S O M E N O T E S O N NOISE T H E O R Y A N D

ITS A P P L I C A T I O N T O I N P U T CIRCUIT D E S I G N * !
BY

WM. A . HARRIS

Tube Department, R C A Victor Division,

Harrison, N, J .

Summary—The mechanism by which noise is produced in an electron

tube and the relation bet-ween induced grid noise and plate noise are dis-
cussed. An equivalent circuit with noise generators supplying voltages and
currents to simulate noise derived from the plate current of a tube, from
the grid by passage of this current, and from the input circuit is then
analyzed to determine the optimum noise factor obtainable under various
conditions. The frequency for which the quantity R g is unity is seen
eq in

to be an appropriate figure of merit for the noise produced by an electron

tube. The frequencies corresponding to chosen values for the noise factor
are presented for several receiving tube types. The paper concludes with a
discussion of the circuit requirements which must be satisfied in order to
obtain noise factors approximating the theoretical values.

INTRODUCTION

~ Y " O I S E generated i n the first tube of a receiving system is fre-

quently the factor controlling the over-all sensitivity of the
-a- N system. A n understanding of the mechanism by which such
noise is produced is helpful i n the design' of receiving equipment, par-
ticularly w i t h respect to the choice of tube types. I f the electrons i n a
tube were to leave the cathode at a perfectly uniform rate, there would
be no noise, or at least, none i n the frequency range i n which a tube is
useful. The rate o f emission of electrons, however, is not uniform. In
r
any given interval of time there are probably a few more or a few less
electrons leaving the cathode than the average number for that amount
of time. The classical shot-effect derivations predict the magnitudes
of fluctuations of this sort. Furthermore, because theory and experi-
mental data have revealed the extent to which space-charge effects can
reduce these fluctuations i n electron tubes, the noise components of the
1
plate current of a tube can be computed i n many instances.
A t h i g h frequencies, the fluctuation current induced i n the g r i d of

* Decimal Classification: E138 X R161.6 X R361.211.

f Reprinted from RCA Review, September, 1948.
1
B . J . Thompson, D, O. North, and W . A . Harris, "Fluctuations in
Spaee-Charge-Limited Currents at Moderately High Frequencies", RCA
Review, V o l . I V : No. 3, pp. 69-285, January 1940; No. 4, pp. 443-473,
April, 1940; V o l . V : No. 1, pp. 106-124, July, 1940; No. 2, pp. 244-260,
October, 1940; No. 3, pp. 371-388, January, 1941; No. 4, pp. 505-524, A p r i l ,
1941; Vol. V I , No. 1, pp. 114-124, July, 1941.
280
 NOISE THEORY AND INPUT CIRCUITS 281

a tube by the passage of the fluctuating plate current through the g r i d

is another noise source which must be considered. T h e magnitude of
the mean square of this current is proportional to the component of
2
input conductance due to transit time.
Since the g r i d noise and plate noise are derived i n part from the
same current fluctuations, they cannot be treated as entirely inde-
pendent noise sources. Nevertheless, valuable w o r k i n g formulas and
principles have been derived for conditions under w h i c h coherence
3
between g r i d noise and plate noise can be ignored. It is possible, more-
over, that the improvements obtainable by taking coherence into
account are not very important for the majority of tubes and circuits
i n current use. Theoretical considerations, however, indicate that a
substantial improvement i n noise factor may be obtained by t a k i n g
advantage of the coherence between g r i d noise and plate noise i f the
conditions assumed for the theory can be realized i n actual tubes.
Experimental evidence shows that at least part of this improvement
can be obtained i n a practical system.*
In this paper, the relation between induced grid noise and plate noise
is illustrated by an examination of the result of the passage of a single
electron through a tube. Then, the conditions giving optimum noise
factors are derived, using the methods employed by H e r o l d and others.
3
Herold showed that the noise factor f o r a tube is a function of the
product R g where R is the equivalent noise resistance and g i s
eq int eq in

the input conductance. I n this paper, the frequency for w h i c h the

product R g is unity is recommended as an appropriate noise "figure
eq in

of merit'' for a tube. The ratios of the operating frequencies to this

reference frequency therefore can be used as the abscissas for curves
of optimum noise factor.

