AO-A096 303 NAVAL POSTGRADUATE SCHOOL MONTEREY CA FIG 9/1

INTERMODULATION PRODUCT SUSCEPTIBILITY TESTING OF MULTICOUPLERS--ETC(U)SP B0 D W SMITH

UNCLASSIFIED NL° IlllEEllllElll

EIUUHEHEEEEIUEEIIIIEEEIIEEIIIEE--mEE-EIIIIEEIIIEEIIIEEEEE*LIIIII7

~.j,,,- ,

NAVAL POSTGRADUATE SCHOOLMonterey, California

THESIS IINTERMODULATION PRODUCT SUSCEPTIBILITY

TESTING OF MULTICOUPLERS USING

WHITE NOISE

by

David William Smith

September 1980

Thesis Advisor: S. Jauregui

Q) Approved for public release; distribution unlimitedC.

i iI

secushry CLASSIFICATION or TwoS 1waOE fhm bate &*tape*________________mKD1S"ThuerNSREPORT DOCUJMENTATION PAGE SIM OPZIGFORm

01PN WMAIM 2. C~?ass5U No a. MCCIPICHNYI CATALGwME

Intermodulation Product Susceptibility 'Septabbe 1980~Testing of Multicouplers Using White Spebr18Noise# 41. pearon"ewa 000. IMPORT MUoomeR

1. AU THORVS B. CONTRACT OR GRANT 0$UUGCRIoj

David William/Smith

9. 009110000 IG ORGANIZATION HAN4 AND ASOREMS t- POGRAM CL CNENT. PMOJ CT. T ASIC

Naval Postgraduate School A 01ARIA&UORK UNIT NUAUIRS

Monterey, California 93940

11 CONTROLLINGe OFFICE MAN A D£omCSS

Naval Postgraduate School eamwW,Monterey, California 93940 1 "

14. wi 0 CING A49C N V NAME &COMSSett #000"flU t Ca"i~lm Ofloe.) Is SeCURITY CLASS. feE this va~erM)

UNCLASSIFIED

IS. ftA~SIICATIO/OVINGRAING

IS. DISTRIOUTION STATEMENT (fe 00. EAeotil

Approved for public release; distribution unlimited

17. DIsY11111uTIOw GTATtMENT rat too aeaosts mimd to atesm, so. to ff.woiAU hoo xAPe"

is. UuppLimamTAny NoTes

It. KEY W0303 (Csooinue an rwieo side 110006 Ande awfetv or e woGA e)

Intermodulation; Intermodulation Product; IntermodulationProduct Susceptibility; White Noise; White Noise Testing;Multicouplers; Testing of Multicouplers

2.AAUTRACT (Cere" A g lm IN o 0 O jdMHUUP W 1001 aob )

1,&Tb-ie problem or ntroaiat ion produce susceptibility and itsspecification in multicouplers is discussed. The two-tone inputand white noise loading methods for testing and determing speci-fications are presented with spectram displays. An improvedmethod of testing and determing multicoupler intermoculationproduct susceptibility specifications is presented.,,-

00 1473 toove' inor9Nv Oso t onaLteT(Page 1) seOI2-O4.et I CLAaSPICArion oP Twes FARE Wasm =awe.

Approved for public release; distribution unlimited

INTERMODULATION PRODUCT SUSCEPTIBILITY

TESTING OF MULTICOUPLERS USING

WHITE NOISE

by

David William SmithLieutenant, United States Coast Guard

B.S., United States Coast Guard Academy, 1973

Submitted in partial fulfillment of therequirements for the degree of

MASTER OF SCIENCE IN ELECTRICAL ENGINEERING

from the

NAVAL POSTGRADUATE SCHOOLSeptember 1980

Author 4

Approved by: j4.Y -A I-T 9 Thesis Adviso-r

Second Reader

Chairman, Department of ing

wDean df Scince and Engineering

2

ABS TRACT

The problem of intermodulation nroduct susceptibility

ard ±ts specification in multicouplers Is discussed. .he

two-tone input and white noise icadig ,ethods for testing

and determining specifications are presented wita spectrum

-isniays An i Droved method of testin- and deter-ninin,

mil ticourler inter-odulation no!-duct susceptibilit-r

specifiiatiors is presented.

I[

f[

it

3

TALE OF COATTNS

Chapter pd-e

I. INTRODUCTION ....................................... 10

A. INTERMODULATION ................................ 10

B. THE MULTICOUPLER ............................... 11

II. TWO-TONE TEST METHOD ............................... 15

A. SPECIFICATION .................................. 15

3. DETERMINING TH7 ORDER OF TH7E IM PRc-UcS .. 15

C. INTERCEPT POINT ................................

2. DISADVANTAGES .................................. le

III. WHITE NOISE TESTING ................................

A. BACK1ROUND ...................................... 21

B. NOISE LOADING TFST ............................. 21

C. NOISF POWER RATIO . .............................. 22

D. APPLICABILITY OF NOISE LCADING TO MULMTICCUPLE RS 24

E. EXAMPLE SPECTRA ................................ 25

IV. TEST RESULTS ......................................... 32

A. ENVIRONMENT ..................................... 32

B. IN-PiASR ADDITION ................................

C. IM TEST RESULTS ................................ 6

1. Two-Tone Tests ..............................

2. ihite Noise Test ........................... Z 6

--

g Comparison of Both Metnois .................. .45

4. CombineO Effects ........................... 52

V. EVALUATION OF WHITF NOISE .......................... :7

A. ADVANTAGES OF WHITE NOISE ...................... 57

B. DISADVANTAGES OF NHITF NOISE ................... 5E

C. NPR ACCUPAC .............................. .... b 9

VI. A NFi SPECIFICATION. ......... ...................... 62

A. FACTORS FOR CONSIDrRATION........................o2

1. Noise Test Set Specifications . ............ 62

2. )ynimic Panze . ......................... .... 6Z

3. Noise Power Input . ......................... 63

1. APOR Level . ........................ ......... 64

5. Stopbands ................................... 35

6. Order of Distortion ........................ 65

B. EXAMPLE SPFCIFICATION .......................... 6

C. T EST PROCEDURE ................................. Z8

VII. CONCLUSIONS . ....................................... 69

APPENDIX A: Signal Voltage Distribution. o1g-normalversus Gaussian ......................................... 71

APPENDIX B: Procedure for Taking an NPR Measurement .... eO

LIST OF PEFERENCES ............................ ......... 2

INITIAL DISTPIBUTION LIST ............................... E3

5

TABLE OF ?IIURFS

F i- ure Page

1. Multicouple- Chain 12

2. Third Order IM at 26 MHz and 23 Miz 17

3. Third Order Intercept Point 1 ?

- Noise Power RatiD 23

5. Ncise Generator Output 26

6. 11.7 MHz Stopband, Am)lifie-r Input and Output 29

7. 11.7 Mhz Stopband, Multicouoler Input and Output 6)i

3. Two-Tcne Test Equipment Confi,uration 35

9. Miultiple Test Tones

11. Input Pnwer versus Output Power and Third OrderIntercept Point 37

11a. *,ulticounler Output, Second and Third Order

Total IM Pcwew versus Input Power 3

lib. Third Order IM Pcwer versus Input Power 40

12. Noise Loadin Test Tquipment Confi.7uration 62

13. FP 4elA Amlifter NPR Curve for 11.7 MHzStopband 43

14. ?igh Dynamic Range Multicoupler NPR Curve 44

15. Tone Inputs and IM Output 47

16a. IM Relative Suppression and NPR versus InputNoise Power -19

16b. IM Pelativm Sippression and NP versus InputVoltage 50

17. IM Relative Sunpression versus NPR 51

6

1Qa. Combined Inputs 5-

lEb Combined Outputs 5L

19. Modulation Fffect -f Tones on Noise 50

2. HF Spect-ur-, Measurement 7quiDment ConffiEuration 73

21. HF Spectrur Sample 75

22. Spectrurr Survey Pesllts 79

23. Log-normal Probability Plot of Si-nal Voltage 76

24. Gaussian Probability Plot of Signal Voltage 79

7

7

rI

TABIJ'7 C? ~BVIIN

S? Bell Systeg Practi-es (Il.,I . pI . 21)

CIRR International Padio Crmmunications Conf.rfn-ce

of the ITU (III.A. p . 21)

CCITT International Conference on Telephone and

Telegraph Communications of the ITU (III.A. * 1. 2

dB Decibel. logarithnic untt of power ratio

d3m Decibel referenced to one milliwatt of power

3CP U.S. Department of Defense

Military High Frequency radio spectrum. 2-Z2 dz

?= Vieh Pass Filte' fi . 5(c.Y

IM l-te -rro d ula t io, n I.A. rg. 1

I? Inte-cept Point (II.C. pg. 16 and fig. 3;

ITU Inte-national Telecoe'inications Union. leneva

ffz Kilchertz

LPF Low Pass Filte- 5(d.) and fii. 8)

M-iz Me,7ane-tz

IL-STD DCD established military standard (lIA. l. 21

AlF ,Noise Power Patio (III.C. p;. 22 ind fig-. 4;

BF Padin ?requercv

r.r,.s. Poot Mean Square (II.D. pf. 20)

PS Relative Supnression of I*! below Iniput si nal level

UU T Lnit Urirer Test, either the 'reneral p)ur,3cSe

Prrrni ier cr the ni:7h d-,ari - ran~e -.it co',)Ier.

