HP ANG3 APPLICATION NOTE 63 Spectrum Analysis HEWLETT PACKARD e9Ny dithd APPLICATION NOTE 63 SPECTRUM ANALYSIS HEWLETT-PACKARD COMPANY HEWLETT-PACKARD S.A. 1501 Page Mill Road, Palo Alto, California, U.S.A. 54 Route des Acacias, Geneva, Switzerland Cable "HEWPACK™ Telephone (415) 326-7000 Cable "HEWPACKSA” Tol. No. (022) 42.8150 Printed: AUG 1968Appl. Note 63 TABLE OF CONTENTS Section I INTRODUCTION a aww ‘Scope Bye Ba tee etl History 8 a.Pe ‘The Spectrum Analyzer I SPECTRAL DISPLAY Pe ‘Time vs Frequency Domain |) 1) 2 1 tt Spectra of Common Signals 2 + + ss tt M_ DESIGN CONSIDERATIONS FOR SPECTRUM ANALYZERS Frequency Coverage. + Spurious Responses emcee Resolution oni tet Sensitivity and Amplitude Response - Dynamic Range . Detector and Video Amplifier « ae Sees General Considerations. . 2... se ee IV APPLICATIONS Pulse Radar Performance Checks Parametric Amplifier Tuning.» Varactor Multiplier Tuning Another Solid-State Harmonic Generator’ FM Deviation Measurement. Radio Frequency Interference Testing... Spectrum Signature and Spectrum Surveillance - Antenna Pattern Measurements. Frequency Measurements with the Spectrum Analyzer APPENDICA. .- SectionT- . viet 3 Waveform Analysis | + i Interpretation of the Analyzer Response Section II- Table of Important Transforms APPENDIXB . apde adage TF Amplifier Response) <2 5} st Distortion. Arar riceAppl. Note 63‘Appl. Note 63 Page 1 SECTION I INTRODUCTION SCOPE ‘Thefirstanalyzers were little more than RF indicators Spectrum Analysis may be defined as the study of energy distribution across the frequency spectrum for a given electrical signal. From this study comes val~ uable information about bandwidths, effects of various ‘types of modulation on oscillators, and spurious sig- nal generation enabling engineers to design and test RF and pulse cireuitry for maximum efficiency. Spectrum analysis is usually divided into two cate~ gories because of measurement equipment capabilities and the use of the resulting information. These are Audio Spectrum Analysis and RF Spectrum Analysis ‘This Application Note will deal with measurements at radio frequencies from 10 megacycles to 40 gigacycles which includes the vast majority of communications, entertainment, industrial, navigation and radar bands: While it is impossible to cover the complete theory and application of spectrum analysis in this Note, we shall discuss the basic principles involved, interpre- tation of spectral displays, design considerations and Limitations of the instrument and finally applications of the # wide band Spectrum Analyzer employing a 2000-Me sweep. HistoRY Spectrum analysis originally received its greatest at- tention during World War Hin designing radar oscil ators with thelr associated modulation circuits, and radar receivers. A radar pulse needed to have rea- sonably fast rise and decay times to provide good res- ‘olution of the time between transmitted and received signals. Furthermore, the pulse duration had to be short with respect to its repetition rate to give ade- quate "listening" time for returning pulses from dis tant targets. A pulse of this description generated a broad band of frequencies which a radar receiver had to accept for faithful reproduction of the pulse. Broad- ening the receiver's bandwidth for maximum return pulse fidelity necessarily meant a serious reduction in gain while degradation of the transmitted pulse sacrificed target distance accuracy. By viewing the spectra of a burst of RF with a newly developed instrument called a spectrum analyzer, scientists at the Massachusetts Institute of Technology Radiation Laboratory and other key locations were able to determine the required bandwidths of receivers for optimum gain and fidelity of a returning radar pulse. Perhaps an even more important application of the spectrum analyzer was in detecting transmitter misfiring andfrequency palling effects. The Spectrum Analyzer was originally developed specifically for this work and it has since found wide application as a sensitive receiver for attenuation, FM deviation, and frequency measurements as well as abasic tool for RF pulse study which lacked calibrated controls or broad spectram coverage but sufficed for the task then at hand. Sub- sequent analyzers have not improved much since the ‘war, yet the need for better instrumentation has grown steadily. The introduction of a calibrated wide band spectrum analyzer by @ now opens the way to more ‘accurate and convenient displays and measurements not even possible with previous instruments. THE SPECTRUM ANALYZER A Spectrum Analyzer is an instrument designed to graphically present amplitude as a function of fre- quency in a portion of the spectrum. ‘The most com- ‘mon and best known approach to this incorporates a narrow band superheterodyne receiver and an oscll- Joscope. The receiver is electronically tuned in fre- quency by applying a sawtooth voltageto the frequency control element of a voltage tuned local oscillator. ‘The same sawiooth voltage is simultaneously applied to the horizontal deflection plates of the cathode-ray tube in the oscilloscope, The output signal of the re~ ceiver is applied to the vertical deflection plates this producing a plot of signal amplitude versus frequency fon the screen. A block diagram of a basic analyzer fs shown in Figure 1 The reception and display of a signal by the instra~ ment is easily explained. Referring to Figure 1, an RF signal is applied to the input of the mixer, As the local oscillator is swept through the band by the saw- tooth generator it will pass through a frequency that will beat with the input signal producing the required IF. ThisIF signal isthen amplified, detected, and ap- plied to the vertical deflection plates of the CRT pro- ducing a plot of amplitude versus frequency, vereeron vives ‘are weurace owes osaon santaort cenERATOR Figure 1. Simplified Spectrum Analyzer of the Swept-Recelver DesignPage 2 Appl, Note 63 runaave supe Te a anPLiTaBe FREQ we Se PLAKE a, Three-dimensional coordinates showing time, frequency, and amplitude. The addition of a fundamental and its second harmonic is shown as an example. b. View seen in the t-A plane. On an oscilloscope, only the composite fj + 2fy would be seen. ©. View seen in the fA plane, Note how the com- onents ofthe composite signal are clearly Figure 2.Appl. Note 63 SECTION Page 3 SPECTRAL DISPLAY TIME VS FREQUENCY DOMAIN ‘To gain a better understanding of the usefulness and application of the Spectrum Analyzer, {t is important thatwe havea clear understanding of what the spectral display is and how to interpret it. We are accustomed to observing electrical functions with respect to time on an oseilloscope. Such meas- urements are said to be in the time domain. Pulse rise time, width, and repetition rate are read directly fon the X-axis of a calibrated cathode-ray tube, Sinu~ Soidal signals up to about 100 megacycles may be dis played on conventional high frequency oscilloscopes and with sampling techniques, up to 12.4 Ge. In all cases the observed paitern is signal amplitude versus time, Tho Spectrum Analyzer, you recall from the earlier discussion, presents signal amplitude versus fre- queney. These are measurements in the frequency Gomain where signals are broken down into their in- dividual frequency components and displayed along the X-axis of the CRT whieh ig calibrated in frequency. ‘The most powerful tool in spectrum analysis is the Fourier integral which provides the means for evalu- ating the spectral display. The application of Fourier analysis is contained in Appendix A and may also be found in a great many texts. In this portion of the ‘Application Note, however, we prefer to discuss the display in more basic terms SPECTRA OF COMMON SIGNALS CW Signals. If the analyzer's local oscillator sweeps through a CW input signal slowly, the resulting re- sponse on the screen is simply a piot of the IF ampli- fler passband. A pure CW signal will by definition have energy at only one frequency and should there fore appear as a single spike on the analyzer screen. ‘This will occur provided the total RF sweep width or s0-called""spectrum width” is wide compared tothe IF bandwidth in the analyzer. As spectrum width is re- ucedthe spike response begins to spread outuntil the IF bandpass characteristic begins to appear in detail as in Figure 3. Amplitude Modulation. When a CW signal of frequeney ‘Fz is amplitide-modulated hy a single tone f,, side bands are generated at Fo #f,. The analyzer display of this will be the carrier frequency Fe flanked by the two sideband frequencies whose amplilude relative to the carrier depends on the percentage of modulation as shown in Figure 4. Note that if the frequency, spectrum width, and ver- tical response of the analyzer are calibrated you are able to determine: 1) Carrier frequency, 2) Modula tion frequency, 3) Modulation percentage, 4) Non- T's mathematical analysis may be found in Appen- dix A, Section I. a Figure 8, Spectrum of a CW Signal inear modulation, 5) Incidental FM (as evidenced by jitter of the spectral lines), and 6) Spurious signal lo- ‘cation and strength. Frequency Modulation. If aCW signal F, is frequency mnodulated at a rate f, it will theoretically produce an infinite nuiber of sitiebands.? ‘These will be located at intervals of Fg 4 nf, where n= 1, 2,3, ..