CURRENT IMPULSES FROM O N E E L E C T R O N

F i g u r e 1 shows the distribution of potential i n a parallel-plate

triode. The potential curve is based on the assumption o f a Maxwellian
distribution of i n i t i a l velocities, w i t h a cathode temperature of approxi-
mately 1000 degrees K e l v i n . T h e dotted curve represents the velocity
of an electron w i t h just enough i n i t i a l velocity to allow i t to pass the
point of m i n i m u m potential and continue to the anode. T h e time of
2
D. O. North and W . R. Ferris, "Fluctuations Induced in Vacuum-Tube
Grids at High Preqencies", Proe. I.R.E., Vol. 29, pp. 49-50, February 1941.
3
E . W . Herold, " A n Analysis of the Signal-to-Noise Ratio of Ultra-
High-Frequency Receivers", RCA Review, V o l . V I , No. 3, pp. 302-331,
January, 1942.
4
M . J . O. Strutt and A . vanderZcil, "Signal-Noise Ratio at V H F " ,
Wireless Engineer, V o l . 23, pp. 241-249, September, 1946,
282 ELECTRON TUBES, Volume 11

F i g . 1—Potentials and velocities in

a parallel-plane triode.

transit for such an electron can

be computed by a graphical i n -
tegration process.
When a charge is i n motion
between two electrodes in a
tube, current flows between the
electrodes bounding the region
containing the charge. F o r a
parallel-plane structure this current is proportional to the velocity of
the charge and inversely proportional to the distance between the two
plane boundaries; it does not depend on the position of the charge
5
relative to the boundaries. The velocity curve of F i g u r e 1 is applicable
for a charge of the indicated i n i t i a l velocity. I f the velocities for
various points in the cathode-grid space are divided by the distance
between cathode and g r i d , and the velocities for various points in the
grid-anode space are divided by the distance between g r i d and anode,
quantities proportional to the current due to the motion of the charge
for these various positions are obtained. Then, the relation between
position and time obtained by integration may be used to obtain a
current-time curve.
The curves of current versus time for the tube structure of F i g u r e
1 are shown i n F i g u r e 2. The solid curve shows the current to the g r i d
and the dotted curve the current to the plate which would result from
the passage of one electron. T h e choice of the velocity of the slowest
electrons which can reach the anode leads to a computation difficulty;
the time required for such electrons to pass the potential-minimum
region is theoretically infinite. Consequently, transit times are com-
puted from the cathode to a point near the potential m i n i m u m on the
cathode side, and f r o m the g r i d back to a point near the potential
m i n i m u m on the g r i d side. The three rectangles between the ends of
the two curves show the times and currents for charges passing between
the terminal points of these curves w i t h velocities exceeded by 90, 50,
or 10 per cent of the electrons reaching the anode. T h e use of one of
these velocities would cause some change i n the remainder of the curve,
both i n the current and the time scale, but the shape of the curve would
be about as shown. T h e effect of a change i n initial velocity on the
current between g r i d and anode would be almost negligible. The
cathode-to-grid transit time for an electron with an i n i t i a l velocity

5
S. Ramo, "Currents Induced by Electron Motion", Proc. I.R.E., V o l .
27, pp. £84-585, September, 1939.
 NOISE THEORY AND INPUT CIRCUITS 283

