|
In the final part of this series, we conclude last =
month=92s=20
discussions on practical audio amplifier systems with =
the=20
final link in the chain between pick-up and =
loudspeaker - the=20
power output stage.
The previous amplifier stages have been concerned =
with=20
producing a faithful replica of the input signal, but =
at a=20
larger voltage amplitude. The function of the output =
stage is=20
rather different and certainly more demanding. Because =
it=20
requires work to drive the speech coil of a =
loudspeaker, and=20
because work equates to power when we introduce the =
time=20
element, the output stage must produce large =
excursions of=20
current as well as voltage - and it must do so with =
the=20
minimum amount of distortion being introduced. This =
means that=20
the output valves must be capable of passing large =
current=20
flows, which implies a heavy emission of electrons =
from the=20
cathode surface in the first place. There are two =
factors that=20
limit the amount of current that a valve can carry =
They=20
are:
- the surface area of the cathode capable of =
emitting=20
electrons, coupled with the power input to the =
heater=20
- the amount of heat that can be dissipated at the =
anode=20
without causing degradation of the valve=92s =
performance, eg,=20
the emission of gases from the surface, which would=20
seriously impair the operation of the valve by =
causing=20
collisions between the gas molecules and the =
electrons in=20
transit; excessive heat could also cause failure of =
the=20
anode structure.
It is obvious from the above points that valves =
intended=20
for high power applications (using the word =91high=92 =
in a=20
relative sense, since one man=92s watt is another=92s =
kilowatt!)=20
must have larger electrodes than those intended just =
for=20
voltage amplification.
Triodes versus Pentodes
These two types are obvious contenders for the role =
of=20
output valve; for the moment we will lump beam =
tetrodes in=20
with pentodes. What we need to consider is how well =
each type=20
of valve will be able to handle the task, given the =
rather=20
stringent requirements that apply Let us consider =
triodes=20
first.
lf the characteristics were absolutely linear and =
parallel,=20
no amplitude distortion at all would result. Both =
half-cycles=20
of the signal input would produce equal swings at the =
output,=20
each being a faithful copy of the input.
The mutual (grid) characteristics for triodes tend =
to be=20
both linear and parallel, which means that swings of =
the=20
signal voltage along these characteristics should not =
cause=20
too much distortion in the output (we won=92t, at the =
moment,=20
quantify how much distortion is tolerable), so this is =
a plus=20
point for the triode. On the other side of the coin, =
however=20
is the low gain of the triode. This means that, for a =
given=20
output signal amplitude, the input signal has to be =
fairly=20
large if the stage is to be driven fully At the same =
time the=20
triode also takes a great deal of current from the HT =
supply,=20
so is not very economical in terms of power =
consumption (when=20
compared with its amplifying capability).
The type of amplitude distortion that the triode =
does=20
produce is largely second harmonic. Thus, an input =
signal at,=20
say 400Hz would have, after amplification, a =
significant=20
amount of an 800Hz component also. This fact is, as we =
shall=20
see later not as serious as it might at first appear, =
since it=20
is relatively easy to reduce this type of distortion =
to=20
acceptable proportions.
Looking at the pentode now, and bearing in mind =
some of the=20
deficits listed above for the triode, one advantage =
that we=20
can call to mind immediately is going to be its higher =
voltage=20
gain, allowing the use of a smaller input signal in =
order to=20
obtain a given output. Also, it takes less HT current, =
and so=20
is more economical with supply power than the triode. =
However;=20
its mutual characteristics are not as linear or as =
parallel as=20
those of the triode, though it is possible to =
compensate for=20
this deficiency.
As far as harmonic distortion is concerned, the =
dominant=20
harmonic produced here is the third harmonic. =
Eliminating this=20
is not as easy as getting rid of even order harmonics: =
2nd,=20
4th, etc. One point of comparison that can be made =
with=20
triodes, that we shall see is significant, is the =
value of the=20
anode slope resistance, ra. In triodes this is quite =
low,=20
whereas in pentodes it is by comparison generally =
extremely=20
high.
The Triode Output Stage
Any valve amplifier, being an active device that is =
delivering a signal to a load, can be considered in a =
general=20
case to be a generator. Any generator cannot avoid =
having an=20
internal resistance, and the value of this internal =
resistance=20
will determine the optimum value of load impedance =
into which=20
the generator can deliver its maximum power. For =
valves, the=20
internal resistance is the parameter ra, this being =
the=20
reciprocal of the slope of the output characteristics, =
as we=20
have seen in Part=20
One and Part Two, and so is quite clearly the =
output=20
resistance of the valve when feeding a load.
