INTRODUCTION

There are many books of circuits, and others giving the theory of
I
various circuits for people with different levels of technical education ;
but there seems to be a lack of books to help to solve the " obscure "
problems that arise when a piece of equipment has been built, but just
doesn't work. This book has been prepared to fill that gap, particularly DISTORTION
in relation to audio frequency amplifiers.

T
HIS chapter deals with various kinds of distortion caused by
Perhaps it can be regarded as the gap between theiJ1y and practice. operating valves under incorrect conditions, in some way or
Books of the constructional type give practical instruction, anc..l books other. T o understand this matter properly, one must be able to
of theory undoubtedly cover all the possibilities that can occur to appreciate the full significance of valve characteristics, acconiing to
stop our particular piece of equipment from working properly ; but their type. The best means to this end is the drawing of what is known
how often bas it been helpful to have an experienced man at hand to as a load line. Engineering text books make the treatment so mathe-
put his finger on the trouble ? The author has often successfully matical that ordinary un-mathcmatical people seldom understand it
played the part of" the experienced man "-even over the telephone. properly, even though they may have passed exams in the subject.
So he realised that what could be achieved this way could well be put So rhe author makes no apology in offering an explanation in simpler
into book form. lt is witb this in mind that the T ROUDLB T RACrNG language.
Gum.B has been prepared. Where the man on the telephone would
say to him " so-and-so happens," the reader turns to the guide to find Different Forms of Valve Characteristics
the appropriate symptoms, and there sees a reference to the chapter In the first place, there arc two ways of showing valve character-
or chapters explaining d1e possible causes in detail. i~tiC!l-thi.~
can be confusing. Figure I shows both ways for the same
But the book should not be left on the shelf until tro4blc comes.
It has been arranged in a logical sequence for reading, so as to gain
valuable permanent knowledge of the subject. When some particular
problem o.rises, and the reader perhaps only remembers that it was
dealt with in this book, then he can use the guide. (It has been
assumed, of course, that the wiring and circuit values have been
checked to diagram, and that still something is wrong.)
Whether the reader has little or much practical experience or
theoretical knowledge, the author believes his book will be of interest
and h elp. He h as tried eo avoid using expressions that those with
inadequate technical knowledge would not understand, and he is
confident that this presentation will also prove helpful to many who
have learned the theory in more technical terms, because often the SO 100 ISO 200 250
ANOOE \()LTS
" classtcal " explanation does not give a really clear understanding (b)
of the matter.
IIG. I. DlfrCRCNT PRLSLNTATIONS OF VALVE (' HARAC rl IU~liCS.
N. H. CROWHt.'RST.
London, 195 I. type of valve. The: 11rrangemcm at (a) gives curves showing how anode
current changes when ~rid voltage is altered, for fixed values of anode
voltage. I f, in use, only anode current fluctuated when a signal voltage
fluctuation was applied to the grid, then these curves would tell the
whole story. This would happen if anode and cathode (or filament)
were connected clirectly to h.t. plus and mmus respectively. But then

II
 AUDIO HANDBOOK No. r : AMPLIFIERS DISTORTION
there would only be anode cummt ftucn1Mions, and the valve would shown doned. However, a practic:~l valve cbaracrerislic is curved, so
be no good as an amplifier, b~::~oause signal voltage is also required, the anode currem ftu~tuations follow the shape at the right drawn in
whether to be passed on to the grid of the next valve, or to be fed to a solid line instead of the dotted sine wave. Thus the wave is distorted
loudspeaker through a Lransformer. As soon as some arrangement due to the " <.:urvature" of t he valve characteristic.
is made so that voltage fluctuations are obtained, the curves of Figure Now turn to the characteristics shown at Figure I (b) for the ~amc
I (a) cease to tell the story prope rly. valve a~ that represented at Figure I (a). Here graphs have been
But for the moment imagine that there are no anode vollag~:: ~ignal plotted showing how anode current changes when a.t1ode voltage is
ftu~tuauons, so one of these C\lrves could be used. Figure 2 shows the altered, for six different fixed grid voltagcs. At (a) the graphR showed
method usually c hosen. T he grid voltage reference 1incs arc drawn how Rnode current changed when grid voltage is a ltered, for five
different fixed a node voltoges. For a triode type vAlve, these two
types of characteristic are similar to look at, and for this reason they
can easily be confused at first sight. To know wlhich one lS being used,
the reference scales must be examined. At (a) the quantity plotted
along the bottom is grid voltage, And rhe curves are " labelled " with
anode voltagcs, while at (b) the quantity plotted along the bottom is
anode voltage, and the curves are " JabcUed" with grid volrages.
Each set of cu1 ves presents the same information in a different
way. In fact, one set of curves can be made from the other. Tllis is
rather laborious, and the reader will probably never bother to d o it ;
but understanding how it can be done helps in sct:ing t heir usefulness.
OISTOQTEO OOTPUT Figure 3 shows the method. This is just a copy of the curves at Figure
(ANOOE) CURRENT. T, with the labelling left ofT the curves thcm~elvcs, so that number5
DOTTED CURVE IS
-5 PURE SINE WAVE

-2 _,
GRID VOLTS
(a)
F IC1. J, FIO.I RLPC\HOTO~IIO\\.l ii ECOVP\I!I~01'00fTlltTW01-0R\1S
FIC. 2. \IHT HOL) 01• I JS INO A CURVh rROJ\1 I IG. I (11}. 01 PR L~~l\ rATION.

down further, and o sine wave is drawn on thrm, representing a pure can be written in identifying the same points on both sets of curves.
signal input voltage to tihe grid. If the characteristic were stra1ght, as At (a) a grid voltage is represemoo by an upright straight line, while
shown dotted, then each value of grid voltage would cause the at (b) it is represented by a curve. Thus the points I, 2, 3 are all at a
corresponding anode <.urreut marked off by the dotted Jines horizontally, grid voltage of zero on both Uiagrams, and points 8, 9, xo, n, xz arc
and extending to the right, so t hat a sine wave representing a pure all at a grid volmge of mim1s 2. At ( b) an anode voltage is represcnled
signal currenc fluctuation in the anode circuit could be drawn, as by an upright straight line, while at (a) it is represented by a curve

12 I3
 AUDIO HANDBOOK No. r: AMI'LIFJERS
DISTOR'r!ON
Thus points 7, xI, 15, 18, 20 arc all at an anode voltage of 200 on both
diagrams. The reader can trace out the corn::~pondencc of other
points on hoth g raphs for hinuclf
Load Lines
But what is a load line ? And how does it help show the behaviour
of the valve? Suppose the anode h.t. ~upply, of 250 volts, is con-
nected to the " top " end of a 25,000 Ohm resistor, the " bottom "
end of which is connected to any valve anode, as at Figure 4· If the
valve does not pass any :~node current,
.. 250 there will be no currem through the
--~------=-v..::O~L~TS r esistor, and the anode voltage \~·~11 be the
same as h.t. plus, 250 volts posllwc. But
25000 if the valve draws to milliamps, there wlll
OHMS be a volt drop of 250 volts in the 25,000
Ohm rcsisror, so its bottom end will be l• lG. ' · L0,\0 Ll~ll IN ITS SIMl'l.f.ST rORM-CI RCUIT O F F JO. 4.
250 volts negative from h.t. .Plus, i.e.,
zuu. If the valve draws say, 4 milliamps, and to the righr so that a sine wave can be drawn as input to
there will be a volt drop of 100 volts, the rid, and a cdrresponding output wavefo.rm be drawn by reference
leaYing the at100e at 250 - IOO = to Je curve. Again, if the ~urve .were stnught, rhe output would be
150 volts pmirive. A Jot more points pure but curvature causes distortion.
could be taken, representing uiifercm In practice this curve need not be drawn to find out whether the
_ _.._____ ~--o anode voltagcs according to the currenr
taken by the valve; but they will all be
FIO. 4. CIRCU I T FOR found to connect up in the straight line
SIMrLI. LQ,\0 liNE,
AB or Figure 5 (a). Whatever happens in
the grid circuit, the anode vol[age and current must be given by some
point along this line, because of the volt drop in the 25,000 Ohm
resistor. For this reason, such a line i!i caUcd a 25,000 Ohm load line.
At Figure 5 ( b), the curves of Figures I and 3 (b) are redrawn
and the zs,ooo Ohm load line drawn on top of them. Now suppose
that the grid voltage is minus r. T hen the combination of anode
cur rent and anode voltage must be somewhere along the curve marked
"grid volts - x." Since it must also be on the load line AB, the only
possible anode voltage and current in the circuit of Figure 4, when tht:
grid volt.age is minus r, is given by point D on figure 5 (b), ahout
120 volrs, S milliamps. Similarly other points along rhe load line,
where the grid voltage curves cross ir, lettered C, E, F, G, H. give the
anode voltage anci cunenr for t he grid voltage represented by each
curve.

How to Use Load Lines

Figure 6 shows how the information obtained frorn such curves
with a load line can be used to produce nn output wavefo{m tor a
pure input wave, in a manner similar to that used in Figure 2. Anode
voltages along the load line AH of Figure 5 (b) are plotted against a
scale of grid voltages. T he reference lines are again extended downwards
Ill:. 6 . ~~~~THOI) 01' FIC.. l USED WITH l.O.\D LI :> U Or 5 (b),

I4 15
 AUDIO HANDBOOK No. r: AMPLI FIERS DISTORTION

load line causes disto(tion. Straightness of the curve depends upon Load Lines for Practical Circuits
~tcps of anode yol~age corr~sponding to equal steps of grid voltage Figure 8 shows load lines representing two types of circuit applied
betng equal. If adJacent gnd voltage curve.s are for grid voltages at to triode type characteristics, and Figure 9 applies the .same type.s of
uniform intervals (in Figure 5 (b) the intervals are all 1 voir), then the circuit to tetrode or pentode characteristics. At (a) in each case, the
curves must cross the load li11e aL cxaclly uniform spacing for there load liuc arraugemem represents resist.auce/capacitam:c coupling,
to be no disrortion. Inequality of spacing, in any way, indicates that shown in Figure Io. The load line AB represent~ the d.c. drop down
distortion will be caused. In Figure 5 (b) the space CD is sljgbtly
longer than GH, and spaces nearer zero grid voltage arc longer than
those nearer H. In this case the uiiference is only slight, and so the
clisronion is only slight, as shown in Fi~re 6.
Now, to apply this idea of load lines ro practical cases, first notice
how the slope of a load line depends upon the resistance it represents.
Figure 7 shows three Joau lines each starting from a h.t. voltage of
250, plotted to a current scale different from that of Figure s. Line

ANODE VOLTS.
(o) f!ESISTANCE/CAPAC!TANCE (b)
COUPLING.
1'10. 8. LOAD LIKES l'OR PRACTICAL CIRCUITS- TRlOOilS.

the resistor Rv as already ex~lained. But when signnl voltages are

received at the grid, changing tts voltage momentarily fmm the steady
bias value represented by point B (- 2·5 volts in Figure g (a), - 2
volts in Figure 9 (a)), the change of anode voltage will cause charges
w on the coupling Ct~pacitor to be changed, producing small currents in
8z R 2 • 11us means that from lhc viewpoint of ~ignal vollagc~, R 3 i:s in
<i
0o~~so~~,oo~--L_~~~u

ANODE
FIO 7. .SHOWING flOW SLOJ'El Of I.OAD UNE DEPBNDS ON VALUC.

