ADO 4 4 4 46 - apps.dtic.mil · ataare fr an pupos other '1,ha vth a eflitel opersion, reltedtheU....

UNCLASSIFIED

ADO 4 4 4 l 46

DEFENSE DOCUMENTATION CENTERFOR

SCIENTIFIC AND TECHNICAL INFORMATION

CAMERON STATION. ALEXANDRIA. VIRGINIA

UNCLASSIFIED

NOT.-E: n ; 'Ivrxm t r obe dxwigsspc/4xt ataare ~se. fr an pupos

other '1,ha vth a eflitel reltedopersion, theU. /

suplie tj erntoother drawi mvngs., speckuiw~m. orote

wisetan in o j .kV ata l are tse fodr oy puros

ohern pers>eu& y L.cr nor conseiltnr any Iorbligatcai, wnu fandCslv4 the fact thtel Gen-maentedy Nth~~2at may ihnxsaydb oreayeayd

p~erec c r.

N°. 21

ELECTRICAL ENGINEERING DEPARTMENTUNIVERSITY OF MINNESOTA

INSTITUTE OF .TECHNOLOGY

Third Quarterly Report

Februarr 1. 1964 through April 30, 1964

It Study of Noise inSemiconductor Devices

AUG 18 194

Contact DA 36-039 AMC-03718 (E)

D.A. Proj. No. 10622001A0"6-0100 DDC'RA C

PLACED BY U. S. ARMY SIGNAL CORPS ENGINEERING LABORATORY

FORT MONMOUTH, NEW JERSEY

REGENTS OF THE UNIVERSITY OF MINNESOTA

MINNEAPOLIS, MINNESOTA

ASTIA AVAILABILITY NOTICE: QUALIFIED REQUESTORS

MAY OBTAIN COPIES OF THIS REPORT FROM ASTIA.

ASTIA RELEASE TO OTS NOT AUTHORIZED.

,,I ' ..o,, ll o,,

60 II.

A2 2O v3 a. 06k 8 A

i8i oil

gin AIII P

" 0 1 lql

IA, !

0 4

Cho

a 0.

as 041 4 "68O

Study of Noise in Semiconductor Devices

Third Qwrterly .ReportFebruary 1, 1964 through April 30, 1964

The object of this investigation is the study of noise

in semiconductor devices and the theoretical

interpretation of the results.

Contract Nio, Dis 36--039 AN-03718(E)

D. A. Proj. no. 1062200lAO56-.010

A. van der Ziel, Chief Investigator

Table of Contents

I. Abstract 1

II. Purpose 3

III. Publications and Reports 4

IV. Work Accomplished under the Program 5

A. Circuitry and Equipment 5

1. Pulse Measurements of FIT Noise atLow Temperatures 5

2. Noise on Transistors at VHF andMicrowave Frequencies 10

B. Measurements 11

1. Noise in GaAs Laser 11

2. 1/f Noise at 1000 Cycles and theReliability of Transistors 12

3. Unijunction Transistors Noise in theNegative Resistance Region 13

4. 2ulse Mesurements of Noise at LowTemperatures 17

5. The Enhancement Mode FIT 18

6. Ecess Noise in FTs 22

C, Theory 25

1. Linear Profile Approaimation for aDiffused Junction, P4hannel FT 25

2. Theory of Themal Noise of anEnhancement FET 35

Table of Contents (continued)

P&1e

V. conclusions 44.

VI. Programa for Next Interval 46

VII, Personnel Ebployed on Project 47

List of Illustrations

Figure&L

1 The method of measurement of thenoise of YET at low duty cycles

2 (a) Jig for FET A2

(b) The balanced gate for 455 kc/s A2

3 Pulse generator and pulse amplifier AS

4 Current vs voltage for GaAs laser atliquid N2 temperature A4

5 Power vs current for GaAs laser at liquid

N2 temperature AS

6 Temperature dependence of Rn in IWP CK913 A6

7 Temperature dependence of Rn in NEUsilicoa 2Li2808 A7

a Rn vs I in MO2808 Nlo 2 at 300*C A8

9 k ve I a in N12808 Mo 6 A9

10 Rn vs time for 2N2808 No 6 AIO

11 Rn vs a 'function of time for 2N2808 ao 5 All

12 eq measured across B, - 2 (Ieq vs IE A12

13 I eq measured across BI B 2 (1 eq vs IB2) A13

14 1ea measured across Bl B2 (I vs. trfrlquency) A14

15 Crystalonics No 44 PT A1S

M, Dependence of 16 and on Vg of theEOA enhancement mode A16

17 Dependence a the frequency spectra ofthe RCA enhancement mode FT on thegate voltage Al?

I.

List f Illustrations (continued)

Figure Page

18 High frequency noise spectra of the RCAenhancement mode PET A18

19 Dependence of I on Vd of the RCAenhancement modjqFET A19

20 Dependence of Ieq on Vd near Vd = 0 A20

21 Noise of RCA enhancement %ode PET when

Vd=0 A2.

22 Penetration of the space charge regioninto the p and n regions for a linearprofile (diffused) Junction A22

23 Theoretical dc charactert.tics,neglecting the diffusion potential A23

24 Theoretical normalized transconductancevs normalized voltage A24

25 Theoretical normalized transconductance vanormalized drain current for varied gatevoltage A25

26 Enhancement mode PET A26

27 I /1 (0) vs I/Inos for enhancement modeFH Jelculated by flermal. noise theory A27

Page I

I. Abstract

Equipment has been designed for measuring noise in

FET's at low temperatures under pulsed conditions to avoid

heating effects. The equipment is performing very satis-

factorily.

Equipment is being designed for trwnsistor noise

measurements at v.h.f• and microwave frequencies.

Preparatory studies of noise in GaAs lasers are'- 6ontin~uing.

The 1/f noise at 1000 cycles has been measured for

various transistors as a function of temperature. Also noise

measurements are being made on transistors that are life-

tested at elevated temperatures.

Noise in unijunction transistors can to a certain

extent be interpreted as being caused by drift of injected

carriers in the region between the two base contacts. This i

especially the case if the base-2 current is large in com-

parison with the emitter current.

Pulse noise measurements on FET's at liquid nitrogen

temperatures reveal a white excess noise spectrum that is

strongly affected by heating effects in the channel.

Noise measurements of enhancement mode FET's with

insulated gate show a spectrum of the form cont/(l + 02-T2)

with tv a 10 - 30 ti sec. This noise is attributed to traps.

A higher frequenoes the moise is practically white, but it

is non-thermal. Various noise mechanisms seem to be present,

and attempts are being made to sort them out.

A discussion of low frequency e;cess noise in lET's

reveals that the present theoretical results cannot fully

explain the data.

The (Id,Vd) characteristic and the (gm,Vg) character-

istic of a diffused junction are calculated both for constant

mobility and for a field-dependent mobility under the asstup-

tion of a linear impurity distribution in the Junction.

The theory of thernal noise in an enhancement mode

F PET with insulated gate is developed. It is found that I eqgoev to zero if the device approaches saturation,

n. Purpose

The purpose of this report is to sumuarize the

progress in the University of Minnesota Electrical Engineering

Department during the period fron February 1, 1964 to

April 30, 1964 on Contract No. DA 36-039 AM-3718(E).

The contract calls for an investigation to determine

the cause and effect of the noise that occurs in semiconductor

devices carrying de current. It can be divided as followas

(a) A study of noise in semiconductor devices such

as e.g., junction diodes, junction transistors, field-effect

transistors and similar devices.

(b) A study of the possible relationship between

1/f noise and reliability.