The maximum power transfer theorem states =
that =91a=20
generator delivers maximum power to a load when the =
source=20
resistance of the generator equals the load =
resistance=92.=20
Before anyone reaches for pen and paper to tell me =
that I am=20
wrong, let me just add that this theorem as stated =
applies=20
when the load is resistive. If the load is reactive =
then=20
maximum power is transferred to the load when the load =
impedance has a value that is equal to the conjugate =
of the=20
generator impedance. If the word conjugate isn=92t =
understood,=20
don=92t worry about it; it comes from complex algebra, =
which I=20
have no intention of going into here!
The valve as a generator; (a) actual triode =
circuit=20
(without bias components) and (b) the AC equivalent=20
circuit.
The figure above shows the valve represented as a =
generator=20
Figure (a) shows the triode (it could equally well be =
a=20
pentode) with an input signal Vg, a load RL in the =
anode=20
circuit and the internal resistance ra. The grid bias =
voltage=20
is represented by a battery and since this voltage is =
DC=20
Figure (b) shows the equivalent circuit which can be =
used to=20
analyse the performance of the amplifier This =
comprises three=20
components: a voltage generator μVg, which =
represents the=20
valve action and two resistances: ra, the generator =
internal=20
resistance; and RL, the load resistance. In the =
specific case=20
of the valve, whether triode or pentode (in theory at =
least),=20
the maximum power will be transferred to the load when =
the=20
latter is equal to the ra of the valve. This raises a =
problem=20
immediately.
The load on an audio output stage is the speech =
coil of the=20
loudspeaker it is totally impracticable to wind such a =
speech=20
coil so that it has an impedance anything like that of =
the ra=20
of even a triode valve, which will be of the order of=20
thousands of ohms! A practical value for the speech =
coil=20
impedance is unlikely to be more than a few tens of =
ohms=20
(eight ohms (8Ω) is a common nominal value). From =
this it is=20
quite obvious that it is impracticable, in the case of =
valves,=20
to connect the speech coil directly into the anode =
circuit of=20
the output stage as the load. There are other =
considerations=20
that make this undesirable anyway, but the above =
argument, on=20
the grounds of load and source impedances, should make =
it=20
clear why this course of action cannot be undertaken. =
The Output Transformer
The answer is to use an output transformer with a =
step-down=20
turns ratio. The required ratio =91n=92 can be =
calculated from a=20
simple formula, which is as follows:
n =3D √(load required by valve) / (resistance =
of speech=20
coil)
Notice that the above formula says load required =
by=20
valve and not actually the valve=92s ra. This is =
because the=20
load presented to the valve by the loudspeaker through =
the=20
transformer is not resistive and it has been found in =
practice=20
that the load that the valve needs to see is =
approximately=20
equal to twice the ra of the valve.
As an example, if the triode in question had an ra =
of=20
8,000Ω and the speech coil resistance was =
16Ω (using=20
convenient figures for ease), then twice ra is =
16,000Ω and the=20
tums ratio needed is equal to √1,000, which is =
approximately=20
33:1.
The penalty that one has to pay by making the =
effective=20
triode load equal to twice the valve=92s ra is an =
increase in=20
the drive voltage required for full output.
Design of output transformers for quality audio=20
reproduction requires great care. In particular the =
primary=20
inductance must be as high as possible (necessitating =
a bulky=20
transformer) if it is to offer a constant impedance to =
the=20
valve at all frequencies of interest.
Modern cores are not made of 'ordinary=92 =
electrical steel as=20
may be found in a mains transformer but of a higher =
quality=20
purer stuff with superior magnetic performance; this =
also=20
makes it possible to keep the bulk down to reasonable=20
proportions, certainly modern examples are a good deal =
smaller=20
than their early equivalents.
