148 IRE TRANSACTIONS ON SPACE ELECTRONICS AND TELEMETRY September

Comments Relative to the Application of PCM toAircraft Flight Testing*

ROBERT S. DJORUPt

Summary-The pending widespread use planned for PCM Therefore, it follows that the odd harmonics are dowintelemetry has caused some confusion in the area of PCM or digital approximately 10 db for the third harmonic 14 db forstandards. It is the purpose of this paper to discuss proposed PCM astandards on a realistic basis and to show by example the organiza- the fifth, 17 db for the seventh, 19 db for the ninth andtion of an integrated digital flight test system using these standards. so on. It is quite evident that a considerable amount ofConsideration is given to the variety of testing applications and broadband energy is present in the raw serial pulse train.ultimate usage of digital flight test data. Transmission of the raw serial pulse train will require

excessive bandwidth because of its harmonic content.INTRODUCTION No information will be lost if the fundamental alone is

L. ~ IKE salt and pepper to a successful dish, two transmitted. This can be done if the serial pulse train isingredients basic to a successful PCM system represented by an equivalent sinusoidal waveform. Inare minimum bandwidth and optimum synchro- this case it is of extreme importance to note that, before

nization. Indeed these two areas have been of great transmission, the serial pulse train must be passed throughconcern to both the military and to industry. The intent a filter to remove all high-order components. Such aof this paper is to offer suggestions in defining these filter is described as a fast-settling time optimum transientareas. filter. This type of filter is used in the recovery of data

involving transients, in data in which time relationshipsTRANSMISSION BANDWIDTH must be preserved, and in the recovery of commutated

The first of these two areas is the transmission band- signals. The amplitude values of the data points in a 50width required to handle PCM/FM data. One of the per cent duty cycle square wave, commutated, or pulsefirst problems arising in bandwidth considerations is code data signal may be recovered to an accuracy ofthat no one formula or rule of thumb can produce a better than 0.3 per cent with a filter pass band of onlysatisfactory answer. It is necessary to define the serial 1.6 times the square wave frequency, thereby greatlypulse train and the successive operations to be performed improving the data signal-to-noise ratio.On it. The allowed settling time error of 0.3 per cent was

In order to keep the rf bandwidth to a minimum, determined by the need to eliminate stored energy effectsnon-return-to-zero (NRZ) transmission is used. The from the previous half cycle or bit interval. If the settlingserial pulse train may now be represented by a square time were to be degraded the received information trainwave whose frequency is equal to one half the bit rate. would be time advanced due to the previous cycle history.This applies to both commonly used types of non-return In effect, time jitter is introduced where the receivedto zero forms. These forms are NRZ (M) non-return bit interval is shortened. This is especially noticeableto zero (Mark); and NRZ (C) non-return to zero (change). with a NRZ (M) code of one-one-zero-zero-one-one; or aThe NRZ (M) method has a frequency change for a one NRZ (C) code of one-one-zero-one-one.... The NRZ (M)and no change in frequency for a zero. The NRZ (C) code of one-one-one-one... is the limiting case wheremethod has a frequency change for a one-zero or a zero- least jitter and maximum transmission reliability isone transition. realized. Since an optimum transient filter yields an

In the NRZ (M) method an all-ones code looks like output in the form of a half sine wave (one peak to thea square wave. The harmonic content is determined following opposite polarity peak) when a step is appliedby fourier analysis of the waveform and is found, in to its input, it can be seen that a degrading of settlinggeneral, to be time error results in an incomplete traversal of this

half sine wave in the bit interval.i(x) = - (cos x - 1/3 cos 3x + 1/5 A family of filters with the properties shown, in the

7 following tabulation of cut-off frequency response and

*co 5-l/7 co 7x + . .. ). settling time, are physically realizable five-pole activefilters. The entire tabulation may be scaled to arrivreat cut-off frequency in a desired range. Again it should

* Prsne at th.99IENtoa ovnin be emphasized that these are mneasured characteristicst Epsco, Inc., Boston, MIass. of a large nlumber of filters.

1959 Djorup: Comments Relative to the Application of PCM to Aircraft Flight Testing 149

20d b ; ; [ X~~~~~~~~~~~~~~~PLIIUDE

-0db_/\

-50db To0.%o| l ll

-20db \2~. .5 dC2b 51

Deci belLoss

INPUT- 30d b

OUTPUT

-40db

T=O Ts-50db Settling Time-

TO ±0.30/o fFinal Value

.2 .5 fc 2 3 4 5 10

Fig. 1-Measured response, active five-pole optimum transient filter.