(c) A theoretical study in support of the experi-

mental work.

Pege 4

III. Publications and Reports

Mr. Konrad Fischer, USAKRDL representative, visited

our laboratory during this period and discussed the research

program.

The Third Monthly Report was submitted on March 2, 1964

and the Fourth Montly Report was submitted on April 1, 1964.

The Second Quarterly report was submitted for approval on

March 10, 1964.

The paper entitledi

"Small-Signal, Highb-Frequency Theory of Fiel.-Effect Transistmmfto

by.'.A. van der Ziel and J. W. Rro was published in the April,

1964 issue of the IEEE Transactions in Electron Devices

(Vol. EDll, 128-135, April, 1964).

IV. Work Accomplished under the .roam

A. Circuitry and Equipment

I. Pulse Measurement of FT Noise at Low

Temperatures.

The measurement of the drain voltage dependence of

the IZT h.f. noise at 770K, reported in the previous quarterly

report, showed that the noise of a PET decreased appreciably

with increasing drain bias voltage (Vd) in the saturation re-

gion. This noise was still of unknown origin, but it decreased

very rapidly with increasing temperature. Because of this,

the decrease in the PET noise with increasing Vd might be

considered in the following way:

(Increase of Vd).(Increase of the power dissipation

in the channel)i(Temperature rise of the channel)-Decrease

of the noise).

To avoid the influence of such heating effects, the

PET must be biased by a pulse for a short tiae and its noise

must be measured within this time before the channel is heated.

The pulse system described in the following paragraphs was

built to measure noise by this method.

(a) Generel description of tbh system

Figure I shows the block diagrun of the system for

the pulse noise measurements. At present, the noise spectrum

between 500 kc/ and 10 Mc/s can be measured by the drain bias

pulse whose width ranges from 1 to 5 msec and whose repetition

frequency is 40 c/s.

Pae 6

This system works in the following way. The pulse

generator generates a train of positive pulses. The polarity

of this pulse is inverted and amplified by a power amplifier.

The negative power pulse thus obtained biases the N-channel

ET from the source side. The terminal voltage of the FET

is a superposition of the noise and the bias pulse voltages.

This voltage is amplified by a preamplifier and than fed

into a radio receiver (National Radio NC-125). Because the

radio receiver is insensitive to low frequencies, the inter-

mediate frequency output (455 k/s) of the receiver is due

to the FET noise and the harmonics of the bias pulse. The

455 kc output is subsequently amplified and then fed to a

balanced gate amplifier. The gate amplifier operates in

synchronization with the drain bias pulse such that it trans-

fers the input signal to its output only when the PET is

biased by pulse. The output of the balanced gate amplifier

is then further amplified by a main amplifier and then fedto a quadratic detector through a cathode follower.

.. (b) roblems in the pulse measuement at- *de

There are three problems to be considered:

i) how to avoid the saturation of the preamplifier

by large bias pulse applied to PET,

(ii) how to avoid the harmonics of bias pulse (such

harmonics cannot be separated from the noise of PET),

(iii) how to calibrate the noise to obtain the

equivalent noise diode current Ieq*

The solution to these problems wis as follows

(1) To avoLd the saturation at the preamplifier,

the PET was biased by the pulse from the source side as shown

in Fig. I and Fig. 2a. The load impedance of the PET was a

LC tank circuit tuned to the frequency at which the roise

spectrum was measured (500 kc - 10 M/s). Because its resonant

frequency was very high as compared with the pulse repetition

frequency (40 c/o), this LC circuit presented a small impedance

to the bias pulse and hence the pulse voltage developed

across the tank circuit was very small and did not cause

saturation of the preamplifier.. The radio receiver (KC-125)

following the preamplifier rejected the low frequency component

due to the pulse so that no saturation occurred in the later

stages.

(ii) There was no practical way to separate the noise

ot the YET from the high frequency harmonics of the drain bias

pulse. Thus in this apparatus the noise spectrum was measured

in the frequencies (500 kc . 10 Mc/) which were substantially

higher than the pulse repetition frequency (40 c/s), In the

above frequency range, the harmonics of the bias pulse were

very mall as compared with the noise of PET and the measuremeitb

were successful.

(iii) Because the PET was biased only for a short time

by the pulse, the noise must be measured and calibrated within

this ahort time. To do this, the amplified noise voltage was

fed to the quadratic detector only when the FET was based.

This was done by means of the balaneed gate amplifier operating

in synchronization with the bias pulse as shown in Fig. 1 and

Fig. 2b.

This technique allowed the calibration of noise by a noise

diode connected in parallel to the FV,.

(c') Details of the circuit

In this section, three important parts of the pulse

noise measurement apparatus will be considered: (i) the jig

for FET's, (ii) the balanced gate amplifier, (iii) the pulse

generator and amplifier.

(i) The circuit of the jig for the FET is shown in

Fig. 2a. The drain of the FET was connected to a LO tank

circuit which was tuned to the frequency at which the noise

spectzn was measured. The PET was biased from the source

side by a pulse generator through a low-pass filter which

eliminated unnecessary high frequency harmonics of the bias

pulse (the time constant of the low-pass filter was such that

the filter did not distort the pulse form). A floating gate

bias supply was provided because the source side of the FET

was not grounded.

(ii) The circuit of the balanced gate amplifier is

shown in Fig. 2b. The balanced type circuit was used to avoid

the occurrence of a large gate transient pulse. The control

grids of the two push-pull sharp-cutoff pentodes -.(6AK5) were

biased to -6 V (cutoff region) in the absence of a gate pulse.

When the PET is pulsed, the control grid voltage was raised

approximately up to 0 V and the input signal was transferred

to the output. Because the output frequency of the NC-125

receiver was 455 kc/s, the input and the output circuit of the

gate amplifier was tuned to this frequency.

(iii) The circuit of the pulse generator and amplifier

is shown in Fig. 3. A pulse generator of a gated beam tube

6EN6* (Fig. 3a) generated a pulse with the repetition frequency

40 c/a. The pulse width and its repetition frequency can be

changed between 1 to 5 msee by changing the values of R and

C in the limiter and the accelerator grid circuit of 6BN6.

The negative pulse taken from the anode of 6BN6 was subsequently

phase-inverted, and was fed to a cathode follower. A diode

clipper circuit was provided at the output of the cathode .

follower to obtain a flat-topped pulse (Fig. 3b). This flat.-

topped positive pulse was used to drive the balanced gate ampli-m

fier. To obtain a power pulse to drive the PET, this pulse

was further amplified and fed into a low impedance cathode

follower (6Y6). A negative pulse whose maximum height was about

90 V was obtained at the output. Because an electrolytic

capacitor of large capacity (200 PF) was used to couple the

*This circuit was invented by Prof. Koichi Shimoda, Departmentof Physics, University of Tokyo. See K. Shimoda, Electronicsno KLoo, Shokabo, Tokyo (1964) pp. 230-231 (in Japanese).

final cathode follower to the FET through a pulse attenuator,

a method to compensate for the do leakage current of the elec-

trolytic capacitor was provided together with the pulse

attenuator (Fig. 3c).

This pulse measurement apparatus is now operating

satisfactorily. Further improvement of this apparatus,

especially the Improvement of the accuracy by increasing the

time constant of the detector, is now going on.

2. Noise on Transistors at VHF and Microwave

Frequencies

The construction of equipment for measuring transistor

noise at vhf and microwave frequencies has been continued during

this period. Among the equipment built is a jig for mounting

the transistor to be measured. The input and output impedance

of this device is 50 oh and tuning of the input and output

is provided by a variable length shorted stub. The bias for

the input and output junction is applied through the shorted

stub.