Other factors such as self-capacitance and leakage=20
inductance must also be optimised if reproduction at =
the high=20
frequency end of the spectrum is not to suffer. This =
is=20
usually achieved by not winding the former in the =
=91normal=92 way=20
like you would with a mains transformer Instead, both =
primary=20
and secondary windings are split up into sections and=20
interleaved with each other: some of these =
arrangements can be=20
quite complex and, to go really =91over the top=92, =
one layout=20
incorporates a split bobbin, where the order of the =
layers on=20
one half are reversed on the other half. Add to this a =
choice=20
of secondary taps for different speaker impedances, =
and it can=20
be appreciated that manufacture can be extremely =
labour=20
intensive, and, hence, valve output transformers can =
cost a=20
small fortune if you want genuine Hi-Fi quality. For =
more=20
information on the subject, see the small book Coil =
Design=20
and Construction Manual by B B Babani (available =
from=20
Maplin, Code RH53H, Price =A32.50) which came out as a =
first=20
edition in 1960, and gives detailed guidelines on how =
to make=20
your own. We wouldn=92t recommend designing and making =
your own=20
output transformer however: there are, fortunately, =
still a=20
few competent manufacturers around.
Tetrodes and Pentodes
In contrast to the triode, both of these types of =
valve=20
have extremely high values of ra that require careful =
matching=20
to the loudspeaker impedance. The rule regarding =
triodes does=20
not work in these cases, and it has been found that =
these=20
valves work best when they see an effective anode load =
that is=20
between one-third and one-sixth of their ra value. =
Care is=20
required in the design of output stages using such =
valves=20
since, otherwise, they can generate an excessive =
amount of=20
distortion. In general, the safest technique is to =
choose the=20
load value which is recommended by the maker of the =
valve.
Although the beam tetrode may often be considered =
as the=20
equivalent to a pentode (especially since they are =
alter-=20
native answers to the same problem of the =
inter-electrode=20
capacitance Cag), there are significant differences =
due to=20
their different modes of operation. One of these =
differences=20
is the type of amplitude distortion that each =
introduces into=20
the amplified signal.
In the case of pentodes, the distortion is =
principally=20
third harmonic with only a little second harmonic; =
this is the=20
opposite to the beam tetrode, where mostly second =
harmonic=20
distortion is produced. Thus, in terms of harmonic =
distortion,=20
beam tetrodes behave more like triodes.
Push-Pull Amplification
There are several ways of approaching the design of =
valve=20
power output stages:
- Single-ended, where the output transformer is in =
the=20
anode circuit.=20
- Parallel operation, where two or more valves are =
connected in parallel in order to boost the power =
output,=20
the output transformer being again in the common =
anode=20
circuit.=20
- Push-pull operation where two valves (or sets of =
paralleled valves) are driven alternately by the =
input=20
signal, each valve having half of the primary =
winding of the=20
output transformer in series with its anode, the =
centre tap=20
of this transformer being connected to the HT supply =
positive line.
We can dismiss both (a) and (b) immediately on the =
grounds=20
that, since the primary winding carries the full anode =
current=20
in one direction only its core would have to be =
excessively=20
large or specially constructed in order to avoid =
saturation=20
arising from the quiescent DC. Again the Coil Design =
and=20
Construction Manual shows how this is done; the common =
arrangement is where the magnetic circuit is =
effectively=20
broken by introducing a thin layer of waxed paper or =
similar=20
to separate the core into independent stacks of =
=91E=92 and =91I=92=20
sections. The gap reduces the core=92s sensitivity to =
DC, yet=20
allows alternating magnetic lines of force to pass =
through. LF=20
response is critically linked to the choice of gap =
spacing, so=20
tolerance must be tight. Many consumer quality =
radio-grams and=20
record players et al used this type of output =
stage,=20
but it is not worthy of serious consideration, =
especially as=20
the type of harmonic distortion introduced by the =
output=20
valve(s) cannot be reduced by this type of =
connection.
However in the case of (c), push-pull operation, =
the DC=20
anode currents flow in equal and opposite =
directions in=20
the half-primary windings, and so their fluxes cancel =
out.=20
Saturation is thus avoided, even with relatively small =
cores,=20
the core size now being dictated by the consideration =
given=20
above, of providing a constant load at all =
frequencies. The=20
second advantage that arises from the push-pull =
connection is=20
the cancellation of all even order harmonics. The =
image below=20
shows the arrangement of one possible type of =
push-pull output=20
stage.
A Push-Pull output stage using triodes.
Valve V1 is the driver stage and valves V2 and V3 =
form the=20
push-pull output stage. The phase-splitting action is=20
performed here by the use of a driver transformer T1, =
in which=20
the secondary winding is centre-tapped to 0V and so =
provides a=20
pair of equal, anti-phase voltages to the grids of the =
output=20
valves as well as DC bias. With no signal input to the =
driver=20
stage, both output valves draw only their quiescent =
current,=20
and this flows in opposite directions in the =
half-primaries of=20
the output transformer T2.