Cutoff Frequency Settling Time Fig. 1 illustrates the measured response of an optimum(-3db response) (to +0.3 per cent Error) transient filter typical of the filters enumerated in the

__ --- above tabulation. The knee of the filter response curve11 cycles 71,000,isecs can be sharpened and the falloff characteristic can be15 52,000 made steeper by the use of additional poles. However,26 30,000 the practical limitations of filter construction restricts45 17,300 the maximum number of poles that can be used in a56 13,900 filter designed to be carried on board an airborne test60 13,000 vehicle. The point of diminishing return is reached when66 11,800 more than five poles are used. As an example, take the71 11,000 case of transmitting a typical serial pulse train. In the77 10,100 particular case where 350,000 NRZ bits per second is to81 9,620 be transmitted over an optimum transient transfer link92 8,480 the frequency of an equivalent square wave is 175,000 cps,111 7,020 with a period of approximately 5.72 ,secs. The time126 6,200 required for the filter to settle is equal to half the period137 5,700 or 2.86 ,secs. From the above tabulation it can be de-212 3,040 termined that the cutoff frequency of the required filter304 2,560 is about 271 kc.349 2,240 Now that the serial pulse train is in a form requiring

minimum bandwidth, the actual required rf bandwidthIf it were possible to construct an ideal filter, the cut-off can be easily determined. It is useful to recall certainfrequency would equal one half the bit rate. Since this properties of FM radio transmission.is a rather difficult accomplishment the difference between For proper reproduction of FMn all the side bands asthe actual information frequency and filter cut-off fre- well as the carrier must be amplified equally where thequency is a measure of the approximation or closeness band of frequencies to be transmitted extends fromof the transmitted signal to a sine wave as determined (fO- flq) to (fO + nfQ), wherein n is equal to the numberby the use of a physically realizable filter. of side bands of sufficient magnitude to be of importance,

130 IRE TRANSACTIONS ON SPACE ELECTRONICS AND TELEMETRY September.954 7

DEVIATION: 75,000 cps

BIT RATE: 350,000 bps

Fq 175,000cps

Mf: .428

2090 2090

.0017 .0228 .0228__I_ _I_

f- 3fq f - 2fq fo- fq fo f0+fq fo+ 2fq fo+ 3fqFig. 2-Frequency spectrum of 350,000-bit example.

an-d the bandwidth is therefore 2nfq. The modulating be determinled from the response characteristic of thesignal is f. The highest order side band which is of previously mentioned filter.importance is about mf + 1, all higher order terms If f, is taken as 271 kc, the 60-db point of the curvebeing less than 15 per cent of the unmodulated carrier, will include all components of the serial pulse train toas may be verified by referring to a plot of Bessel functions 4.7 times the filter cutoff frequency of 271 kc. Thesefor small values of the modulation index mf. An exact components go out to 1,273,700 cps. The required band-expression for bandwidth is width with 75 kc deviation is therefore

bw = 2(mff +±f). -60 db bw 2(75,000 + 1,273,700)

The bandwidth required for the transmission of 350 k -60 db bw = 2,697,400 cps.bits over an optimum transient FM1 transfer link is asfollows:

It must be remembered that, since the coding of the-3 db bw 2(125,000 + 175,000) serial pulse train is random in nature, the transmission

contains all frequencies to the left of the filter characteristic-3 db bw 600,000 cycles per second. curve. The lower frequency limnit is determined by a forcing

of the code distribution such as that caused by the inser-A deviatioln of 125 kc has been assumed. In this condition tion of word sync bits. The only conclusion that cani bathe link must have a bandwidth that is restricted to drawn is that in order to remain compatible with govern-500,000 cycles when using standard IRIG telemetering ment regulatory requirements a higher frequency tele-channels in the 216-260 MCS VHF telemetering band. metering band must be used or the bit rate must beClearly the deviation will have to be reduced to less materially reduced in order to use the VHF band. Athan 75 ke. The line spectrum for this case is showin in difficult alternate is to request assignment of additionalFig. 2. adjacent 500 kc VHF telemetering channels.