AR passes through :250 volts at A, when no c urrent is flowing, and

through zero voltage (meaning that the whole 250 volts j:; dropped in
the reslslor) al a current of ro milliamps. This represents a load of
25,oooOhms,as before. The other load lines correspond to ro,ooo Ohms
and s,ooo Ohms. 250 volts dropped across I o,ooo Ohms produces
Z5 milliamps, so AC represents xo,ooo Ohms, and 250 volts across ANODE
5,000 Ohms produces 50 milliamps, so AD represents 5,000 Ohms. (<1) RFSISTANCF/C.APACITANCE (b) Cl-k. ·c OR TRANSFORMER
l•rom these eKamples, it is clear that lower resistance values arc repre- COUPLING COUPLING.
sented by steeper sloped lines, and higher resistances by lines nearer FIO. 9. LOAD LINt:S FOR PRACfiCAt:. CIRCUITS-PENTODES OR OI!A \1
horizontal. Tt!TRODES.

!6 17
 AUDIO HANDBOOK No. 1 : AMPLIFIERS DISTORTION
parallel with Ru reducing the effective lond resistance; but the part of d1c load line, before the other is reached (this is what happens
operating poim is still fixed by the volt t.l rop down R1 only. So the if incorrc~t grid bias is used).
operating load !me. (;1), will have a steeper slope, but will still pass Too little bias means that the cntl where the zero grid volts curve
through point B. cros~e.~ will he reached fi rst. Unless the vnlvc ts used under " power
drive" conditions (explained later), when the grid voltage runs past
zero, Lht [a~l that the grid starts to collect electrons, as welt as the
anode, means that grid current '>"ill cau~c a volt drop in the grid
circuit rcs i~tances ; this volt drop will start suddenJy at one point
in the waveform, where the grid passes zero voltage, so the volt drop
will not follow the waveform, as it does in the anode resistor, but will
., come on suddenly, causing distortion. In effect the grid circuit
resistance i ~ 11Su111ly so high that all the voltage above zero is dropped
in it, and the waveform at the grid is flattened off as if chopped off
at the voltage when: the g lid passes through zero.
Too much grid bias results in distortion at tht: other end of the
load line, before the positive excursion from btas voltage reaches zero.
This is known as " anode bend " distortion, becau~e it i~ due to the
curvature of the anotle ~:u nent cnaractcristk when the grid voltage is
almost cutting oft the anode current. This form of distorlion is not so
drastic as grid current di~tortion . Figure l i (a) shows a typical wave-
form produced by insufficient bias, while (b) shows one produced by
too much bias, as shown on an oscilloscope screen wh en the input is
pure (shown dotted so that departure from the proper :ihape can be
fiO. 10. R ESIST A 'lCL,CAPACITAM; I COUPLIN(l C: IRCU II .
seen better).
In (b) of Pigmes 8 and 9, the load line arrangement is for choke/
capacitance or transformer coupling. rn each case AB represents the
volt drop ciue to rhe·rcsistance of the winding connected in the anode
circuit. But the winding wiU have a very high inductance, which is
not cousidered in the load line diagram for the mome.nt. T his
inductance is used to provide coupling to a load resistance that is
much higher rh~n the winding resistance, and it is this load resistance
that fixes the slope of the load line, represented by CD, still passing
~
through B, because the winding rc5istancc fixes the operating point ~
due to grid bms. (o) TOO LITTLE BIAS. (b) ( c ) REACTIVE LOAD
The slope is fixed by the actual load resistance value tor choke
coupling, nnd by the transferred, or matched, resistance for trans- FIG. 11 . OISTOR IION CAUSLO B¥ V"-RIOVS I NCOilREC'T OPt R .\TJNr.
former ~:oupliug, a& explained in Chapter 9· (.0N11tnONS.
Allernotove causes arc alvtn 1n rho tc•l.
Wrong Grid Bias
I n Figures 8 and 9, the load lines shown are about the" optimum," Wrong Load Resistance
or bc~t values, for the curves of the valves as drawn. ~ow turn to see That is what can happen when the correct load is there, but the
the effect of operating umlt:r incorrect conditions. T o obtain the full grid bias is wrong. Now to sec what happen~ when the load goes
use of the load lines shown in Figures 8 and 9, not only musr rhc load haywire. The simplest way tor this to happen is by using the \Hong
line have the correct slope, but the opcrntil1g poinl must be in the resistance, or combination of rcsb tancc~, for the valve i n question.
correct place along it, otherwise signal voltage applied to Lhe grid will Suppose first that the valve is a triode, with characrerislics similar
swing the anode ~urrcn 1 and voltage beyond one end of the useful to those in Figure 8, repeated to show this point in F igure 12 : with

I8 I9
 AUDIO HANDBOOK No. 1 : AMPLI FIERS
DISTORTION
the correct load line AB, there is no distortion ; using a higher value,
passing through the same operating polnt, gives a load line such as those shown in Figure 9, repeated for t:hls purpose in Figure r3.
CD, representing twice the resistance of AB; this still gives little or The optimwu load line is AB, and CD and EF represent loads of

UJ
g r---~~~---~~~--~4
z<(

FIQ. 12. Oll I I' RENT LOA I) RI SIS I ANCCS- TRIOOCS. 11(.,, IJ. Oll H KtN I I UAO Kl .'>l~l ANl' U - PENTOUL S 01{ JJCAM iiTROI>L.S.

no d istortion, and produces a slightly larger anode voltage swmg, as double an~ half optimum respectively. The half "alue, E F, reduces
shown by the dotted lines. Using a lower value, however, not only "?ltag~ swtng at the output to about haJf; but there is little or no
drops the available swing, but also causes distortion, as shown by the dJstomon. The double value meets the zero grid volt curve where it
line EF, representing a value one-third that of AB . bends, over ami begim to meet _tlu; others, while the other end of the
For voltage amplifier stages, it is good to use a higher value than l~ad lme reaches up ro .v~ry high anode voltages (off the right J1and
the " optimum" to get higher Rain; but a higher h.t. supply voltage ~1~c o~ lh~ figure). 1 his shows two bad features: (a) excessive
is needed, because of the greater volt drop in the ll.igher anode resistor dJStoruon ~s caused; (b~ hJgh anode voltages occur, which may do
value. But for power amplifiers there is no advantage in matching to a damage. 1 he resuJL of usmg too low a value does not cause distortion ·
load higher than " optimum," although it will not cause distortion. bll:t the output is roughly proportional to load resistance, so valuabl~
To see why, take the following typical figures. Operating anode voltage gam or power output may be lost.
250, anode current 40 rnilliamps, oprimum load 4,000 Ohms: with T he \~avc_to~ caused by working a !J'iodc with too low a load
this optimum load, the voltage ~wing will he from ahout r2o volts rcs1stance JS similar to t~e anode bend distortion of Figure u (b),
72·5 milliamps, to about 380 volts 7'5 rnilliamps, a signal output of w~lc that caused ~y wo;kin~ a ~etrode or pentode with too high a load
about 2 · 1 watts. Using double the load, 8,000 Ohms, the swlng changes restsrance causes dJstortJon stmilar to that shown in Figure rr (a)
from about 95 volts 6o milliamps to about 405 volts 20 milliamps, a . W~th RJC coupling it must be remembered that the load resis~n e('
signaJ output of about 1 ·5 watts. IS £!Ot Slf!!ply the ano~e coupling resistor, but is effectively the coupling
(The formula for power output is rcststor m parallel w1th gnd leak or input resistance of the following
s ta~e.
voltage swing current swing
- 8ooo
Elliptical Load Lines- Reactive Load
current swing in mllliamps, power in watts.) Tho~e cases Ileal with wr~u~ ~ias or wrong Joad, always supposmg
For triodcs, the exact load is not highly important; but it is the l?a<.l m th~ valve anode CJrcwr behaves as a pure resistance. Jn
better to be on the Wgh side than the low side of the optimum value. J?rac."tlcc, cspeCJ~Y at the ends of the frequency range, it is seldom
Turning to tetrodes and pentodes with characteristics similar to like a pure rcststance. Inductances and capacltaoccs in the circuit
make the load line er elliptical." Tllis is not so difficult to understand
20
21
 AUDIO H ANDBOOK No. I: AMPLIFIERS DISTORTION

as is often supposed. The property of a pure reactance, t.c., a capacitor

or high Q inductance by itself, is that the peaks of current coincide
with zeros of voltage anJ vice vet-sa.
A load line due to pure reactance takes the form :.hown in figure
' 4· The arrowheads shown represent the direction taken by the
operating voltage and current for a high Q inductance. When voltage
1s a maximum positive, at A,
current is zero; quatter of a
+ cycle later current is at maxi-
mum, and voltage is zero, at
B; another quarter cycle, and
voltage is at maximum nega-
tive. current is zero, at C ; ANOOE
tzw after the third quarter cycle, (Q) REACTANCE IN PAI'lAI..Lil. (b) nEACTANC£. IN SEAIFS