- As a practical matter only the sidebands containing ‘significant power are usually considered. For a quick approximation of the bandwidth occupied by the sig- nificant sidebands, multiply the sum of the carrier deviation and the modulating frequency by two: i.e. BW = 2(AFo+f,). An FM display is shown in Fig- 40 tN . i VG Figure 4. A Single Tone Amplitude Modulation, a) Time Amplitude Plot b) Frequency Amplitude Plot, ?'c.L, Cuccia "Harmonics Sidebands and Transients in Communication Engineering’ First Ed. p. 255, MoGraw-Hitl.Page 4 tn ke b 3 —— Z ze Figure 5, Amplitude Spectrum of Single Tone Frequency Modulation Pulse Modulation. The formation of a square wave from @ findamental sine wave and Its odd harmonics is a good way to start an explanation of the spectral display for non-sinusoidal waveforms. You willrecall perhaps at one time plotting a sine wave and its odd harmonies on a shoot of graph paper, then adding up all the instantaneous values. If there were enough harmontes plotted at their correct amplitudes and phases the resultant waveform began to approach a ‘Square wave. The fundamental frequency determined the square wave rate, and the amplitudes of the har- monies varled inversely to their mmber. A rectangular pulse is merely an extension of this principle, and by changing the relative amplitudes and phases of harmonies, both odd and even, we can plot fan infinite number of waveshapes.? The Spectrum Analyzer effectively "un-plots” waveforms and pre~ sents the fundamental and each harmonic contained in the waveform, 3 M1. Radar School Staff "Principles of Radar", Second Ed. pp. 4-12, McGraw-Hill, 1946. -t & c= — Figure 6a, Periodic Rectangular Pulse Train AVERAGE FUNDAMENTAL aise SUM OF FUNDAMENTAL, 2ND HARMONIC AND ‘AVERAGE VALUE Figure 6b, Addition of a Fundamental Cosine Wave and its Harmonies to Form Rectangular Pulses Appl. Note 63 Consider a perfect rectangular pulse trainas shown in Figure 6a, perfeet in the respect that rise time is zero ‘and there 18 no overshoot or other aberrations. This pulse 4s shown in the time domain and we wish to ex- Amine its spectrum so it must be brokendown into its individual frequency components. Figure 6bsuperim- poses the findamental and its second harmonic plus a eoustant voltage to show how the pulse begins to take shape as more harmonics are plotted. If an infinite number of harmonics were plotted the resulting pulse ‘would be perfectly rectangular. A spectral plot of this would be as shown in Figure 7. SPECTRAL LIN we eS AMPLITUDE 3 FREQUENCY, F Figure. Spectrum of a perfectly rectangular pulse. “Amplitudes and phases of an infinite number of harmonies are plotted resulting in smooth envelope as shown. ‘There is one major point that must be made clear be~ fore going into the analyzer display further. We have been talking about a square wave and a pulse without any relation to a carrier or modulation, With this background we now apply the pulse waveform as am- plitude modulation to an RF carrier. This produces Sums and differences of the carrier andallof the har~ monie components contained in the modulating pulse. ‘You recall from the discussion on single tone AM how ‘the sidebands were produced above and below the ear- rier frequency. ‘The idea is the same for a pulse ex- cept that the pulse is made up of many tones thereby producing multiple sidebands, which are commonly referred to as spectral Lines on the analyzer display. In fact, there will be twice as many sidebands or spectral lines as there are harmonics contained in the modulating pulse. Figure & shows tho spectral plot resulting from rec~ tangular amplitude pulse modulation of a carrier, ‘The individual lines represent the modulation product of the carrier and the modulating pulse repetition frequency with its harmonics. Thus, the lines will be spaced in frequency by whatever the pulse repeti- lon frequency might happen to be. The spectral line frequencies may be expressed as: FL = Fotnly where F, = carrier frequency 4 = qulse repetition frequency n= 01,23,Appl. Note 63 =" & =a Figure 6. Resultant Spectrum of a Carrier Amplitude ‘Modulated with a Rectangular Pulse ‘The ""main lobe" in the center and the"side lobes" are shown as groups of spectral lines extending above and below the baseline. For perfectly rectangular pulses: and other funetions whose derivatives are diseontinu- ‘ous at some point, the number of sidelobes is infinite. ‘The main lobe contains the carrier Lrequency repre- sented by the longest spectral line in the center. Am- plitude of the spectral lines forming the lobes vary as lll, TTT TTT Page 5 a funetion of frequeney according to the expression ned oF t for a perfectly rectangular pulse. Thus, the points where these lines go Uirough zero amplilude are de- termined by the modulating pulse width omy for a fiven carrier ireqieney. "Aa pulse” wicth becomes Shorter, minima of the envelope become further re~ moved in frequency from the carrier, and the lobes become wider. The sidelobe widths in frequency are related to the modulating. pulse width by the expres= sion = 1/7. Since the main lobe contains the origin of the spectrum (the carrier frequency), the upper and lower sidebands extending from this polnt form a main Jobe 2/7 wide, Remember, however, that the total number of sidelobes remains constant so long as the pulse quality, or shape, is unchanged and only its Fepetition rate is varied. Figure 9 compares the spectral plots for two pulse lengths, each at two rep- tition rates with carrier frequency held constant, mm im” Figure 92, Narrow pulse width causes wide spectrum lobes, high PRF results in low spectral line density. iLL il | iLL Figure 9b, Wider pulse than 9a causes narrower lobes but line density remains constant since PRF is unchanged. Figure 9c. PRF lower than Ga results in higher spectral density. Lobe width is same as a ‘since pulse widths are identical Figure Od. Spectral density and PRF unchanged from 9c but lobe widths are reduced by wider pulsePage 6 Table 1. Appl. Note 63 Summary of Pulse Spectra Characteristics scteoe it Line Spectrum Spectral Intensity Plot Number of responses per em on CRT Requirement for each type of spectrum at<<1, Lecatccd T eT Felt atts AGN Sei Amplitude of spectrum display at w= reff gt K reff af E = aan B = TR, changes with Ts changes with sweep spectrum width, not sweep time. time, not spectrum width. Approximate time duration of each response | af Ts 1 independent of fr ar spectrum width and sweep time, response on CRT due to CW signal At = IF bandwidth (8 ab) Bpposwet. K = constant of IF amplifier Ts = sweep time In sec/em Fett = width of rectangular pulse of same height and area as pulse applied to Fs = spectrum width in eps/em analyzer = ft p(t)at on Ep Figures 8 and 9 bring about an important consideration about the actual display on a spectrum analyzer. Notice in the drawings how the spectral lines extend below the baseline as wellasabove. This corresponds to the harmonies in the modulating pulse having a phase relationship of 180° with respect to the funda mental of the modulating waveform. Since the Spec- trum Analyzer can only detect amplitudes and not phase, it will Invert the negative-going lines and dis- play all amplitudes above the baseline, Thus the Phase spectrum is lost and the display 1s not unique for a given function. Further discussion of this may bbe found in Appendix A, Section IB. This does not seriously limft the usefulness of the Spectrum Ana- Iyzer, however, as will be evident in Section IV deal- ing with applications: ‘A special case of the rectangular pulse spectrum is, the spectrum of an “impulse”? function. Ifthe process, of narrowing a single pulse could be continued far enough, the result would be a main lobe infinitely wide and