For measurement of noise in lower frequencies up to

50 Mo/s a different biasing circuit has been built. For these

frequencies the input and the output of the tranisstor is tuned

by lumped tuned circuits.

In the frequency range between 50 Mc/s and 200 Mc/s

the noise of the transistor stage under test drowned in the

noise of the TV toner used as a preamplifier. We are now

building low noise preamplifiers for each channel.

I

Some low frequency noise measurements were made on the

transistors supplied by the Signal Corps. Since the results of

these measurements are not complete, they will be reported in

the next quarterly report.

B. Measurements

1. Noise in GaAs Laser

It has been reported in previous quarterly reports

about the noise measurement problems posed by pulsed operation

of the GaAs laser. The first of these problems is that of

background noise, which is always present, drowaing out the

laser noise, which is prebent only when the'laser is turned

on. A gating circuit has been devised which can be synchronised

with the pulse driving the GaAs laser, and can turn off the

background noise when the laser is off.

A second approach to the problem is that of attempting

to operate the Ga" saser contituously. Fig-res 4 and 5

show very rough measurements of the current vs voltage charac-

teristic of the laser at high forward currents, and of the

power absorbed by the laser at these currents. Both curves

refer to the laser cooled to liquid nitrogen temperature. If

the laser were cooled to liquid helium temperatures, or even

lower, the current threshold for losing action should be

reduced by a factor of 10 from the 20 amperes required for

liquid nitrogen temperature. This reduction in threshold

current will have an even greater effect on the power that

must be dissapated by the laser. Our hope is that the lower

temperature will permit continuous action. But even an

increased pulse length at a higher repetition rate would be a

significant improvement.

Experiments using continuous operation of the laser

must be carried out with great caution to avoid destruction of

the laser.

2. 1/f Noise at 1000 Cycles and the Reliabil_t

of Transistors

More intensive measurement of the noise as a function

of temperature was made for the Germanitu transistor CK913.

As shown in Fig. 6, the noise stays fairly constant until a

certain critical temperature where the noise starts to increase

rapidly.

The sane kind of measurement was made for the silicon

transistor 2N2808. It shows the same behavior (Fig. 7).

The critical temperature of this kind of transistor can be

seen fran the figure to be higher then that of the above men*

tioned Germanium unit.

For unit #3 in Fig. 7 near 3000 C9 I C failed to reaoh

1.0 mA (all other measurements were made at Ic 1.0 mA) so

that the noise was measured at Ic = 0.7 mA and was found to be

below the expected value. Then a measurement of noise as a

function of collector current was made on two different units,

(Figs. 8 and 9). The noise was found to increase with

increasing collector current. This suggests the dotted curve

in Fig. 7 for unit #3.

Having estimated the critical temperature of silicon

transistor 2N2808, two transistors with nearly the same

I c - V., characteristic curve were heated up to the estimated

critical temperature; namely, 2800C. Then a measurement of

noise as a function of time was made. The noise of unit #6

stays failrly constant and has nearly the same value as that

at roan temperature (see Fig. 10). This suggests that 2800C

is below the critical temperature for this unit. On the other

hand, the noise of unit #5 (Fig. 11) is found to increase

rapidly at 2800C as compared with the noise at room temperature.

This suggests 2800C is either the critical temperature or above

the critical temperature of unit #5. We see that critical

temperatuw's ire lifferent. even.for the same type of transisM

tore with very close I c - Vc, curve.

Further investigation on the cause of this difference

in critical temperature will be carried on as well as the

accelerated life test of the transistor.

3., UniJunction Transistors .Noise in the

Negative Resistance Reion

Under certain operating conditions the input impedancee

between emitter and base-one is negative. This corresponds to

the condition that IB2 is large compared to I.

For small Ia and large IB29 the operation of the

unijunction transistor can be understood in terms of a very

simple model. It is assumed that the electric field in the

base two region is constant aid give, byt

i

2 A q1L kpP + 'bn) (1)

where b = I/Ip.

If the diffusion current is neglected, the continuity of hole

and electron currents at the emitter yields:

q n no E2 = q %(no + po 0P)E " (2)

q p p E2 + I/A =q p p" E (3)

Here E" and p* are the electric field and the hole concentration

I repsectively at the emitter. Solving for p' and E' we obtain:

SIB2 - IE n + p0)/(no - 4)A qppo(bn 0 + p0 )

P°= P0 + Ip 2 ( o o /0 (po + b o .- (5)

For extrinsic material# no)) p0 and (14) and (5) reduce tot

I(B2 - b6)mom-- AW -pp bin 0

P" ' Po bZ (7)0 1B2 .M W

If the time required for the injected minority carriers

to drift through the base is small compared to the minority

carrier lifetime, then the electric field and the bole

Pa ge 15

concentration will be constant in the region between emitter

and base one and will be given by (6) and (7).

For this case the noise can be calculated using the

theory of Hill and van Vliet. They obtain the results

2(1 - co o.a)S (f)=i4VPTr (8)sp( a, " ')2 )

OaL

Here ra - = ambipolar drift timep

p = hole concentration

V = volume

This can also be written as:

F sin(¢/2) 7 2S (f) = 2V a- (ua2) 2 (9)

Since:

SI(f) =( b o . S p (f) (10)

0

Then substituting into (11) the following:

bn0 IEY s2 .b m (12)

L q tki0 V

Ia = -b (13)

=B2 + IE (14)

Pa ge 16

We obtain:

Sf) ( 2 1 (15)I B2E '1 ri(dora2)1

But for IB2> 1E, this becomes

i .sin w./2 -S(f) 2q I [ /2 1 16)

LetS1 (f) = 2q 1 eq (17)

Then

eq e "I (Oa/2) (18)

By examining Eq. (18), it can be seen that the spectral

intensity should depend linearily on IE and should be indepen-

dent of IB2* Moreover since -v a depends inversely on 1 B2, the

frequency at which the noise begins to decrease should depend

linearily on IB2*

Figure 12 shows a plot of Ieq measured across B1 - B2

as a function of a. t 25 kc. The noise measured across B - B2

with I E = 0 has been substracted. This means that part of the

thermal and excess noise has been substractedo This is

peruissible only if the excess noise is uncorrelated with the

generation-recombination noise. From Fig. 12 it appears that

1eq depends linearily on I E from 10 a to 200 .a. The magni.

tude of Ieq is slightly greater than I

Figure 13 shows a plot of Ieq as a function of IB2 at

25 kc. Note that in the region IB2> IE, the noise is almost

independent of IB2 * The agreement is best for very small I

SPu IB2 C 1 E' the-noise is essentially that associated with

the diode.

Figure 14 shows the spectrum of Ieq for three different

values of IB2. If we define a = 1/,T, it can be seen that

CO dspends heavily on IB2 as expected. Hqwever, the spectruma B2/sin x 2

does not show the characteristic dependence at high

frequencies. This may be partly caused by the fact that the

thermal noise level is appreciable compared to the generation..

recombination noise level. It also could be caused by the

geometry of the device. All the formulas used were derived

on the basis of a one-dimensional geimetry. In the device

used, the length of the region from emitter to base one was

only about twice the width of the base region. This may mean

that the distances the individual carries drift are quite

different. Thus, T a will be different for each carrier. This

could cause a smoothing of the characteristic "baps" in the

sin x 2 function.