~0 current is maximum negative,

and voltagt: i:~ .Gero at D ; 110. IS. "llli'Al." I O\l1 tiN I'\ I OR REAC I .V-/CI ADI)rD TO FIXI:D
=>
(.)
R t.'>l'l f \ N<:C. i\ i
finally at'ter the fourth quarter
cycle the pMition is back at
A. Current reaches any rc~cta~ce n~rmally m~t in these circuits is the coupling capacitor, so
position in ils cycle, quarter thiS distoruon can anse at low frequencies if the coupling capacitor
ol a cycle after voltage has following a letrode or pentode stage is too small. A commoner cause JS
+ passed the same position. For t he speech coil reactance of n loudspeal:er, used as a load for a tetrode
a capacitance, the direction or pentode output. Thi~ is partly why tetrodes or pentodes need
ttG. 14. r uPnC\r r O.\n r rNr nur To of the arrowheads would be correction capacitors connected, either from anode to earth, or acru:.s
PURP. 1'1/0Ul-1,\I"ICI' reversed. the secondary of the output transformer.
I n valve circuits pure In triodes, the kind of shunt reactances that may cause trouhle
n:a~Lancc loads can never occur, because there must be coupling arc the gr~d input capnc}tancc of t.hc following stage, which is unlikely,
reslSt~ces, etc., and the volta~e and current never become actually an? the 1 ~\ductance ?1 a coupling. choke or coupling tran~forrncr
negatsve, our fluctuate, always m a positive direction. The reactance pnmru;y. 1 ht> l arr~r will C!luse d1stort1on particularly at low frequencies,
component appears either in shunt or in scr ies with 1he known resistance sf the mductonee 1s not hrgh enough for the purpose Notice that 1his
load. <.li:.turti_on i~ quite different from that produced by saturation of the
The elfect of /'arallel addition of reactance w tt resistance load, core of the c~~ke or ua~sfon;ner, due to too hig h ~ level at too low
AB, acrMs " idea " valve characteristics (i.e., in which the g rid a fr·cqucncy. 1 he latter JS quttc another source of d1slortion, and one
voltage " curves" are all straight, parallel, ancf equally spaced), U6ually less often met in modern practice.
working at the operating point C, is shown in F igure rs (a), whHc (b) The kiml of waveform produced is rather different from anode
represents the effect of reactance added in series with a rc:>istance load. bend distortion, although it does not sound very dilfcrcnt. l~igurc
Tt is not easy to show elliptical load lines on practical valve character· 11 (c) shows the waveform produced by a coup ling c hoke of too low
istics where the effect of reactance causes distortion, because the ellipse on inductance value.
• t
~oc:. out of ~hape, the current waveform being d istorted differently
from the voltage waveform. But from the ideal ellipses, applied in Push-Pull WorklJtg Gives Bigger Outputs
imagination to actual valve characteristics, it is easily understood that Push-pull operation, particularly of triodc type valves, enables a
shunt reactance, in the case of uiodes and series reactance, in the case greater output to be oblaincd from each valve with less ovcraU
of tctrodc:. or pentodes, cause distortion. distortion. It was noted that use of too low a load value with triodes
causes distortion similar to the anode bend type. Using two valves in
Practical Cases of Elliptical Load Line push-puH allows both of them to work with a lower load resista11cc
In tetrode-; or pentodes, series reactance causes distortion verv giving a bigger power oulpul, bccau:.e the bend distortion cancels i~
like that en used by too high vnlue of resistance load. One kind of series

22
 AUDI O HANDBOOK No. r: AMPLIFIERS
the output. Tllis is known as " low-loading." By working at a greater
negative bias, even bigger outputs can be obtained from the same pair
of valves by swinging deliberately further <; round the bend," the 2
greatest posslbiliry being achieved when the operating point r~::achc)
the middle of the " bend "-the arrangement known as " Class B
o perated." Intermediate values of bias give coudltiMs known ll~
"Class AD operated." These have the advantage of being less critical INSTABILITY
to adJUSt for satisfactory working than rhe full Class B condition.

S
Still bigger outputs can be obtained from some rypes of valve by OME of the results of instability arc not unlike those produced by
workmg them in push-pull like this, with a slightly higher h .t. supply the forms of disronion dealt with in rhe last chapter, which is
voltage, and by driving the grids positive over part of the cycle nt why rhis one is put next to it. A variety of effects come under
maximum outpur. This requires special atrention ro chc lilagc before, tbe heading of instability. The word means unwanted oscillation, or a
to see that the necessary power to supply the grid current is available t7ndcncy to os~illatc. The frequency of unwanted oscillation is usually
withou t causing distortion. As much literature has been devoted to c1tber low or high, ofr~n below or abov~ rhe audio ranpe of frequencies.
circuit d~igns for this purpose, this book will not go into details of Below the low end Jt causes what 1s known as ' motor-boating."
such circuits. Above the high end, ir causes " h.f. blocking." Sometimes it is not
above the range of audible frequencies, when a high pitched squeal
Recognising and Locati11g Cause of Distortion will be heard.
From the information in this chaptre· it i ~ evident that a variety of
causes can introduce distortion of one of the types s hown in Figure 1 1, Motor-boating
and careful checking up is necessary even after the uffcmliug stage has Take motor-boating first, because the possible cau~cs are fewer
been located, in finding just why the distortion appears. If an oscillo- than for h .f. forms of instability. It w.ill not occur in amplifiers having
scope is available, the distortion is easy to recognise from this figure ; fewer than three :;cage~. It i~; due to positive feedback at a low frequency.
but in the absence of an oscilloscope, it may be necessary to rely on The mos~ common cause is i?Sufficien.t decoupling of the h.t. supply.
listening rests. Waveforms as in Figure n (a) result in sound very Use of e1rher larger drcouphng rapawors, or smaller coupling capa-
like that produced if the speech coil is knocking against the pole piece
or some other object at one end of its trAvel. Having checked that this
is not happening, it will be known that one of the types of distortion
resulting in this kind of waveform muM be occurring. Wavefurms
shown at (b) or (c) do not cause such nouceablc distortion to the ear,
but can be recognised best on certain programmes as producing a
rather sharp reproduction.
If all sounds above a certain level, regardless of frequency (pitch),
are distorted, then the trouble is incorrect loading or biasing of a valve
somewhere. If the low frequency or high frequency sounds are
particularly distoned, then the trouble :Irises due to reactances, the
low due to a couplin~ inductance (choke or transformer primary)
of too low value, t he high d ue to insufficient correction capacitance on
the output tctrode or rentorle.
A simple check for locnlising the cause of distortion is to measure
·~ ,
anode currents, or volts acro~s bias rc:.i:.tors, while the amplifier is in
use. Anode bend causes a rise in current or bias volts when 9gnal is
prtSent. Grid current causes a drop in current or bias volts. Slight
changes proportional to signal level are normal; but a sudden change
when a certain level h reached indicate~ distortion. Dias or load
resistors should be a ltered accordingly to correct the defect.
1'10. 16. TYriCAL CIRCUIT LIADL C ro ~10TOR·00A'f1NO.

2-f 25
 INSTABILITY
AUD IO HANDBOOK No. r: AMPLIFIERS
citors, may effect a cure. Figure 16 shows one circuit where this may be
the case, and the capacitors marked Cc may be made smaller, or those
marked cd larger. . . .
Another circuit that can be unstable IS shown m F 1gure 17.
The cause is usually unbalance betwee.n the output valves at extremely
low frequencies, due to slightly different values. of. the. lettered co~­
ponent!> that should be identical. Although this c!rc.mt has been ID
favour with quality enthusiasts,, a g~od p~~se-sphttmg transfor~er
saves a valve avoids this cause of msrab1hty, and generally g1ves
quality not inferior to that from the circuit of Figure 17.

PIG. t8. REDUCTION OF H.T. SUPPLY UI:'HT IMPBDANCB. USING TillS

CIRCUIT HELPS WITH MOTOR·ROATING AND OTIIllR PROBLEMS OF
INSTABILITY.

the oscillation is continuous, the effect of both is somewhat similar.

The ampHfier is worlcing at full output, or nearly so, at a frequency so
high that it is inaudible. ln the absence of signal, the output appears
unusually quiet. The h.f. oscillation may be strong enough to prevent
any desired sounds coming through at all, or it may be that just the
strongest sign als will temporarily reduce the strength ofb.f. oscillation,
allowing a little sound to break through. Sometimes, where the h.f.
oscillation is not very strong, practically all the sound gets through ;
HG 17 AN AM PI IFH!R WITII PUSH-PULL OU rPVT FED FROM A I'HASe· but it sounds as if it is breaking through-quite a similar effect to that
. SPI miNC STAGS CAN SUH'l' R MOTOR·BOA1 JNO IROUllf R
produced by an intermittent connection just touching.
In amplifiers with h~gh gain it may be found that !n crcasing the The kind of oscillation that occurs through several stages, usually
value of de coupling capacitor only reduces the motor-boatmg frequency, the whole amplifier, is due to feedback from output to input at a very
while changing the coupling capacitors has. little effe~r•. or ma~ raise high frequen cy, beyond the audio range. Amplifiers with very high
the frequency slightly. The amount of gru~ mak~ 1t 1mpract1cal to gain are particularly susceptible to this trouble. If the frequency
increase decoupling enough to prevent the mst~bli1ty. Under these should be within the audio range the trouble is more obvious. To
circumstmccs, a useful method IS to reduce the 1mpeda~ce. of the h.t. prevent stray coupling, which might cause such oscillations, each
supply unit. A good circuit that reduces h.t. supply ~ntt 1mpeda!lce, stage sh ould be separated from the next by an earthed metal screen,
and at the same time supplies a high degree of smoothing, as requ1red and the wiring should be arranged so that connections belonging to
for high gain amplifiers, is shown in Figure 18. each stage arc kept separate. The layout should keep stages near the
output well away from those near the input end of the amplifier. It
H.F. Blocking . .. . . is good to arrange the wiring so that " hot " leads are as short as possible.
There are two main forms of h.f. mstabihty: those mvolvmg
several stageas, nd those where just one valve oscillates by i'tself. Where