hence would exhibit constant amplitude across the spectrum. A perfect impulse is a rectangular pulse of infinite height and zero width withunitarea. If such ‘a function could be produced it would contain all fre- quencies in fhe spectrum equal in amplitude, ‘While it Is not possible to generate a perfect impulse it may be approached for many practical purposes. Innpulse testing has been found useful in amplifier and network response testing and RFI meter calibration. In Section IV, we will show some applications of the spectrum analyzer to impulse measurements. Line Spectrum & Spectrum Intensity Plots. The Spectrum of 2 periodic function has been shown to consist of a series of harmonic components sepa- rated in frequency by the repetition frequency. If the IF bandwidth of the spectrum analyzer is narrow compared to the pulse repetition frequency, the an~ alyzer will respond to only one component at a time. If the analyzer 1s swept slowly past the spectrum, a true amplitude spectrum display of lines will result (true in the sense that there is exactly one response or Une for each frequency component within the swept region). Figure 10a.Appl. Note 63, In many instances, it is neither possible nor desirable to make a fine grain line-by-line analysis of a spec- trum. A good example of such a case is a train of short RF pulses at a low repetition frequency. Not only must the IF bandwidth become inconveniently narrow, bat often the frequency modulation on the pulsed carrier would be so excessive as to make the resulting display confusing. In such a case, it is possible to oblain a plot of the spectral envelope, but without individual Lines. To do this, it is only necessary that the analyzer IF band~ width be narrow compared to the reciprocal of the pulse width rather than the pulse repetition rate. As the analyzer is swopt slowly past the pulse Page 7 spectrum, the IF amplifier responds to several spectrum lines at once (Figure 10b). If the pulse repetition rate is low enough, the display is essen- tially the impulse response’ of the IF amplifiers (see Figure 10c). When the analyzer is swept past the spectrum, the response appears as a series of “pulses”, ‘rather than lines, having amplitudes corresponding to average values of the magnitude of ‘the input pulse spectrum lines. ‘The number of these “pulses” is a function of the PRF and the scan time, In fact, reducing the frequeney seanto zero will enable you to see the time domain pulse waveform. This is the spectral intensity plot. Because of the importance of spectral intensity analysis, the display character- istics are summarized In Table 1. thts Lith ‘ li, wlll “Line Spectrum the spectral intensity plot Fintfion Bandwidth containing several spectral Lines ‘The Lines within the bandwidth of (b) are averaged when the pulse is "on" to present "'$" is a funetion of scan time and PRF. Figure 10,Page & Appl. Note 63, a. Spectrum Analyzer using swept second L. 0, b, Spectrum Analyzer using swept first L.0. This facilitates using the analyzer as a general laboratory tool Figure 11Appl. Note 63, SECTION Page 9 DESIGN CONSIDERATIONS FOR SPECTRUM ANALYZERS FREQUENCY COVERAGE Spectrum Width and Tuning Range. The widest range sFteequencies that can be Sbserved ina single sweep is known as spectrum width. ‘The basic limitation to spectrum width is the sweeping capability of the local oselllater, ‘The maximum sweep width of various oscillator types is reasonably well fixed by their inherentdesign. Kly- stron local oscillators have long been employed be~ cause they offer fundamental frequencies in the radar band and are capable of limited electronic sweep. Using a klystron local oscillator in the arrangement shown in Figure 1, reflector voltage sweep will typ- leally produce maximum spectrum widths of 50 to 80 megacycies. In recent years most spectrum analyzers have used a double conversion scheme with a wideband first IF and ‘a swept second local oseillator of the triode-reactance tube modulator variety. (Figure 11a) The advantages here are that stable sweep circuits are easier to de- Sign at the lower IF frequency and sweep calibration does not change as the input frequency is tuned, With such a system, however, the spectrum width is re~ duced to the bandwidth of the first IF whlch is again about 80 Me. ‘Basic input tuning range is restricted to that of the klys~ tron first LO which must be tuned through mechanical linkages to the cavity, repeller tracking potentiometers and a frequency dial. ‘This method is subject to the age-old problem of tuning backlash and mode track- ing. Ta molti-band analyzer is desired, there either must be a local oscillator for each band with its cor~ responding tuning mechanism or a broadband mixer employing local oscillator harmonies to extend the frequeney range without additional local oscillators. The latter method results in some loss of sensitivity and increased possibility of spurious responses (dis~ cussed later). By careful design of the mixer, these problems can be largely overcome and the convenience ‘of harmonic mixing enjoyed. ‘The newest approach to solving spectrum width prob- lems employs a backward-wave oscillator (BWO) asa swept first local oscillator. (Figure 11b) Hewlett Packard has chosen a 2 to 4 Ge BWO for the 85914 Spectrum Analyzer which allows the instrument to sweep a wide as it can tune and results in 2-Ge spectrum coverage in a single display. Frequency 1s determined by the BWO helix voltage virtually elimina ting tuning backlash and providing the required ability to be electronically swept. A BWO inherently exhibits more residual FM than’ a xlystron which becomes serious when analyzing narrow spectra, The ® analyzer overcomes this with aunique frequency stabilizing cir cuit that reduced BWO residual FM from 90ke to less, than ke. Ahigh-frequency, wide-sweeping local oseil~ lator plus careful selection of the IF frequency will result in more usable spectrum width, Usable Spectrum Width. The mixer and swept local oscillator jointly translate a signal at frequency F, to @ response - producing signal at the IF frequency Fyy whenever mFy = Fig + Fig « 2 Sex 233). al naeillator frequency. If the amplitude of signals at frequency F, are small ‘compared to the local oscillator voltage, ‘the system 4s said to be operating linearly and m= 1 Assuming m = 1, two families of analyzer tuning curves may be plotted from equation (1) for two choices of IF frequeney. ‘These curves aro shown in Figure 12. Figure i2a represents the response loca- tions when the IF frequency 1s chosen to be equal to the minimum local oscillator frequency which is 2 Ge in this case. Figure 12b Is the result when the IF is chosen to be one tenth of the minimum local oscillator frequency or 200 Me For each n there are two possible bands of signal re- ception: one above and one below the frequency nfo. ‘These are shown as nt and n= respectively. ‘The sep- aration of these bands of possible responses is twice the frequency Fig. In most applications, the presence ff the so-called Mimage” signal is suiflolently annoy- {ngto reduce the usable spectrum width to half of this value or Fy. Note the comparatively wide separation ‘of responses for each n when Fi is chosento be equal to fig (min). ‘This, of course, allows a much greater spectrum widthto be displayed that is free from Image responses. For this reason a 2-Ge IF is used in the & analyzer. SPURIOUS RESPONSES Whenever a nonlinear resistive element such as a diode mixer is excited by an RF signal of sufficient amplitude to vary the mixer's conductance, there will be harmonies generated. This can be advantageous in the case where harmonic mixing is desired to extend the frequency range of the spectrum analyzer. In this case the local oscillator voltage is large enough to vary the conductance at the local oscillator rate re- ssulling in harmonic generation and mixing action. Unfortunately, the conductance can also vary at thein- put signal rate. As the inpat signal amplitude ap- proaches the local oscillator voltage level, vigorous harmonics are generated which ean combine with the Tocal oscillator harmonics to produce the IF frequency. ‘This is then passed through the IF amplifier to the de- tector and subsequently appears on the analyzer dis~ play. Such responses are termed "spurious" since they do not represent true Input signals of the indicated frequency. The amplitude of spurious responses is,Appl. Note 63 ‘0 2 : 3 . ss a & : 2 2 2 ' ' 3 { °. ag ga ES Mae ae a aaa eG gee a Fo Fo secon Figure 123, Analyzer Tuning