4. Pulse Measurements of Noise at Low Temperatures

The system for pulse noise measurements described

before was used to measure the drain voltage dependence of FE?

noise, Ieq, at 770K. The PET measured was Crystalonics #2

which showed a white noise spectrum for frequencies between

50 kc/s and 8 Mo/s at 770K. The results of dc measurements

were reported in the previous quarterly report.

Figure 15 shows the result. Measurements were made

for V = 0 and -3 volts. The remarkable point is that the

results of pulse and do measurements are considerably different

frm each other. For small drain voltages, these two methods

give the same Ieq However, for Vd)2 volt, Ieq measured by

the pulse method is larger than that measured by the dc method.

Also the peak of the curve shifts toward higher drain voltages.

This result shows that a considerable heating effect

of the channel occurs at low temperature measurements. As

discussed in the section on equipment, the channel is less

heated by the pulse method than the do method, and hence higher

Ieq values are obtained.

This result indicates that the pulse method must be

used for an accurate measurement at low temperatures. This

suggests considerable technical problems at low frequencies.

A preliminary attempt to extend ",he measurements to low

frequencies is now going on.

5, The Enhancement Mode FET

(a) Measurements

Measurements on two RCA enhancement mode FET's of

silicon were made during this quarter. These FET s showed do

drain characteristics similar to those of ordinary FET' if

the gates were biased positively a few volts.

Their trensconductances in saturation (g.) and zero

voltage drain conductances (go) vary with gate voltage as

shown in Fig. 16.

The noise spectra in saturation region are shown in

Fig. 17 (between 2 and 800 kc/s) and in Fig. 18 (between

0.8 Mc to 22 Mc/s). The low frequency noise spectra are of

the generation-recmbination type whose characteristic

frequencies are approximately 10 kc/s. For frequencies between

3 Mc/s and 15 Mc/s the spectra are essentially white. Above

15 Mc/s9 a slight increase in the noise spectra with frequency

is ob-erved, Figure 19 shows the dependence of I on theeq

drain voltage (Vd) at 5.5 Mc/s. As is seen in Fig. 18, the

noise spectra are white around this frequency. For gate vol-

tages between 1.5 and 3 volts, Ieq(Vd) decreases slightly with

increasing Vd (for small Vd). This noise is the thermal noise

of gdo'

If Vd is increased further, the noise increases sharply

and finally approaches to another constant value when Vd = Vg

(Vd(sat) = Vg is the drain saturation voltage). This voltage

is indicated by arrow marks in Fig. 19. This sharp rise of

Ie(Vd) occurs approximately V = 1 V. If Vd is increased

further beyond saturation, Ieq increases again because of

the breakdown noise of the insulator (gate leakage current

appears). For V larger than 4 V, Ieq(O) is considerably

higher than the tbermal noise level of gdo .

Ieq(V,) decreases quite rapidly with increasing Vd

(for small Vd) as shown in Fig. 20. This indicates that a new

noise mechanism occurs for small Vd" Because this noise

becomes prominent for high gate voltages and decreases rapidly

when the drain voltage is applied, this noise may be due to

some breakdown mechanism sensitive to the relative voltage

between the channel and gate. In Fig. 20, the drain voltage

for zero drain current is not zero. The arrow mark indicate

such points.

Figure 21 shows the plots of 1 (0) vs Botheq e gd~ otof 1eq() and gdo are the functions of V. The points on

the curves are taken by changing V The straight line shows

the thermal noise level. This figure shows that for medium

gate voltages I (0) is higher than the thermal noise leveleqby only 20 - 30%. Thus this noise is considered to be a

thermal noise. For lower and higher gate voltages, I eq(0)

is much higher than the thermal noise level. The peculiar

drain voltage dependence of the excess noise that occurs

Zor large Vg was shown in Fig. 19. The nature of the excess

noise that occurs for small V is not mown.g

(b) Classification of the noise of enhance.

ment mode PET

Based on the experimental results discussed before,

the noise of the enhancement mode FET can be classified into

five kinds. To make the classification clear, the noise of

?ee 21

the device is regarded as a function of Vd and Vg The

follcming table gives the regions where each kind of noise

is impottant.

Table 1

Vd I Near BeyondVd Saturation Saturation

small III

medium I* IV* V

large II

*Essential noises of the enhancement mode FET.

These noises are explained briefly in the following.

The noise of the lot kind is the thermal noise of

go" Actually the measured Ieq is about 20 - 30% higher than

the tnermal noise. This is probably due to the noise of the

lind and Illrd kind, which is not exactly zero in this region.

The properties of this noise will be analyzed in the theory

section (Ca).

The Noise of the lind kind is characterized by a

strong decrease of noise when small drain bias is applied.

This noise is probably due to some kind of leakage between

the gate and channel because this noise occurs only for large

gate voltages.

The noise of the Ilird kind appears for small Vg.

Its nature is not known yet.

The noise of the IVth kind appears in the saturation

region. This noise has a white spectrum above 3 Nc/s. As one

possibility, this noise may be regarded as a suppressed shot

noise. If this is true, its suppression factor at the satura-

tion is about 0.25.

The noise of the Vth kind is due to the breakdown of

the insulator by high drain voltages. This noise is accompanied

with a gate leakage curren .

Only the noises of the let, Ilrd and IVth kinds are

essential to the enhancement mode FET because all the other

noises are more or less related to the breakdown of the insulat=

6.- EEcess Noise in FETV.

In an attempt to investigate further the origin of

the excess noise in IPT's, we have first tried to better

understand the dc operation. In previous reports, the theore-

tical approaches have been presented for the alloy junction

units using an abrupt junction model. These approaches con-

sidered a field dependent mobility and a combination of a

constant and field dependent mobility for particular regions

of the Id - Vd characteristics. However, since several units

under investigation are diffused junction PET*a, an approxi-

mate model in this case might be to assume a linear profile

for the charge distribution at the gate junction. In the

theory section of this report, this model is used to calculate

the do characteristics both for a constant mobility and a

field dependent mobility.

If we now make a comparison between the experimental

results and those found theoretically, it becomes apparent that

a combination of factors must be taken into account if the

Id - Vd characteristics are to be explained for a particular

FET. The first of these appears to be whether or not the

device is an abrupt (alloy) junction device or a gradual

(diffused) junction device. As illustrated on Fig. 23, the

effect of the latter consideration appears to make the initial

slope of the Id - Vd characteristic more nearly linear.However, a second effect to be included is the presence of a

field dependent mobility. As was reported previously, the

effect of this consideration appears to increase the initial

slope and provide a sharper pinch-of. Fram the experimental

results, it appears that these two effects are the most signi-

ficant factors contributing to the Id - Vd characteristics

deviation from those suggested originally by Shockley. In

some instances, there still remains sme discrepency between

theory and experiment but because the construction may be

non-unifovm, it may be necessary to represent a FET by two

equivalent PET's in pprallel with different cutoff potentials.

Thus to completely match the experimental Id - Vd characteris-

tics, these three factors must be considered, making a

general theoretical expression for the characteristics of all

FET's virtually impossible. However, in each particular case,

by comparing the Id - Vd characteristics found experimentally

with toose found theoretically, at least some idea can be

made concerning the mechanisms contributing to the dc operations

Also of significant interest is the dc behavior of

an PET, biased in saturation, for varying gate voltages.