26 27
 AUDIO HANDBOOK No. r: AMPLIFIERS INSTABILITY
This can be achieved by seeing that the anode connection of one stage " line." Figure 19 shows the method diagrammatically. If metal-cased
is near to the grid connection of the following stage. The coupling electrolytic capacitors are used, it may be advisable to use a chassis
capacitor, or transformer, should be quite close, so the leads are short, earth, where the chassis mounting forms the earth return ; but care
and grid resistors, anode coupling resistors, and other components should be taken to sec that the capacitors are near to the stage to which
connected to " hot " electrodes, should be kept close to the valve pins. they are connected. It is simpler, for high gain amplifiers, to use one
Where for some reason, e.g., the grid connection is on the top cap, of the carton-packed electrolytics, so that mounting position is not so
the coupling lead cannot be kept short, and may not readily be taken important, and the necessary care can be taken by means of the negative
up inside the valve screen, a screened lead should be used, and also a return lead.
screened valve cap where the gain is high.
Electrodes that are decoupled to earth may have longer leads, if Turning to h.f. oscillation in single valves by themselves, one
that assists the layout for keeping " hot " leads short. But there is possibility occurs with beam tetrodes, such as the KTZ63, as high
another point here that requires care : a lead may be taken from a gain amplifiers. Some of the recommended circuits for these valves
screen or cathode pin to a dccoupling capacitor, mounted a little use a high series screen feed resistor with a decoupling capacitor to
distance away ; but the return lead from this decoupling capacitor earth, as in Figure 20. But under some conditions this circuit will
should be taken back to an earth point for the stage it is decoupling. produce an oscillation of extremely high
It must be realised that although there is no signal voltage on a Jrequency, due to the geometrical
decoupled electrode (so that it is not " hot " in the usual sense) there structure of the beam electrodes. When
is a signal current, and if this is not returned to the correct point, it this happens, the anode current falls
can be responsible for unwanted coupling that will cause instability. almost to zero, and the screen volts drop
For this reason it is best to arrange a separate earth line, only con- down nearly to zero. The frequency of
nected to chassis at one point, preferably at the amplifier input, and oscillation is so high that it is not passed
connect all earth returns from each stage back to one point on the earth on to later stages, but it blocks signal in
the valve concerned, giving the break-
through effect. The cure is usually
., effected by using a lower value of screen
feed resistor, with perhaps a somewhat
lower value of anode resistor, to main-
rain the correct balance of anode and
screen cuxrents.
The next possibility occurs generally
in high slope output valves. Both
triodes and pentode or tetrode types are
subject to it. The cause is that the grid
wiring forms a h.f. tuned circuit, and the FIO. :20. 1 IllS BRAM TIIT·
anode load is inductive at the frequency RODE CJRCUI r CAN CAUSE
ofoscillation. A capacitor from anode or H.P. IJLOC'K INO.
cathode will sometimes cure this ; but
the more general cure is the insertion of resistors known as " anode
'>toppers" and "grid stoppers." For pentodes or tetrodes, sometimes
1.creen stoppers arc necessary as well. Suitable values arc : 50 to
500 Ohms for anode stoppers, according to valve impedance ; and 5,000
to 15,000 Ohms for grid stoppers, again according to size of valve
(larger valves requiring lower values because of the grid input capaci-
tance). Figure 21 shows a typical push-pull output stage with grid and
anode stoppers fitted.
IIG. 19 MHIIOD OF I:ARTH WIRING 10 AVOID INSTABILITY AND It is important that " stopper " resistors should be connected so
O'IIIER TROUBLES.
the resistor itself is right against the valve-holder pin.
28 29
 AUDIO HANDBOOK No. I : AMPLIFIERS INSTABILITY
A problem occurs where this form of oscillation starts up in an Locating H.F. Oscillation
output intended for power drive operation. Grid stoppers cannot be OsciiiJation originating in a single stage can usually be located by
voltage and current measurements. A valve, known to be up to
standard, is found to be working at low anode current, with apparently
almost zero bias and full h.t. volts on the anode. All circuit values
check, so it seems as if Ohm's Law has gone wrong! H.f. oscillation is
the explanation--only parr of d1e actual bias being produced across the
self-bias resistor, the rest caused by grid current. If the oscillation
causes blocking, the voltages will show a change whenever sound
breaks through.
Parasitic Oscillation
There is another type of oscillation that appears, sometimes in
Class B amplifiers using power drive, only when a signal is being
passed. The waveform on an oscilloscope is shown in Figure 23. A
shock-excited wave train is set up
at each point where grid current
ceases. Usually the frequency of
FIG. 2 1. METHOD OF W IRING GR ID AND ANODE STOPPERS. this wave train is too high to be
audible, and if only rhe one note,
inserted as in Figure 21, because they would defeat the object of shown in Figure 23, were being
providing power drive. The only alternative is to provide a stopper passed through the amplifier, there
resistor connected in parallel with the drive output, as in Figure 22. would be no audible evidence of
The appropriate value can only be determined by experiment, as it the effect. The trouble becomes
evident when mixed signals, such as
music, arc passed through the
h.t .+ amplifier. The h.f. oscillation
during part of the wave has dte
effect of choking higher frequencies
in d1e Signal for part of the low FIG. 23. J'ARASITIC OSCJLLA rtON.
frequency wave only, so the repro-
duction is like a rather bad case of intermodulation. When a low
frequency is reproduced at high level, so as to cause this effect, t he
higher frequencies sound " dithcry."
To cure this trouble, lower value shunt grid stopper resistors
(Figure 22) should be tried. I f reducing this value fai ls to stop parasitic
oscillation before it causes distortion in the drive stage a better drive
transformer is needed, probably with less step-up, or greater step-
down.
.___ _ BIAS
L-----CO!'INECTIONS
PlO. 22. SrQPPER CONNECTIONS WHI I POWER DRIVB (CLASS B,).

depends on the design of the transformer as well as on the types of

valve used for both drive and output.

31
 I '

1000/1
w
u
z
~
800/1
700/1
600/1
2001
os w S00/1
0 "6 ~

~i
4 00/1
07-
o·a ,......... 300/1 _ so ~
a. 70
I :::> er we./)
~ 200/1
UJ u~ so
u 0 45 ~ z::t
~0
~
a. 40
~ ~ (/)
w
a.. 110. 27. AN ABAC I O R CALCU1 Arl '-IG
0 (Y
~u..
30 MATCHII\Ci RA110.
'---"'
c./) 100/1 40~ -0 (See page 49)
UJ
z
z er t-
a: 80/1 t- 20
:::>
~~
:::> 70{1 ....
t- t- Uc.J)
60~ -
~ a: a

101
O..I
oo a: - 35 -z
u<C
w SO/I z

J
a:
u ~
a: w~
30~
...:t...
~ 0 ~0 8
~~
IJ... 30/1
a: c./) z 7
0 z <( 6
a:
w
~ ~
5
t- 25 w
~ ~
4
a.. t-

20~
c./) 3
a:::>
9 2

IS

32 33
 HOW MUCH GAIN ?
not contribute appreciable gain. In fact, sufficient voltage to swing
one of the output valves must be provided at the anode of the stage
3 before it.
Choice of Valves
The gain contributed by each stage depends upo~ the type of
valve used and the circuit in which it is used. lncreasmg the anode
riOW MUCH GAIN? resistor will increase gain to a certain extent, but often at the expe~e
of quality. High gain triodc valves, such. as t?e H~3, can opera~e w1th
an anode resistor of 50,000 Ohms, and w1ll g1ve a hule more gam, but
not double by using one of 10,0000 Ohms. An h.f. tetrode or pentode,
ilE provision of m~r~ gain than is needed makes the job f!lO~e such as th~ KTZ63 or 6}7, will give more gain than the H63 with
T difficult. So a dec1s1on must be made as to how much gatn IS
necessary.
Inputs Used
50 ooo Ohms. It will give ten times as much gain with an anode resistor
of'500,000 Ohms, but the quality will suffer. An inte~mediate val~e
can be chosen, according to the balance between gam and quality
desired.
An amplifier to be used for reproducing only gramophone records These high gain valves wlll not usually provide a larg~ voltage
need not have much gain. If it has to amplify from a microphone, swing without distortion, because of their curvarurc. For th1s reason
then the gain required will depend ou the type of microphone used. it is best to use high gain type valves for early stages and. change to a
A carbon microphone (which of course needs polarising) requires lower gain valve, with a lower anode impcdanc~ an~ working at rather
little more gain than a gramophone pick-up-one extra stage would higher h.t. on the anode, for the later stages .. Cucu~t values are usually
be suOicicnt for most purposes. Moving coil microphones arc not so given in valve tables or data, such as coupling rcs1stors (anode load),
sensitive, so one extra stage would only provide enough gain for close bias resistors, etc.
talking into a moving coil microphone. Two extra stages would be
needed for long range work. A ribbon microphone is usually less
sensitive than a moving coil type, and so may require yet more gain.
I t is when these higher gains are used that more difficulties arise, in
both stability and hum pick-up. However it is quite possible to build
very good equipment for high gain, by careful anention to the points
mentioned in Chapters 2 and 5·
Type of Output Stage
Another aspect of the gain question is the type of output stage
to be used. If long grid base triode type output valves are used-the
ones needing a large grid bias-an extra stage will be needed to get
enough voltage swing to drive them. A point to watch here is that the
drive stage must be capable of delivering the required swing without
distorting. Usually distortion occurs first in the output; but it can
oflen be in the drive stage, unless care is taken to see that it can produce
enough swing. A drive stage will need a higher h.t. supply than all
earlier stages, because this is one essential for getting necessary swing.
If the output \ISes power drive, then the drive stage often produces
no voltage gain at all, or very little. It is devoted to convening the
voltage swing produced by the previous stage into power. Then, care
is necessary to sec that the stage before has enough swing to drive the
driver.
A point that may be overlooked in working out gain is that a valve
used for phase splitting, such as in the circuit shown in Figure 17, does

34 35
 RESISTOR VALUES

20°/o
4 TABLE I. TOLERANCE PREFERRED VALUES

1--HHHI J-
PREFERRED LOWER UPPER
VALUE VALUE VALUE
RESISTOR VALUES l --.
BROWN BLACK 100 80 120

I N this book, as in many books, a circuit value is specified in

" round " numbers, say 50,000 Ohms. In practice a nominal
value of 47,000 Ohms will be used. Some readers may wonder why
books do not refer to the value as 47,000 Ohms, so the understanding
BROWN
RED
ORANGE
GREEN
RED
ORANGE
ISO
2.20
120
176
180
264
as to what rated values mean is defutitely a matter needing attention. 330 264 396
All commercial resistors are manufactured to a tolerance-they YELLOW MAUVE 470 376 564
arc not " spot-on." The standard tolerance is 20% . This means that
a resistor marked 100 Ohms may be any value between 8o Ohms and BLUE GREY 680 544 816
120 Ohms. Correspondingly a resistor marked xo,ooo Ohms could BROWN BLACK 1000 800 1200
be any value between 8,ooo and 12,000 Ohms. The example chosen
in the previous paragraph, marked 50,000 Ohms (as they used to be
at one time) could be any value bet\veeu 40,000 Ohms and 6o,ooo
Ohms, so that 47,000 is well within the tolerance of resistors marked
50,000 Ohms. The ridiculous situation arose that resistors might be 10°/o
marked, say 40,000 Ohms, 50,000 Ohms and 6o,ooo Oluns, and aU TABLE 2. TOLERANCE PREFERRED VALUES
have the same actual value, within the allowed tolerance! To put an
end to this situation the range of preferred values was instituted, so
that a minimum number ofdifferent ratings is needed to cover the wide PREFERRED LOWER UPPER
range of resistors necessary. ! ---1 ann r- VALUE VALUE VALUE
For the range from 100 Ohms to r,ooo Ohms, this preferred l '--------
value range is shown in Table I. Against each rated value the limits of BROWN BLACK 100 90 110
value set by the 20% tolerance are given. All the other preferred values
are obtained by using more or fewer o's, and altering the third colour BROWN RED 120 110 134
in the code accordingly. BROWN GREEN 150 134 168
Notice 1l:hat the range falling into each rating meets or slightly BROWN GREY 180 164 200
overlaps the next rating. Sometimes .it i~ iruportant that a value shall
be closer to its rating than such standard COfl\lllercial resistors can RED RED 220 200 244
guarantee ; then the ro% range comes in useful. The IO% .range of RED MAUVE 270 243 297
preferred values, costing a little more, is shown in Table 2.
For the majority of purposes one of the 20% values will give ORANGE ORANGE 330 297 363
results of sufficient accuracy. Suppose, for example, that an anode ORANGE WHITE 390 351 429
resistor of 50,000 Ohms is decided upon. Then the preferred value YELLOW MAUVE 470 423 517
will be 47,000 Ohms, and the actual resistor may have any value
between 37,6oo Ohms and 56,400 Ohms, which variation will not GREEN BLUE 560 504 616
greatly affect the performance. BLUE GREY 680 612 748
GREY RED 820 738 902
BROWN BLACK 1000 900 1100