Curves for Figure 12b, Analyzer Tuning Curves for Fit = Figlia wy = Enis Fram shee F, = nF, + 200 me 3° "Fly nonlinearly related to input signal level and can ser~ fously clutter the display of 2 spectrum analyzer mak- ing it impossible to distinguish true inputs from the false signals produced by the mixer. No spurious re- sponses would be visible onthe analyzer display if the Input signals were kept extremely small. Doing this, however, would severely limit the amplitude range of the instrument, making it of little practical value. Many analyzer discussions ignore spurious responses or define them as those responses that appear with no input signal applied ata Such phenomena are better defined as "residual responses” and should not be con- fised with spurious responses when discussing ant— lyzer specifications By careful design of the coaxial mixer in the @ ana- lyzer, spurious responses are greatly reduced over those in other analyzers. ‘This is accomplished with- out imposing impractical Limitations to the inpat level, fand the 60-db dynamic range of the instrument 1s maintained. Figure12c is a plot of the @ analyzer input tuning curves from equation (1) with the spurious producing, frequencies added. ‘The heavy solid lines represent the best tuning range for each local oscillator har- ‘monic number n. The lighter solid extension of these Mines merely shows that primary signal reception is possible above and below the heavy Lines, but not use- ful because the tuning curves become’ too closely spaced at either end. This narrow spacing would ‘cause a single frequency at the high or low end of the sweep to appear twice at close intervals on the ana~ lyzer display. ‘The dashed tines solve equation 1 for input frequencies of -380 dbm amplitude which will produce spurious responses less than 60 db below the input reference level. The figures in parentheses at the end of each Line represent the n, m and +or- terms respectively, satisfying the equation mfs = nfig + fig. The db figure ‘Shown along each line is the typical amplitude of the spurious response compared to a -90 dbm signal frequency of mf if such a signal were applied to the input, Stronger input signals cause larger spurious responses and more of them. It is recommended that the analyzer input attenuator be used to keep signal snput to the mixer at -30dbm or less for minimum spurious generation and full 60 -db dynamic range. Here is an example of how to use Figure 2c in pre- dieting spurious responses for a given frequency in- ‘put. We are operating the analyzer on the n= 2 range because we are interested in aome signal in the 210 10Ge region and therefore want to usethe second har- monic of the local oscillator to tune this band. As— sume there is also a -80 dbm signal applied to the mixer of 1.8Ge. To find the frequency of the spurious response caused by the 1.8-Ge signal, enter the graph at 1.8Ge on the ordinate. Move across the graph untilAppl. Note 63 InpUT FREQUENCY 6c 30H m8 aS LOCAL OSCILLATOR GC wee mF, =nF,,+2Ge igure We, Spectrum Analyzer Tuning Curves. Heavy solid lines Indicate desired response locations {or input signals of 10 Me to 10 Ge. Spurious ‘responses are indicated by the dashed lines. intersecting the dashed line at a local oscillator tre- quency of 2.8Ge as indicated on the abcissa. Note the figures in parentheses at the end of the line to see which mixing product results in a spurious response. In this example the figures are (2, 2-) which indicate the lower mixing product of the local oscillator second harmonic and the second harmonic of the 1.8Ge input causea spurious response. Now move vertically from the intersection just located until intersecting the main heavy tuning curve for n=2-. Read the ordinate point directly opposite this Intersection for the fre- quency of the spurious response as indicated on the display. This response is 60 db below the normal re- sponse to a -80 dbm input signal at 3.6 Ge. Notice that some of the dashed lines in the graph do not extend the full range of the local oscillator sweep. This indicates that spurious signals produced above these polnts are more than 60d down referred to the Page 11 input, thus making them insignificant since they fall into the noise level on the analyzer display. See Sec- tion IV on applications in RFI testing to see how pre— selection filters may be used to provide virtually spurious-free analyzer displays RESOLUTION Ih general, resolution is the ability of the analyzer to give an accurate presentation af the frequency distri- bution of the signals present at the input. A more pre~ cise definition would be purely arbitrary. Several factors affect the resolving capabilities of an analyzer. Since the response to a CW signal is 2 plot of the fre~ quency response of the IF amplifier, the most serious Imitation to resolution is the width and shape of the passband. Two signals with a frequency separation much less than the IF bandwidth would not be individ~ ually distinguishable. Another consideration is the sweep rate, If the local, oscillator is swept past the input frequency at too great a rate, the apparent bandwidth of the IF ampli- fier will be wider than the actual bandwidth.? This ‘ean be explained in simple terms as follows: in Fig ure 13 we see two response curves for a gaussian IF amplifier superimposed. If the input frequency is ‘swept slowly through the IF amplifier passband, the ‘output level will have time to reach full amplitude, tracing out curve A. Note the 3-db points of this, curve with respect to frequency --this is the true bandwidth. If, however, the local oscillator sweeps through at a high rate, the amplifier output will not have time to reach its full amplitude before the input is gone. This results in the lower amplitude curve B. Here the 3-db points are much wider with respect to, frequency giving the impression that IF bandwidth is, wider than it actually is and overall sensitivity has been reduced. If frequency modulation is present on the local oscil~ lator (other than the Linear sweep tuning) the effect is, equivalent to having a signal input that is frequency- modulated. Asa result, the analyzer will present an FM spectrum rather than a single response, seriously Umiting the resolving ability. The effect of local os~ elllator FM on a CW signal is shown in Figure 14 Let's consider an FM spectrum for a moment. As a general rule, the approximate bandwidth of an FM Spectrum is that of the peak-to-peak deviation of the signal provided the frequency deviation is much larger than the modulation rate. If, then, the peak-to-peak deviation of the local osesilator is lange compared to the IF bandwidth in the analyzer, two CW signals must bbe separated in frequency by an amount greater than the deviation in order to be displayed as separate sig- nals.2 The local oscillator stabilizing circuit in the ® spectrum analyzer, mentioned earlier, enables full uullization of the narrowest IF bandwidth setting whieh is 1 ke. Thus, resolving power is about 3 ke. T See Appendix B. 2 pia.Page 12 Eo iS ae ape: sar ‘ona FREQUENCY om Figure 18. Curve A shows actual bandpass for gaussian IF amplifier. Curve B represents ‘output when local asoillator frequency is swept too fast for IF amplifier to fully respond. SENSITIVITY AND AMPLITUDE RESPONSE Sensitivity is a measure of the analyzer's ability to detect small signals. There are several common ways to measure sensitivity in receivers depending pri- ‘marily on the input stage. In communications, the in~ put stage 1s usually an RF amplifier and the method of Yerifying sensitivity is to measure signal -to -noise ratio under specific conditions. In radar the Input is usually direct to a diode mixer whose nolse behavior is uncertain making signal-to-noise ratio difficult to accurately determine. Here, required input power to the receiver for minimum discernible signal out of the video amplifier is commonly referredtoas sensitivity. This, however, is too arbitrary for consistent meas- urements. A’ more specific measure would be in terms defined by the system noise figure. In any case, Inherent noise is the ultimate limitation to receiver sensitivity and since the spectrum analyzer is a re- ceiver, let's see how sensitivity may be defined with relation to noise. ‘The mixer and IF amplifier comprise a system of stages in cascade, each contributing its own noise. We can therefore express the overall noise figure of the analyzer in db as follows! Fy + Fm + Le (Fig-) @ where F, = overall noise figure Fyy = noise figure of mixer Le = conversion