On Fig. 24 the results of several theoretical approaches for

vs V$/VgO are presented, but such results, while givingreasonable agreement with experiments for low gate voltages for

some units, do not account for the experimental data for all

cases for high and low gate voltages. A preliminary investiga-

tion of a plot c Id/s~ vs V suggests a dependence of the

current on the gate voltage given by

nId/Zdo = (1 - Vg/Vgo)

where apparently n is a rational number (in most cases n = 3/2,

2, or 4). However, as the limiting condition Vg.-V ° is

reached, this expression fails to agree with the experimetal

results. A point of ambiguity is in the choice c IT andgoactually a modified exponential dependence may be a more

accurate estimate. Thus before the noise an a function at

gate voltage could be discussed, it is necessary to better

understand what is happening with the channel in pinch oft

condition. Without giving any experimental comparisons at

this time, it should merely be mentioned that there appears to

be an indication that the edge of the depletion region can not be

considered as an abrupt transition or equivalently that the

depletion region is not entirely void of carriers.

Appazently this effect is of second order for most

cases when the FET is in saturation, until the gate voltage

is such that the characteristics are altered. At this time the

exact processes involved are not clearly understood and care-

ful investigation is now being undertaken to find the correct

theoretical approach. It is hoped that through this investiga-

tion that the actual noise processes contributing in each mode

of operation of an FET will be more completely understood.

C. Theory

I. Linear Profile Approoimation for a Diffused

Junction, P-Channel FET

(a) Constant mobility; Id - Vd and s. vs Vg

characteristics.

Consider now the carrier concentration in a FET soawn

in Fig. 22a such that the charge density in the depletion

region at any point y may be represented by:

P(y) ='- P (I- .y/a) y > b

where y is the distance from the center of the p-region.

Now letting E,(Y,) be the electric field and V(y) the

electrostatic potential in the depletion region, Poisson s

equation becomes a

I

d2 Vks° -= " k% = -p(y)

Thus

dEj a p/ke = o(I. y/a) y>bCY 0 0

Ngow using the boundary conditions Ey = 0 for y = b and y = 2a-b,

the electric field becomes

Cy- b) 1 ( y b)2J for bya

EY 2 1y 2 {L(a - b) -1 (a- [ y q'a (2'a-y j

for ay(2a - b

Now using V f-u dy and for the total potential

V acre.ss the depletion region, V = V1 + V2 where the voltages

V1 and V2 are f oe the regions b

Pae 27

V(b)j Po l 2- b/a)2 - ( - b/a)

o' after simplification and taking the potential W> 0 to be

the reverse bias we have

3 b 3W -V(b) -W0 (1 - b* y ) (1)

a

(2where W= 20 =80-

V(y) and E(y) are shown on Figs. 22b and 22c respectively.

Now the current in the channel is given by

dWI = g(W) e (2)

where g(W), the channel conductance, for the model being con-

sidered is given by

g(W) , q N (a - b)/a bw

where we will now take p = p, a constant. Thus,

,,,-.0 E.-

where

gO- 2 6 "oa =1 qP, odoaw

Nw rewriting Eq. (1) we have

b ob2(1

W 0 0

This is a cubic equation in (b/a) and thus mathematically has

three possible solutions. For this particular equation a

trigonanetric solution is possible1 and coupling with the

boundary conditions

(b/a)-41 as W-O

(b/a) -*O as W -. W

tthe only solution permissible is then(b/a) - 2cos(a/3 + 600)

where cost = - (0 )

Thus substituting in Sq. (3) f£or (b/a) we have

g('W) = 4go [i + cos(/3 + 600)Jcos(/3 + 600) (4)

for cosa -(1 w/Wo)

XI) Drain characteristics

Frzn Eq. (2) integrating over the length of the channel

we have W

IoL g(W) dW (5)Ws

or in terms oE the tr.gonometric parameter a,

1 6athematical Handbook for Scientists and Engineers, Korn andKorns, Mozaw IMLT Go., 196., p.23.

0-W sina g(a) da

Then introducing the physical constrai.ts (10 d)

1 0 if W = OW 0

d Tif Ws 0, "d W o

Id > 0 for Wd >0 for a p.channel PET

ye then have

IdL - + - 5%)L'~d 1 (coy V3 si V

+50- 2.5 o ,,d -Toi ad + d

+3 2Co +Iaf- 3sin~a 34VYc03 Is tad ( 3 V ~

3 22 d- g0W/d

The results of this fomlula, taking Ido = 'dSAT= goW*/L"

oapared to the abrupt junction result given by Schockley is

shown on Fig. 23. It may be noted that the effect is to

decrease the initial slope, finally reaching the same lmit

in slope near saturation.

(2) Transaoonductance

Consider now Eq. (4) for the current in the channel

where it should be noted that

Wd =Vg + Vdf Vd

W V +

Now the transconduotance is given by

bd

gmg

so that assmuing Id is continuous and differentiable, we have

using Eq. (4),

= 1 + cos(ad/3 + 600) cos (ad/3 + 60)

- [1 + cos(%s/3 + 600)] cos (a/3 + 600).

where

co da" (1 aWd/o), cos % =- (1- Wa,/ o)

Considering now Wd/Ao = 1, i.e., the PET ia in

saturation, then

[I + f oos(%/3 + 60*) cos(%/3 + 60')

where coo % =-.(I- W/5 / o ) and

for o =W/W15 1, 1" 2%&-"g*a %o-a

The results of this calculation compared to (1)

Shockley's results, (2) Daey and Ross field dependent

mobility results, and (3) a mobility dependence

Pi p O( E)/2W

is presented on Fig. 24, neglecting the diffusion potential

(Vdif). From this normalized plot the most significant result

appears to be that the initial slope is significantly decreased,

perhaps reflecting the predominant effect of spreading the

depletion region back toward the gate for low gate voltages

for this model.

(b) Field dependent obility; am V Id(Vg)

characteristics

Now it has been found for the linear profile apprai-

mation for the gradual junction PET that the channel conduc-

tance is given by

8(") - o [ + cos(q/3 + 60)] cos(/3 + 0

where coo a - /(la 0' )

and go - Nqp Ndo aw

Suppose now that the mobility is field dependent such thatP &0o(E/2)1/2 where E is the Leotric field. Then

I

1/2 1 1/20= q(E0 ) co aw(V

or

1/2 dx 1/20 Iioado aw(aV)

.ow using the definition for the current in the channel

dW

Ig(W)

we have

o2 + os 2I0 dx = g+ aos(/3+600)] cos(./3+60 ° ) oW

where

1/2

90~ q(E0) ILo Ndo aw

If we then integrate over the length of the channel for a

fixed current 10 1d' we then find

IdL2Ln (4g0') 2 [-cos(a/3+60") I-coa(/3+600) 12 d

(1)

(1) Trans conductance

Now using the definition of the transconduotance

6 - - and using the definition. of Wd and W., we haveavg

after s~e manipulation

(4g0 )211+ cos(cad/360)j 2ECo0s(cd/3+60*]l 2

+ coo

~ + os%3+Q4)JCos( %13600)]I7j

CS ad u-( Wd/W o ) ; o % = -(1- W/W o )

(2) DC characteristics

Now returning to a consideration WC Eq. (1) and using

cosa .- (l - wA 0)

such that

dW=- W sin a da

Then

cos (Wd/Wo - 1)

Id2L - (4 go*) 2 w JO sin a g1 (M) dacos,1 (a/J o -1)

where

91(a) + coa (/3 + 600 [ cos(a/3 + 6010)

Tbus expanding sin a gl(a) and after considerable

manipulation and utilisation of trigonometric identities, we

obtain upon integrations

CO

I LI

d o 1d 3 7( d ) o ( J o

3 s w oo ha coe 2o 00

If we now return to the transconduotance given by

Eq. (2) and considering Wd =o we thus finally obtain:

(4gD2 3a

a%~Y3in-3 - + cs08- + ( 3 sin-y'T-Cos04m)

where

Co% =0 -( We/o)

for O2wa/_ , l8o aa= 90

These results are presented on Fig. 25 together with

the constant mobility, linear profile approximation, and the

results of Shockley and Dacey and Ross. Here it may be noted|that the effect is to decrease the initial slope but the final

slope is the same as in the constatft and field dependent

mobility oases of Shockley and Dacey and Ross, respectively.