corresponding to one of the rectangles i n F i g u r e 2 is, therefore,

approximately the sum of the transit times represented by the two
curves and the appropriate rectangle. The indicated range is from 6 to
10
8 X 10- seconds. The t r a n s i t time from g r i d to anode for the condi-
w
tions of F i g u r e 2 is about 1 X 10~ seconds.
The curve of F i g u r e 2 does not show the compensating effect which
takes place when an extra charge passes through a tube. The potential
minimum is depressed by an amount depending on the position of the
added charge, d u r i n g the whole time this charge is between cathode
and g r i d . The result is a reduction of the current, which can be con-
sidered equivalent to the passage of a series of charges of opposite sign
between potential m i n i m u m and g r i d , and the passage of charges of
the same sign but of opposite direction between potential m i n i m u m
and cathode. These effects account for the shot-effect reduction factor
1
computed by N o r t h . It appears, however, that between the i n i t i a t i n g
pulse and the compensating current there is some time delay which
may be important i n the determination of the noise at very h i g h
frequencies.
A s soon as the extra noise-producing charge leaves the cathode a
small effect on the m i n i m u m potential w i l l be noted; some electrons
which are reaching the potential m i n i m u m at this instant t u r n back
instead of continuing toward the plate. The effect of the extra charge
persists until i t reaches the g r i d . The compensating effect cannot be
completed until the time at which an electron, turned back because of
the depression of the potential m i n i m u m when the extra charge was
near the grid, would have reached the g r i d had it not been turned back.

TIME - SECONDS X 10

F i g . 2—Grid and plate currents due to passage of a single electron.

284 ELECTRON TUBES, Volume 11

The compensating charges i n motion after the passage of the causing

charge, however, must themselves have an effect on the minimum
potential. The complete result, consequently, is probably of the nature
of a damped oscillation w i t h a period related to the transit time of an
electron from cathode to g r i d . Thus, the curves of F i g u r e 2 do not
present a complete picture of the generation of noise i n a parallel-plane
triode, but they do show some of the characteristics of the basic noise
impulses.
It is pertinent at this point to discuss the extent to which g r i d and
plate noise currents can be made to cancel each other. The pulse shapes
are quite different, so i t is evident that complete cancellation cannot be
expected. P a r t i a l cancellation may be obtained i f a voltage is developed
at the g r i d by allowing the g r i d current to flow into a capacitor. T h i s
voltage is proportional to the integral of the grid current and has the
effect of momentarily reducing the plate current. A suitable choice of
capacitor value can give a plate-current pulse of zero net area for
electrons of a particular i n i t i a l velocity. However, there w i l l always
be some conductance i n the g r i d c i r c u i t which w i l l result i n a grid-
voltage component tending to increase the noise output. Moreover,
electrons leaving the cathode w i t h velocities too low to allow them to
pass the point of m i n i m u m potential produce pulses of g r i d current
without producing corresponding plate-current pulses.

D E T E R M I N A T I O N OF F R E Q U E N C Y SPECTRA
T
The curves of F i g u r e 3 illustrate the method by w hich the frequency
spectra corresponding to the grid-current and plate-current pulses may
be obtained. The current which could be measured i n the small fre-
quency range represented by do> at a frequency « / 2 T T is obtained from
the F o u r i e r i n t e g r a l :

2 r 2 r
Ad& = ~
— /I F,y.
-FV\ cos t&\
} to\ dX cos o>t
o\t d
du\ + —
M / iF (\> sin wA dX sin «>t du.

(1)

The function F ( represents the pulse.

M

When the frequency is low i n comparison w i t h the reciprocal of

the transit time, the value of the term cos o>\ i n Equation (1) is nearly
constant over the region i n w h i c h F has a value other than zero.
( X )

In addition, the value of the term sin o>\ i n E q u a t i o n (1) can be repre-
sented as a straight line w i t h a slope directly proportional to the
frequency over the same region.
Because the g r i d pulse has equal positive and negative areas, the
 NOISE THEORY AND INPUT CIRCUITS 285

Pig. 3—Development of frequency GRID CURRENT PLATE CURRENT

spectra.
1
integral containing the cosine
terms is zero for any frequency
low enough so that cos u>A can be
I cosw\
1
\

considered constant. The inte-

MtfiGUCT
gral containing the sine terms MOO OCT

\
will have a value w h i c h i s repre- * 1
sented graphically by the area under the curve labeled "product" i n the
grid-current curves of F i g u r e 3. T h i s area will be directly proportional
to the frequency because the slope of the sin <n\ line is proportional to
frequency. It w i l l also be proportional to transit time because, i f the
areas and shapes of the parts of the F curve are maintained constant
( X )