1I 1 JIOYTCT I 0

A. I TFMOrTWATION

Interrodulation products (IM) are spurious signals

occuring at the output of a device, which are not in the

input of the device or network. They are caused by nonlinear

processi- (e.g. overdriven amplification) of two or more

inDut signals. The sp',rious signal Is called "in-band" if it

occurs within the bandwidth of interest and called"out-of-band" if it occurs outside the bandwidth of

interest. In-bard I" can cause degradation of system

performance. In-band IM can occur on the frequency of a

I4sired signal and with a larger rgaritude mesk in- tie

desire! signal. In-band and out-of-band IM utilize part of

the network's output power. This can dec-ease the device

dynamic ran.!e, degrade desired signals and cause the networK

tc go into saturation or other nonlinear operations. This in

turn causes the generation of additional IM which yields

further degradation.

Two or more stronp input signals maf be enough to start

the generation of Intermodulation products. I%1 generally

occurs in any device when it is operated in or near tae

nonlinear region and has several input signals wittin its

input bandwidth. The point of concern Is now large are these

!M products and bow much can the systen tolerate.

. . . . . ... - . . .. .. . .... ,., , ,. ; ,_.. -. ,. . . .. .. 1 0 :

'.THE' ILTI OU?LF

In any'r area o L commiiniCations. ili tar,,. commercial, o r

amateur. one of the liaicr buildin", blocks of a system is ar

amplifier. This device may appear in many applications. It

may also have many por ts . One sucli device is called the

mn~ilticoupler. It connects multiple receivers or transmitters

to a single antenna or single operating position.

The Darticulir device of interest here is the receiving

multicou-pler which coupoles a singl1e i npu t t o several

biilanced outputs. The rulticoupler operates at radio

frequencies in the military high frequency (iF) band. 2 - 32

MAz. Single Inputs may be from an antenna, preamplifier. or

t'ne outvut of another miiltic-upler. An amnpl if Ier sta --e i s

reeded t o overcome the losses and orovide the overall 1 d,9

(-.3in figure (plus or minus 1 d3 the obiect being to

provile a sin.gle snurce to multiple users and not sig nal

amplification.

This device Is widelyr used in military electronics on

fixed, mobile, shipboard and aeronautical platforms. Cf

particular interest is its use at U.S. N4av.' High Frequency

Direction Finding (HFDF) sites. lere a single element or

group of elements from an antenna array, such as a

'Wullenweber circularly disposed antenna array (CDAA), have

to le fed to numerous users or prncessing points. Typically,

these multicouplers hav)e one input and eight outputs. They

are often cascaded two or three deep (fiqure 1). This aliows

8 OUTPUTS

UP To 6~4 OUTPUTS }UP To 516 OUTPUTS

Figure 1. Multicoupler Chain

12

one input to be routed to as many as 61 to 512 pricessing

points. Obviously, any errors or deficiencies occurrin4

early in the chain will propagate to all sutsequent

processin, points and degrade the entire chain.

Reference Ell discusses noise surveys made at varicus

N avy FFDF sites. Results of tnese surve-s show trlat two

major sources of system noise were III products and parasitic

oscillations. Parasitics have a two fold' effect. fhey not

only put unwanted noise or oscillations into the system. but

the resultant jower usa,,e of the oscillations reduces the

multicouplers dynamic range and produces IM as a ty-product.

The studies also showed that even in properly operating

rulticouplers rthose which conforned to specifications; IM

products and listortion were prevalent due to the aiga

density of signals in the environment. vita the next

generation f multicouple-s being realed for contrazt,

there is i need to reexamine the current methods of

specifying IM product ssceptibilitv, the adequacy if such

specifications, and the methods of testing for -orpliance

with these specifications.

Current specifications usually call for the use of two

tones or si7nals of a given voltage to be applied to the

rrulticoupler Input and the IM resultir., shculd be a given

number of de-.ibels (d) below the input siznal level. A

newer method of testing might be more realistic. For several

years a technique of loading a circuit with white noise Las

one input to be routed to as many as 61 to 512 pri-essing

points. Obviously, any errors or deficiencies occurrin3

early in the chain will propagate to all sutsequent

processing points and degrade the entire chain.

Reference [12 discusses noise surveys made at various

Navy .FDF sites. Results of tnese surve.,s show that two

major sources of system noise were IM products and parasitic

oscillations. Parasitics have a two fold' effect. 2hey not

only put unwanted noise or oscillations into the system, but

the resultant power usa,,e of the oscillations reduces the

multicouplers dynamic range and produces IM as a ty-product.

The studies also showed that even in properly operating

rulticouplers rthose which conformed to specifications; IM

products and istortinn were prevalent due to the higa

4ensity of sijnals in the environment. .ita tie next

generation -f multicouplers being real ed for contrazt.

there is i need to reexamine the current methods of

specifying IM product susceptibilitv, the adequacy of such

specifications, and the methods of testing for corpliance

with these specifications.

Current specifications usually call for the use of two

tones or signals of a given voltage tc be applied to the

multicoupler input and the IM resultir.- shculd be a given

number of de-3ibels (d3) below the input sicnal level. A

newer method of testing might te more realistic. Fcr several

years a technique of loading a circuit with white noise nas

' 1.

been use! f i- tes ting io~es ti and i rterna ti or:l telepj.,one

lines. The noise is anol e ! to ths? -i-ciiit wdith a na ! o w

f requenc y banl whicb Iias been not-:hed nut or reje, tei. At

the circ'it outp',,t. the narrow stopband is reexamined and

the increase in noise in the notch is used as the measure of

IM. thermnal noise and cross-talk caused ty the circuit. This

eff.Lectively simulates a single unused channel arron, many

loaded channels and indicates the amount of interference the

user of the clear channel would experience. Such methods can

logiCally and, realistically te extended to multicoupiers,

amnlifiers and other wideband devices.

Section It addresses the two-tone mnethcd and section III

addresses the noise loading method. Section Pr contrasts

seitions Ht and 11t. Section V -gives an evaluation of the

white noise method. Section VI offers a method of specifyinFg

IM1 p-nduct siisceptibility in multi, .ouplers and i metnod for

testing the devic e for compliance.

14

II T0-TONr T

A. SPECIFICATION

Typical specificitions fnr IM susceptibility for Navy

multicouplers ire:

CU-1382/FRR 2 - 32 MHz

Better than 65 dB below two 0.5 Volt r.m.s. input signals

CU-1399/FPR 2 - 32 Mhz

Better than 60.5 dB below two 0.25 Volt r.m.s. input

signals

The frequency or separation of the two signals are not

specified although they are chosen so that the second and K

third order IM products fall inside the bandwidth of the

device (e.-. - - 32 MHz).

B. DETERMINING THE CRDER OF TIE IM PRODUCTS

The order of the IM product is an indication of tie

combinations of the two test signal freque'ibies that result

in the IM frequency. If two signals of frequencies fl and f2

are used ther the Order of the IM products are as follows:

QA

15

2nd Crde- f = fl f2 or f -

3rd Order f 2(f 1 f? or 2(f ) - f2

=2 ft or 2(f2) - f1

4th Order f = 3(fl) - f2 or 3(fl) - f2

= -(f2) + 11 or 3(f2) - fl

= 2(fl) + 2(f2) or 2(fl) - 2(f2f

and so on.

The total number of times the frequencies enter into the

calculation indicates the order of the IM. Usually fourth

order and higher are not considered sionlficant due to their

reduced amplitude and being far removed in frequency. IM

products are 'sually identified by taeir order and relative

suppression (RS., which is the -umber of d3 (given as a

positive number) that they are down from the two equal

amplitude input signals. The IM products can be convenlentl,

measured on a spectrum analyzer presentation of tie

multicoupler output, with a reasonaole de.-ree of accuracy.

Figure 2 shows an eyample of a spectrum display with second

and third order IM. Harmonics of the two input tones are

also present at the input.

C. INTERCEPT POINT

Another way of specifvrn IM susceptibilit: is wit tqe

third order intercept point (IP). This can be determined

graphically. Output power of the fundamental versus input

16

0 DBVi

...........- 80 DBt'1

0 50 l'lHz(A.)

-80 DB.

0 50 MHz(B.)