In order to state the total bandwidth required for a An extension of the reasoning that led to the above-given telemetry transmission, it is useful to consider the conclusion shows that the maximum bit rate that can be60-db bandwidth. A tentative TM standards document, accommodated by a 500-kc TM channel, when using 75referring to the VHF band, states that "The bandwidth ke deviation, is in the order of 48,000) bits per second. Aof the maodulated rf carrier including all components decrease in deviation to 50 kc will allow the transmissionattenuated less than 60 db will not exceed 500 kes." This of 55,555 bits per second. Fig. 3 shows the 60-db bandwidthpoint is of utmost importance when considering transrnis- required versus selected bit rate. These curves are validsion of high bit rate data. The example of 350,000-bit only when a filter, such as that described above, is usedtransmission discussed above is valid only when the 3-db before the transmitter and a broadband transmlitter ispoints occur at 500 kes. The total 60-db bandwJidth can used without the use of additional filtering.

1959 Djorup: Comments Relative to the Application of PCM to Aircraft Flight Testing 13110 Mc -

-5 Mc

11C4 MCDVITO

25MC E >

0

-0 0K

/ - 60 db R F BA NDW IDT H/ | ~~~~ForPCM/Fl\ TRANSMISSION|

-20KC C DEVIATION lllll

10,000 20,000 50,000 100,000 200,000 500,000 1,000,000

BITS PER SEC

Fig. 3-A -60 db rf bandwidth for PCMI/FM transmission.

SYNCHRONIZATION 1) the number of digits must be one less than an integralmultiple of four;

The second major area of concern is the subject of 2) if digits are paired counting from both ends, the oddsynchronization. Four basic sync considerations must pairs must be different, even pairs alike;be made. The first is the frame syilc word which tells 3) the simplest patterns arewhere the frame begins or where time zero is. The second n = 3 1 1 0is the word sync bit (important especially when long n = 7 1 1 1 0 0 1 0frames are used) which tells where the word begins. The n = 11 1 1 1 0 0 0 1 0 0 1 0; andthird is subcommutated frame sync which is identifiedas to position, or where to lookforit,bythecombinat 4) no longer patterns have been found and it is consideredas to position, or where to look for it, by the combination ulkl htte xs,atog hshsntbeof frame and word sync. The fourth is bit sync which proved.tells when to look at each digit to recognize its sense. proved.The received sequence of binary digits has very little It is possible to obtain longer patterns which are very

meaning unless the significance of the individual digits nearly ideal. Patterns whose auto correlation sequenceis known. It can be shown that transmission of a pre- terms do not exceed +1 (except at the center) can,arranged synchronizing pattern can perform a labelling however, be constructed by combining ideal patterns illoperation as in the case of frame synchronization. Certain ideal groups. The other possible combinations usingpatterns are especially suitable for this purpose under the above ideal patterns have the number of digitsvarious specified conditions of operation. Practical usage n = 21, 33, 49, 77, 121.dietates some deviation from what ean be shown to be The teehnique of using codes that are forbidden fromtheoretically the most desirable synchronization codes. appearing at any point in a frame other than at the

-It can be shown (the proof is beyond the scope of required point is the most desirable. To generate suchthis paper) that there exists a series of optimum group codes without duplicating or removing data codes alwayssynchronization patterns. These have been determinled means the addition of bits to the message word with aby inspection of autocorrelation sequences for various consequent bandwidth penalty. Some have suggestedcode groups. The following general rules ensue for these that analog-to-digital encoders could be restricted inoptimum code groups: range of encoding so that higher weighted codes are used

152 IRE TRANSACTIONS ON SPACE ELECTRONICS AND TELEMETRY September

for frame synchronization and sub-commutated frame digital and analog techinques in a closed loop feedbacksynchronization. It is dangerous from a transmission arrangement where the actual ground bit rate is generatedreliability standpoint to use such a technique because by a voltage controlled oscillator which is in turn excitedthere is a possibility that the random digits preceding by a digital to analog converter tied to the output ofthe pattern will combine with the first part of the pattern a forward-backward counter. The forward count line isto produce a synchronizing signal that is slightly too tied, through appropriate logic, to the serial pulse trainearly. For example if the pattern is 1 1 1 1 1 1 1 there output of the radio receiver. The backward count lineis a 50 per cent chance of it being preceded by a 1 with a is tied to the voltage controlled oscillator output which isconsequent synchronizing error. The form of the pattern operating at, say, twice the bit rate. The VCO outputshould be such that the probability of this type of error is now indeed slaved to the received serial pulse train.is minimized.A reasonable solution is to transmit message words