and the base line is extended, the area under the "product" curve w i l l
increase i n proportion to the increase i n base-line length.
The plate pulse w i l l give a zero value for the integral containing
sine terms i f a suitable point of o r i g i n is chosen. T h e integral con-
t a i n i n g cosine terms, then, gives the current, which is independent of
the frequency when the frequency is low. Because the area of the
plate-current pulse represents the amount of charge producing the
pulse, the current i n a small frequency band resulting from a given
amount of charge is also independent of the transit time.
The mean-square noise current measurable i n any frequency band
results from large numbers of pulses distributed at random w i t h
respect to time. F o r the plate current, consequently, the mean-square
current dP in a frequency band of w i d t h df can be represented by the

3
equation di = k df 2 (2)
2 2
and for the g r i d current, by dp= k w - dfx (3)
2
where r is the transit time, or dP = fc B df t (3a)

where 9 is the transit angle.

The electronic component of input conductance is proportional to
a
the square of the transit angle , so a proportionality between the mean-
square noise current and the input conductance is indicated, thus:

dp = k s 0! df. (4)
2
N o r t h and F e r r i s found that the complete relation f o r grid-current
noise is dp — Q g • 4k T df
x x 0 (5)

* D. O. North, "Analysis of the Effects of Space Charge on Grid Im-

pedance", Proc. I.R.E., Vol. 24, pp. 108-136, January, 1936.
 286 ELECTRON TUBES, Volume 11

where & has a numerical value of approximately 5 when the cathode

l

temperature is 1000 degrees K e l v i n and the reference temperature T 0

is approximately 300 degrees K e l v i n . The dependence on temperature

T
is discovered only w hen the analysis is extended to include the com-
pensating currents.
The plate current noise from a tube may be represented as i f it
were derived from a noise voltage at the g r i d sufficient to produce the
1 2
noise current. The appropriate equations are de = AkT R df (6) cq

where, for oxide-coated cathode tubes, theory indicates approximately

that, for triodes, R ~2.5/g eq m (7)

and for pentodes (8)

F i g . 4—Equivalent circuit.

CIRCUIT A N A L Y S I S

The circuit of F i g u r e 4 represents the replacement of a real tube

by a fictitious noise-free tube with zero input admittance and suitable
noise generators and external circuit elements. The plate noise is intro-
m
duced by a constant-voltage generator delivering a voltage e series
2

w i t h the g r i d . The noise current to the grid, i , is represented by a

x

constant-current generator across the g r i d circuit. The noise from

the input system, i 0)is represented by a second constant-current
generator. The plate-noise generator can be replaced by another
constant-current generator; the voltage output of the plate-noise
generator is multiplied by the total admittance of the input circuit to
give the required current u- The relations between the noise currents
and the tube and circuit parameters are given by the equations:

(9) (10)

(ID

i =
0 K\/e g , Q 0
(12)

where K = V^kTAf
g v is the electronic component of input conductance;
 NOISE THEORY AND INPUT CIRCUITS 287

R eg is the resistance equivalent for the plate noise, referred to

the g r i d ;
$ t is a multiplier relating g r i d noise to input conductance; its
value is approximately 5 for tubes with oxide-coated
cathodes;
6 is a multiplier representing the ratio of antenna noise to
a

the noise i n a resistor at room temperature;

g 0 is the antenna conductance, referred to the g r i d ;
B a is the net susceptance of the circuit at the operating fre-
quency.

It is assumed that the conductance g can be varied a r b i t r a r i l y by

tl

some such means as a variable-ratio transformer between antenna and

g r i d . Also, it is assumed that means such as a t u n i n g capacitor are
provided so that B can be varied a r b i t r a r i l y . Ohmic losses i n the input
0

circuit are neglected.

The quantity (— /) in parenthesis i n the expression for i indicates x

that % may be in quadrature w i t h e over a specified frequency range.

