 Germanium and early junctions

Figure 4.4

Topology 2: Push-Pull Transformer-coupled, CE type.



those simple-minded enough to believe that the map is the territory. In reality, making a good input (interstage) coupling transformer is no easy task. Operation could

be class A or B. But in class B, the output transformer gave problems, as the stored

magnetic energy was left to ring each time one side cut-off. The ringing caused

voltage spikes that could cause the transistors to experience C-E voltages that were

much more than the rail voltage, and their own ratings. Death was instant and there

were no warnings. An unexpected ‘free lunch’, due to the maths of dissipation, or

really, the high efficiency of a push-pull transistor output-stage operated in class B,

is that transistors rated at 10 watts each, could deliver 49 watts into a (perfectly

resistive) speaker before exceeding their ratings! This was later exploited in Sinclair’s

infamous Z30 DIY hi-hi module and relatives (1969), which pushed cheap, low

power driver transistors to their limit.

Figure 4.5

Topology 3: Push Pull TX Coupled, CC type.



Figure 4.4 employs the output BJTs in the common emitter (CE) configuration. This

offers voltage gain, but lower current gain. The transformer’s step-down ratio would

have compensated for this, offering one excuse for employing a transformer. Figure

4.5 shows the alternative emitter follower (alias common collector) configuration.

This configuration is possible with valves, but wasn’t uppermost, as it stirs up trouble

with heater-to-cathode voltage limitations, or else demands the expense of multiple

floating heater supplies. In this configuration, voltage gain is just below unity (x0.95

typically), but current gain equals beta. Alas, beta is very variable with individual

BJTs and with junction temperature.



4.2.3 Sub-topology: the Darlington

Early on, the cascade connection of two BJTs to achieve very high current gain,

was discovered and patented at Bell Labs, birthplace of the BJT (see Appendix 1).



89



Topologies, classes and modes

Figure 4.6

Sub-topology 1: The Darlington



The particular connection, shown in Figure 4.6, has long been known as the

Darlington, after its inventor. The two transistors may be identical, but for most

utility, the left hand (input) device may (and often need only) be a ‘smaller signal’

part with a lower current rating (but no less high a voltage rating), while the second

or ‘final stage’ device would be rated commensurate with the maximum current. To

suggest the likely difference in maximum current and power handling capacity, the

transistor symbols are sized accordingly on this occasion.

In early transistor amplifiers, where inadequate current gain (beta) was quite a problem, the Darlington reduced distortion, and also production variability, caused by

the naturally wide spread of beta in early and any unselected BJTs. In modern high

performance power amplifiers, it is still used and is often still made (like the original) from two discrete transistors, despite the existence of many Darlington transistors, which are monolithic ICs of a power and driver transistor, in the Darlington

connection. These types are not much favoured, as one or other of the constituent

BJTs may be sub-optimal for the designer’s requirements; may be unsuited to audio, inadequately rated, bad sounding, or they may have the wrong value and quality of flushout resistor connected internally. The flushout resistor is essential to

ensure the second, larger BJT turns off. Other resistors and diodes may be required

to ensure clean, benign operation under all conditions.

There are other forms of two-transistor marriage able to create a high gain

supertransistor – all loosely termed ‘Darlington’. One, often called ‘the compound

pair’ but far better named after its Americanized-Japanese inventor, Sziklai [2], is

particularly effective [3,4] (Figure 4.7] as it has self-feedback.

Figure 4.7

The Sziklai (Compound Pair).



90



4.2 Germanium and early junctions



4.2.4 Transformerless push-pull (transistor OTL)

Elimination of the output transformer was a major step forward in high performance audio. This had long been recognised, since the 1930s, when designers first

struggled to make output-transformerless (OTL) valve amplifiers [5]. Valves have

the difficulty that they are intrinsically high voltage, low current devices. The amount

of current available is out-of-step with the requirements of ordinary electrodynamic

speakers of nominally 16Ω and below. The output transformer was a logical and

theoretically elegant way to compensate for this. The reality is very different though.

Even today, with computer design, good power stage output transformers that do

not wreck measured and/or sonic performance are complex and expensive, both to

design and manufacture. They can account for as much of an amplifier’s weight and

size as the usual 50/60Hz power supply transformer.

Figure 4.8 shows prototypal OTL transistor topologies. The small differences in

where the ground is placed, and thus how the two, floating push-pull (differential)

input signals are connected, has a large effect on the behaviour – with or without

global NFB. The first configuration (a) is half-common emitter, which has moderate voltage and current gain, and also a moderate output impedance; while the other

(b) is half-emitter follower (common collector), with unity voltage gain, but potentially high current gain, and a low output impedance. In both cases, the transistors

might be Darlingtons. On the other side of the coin Figure 4.8 (a) might require a

substantial current from the preceding driver stage, while (b) will in all cases require an input signal voltage slightly larger than the anticipated output swing.

Figure 4.8

Topology 4: OTL Types.



CE



CC



CC



CE



4.2.5 Sub-topology: diode biasing

Biasing is essential, but omitted in the surrounding topological discussions for simplicity. A snag in the early, elementary class B (and A-B) transistor power amplifier

topologies where capacitor input coupling replaced the interstage input coupling

transformer, was that music’s asymmetry caused a DC voltage to build up, as the

output or drive transistors’ junctions rectified the signal [99]. Extra parts, typically

a BE reverse-connected, parallel diode with a series resistor, were needed to prevent this. Long after coupling capacitors were abandoned in output drive stages, the

same diode connection has since reappeared in cautious designs to protect BJTs;

once again to do with charge control under overload.