When a signal is being amplified by V1, both ends =
of the=20
driver transformer T1 are at opposite potentials with =
respect=20
to the centre-tap, the latter being connected to the =
cathodes=20
of both output valves via the cathode bias components =
R1 and=20
C1. Thus, as the grid of one output valve is being =
driven in=20
one direction, say positively, the grid of the other =
output=20
valve will be driven in the opposite direction, in =
this case=20
negatively. The terms positively and =
negatively=20
are here used in a relative sense, since whether the =
signal=20
voltages are actually positive or negative with =
respect to 0V=20
will depend upon the way in which the output valves =
are=20
biased. The classes of bias possible with output =
valves are=20
illustrated by the mutual characteristic shown =
below.
Classes of bias.
Classes of Bias
These are more commonly used to describe 'classes =
of=20
amplifiers'. Taking Class A first of all, this is the =
bias=20
method commonly employed with most single-ended =
amplifiers, ie=20
not wired in push-pull. The valve is biased to the =
mid-point=20
of the characteristic such that the quiescent anode =
current is=20
large, and the signal swings cause equal changes in =
this=20
current for both half-cycles. Distortion is minimised, =
but=20
efficiency is low because of the high DC power input =
required=20
to obtain a given AC power output. The available AC =
power=20
output is also reduced, because the valve is =
contributing to=20
both positive and negative half-cycles of the output =
signal,=20
and having to do it within its available total signal=20
excursions.
If, instead, the valve is biased to the point of =
projected=20
cut-off, the valve is then operating in Class B. The =
quiescent=20
anode current is extremely low giving very high =
efficiency but=20
the amount of distortion introduced is high, making =
this mode=20
unsuitable for quality reproduction. Its main =
application is=20
in public address systems where quality is less =
important than=20
cost.
A compromise class of bias then is Class AB, where =
the bias=20
point lies between those for A and B. This gives an=20
improvement in efficiency and possible power output =
over Class=20
A, with better quality than can be obtained with Class =
B.=20
Class AB actually divides into two sub-classes, =
depending upon=20
how hard the grids are driven. In Class AB1, drive is=20
restrained so as not to cause grid current to flow; in =
Class=20
AB2, grid drive is increased and grid current flows at =
the=20
peaks of the positive half-cycles.
Finally, in Class C operation, the grids are biased =
well=20
beyond cutoff and are driven very hard in order to =
make anode=20
current flow in short pulses, just at the peaks of the =
positive signal half-cycles. This type of bias is =
restricted=20
to radio-frequency operation, where the pulses of =
current=20
merely excite a resonant circuit in order to produce a =
full=20
sinusoidal output. So now you know!
Harmonic Cancellation
Because of the fundamental way in which push-pull =
output=20
stages work, the signal currents in the output valves =
flow in=20
opposite directions in the output =
transformer=92s=20
secondary winding. The magnetic flux that links with =
the=20
secondary winding of this transformer induces =
voltages=20
that are additive in this secondary Thus, a =
positive=20
half-cycle developed in valve V2 at one instant causes =
a=20
corresponding voltage to be developed in the secondary =
winding; this is followed by a negative half-cycle =
produced by=20
valve V3, which also induces a similar voltage in the=20
secondary winding. As a result, the secondary voltage =
appears=20
as a continuous voltage, compounded from the =
successive=20
efforts of the two push-pull output valves working =
separately=20
but in co-operation.
However any even order harmonics, 2nd, 4th, 6th, =
etc.,=20
generated by the output valves cause opposing =
magnetic=20
fluxes in the primary winding which, consequently =
cancel out.=20
No even order harmonics appear in the secondary =
winding. This=20
is a major advantage of push-pull operation. It is =
obvious=20
that this is of more significance when triodes or beam =
tetrodes are used, since this is the type of harmonic=20
distortion that these valves generate. This somewhat =
dampens=20
the popular idea that the main characteristic of such=20
amplifiers is that they generate lots of 2nd order =
harmonic=20
distortion!