2

The preceding discussion suggests that this assumption is legitimate

in the case of a triode, when the frequency is not too h i g h and the
frequency band is not too wide. The assumption is not valid, however,
for a pentode because i n that case the larger part of the plate noise
1
results from the division of current between plate and screen g r i d ,
and consequently i t cannot be correlated w i t h the g r i d noise.
The total mean-square current from the three generators of F i g u r e
4 can be found as follows: A d d i and i , t a k i n g coherence, i f assumed,
x 2

into account. Then, determine the sum of the squares of i , the real 0

-
part of (ti +* t ) , and the imaginary part of ( h 4- iz) When coherence
2

is not assumed, simply add the mean-square values of i , i and i 0 u 2

The results follow:

When a quadrature relation between g r i d and plate noise i s assumed,
the mean-square current is

2
** = K*{g B + R , a 0 e + g ) * + (B \/R~T -
0 9 Q V^ffi) } (13)

When no coherence is assumed

i
¥ = K*{g 9 + E „ 0 0 +g ) 0 + R e<1 B;- + 0,g,} (14)

O P T I M U M NOISE FACTORS

Optimum performance w i t h respect to noise is obtained when the

term g $ is as large as possible i n comparison w i t h the other terms,
6 g
288 ELECTRON TUBES, Volume II
7 8
and, i n fact, the noise factor as defined by N o r t h , F r i i s , and others is
2
obtained by d i v i d i n g E q u a t i o n (13) or (14) by K g 0 and assuming 0 o

e — 1. T h e first step i n finding conditions for m i n i m u m noise is the

0

adjustment of B to eliminate the term i n which i t appears i n either

0

equation.
Then, either equation can be differentiated w i t h respect to the ratio
a n
g^Qo d an optimum value of noise factor can be obtained. T h e noise
factors after adjustment of B are given by the equation 0

/ 9i 9o \
NF^-l+R^g! — + — + 2) (15)
\ g ffi0 /

when coherence is assumed, and the equation

/ 9i 9* \ 9i
NF = 1 + R 9i eq ( — + — + 2 4- 0i — (16)
\9o 9i I 9o

when coherence is not assumed.

The m i n i m u m noise factors, w i t h the conditions for obtaining them
are

^ = 1 + 4 ^ ^ (17) — = 1 (18) B ^ = e g yR g
0 l 1 m 1 (19)

when coherence is assumed; and, when coherence not assumed

2
NF = 1 + • 2 R ffi + 2 VOiR^g,
etf + U^ffi) (20)

— = VR eq 9t/ih + K 90
Q (21) B =Q0 (22)
9a

The quantities R^ and g are both tube parameters Since they

x

appear as the product R g i n Equations (17) and (20), the magni-

eq t

tude of this product indicates the noise performance obtainable from

a tube. T h e quantity g however, varies w i t h the square of the
lt

frequency. F o r purposes of computation, i t is preferable to use as a

reference parameter the square root of the product R^ g which lf

f D. 0 . North, "The Absolute Sensitivity of Radio Receivers", RCA

Review, V o l . V I , No. 3, pp. 332-343, January, 1942.
* H . T. Friis, "Noise Figures of Radio Receivers", Proe. LR.E., V o l . 32,
pp. 419-422, July, 1944.
 NOISE THEORY AND INPUT CIRCUITS 289

Fig, 5—Minimum noise factor.

u \ 1
varies w i t h the first power of 8)2 IN C
1
|
the frequency. The curves of K B -
AS S U M t D "

- -r - /
A
y
F i g u r e 5 show the optimum noise - - —j—y- ULWU "URE
/ r / A S S UM £ D
factors for the two cases con-
sidered, plotted against the quan-
/
D.UI 0.02 O.OS 0. Q.2 OS I

tity VR ea ffi for the condition

|
6 = 5. I f the frequency for w h i c h V #
1 w / 0 i is unity is designated as
/,„ the quantity VH eq #i for any frequency / is equal to the ratio / / / „ .
The method of analysis described above is essentially the same
s
as that used by HeroId . T h e curve for the case of no coherence
( F i g u r e 5) can be identified w i t h one of the curves ( F i g u r e 5) of
Reference 3 when differences i n the coordinates used are taken into
account. E q u a t i o n (21), g i v i n g the required ratio of tube input
conductance to c i r c u i t conductance, is equivalent to E q u a t i o n (7) of
Reference 3.
The suseeptance required f o r the case of quadrature, as found
from E q u a t i o n (19), is obtained by the same amount of capacitance
at any frequency. E q u a t i o n (19) can be rewritten

B0
2
= e g /B .
l 1 e<1 (19a)

Because g is proportional to the square of the frequency and R

1 eq

and 6-y are independent of frequency, i t is evident that the suseeptance

B is directly proportional to the frequency and, consequently, can be
0

produced by a fixed capacitance.