2. Theory of Thermal Noise of an Enhancement FET

The noise of the 1st kind is a thermal noise of the

induced carriers in the channel. Thus the property of this

noise must be analyzed by taking into account the channel

modulation effects as in the case of ordinary PET's. In this

paragraph, the characteristic quantities of the enhancement

mode 1ET are calculated based on the model of H. Borkan and

P. K. Weimer end then the theory of the thermal noise is

developed using these quantities.

(a) Characteristic quantities of the enhance-

ment mode FET

The cross sectional view of an enhancement mode PET

is given by Fig. 26 together with the definitions of necessary

quantities. The induced surface charge density, Q(Z), is

given by,

9(Z) e V(z) (1)

The channel current I is given by,

I =Q(Z) S -L-Z) S ellV(Z) EL(Z) (2)

where EL(Z) is the electric field in the channel as a function

of Z ° V(Z) and EL(Z) are related byZ

V(Z) V ) EL(x)dx (3)0

It is well known that there are energy levels on the surface

of a semiconductor that can trap and immobilize the carriers.

To take. this effect into account, V(x) must be replaced by

V(x) - V everywhere. The physical meaning of Vo is the gate

voltage needed to fill all the "surf-ace states" of the channel.

The above replacement is equivalent to using (VK - V ) instead

of Vg The effective gate voltage V * is then defined by

Vg Vg Vo (4)

From (2) and (3),

V(Z) = *2 -21d z (5)

VD9 the drain voltage, is given by

Id dZ :d j rVD L(Z) dZ~ I ~) TL 7]VD Vre

0 0 (6)

where X V *2

0 I

Fram (6), the following two relations are derived.

VD]-i. I V'X.L

(7)

L 21d VXo-LWD ett VXO +

By adding these two relations,

VD + L i- -- 2 V/X 0 2 Vg*

Solving (8) for I

I VD( 2 Vg* -VD) (9)

If VD = O, =. I attains a maximum when VD = Vg*. This

corresponds to the saturation of the PET.

Peae 3 8

Isat m(* ) Vg*, a vt 0 9

The transconductance a end the channel conductance g are

given by

D(11)

Therefore

+ V g* (12)

and

do= mSax : Vf Vg* 3)

Next, d2 is expressed by I in the following way

Sd W(VgW~ VD) = 2e (I*

(ss )2 : 2 (V,

2 ier I is 1 2(a 1Is \1 -Jy. in- )(I='N) (14)This relation will be used later.

(b) Theory of the thermal noise of an en-

hancement m9de FET

To calculate the thermal noise of an enhancement mode

PET by taking into account of the channel modulation effects,

the channel is divided into small sections and the contribu-

tions fron each small section are sumed up by integration.

The contribution from the section AZ at Z is calculated

as follows. The do parts of V(X) and ZL(X) are represented

by V0 (X) and Lo(X) respectively. If there is a thermal

noise AV(Z,t) in the section AZ (see Fig. 26)

vCx) v(X) o x

Therefore L-.f 27)2 I .d _dX

z

L (18)

because VDo ELo(X)dX

0

hV(Zt) 2 is given by the Nyquist formula,

AvZ,t)2 = 4T Z (per 1 cis 1 (19)

mext, the contribution fran every part of the channel is

sumed up by integration to obtain the total thermal noise.

__L L 12dZ&D) a .I+Id dX

(20)

This expression is reduced to

L L(AVD) -r- + kT S__ dZ FdX f"A " od o V V f w % 2

0 Z 0

+ 4kT 1 ~ 2 S z L] 21Vow %X)2(1

where VowZ~ e;Ig* Z1

The integrals are calculated as follows.

dZdX 2Soil 3/2

0 U[J=0

,.og ., 2 . Vx .XV 0 Lj(22)

where x~ (Sf~ Vg* 2

00L aZ Ld /LVvx)W 2= 0S2

(23)

2FX.~~ IIg+2 X

By substituting (22) into (21),

kT +3 l 2

(24)

Using the relation 2o i-sr =

gd TT is (25)

... ) ginj2 Ts [.o8 3i~ ~ ~ k [ 1V) =08(o1 I o~(26)

- 3 log. - + 6 +6 - (27

To obtain the expression for the noise current generator,

(A)2 2 2(124kT 1 +m4~

-3 log (1 4- + 6 vim I. (27)

where the relation (14) in the previous section was used.

I

) Flog~l- 3 logCl--X)-6+6 V1i-jjfO.) -2X

1 I x + 0(k2 ) (for small X)

(AI) = 4kT gdo f( ) (28)

If the equivalent noise diode current, Ieq, is introduced,

I (eq (0) - 2kT (29)f£(X), -eq gdo

t, This relation is plotted in Fig. 27. The thermal noise of an

enhancement PET becomes xero at suatration. This means that

the channel modulation effect is not strong enough to cancel

the decrease of 9d Because of this, the noise of an

enhancement PET at saturation is not due to thermal noise.

The noise of an enhancement mode PET is essentially due to

two different noise mechanisms. The thermal noise is important

for small Vd, and another (still unknown) kind of noise

predominates over the thermal noise near saturation.

Sv. aConlUatonw

The equipment designed for measuring noise in WET's

at low temperature under pulsed conditions is operating

satisfactorily and gives reliable results.

The program of studying the correlation between 1/f

noise and reliability is continuing satisfactorily. Noise

measurements have been carried out after life tests at

elevated temperatures. No conclusions can be drawn as yet.

Noise in unijunction transistors can be successfully

interpreted by drift of injected carriers when the base 2

curzent is large in comparison with the emitter current,

The interpretation of the frequency dependence of the noise

is still a problem, however.

Pulse noise measurements in FET's at liquid nitrogen

temperatures show a white excess noise. Heating effects in

the channel strongly affect the noise output. The noise

might be caused by non-ionized donors in the channel.

Noise measurements on enhancement mode FET's with

insulated gate show a spectrum caused by surface traps with

a time constant of 10-30 Ipsec. The noise at higher frequencies

is white but ncn-thermal. Sane of the noise processes seem

to be associated with the insulating layer whereas others

seem to be associated with the channel.

Present theoretical results on low-frequency excess

noise in FET's cannot fully explain the observed data.

The (IdVd) characteristic and the (g,,,V ) charac-

teristic of a diffused junction PET was calculated both for

constant mobility and for a field-dependent mobility under

the assumption of a linear impurity distribution in the

junction. The results obtained agree somewhat better with

the experimental data than the Shockley model, but the agree-

ment is not perfect.

The theory of thermal noise in an enhancement noise

FET with insulated gate shows that Ieq goes to zero if the

device approaches saturation.

VI. Progrim for Next IntervalI

It is hoped that the work in noise in unijunction

transistors can be written up in its final form.

It is also hoped that the work on lw-frequency

excess noise in FET's can be written up in its final form.

It is planned to collect noise data c sacie transis-

tors over a wide frequency range.

The work on the correlation between 1/f noise and

reliability will continue with longer and more extended

life tests.

An attempt will be made to measure noise in GaAs

lasers under pulsed conditions. Plans will be made to

measure GaAs lasers at liquid R2 temperature under C.W.

conditions.

Work on enhancement mode FBT's with insulated gate

will continue and an attempt will be made to better localize

the noise sources.