that are two bits longer than the data. In the case of a APPLICATIONSframe sync word all l's can be used. All other words The considerations set fourth in this paper have beeniwill begin with a 1 and end with a 0 or vice versa. In developed as a result of EPSCO's work over the past threethis manner, the frame sync word can never be mistaken. years in pulse code telemetry. Currently systems areThis can also be accomplished with the addition of a being constructed for Republic Aviation Corporation andsingle bit, but now careful attention must be paid to the Electronics Test Division, Naval Air Test Centerthe organization of the pattern recognizer because the at Patuxent River, Maryland, that will be in operationsync code does not now sit apart as before, although this year. These installations include both the vehicleit is still a forbidden code. instrumentation and the receiving or recording groundFor example, stations.

The airborne systems, in the case of the Republic F-105,2 extra bits Word 1 extra bit accomplish inflight digital recording of test data gathered

during a complete weapons system demonstration. Inputsare digital words (computer outputs), analog data points,

1 1 1 1 1 1 sync 1 1 1 1 1 quasi-static event signals and voice signals. The inflight1 0 0 0 0 0 data 0 0 0 0 0 tapes are played back on a ground station; data is selec-1 1 11 1 0 data 0 1 1 1 1 tively stripped and is then ordered into IBM-704 format

with the preparation of IBM-727 tape.data data In the case of the Patuxent systems, inflight recording

is done when the size of the aircraft permits it. The inputsThe use of two extra bits will allow the identification are similar to that of the previously mentioned systems.of the beginning of every word including the sync words When a tape recorder cannot be carried, digital tele-to be the same. This will materially aid sorting and timing metering in the 1435 mc to 1535 mc band is done. Bothoperations that are performed in the ground receiving telemetering and recording can be carried on simultan-station programmer section. eously if the test conditions require this. The aircraftSubcommutated frame sync can be a discrete word using the Navy systems will vary from extremely small

that is of any value provided that it cannot be repeated high-performance fighter aircraft to the larger recon-during the subcommutation cycle. The particular code naissance type aircraft. These PCM systems are intendedchosen can be repeated anywhere else in the frame since to facilitate overall evaluation of these aircraft and theirthe subcom pattern recognizer is, or can be, enabled by installed electronic systems.counting word sync bits, after the frame sync word, The Patuxent PCM ground station will accept bothuntil the proper time slot appears. Then and only then the airborne tape and the serial pulse train as inputs.will the recognizer examine the message word to determine The synchronizer will track any bit rate from 10,000 toif the beginning of the subcommutated frame is at hand. 1,000,000 bits per second with dynamic variations to

Proper bit synchronization often can be the toughest 50 per cent at rates in excess of 100 cps. The word lengthproblem of the lot. Practical vehicular PCM telemetering can be of any value from four to sixteen bits. The syncsystems are never in ideal environments; therefore, the pattern recognizers, operating from the serial-to-paralleluse of matched crystals, delay lines, shock excited oscil- converter, can be set to any desired code by toggle switchlators and the like are out as far as ground bit rate genera- control. Provision is also incorporated to insert auxiliarytion is concerned. The bit rate generator or sync separator data into the primary sampling format, in real time,must be capable of searching for, locking on, and tracking from auxiliary digital ground equipment. Auxiliarythe received bit pulses that are implicit in the serial analog data mnay be entered into the ground systempulse train. The tracking time constants have to be while the test flight is in progress by the use of an addi-adjustable in order to account for short term breaks in tional digitizer slaved to the ground decommutationtransmission or signal interference, programmer. Provision is included to display any selectedA way around this problem is to combine operational digital words, "quick-look" desired data channels and

1959 Contributors 153

strip and edit any of the test information. Vehicle time to their flexibility can be changed, to meet the subtlecode, also telemetered, may be used as a basis for editing. variations that are often found necessary to be incor-Similarly, quasi-static event data can be used. In both porated in standards for new research and developmentthe Republic and Patuxent systems the airborne digital test equipment.time code generator is also used as a block counter.The maximum volume of each of the airborne PCAM BIBLIOGRAPHY

systems is less than three cubic feet. There is a great [1] H. S. Black, "Modutlation Theory," l). Van Nostrand Co., Inc.degree of flexibility and versatility built into the systems New York, N. Y.; 1953.