COMPARISON OF T U B E S

The data for Tables I and II were obtained by calculating values

for the equivalent noise resistance and using measured values for input
conductance for the tube types listed. Table I gives the reference
frequency for noise, / „ , and the frequencies f o r which noise factors of
1, 3, and 10 decibels are calculated for a number of pentode types. N o
coherence is assumed between plate noise and g r i d noise for this case.
Table I I gives similar data for two triodes and for several pentodes,
connected as triodes, under the alternate assumptions of no coherence
between plate and g r i d noise, and a quadrature relation between plate
and g r i d noise. The 10-decibel column for the quadrature case is omitted
because the indicated frequencies are too h i g h to make the assumption
appear reasonable.
290 ELECTRON TUBES, Volume II

Table I—Pentodes

g tH Frequency for Noise Factor

Rgq (100 mcs.) fn (Idb.) (3db.) (10 db.)
Type oh ma micromhos mcs, mcs. mcs, mcs.

6SK7 11,600 440 45 2.5 8.9 56

6AC7 650 1,730 94 6.3 19 119
6BA6 3,800 580 67 3.8 13.4 85
6AG5 1,900 300 133 7.5 27 169
6AK5 1,900 125 208 11.6 41 262
6BH6 2,360 340 12*2 6.8 24 154
6BJ6 3,800 275 98 5.5 19.3 124

The input conductance values used i n Tables I and II were meas-

9
ured by the susceptance-variation method and include the effects of
lead inductance. F o r pentodes, the predominant lead effect is that
of the cathode-lead inductance, which tends to increase the input
conductance. F o r triodes, inductance i n the plate lead tends to reduce
the input conductance and this effect may be equal or greater than
the effect of cathode-lead inductance. F o r the triode-coimected
pentodes, the input-conductance data obtained w i t h the tubes con-
nected as pentodes are used.
Triode " A " i n Table II is a developmental triode, designed p r i -
marily for use as a high-frequency oscillator. The low input con-
ductance and the consequent high " / „ " value recorded for this type
9
"Input Admittance of Receiving Tubes", RCA Application Note
AN-118, RCA Tube Department, Harrison, N. J., April, 1947.

Table II—Triodes and Triode-Connected Pentodes

Frequency for Indicated Noise

Factor
No Coherence Quadrature
R,q g>« fr, Assumed Assumed
(Triode) (100 mcs.) (Triode) (ldb.) (3db.) (lOdb.) (Idb.) (3db.)
Type ohms micromhos mcs, mcs. mcs. mcs. mcs. mcs.

6SK7 970 440 72 4.0 14 91 18 3P

6AC7 214 1,730 164 9.2 33 207 41 82
6BA6 410 580 204 11.5 41 258 51 102
6AG5 380 300 294 17 59 374 74 148
6AK5 380 125 476 26 92 580 116 230
6BH6 390 340 274 15.4 54 345 68 137
6BJ6 485 275 274 15.4 54 345 68 137
6J6 470 195 320 18.0 63 410 80 160
"A"* 360 50 747 42 147 940 186 373

* Developmental triode.
 NOISE THEORY AND INPUT CIRCUITS 291

is probably accounted for by close spacing, h i g h current density, and

a symmetrical cylindrical structure which contributes to uniformity
in the cathode-to-grid and grid-to-plate transit times.