Figure 2. Third Order IM at 26 MHz and 239 MHz

17

power of the fundamental is plotted in di. then output

po,.:er of the third order I' versus ,n-ut power cf tze

f,:ndarrental is also pltted in i3r,. The st raight line

extension cf the linear portions of each curve will yield a

-oint of intersection. The value, in d3m. on the input power

axis is the third order inut I?. The value can be rpa of'f

the output power axis and the point is called an output I?.

The input IP value will be used here. The larer tae value

of the IP the more resistant tie device will be to IM

distortion. Fig,'-e 3 shows the IP for the high dynamic range

multicoupler being evaluated. Feferen-e 12j gives a more

complete discussion of this topic. "he intercept point is

gaining in usage since it is mathematically indepenlent of

the input signal amplitude and can be specified for any

order IM desired. This technique still relies on the use of

the two-tone input.

D. SHORT COMINGS

One short coming of the two-tone test is that it is not

a good model of the real world H ' spect-2rr. The multicoupler

of interest is a wideband device, coupling the full 2 - 32

MHz range. The determination of the two-tone specification

which gives the multicoupler sufficient IM resistance,

becomes an empirical and inexact nrcblerr. Reference [11

presents survey results that show a receiving site on the

Fist Coast of the United States sufferlnx severe : problems

' 18

OUTPUT iO DBMI__ .2 B

FUNDEMENTAL

/m

-50 BM / HIRD ORDR I

-50 -5BMDB

Figure 3. Third Order Intercept Point

19

from sinals originating in Europe. The rmlticoupler at

fault was the CU-13?9/FRR whose specif ication aas teen

Riven. The specification is inadequate for the si6nal

environment encountered.

Secondly, the device has to be viewed as peaK pcwer

limited fo- both the input power it can receive and the

output power it ca.n provide. Whatever the device's iynaric

range, there will be a maximum limiting value of Input power

which produces the maximum available output power, bringing

the device to the boundary of nonlinear operation. Any

additional input will drive the device into the nonlinear

region. Complex waveforms will randomly dd in-pnase due to

the large number cf freqluencies and paases involved. If t.iis

in-phase addition exceeds the peasc input power, the device

will again be drilTen nonlinear. Hence, the pease

characteristics of the input become important. The r.m.s.

voltage or averav e power indications do not inclitde this

phase information. Therefore, the instantaneous voltage or

peak power must be used to indicate the level of complex

waveforms. If the period between input peaks, that drive the

device nonlinear, is eq'ial to or less than the relaxation

time of the device, it will saturate and be held in

saturation. The impact of this periodicity is not evident

from a two-tone test source.

20

H I ui I S S I IN

A. 3A C G .ROU D

White noise loadinr is widely used on telephrne sytstems

to measure the interrodulatior and 'rtnss tali tbat occur I.r

the system. Standards for the amount n f noise tnat is

allowed to exist have been established. These have raned

from the subjective opinion of the user to current Bell

System Practices (BSP ,. International Telecgrm,7uoications/

Union (CCIR and CCITT) recomendations, and U.s. -Alitary

standards (tIL-STD-18e). Reference [31 presets a detailed

discussion of the apnlicatcns-cf noise loading to telephone

sys tems,.

3. NCISE LOADI 1,7 TEST

The principle of the test involves loedin-; a system,

cable, repeater or comnonent with wnite noise at tne input.

The white noise is -ene-ated by a noise generator su:n as a

Marconi TF 2091B and is received with a noise receiver,

Marconi TF 2092B. The white noise is renerated over a

bandwidth wider than the systems specif.ed bandwidta and

then bandlimited with the highpass (HPF'. and lowpass (LP?;

filters to equal the system bandwidth. The noise amplitude

is uniform. TAe noise is applied to the unit under test

(UUT). A bandpass filter on the roise receiver is selected

21

and switched In line. The receiver reference meter is tnen

set to a reference level. :ext, the !enerator stcoazd

filter. which corresponds to the selected tanipass filter.

is switched in line. The total out'put noise Dower is held

constant by an output monitor (alc . The receiver reference

meter is retuirned to the reference level bj adjustin,- a set

of attenuatcrs on the receiver. The attenuator readiag .,iven

is the noise Power ratio (N?R) of the cross talk,

intermodulation. and thermal noise that exists in the

receiver jassband.

C. NOISE POWER RATIO

The ki'PR is read from the noise receiver attenvatcr. in

decibels (dB) or picowatts. The N?R is the ratio of tie full

loading noise Dower, in a narrow 'assbard, to tn.e ncise

power in the passband after a natchin:7 stopband filter has

been switched in. The rest of the system remains fully

loaded. fi.u-e 1. The NPR can also be thought of as the

number of dB an unused channel's noise is below the systerm

loidinp level. The NP includes the thermal noise power,

also called baseband intrinsic noise !3], of the system.

Sophisticated techniques are available tc measure tne

portion of the NPR that is thermal noise, as well as tnat

portion of the 11 that is below the thermal noise level.

ref. f,31. I' and 151 . For this licussion. the total IM,

cross talk, and tharmal noise will be considered since all

22

POWER

fRECEIVER PASS BAND

-PN ESPECTRUM

F

POWER

STOPBAND

F

NPR = P/P'

Figure 4. Noise Power Ratio

23

cf taese utilize the devices li-ited Dcwer ayd cor trib,,te tc

7 iePv4 n - -enlI near -,n iti ons.

D. AFPICA31:.TY OF NOIS". T OADIN& TO N 1 UL:TICUPL -R S

'vAite noise testin ; has been. cf valu e in testing

telephcne transl2tior equipmert, ,icrowavo re.eaters. 2rd

satellite transpcnders with uD to 27?.2 channels or

bandwidths up to 12 MHz. It is reasonable to e.ten this

type of testing tn multicoiprles. The twr'-toe test measures

IM pr 1uced by just two Input sI-nals. This is not a

reasonable nodel of the HF spectrum. How rany signals should

be modeled in the )assband? App)endix A shows the reslilts of

son7e samples taken of tne ? - 7-2 MHz snectrr., An "iven h.X

[Hz seg-ment. containe4 an gverece of 2ZJ sizrals for trhe a-r

and niqht measurements taken. Perhaps. a -- 1 tone test mould

be mo~e realistic. If numerous ??Z tone tests are made and

averai-e4 to develop sorre statistical descriptors. then wa at

has been achipve is the white noise test, done piecewise.

'xaminetions or processinp of received HF sigznals is

most often dore over a Pinite bandwidth ind nnt at i

discrete freoliency. The process will usually examine a small

bandwidth (e.,. 3 Kz) of the 2 - 32 MHz availatle th-ou;.h

the multicoupler. The concern Is for low I, in the : iz

tandwldth of interest. Th= sirnal of_ interest mi.st be

ava.ilable and not buried in I jenerted bTr the otner %.99'

Miz of sii5ls at the nulticouple- input. The white noise

21.

test par4llels this c ex t In. Tt uses a wi ean , s1~na!

oe: -3 a' e7ires a Sn~ll t.a-wita r f0 .

E. EXAMPLE SPECTRA

The followin<, fi-ures will show the use of tae noise

Fe-erato- and receiver set. The set used was the 1!arconi CA

2OZ-B, consistiniz of the T? 2?913 Noise ;enerator. tie T-''

21o2B 3 is re.-eive-, and the B version filters. The

;a-coni " version filters e-e those that iorpl with the

current C CIP reco -medation 39G-1 (I 9,. 2he

characteristii5 of these filters a-e riven in ref. 13' and

[5. As an eiarple, the stopband filter f- i 5..34 :iiL ha a

-KFz bandwidtei at -7. d, E -KHz at -3V d.E and 2 :*-z at

d3. The correspondin bandjass filter has a 2 K -z banddidtn

at ? dB plus cr minus 0.2 dB and at -25 d3 the bandwd th

must be less than 2t percent above or belcw the nomiral

iutoff frequencies [31.. F iz-t e 5 shcws the sOectral

characteristi-s of the noise generator output; Fi-,-res 5

'a.) an4 (t.) show the fili and bandlirnited nutputs. igures

5 (i., and (d.) show the high pass (.iPF) and low pass (L??I

filter output characteristics and fig-ves 5 (e.) and (f.

snow the 5.34 MHz and 11.7 MBHz stopband charicteristics.

Two examples of the effects of passin, tne stobanl

through a device are shown in figures 6 and 7. Figure 5(a.)

siows the noise spectrum of the 11.7 ",iz stopband at the

innut to a enerzl purpose amplifier (HP 461At and fi : ,ure

25

O L---- - L

o DB[1,

-80 DrIl

0 100 fMHz

(A.)

o DB&1

-80 DBf1

0 100 MHz

(B.)

Figure S. Noise Generator Output (A.) Full Spectrumn(B.) Bandlirnited White Noise

26

-20 DEMI

-100 DBM

(C.)

