Effect of the Cathode Capacitor
on P-P Output Stage
ROBERT M. MITCHELL
Because of the fact that some amplifiers use a bypass capacitor across the cathode resistor in the push-
pull output stage and some do not, the conclusions reached in this paper should be of considerable inter
est, Note, however, that these conclusions refer only to Class A amplifiers, while most modern ampl
fiers are designed to work up into the Class AB region.
is one of the resolts of « balanced stage.
‘The Initial Avereach
Figure 1 shows the basic push-pull
‘output cireuit with signal voltages
labeled and the bypass capacitor in
question, Cy, shown in series with a
witch whioh will allow it to be Snaerted
for removed at will during the course
‘of measurements, Tt should be noted
‘hat this ia the eommon type of push pull
circuit, and the entire discussion which
follows is condned to this eirenit and
tio relly and devin, it has
Sroowe swneite forthe auto
soginer to examine some af hep.
Cedars hat formerly were tke largely
for granted. Among thee procedures i
the te of the by-pass capacitor in
Chae pope tpt tage "The pe
oor of thin captor seems evident
fone, yet the engineer i constantly
Saing eres without this component
fd jal ae froguenly fading eeuls
with it 'To make the mater tore Dua
2g aly be wl ead ope authors ed fellows ig conned to thle, cen sod
Socio hat ou abuaately noses Ga Sa RGSS aire ieee ee on erp
sary, and another's that its use is ranged to switch capacitor in or out of | cireuit.
specially to be avoided. ne revit for measurements described. ‘The actual amplifiers used in the teat
Tp view of thia ambiguity of opinion, were the UTC W-10 Williamson ampli-
58 was felt that a test af the difference sommop onthode, it is sino tron that ers, nines these conform almost exactly
would be interesting. ‘Tho reli, whieh since they appear as in-phase inputs at to Williamson's circuit. (See Fig. 2.)
tro not. always, co ntcpaied, are ‘age ty appear as i-phne ipl
Presented ‘in this article, along! with Gh Sri? tes ed 19 he pict Teste
same evaluation of the outcome. The "Another and more serious change is _ The first messurements made were of
nature of this investigation neccesitated ya the even-order distortion terms of total harmonie distortion, using « Geu-
the examination of some of the effects 1 kinds which pass through the eathode eral Radio Type 1032-A Noise and Dis-
of balancing techniques in the Out ution may ers molulte with the lotion Meter, and a low-dstorton
Sage tn well and the ests of tis Dat vata and ne ite Thiet tequency wee 80
ae ot the undertaking are also pre: i2emt wallace and th Go and. messurements were made st
oer which, not necessarily being in phaoe from 4 wat
‘Hib cach otter, may appear it the ona te wer de
J Peblom tpt sampler ing 561
ee ‘As a representative of the other ("‘un- stage, and gave very consistent results.
Proponents of bypassed cathode op- bypassed”) school of thouzht, William- The addition of the bypass espacitor
qation maotain, among other thingy, san sate {Wirese Word, May sleeve icrewed the ifrton, rgard-
‘hat since the even-order harmonic terms 947) “A feature of this arrangement less of power level. The increase was
which are generated in the output stage js that the valves operate with a eom- very slight, but nevertheless very definite,
of necessity pass through the common mon unbypassed cathode bias resistor, The total distortion was small, so it be-
eathode resistor, they therefore sppetr which ussists in preserving the balance came very difficult to measure, expecially
between grid and cathode of both tubes, of the stage under dynamic conditions,” since the distortion of the oscillator it-
sre thus introduetd at igs, e000 Sod inthe August Toad ‘neue’ "Due fo wef wes of te sume order of magni
fear in he output. While i gute trie tke ae of common unbyresed ressors as that of the amplier. Consequently, it
Sha ta even harmonic tes do appear for the posh pl sage, the amplifer mas deed to mean theperformanee
ss voltage drop from each grid tothe yen elf balancing to sigaal.--” widoat feedback. This ctange would
Here, then, are te two opporing views Increase the distortion, of course, bat
vu tie fabs ual te ensoor nivenced would not offek the ction of the
: foreach by their respective proponents. capacitor. ‘The revlls of this tit. were
The meaturements discribe bere ‘ithongh the retsns advanced sem to similar, with distortion increasing When
ese alt i Iterator of the Satya, iy semonnble to auune the apace wn ued, (See Fg).