Value of Anode Load for the Output Stage
The actual value of the anode load in push-pull =
operation=20
is not the same as that calculated for single-ended =
operation.=20
Again, the best bet will be the figure provided by the =
valve=20
makers since this will have been computed to give the =
minimum=20
amount of odd order harmonic distortion. In pentodes, =
it is=20
third harmonic distortion that is the major type, and =
it is=20
possible to minimise this by a suitable choice of =
load. In=20
fact, it is essential to do this, since push-pull =
operation=20
does not result in any reduction in the odd order =
harmonic=20
content. It has been found that reducing the value of =
anode=20
load substantially below the nominal value reduces =
third=20
harmonic distortion and increases second harmonic =
distortion,=20
the latter then of course being cancelled by the =
push-pull=20
connection.
Negative Feedback
It is possible to write a book on this subject; =
many have=20
already done so and I have no intention of adding to =
the=20
published material. Suffice it to say that it is a =
matter of=20
great importance in the design of audio amplifiers, of =
any=20
type. Thus, the theory is not specific to valve =
amplifiers.=20
Its benefits are well-known in reducing distortion of =
all=20
types, in giving gain stability and in coping with the =
resonance problems of loudspeakers. That is all that =
need be=20
said about it here.
Examples of Hi-Fi Audio Amplifiers
We shall now have a look at a few audio amplifier =
circuits=20
which will illustrate the various approaches that =
designers=20
have adopted in the past.
The Garner 10W Main Amplifier
The Garner 10W Main Amplifier
The first of these is shown above and was designed =
by Major=20
H H Garner in the early 1950s. It doesn=92t have any =
pretensions=20
to exceptional performance but is, nonetheless, a =
creditable=20
performer according to G A Briggs, who tested it =
exhaustively=20
at the time. It was paired with a preamplifier; which =
had the=20
obligatory bass and treble controls as well as input =
switching=20
for pick-up and radio, and compensation for LP and =
78rpm=20
records.
The circuit is fairly simple and the line-up is =
quite=20
conventional for those days. (The =91line-up=92 was =
the collective=20
expression used for all the valves.) No miniature =
valves were=20
used; these were still a little way off in time for =
general=20
use. The output valves were the popular 6L6s=20
and these were somewhat under-run. The distortion =
figures may=20
not seem exciting but were seen as being adequate for =
the=20
purpose. At an output power of 5W the THD was 0.8%, =
which rose=20
to 1.5% at 10W output.
The pentode first stage, V3, included overall =
negative=20
feedback from the output transformer secondary =
winding. This=20
could be varied between zero and 22dB of NFB (Negative =
FeedBack). Apart from the fact that this varies the=20
sensitivity of the amplifier no justification was =
offered for=20
this arrangement. The points that were made were as=20
follows.
- Since the 6L6 output valves made no great =
demands upon=20
their input drive, a concertina phase-splitter was=20
considered adequate, with balancing of the bi-phase =
outputs=20
being achieved simply by matching the values of load =
resistors R24 and R26.=20
- Because of the high input impedance of the=20
phase-splitter; it was possible to use a pentode =
first stage=20
to give both high gain and good top response. =
A 10W Amplilier for Mobile Public Address
A 10W Mobile Public Address Amplifier
The circuit above shows an amplifier that appeared =
in the=20
American magazine Radio-Electronics in March 1957. =
This was=20
published in response to a letter from a reader who =
wanted a=20
power amplifier both for his FM tuner and for mobile =
PA=20
requirements. Since it had to be mobile, it couldn=92t =
use a=20
mains supply, so a feature of the design was a =
vibrator pack,=20
to convert the low level vehicle DC (6V no less) into =
the=20
higher level HT required by the amplifier.
The first stage, V1, was a 6SJ7=20
pentode, used purely as a microphone preamplifier and =
giving=20
enough gain to allow any high output, high impedance =
crystal=20
or dynamic microphone to be used.
A separate tuner input was provided that tapped =
directly=20
into the grid of one half of the double triode V2a. =
Here we=20
see a choice offered between the older and larger 6SL7GT=20
valve and one of the newer all-glass the 12AX7=20
(ECC83).=20
The output valve is another popular choice of the day =
the 6V6;=20
or a 6AQ5 could be used instead. The phase splitter is =
of the=20
paraphase type with V2a as the first stage and V2b as =
the=20
second stage. Anode loads are 150k and these feed the =
purely=20
conventional output stage comprising pentodes V3 and =
V4.=20
Overall NFB was taken from the 8Ω tapping on the =
output=20
transformer secondary winding to the cathode of V2a. A =
grid=20
signal for V2b is contrived by tapping the grid leak =
resistor=20
chain for V3.