0 DEM

-80 DBM

0 5 ~

(D.)

Figure 5. Noise Generator Output (C.) 316 KI-z H-PE

(D.) 12,360 KHz LPF

27

...... .. ..- 50 DBMI

...... . . .- 130 DBM

5.29 5.34 5.39 M1Hz

-50 DBt"

-130 DBM

11.65 11.7 11.75 MHz(F.)

Figure S. Noise Generator Output CE.) 5.34 MHz Stopband(F.) 11.7 MHz Stopband

29

-35 DBrI

-115 DBM

11.675 11.7 11.725 MHz

-35 DEN

-15DBM

11.675 11.7 11,725 MHz

(B.)

Figure 6. 11.7 m4I-z Stopband, Amplifier (A.) Input

(B.) output, NPR 45.6 dB

29

rb.) s.hows tae output response. ThIe stopband hs been

fille! in" wi th irte-modiilatiin products. 're ie?. -eastred

bi thp ajbate ro is& re..ive was 15 .. A2. .e s:ec tru.'

aral:zer displa-, inlicates an NPR a-ou.r 31 dJ3.

NPR = (-22 tBm Input) - '-53 d3m koise in notch.. = 31 dz

Tbe spectrum analyzer horizontal sweep of 50 KiIz with an IF

bandwidth of 1 KHz resulted in the displ,,y averaging some of

the noise at the edges of the stonbare into the center,

givi ne the ppei-ance of a higher" NR. The fixed, center

frequeniy 2 Liz bandpass filter in tae noise receive - is not

effected in that war and gives the more accurate value 'f

N?P. Figure 7 shows a similar lisplay r the innut and

output of a high dy.ami1 rang e rulticoupler that was

underxoi. test and e-raiuatic . Th e _?. for the Tulti-oujler

was 48.2 1.E.

30I

aim'

-10 DBII

-90 DBN

(A.)

-10 D314

-90 DBM

11.65 11.7 1175 M$z

(B.)

Figure 7. 11.7 MHz Stopband, Multicoupler (A.) Input(B.) Output, NPR =48.2 dB

31

I

I'. T'IS T P 3S'UL '

A. 7'VIPONMENT

Appendix A contains a summary of data gatnered fron a

spectral analysis of the {F spectrum from 2 to 32 M!z. The

data was examined to determine the distribution of signal

arnolitudes versus the number of observed signals. The

combination of sDectrum analyzer sween ti-ie. displa:r screen

persistence and photographic film speed yield an average

power or r.,1.s. voltage sarple. kn, phase infcrmation is

lost and determination of the peak amplitudes of tae in

phase component additions is not possible. ?eak power or

instantaneous voltages will exceed the levels shown. The

data showed the mean signal amplitude was -i? dBm for tne

daytime survey and -95 dBm for the niihtire survey. )verall

the mean was -92 d3m. The dynamic ranse of the si.,nals was

F5 dB (-3 to -115 dBm)

The total number of signals observed was 12?95 of wnich

8214 were observed in the daytime and the rema!nder at

night. No signals in excess of -30 dBm were observed. Tie

noise floor of the spectrum analyzer was -120 dnr, !ie':>e

signals below -115 dBm were not counted in order to ex:lude

noise from the analyzer. The data was found to be adequately

modeled by the loe-normal distribution.

32

P. IN-?HAS- iDMTICN

In-phase aldition of the si-i-a51s is an ir;portant

consideration. The signal raa cs will eyceed tae r.-n.s.

values and ray drive the device into nonlinear operation.

This reouires that additional dynarric ran.-e be desirz'ied int

the device. This is the crux of the intermodilation p-cblem.

Sufficient dynamic range and input powe- capability will

prevent nonlinear cperation and thereby prevent

interroeulation. This dynamic range :;st account for the

total In rut instantaneous voltage or neac T~wer levels.

Iften the i-.-phase additon occurs and what are t.e maximurr

levels, are the noints bein- ar-aed. These t\,o ',vestions can

be answered fror a ;'nowled4re of the instantanecus signal

volta.e distributicn (A~Dendix A).

Ancther Perimeter effected by the nature of tae

eistributio, is the reqired relaxaticn ti.e. The relaxaticn

t irre rust ba less than the average period c.0 the in phase

adclitions that exceed the dyraric -arpe. otherwise the

device may be held in a ronlinear mode, orce driven into it.

C. IM TEST Rw3ULTS

An IM test using a single tone or freauency can be

condjcted. although this type of test is generally

inadequate. 1,o sinrle tone tests were conducted. The

twc-tone and white noise loading tests were the two types -f

tests carried out.

33

1. wc-Tcne Test

Tests using the two-tone method were conductei %%itn.

two R? sinPls at 9 and 10 MHz, w"hicn were combined in a

summing network and anplied to tne input of tie unit ur.:e r

test (T'T). as shown i r figure F. The UUT output was

otserved or a spFctr,-m anaIyzer and th- displav was

photographeO. The si,,-al levels of both the input and output

were measured from th- display ard wit a n a ? n ...

voltreter. Frequencies were verified with a di;i tal

F-'quency c unter .

Figure 9 shows a typical innut ad )utput s',ectrum.

Multiple tones were nresent at the input due to the

harmonics of the fundamental fre,,encv generated i* t.ie

si..nal .3enerator. These were partially reduced Ly filtering.,

but were nct filly eliminated. Elimination resulted in

insufficient innut power to arive the test m:lticoupler

nonlinear. Figure 1' is a -raph of outnut Dcwer versus input

power and thi-d order IM power. The third order irtercept

point is also plotted. The mIticoupler tested had a 1 d3

compression point of 24 dBM. The test equipment used could

not provide more than +12 dBM of input drive.

The power In an IM product deperds on the power in

tne fundamental frequency and the f requeniy of the IM

product itself. IM power Is assumed to be a smooth ard

quasi-linear Ounntion as is the multicorpler outptt. Figure

11(a.) shows this in a graph of the spcond and tai-d orde-

34

- .

li w1 4

0

U, W

~~C3-

00-

0

b)

CI.4

35

-0 DBN

0 50 MHz(A.)

0 D BI

-80 D -1

0 5 0 114Hz(B.)

Figure 9. Multiple Test Tones (a.) :nzut.(b.) Output.

36

OUTPUT II' = 2 DBI

O DBM

FUNDEMENTAL

-50 D/

THIR ORDR I

-50 DBM DB

Figure 10. Input Power versus Output Power and Third

Order Intercept Point

37

tital IM nwe" of the test ?lticouoler outnut. >e test

si.'nals we-e MHz and I2 1Hz. Fi,:re 11(b. shows tae

i n ividual p-- in eac nf tae f u" third orde- I'

p-oducts. These curves show the variatiors that occur in the

individual products. I, powe- should be measured by a total

for the farily of products ani not by usin? one produzt as a

representative sample.

2. White Noise Test

The output spectrum of the Marconi OA 2C 9 B Noise

Test Set is shown in fi.lure 5. Technically the white noise

driving the unit under test is bandlimited white noise. Fcr

trevity it will be referred to as white noise or 5.z-;l

noise, when no abigity will e'rist. The waite noise

produced b, the TF 20912 nlse generator is '.andlimited Lj a

316 K Hz hihpass filter and a 12.360 MHqz lownass filter.

These filters are used in testing 27 channel telepnre

systemrs and give the widest bandwidth of white ncise

commercially available. Ideally the bandwidth of the noise

should equal that of the system under test. Tne ilill syste-'

bandwidth must be loaded to sirulate full load conditicns.

ref.[3. sec. 6.21. and it is thit full load condition which

causes the saturation and I' products, ref. [3. sec. 6.1i1.

In the tests conducted. full system bandwidth loading was

not achieved. The white noise occupied only 35 perc.ent cf

the system bandwidth of the multicoupler and 7 per-ent of

the peneral purpose amplifier bandwidth. Howeve-. due tc

38

" -" ' Ill~/ i, ... .."- - " " : '' " "" "'.

D- - SECOND ORDER

DBM ____THIRD ORDER

-30

-40

W -50/X

~-60

-70

16 17 18 19 DBM

INPUT POWER

Figure 11(a.) Multicoupler Output, Second and ThirdOrder Total IM Power versus Input Power.

39

6 MHz--- 12 MHz

DBM - - 6~-30 26 2MHz

-40/

-50/

-60

-701

-600

differences in dynamic range. the aMDlifier could be driven

well into saturation, while tae miulticounler remained in a

linear iode. Hence. the actual IM nrcducts in bota cases i,-ay

te more severe thar those shown.