loss of mixer* noise figure of IF amplifier. * Note: Ly o7 ain Since Fy, 1s not measurable, a noise ratio Ny is gen- erally specified for diode mixers and may be substi- tuted in the above formula. With appropriate factor ing the equation becomes: Fy = Le (Np + Fig- 1) T gee Terman & Pettit "Electronic Measurements" 2nd edition, p. 861, Noise Figure of Systems in Cascade, McGraw-Hill. Appl. Note 63 Figure 14. Appearance of a CW spectrum when analyzer's local oscillator has excessive FM. Now, considering the overall noise figure F,, we ean show the equivalent noise power to the input of the analyzer as Pi, = FL KTB ® Boltzmans' constant of 1.37 x 107% joule/? Kelvin ‘T = absolute temperature in * Kelvin B = equivalent noise bandwidth of the IF amplifier. where K Pi, a8 defined in equation (9) Is generally accepted as, receiver sensitivity. A signal of this amplitude would produce an output signal-to-noise ratio of unity if there were no deleterious effects such as local osetl~ lator FM or excess sweep rate. In specifying analyzer sensitivity, it ts essential that the associated IF bandwidth is also known as ean be seen from the formula. Only under these conditions can one prediet the signal-to-noise ratiofor pulse sig nals or determine an upper limit on sweep rate and the tolerable local oseillater FM. The equivalent noise bandwidth is defined by the equation on? pe jah a ‘where Hy) = frequency response characteristic of IF amplifier H, = center frequency gain of IF amplifier. As a rule of thumb, the equivalent noise bandwidth is about the same as the 3-db bandwidth for a gaussian IF amplifier, and the value of KT at 290° Kelvin (roomAppl. Note 63 temperature) is -114 dbm/Me. Thus the above rela~ tionships become quite simple for calculation of ana~ lyzer sensitivity, viz +0106 Bae sent (ba) =, 2840 Using typical values for the analyzer we can caleu~ late sensitivity for any LF bandwidth setting and sys~ tem noise figure, For fundamental mixing the @ ana- lyzer's noise figure is typleally 28 db. I we choose the 10 ke IF bandwidth, the sensitivity of the analyzer is ol Me 29 - 114 + 10 tog OF Me 9 - 114 - 20 = -108 dbm. Amplitude response of a spectrum analyzer is closely related to sensitivity. I'the system isto present equal amplitude deflections for signals of equal amplitude independent of frequency, it is vital that the conver- sion loss L, be independent of frequency. Earlier mention was made of the use of harmonic mixing to extend the frequency range of a system. In this case, we define a quantity Ley 2s the conversion power loss of the mixer when mixing with the nth harmonle of the local oscillator. It has deen shown that if the oscillator voliage is rel- atively large compared to the input voltage, the con- ductance of a diode mixer will be time varying at the local oselllator rate. During the period when the local oscillator biases the mixer aff, an average mixer con~ ductance level exists from the forward bias effects of stored charges in distributed circuit capacitances. Conversion loss 1, 18 dependent on the ratio of con- version conductances to this average conductance. 1 Un- less special means are taken to control Ley (such as extremely small conduetion times or conduction angle adjustment), the possibility of large variations in Ley is very high. As anempirical rule, if conduction angle 4s optimized, Ug, ean be expected to fal of ata rate of about 1/n2 DYNAMIC RANGE ‘There are commonly two measures of a spectrum analyzer’s dynamic range, The first is the ratio of largest to smallest signals which canbe simultaneously displayed on the analyzer screen, This is usually extended by providing the IF amplifier with anoption~ fal logarithmic response. In the HP 851/8551 the dynamie display range is 60 db. Distortion due to saturation in the IF amplifier stages can also limit dynamie range if it occurs prior to overloading the nuxer with large inputs, A well designed analyzer ‘ill have sufficient IF gain control so the IF amplifier will not be the Lmiting factor in dynamic range. In the HP analyzer any on-scale display will not saturate ‘the IF amplifier and signal levels of 60 db difference may be simultaneously observed without distortion. 4 B.W. Herold, "Frequency Mixing in Diodes," Proc. IRE Vol. $1, October 1943. Page 13, The second measure of dynamic range is the ratio of the largest signal that can he applied at the input with- out serious amplitude distortion, to the smallest sig- nal that ean be detected (sensitivity), Distortionoccurs when the applied signal is large enough to cause the mixer conductance to vary with the applied signal. This happens when the peak signal voltage to the mixer approaches the magnitude of the local oscillator drive voltage. It is therefore desirable to limit large input signals to the mixer with a suitable input attenuator. An input attenuator of 60 db will inerease the saturation to noise dynamic range by 60 db. This is the input range. ‘The dynamic range is greatly reduced when analyzing pulse signals since the mixer saturates at peak levels Dut the displayed response is attenuated by the slow rise time of the IF amplifier. ‘The exact vaiue of this, attenuation may be calculated provided the IF band- ‘width is knowa. See “Amplitude of Spectrum Display’” in Table 1. This again emphasizes the importance of knowing the IFbandwidth at which the sensitivity of an analyzer is specified (or preferably that the overall system noise figure be known. DETECTOR AND VIDEO AMPLIFIER For most purposes, it is proferable to use an enve~ lope (peak) detector at the output of the IF amplifier. In some of the earlier spectrum analyzers, square- law detectors were used which resulted in a response that presented a power spectrum display; however, detector square law characteristics cannot be relied uponfor accurate spectrum measurements. It is gen- erally better to use a linear onvelope detector and shape the gain characteristics of the IF amplifier to obtain a response other than linear when required. This is the approach taken inthe design of the ® spec- trum analyzer. Linear, logarithmic (éb), or square law (power) displays may be selected by'a "vertical display" control on the front panel. ‘The basic duty of the video amplifier Is to drive the CRT without excessive loss in amplitude because of its bandwidth. It should be the IF bandwidth and not the video bandwidth that determines the resolving capa~ bilities of the analyzer. Since the fastest transient available from the IF amplifier is its response to an impulse, the video amplifier bandwidth Is usually of the same order of mangitude as the IF bandwidth. Other than this, the characteristics of the video am- piifier are relatively unimportant. GENERAL CONSIDERATIONS ‘To perform meaningful measurements with a spec- trum analyzer, it is necessary to be ableto accurately determine amplitudes and frequencies. For amplitude comparisons, an accurate IF attemator is generally uused since it can be calibrated very closely ata single frequency, that of the IF. The @ analyzer employs a 0 to 80 db step attenuator calibrated in terms of IF gain. Accuracy is 0.5 db on the 10 db/step section, and 20.1 db for the 1 db/step. A vernler is also in- eluded for continuous adjustment over a 1-db range. ‘To avold distortion in the mixer at high input signal levels, an RF attenuator is also desirable as prPage 14 viously explained. This attenuator need not be particularly accurate but must be flat to maintain the Overallamplitude response of the system. Waveguide- deyond-cutolf type attenuators have usually been em— ployed in analyzers for simplicity and wide frequency range, The problem with this is that insertion loss is typically 20 db or more requiring an RF patch panel to bypass the input attenuator for weak signal Inputs. Also, this type attenuator is not flat with frequency. A recent advancement in film attenuator design by Hewlett-Packard allows both low insertion loss and flat frequency response which ts ideal for this work. The @ analyzer uses a0 to 60 db RF step attenuator of this type that has zoro insertion loss at 10 Me in- creasing to only 2 db at 10 Ge. A calibrated input frequency dial and accurate sweep Iinearity is essential for properly determining energy Appl. Note 63 distribution. Many analyzers employ built-in exystal tnarker generators which prodce responses at aceu~ Tate frequency intervals. ‘This technique makes fe quency difference measurements more accurate by ‘Siiminatingthe analyzer's sweep Linearity error. ‘The Iimitation