37
 HUM
most noticeable at low levels, but it will certainly not be more notice-
5 able at high levels, as is the effect of interm~d~lation or parasitic
oscillation, which otherwise sounds somewhat Similar.
Hum Pick-up . .
T he foregoing forms of hum are caused by fatrly obvtous means.
HUM Now follow those forms that are more difficult to track down, due to
various forms of pick-up.
(t) MAG NET I C

M
AINS hum in the output from an amplifier can be very objection-
able, and is sometimes difficult to track down or eliminate. One Where an input transform~r is used, or an inte~-valve transform~r
thing that often helps i s recognition of the " type " of hum. between the earlier stages, thts component can pick up a magneuc
The reader may have noticed that there are different qualities of hum, hum field from either the mains transformer or the main smoothing
as well as a difference in loudness, although they are all basically choke. Such hum gives a deep note, almost pure 50 or 100 cycle.
" 50 cycle." To track it down, it is best to earrh the core (and case, if metal)
of the suspected transformer, rem~ving it f~om. the chass.is, k~eping the
Heaters connections intact. Then rotate 1t gently m different dtrecuons as far
A true 50 cycle hum, which is a very deep note by itself, is seldom as the leads will allow. I f the hum is caused by magnetic induction in
loud enough to be objectionable. The most common source is valve the transformer core, this movement will vary the hum level. If the
heaters. If the hum is due to unbalanced earthing of the heaters (if the magnetic field causing the induction is strong, it will be difficult to find
heaters arc not earthed at all they cause a hum that is higher pitched a position where the hum can be eliminated, and pro?ably even ~ore
or " ticky ") then a humdinger across the heater supply, with the difficult to contrive a fixing for the transformer m the re(jmred
slid er to earth, will provide means of adjusting out the residual 50 cycle position! . . . .
hum. I f the hum is due to magnetic induction inside the earlier s tages Input and intervalve transformers for hig~ gam appl!cauoas .are
of a high gain amplifier, demagnetising the valves in a large de- usually mounted in Mumetal cases to prov1de screerung ~gamst
magnetiser, such as watchmakers use for demagnetising watches, may magnetic hum pick-up. This does not. completely prev~nt p1c~~up,
remove the hum. But this process may have to be repeated whenever so it may still be necessary to pay attentiOn to the mountmg postuon.
the valves are changed, so tl1e most satisfactory solution where an Layout is a great help. The input sta~e should be kept .as far as
amplifier has enough gain to cause this particular trouble is to run the possible from the supply end of the chassts, where the mams tra?-
fi rst stage heaters from a rectified and smoothed d.c. supply. formcr and smoothing choke are-if they are on the same chassts.
Aluminium chassis are better than steel ones for this purpose. Another
H.T. Smoothing useful aid in an equipment where the voltage amplifier is on a separate
A bum that is oiten mistaken for so cycles is really chiefly 100 chassis from the output section, is to use mains transformer and chokes
cycles, and is often caused by insufficient smoothing on the h.t. that work at low a.c. flux density. Some manufacturers specialise in
supply, particularly to the early stages. If this is the cause, provision components for this purpose. It is nor practical to work supply com-
of extra decoupling for the early stages is the simplest remedy; but ponents for power stages at low flux density, so if the whole amplifier
before making drastic alterations to include this, it should be " lashed is on the same chassis, the alternative is to mount the supply com-
on" to see whether it proves a remedy. ponents-mains transformer, ch~kes and ~pacito~s-on a. separate
Insufficient smoothing on the h.t. supply can cause another chassis, which can be kept at a sausfactory d1stance from the mput end
trouble, not recognisable as hum. This occurs where practically of the main chassis.
unsmoothed h.t. is used to supply the anodes of push-pull valves. Due (2) I N D UCT I VE
to the push-pull action, no hum is heard, because the two hum voltages Other forms of hum pick-up occur in the wiring of the amplifier.
cancel. (Adequate smoothing may prevent hum being caused in earlier There are two types of pick-up that ea~ caus~ trouble,. and t~c. hum
stages.) But because of the large hum voltage on both anodes of the picked up differs in character so that w1th a httle ~xpenence 1t IS not
output valves, the gain of the valves is modulated at hum frequency, difficult to recognise the type, a~d he~cc go ~ore ~uectly to the cause.
usually 100 cycles. The result on music or speech is that the repro- One type is called electromagnetic, or mducuve, p1ck-up, and produc~s
duction sounds "chopped" or "dithery." This will probably be a deep hum similar in character to the others dealt with so far .. I t tS
caused by the wiring forming a loop somewhere through whtch a

39
 HUM
AUDIO HANDBOOK No. 1 : AMPLIFIERS
(4) EARTHING
num field can induce a voltage in the circuit, which gets amplified Another form of pick-up is caused by incorrect carthiH~. This
along with the signal volt::tges. This type of pick-up is particularly fonn can be very puzzling until understood. Often an amphfier that
~ikely in low impedance circuits, such as the primary of a step-up performs quite satisfactorily on test develops a severe- hum when
mput transformer; but it can also occur in high impedance circuits, installed. Returned to test, the hum has disappeared. The commonest
especially where there is high gain.
cause of this trouble is the method of earthing.
To avoid this hum, the leads from the primGry of the input There must be only one earth path.. All earth ~onn.ections sho.uld
transformer should be twisted all the way from the transformer winding be kept separate from the chassis, to wluch ~onnecuo~ 1s ma~e at JUSt
to the input lead itself. Other leads that can cause trouble of this kind one point, preferably at the input. Connection to an tnstallauon earth
arc Lite anode decoupling and screen decoupling circuits. Leads from should be made at this point only. There may also be an earth con-
capacitors used for this purpose in early stages of a high gain amplifier nection through three-core mains lead. If so, one or other earth
should bt:' twisted together, and the earth retum made as close to the connection must be removed, leaving only one good earth.
" hot " terminating point as possible. Poor earth return arrangement In all mains installations there arc capacitance currents to ea.rth,
can also cause hum for a different reason deAlt with later. and these currents cause small voltages in the earth connecoons
themselves. If the amplifying equipment is c.onnected to tw~ d!fferent
" earth " points in the building from n~o different places m m own
(3) CAPACIT IVB earth wiring, then the small difference m voltage at hum frequency
T he other type of pick-up is called electrostatic or capacitive, and between the different installation earths will send a small current
produces a " ticky ,. hum as a rule. It is due to stray capacitance througn the amplifier earth. Line. . -~his cu.rrent w~l cause voltagcs
coupling from places where there are a.c. hum voltages to signal wiring, along the earth line that w11l be mJCCted mto earlier stages of the
usually at high impedance, but not necessarily. amplifier. . . . .
Complete screening of the amplifier prevents this kind of pick-up The usual effect m a good eqwpment IS : no earth-slight
inside the amplifier, as well as ensuring stability. But this kind of '' ticky" hum; one carth-O.K.; two earths-deep hum: I f there
hunt urn still get picked up. If are earth leakages in the amplifier .itself-say fr.om one s1d~ of the
the amplifier is not earthed, a mains-then anything may happen tn the way .ot h~. ~t will often
tick.y hum may appear; a be found that reversing the mains plug connecuons w1ll gtve less h~m

I
connection to a good earth will one way round than the other. This is particularly the case w1th
cure it. a.c.fd .c. amplifiers.
If an input lead is not
screened, or the screening not
L...---+--' connected to amplifier chassis,
capacitance transfer may take
place through the input Wins-
former, even though rbe input
is low impedance. If screened
fiG 2•. A K~ISTOR IN SfRIE!l WITII lead is nOt available, earthing
TIIIJ RI:SI!RVOJR CAPAClTOR, l'(LOUCES one Sl.dC of the IOW l'rnpedance
ll.AOIA'flON OP" riCKY" HUM.
V>luc> •ho"n are 10 ll"e d roJth auUo only. input is a good second best .
As with magnetic, so also
with electric, it is possible to reduce the effect by cutting down the field
at its source. Reducuon of the magnetic field is achieved by working at
lower flux density in mains transformer and chokes. One source of
static field is the sharp current pulses that charge the reservoir capacitor,
producing peaks of voltage round the supply drcuit, and even back
through the mains. A suirahle resistor connected in series with the
reservoir capacitor (sec Figure 24) will reduce this effect, and at the same
time lengthen the rectifier's life.
41
 SCREENING

6 process), he should see that any Mumetal case he uses has all the
necessary holes, etc., already, so he need not work on it further .
.M.umetal is not s uccessful as a screen against very strong fields,
because it saturates at a fairly low flux density. For tbi5 reason it is
sometimes an advantage for the mains transformer, and perhaps the
smoothing choke as wcU, to be mounted in steel cases. Some types arc
SCREENING finished in steel cases as routine. The steel case round the mains
lrausfonm:r or choke ensu res that the field outside the case is small
HERE arc three basic kinds of screen, ami Lhe requirements for enough for the Mumctal ca~c ruuml Lhc input Lr<lnbformer to l.kal with