Figura 12 shows the test equiprent cnnfiguration for

the noise loain '- test. The test was conducted as previously

described in section III.P. The NPR for several levels of

loadin- can be plotted to gIve an N?P 'curve similar to

figure 13, for the general purpose amplifier, and fiure I!,

for the high 1,ynacmic range multicoupler. The NP1iP curvt can

be plotted in values relative to the maximum NFR loading

level, fipure 13. or in absolute values, fig-ure 14. There

are three regions on the NP? curve. They are the linear,

maximum NPR aid nonlinear repions. Reference j31 contains a

detailed discussion of these.

The tqree rep-ions are shown in figure 13. In the

linear region, the signal to noise ratio remains constant.

The therral noise is sipnificant compared to the

intermodulation and a I dB increase in loadinr- should resUlt

in a 1 4d increase in NPR. The signal to noise ratio can be

expressed by:

S lPr =-10 loi f a/(a-1) ]

SN~r is the SNR relative to the SAE at max. N?'.

"a" the predominant order of distortion.

LU LU

C/) (0

owoLU 0. Z ..)4-4

.

bO

0

0

00

w0z'wH

420

C~4 604

CD,

+ Cw z

z

z_ 04i04

w

4-4w w j

z '-I

C144

0 C=

43

NPR10 DB --

30 DB-

11.7 MHz STOPEAND

50 DB-

70 DB-

0 10 20 DBM INPUT LOADING

Figure 14. High Dynamic Range Multicoupler NPR Curve

44

In the nonlinear region, the thermal noise is

negligible compared to the intermodulation. A 1 d? increase

in loading decreases the SNRr b7Y ( a ) d3 an! iecreases the

NPR by ( a-1 ) dB. The value of "a", the predominant order

of distortion can be determined from the nonlinear region of

the NPR curve. For example, in figure 13 "a equals tnree

since:

Nonlinear Slope = 2 = ( a-i ) ; hence a = 3

In figure 14 the value of "a" can not be found since a

nonlinear region could not be measured due to insrfficient

d -i ve.

In the maximum NPR region. figure 13. the thermal

noise is bein4 overtak:en b, the Intermodulation [31. In tae

amnlifier ( a = 3 ) measured in figure 13. the thermal noise

is twice the IM at maximum NPR.

Thermal Noise Power = ( a-i ) intermodulation power. [31

In a second order system ( a = 2 ) the two would te equal at

meximum NPF. In figure 14 the maximum NPR of the

multicoupler was not reached.

In figure 13. the amplifier maximur, NPR was reached

at -23 dBm of input power. The amplifier had a specified

maximum output of 0.5 volts r.m.s. into a 50 ohm load. that

is +6.99 dBm. The 20 dB (plus or minus 1 dB) g2in settin;

was used. Thus, the maximun input power needed to Droauce

the maximum output was -1.1 d3m (plus or minus 1 dS). The

45

Ai

N? . curve snows the practical peak was reached 1 , d3 sooner

at -23 dBm. It is this type of differenmce between the

specification and actual operatlonal cajability that wnite

noise testin should eliminate. The result will be mcre

realistic sPecifications and device perforiance.

In fiiire 14. a 1 dB in.rease in noise loadinri,

produced a 4.5 dB inrease in JP instead of the expected 1

dB increase. It appears that the reascn for this is that the

thermal noise was not significantly lar7er than the

interr oulaticn, as is usually the case. The thermal noise

of this quality device was below the noise floor of the

spectrum analyzer ( -120 dBm ) and could not be measured. A

source of the IM products. which would exceed tiis low

thermal noise level, appeared to be higher orie" I'!

products. Figure 15(a.) shows a low level input to tne

multicoupler. Fizure 15 (b. ) shows fourth, ei.Yhth and ninth

order IM products just appearing at -80 dBm. This level of

IM would readily exceed the thermal noise level.

3. Comrarisor of Both Methods

It is difficult to carry out a strict comparison of

the two-tone and white noise methods due to the different

characteristics of the signal types involved. A ,reliminary

comparison was made to identify areas of superiority of one

method over the other. The comparison was conducted by

separately applying either the two tones or the white noise

cf equal total r.m.s. voltage to the nigh dynamic ran-,e

46

C. -,Le

-5 DBfl

-85 DB.11

0 100 MHz

(A.)

-5 D B."1

-85 DBII

010Mz

(B.)

Figure 15. Tone Inputs (A.) and Low IM Output (B.)

4~7

ul ticoupler. The NF' an -i the re1at. t v, 5-

the third order IA 'rodcts ere

7i-ure 16(a.) shcws a r i :f t ie K. : r .ze .,.A

P4 z and 11.7 MHz and tie 1i .3 nlrttec a-a -st tza ie . t

Dower. Fig-ure 15'b. shows the s,-,e I?1 in . I1 13 plotted

7ainst the i-rut voltage. There is rc radical Ii:fer n.e in

these cirves is w(,,ld be exoectpd wit. the rmlti~zupler

operating well within the lnea- re-ior " t-e av.ar ic

range. Figure 17 shcws the relationshi if the 11 ?3 to tie

N'PP granhicelly. This relatiorship is near linear wita the

correlation of these curves to a strairht line bein,; .

for the 5.34 MHz curve and 9.P9 for the 11.7 M-!z curve.

Therefore. for linea. operation either netnod of testing,

right be sufficient, but this is rot the case for nonlinear

cppration.

The level of IM in a device varies with the

frequency. For the multicoupler the NPR had different values

in each of the two stopbands. These were closely related as

shown in figures 16 and 17. The NPR at 5.,4 {Hz ard II.7 ,-IHz

differed by 0.4 to 2.2 dB. This means multiple samples must

be taken. With several sets of stcpband/bandpass filters,

several frequencies within the passband of the device can be

measurei to check for radical variations in the 1PR. Unless

several two-tone tests were made at a variety of

frequencies. a significant abberation could easily be

missed. Such an increase of IM in one section of the

48

C14-

00

Ln Qn

'-. I:Z

w-

/L 04

4-J

r-4

49

4 -J

an 0

_- 04

0 .~ III

N N EIii 04

r-4 Ln

ILI(D

//<

~ ci)

II 50

- - 11.7 MHz NPR

IM R DB - -- -5.34 MHz NPR

50

40 /

30/

40 3040/

NPRFiue 7 I eltveSppesinvess:P

(1

Dassband of the device, while operating in a linear mode,

could be sufficient to obscure a signal of interest.

Using the relative suppressior of an order cf I,

(usually the second or third crder) as a specification,

presumes that it is the predominent order of distortion.

Measuring the N?R to deve!Op an NP. curve -ives the order of

the distortion without any presumptions. As already shown

the slbpe of the ronlinear region gives that value. ToIdtermine the order by the two-tone method. woull a,,aln

require observation and analysis of severel orders of IM

produits at various input levels and frequencies; a lengthzy

procedure comprared to the white noise rethod. rhere is alsc

the nontrivial problem of filte-ini- out the harmonics of the

test tones at the input.

The use of noise, to load the input, as a more

realistic model of the actual HF spectrur has alread! been

discussed. The partial in-phase addition of signal peakcs, as

is true in the HF spectrum, is more licely to occur with the

widebard. multiple frequency noise than with tae twc-tone

input.

4. Combined Effects

In-band intermodulation produces two effects: one is

spurious signals which may mask desired signals; the second

effect is piwer consumption and the attendant reduction in

dynamic range. Out-of-land IM produces only the second

effect. Out-of-band IM can be elirinatel by output

5 I

52 i

V:,

filtering, but this just decreases tne lirited Dower of tae

A evice. Desi:-nin-17 an adenuate dynamic range will also

eliminate this IM. but cnce the device is driven noniinear.

out-of-band IM may occur. The amount of out-of-band IM car

serve as a reasure of the design adequacy.

An example of out-of-band IM is shown ir fiimlre 13.

Nide band noise and a group of signals were applied to the

input of the multicoupler, figure 19,a.). The multicoupler

was designed for the 2 - 32 Miz HF band, however it did not

incorporate any output filtering. As a result, the device

showed an output response to 10 M!Hz with negligible roll

off. This provided an examination of out-of-band IM. Fig;re

l3rb.' shows the output spectra. The noise only ani signal

only outputs above 80 MHz are less than -67 d.m. The output

with both signals present at the input shows hirher levels

due to the out-of-band IM. The signal at £9 Mfz is -65 d3m

vice -69 dBm and the noise is at -67 dEm vice -75 d3m. a

sinificant portion of the aveilable power is being spent

out of band.

By choosing the noise and signal levels p-operly.

the noise was given an envelope or modulation effect that

followed the envelope of the signals. This can be seen to

some extent in figure 19 by comparing the output to the

innut at 70 and 88 MHz.

53

.... ...

-5 DBfl

NoiSEINPUT

-85 DBI1-5 DBPII

I IGNALNPUT

-85 DBM

-5 DBM1

ITOTALNPUT

-85 DBM

Figure 13 A. Combined Inputs

54

-5 DBrI

NoiSEOUTPUT

-85 DDM-5 DBM

S IGNALOUTPUT

-85 DBM

-5 DBMI

TOTALOUTPUT

-85 DBAN

0 100 MHz

Figure 18 B. Combined Outputs

55

-5 DB~I

-85 DBli

(A.)