91



Topologies, classes and modes

Jones and Hilburn demonstrated the uses of diodes as biasing elements in the late

50s. They used a silicon diode followed by a voltage divider, to attain the optimum

bias voltage[99]. Later on, series strings of two or more diodes were used, sometimes with ad-hoc shunt resistors. The inability of a small, glass-encapsulated junction to follow the junction temperatures of the power devices, alias ‘thermal lag’,

was recognised in 1957, with a paper that described a diode bonded to the base of a

power device [6]. Forty years on, there still aren’t more than a few very specialised

power MOSFETs, and no BJTs, with integrated junction temperature monitoring.

The world has more or less given up waiting for semiconductor makers to launch a

range of power devices incorporating this most elementary engineering improvement.



4.2.6 Complementary push-pull OTL

Beginning in the early-to-mid 60s, the arrival of complementary (matching npn and

pnp sets of) power transistors overcame the need for floating push-pull (differential) drive. Even if two drive signals were still required, at least they could now be

referred to a common point, eliminating the argument for an interstage transformer.

Figure 4.9 shows the all-emitter follower (alias Common Collector, CC) connection with single (a) and dual (b) supplies. Note how symmetry is lost in (a). Around

1962, while the most practical OTL topologies were still being worked out, there

was even a proposal to retain symmetry and still have only one power source, by

using a special, 3-wire speaker with a centre-tapped voice coil, but this didn’t help

speaker efficiency, nor did it catch on [7].

Figure 4.9

Topology 5: Complementary

OTL, all-CC Type.



Figure 4.10 shows the all-common-emitter (CE) type, with a single supply (a) and

with dual supplies (b, c) with different grounding arrangements. With the superficial similarities, confusion is easy. To distinguish between CE and CC (follower)

stages, look to see if the speaker is (or would be) connected between the emitters of

the output transistors and the input signal reference (ground). If so, the stage is a

follower (CC); if not, it is CE.

Figure 4.11 is an example of how the basic follower-connected BJTs could be replaced by compound devices, either Sziklais (a) [8] or Darlingtons (b).

Figure 4.12 shows a complementary pair of Sziklai connected in CE format, the

natural development of Figure 4.10 (c). With the Sziklai, don’t forget that the effective emitter terminal is the actual collector terminal of the final BJT. In this circuit,



92



4.2 Germanium and early junctions

some of the real world elaborations have crept back in: R1a,b are flushouts. Re1,

Re2 provide local feedback, ideally for thermal stability. Other descriptions are

possible: Cherry calls the configuration ‘Push-Pull folded totem pole’, while Roddam

describes it as comprising ‘Cascaded complementary pairs’.

Figure 4.10

Topology 6: Complementary OTL, CE Type.



Figure 4.11

Topology 7: Complementary OTL, Compound OPS variants, CC type.



Figure 4.12

Topology 8: Complementary OTL, Compound CE type.

E = effective emitter!

E



E



93



Topologies, classes and modes



4.1.8 Quasi complementarity: the faked match

In the event, many makers eschewed these perfectly symmetrical schemes, preferring for example Figure 4.13, a possible and once popular arrangement of the infamous quasi-complementary scheme. The latter was developed as a smart solution

to the lack of good (or any) pnp complements, in the early days of silicon transistors. Those pnp opposites that existed usually had a lower ‘speed’, ie. transition

frequency (fT), and they were not so rugged as their npn sisters. The outcome was

(i) asymmetric (dis-complementary) performance at high frequencies, and respectively, (ii) sudden silence – except for a loud buzzing.

Figure 4.13

Topology 9: Quasi Complementary

Compound, CC type.



This time, the preceding, CE driver stage (TR.1) is shown. TR.2,4,6 act straightforwardly as a giant Darlington, or triple (section 4.3.4). For the lower half however,

TR.3, the only pnp transistor in the circuit, acts as a so-called ‘complementary phase

splitter’. Really, it is more a level shifter, and certainly the first part of some kind of

three-stage Sziklai (TR3,5,7), which may be seen also as a conventional Sziklai

(TR3,5) combining with a final follower (TR.7). Keynote reasons for using circuits

such as this one (Ca. 1960 – 65) were fourfold, three of them interwoven. First, all

power transistors were at the time high cost, metal-canned items – there was no

plastic packaging. Second, the saving in not having to buy the pnp type, and third,

instead being able to buy twice as many of the cheaper npn type, added up to a

tempting increase in profitability. Also, far more so than today (1996), silicon power

transistors of any kind, let alone pnp complements, were only available in a limited

range of voltages and currents, particularly for the output stage.



4.1.9 Sub-topology: paralleling

Output stage transistors devices have long been paralleled to increase current handling. As brute high-current handling does not square altogether with speed or high

fT, parallel connection is essential beyond a point. Without it, there would still be no

high performance, rugged audio power amplifiers with output power capabilities

much above 100 watts. Paralleling may even be beneficial when not essential, since

‘n’ smaller ‘lighter’ parts will likely have a faster response time, and wider bandwidth. The outcome can be a species of supertransistor with a heightened bandwidth x current-handling product. Figure 4.14(c) shows the symbol invented by the

author to denote one such, the ‘Triplington’.



94