R E L A T I O N OF R E F E R E N C E F R E Q U E N C Y TO T R A N S I T TIME

The reference frequency for noise for a triode depends p r i m a r i l y

on the electron transit time between cathode and g r i d . The noise
equivalent resistance for a triode is approximately R ~2.5/g (7) Kfl m

6
and the electronic component of the input conductance is approxi-
mately g = g ( » r , ) 720.
x m (23)
The product, therefore, is R ev g = (WT, ) 78.
y (24)
T h i s product is equal to unity when MT = 2.83
X (25)
so f = 0.45/n.
n (26)
The values of / „ obtained from E q u a t i o n (26) are even higher than
the values given in Table II. The eathode-to-grid transit time for a
10
tube such as Type 6 A K u is of the order of 7 X 10" seconds, so the
value of /„ from the above equation is
10
/„ = 0.064 X 10- cycles = 640 megacycles.
The value obtained for Type 6 A K 5 from input conductance data (Table
II) is 476 megacycles.
It appears that the only way to increase the frequency for a given
noise factor w i t h electron tubes of conventional design is to reduce
the transit time, Triode types such as the 6J6, 6J4, and 2C43 are
designed w i t h close enough spacings and, consequently, short enough
transit times to give promise of good results i n equipment designed
for m i n i m u m noise.

E F F E C T OP CIRCUIT LOSSES

A n important question w i t h reference to the application of the

curves and tables presented is the attainability of the circuit condi-
tions assumed. The conditions are not hard to realize in practice, as
the following examples illustrate:
1. Consider the use of Type 6 A K 5 as a pentode amplifier at 40
megacycles. The reference frequency f is 208 megacycles, so the n

ratio / . / / „ is 0.192; the product # „ f f i is 0.037. The calculated noise

e

factor is 3 decibels. The required ratio g /g„ is 0.046. Because thex

tube input conductance for 40 megacycles is 19,7 micromhos, the re-

quired antenna loading is 230 micromhos. F o r a tube input capacitance
of 6 micromicrofarads, the quantity <*>C is 1500 micromhos; because
292 ELECTRON TUBES, Volume II

the total conductance at the g r i d is 250 micromhos, the m i n i m u m value

of Q is 6, H i g h e r Q values may be obtained by adding more capacitance
w i t h appropriate inductance values. It is evident that there w i l l be
no serious increase i n the noise factor until the conductance of
the added elements becomes appreciable i n comparison w i t h 250
micromhos. I f the Q is improved to 50 by addition of a resonant
circuit w i t h a Q of 200, the added conductance is approximately 60
micromhos. The noise factor would be increased from 3 to 3.5 decibels
by the added c i r c u i t losses.
2. Consider the 6 A K 5 or an equivalent tube connected as a triode
used at 200 megacycles. Neutralization may be used to avoid feed-
back, but feedback generally docs not have an important effect on
the question of obtainable noise factors. The reference frequency
/ „ is 476 megacycles; the ratio / / / „ is 0.42; the expected noise factors,
from the two curves of F i g u r e 5, are 5.1 decibels for no coherence,
2.3 decibels i f the quadrature relation holds. In the first case, the
required antenna loading is 2600 micromhos and the resulting Q for
the input circuit is only 2. Adjustment of Q to any moderate desired
value can be made by the addition of circuit elements as before with-
out materially affecting the noise factor. I n the second case, the
antenna loading would be adjusted to equality w i t h the tube con-
ductance, which is 500 micromhos for this frequency. Then, the
susceptanee which must be added is 5700 micromhos, corresponding
to a capacitance of 4.6 micromicrofarads.

CONCLUSIONS

T h e conclusions which may be drawn from this discussion may

be summarized as a set of principles to be followed i n the design of
amplifiers for low noise.
1. Choose an input tube w i t h low transit time. F o r frequencies
above 30 megacycles, use a triode or a triode-connected pentode.
2. Adjust the input circuit w i t h signal-tc-noise ratio as the
criterion. T h i s adjustment is most readily made by using a noise
generator, such as a diode, as a signal source.
3. T r y the effect of detuning the input from resonance and the
effect of increasing the coupling to the antenna beyond the value
for m a x i m u m gain.
When theoretical considerations indicate a very low noise factor,
it may be necessary to pay considerable attention to the design of
the load c i r c u i t of the first tube and the input circuit for the second
tube to obtain optimum results.