The Craftsman C-500
The Craftsman C-500 Williamson Amplifier
In the May 1956 issue of Radio-Electronics =
there=20
appeared a circuit for a typical =91Williamson=92 =
amplifier, the=20
=91Craftsman C-500=92, shown above. For those not in =
the know,=20
Williamson was a famous name at the time in the field =
of Hi-Fi=20
amplifiers and belonged to D T N Williamson who was =
employed=20
as an engineer with the British firm of Ferranti. His=20
Craftsman C-500=92 amplifier uses a pair of KT66s,=20
which were very popular high power valves in those =
days.=20
Although they are actually beam pentodes, they were =
connected=20
as triodes by wiring the screen grid to the anode via =
a 100Ω=20
resistor (The use of the KT66 in triode mode is =
validated by=20
applications data; not all pentodes and tetrodes may=20
automatically be connected up this way).
Note also the circuit symbols used for them. Do you =
remember in Part Four mentioned that often beam =
tetrodes are=20
given pentode symbols in circuits, whereas the beam =
tetrode=20
version should really be used for clarity? Although =
the KT66=20
is classified as a pentode, it has beam forming plates =
in=20
place of a suppressor grid, and is practically=20
indistinguishable from a beam tetrode in construction. =
Some=20
would say the KT66 is a classic beam =
tetrode-ed.
In this circuit the first stage, V1a, is a =
pre-amplifier=20
whose output drives a concertina phase splitter V1b, =
thus=20
neatly putting both stages into one double triode. =
Note also=20
that, as was mentioned in Part Six on the subject of =
phase=20
splitter configurations, that here is an example of =
the signal=20
grid of the splitter stage, V1b, being DC coupled to =
the anode=20
of V1a, which then DC biases V1b. This makes a local =
bias=20
resistor network and a coupling capacitor unnecessary, =
a=20
saving in component count.
The third stage driver is a push=B7pull type whose =
antiphase=20
outputs directly drive the push-pull output stage =
grids, and=20
is included to develop a healthy driving voltage swing =
and=20
increase the open loop gain of the whole system, =
bearing in=20
mind the modest gain of the output valves, operating =
as they=20
are in triode mode (see discussion about output =
triodes=20
above). A =91balance=92 preset is provided to equalise =
the anode=20
currents of the output pain while the =91bias=92 pre- =
set sets the=20
net bias current in the output stage. In the =
Williamson design=20
this is typically 125mA (62 to 63mA for each valve). =
These=20
adjustments have to be made with the aid of a =
multimeter after=20
temporarily removing the link shown. Overall NFB is =
taken from=20
the 16Ω tapping on the output transformer back to =
the cathode=20
of V1a.
It was a feature of the Williamson designs that the =
output=20
transformer was of a superlative specification. The =
bandwidth=20
was enormous, sometimes in excess of 100kHz, with =
extremely=20
low distortion and of massive size. It is said that =
Williamson=20
amplifiers were designed around the output =
transformers. The=20
wide bandwidth introduced in stability problems, =
however in=20
the presence of NFB. It was possible to alleviate this =
by=20
including a bypass network within the loop that was =
only=20
effective at frequencies at or above 100kHz. This is =
the RC=20
network R1 / C1 between anode and ground of V1a. Its =
function=20
is to load V1a at high frequencies, reducing its gain =
and=20
ensuring a top end roll-off. As a further =
anti-instability=20
measure, a small value capacitor C2 was also added in =
parallel=20
with the feedback resistor R2.
The Ultra Linear Amplifier
The Williamson Ultra-Linear Amplifier
In 1952 or thereabouts, an unusual output circuit, =
which=20
had been patented some years previously was =
resurrected by=20
David Hafler and Herb Keroes. This circuit combined =
the best=20
features of triode and tetrode/pentode operation and =
was known=20
as the Ultra-Linear configuration. The =
important change=20
to the normal push-pull connection was minimal and =
consisted=20
of connecting the screen grids to taps on the primary =
of the=20
output transformer Oddly enough, opinions are divided =
on how=20
the circuit actually works, but the secret is believed =
to rest=20
on the introduction of NFB. Whatever the modus =
operandi, there=20
is no doubt about the benefits. The operating curve is =
more=20
linear than for triode operation even, and the output =
power=20
capabilities are about one-half those of pentodes and =
double=20
those of triodes. 'l`his means that, for the same HT =
supply=20
and signal drive, the output of an ultra-linear =
amplifier is=20
double that of the same amplifier when triode =
connected. In=20
fact, distortion at low levels is reduced as is the=20
phase-shift at high frequencies. This makes it =
possible to=20
increase the degnee of NFB without incurring any =
further=20
instability penalties. The Ultra-Linear Williamson =
ampliiier=20
is shown above.