T each to be effective differ considerably. It is helpful to under-

s tand how each kind does its job, so that screens may be con-
structeu or used intclligcmly to get the best res ults.
satisfactorily.
Electrom agnetic
Electromagnetic screen.i.ug keeps a magnetic field out by the
Magnetic famous principle of " elecrrornagnctk indu~.:tion." Th~.: variations in
the magnetic field cause currents to ftow in the low resistance material
A magnetic ~creen is made of magnetic material such as steel or of which the screen is made, and these currents create a magnetic
~umeral, and its purpose is to " capture" the magn~tic field and lend field in opposition to the original, so that there arc no magnetic fleld
lt ro~d .the obj ec~ to be scrcen~d, usually a transformer, without fiuctuations penetrating the screen. Figure 26 shows (a) the relation
affecrmg 1t magnetically. M agncttc screens for input and intervalve
between the component fields and current at one instant during an
transformers are usually made of Mumeml. An important feature for
their satisfactory working is that any lids or joints in the screen should
be a good close fit, so as to provide good magnetic " contact." Figure
25 shows (a) the way a magnetic field is led round the screened space by

--- ---
.......... ......
- - - - -- -- -- --
..,.... -·- -~---- ·

---INTERFERING FfELD.
- INDUCED CURRENTS
--c> FIELD DUE TO CURRENTS

(a) COMPONENT FIELDS (b) RESLLTING FIELD

l'IG. 26. eHCCT OF E L ECTR.OMAGNI!TIC SCREEN ON IN rERPICRING

I' IPI.r>.

alternating cycle; and (b) the resulting field, due to combination of

the magncric screen ; and (b) how a poor joinL allows some of the field the components as shown at {a).
to appear inside. So electromagnetic screens are made of copper or aluminium.
Mu~etal has to be heat treated after drilling, filing, or any other With these it is important that any lids or joints should make good
work on Jt, so unless the teader has facilities for getting this treatment electrical comact. There must be no gaps of any kind in the screen.
done (Te/con Metals, the manufacturers of Mumctal, have their own Sheet tinned iron wiU not do for an electromagnetic screen, because
an clecuomagnetic screen must not be of magnetic rnatctial.

43
 AUDI O HANDBOOK No. r AMPLIFIERS
Electrostatic
Electrostatic screening keeps electric fields out (or in). It is
~~cnlial for an electrostatic screen to be well earthed. This is not
7
essential for the other two types for lhcir own purpose; but by seeing
th:~t they are earthed, the one screen mny serve a double purpo~.
For an electrostatic screen it is nor essenNtl that lids or joints should NOISE
make good contact all round, and the screen need not be continuous,
it can have holes in it (e.g., for ventilation), without destroying its wo lcinds of noise wiU be dealt with in this chapter: that known
electrostatic screening properties. (Some valve screens are an example
of this.) Sheet tinned iron (such as a cocoa tin) will serve for electro-
static screening provided there is no electromagnetic effect to be dealt
with as well.
T as valve hiss, and various other noises in the form of crackles
and bangs.
Valve Hiss
Valve hiss is quite a normal form of noise, and cannot be elimin-
Applications ated altogether from high gain amplifiers. H owever, rhere are ways of
The reader should never look upon any :;creen as being perfect. keeping it to a minimum. In the first. place, whatever noise .th~ first
Whatever type is used, it only reduces the eftect by a certain amount. stage develops will always be present af the voluJ?e contr~l .ts an the
If the correct type of screen is used, the amount i~ mually adequate, grid circuit of the first swgc, because the amplifier ilselr JS always
and so efl:cctive that it appears to remove the trouble completely. But if working " flat-out." By having the volume control after the- first or
the field requiring screening is too strong, any screen becomes inade- second valve, the noi~e dne to the early stages will be turned down when
quate, and further steps must be taken. volume is turned down.
It should be emphasised that clcctromaguclic screcnjng only I t should be mentioned that the volume control should not be
excludes a/temnting fields, and in fact is more efficient in doing so at put too late in the ampliJicr either, tor another reason. Suppose
higher frequencies. This characteristic is the reverse of that for different type~ of mic.Tophonc arc to be used, or the same one used
magnetic screening, which is most efficient for steady, or d.c., fields alternatively for long range pick-up or close speaking, then wi~cly
and low frequency alternoling fields. different setting~ of the volume conrrol will br needed. If tl1e amplifier
The purpose ur Sl.tetning has been referred to in earlier chapters, has enough gain to cope with the weakest microphone, or l?ng range
but it \\ill be well here to point out the applications of each cype of pick-up, then it will have plenty in reser~e for thest ron~cr nucrophone
screen. Magnetic screens are almost exclusively used against magnetic or close talking. If the volume control 1s, say, JUSt beiore the output
hum pick-up. Electromagnedc screens help reduce inducrive hum stage, as it is turned well Jowu for the lauer job, so the previous s tage
pick-up, and also remedy instability if it is due to inductive coupling. "'ill be overloading and distorting, before the output from the \'Olumc
Electrostatic screens arc used against electrostatic hum pick-up, and control is high enough to give full output.
provide a remedy for in&tability due to capacitive coupling. In high Valve hiss is proportional to anode current for a ~ivcn type of
gain amplifiers, screening between stages to prevent instability muH valve. At low levels, in the early stages, little gain will be lost by
be both electromagnetic and clectro~tatic. working the valve at lower anode voltage and current; but 1t will
cause considerably less noise. The anode v~ltagc ~.:<~n bl.! dropped by
providing extra stages of dccoupling which will also be advnntngeous
in reducing hum in the b.t. supply.
Crackles and Bangs
T hese arc another sort of noise, and arc not normal. They can
be due to defective valves, but arc far more often due to faulty joints
in the wiring-dry joints. These arc caused by poor soldering, :~nd the
remedy is to use greater care in cleaning the lead and contact point
to be soldered, and in the use of soldering flux and the iron. A good
resin cored solder, with no separate flux, is recommended, after care
has been token to sec that the parts to be joined are cleaned. Both

45
44
 AUD IO HANDROOK No. 1: AMPLIFI ERS
pans should be tinned before being brought together. Ailcr this the
two parts should be brought together, a good mechanical joint made
(by bending over, or inserting through a hole or slot, if provided), and
8
the iron together with enough l>oldcr applied to the juim for just long
enough for the solder to flow all rou nd the joint Do not use too much
solder to make a large " blob/' or too little, ~ that the joint is not
adequately filled with solder. INTER MODULATION
Other contacts beside soldered ones c:an be the eau~~: of lTackles
and bangs) and often are. The shdmg contacts of volume controls N'I'RRJ\.tonutATfON mt>ans that low frequencies in the signal modulate
and switch contacts arc frequent offender~. Cleaning the contacts will
at least provide a temporary remedy. Dust from the resit>tance track
may collect on a volume control slider. Removing th~.: ~.:over, blowing
away rhe dust, and replacing the cover, often gives a new lease ot lite
I high frequencies in the same signal, in the same way as a broadcast
carri~.:r frequcm..y is modulated by tht: audio frequencies in the
pro~ramme. The result is that the intensity of the high frequencies
goes ur and down during l.':tch cycle of the low frequency cau~ing the
to an old control. Noise can also be due to a break or hre1ks in the track modulation. The effect 1s that reproduction goes " d ithcry " or
itself, for which the only cure ts a new control Switch contacts can be " ch uppctl up " when a loud low frequency signal appears.
cle~~ned from time to time, but a good ~witch should uot !Cl!uin: mu~.:h Intcrmodulation, m sul1lctcnt de~ree to cause aurally nouceable
aucntion1 especially if it is enclosed so that dust cannot reach the eiTect, seldom occurs in amplifiers, provided they are free from other
contacts. In a wcll~designed swi tch, the con:acts are self-cleaning in form s of distortion ; but it quite frequently occurs noticeably in
operation, and good contact pressure ensures good cont:\Ct. pick-up~, Jouuspeakcn; 1 anu Ul.:l:l!Sionaliy in mkrophoncs.
Most pick-ups arc ol the magnetic type, using an armature lhnt
moves between two pole pit'Ccs The output from the pick-up depends
upon the changing magnetic field due to this movement. To gain
~e n~i tivily, the pick-up i:s dcl>ignt:d :so that l>mall movement caused by
the needle m the grOO\'C of rhe record prod uces as much change in
magnetic field as pos-.ihlt' Rtlt the bigger the change, the more curved
does the characteristic become) and ~o relative movements due to low
frequencies modulate the intensity of the clc~.:tril.:al output at l1igh
frequencies being reproduced at the same time. It is a general principle
for magnetic type (or variahlc reluctance,~~~ they nrc technically calle<l)
pick-ups, that improved freedom from distortion requires lower
~cnsitivity, or lower output. Another cau~c of intt:rmodulatioo is
non-linear behaviour of the control used for centrmg the armature.
This is usuallv made of rubber or some subHi tutc, which may pcnsh
wi th age, eau~ing distortion not present when the pick-up was new.
Dccausc of non-linear bdlavioul' of the \.Tysutll many piew cly:.tal
pick-ups cause intermodulation at high levels. The modern moving
coil or ribbon pick-ups eliminate intennod ulalion, but have the di~­
advantage of extremely low output, or sensitivity. The light-weight
pick-ups reduce intcrmodulation to a very low order, and have a
reasonable output, although much less than that of the older types.
Loudspeakers can cause intermod ulation for two reasons: ( r)
Uneven magnetic field in the air gap, over the distance Lhrough which
the coil moves; and (2) uneven control fur<.c from the centring
spider. Distortion due to (1) is not easy to check. Distortion due to
(2) cru1 be detected by careful feeling of the action of the cliaphragm.
As the cone is moved back and forth with the fingers, the pressu re

47
 AUDIO HANDBOOK No. r: AMPLIFI ERS
against the fingers, pushing it from its central position, should steadily
increase in both Ulr~:clions. Somctime;:s the spider uevelops an " oil-
can " action, so that it avo1ds the centre position, travelling past it
quickly; sometimes it is not as had as that, hut there is central ;~,one
9
of movement practically without any opposition due to the spider, and
beyond this movement the spickr sud1.h:nly begin~ to show greater
opposition to further movement. Either of these cases produce con- ~1ATCHING
siderable modulation distortion. C.arcful refitting of the ~pider can
HllRB are tWO place-s Wan amplifier where matching is important-