in oderto inimze he orkof te figh tes intruen- [21, A. V. Eastman, "Fuindamentals of Vacuutm Tubes," McGraw-in ordler to minimize the work of the flight test iiistrumen- [21Hill Book Co., Inc., New York, N. Y., pp. 536-544; 1949.tationi engineers in in-itial installation anid operation. [3] S. Goldman, "Frequency Analysis, Modulation and Noise,"

a h bMcGraw-Hill Book Co., Ine., New York, N. Y., pp. 146-154.Carefl atentin ha bee pai to he fnal anne in 4] I. H. Barker, "Grouip synchronizing of binary digital systems,"which the systems will be used when installed. "Communication Theoiy," W. Jackson, Ed., ButterworthsThe systems have been designed in accordance with Ptublications, London, Eng., pp. 273-287; 1953.

[5] E. Jahnke and F. Emde, "Tables of Fuinctions," Dover Publica-the pending tentative IRIG PCMI standards and owing tions, New York, N. Y., pp. 154-188; 1945.

Con tributors

CGeorge F. Anderson (S '50-A '51) was for Radiation, Inc., Philco Corporation, raeli Ministry of Defense, Scientific Depart-born on December 10, 1924, in Mt. Carmel, Indtustrial Electronics, and Dowell, Inc. ment, as the leader of a computer researchIll. He received the B.S. degree in electrical His fields of work and study include counter- group.

engineering from the measures systems, He has published several papers in theUniversity of Miami, chaff materials, chaff field of network theory, and lectures onCoral Gables, Fla., in dispensing systems, this topic in the Israel Institute of Tech-1950. He was em- and anti-jamming nology, Haifa.ployed by the Radio techniques; radiquad Dr. Cederbaum is a member of theC o r p o r a t i o n o f multi-helix arrav sat- Association of Engineers and Architects,America, Indian- ellite communica- Israel, the Tensor Club of Great Britain,polis, Ind., where he tions antenna, as well and the Association of Applied Geometry,engaged in video cir- as directive array an- Japan.cuit design. He then tennas and propaga-joined Radiation, tion studies forInc., in Melbourne, forward propagation

G. F. ANDERSON Fla., where he was R. BAKER ionospheric scatter,appointed project forward propagation

engineer and then section manager re- tropospheric scatter point-to-point com- Samuel Cogan (S '50-A '51-M '56)sponsible for design and development munications, and circular-polarized lens was born in Atlantic City, N. J., on De-of digital data processing systems and scanning feed for TLM-18 automatic track- cember 24, 1923. He received the Bache-associated telemetry equipment. In 1957, ing telemetry antenna *and 108 mc high- lor of Science de-with four engineering associates, he founded gain r e-eivinT antenna. gree in electricalDynatronics, Inc. as an electronics develop- engineering frommenit firm in Orlando, Fla. He functioned D)rexcl Institute offor some time as vice-president and chief Technology, Phila-engineer, having both technical and man- (lelphia, Pa., in 1951.agement responsibility for a number of Isratel Cederbaum (SM '53) was born in From 1951 to 1957United States Defense l)epartment con- Warsaw, Poland, on February 4, 1910. He Mr. Cogan workedtracts. He is presently vice-president and received the M.S. degree in mathematics on digital computergeneral manager of I)ynatronics, Inc. from the University eircuit development

Mr. Anderson is a member of the Ameri- - of Warsaw in 1930, and was placed incan Ordnance Association and( the American 0 the degree of elec- charge of a computerRocket Society. N trical engineer from S. CO(AN reliability program

the Warsaw Poly- at the Burroughstechnic in 1934, and Corporation, Research Center, Paoli, Pa.

3i__F X the Ph.l). degree In 1957 he joined the D)atalab Division offrom the University the Consolidated Electrodynamics Corpora-

Richard C. Baker (M '58) was born on of London, England, tion, Pasadena, Calif., where he has workedFebruary 14, 1929, in Kansas. He received in 1956. on advanced airborne digital data acquisi-the B.S.E.E. degree in February, 1952 Sitnce 1950 Dr. tion systems.from the University of Oklahoma, Norman Cederbaum has been Mr. Cogan is a member of the Research

His seven years experience includes work I. CF.I)ERBAIU employed by the Is- Society of America.