Work on depletion mode FET' with insulated gate

will be extended to higber frequencies.

An attempt will be made to pinpoint the noise source

giving the temperature dependent white noise in nomal FZT's

at liquid H2 temperature.

VII. Personnel E lployed on Project

Man Months

A. van der Ziel .45

D. C. Agouridis 1.50

R. D. Baertsch 1.50

Henry Halladay 1 .50

L. J. Prescott 1.50

M. ShoJi 1.50

A. Tong .i50

Laboratory Technical Assistants 321.3 Man Hours

Al

I.. -NoiseI diode

5.555McT

I NC15___Prap55Vc

455Kc/ Gacever no.metean

thermocoupl

The ~ ~ ~ mehoathesueenod h ni e Eat~~Fo loodtycyle

FI I

I-

A2

To prearmp

gate supply

From pulse attenuator

Fig. 2 (a) Jig for FETj(b) The balanced gate for 455Kc/s

6AK5

Gae use4-6K

(b)

A3

in

w

I--

CDL

4--

N- a-a

IL 0d

-AI -A AAU) v -11

A4

200

0.

~l0

Fig. 4Current vs. voltage for GaAs

S7 _laser at liquid Np Temperature

I 02.EIc:

2

I Voltage in Volts

70~~~~~~ R___ ___ _ _

700

40 -0

44

20 - NTeprtr

Power vs. current for GaAsLaser atliudN Teprte

R0O 0 __ ____

4

2--2

0 4 8 12 16 20Current in Amperes

A6

10K

PNP CK913

II

100

0 50 100 150 200 250-> Temperature 00

Fig, 6 Temperature dependence of Rn in PNPCK913

A7

, :Rn NPN Silicon 2N2808 (*4)

• , IOK/

10K 1c = ,OmA

I t0 50 100 150 200 250

300 350

Temperature *C-

Rn

NPN Silicon 2N2808 (3)I

1OK I JmA c

K( - , K ,m .....,.........,......

1(

IK Temperature tC0 50 I0 150 200 P50 300 350

Fig 7 Temperature dependence of Rn in NPN Silicon 2N28O8

A8

Rn NPN 2N2808 M'2)

Rn vs Ic at 3000C

lOOK,

90K,

80K

70K x

60K

50 K 0 0

40K

300

20K

IOK

0 . , I.0 . .2 .3 .4 .5 .6 .7 .8 .9 1.0

I c (mA)

Fig. 8 Rn vs. 1c in NPN 2N2808 No. 2at 300C

A9

10Rn vs. IC (300*c) (*4)-

II A

Fi89 R sI nNN220 o

AIO

(NPN) 2N2808 *6

lOOK Rn vs. Time at 28000__ ____

IOO

tI 1Rn=14KrL. (room temp.)

I Rn (2800C)

10K

IK _ _ _ _ _ _ _ _ _ __ _ _ _ _ _ _ _ _ _

0 10 20 30 40 50-0 Time (hrs)

Fig. 10 Rn vs. time for 2N2808 *6

All

n~(n) NPN 2N2808 *5

IO O KJ R n a t 2 8 0 00

Rn=27K~n at room temperature

1K K 031020 040 50

-- Time (lirs)FiglII Rn as a function of time for

2N2808 No.5

1000

Fig. 12'eq measured across

2N490A No 24 f=25Kc. Noise withIE=0 Subtracted

Cr 100 1 92=3.8mA0 I162 =2.OmAI A I8,=I.OmA

lOpA IOOjjA ImA'I

A13

Fig. 13 'eq measured across B, -B 2______ I:4mA2N49OANo.24 f=25KC

10 _ - -Noise with I =0 Subtracted

IE 2mA

IEIm0

E

Ex200pA 0

0..1.111100 mA

A14i

10Fig 14 'EQ measured across BI-Bz. IE= IOO)JA

2N489A No 12. Subtract noise with Ie:Q

I8 3.&nAI 19

I0~~C 0_____ _-0I

I Frequency Kc/si3

A15

0

00

.01 f / N

I Cj

3 A3UN -0 0

b o o0-U)

000

0i ~~00 0-

0 co w* -. 1 0

A16

800.go

6001

4001A

E 2 /

1~00 *-.0~

E8 0 * do60

gn-r

40 295*K

200

0vg(ols

Fi 76Dpnec fg n 9oo fteRAehne20tmoeFE

A17

RCA *2

2950 K

Vd=8 volt

=g3

r9

[ .--.104

=g2

V=I.5

Vg= G

I0 "..

I 10 100 1000Frequency in Kc/s

Fig 17 Dependence of the frequency spectra of the RCA enhancementmode FET on the gate voltage

AI

0CN~C

0

w W

00 0 0'.-0 0 qt

II f)be

A19

o 0 0 0 0

11 +0.

E

a)

U.

N0 0

000 03

0 0

o U-

baa

A20

0

I-I

0 cn

00- 0

doo

0 0

(Vr be,

A21

A

Vg 5~

Scale A

i001 2 95 0KVd-O

excess noise V4

above thermal noise

Vg 3I

-. Scale BU, 6LE~~ 10V-5.100

0* 0

\/gQ5 Thermal noise /RCA#I* level '

0

og

RCA#2 .W

~~10I I i I I II

I I0ie q (? A ) = 2kT.e gdo

Fig. 21 t,.,ise of RCA enhancement mode FETwhen Vd=O

A22

-y

lb b a ab2Fig.I Peerto (b)esoechrergo it h n

nI rein fo ierpoie (ifsd ucion

oa --- 4-age-----.fel c)Eecrsttc oenil

A23

Theoretical d.c. characteristics neglecting the diffu-

sion potential. Vg =0, .

1.0 _ _ _ __ _ _ _ _ _ _

.8 - - _ _Ideal

.6

.4

.2I

00 .2 .4 .6 .8 1.0y

Fig. 23

A24

1.01

Diffused junctionlinear profile

'Doceyp a Rss (Shockley)

EoL/W.. =0.5

Fig. 24 Theoretical normalized transconductance0.01 _____ vs. normalized voltage.

(Wd/Woo =l.0; Neglecting Vdif)

0 .2 .4 .6 .8 1.0z

A25

o0 C

4

ClC,-

,

IA- -

Cj

-U-

00

A26

i (Gate

Metal //

\ "I AZ ELX

sourceI

!!. ' P" ro, n

I IX

Oxide / insulator 6(X

' f EtlX)

Semiconductor L

Fig. 26 Enhancement mode FET

E: dielectric constant of insulator, Q(Z): surface charge densityat B, P(Z): volume charge density at B, y: surface mobility,V(Z): voltage difference between points A and B, Vd drainvoltage, Vg: gate voltage.