to this approach Is that the actual frequen~ ties of spectral components eannotbe determined any tore accurately than the basie tning accuracy of the inalyzer. The @ analyzer was designed to provide good frequency accuracy {1') and specteum width ln arity sufficient for most spectram mensirements. When increased accuracy is required, an external marker generator may be fed into the analyzer's in put. By doing this, both frequency difference and actual frequency measurement accuracy is increased. ‘The use of an external marker generator also allows tnore flexibility for frequency measirement of Pas sive devices as illustrated in Section 1V.Appl. Note 63 Page 15 SECTION IV APPLICATIONS Jn Section 11, we learned how to interpret various dis~ plays without concern as to their source. Now we ‘shall apply this Information to Hllustrate the value of ‘a wideband calibrated analyzer. The examples shown in this section are only a few chosen from many, with ‘emphasis on newer applications made possible by the HP 851/8551. Because of thelr continued impor- tance, some of the older applications of spectram ana~ lyzers are also included. PULSE RADAR PERFORMANCE CHECKS One of the earliest applications of a spectrum ana- Iyzer continues to be of major importance. This in- volves measuring pulse radar's magnetron output to ensure stable oscillation,free of moding and spurious signals. The HP 851/8551 is ideal for radar meas- urements for several reasons, one being its wide range, Any radar frequency may be received with this one instrument and broad spectrum widths allow complete displays of spectrum signatures. Wide image separation (4 Ge) and comparatively few spur- ious responses keep the display uncluttered for aceu~ rate presentation of radar spectra The spectrum width may be quickly reduced for detailed examination fof the main or side lobe structure, if desired. A. Magnetron Pulse Operation. Connect the analyzer RF input to a sample of the radar transmitter power through a directional coupler of at least 30 dl coupling. Most radars have built ~in couplers providing a test output of a milliwatt or so. Figure 18 shows this setup. With the radar high volt~ age on, tune the analyzer to the transmitter frequency ‘which will center the main lobe of the radar's pulse spectrum on the analyzer screen. Check for over~ Toading of the analyzer's mixer by increasing the RF input attenuator of the analyzer by 10 db. The ampli- tude of the lobes should decrease by 106) 16.07 + r(n-6) Order of Carrier Zere Figure 22 is a pictorial diagram of an FM signal gen- erator belng checked for accurate frequency deviation ‘The generator is first placed in CW operation by dis— connecting the ( 204B from the external FM jack of the generator. The analyzer and generator are tuned tothe desired carrier frequency producing a single CW response in the center of the analyzer display. ‘The frequency of the analyzer is then stabilized and the attenuator and IF gain controls adjusted for a good display. TheSPECTRUM WIDTH control is switched to the 10 KC/CM position and its vernier used to farther reduce spectrum width to about 2KC/CM The IF BANDWIDTH is switched to 1 KC for good resolu of the carrier and sidebands. Now the 204B Oscillator is connected to the FM gen- erator as shown In Figure 22 and its frequency accu- rately set to 6250 cps as indicated by the electronic counter. This frequency when multiplied by a modu- lation index of 2.40 results ina carrier deviation of 15 ke whieh is full seale on one range of the Boonton 292-1" deviation meter. ‘Then, with the 2048 amplitude control, the audio mod~ lating signal voltage ts slowly increased from zero until the amplitude of the carrier response onthe ana~ lyzer display first goes to zero. As the carrier am~ plitude is decreasing, sidebands begin appearing at intervals of 6250eps above and below the carrier, and when the carrier amplitude is zero all of the energy i contained in the sidebands. At this point the gen- erator deviation is 15 ke and the deviation meter is checked for full scale accuracy. If the 204B ampli tude is slowly increased further the carrier amplitude will increase again to some maximum and then begin Gecreasing until it goes to zero amplitude for the second time. At this point the modulation index is 5.52 as indicated in Table 2 for a second order car rier zero. Note that any deviation, within limits of the generator of course, may be set up by choosing the correct combinations’ of modulation index and mod~ ulation frequency. This technique is known as the Crosby Zero Method for measuring frequency deviation. ‘The wide dynamic range of the HP 851/8551 makes it possible to accurately determine when the carrier is zero because the IF gain may be increased, as carrier zero Is approached, for maximum sensitivity without the lange adjacent sidebands saturating the IF. Accurate carrier zero Is essential for correct devia- tion measurements, It should also be stated that the analyzer must have a resolution at least three times better than the modulation frequency to be used in order to distinguish between the carrier and first sideband responses. Resolving capability of the ana~ lyzer, you recall, is largely determined by the TFAppl. Note 68 SSE 0.5 ns/cm TIME FUNCTION bandwidth. The narrowest bandwidth on the { ana- lyzer is 1 ke allowing modulation frequencies down to about 3 ke to be used with this method of frequency deviation measurement. Modulating frequencies below the analyzer's resolution may be used, and approxi mate deviations inferred by the width of the spectrum produced by the modulating signal Residual FM of signal generators may alsobe checked provided the peak-to-peak deviation is 10 ke or greater. This is done by switching the SYNC control fon the analyzer to INT and setting the SWEEP TIME slightly different than the power line frequency or submultiple (3 msec/em for 60-cps Line). With the generator set for CW operation and the analyzer tuned to the operating frequency, residual PM will be indi- ‘eated bya slow periodic movement of the CW response back and forth on the analyzer display. This movement will be at the differential rate of the line frequeney and the analyzer's sweep time. The peak-to-peak de- vlation of the generator's residual FM is then meas~ tured by noting the maximum horizontal excursion of Apassiracin al masa Ors 2048 =i Pum oa’ Leiner a Figure 22. FM Deviation Measurement Using & Spectrum Analyzer to Monitor Carrier Zero Page 21 Mt 0 166 266 FREQUENCY FUNCTION Figure 21, Spectrum of a Boif Step-Recovery Diode the CW response, and reading the frequency from the calibrated SPECTRUM WIDTH control RADIO FREQUENCY INTERFERENCE TESTING A. Background Some forty years agothere were only 30 standard AM broadcast stations transmitting in the United States. In 1925 only eight per cent of all U.8. homes had radio receivers, but by 1940 this figure had risen to 17 per cent, representing over 45 million sets. As the use of the’ electromagnetic spectrum grew more intense with the addition of more communications, TV, Facsimile, Radar, and navigation aids, the prevention of radio frequency interference (RFI) became ex- tremely important Aside from the obvious importance of careful channel assignments and frequency control of transmitters, there is the need for preventing RET from other sources. Spurious signals, harmonics, RF leakage and transients are some forms of interference that ‘ean be caused by an electronie or electrical device. ‘The U.S. Government establishes standard methods for determining RFI emanating from a device and documents them in the form of Military Standards Careful testing and control of RFI is required on all electronic equipment to be supplied for Goverament installations. Interference may occur anywhere in the spectrum and the problem of checking for its presence and intensity is manifold. ‘The common approach has been to tune a number of receivers, with various antennas or probes, through their frequency ranges searching for undesized signals indicated on a calibrated signal ‘strength meter, The process is long and tedious with bband-switching, slideback detector adjustments, me- chanical cranking and the constant concern that you may have "blinked" when you went through a signal missing its meter deflection completely. Some im- provement has been made, however, by using a motor Grive and mechanical linkage to alleviate the drudgeryBes Appl. Note 63 Gp oasen dpeassa INPUT FREQUENCY GC (enact use pessoa SaNopass ALTER \ \ $ dprcoe 2248 8 ets LOCAL