T
often improve such a speaker.
Jmermodularion dinonion i~ nuL t:a~y Lu track uowu witl1ouL the input and the output. The general idea of output matching
special equipment, since it can originate almost anywhere in the chain. is well known, bur often only partly understood. The require-
Any measurement to detect it relies on the use of equipment known ment of input matching is often not realised at aJl.
to be free of ir. The method used to check loudspeakers for inter- The formula for turns ratio in matching is well known,
Z2 T Z ;- -
modulation distortion consists of feeding in two fairly high freque::m:ie::s ~.. = 1 or 1 = '\1 T l
c

at the same time from an amplifier known to be free of distortion, Z2· r~ Z2 T2

coupled to two pure tone audio signal g<!nerators. Distortion in the The abac shown in Figure 2.7 on pages 32/33 is pro\'ided to facilitate this
speaker shows up by an extra note of lower frequency, equal to the calculation. As an examplt: of its u~e, to mau;h 15 Ohms to 6,000 Ohms
difference of the other two, although this f requency is not applied to requires a talio of 20/ I.
the speech coil. Having checked that the distortion is due to the loud-
speaker, it remains to see how it is caused ; but since the. reader may Measuring Transformer Ratio
not possess either the facility to check for distortion in this way, or It is easy to find which windiugs of a transformer are the high
the means of altering lhe loudspeaker, no more will be said here. and low number of rurns respectively> by measuring rhelr resistances
with an ohm-meter. But the ratio of wincling resistances does not
necessarily give any indication of turns ratio, so if the ratio of a trans-
former is not known, another method musr be used. The simplest is a
check with input from d1e mains. A voltage should be connected to
the Wgh nuns side, taking care that the voltage is not too high for the
purpose of the transformer. Then the voltage is measured on both
sides, and the ratio calculated directly from the voltagcs.
Input Matcbing
At the input end of the amplifier, the microphone or pick-up
should have a transformer of the correct ratio to get best results. Use
of too h.igb a step-up ratio will result in loss of quality, although there
may be greater galn : while use of too low a ratio will merely lose gain-
there will be no deterioration in quality. However, too much loss of
gain may mean that an amplifier, that otherwise would have sufficient
gain> has not. This will appear ftom the ex.amplcs that foUow. The
abac of Figure 27 may be used to find a good ratio for any specified
microphone or pick-up impedam.:c. If the impedance is mat<.:bed to
less than so,ooo Ohms, useful gaiJl will be lost; but quality becomes
poor if an attempt is made to match to over 2oo,coo Ohms. The best
plan is to find. a standard ratio that matches somewhere between these
figures.
For example, a certain moving coil microphone has an impedance
or or
50 Ohms. A ratio 40/ I will match it to 8o,ooo Ohtns, OI: one of
sof r to 125,000 Ol1ms, either of whkll would be reasonably suilable,

49
 AUDIO HANDBOOK No. 1 : AMPLIFI ERS MATCHING
only making a dillc rcncc of 2 db. (There is a scale provided on the Tlu.: value usually given by the manufacLU rcr of the loud~pcakc!'
opposite side of the ratio line, giving db gain due to n step-up trans- is not mud1 above the lowest n:ading-ofu~n it~ illlpc.:druJcc at t,ooo
former.) cycles. It is quite a ~implc..:
As another example, a ubbon microphone has an impedance of matter to measure the loud-
! Ohm (0·5). This requires a ratio of 400 I to match it to So,ooo Ohm~ speaker impedance at a few spot
or 500/ r to match it to us,ooo Ohms. If LliD microphone were used frequenc1es. It is connected to
with the same transformer as that used for the moving coil microphone, the output of an amplifier, with a
it would only match to 800 or r,250 Ohms, and there would be 10/ J resistor in series With it, of a
too little step-up-a deficiency m gain of 20 db. On the other hand, value somewhere near to the
using a 400/r or 500[1 rransfol'mer for the 50 Ohms microphone1 expected impedances. A con-
would match it to 8 or nk Megohms, which would result in very poor stant frequency disc, or on audio
quality indeed, giving an extremely " woofy " reproduction. signal generator, provides a con-
Input matching is not highly critical, bur a very w1de range of stant note input to the amplifier
impedances is encountered, so that different transformer ratios are a of ~ve ra! known frequencies,
definite requirement, according to purpose. and at each frequency voltage riG 1J 1\11 I 100 Or ~I EASU R il'\(J
LOU I>\ II A" I R I \1I'LOANC. L.
readings arc taken acroc;~ horh
Output Matching the resistance and the loud-
Output matching requires more careful attention, because in-
speaker. FJgure 29 shows the method. The loudspenker impcdnncc is
correct ratio results in loss of output and considerable distortion. The
given by
problem is complicated by the fact that the impedMce of a loudspeaker
change<; for difT~rrnt frequencies. Z - R X V.
Figure 28 show!. a rypic::tl moving coil loudspeaker impedMce Vr
characteristic. The lowest value of impedance shown, nt about 6oo where R is the value of the :.crie:. resistor of known value.
A fact often not realhcd i:; t.hat the q uality of a loudspeaker
depends upon the output impedance of the amplifier, quite apan
from the matter of matchjng, T he ratio has been calculated so that

I
~~
4ll
when the loudspeaker is connected to the secondary of the transformer,
the output valve(s) " sec" their optimum load impedance in rhe
primary, according to the iormula
zl
Zm
.jT1
T2

L~t~----'-:-='--1-~:---1-:.
where zl is the optimum load impedance, and zl that of the speech
coil. But the extra factor determining the quality of the loudspeaker
reproduction, even if the amplifier has a perfectly flat response and is
50 100 200 matched into its correct load, is the a.c. resistance of the output
valve~.
u~ing pentodes or tetrodcs the a.c. resistance of the valves is
~ 1 \J , !M IMI'II),\ NC"~ C'H ARA<. ll KI S II< OF A TYPICAL \IOVINO COI L
lOUDSPI::A KI R several times the optimum load resistance, usually about five times.
Using triodes the a.c rcshtam.e is usually between one-fifth and one-
cycles, is not much more than the d.c. resistance of the speech coil third of the optimum load impedance. This a.c. n:~i~ tancc is
T he rise in impedance at the low frequency end, sh0\\1\ centred at " matched " through the output transformer; thus, if the a.c. resistance
llhclut roo cycles, is due to the diaphragm resonance. If the diaph ragm is one-quarter of the optimum load impedance, the speech coil
i~ clamped n:sonance disappears, and any alteration to cabinet design impedance will " see," or be looking out of an impedance, on the
that alters rcsonnnce will alter the impedance characteristic at this end secondary or the transformer, one-quarter of its own.
in a corresponding way. The stC}ldy rise aL the high frequen cy end is Figure 30 shows the effect of two typical types of output stage
due to the inductance of the ~rrech coil. The question is: What on the response of the loudspeaker whose impedance characteristic is
value is to be taken as the loud~pcaker impedance ? shown in Figure 28. It is assumed that the loudspeaker gives a perfectly

50 51
 AUD I O HANDBOOK No. r : AMPLIFIERS MATCHING
time or two at .its resonant frequency before coming to rest; noQw
short the speech coil terminals and repeat the action : the tendency to
oscillate will be notict!ably reduced, in fact there will probably be no
overshoot at all. This difl'erence in movement wiil occur whatever
form the impedance conncctecl to rhe loudspeaker terminals takes.
When it is connected to an amplifier, the impedance is the a.c. resistance
of rhe valve(s), matched down by the o11.1tpuc transformer.
Thus if the a. c. resistance of the output valve(s) is low compared
to the optimum load; the speech coil movement will he damped, so as
not to over-shoot. But if the a.c. resistance is high compared to t he
optimum load, the llpt:edi coil will nol be damped appreciably, and
overshoot wi!l occur. The effect of Insufficient dampmg on repro-
duction, as well as over-accentua1ing frequet\cies in the vicinity of
HC. 30. E~H!C J OF Ol. TPUT A.C. RF;SIS lA I'<.:£ ON FltEQUP.~CY RI:SPONS D resonance, will be to make the diaphragm "ring " on its resonant
OF LOLDSPCAKllR. Wl lJI IMPEDANCE ClfARAC'rERJSTIC OF Fl(l. ~.
frequency, rather than play the t.rue note, and, also, whcneve.r a tranllient
appears, the diapb.ragm will " ring" on resonance. .
fiat response when fed with a volLagt: Lhat is constant for all frequencies. From this it is evident that rriodes will give hetter reproductmon
Curve A shows rhe response when fed from a triodc whose a.c. resistance th:.n tetrodes or pentodes, although the: latter are much more efficient
is one-quarter that of its optimum load, while curve B shows the typt:s of valvt:. But mu<.:h ~.:an be done lo ov~;:n;omt: lhdr dcfidc~cy
response fed from a tetrode wJlose a.c. resistance is five times its from the quality viewpoint b)' the use of negative feedback, wht cb
optimum load. In both cases the optimum load is marched for the forms tl1e Sllbject of the ne}.'t chapter.
value of loudspeaker impedance at 6oo cycles. l t will be noticed that
the pentode over-accentuates both low and high frequencies. High
note correction in the form of a capacitor from nnode to cathode can
modify the response to d!at shown in Figure 30, curve C.
At low frequendes in the region of resonance, the actual load
resistance " seen , by the tetrode wiU be much larger than optimum,
which, as seen in Chapter r, will cause distortion. To get over this
trouble the impedance of the loudspeaker al n:sonance should be
matched to the optimum load value. Th.is would mean that the load
impedance at other frequencies will be even lower compared to a.c.
resistance, further exaggerating the poorness of the response shown in
Figure 30 for pentodes or tetrodes.
Dampins
There IS another aspect of this matching question closely related
to this a.c. re~istance/load impedance ratio, cnJled " damping." It has
been stated that the r ise in impedance at the Jow fre<!ut:nde~ is due ro
resonance, which means it is due to diaphragm movcmem. Since
diaphragm movcmcnil has such a pronounced effect on impedance at
these frequencies, it is not surprising to find th:.t the i mpedance
connected to the loudspeaker, resulting from the a.c. resistance of the
amplifier output stage, matched down by the transformer, altecLs the
movement of the diaphragm.
This facr can be veLified fairly simply by disconnecting the
Joud~peakcr altogether, and pushing the diaphragm iu, amllt:Lliug go
suddenly ; it will be seen to overshoot-perhaps it will oscillate a

53
 NEGATIVE FEEDBACK

IO Figure 32 shows a case where negative feedback cannot increase

output at all. Overloading occun ~uddenly. Whatever happens, the
output valves cannot be driven beyond certain limits, where the wave-
form completely flattens. Ap-
plication of negative feedback
NEGATIVE FEEDBACK could not extend this swing in
the output, and so could only