A27

1.0

0.8

0.6t

0.4

Q 2

00 0.2 04 0.6. 0.8

I/Isat

leq I'eqs(0).vsfor enhancement mode

FET calculated by thermal noise theoryFig. 2 7

DA36-039 AMC-03718 3rd Quarterly ReportUniversity of Minnesota 1 Feb. - 30 Apr. 1964

Distribution List

Number of Copies

OASD (R and Z),Attn: Technical LibraryRa 3E1065, The PentagonWashington 25, D. C. I

Chief of Research and DevelopmentDepartment of the ArmyWashington 25, D. C. 2

Director, U. S. Naval Research LaboratoryAttn: Code 2027Washington, D. C. 20390 1

Commanding General, U. S. Army Material CommandAttn*: R and D DirectorateWashington, D. C. 20315 2

Commanding Officer, Harry Diamond LaboratoriesConnecticut Avenue and Van Ness Street, N.W.Washington 25, D. C. I

Mr. A. H. Young (Code 681AIA)Semiconductor GroupBureau of ShipsDepartment of the NavyWashington 25, D. C. 1

Commanding Officer and DirectorU.S. Navy Electronics LaboratoryAttn: LibrarySan Diego, California 92152 1

Air Force Cambridge Research LaboratoriesAttn: CR)M-RL. G. Hanscom FieldBedford, Massachusetts 1

Systems Engineering Group (SEPIR)Wright-Patterson Air Force Base, Ohio 45433 1

Commanding OfficerS U. S.Army EZlectronics R and D ActivityAttn: MBEL-RD-WS-AWhite Sands, New Mexico 88002 1

Electronic Systems Division (AFSC)Scientific and Technical Information Division (ESTI)L. G. Hanscom FieldBedford, Massachusetts 01731 2

DA36-039 AM-03718 Number of Copies

Commanding OfficerU.S. Army Combat Developments CommandCommunicatione-lect-,onica AgencyFort Huachuca, Arisona 85613 1

Commander, Defense Documentation CenterAttn: TISIACameron Station, Building 5 L--Alexandria, Virginia 2 314 20

Chief, U.S. Army Security AgencyAttn: AlofS, G4 (Technical Library)Arlington Hall StationArlington 12, Virginia 2

Deputy President, U. S. Army Security Agency BoardArlington Hall StationArlington 12, Virginia 1

Commanding Officer, U. S. Army Combat Developments CommandAttn: CDCR-ZFort Belvoir, Virginia I

Commanding OfficerU. S. Army Eingineering Research andDevelopment Laboratories

Attn: STINFO BranchFort Belvoir, Virginia 2

Rome Air Development CenterAttn: RAALDOriffiss Air Force Base, New York 13442 1

Commander, U. S. Army Research Office (Durham)Box C-Duke StationDurhm, North Carolina 1

USARL Liaison OfficerRome Air Development CenterAttn: RAOLGrifias Air Force Base, New York 13442 1

APSC Scientific/Technical Liaison OfficeU. S. Naval Air Development CenterJohnsville, Pennsylvania 1

Advisory Group on Electron Devices346 Broadway, 8th FloorNew York 13, New York 3

Sprague Electric Transistor LaboratoryMarshall Street, Building 4Attn: Wr. Kurt LehovecNorth Adms, Massachusetts I

DA36-039 AMC-03718 Number of Copies

Sylvania Electric ProductsAttn: R and E Librarian100 Sylvan RoadWoburn, Massachusetts I

Semiconductor Components Library.Texas Instruments, Inc.P. 0. Box 5012Dallas 22, Texas 1

Battelle Memorial InstituteAttn: Librarian505 King AvenueColumbus, Ohio I

TRW Semiconductors, Inc.Research and Development DepartmentAttn: H. Q. North10451 West Jefferson BoulevardCulver City, California 1

t Prof. John G. LinvillElectronics LaboratoryStanford UniversityStanford, California 1

U. S. Naval Research LaboratoryCode 615Washington 25, D. C. I

Director, National Bureau of StandardsAttn: Chief, Section 1.6

(Engineering Electronics)Washington 25, D. C. I

General Electric CompanyElectronics ParkAttn: H. M. SullivanSyracuse, New York 1

University of IllinoisAttn: Dr. J. BardeenUrbana, Illinois 1

Brown UniversityAttn: John TruellProvidence 12, Rhode Island I

Minneapolis-Honeywell Regulator CompanyAttn: Mr. L. Griffith2753 Fourth Avenue SouthMinneapolis, Minnesota

Raytheon CompanySemiconductor DivisionAttn: Semiconductor Division Library150 California StreetNewton 58, Massachusetts 1

flA36-039 AMC-03718 Number Of Cop2ief

Lincoln LaboratoryMassachusetts Institute of TechnologyBox 390Attn: A. L. MicWhorterCambridge 39, Massachusetts 1

Radio Corporation of AmericaAttn: D. 0. NorthPrinceton, New Jersey 1

Bell Telephone LaboratcrieaAttn: H. C. MontgomeryMurray Hill: New Jersey 1

Philco CorporationC and Tioga StreetsAttn: Dr. C. H. SutcliffePhiladelphia, Pennsylvania1

Motorola, Inc.Attn: Dr. Bottom5005 East McDowell AvenuePhoenix, Arizona1

The Ohio State UniversityAntenna Laboratory2024 Heil AvenueColumbus 10, Ohio1

New York UniversityCollege of EngineeringAttn: Dr. K. CadoffUniversity Heights, Nov York

Director, National Bureau of StandardsAttn: Radio LibraryBoulder, Colorado1

Purdue UniversityPbyuics LibraryLafayette, Indiana

The Moore SchoolUniversity of PennsylvaniaAttat LibraryPhiladelphia, Pennsylvania1

high.. Aircraft CompanySmkiconductor DivisionP. 0. Box 278Newport Beach, California

Crystalonicm, Inc.Attn: Mr. B. B. Frusutajer14~7 Sherman StreetCambridge 40, Massachusetts1

DA36-039 AMC-03718 Number of Copies

Siliconix, Inc.Attn: Mr. Richard Lee1140 West Evelyn AvenueSunnyvale, California I

Radio Corporation of AmericaSemiconductor and Materials DivisionAttn: Dr. Robert JamesRoute 202Somerville, New Jersey 1

Mr. A. D. Bedrosian (SELRA/LNM)USAEL Liaison Officer77 Massachusetts AvenueBuilding 26, Room 131Cambridge, Massachusetts I

Marine Corp Liaison OfficeU. S' Army Electronics LaboratoriesAttn: A4SEL-RD-LERFort Monmouth, New Jersey 07703 1

Director, Monmouth OfficeU. S. Army Combat Developments CommandCom nications-Electronics AgencyFort Momouth, New Jersey 07703 1

Commanding OfficerU. S. Army Electronics Research UnitP. 0. Box 205Mountain View, California I

Director, Material Readiness DirectorateHq. U. S. Army Electronics CommndAttn: AMSEL-MRFort Monmouth, New Jersey 07703 1

AFSC Scientific/Technical Liaison OfficeU. S. Army Electronic LaboratoriesAttn: AMSEL-RD-LNAFort Monmouth, New Jersey 07703 1

Scientific and Technical Information FacilityAttn: NASA Representative (SAK-LD)P. 0. Box 5700Bethesda, Maryland 20014 1

Research Materials Information CenterOak Ridge National LaboratoryP. O. Box XOak Ridge, Tennessee 37831 1

DA36-039 AM.-03718 Number of Copies

Director, U. S. Army Electronics LaboratorieslIq. U. S. Army Electronics CommandFort MoTmouth, New Jersey 07703Attn: AMSEL-RD-DR (Research Rpts Only) 1Atta: AMSEL-RD-ADT 1Attn: ANSEL-RD-ADO-RHA 1Attn: AMSEL-RD-PF (Reports Distribution Unit) 1Attn: AMSEL-RD-PPS 1Attu AMSEL-RD-PP 1Attn: ANSEL-RD-PF (Mr. V. J. Kublin) 1.Attn: ANSEL-RD-PFS(K. Fischer)11

ADO 4 4 4 46 - apps.dtic.mil · ataare fr an pupos other '1,ha vth a eflitel opersion, reltedtheU....

Documents

Transcript of ADO 4 4 4 46 - apps.dtic.mil · ataare fr an pupos other '1,ha vth a eflitel opersion, reltedtheU....