Another aspect of the design of amplifiers such as =
the=20
Williamson types is interesting. At the time, valves =
might be=20
classified loosely as being either transmitting types =
or=20
receiving types. The former type of valve was =
obviously=20
designed for handling high powers, a prime requirement =
for=20
audio output stages. It was natural, therefore, to =
adopt=20
transmitting valves for Hi-Fi purposes, with =
consequent=20
advantages. They didn=92t make too stringent a demand =
upon the=20
driver stages as far as the amplitude of the driving =
voltage=20
was concerned, and they were operating on the linear =
parts of=20
their characteristics when delivering the high powers=20
required.
A Low-distortion 12W Amplifier
Design for an Ultra-Linear 12W Amplifier
Finally another ultra-linear amplifier circuit, but =
rather=20
simpler than the Williamson version. The power output =
from=20
this is a modest 12W but even this is often considered =
to be=20
adequate for the average living room. I reproduce here =
the=20
designer=92s original criteria, taken from the August, =
1958=20
issue of Radio-Electronics.
- lnaudible distortion at all feasible levels (in =
a 10 x=20
15ft. room).=20
- Low source impedance.=20
- High efficiency.=20
- Best possible stability characteristics.=20
- Hum and noise below audibility (under specified=20
conditions).
The designer concentrated most of his efforts on=20
consideration of the degree of NFB needed to meet his =
criteria=20
1 to 4 above, and opted for the ultra-linear mode of=20
operation, using a pair of 6V6s.=20
To obtain this type of operation with these valves =
requires=20
the output transformer primary winding to be tapped at =
25% on=20
each side of the centre=B7tap. The choice of output =
transformer=20
from what was on the market at the time, fell on the =
Acrosound=20
TO-310, which had a rating of 10W over the bandwidth =
20Hz to=20
30kHz, and 20W over the reduced bandwidth of 30Hz to =
20kHz,=20
the latter being adequate for good reproduction.
The circuit itself uses a 12AX7 (ECC83) double =
triode, with=20
one section, V1a, being a voltage amplifier directly =
coupled=20
to a concertina phase-splitter This gave a number of=20
advantages.
The DC coupling eliminates one stage of capacitive =
coupling=20
which gives greater frequency stability Placing the =
phase=20
splitter directly before the output stage (which is =
possible=20
due to the low drive requirements of 6V6s) eliminated =
the hum=20
frequently resulting from the large potential =
difference=20
between the heater and the un-bypassed cathode of V1b, =
being=20
amplified by a further push-pull driver stage. The =
anode load=20
of the voltage amplifier acts as the grid leak of the =
phase=20
splitter thus allowing the voltage amplifier to work =
into a=20
very high impedance, so that high gain with low =
distortion are=20
obtained from the first stage of amplification.
Efforts were made to improve the performance by =
matching=20
pairs of resistors and some capacitors (if possible) =
to within=20
1% of each other These are indicated on the circuit =
diagram.=20
It seems, from the comments made upon completion and =
testing,=20
that the amplifiers performance lived up to =
expectations.
That just about concludes this series on Valve =
Technology,=20
which has been restricted to audio applications only, =
and this=20
is now one of the few fields left where valves still =
survive -=20
we haven=92t even mentioned radio, TV, Instrumentation =
and=20
industrial applications, which of course at one time =
valves=20
had to cater for also! The only other area where =
valves are=20
still in common use is for radio transmitters (a very =
large,=20
high voltage, high power triode operating in class =
C).
We hope you have found the series both interesting =
and=20
informative, and that it has answered a lot of =
questions you=20
may have had about these devices. You may now be =
sufficiently=20
better informed to try some experiments of your own. =
Elsewhere=20
in this issue you will find details for Maplin=92s own =
valve=20
power amplifier project, which closely resembles =
Mullard=92s=20
=91520=92 ultra-linear design of the late 1950s and =
early 60s.=20
Happy metal-bashing!
Return to=20
Series Contents or back to Phase=20
Splitters =
|