-5 DBIM

-35 DBr1

Figure 19. Modulation Effect of Tones on NoiseCA.) Input and (B.) Output

56

A. ADVANTAGES Of NOISE LADING

The value of the NPR in tne stopband Drovide,, an

excellent measure of the IM a desirei signal would ce buried

in, when the device is used. The 'cise generator and

-eceiver need rnt be colocated. thus actual si:-nal patas or

links can be measuired The NPR figure includes

intermolulaticn. cross taIh. ?nd thermal noise, all of waich

contribute to maslinr a desired signal at the output cf± the

multicouple-. The NP curve dewived frcn the noise loadine,

test provides inforration on both linear and nonlinear

operation. It also allows for the determination of tne

predominent order of the IM distortion. The noise test set,

consisting of tae noise generator. noise receiver, stopband

filters, bandlimiting filters, and bandpas s filters wita the

accompanyin, local oscillators are .ontained in two units

which are compact, weigh approximately 40 pounds each and

are relatively easy to handle. The white noise loading

technique can be used to measure In-band or out-of-band IM

usinr filter sets for the approDriate frequency. There is no

problem of filtering harmonics from the CUtput as with tne

si.znal ,enerators. Noise test set costs range from $6,Z01

for the basic mainframes to $12,0eJ for the complete set

with a nominal set of filters.

57

m om-

B. DISADVANT.7E3 OF NOI3F LOADING

No specificatiors or standards fnr noise loadin.7 tests

fr HF multicouplers are available. rnese essentially will

have to be developed from erperience and increase! usage of

the noise loading technique. The white noise used in tne

test is uniform in amplitude, the -? spectrum is not. T ris

appears somewhat unrealistic, however It is much closer to

the truth than two. sineular test tones. This will result in

a fairlyr accurate, perhaps conservative filure of IM Droduct

susceptibility. To achieve full loading of the entire

rulticoupler bandwidth, a noise ,,ererator with a waite.

bandlimited output fron 2 to 32 MLz will be needed. ',urrent

c-nerato-s are Ceared toard telephone system baniwidtaswhich are much less tan 37 M{z. The cost of the sstem will

increase if the additional bandwidth is provided.

Figure 14 shows another problem. which is inadequate

output (+,, dB') to drive the high dynanic range

multicoupler into nonlinear operation. This ray be a trade

off with the increasinr of the output bandwidth, sin .e pore

total power will be present tn the 33 'IHz bandwidth. In

either case sufficient output to cause ronlinearity is a

must. This problem was also encountered with the 2 - 32 M'Hz

laboratory signal -enerators used in the two-tone tests.

These were also lirited to a maximum output of +20 d 7. Tnis

mi,*ht be solved by additional amplification provided by a

hieh dynamic range, low noise Dreamplifier with a high tnird

58

*

[I

order irtercent. The unit under test 177= zould be measured

ty switching in the stopband and bandpass filters, ind

setting the receiver reference to zero at the output of the

preamplifier. The unit under test output is then measured to

-et the actual APR reading.

C. NPR PCCURACY

Reference [3, sections 7.7 and 9.11] contains a detaileddescription of NF acurai.les. This section will ,ie a

su-mary of that reference. The overall accuracy of the N??.

will vary with the bandwidth of the white noise, the center

frequency of the stopband and the level of tne stopband NP?..

if a s:rste has a high NP (very little IMK tnat aproaches

the NR of the sto,bmnd at the generator output. then a

correction ranging from J to 10 dB is added to the reasured

system NPR. giving an improvement in the system result. This

co-rectic - is Piven in ref. E3, section 7.7, fi-ure 7.7.1.

In alditinn there are several corrections taat are

attributable to the noise generator and noise receiver.

These will vary with the equi p rent that is usel. !he

corrections for the Marconi OA 2Z90B set are:

Accuracy of R-enerator output:

accuracy of alc mor itor .0. +-B. &3

accuracy of ortput attenuator .= +/- 2. d3

59

L( '-

Accuracy of bandlirritin filters (neligible

Flttress of frequency response f +/- .5 db

AccurAcy cf bandpass filters (ne-liu-ible)

Accuracv of receiver attenuators . +- 1.4 d

Error due to switching i. stopband:

For wideband systems ( M>7.8 z s 0.1 d3

The error iue to switching in the stophand results f-om the

stopband filter eliminating. from the inpit, sorne )f the

frequency -orponents that contribute tc the overall .eR

This could result in a better PF.. For this reason thenarrow bard "B version" filters are now reconmended. Only

one stopband filter should be switched in at a time. The

equations for combining the above factors are given in ref.

f3. section 9.111 an4 are restated below. ior wideoand

systers, bantwidths of 7 .C MHz or mo"e. the total NP errcr

is on the order of plus or mirns 1.4 43.

Linear Rerior.:e f r

Nonliiear Reion

e(a- )I ) + f r + S

The slot width error "s*" only effects the intermodulatio

products and is not a factcr in the linear reazicr.

IrI.

Tne overall acci .icy o? o NPR !-eastrer-ents is waollv

aceptable ar can be closely controlled. The meas uremerts

arp straight f±-wa-d and easily condiicted in the laboratory

or in the field.

61

I. A N'.-., S? EC 1 F117ATICON

Jiver the significance and viability of using noise

loading and NPR measurements to specify interrodulation

oroduct srsceptibility, some guidelines and requirements for

writin,, an NPR type of IM specification will te presented.

Any specific fi:ure of NR will depend on the specific

device and its application. The exarmule offered is for the

hign lynamic -anre multicounler, already discussed.

A. FACTORS FOR CONSIDERATION

I. N Test I Specificatlons V

Current noise test set specifications have been set V

forta by the International Teleccmmunications Union

committees. CCIR and CCITT. and are quite detailed. Early

problems with white noise testing nave been resolved. To

avoid relocating many of these pitfalls, it would be prudent

to utilize test sets that meet recommendation CCIR 399-1.

These correspond to the Marcori "B" version equipment. The

ITU recommendations for noise levels in telephone channels

are cf little use. but the sections of the recommendations

dealing with the bandlimitin,;. sto-band a d bandpass filters

are applicable. The equipments now available are >-eared to

the customary telephonp test frequencies. In .eneral, there

52

~ . -i

is an adequate sele.tior nf filters available whiih will

rer.uce the costs arl confusion of a proliferation of aew

filter freouencies.

2. Dvramic Range

The dynamic range of the device impacts directly on

the values of maximum .APR and maximurr N?P loadi. r".

Specifying naximum NPP loading will fix the upper eni of the

dynamic range. The lower limit would then need te be

specified by some minimum signal detectability or otner

mreasure. It would probably be test to continte to specify

the dynamic rare of the device separately. keeping in m ind

that it is closely tied to the IM product specification.

. Noise Power Injut

Noise power input levels could be snecified. If the

drnami c rarpe of the deirice and the range of NP? over the

dynamic range is specified. then the roise p wer input level

is irplied. If the IM is specified b;, a s!ngle lesired NPR

figure, then the noise power loadinp for that level of '??

should be riven. This value of noise power input could be a

nominal value for the innut loadinrg expected (also called

the zero transmission level), or it could oe the level of

input power needed to reach the 1 dB compression point.

The bandlimits also effect the input noise power

level. For iF multicouplers the band lrits are 2 - 32 MHz

an! the available noise banewidth is approxirately 12 Mi:

with a flatness of plus or minus a half of a decibel.

63 A-

Srecifications. written at this time, will nave to rezlect

this bandwidth limitation.

4. MP? Level

The NPR level could be specified in several ways.

The NPR could be eiven as a si,,le value for a given i nput

lcadinz level. such as the maximum NP. loading level or 1 dB

conpressior poirt lerel. Specifying ar interrediate value

shonld be avoided unless the slipe of tne linear region of

the NPR curve is known or also specified, and that median

value is in the linear region. In this case thie dynamic

ran-ge specificatior would have to be ziven also.

If the MPR was specified for a maximum NPE oadino-

level, a ;iven slope of the linear ran.--e and a ninim um

detectability. then the dynamic rai'e would. not need to be

separately specified. The NP could be stated as a -inimum

to maximum range. with a separate dynami! ran,-;e

specification. The coincidence of the 1 d2 compression point

loading level and the maximum NP. loadin. level should nct

te assumed. At 1 dl compression, the device may be operatin ;

in either the maximum NP? or nonlinear region. Altaough

additional study is needed on this point, section IV.C.2

indicated that maximum rNPR may occur before the 1 dB

com.vression level

Determining a reasonable value of AJPF will be

difficult initially. A selection can be maae by comparin;

the NPR of a device with its known tnllrd orler IM

F4

specif I attio. Errni-ica.lr rlaticrsnV-s such as fi-ure 17 may

be usable until exact relationships and experien:e witn

uisage have been developed.