ed caraptormer Company Oring that redotion of distortion iv alse the These reals were obtained consent
Me pee rien t stat sved hy Watinnon, nines thin after ourerour heske and reebok‘ding tests at higher frequencies (600
tnd 2000 eps), ‘Several ‘curves were
rawn and the date wan about to De nae
tembled for write-up when it waa decided
{o eubstitate woe other type of ontput
‘tubes and se if there was any difference.
ly, a pai of 1618's was subst
the teste re-ran. The revlla
‘were as complete a reversal of the trend
tr could be imagined! Almost every teat
showed lower distortion withthe eapaei=
for én the cireit. A typical messuremeat
is Ulasteated in Fig. 4. In view of such
‘onfliting resis, ib was decided to re-
measure with as) many different (ube
types and amplifers an possible. Conse-
‘quently, the tect sehedule outlined below
ras evolved.
(1) Pour different stock ampliers
with 1614 tubes were checked for total
bbarmonie distortion ona distortion
ameter,
(2) The same four were checked for
Individual harmonie distortion compo-
‘nents on a wave analyzer,
(8) The meamurements of (2) were
made with different degrees of current
‘unbalance
(2) Ove amplifier wae checked with
four diferent ets of utp ties al of
face diretlyinterebangeable in the
UTC W-10. These four seta indaded
fo pairs of 1ais's and one each of
KsTet's and 5881's
(3) The amplider of (4) wae
checked for intermodulation distortion
with the three tube types mentioned
bore,
18) The amplifier of (4) was ex
tunined for transient distortion by the
square wave method, with differing de-
ies of current unbalance,
(7) All measurements were made
with the 100.R8 bypess capacitor
switshed in and out of the clveuit and
the couparisoas made pont-by-pofat on
fs ASD lai Care wa taken Uo prevent
‘any transient disturbance during eonnee-
tion or diseonnestion of the expacitor
from being ineluded a recorded data,
eats
‘The outcome of the harmonic distor-
tion measurements of (1) was very in-
conclusive at low frequencies. The addi
ton of the capacitor either reduced oF
Increased the isbortion depending va
‘which type of tobe was used, and even
‘Varied among tubes of the same type. At
the higher frequencies, however, the
nddition of the capacitor quite consi
svaced the distortion.
‘This ambiguity of reslts pointed up.
the need for'a more refined analysis, so
the next step was to measure the individ-
unl harmenie components on a ware
analyzer. A fundamental frequency af
50'eps was chosen, and the second and
third harmonies were checked, with out-
put tube currents adjusted to" produce
there four diferent conditions:
(2) Misimum wubalaueed current
(2) Minimum Srdbarmonie in the
oxtpat
(3) Maximum unbalanced current
in one direction
(2) Moximom unbalanced corrent
in opposite direction
Again the reslta were indefisite, with
diferent tube types or diferent tubes of
the same type giving different results,
fia wo prepuuderauce uf senlle une way
for the other. Figure 5 shows a graphic
‘comparizon of the results of this test for
‘to partieular pairs of output tuber,
‘An interesting, and rather unexpected,
Sniling was that tho oapacitor made ®
(greater relative change in dstortio, thc
lover the system was to balance Wb
(greatest aliferencas inthe two eon!i-s
{king place when the currents wer =
{usted to produce « minimum of 'h
hharmonie, This is shown clearly in
‘The condition for minimum thini lr
tmonie waa ali found to oatar very ow
to that for minimum second harmon
close in faet that the two were >
cally coincident.
or example, when the currents were
sdjonted for minimsm wnbalanen oF
‘inimam third harmonic, addition ofthe
capacitor produced changes of the order
of two or four to ase, When the enrrents
were greatly out of balance, however,
the addition of the capacitor caused
changer of the order of only 10 per cent
for less, although the distortion terms
Were muek larger, of course, Ie was ale
Found that the coadition for minimun
Sreramenaw eT
Do ma
Bes m eames,
Onn mrarel
ce
Cols
Fig. 3. Hormonie distortion curves on am
plier of Fig. 2. with and without bypors
capaci.