OSCILLATOR GC 7 Figure 23. HP 861/8551_ tuning curves allow selection of RF filters and local oseillator sweeps for eliminating spurious responses. Shaded area represents spurious-tree display range of analyzer using an 84304 filter and 3-4 Ge local oscillator sweep.Appl. Note 63 of manual tuning, and X-¥ recording of the detector output has solved the "blinking" problem. But what about the case of transients or closely spaced signals swept through too fast for the X-Y recorder to re- spond to their true amplitudes? Many times the clos ing of relay contacts or a switch will produce serious transients and continuous monitoring of the entire spectrum of interest is the only way it ean bedetected ‘The HP 851/8551 Spectrum Analyzer with its rapid electronic spectram sweep over wide ranges at high sensitivities has an important role in RFT testing and spectrum surveillance work. This single instrument presents continuous visual coverage of the spectrum from 10 megaeycles to 12.4 gigacycles for rapid loca~ tion of interference, Two external waveguide mixers extend the frequency range to 40 Ge. B, Preselection Filters. RFI can be of two general types: narrow band, such as transmitter harmonies which are sharply de- fined In frequency, and broadband, such as electrical noise produced by'electrie motor brushes, switches, ‘and high-voltage discharge. Spurious responses in a spectrum analyzer used for RFT testing must be con- trolled to prevent their being mistaken for actual RFT signals. ‘This is particularly true when measuring proadband noise where it 1s virtually impossible to separate the spurious signals from true inputs. Recall from the discussion on spurious signals in Sec lon Hi that the amplitude of a spurious response is not Iinearly proportional tothe input signal causing i ‘This makes it easy to identify a spurious response if there are relatively few signalson the display. Simply note the amplitude of a signal on the logarithmic dis~ play of the ( analyzer and switch in 10 db more Input, attenuation. Ithe amplitude of the response decreases by 10 db (1 em on the log display) the signal is gen- nine, Ifthe amplitude decreases by more than 10 db ‘the response is spurious. It 4s best to eliminate input signals that can cause spurious responses by using preselection filters at the analyzer input when making specific RFI measure~ ments, By careful cholee of filters and local oscil- lator sweep widths, spurious responses can be com- pletely eliminated, Generally total elimination im- poses heavy restrictions on spectrum width and requires more filters than is economically feasible. ‘The & 360B Low-Pass Filter (1200 Mc cutoff) and 8430 series of interdigital bandpass filters comprise a good working set of filters for the @ analyzer, ‘allowing reasonable spectrum width and virtually clean displays. ‘The Information given in Figure 23 enables you to choosea preselection filter for spurious-iree displays in any frequeney range of interest. Figure 28 shows, for example, haw a 1- to 2-Ge filter can provide a completely spurious-free display on the @ Spectrum Analyzer. ‘The & 840A Filter passes only those fre~ queneles from 1 to 2Ge, rejecting all others. The only spurtous-producing signal crossing this band on the graph is the (2, 2-) curve. We have eliminated all others by preventing their reaching the mixer withthe filter. Now we see that we can cover the octave range from 1 to 2.Ge by sweeping the local oscillator from Page 23 3.0 to 4.0 Ge with primary signal reception along the heavy n= 1- tuning curve. By setting the local oscil~ lator sweep for these limits (main tuning at 3.5Ge and spectrum width at 100 Mc/em)a houndary is setonthe graph which excludes all spurious responses. The Sputious-free area is shown by the shaded rectangle in Figure 23. The bandpass filter requirements and local oscillator sweep limits for any region of the spectrum may be quickly determined in this manner. Table 3 inchides the proper @ filters to.use for the RFI tests that follow. The -hp- 8441 Proselector is an electroniely tunable filter designed for use with the hp- €51/8551, Information on this unit is available in ANOSB, C. Measuring Absolute RF Levels. ‘To make absolute measurements of RF levels with any instrument, iis necessary for its input to be cali- brated with known standards. Conventional RFI meters cover a comparatively narrow band of fre- quencies, yet exhibit large variations in gain versus input frequency. This greatly complicates absolute amplitude calibration. Some RFI meters utilize ex- tensive calibration charts for setting gain as frequency 4s changed so that levels may be read directly from a meter indication. Other RFI meters are used a5 (ransfer standards employing a calibrated impulse generator. In this case, the meter indication serves only as a reference and the actual RF voltage is read from the output dial of the impulse generator. This eliminates the requirement for charts but does not solve the problem of continual calibration for small frequency changes. ‘Tho @ Spectrum Analyzer has a remarkably flat input response (including the input attenuator) across any range plus accurate relative IF gain calibration and known IF bandwidth These characteristics greatly simplify absolute RF measurements using a form of the transfer technique descrined for impulse calibration. Since there have hbeen some doubts raised about the calibration acca~ racy of impulse generators, we will show the use of signal generators at CW in this application. Ifno preselection filters were used at the analyzer i put, only four calibratingfrequencies would need to be used for amplitude reference: one for each local os- cillator harmonic used in tuning the analyzer from 10 Me to 10Ge, Since we are to inelude preselection filters in the absolute measurement of RFI, we must be able to provide signal generator outputs within the passband of each filter; thus, variations in the filter's response will not introduce error into the measure~ ‘ment. Also, transmission losses can be taken into account with this technique. ‘One further consideration must be made about the analyzer at this point andthat 1s pulse desensitization When measuring pulsed signals and comparing Wel? spectrum amplitudes with those for CW signals, the shape and bandwidth of the IFamplifier inthe analyzer must be known. Because the analyzer must be selec tive for resolving spectra, its IF bandwidth does not admit all of the frequency ‘components contained in a pulse at one time. ‘Therefore, the peak amplitude of the main lobe of a pulse's spectrum represents only a portion of the total energy contained in the pulse. The main lobe of a pulse's spectral display is typicallyAppl. Note 63, Page 24 juoqeagnbs £0 “@{-419 sorueysaoNoaTE ‘I-16hE6 TOPO KEPPOIS ey “yuatwaqnbs 10 “y]-q'T9 “09 SopuByDOMONDAT “T-9GFEG 19PON “OD OIEY WRIONY MPPs + suopoas sode) pur sox apinfaxem pueE U-H-a VLISIT @ 9en, OE oprniioseyg PrEE-X VIZSIT H a8 in t (sqowses 05-2 1) 95 OF 9% (squowes 99-2 m1) 90 92 - 8 29 81- oF 95 91- FL 99 PL er 2a ep 0-8 208-9 2097 2DF-e 9D%-1 oD r-0 m1 008 - 0 aw se - wo /oW 002 9 /0ny 00g w9/o¥n 008 9/9 008 19 /aNt 002 wo /ow 008 10 /9W 008 9/OW 00% ‘19 /9W 002 (wo/oW 00g 9 /2¥% 00T 9/29 OOF w9/oW 08, moon § (sdays 29-z 1) 99 6f- Le 90 sa ‘ea ‘Te ‘6r 2D LE 25 &t (oun 99 81-8) 99 &r (ones 99 81-8) 99 11 206 “i. api (ower 95 6 2D 0-2) 70 20 9°r wi 008 2W 0st ow 08 uou sseg pura vbEre @ sseq piva veers @) seed PULA VERE Souq Pu VIEFE ‘ssc Pima Vote @ ssea-#07 BORE @ SStd-MOT BOE sseq-morl gone (eve w) Wao apInianea (preg 30) xO APES AEAA (oueg a) xo apinaavay (reg x) xO apInFAE A ayumypae30"y [e9]U03) Ow $e 03 pour atodicr wn8u9q “9972 OPH Z/L ‘PO! ze z 25 00h - 9°92 20 $92 - 0°8t 95 0'8T- FET power- oF 2D 0r ‘4x o00r ~ 002 OW 002 - sz OW $@ - OF poserdsig wnaoadg wapEM wnaypeds Aouanbaag sozsreuy souns vonoateseaa way 98uvy Aouanbor paanseay, ‘Syuowo.msvay Ia poveIpeY 10} Sues ZORATEEY UTES H pu SEMIOTEY -p ATPL,Appl. Note 63 i | | sia 95st 05th ‘SPECTRUM O ° “h cra) ypu S| wen 4 | TOKFo =p | $1 weren | Figure 24, Suggested setup for measuring radiated RFT with the @ Spectrum Analyzer anc 20 db lower than the response to a CW signal of equal peak value to that of the pulse. This is termed pulse desensitization andis given in terms of db loss by the equation a = 20log 7 BW, attemzation of pulse spectrum main lobe relative to a CW signal of equal strength Impulse bandwidth for the particular 851 being used* ‘Measured pulse width in sec Af = IF bandwidth selected on the analyzer. ‘An example of how to use this information is shown in the following example for an RFI measurement. where @ * May be determined by measuring a withan @ 86144, Signal Generator and 87144 Modulator as follows: 2) Set up a known CW reference level and note ite spectrum amplitude. 