5. Stcpbands

The stopband filter specifications saould be taken

fr~ro the current ITU recomendations. A specification saculd

state the number ar.d center frequencies of the stopbands at

which N PP will be measured. The variation of N PR with

frequency calls fnr a sampli of a minir-u, of three

stopbanOs. CCIR/CCITT recommendations show sampling at

intervals of 5Z. K~z tc 7 Mhz, deperdi., on the device

bandwidth. Spacing between stopbands should -ot exceed seven

to ten megahertz. The use of those stonband frejuencies

available on the market would produce satisfactory results

even with the ]ack of filters for frequencies above 15 M-iz.

Out-of-band stopbar's can be used tn tare out of

tand measurements provided th!e filter sets are available fcr

the frequencies desired. Out-of-band IM which right legrade

total systen performance could be observed on a spectru'

analyzer if the stopband and band-pass filter sets are not

available.

e. Order o- Distortion

A new area of specificatior tray be developed around

the NR curve. Since the order of p-edomirent distortion can

te determine.. from the NPR curve nonlinear rezion, it mi:ht

be desirable to specify this order cf distortion. This

65

sfecification would reflect the desired 2racefulress or tne

5:,stem degradation. Tbis figure would nave a major effect on

t'e device desizn and if noorlr chosen co i'd l ead to

irnerent expense or capability not required. In tne area of

class C amplifiers or aardliritin? transpcnaers. this tf pe

of specificatior colild have important impact on specifying;

,r quantizing the systepr intrmodulatior.

B. EXAM1PLE SPECIFICATIC S

?rcm te discussion and data presented, a aypetnetical

specificatior for tle experirental hi7h dynamic ranfe

rrulticoupler world bi witten. The 1 d2 compression point is

krow- to be +?-& 1qr'. The signals likely to be encountered

rane from +5.5 dBm to -60 dBm. To account for tne in-phase

aiditions, a lost--normal signal voltage distrituticr,

Appendix A. requires an additional 21 d3 of dynamic range

for a total of 86.5 d? . Assume that with nominal loa in,- of

-5.5 . Brr. a sirnal of -61 dB nust be reccveraole. This

would recire an NPR of 65.5 dP. Comparing this to figure 17.

an equivalent third order I:1 RS would be -?. 13. This may

seen to be an excessive figure wlien corrparei with zu-rent

multicoupler spciifications. However, severe IP prcblens

exist in these multicouplprs. It is wo-thwhile to realize

that the back to back (noise generator connected directly to

the receiver) NPR of the test set is around 7 to 75 d?.

Measurements at a 65.5 dB level of NP.R require close

66

ettenticn tc tne errors already disc';ssed..k summary oif tne

a rIve Aata is"

1 d3 cor.-ressio.-, piort +21 d3-i

Dyn .nic ,anpe E-.5 d3

N?F. at 5.5 eBm loalinr; -5.5 13

The sample specification might -ead:

The measred stopbanO NPR for internodulation., tnermal

ncise and crcss tali shall be not less ttan 65.5 dB for

an input loading of waite noise of a bandwidth firom 316

KHz tc 12360 KEz with a nower level of *5.5 dBm. The Ni

shall be sampled at stopbands with center frequencies of

5.34 ;1Hz an 11.7 MHz. The noise test set used snail

comply with recommendations Cc[R 399-1.

The measured stoDband NPR at 5.3M M~z and 11.7 MHz snall

be greiter than 75 dP for bandlimite4. white noise

loading less than -40 dPm and a maxirnur 4?P of 65.5 d2

for noise loir4ing of +21 dBrn. The NR shall nave a

linear characteristic between these points. The white

noise will be bandlimited at 311 KHz and 12350 l".z. The

noise test set shall meet current CCIA/CCITT

recommendations.

67

.TFST ?RCC DU'0I.

Th!e tes t D-ocedv re is s traipt i f-wa-d. The be a to t;a :!<

N?F of the test set should be chee usin. t-e -equired

input power at each stopband to be measured. These are

recorded and would be used to apply the hitch NP correction,

if± needed. The unit under test is ther. connected tetween tile

,enerator and receiver. figure 12. The bandlinitin,; filters

and noise loading level are set. For each stopband/bandpass

filter pair. the receiver reference level is set with the

bandpass filter switched in. Thea the stoptand filter is

switched in and the receiver attenuators are adjusted to

restcre the reference level. The NFR is read from tne

attenuatcr dial scales and tae necessary corrections are

applied. This uracedure is based vzon a rn cwled- e of te

Marconi Instruments OA 209eB Noise Test Set an. should be

verified agairst manufacturers instructions if other test

sats a-e used.

68

VII. CONCLUSIOIS

The noise loading techniqre has been widely accepted by

the ITU, !ell System. Intelsat. DOD and other agenc ies

involved with international and domestic telephone systems.

This method has been used. to successfully test multiple

channel. wideband translation equipment, repeaters.

transponders and cables. The noise loading technique has

been developed, tested and standardized in these respects.

Application of this technique to wideband HF devices is a

new development. No established standards are available for

HF amplifiers or multicouplers, but the application of the rtechnique is wholly valid and reascnable. Specifications.

practices and expertise will develop as usage and acceptance

increase.

The measured NPR gives an accurate value of the

intermodulation products that are generated by the

nonlinearities or power limitations of the multicouoler.

Methods for identifying both thermal noise and that mortion

of the intermodulation below the thermal noise level are

available.ref. r3], f4] and f5l. Error correction tecnniques

can reduce NPP errors to less than one decibel.

Additional research should be di-ected toward thqe

determination of the relationship between tie maximum ?R

loading and 1 dB compression points to determine if the

specification of the upper limit of the dynami- ranrge by

usinp the maximum PP lnading point, is preferable.

Determination of the relationship of the third order

intercept to the NPR will allow conversion of old IM product

susceptibilitv specifications to the NPR type of

specification. This wnuld allow a rapid brratening of a base

for NPR specification. if the relationship can be

quantified.

Intermodulation oroduct production is dependent on the

nonlinear action of the device. Additional knowledge about

the nature of instantaneous signal voltage distribution and

the frequency with which in-phase additions occur, will

broaden the understanding of how these devices Are driven to

thp nonlinear region. Combining this with a knowledge of the 1levice relaxation time and maximum NPR loading will provide

a basis for exact specifications of IM product

susceptibility that will minimize costs and provide maxirun.

performance.

Additional research should also be conducted to

determine if IM product tests are sufficient witn the

current bandwidth limits (12 MHz) of the white noise or

whether the full 2 - 32 MHz bandwidth will be necessary.

Sufficient justification, in performance gained. will be

necessary to validate the cost increases for the 30 Miz

bandwidth test sets.

4 4 -

APPiNDIX A

The distribution of signal peak voltages impacts

directly upon the dynamic range of an PF device. Two

hypotheses in this field are currently being debated.. The

debate centers on the nature of the distribution of peak

signal voltage versus the frequency of occurrence of that

voltage. The distribution determines the additional amount

of dynamic range, ever the observed signal values, tnat is

needed to withstand the in-phase addition of srall signals

that results in a sudden overload of the device input. The

two distributions being suggested as models are:

1. That signal voltages are log-normal distributed, and

an additional 21 dB of dynamic range is needed.

2. That signal voltages are .aussian distributed, and an

additional 13 d? of dynani, range is needed.

The points to be considered in arguing these two views from

an HF communications viewpoint are many and varied. They

involve consideration of the near field, far field, Rayleigh

fading, D layer absorption, geopraphic location,

polarization, atmospheric and man-made noise and other

factors.

71

A cursory look at the probler was undertaken as part of

the preparation of this paper. Figure 2? shows the test

equipment confi-uration. The procedure was to use the

spectrum analyzer to examine tae HF spectrum for 2 to 32 MHz

in 5eZ KBz segments. Fach segment was photographed for

record. The entire scan from 2 to 32 MHz was done cnce

during daylight and once during the night, to record two

distinct propagation conditions. The entire scan took

approximately four hours to record. Therefore, the frequency

scan also represents a time scan. The signals recorded on

film were broken into 5 dB ranges. The total numler of

signal peaks in each 5 dB range were counted.

Several inaccuracies were induced using this form of

measurement. No error correction was attempted due to the

overall accuracy desired. These errors include: the "Q- of

the wire antenna was insufficient for the bandwidth

measured; the antenna was ho-izontally polarized; anI hi;h

signal peaks masked low level signals on nearby frequencies

due to the nature of the photograph and the CRT trace. An

signals above -30 dBm were observed. The spectrum analyzer

noise floor was -120 dBm. Signals below -115 dBm were

omitted to prevent inclusion of spectrum analyzer noise.