‘unbalanced earrent wat generally net
‘the condition for minimum harmonie ds
tertion, This isnot surprising, sine the
{net thet the two tubes are in tate ha
lance (Ge. conditions) does not mean
that they are also balanced dymanicalle
(ae conditions), The purpore of the
Delaneing errangement in the “Wiliam:
fon” amplifier is primarily to winimiae
‘the unbslaneed de. eurreut in the outpet
transformer primary, and thereby ine
erence the lov-frequesey respons, while
simultaneously reducing core saturation,
Invermodulation Distortion
‘The ment step was to measure the fa
termoduletian distortion of the sxpliir,
sand since tho method of measurement
‘was not that most widely used, a brief
disevssion of the technique will be. of
Tnlermodalation distortion oocura when
two or more frequencies interact fo ns to
produce frequency components whieh are
proportional to the peodost of the input
Frequencies. One result of such a rele-
tionship isthe production of frequenciesTYPE Je pain
rowr
eee
ay
:
i
5
z
qual to the som and difference of the
tro input frequencies, or of one input
and harmonies of the olber, ele, This
tution is the basis of our common system
ef amplitude modulation, where the pro-
dustion of such sidebande is « necessary
Telace of Whe process, Ta the commonly
‘sed (and niost familiar) method of in-
ferwodulation tests, a high-frequency
tae {the carrier) is modulated by a low-
requeney tone, and the resulting side-
inde dotocted and mensured hy site
‘le circa,
Although this method is an exelent
indication of nonlinearity which gon
ally correlates elosely with listening tests,
‘t vas found that another method gave
resus whieh sometimes seemed t0 corre:
lste even more closely. This second
nnethed of intermodulation distortion
measurement also vtilizes the fact that
tilference frequencies are produced, Im
‘bie method, gunorelly ealled the COIP
rurthod, two high-frequency tones of
qual amplitude end separated in fre-
‘qeney by a fixed amount are applied to
the device under test. The output is
rueasured by a wave analyzer tuned to
‘ne of the distortion frequencies such as
Je fo Be fv 08 2, =f, ad the distor-
ion voltage ia expressed aa a percentage
of the sum of the two input voltages
Figure 6 shows the two types of inter:
‘uudulation is graphle form. One notiew-
thle fenture of the CCIF type is the
location of the differenoe-frequeney com
potent f,~ fe. Sinee thin component is 00
far removed from the two input frequen
cvs thar ig Inea tondoney for the nignsl
tomssk the distortion as there might be
in the SMPTE method, or in the CCTF
nuethod for the 2f,~f, und 2f,~f, terms.
This difference tone is particularly ol
Jectionable when it falls in tho region of
iaarimum sensitivity of the ear (400 to
5000 epa), and is notieeable in amounts
sa all asa fraction of one per cent. Tt
must be pointed out, however, that doe
tthe difference in method of expressing
percentages in the two methods, equiva
lant peremtsges are wnt indicative af
equivalent degrees of nonlinearity. The
SMPTE values appear relatively high
while the COIF values appear relatively
low for the same nonlinearity?
Ta using the CCLF method of measure
ment two procedures are commonly used.
One is to select two fixed input frequen,
cies aud measure the distortion as the
power output is varied. The other is to
Select a fixed power output level and to
‘ary the two luput frequsscies sinule
taneously, maintaining a constant dif
ference frequency. The frst is rather
readily accomplished, whereas the second
requires cither a special omilator such
fas the General Hadio 1303-4, oF two
‘seillators calibrated with sufficient e-
PERCENT Tmo HARREME
5
Fig. 5. Comporative distortions with dif
ferent typer of tubes in the output stoge
‘show different results
earacy to enable the difference to be
readily distinguished from the dial set-
tings. ‘The stock oscillator in the UTC
laboratories is a deende-type oscillator
with an securacy of four places, and
‘consequently is ideal suited to any such
application.
‘Two such oscillators are used with the
simple mixer-potentiometereireait shown
in Fig. 7 to provide control of Erequency
and output of either oscilator and the
overall voltage outpet, The isolating
resistor networks prevent any interac:
*A discussion of the two methods of
intermodulation distortion measurement
will be found in the General Radio Com.
any “Technical Publication B-3, The
Measurement of Nowlinzar Distortion,
by Ammold Peterson.