T HrS is a wide subject indeed, and is to have a sepatate book in

tl1is series devoted to it. The purpose of this chapter is to explain
its effect on performance. Many claims have been made for
negative feedback, most of wlUch can be substantiated in appropriate
reduce the distortion until the
output js at the same level as
for the same distortion with-
out feedback. Use of feed-
cases ; but the claims have often been presented in such a sweeping back will cle.an up any small
manner as to lead the reader without sufficient knowledge, to th~: distonlon present before fu ll
conclusion rhat negative feedback provides the answer to all his output is reached ; but when
problems, like a fairy godmoth er. fuJl output is reached-no
bigger than without feedback
Claims for Negative Feedback -distortion comes in just as
One claim that must be seriously limited concerns tJ1e abiJity of
negative feedback to straighten out a poor frequency response. More
as
suddenly without feedback. ~WA~~u~~t.~'Mie~~~'lJ<:?..f.f'l'~e ~~~~
Negative feedback i S use- BACK C,\NNO' I IMI'llOV£ MAX I MUM
accurately stated, negative feedback can make the response of a good ~ b'J' . . OUTPUT.
ful 10r sta lismg gatn, espe-
amplifier better; but, applied to an amplifier with a poor response, it cially when using high gain .valves where the e~act sa.in j~ dependent on
will more likely make it worse. operating voltages to a constderable extent. Thts applicatton offe~dback
The next claim is that negative feedback reduces distortion. This is rather specialised, and does not concer~ the commoner aud1o fre-
is quire true ; but it does not necessarlly mean that it can make an quency applications, so further e:<planauon does not find a place
amplifier give a bigger output. The effect of negative feedback on in thi:. chapter.
output depends on the way distortion builds up as output increases.
Thls can best be nndcmood by examining some typical waveforms. Cathode Follower
Figure 31 s hows : (a) rhe waveform at two different output The most important pr?perty of ne~ative feedback fro.m the view-
levels, from an amplifier whe re negative feedback can inc.rease output point of this book, is its ab1ltty. to alter mpu~ and output tmped~ce~.
slightly; (b) shows a repeat of the higher level, distoned outpul of A wt:ll-known circuit that achteves both ObJects at the same time IS
(a), and also rhe improved waveform p roduced by feedback. the cathode follower, shown in Figure 33·
R 2 is the self bias resistor required by the
valve used. Ro is the "anode" load
transferxed to the cathode lead . Tlteinput
impedam;c, which if R 3 were in the anode
would be R 1 in parallel with the effective
grid-cathode capacitance of the valve, is
multiplied by the gain of t11c valve using
R 3 as anode resistor.
The output impedance (effective a.c.
resistance) can be found approximately,
(a) DIFFERENT (b) HOW NEGATIVE in Ohms, by dividing the mutual con-
OUTPUT U!VELS l=EEC\CK CLEI\NS ductance of th.c valve into x,ooo. Thus
UP HIGHER LEVEL
AT (a)
a valve with a mutual conductance of
2 milliamps per volt w.ill have an effective
FIG . ~l. CIRCUI f 0 1
riG . 31. SAMPI."- WA VbFOK ,\ 1S 1-RQI\1 AN AMPI..IFlE'R WHERE N.EiGATIVP. LA I IIO UE FOLLOWE.It, a.c. resistance of 500 Ohms as a cathode
1-EEOOACK CAN IMI'ROVt: MAXIMUM O U'fPUT Sl.10HTL)', follower. The " anode " load, R 3 , will be
54
55
 NEGATIVE FEEDBACK
AUDIO HANDBOOK No. 1 : AMPLIFIERS
much higher than this, and so need not be taken into account from the
Output Impedance, or A.C. Resistance
Use of negative feedback to reduce output impedance or a.c.
viewpoint of output impedance. resistance is probably its most important feature. From the output
The gain of a cathode follower stage is very nearly one. That viewpoint, there are two kinds, known respectively as voltage feedback
means one volt input gives very nearly a volt output. So the valve and current feedback. Voltage feedback always reduces effective a.c.
does not give any gain in the normal sense of the word. Instead it is resistance, while current feedback increases it, each by a ratio equal to
devoted to changing the impedance at which the voltage is provided. the loss of gain caused by connecting the feedback.
Input Impedance Current Feedback
In effect feedback always exchanges one advantage for another- The simplest, and only popular form of current feedback, consists
it never gets " something for nothing." To raise, or lower, input simply of leaving off the decoupling capacitor across the cathode bias
impedance requires negative feedback that will lose a gain of the same resistor. It should be noticed that while the small negative feedback
ratio as the change of impedance desired. Figure 34 shows methods of thus caused will reduce distortion slightly, it actually raises effective
a.c. resistance, aggravating the effects described in the previous
chapter. It is usually far better to leave the bias decoupling capacitor
in, and use voltage feedback to reduce effective a.c. resistance.
Feedback and Valve Hiss
FEEDBACK There is little point in applying negative feedback over the whole
CONNECTION of a high gain amplifier. The earlier stages do not require feedback to
reduce distortion, and there is the disadvantage that feedback does not
reduce the effect of valve hiss caused by valves within the feedback
section. As more gain must be provided to make up for that sacrificed
in feedback, overall feedback will give more noise at the output than
I !C . ~4. 11 Ll)lli\CK CONNLC l iONS G IVING IN(. IHJASLO
an amplifier without feedback having the same gain. Thus it is better
INPU'l IMPI!OJ\NCC. to .leave the carl.Y s~agcs without feedbac~, an~ only use it over stages
where the levelts Jugh enough to cause dJstoruon and to swamp noise.
~onnection fro'? th~ input viewpoint .tha~ result in increased input UsuaJJy the last one or two stages are sufficient for this purpose.
~mpedancc, whtle Ftgure 35 shows a ctrcuJt that gives reduced input
tmpedance. Voltage Feedback
Figure 36 shows a variety of feedback circuits, all employing
voltage feedback. The feedback connections are d rawn blacker to
FEEDBACK simplify following the diagrams.
CONNECTION At (a) is a circuit suited for single pentode or tetrode output
stages. The stage before may be a pentode or tetrode without involving
any difficulty; but if a triode is used distortion may be caused, as the
feedback reduces its effective anode load resistance.
At (b) is another circuit for a single output valve, having the
advantage of including the previous stage in the feedback, and also the
output transformer. Care must be taken to sec that the transformer
is connected the correct way round, otherwise the feedback will be
positive and oscillation will occur, usually violently. Even connected
the right way, oscillation may occur, if too much feedback is used.
R 1 .and R 2 determine how much there is. R 2 is fixed by being the bias
reststor for the first btage of the pair. In some circuits there will not
be too much feedback if R 1 is eliminated altogether.
I l(i 35. 11 r I)IIAC'K CONNEC·
I ION G I VI NU RfDUC ID INPUT At (c) is a well-known arrangement for push-pull output and
I 'vll'f I >AN( L

s6 57
 AUDIO HANDBOOK No. r : AMPLIFI ERS
NEGATI VE FEEDDACK
drive stages. A special feature is the resistor R 6 which makes the
circuit self-balancing, to compcmatc for inequalities in individual
valves, so that " matched pairs " are unnecessary. The resistors
R1 to R 1 should all be 10°•0 tolerance values, as the balancing action
depends on their accuracy.
At ~d) is a circuit that can be applied by winding extra turns of
fine wire gauge on to an existing output transformer tor the cathode
windings. The two set!> nf wrns must be equal in number, and again
care should be taken llo see they are connected the correct way round.
One way will increase gain and maybe c.:au:.c oscillation, while the other
way will reduce gain. The latter is the required method of connection.
Incidentally this is voltage feedback, not current tcedback, because the
voltage is fed into the cathode circuit instead of being taken from it.
Another way of looking at the arrangement is to think of it as being
partly a cathode follower, due to the cathode windings.

(a) (b)

(d)

I 10 . J6, NEOA'IIVF I l'll)llACK C'IH.CU ITS FOR REDUCING OUTPUT A.C.

RI'S!Sl ANCr.

59
 HO~r 1\llANY
II
WATTS ?
AUDIO HANDBOOK No. 1 :"AMPLIF IERS
'
T HB answer to this question depends largely on another : What
!kind of watts ? The previous chapter explained how negative
feedback can be used to reduce effective a.c. resistance so that
rerrodes and pentodes can be used for outpllt stages, giving quality
very little inferior to that from triodc stages. One poim to be remem-
bered in using negative feedback in this way_, is that it does not really
alter the valve's characteristics. All the while the part of the character-
isti~o-s being used is reasonably near perfect (known a.<\ linear), negative
feedback "does the trick." But as soon as the c urved parts come:: inro
play, negative feedback ceases to be so wonderful.
Notice what happens when feedback is applied to tetrodes,
particularly at the low fi:cquency resonance Qf the loudspeaker. The
load line is still too near horizontal, because of the rise in loudspeaker
impedance. Negative feedback cancels some of the gr\lf input at this
frequency-more than it does at other frequ.encies-so that a shorter
part of the load line is used. Negative feedback helps rhe damping
problem, because when the diaphragm goes to move more than it
should fo.r the current passed to its speech coil, a voltage is caused by
this movement wltich is fed back along with negative feedback voltages.
This produces a current in the output opposing Lhc movement and
preventing overshoot.
But overloading still occurs l'ery suddenly with tctrode or pentode
type valve~, especially when too high a value of load line occurs, as at
loudspeaker resonance, due to the fact that grid voltage curves converge
together at the " knee " of the characteristics. In a triode type valve,
the effect of too much grid voltage input is tOo start grid current, or go
round the anode bend. Neither of these produce such sudden stoppage
o f swing as does the zero grid end of a load line of too high value with a
tetrodc, so distortion does not s lart ~o suddenly. For this reason triode
type output stages give " nicer watts " than tetrodes or pentodes with
the same nominal output.
This affects the How many watts ? problem. Sudden distor1jon
gives the same effect as the speech coil knocking agaiu~t a pole piece in
the loudspeaker, which is very noticeable and annoying. Less sudden
distortion can be tolerated more readily. Consequently, if the output
is limited eo a cerraln absolute maximum, as in 17igure 32, the maximum
should be sufficient to ensure that even sudden peaks do not" go over."
I f suddcu peaks do nor cause such drastic distortion, a large margin
is not needed. This aspect, too, favours Lriodes for quality, although
cost is against thcni. Even so, it is possible to get good quality in small
amplifiers with terrodes more cheaply than witl1 triodes, but a large~· o .. 0
nominal .output must be speet'jiucl. z
00
....

60 61