72

160 FoOT

WIRE ANTENNA

I SHIELDED

LEAD

SPECTRUM

ANALYZER i!!,1CAMERA

Figure 20. HF Spectrum Measurement EquipmentConfiguration

73

The spectrum analyzer settings were:

IF Bandwidth 0.3 KHz

Verticle scale 10 d? / division

Log dB r'eference -30 to -60 dEm as needei

Horizontal scale 50 KHz / division

Horizontal sweep 5 sec/ division

Total horizontal axis 5eO Kiz

Figure 21 contains a sample of the photographs taken during

the daytime survey. A total of 12095 signals were observed

during the day and night surveys. The statistical data are:

Day

mean -100 dBmr

stnd. dev. 14 dB

signals counted 8214

Night

mean -95 dBm

stud. dev. 15 dB

signals counted 3881

Total

mean -92 dBm

stn4. de. 15 dB

signals counted 12095

Figure 22 shows the total distribution of the signals

plotted in dBm on linear scales. If the sijnal voltage value

74

UL-

Ln

LA)

C,,

C-)

CD J w

LLU-

LA -

75

NUMBER OF SIGNALS

4000- 3776

3200-

2400-

1600-

800-

-30 -60 -o ' '-1 oLEVEL OF SIGNAL PEAK

DBM

Figure 22. Spectrum Survey Results 2 - 32 MHz

76II

i

is scaled and then plotted on log-ncrmal or gaussian

probability paper, the data will give a straight line on the

paper of the proper distribution. Reference F6] was used for

this procedure of scaling and ordering the data. The data

was plotted in straight line segments represeating the 5 dB

range and the range of the number of signals in it. Figures

23 and 24 show the log-normal and gaussian plots

respectively. The conclusion that can be drawn from such

plots is whether the represented distribution is adequate,

questionable or inadequate as a model of the actual siznal

distribution. When triuncated data or samples are taken, tne

curve will usually show deviation from a straight line near

the enis f[ .

Figire 23 shows that for the signals measured the

log-normal distribution is an adequate statistical

descriptor. Figure 24 shows that the normal or gaussian

distribution is an inadequate descriptor. It would be

concluded from this, that the dynamic range of a device

receiving; the 2 - 32 MHz spectrum should include an

additional 21 dB as opposed to 13 dB.

77

100,000

10,000-

1,000-

100-

:10-

1 10 50 90 98 99.99CUMULATIVE PERCENTAGE

Figure 23. Log-normal Probability Plot of Signal Voltage

78

VOLTS CUMULATIVE PERCENTAGE

x10-7 V 1 39J 1 6p

40

.70

20

0 -60

-50

040

30 I.-

-20

10

-070 80 90 98 99.9 99.99

CUMULATIVE PERCENTAGE

Figure 24. Gaussian Probability Plot of Signal Voltage

79

I..

APPFNDI. P

The step by step procedure for taking an NPR measurement

is presented below. The procedure is for the Marconi

Instruments CA 2?92B White Noise Test Set. Reference [51

should be consulted when using this equipment. The procedure

may vary for other noise testsets.

A back to back NP should be taken for each generator

cutput level and stopban4 to be used. This value may be

needed to apply the errer corrections. The back to back test

is made by connecting the generator output directly to the

receiver input. The procedure below applies to both the bacK

to back test and the system test.

The connections for the system test and back to back

test must be made with cables that match the generator

output and receiver input impedances (75 ohms). For system

tests the generator and receiver can be ac separate

locations, such as at opposite ends of a rultiplexed cable

or microwave link. A minimum of three, widely spaced

stopbands should be sampled.

The procedure is:

1. Turn on the generator and receiver.

2. Switch in the desired highpass and lowpass bandlimitin,,

filters on the generator.

80

3. Switch on the roise output and adjust it to the level

that will be used for the system test.

4. Switch in the desired receiver bandpass filter.

5. Using the "Reference Set" knob, set the receiver meter to

the zero reference level. Adjust the attenuators if

necessary. The attenuator scales can be set to zero,

independently, without changing the attenuator knob

position.

6. Switch in the generator stopband filter corresponlin4 to

the selected reielver bandpass filter.

7. Adjust the attenuator knobs to restore the zero

reference. Do not change the "Reference Set" position. The

NPR can be read from the attenuator knobs and the receiver

meter to an accuracy of 0.2 dB.

2. Apply any error corrections that are needed.

9. Pepeat the above procedure for the remaining

stopband/bandpass frequencies to be tested.

DO NOT SWITCH MODE THAN ONF STOPBAND FILTER IN LINE AT A

TIME. Otherwise a large error will result (CCITT 228).

81

9 1.

TIPST OF PFF7-PFF

1. Cummins, '. J.. Hizh Frequenc, Radio Interferenca . E.Thesis, Naval Postgradate School. 1979.

2. *atxin-Jcqnson Cc., Technical Note Vol. i.o. 2, High"'Ynamic Rathe Receiver Parameters. McDowell. R. .. D.4 - 7. MarhAprIl 1990.

3 Tent, M. J.. The qhite Noise Bock, D. 37 - 85. MarconiInstruments, 1974.

4. Moseley, R. H., Pirkai,. S. B. and Swerdlow. R. B..Psuedo-andton Noise Loadini for Noise ?owe. Ratio andLaw of AdAition , Bell System Technical Journal, Vol.56 No. A. p. 511 - 533. ADril 1971

5. Marconi Instrurvents. White Noise Test Set 0. 2 9_BInstruction Manal. Revised J an 197;. Marconi

Instr.uments, 170.

6. Hahn. G. J. and Shaviro, S. S., Statistical Mo eli itrvrirperin-P. p. 26e - 2E4, Wiley. 1967.

82

INITIAL DISTRIBUTION LIST

No. Copies

1. Defense Technical Information Center 2Cameron StationAlexandria. Virginia 22314

2. Library. Code 0142 2Naval Postgraduate SchoolMonterey, California 9394T

3. Department Chairman. Code 62 1Department of Electrical EngineeringNaval Postgraduate SchoolMonterey, California 93940

4. Professor Stephen Jauregui. Code 62Ja 10DeDartment of Electrical Engineering

Naval Postgraduate ScbholMonterey. California 93940

5. Dr. silliam Tolles, Code 023 1Dean of PesearchNaval Postgraduate SchoolMonterey, California 93940

5. CDP. Pobert Spencer. Code 62 Sp 1Department of Electrical fntireeringNaval Postgraduate SchoolMonterey.'California 93940

7. Commandant (;-PTE-1/72) 2U. S. Coast GuardWashington, D.C. 20593

8. LT David W. Smith. USCG 1Commandant (G-OTM)U. S. Coast GuardWashinoton, D.C. 20593

83

9. C om' ander 1Aaval Se7.urit:r Group Comran4ZIE1 Netraska Avenue t"WWashington. D.C. ?0390Attn: CDR G. Preston (GSOX)

10. Commander3801 Nebraska Avenue NiWashinpton, D.C. 2039,1Attn: CDR R. K. Shields (G8 2)

11. Commande-3801 Nebraska Avenue NWWashington. D.C. 2q390Attn: Mr. R. Rich (Ge3)

12. Comiande-Naval Electronic Systems CommandWashington. D.C. 2036eAtt: MR. Ray LeSage (PME-lS?)

13. DirectorNational Security Agenc.Fort George G. Meade, Maryland 20755Attn: Mr. James Boone (R Group)

14. Commander 6Naval Electronic Systems CommanlWashington. D.C. 20360Attn: Mr. T. Lutz (PME-107-6)

15. Interstate Electronics CorporationBox 3i17Anaheim, California 92083Attn: Mr. James Counter

16. Sanders Associates Inc.95 Canal StreetNashua, New Rampshire 03060Attn: Mr. 4. Wing

84

- -

17. Sanders kssociates Inc.G5 Canal StreetNashua. New Hampshire 0336?Attn: mr. ialt Zandi

1P. Southwest Research Institute.5200 Culebra RoadP. 0. Drawer 2B510San Antonio. Texas 762P4Attn : Mr. William Sherrill

19. RCABuilding 13-5-3front and CooperCamden. New Jersey 08102

Attn: Mr. T. F. J aelev

20. Naval Ocean Systems CenterSan Diego. California 92132Attn: Mr. Jerry Pobinson

21. GT9 SylvaniaP. C. Box 205Mountain View, California 94042Attn: Mr. R. P. Cassam

Pecon. Systems Division

22. Professor W. R. Vincent, Code 6?JaDepartment of Electrical EngineeringNaval Postgraduate SchoolMonterey. California 93940

23. BDM CorporationNaval Postgraduate SchoolMonterey, Califo"ntia 93910Attn: Dr. Richard Adler, Code 62Ab

85

0 1