7. Method of connecting two au
tnillatrs 10 provide CCI type of UA wig.
ok without Introducing intermodulation
in the mixing circu,
tion, an consequent intermodulation,
Tetween tbe two enistors
“The CIP inermolelaton tert were
performed sing bath methods tentioned
hove. When tortion was measured
i Tanstion of power lve, te following
Felttne were taken into einsieretion!
Sine to Eeauencna. ane ieniend,
conpies {wonsiuscdal) wave 1s pro-
Shue, and the indieation of a2 ordinary
‘zcauin tube voltmeter i, therefore, not
ntti detaining tbe power level by
to enstommry formula P= B*/R. Fore
thermore, unc one wave rides fe ole,
the peak value of the two waver may
race value egoal fo the algebra sum
OF the invidual waves with the result
{bat overuad en occur for tno fregien-
fhm wl ac i ony onechld fhe ams
Pitde ruined for overload by a single
frequeney. (See Fig. 8) Since cach wave
isonly one-taf the taxinain ampli,
it tam. produce only one-quarter the
Jower-of nance maximus amplitude
Mave. The power ssl fon feo sich
Mraves withost aay pomblity of over
Toul ior therefore, only one-half. the
jer avaible from a singlestrequeney
Rave The carves forthe CCT intermods
itdow tet are caravan tee of
Nolte av read onan aio trequeney
‘Continued on along pe)
Fig. 6 Comparison of frequencies pres
Cent in two different types of intermody-
lation measurements.ag
[caer
maximum voltage resulting from mixing
two. signols of lightly different fre-
‘quency and of the same amplitude, ot In
the CCIF method.
vyecuum-tube voltmeter. In seeordance
‘with the above, maximum power level is
equivalent to an output vollage reading
‘of about 933 volte on these curves,
The eurves of Fig. 0 were obtained by
‘maintaining the indjeated input frequen
cies constant, varying the input voltage
up to overload and above, and messuring
Ue Msst-order differenes frequency. The
tubes used were 1614's and the difference
frequency was 400 cps. Notice that al-
though the difference in distortion i
slight, it is slmost always lower when
the authode resistor jn bypassed. ‘The
same revulte wore obtained with 5881's
‘hd KT-66's,
"The curves of Fig. 10 were obtained
by Keeping the inpit voltages constant
and varying the input frequencies. This
vwas done for three dfference-freqaen-
cies, providing a consideruble amount of
ange overlap as shown, The tubes used
were KT.60's. Again itis seen that the
‘capacitor effects a slight but definite im-
provement. (Note thar cach difterence
frequency has separate distortion ordi
‘As a final check the amplifier square
wave response was observed for different
(Amounts of current unbalance, With wn
balanced eurrents up to 10 ma in either
‘ireetion there was no discernible differ
‘ence in the ovtpat wave shape ar the
fapacitor was added or removed a! fre-
9. Effet of copactor on CCIF inver-
‘modulation mearurements at two differ
‘ont oreat ofthe framuency spectum,
Fig. 10. Effect of capacitor on CCIF inter
‘modulation meosurements whan differ
fence Frequency tr varied. (The three pairs
of curves are plotted to different base
Tinea)
‘quencies from 20 to 20,000 eps. At ios
frequencies there was a noticesble round
ing of the trailing edges as the unbalance
exceeded 10 ma. (See Fig. 11} These re
sulls were obtained with any of the three
types of output tubes.
From the foregoing experiments st
least one curious resolt stands oul: fn
Class A amplifier the use of a bypass
‘capacitor across the output eathode gen
erally reduees the intermodulation dis:
tortion, although it may either decrease
or increase the harmonie distortion,
The decision ss to whether or not to
tse such eapmritvr depends unity om
the magnitade of the distortion. IE itis
‘very small, then it may be safely left off
‘with no possible audible difference, Tie
is only moderately low, then the use of
bypnea capacitor ie nlvienle
All of the foregoing applies to a Class
A-amplider only. In the case of a Clase
‘AB amplifier the bypass capacitor it ab-
‘olately necessary if the amplifier ie to
Pegg wih the odes nits of
igh deity performance.
aa
1
Effect of current unbolance on
Ware waves, Note that only the low
frequency is affected, as indicated by
‘ounding of wailing adge of wav