3) Pulse modulate the CW signal with the 87144 PIN lode modulator. 3) ‘Note the attenuation of the pulse spectrum main lobe relative to the CW reference amplitude 4) “calenate BW, by tranapoeing, the" desensitization formula to i a5) 5 20 using the value of attenuation from step 3 in db, pulse ‘width selected on the 8T14A, and TF bandwidth of the hualyzer. This value should only need to be deter mined once for each bandwidth unless changes are made in the TF alignment + tog” [ id appropriate antennas. D. How to Measure RFI with the Spectrum Analyzer Hereis how you may use the analyzer in a typ- ical RFT measurement. Set upthe equipment as shown in Figure 24, selecting a suitable test environment of low ambient RF signals. A screen or shielded room fg ideal if one is available or ehoose a location where the ambient noise is at least 6 db below the maximum allowable RFI from the device in test. Follow the manufacturer's recommended turn-on procedure for all equipment to be usedin the test. Tune the @ Spec trum Analyzer to 1.5 Ge and set SPECTRUM WIDTH to 100 MC/CMIor this example This sets up a spec~ tral sweep from 1 to 2 Ge. Initially set IF BAND- WIDTH to 100 KC, IF GAIN (DB) to 80, VERT DIS- PLAY to LOG and turn the input attenuator to the "0" db position for maximum sensitivity of the analyzer. ‘Table 3 gives the frequency settings of the analyzer along with the recommended antennas and filters for ‘each band throughout the analyzer's range. From this table, determine the appropriate antenna and filter for 1-2Ge. Connect the antenna to the analyzer input through a low loss coax cable andthe preselection f- ter (fp 84304) Now move the antenna slowly around the unit intest at ‘a distance of about one meter, watching the analyzer display for a signal indication. If signals appear, maximize heir amplitude with antenna positioning, Keeping it one meter away from the tested unit. Re. duce the analyzer IF gain and increase input attenu- ation as required for an on-screen display of the in terference signals. Check for overload of the mixerPage 26 following the instructions given previously in "Pulse Radar Performance Checks", Paragraph A. Note the vertical deflection and frequency of the sig- nals onthe analyzer display and adjust the IF gain un~ til the maximum amplitude of the RET is at some con- venient reference on the display. Now disconnect the coax cable from the antenna and connect it to the OW output of a calibrated signal generator in the 1-2 Ge range ( 86144). Tune the generator's frequency to approximately that of the largest RFI signal and ad- just the output attenuator until the CW response equals the reference amplitude previously set. At this point, the attenuator indicates directly the absolute amplitude in dom of the largest CW RFI signal. ‘The amplitudes of weaker RFI signals appearing on the display may how be read directly from the gratieule. When in log display, 1.em represents 10 db, enabling direct read- out from the reference point in dbm, ‘The Linear or square display may be calibrated and used, if pre- ferred, to readout In terms of voltage or power. If pulsed RFI signals were observed with the antenna connected to the analyzer, their peak amplitudes may be determined making use of the desensitization for~ mula as follows: 1) Caleulate pulse width from measured lobe width. 2) Note indicated main lobe amplitude in dom. 3) Calculate peak pulse power in dom from the formula Py = Py +20 log Krat ok where Py = level noted in step 2. Now that the absolute calibration is known, you may record the interference measured. The best technique isto photographthe display with an oscilloscope cam~ era. If the shutter of the camera is held gpen long enough for about 100 3-msec/em sweeps of the ana~ lyzer trace, the photo will include transients, inter- mittent noise, and random spikes that would be impossible to see In an X~Y¥ plot or manual tuning ‘operation. E, Spectrum Analysis Photographs. Many people have their own "best way” for making photographs of CRT displays but the following sug- Bestions may help when working with the analyzer and @ 196B camera, 1) Remove the blue filter from the CRT face when photographing spectra. 2) With trace intensity turned off, expose the CRT graticule at full intensity of the ultraviolet light source inthe 196B. Using 3000-speed film, use an aperture ‘opening of about f:8 and a shulter speed of 1/5 sec. 3) Tum olithe U.V. lignt source and set the exposure time to bulb. Adjust the trace intensity for normal viewing and analyzer SWEEP TIME for 3 MSEC/CM. Leave the aperture opening at f:8 and with the RFI displayed, open the shutter for about 3 seconds. 4) Develop the print and log the absolute scale calt- bration and frequency range swept on the back. Appl. Note 63 SPECTRUM SIGNATURE AND SPECTRUM SURVEILLANCE ‘An extension of RFI measurement is the vast spectrum ‘signature work performed for the government by var~ fous companies and institutions. Spectrum signature refers tothe energy distribution, both desired and un- desired, emanating from a device. For example, It ‘would include the mainand side lobes of a pulsedtrans~ mitter or signal generator plus any undesired outputs. ‘This information is useful in statistical prediction of interference caused by the outputs of various devices. Previous spectrum analyzers could not look at eom= plete spectrum signatures because of their restricted Sweep width and narrow image separation. Figure 25. Spectrum Siguature of a 0.5 jisec Pulse Figure 25 is a photo of the spectrum signature of a 0.5-mlerosecond pulse with a rise time of less than 20 nanoseconds as seen on the § analyzer display. Here the spectrum width of the analyzer is set to 30 Me/em giving 2 300-Me wide display of the main and side lobes. ‘The logarithmic vertical display has been selected so the side lobes down to 60 db below the main lobe are visible. Notice that the spectrum of the pulse is over 200 Me wide 60 db down: ‘The HP 51/8551 is invaluable for spectrum sur- veillance worlealso, where the relative amplitudes and frequencies of all electromagnetic radiation ata loca~ ton are desired. An application of this would be in a imissile or spacecraft Imunch site or tracking station where multiple command, communications, and radar tracking signals are transmitted at various times, often simultaneously. ‘The far-ranging sidebands of radar transmitters, in~ termodulation products of telemetry and communica~ tion transmitters, or spurious signals can blot out reception of valuable data. It can also be responsible for accidental triggering of control Links for detona~ tion devices in missiles and retro - rocket firing in spacecraft. To spot such signals and eliminate them before they cause trouble the & Spectrum Analyzer ‘ean be set up with appropriate antennas in the launch {ng vieinity and used as a spectrum monitor. Figure 26 depicts such an installation where several antennas for the analyzer are selectable by theAppl. Note 63 \ Re TELEMETRY ‘GROUND aN ‘COMMUNICATIONS RADAR ‘CENTER Page 27 ANTENNA SELECTOR AND ROTOR SPECTRUM ANALYZER MONITOR Figure 26 Broadband Spectrum Surveillance using @ Spectrum Analyzer Operator monitors electromagnetic radiation in vieinity of launch area to ensure clear data channels and tracking frequencies. ‘operator througha coaxial switch. Withthe broadband, omni-directional antenna connected to the analyzer in- put the operator scans wide segments of the spectrum ‘watching for interfering signals. If any are seen, he can quickly narrow down the analyzer sweep for a de- tailed examination of the interference, determining I it is of pulsed or CW nature. The operator can then ‘select a directional antenna whose positioning is eon- trollable from the monitoring point and determine the direction from which the undesired signal is ema- nating. For additional sensitivity, RF preamplifiers may be connected ahead of the alalyzer input. By knowing the nature of the interfering signal, its fre~ quency and direction, the monitoring station ean dis- patch technicians to determine and eliminate the problem if possible. If the problem cannot be cleared ‘the monitor could advise the use of other frequencies away from the interference and proceed with the launch, ANTENNA PATTERN MEASUREMENTS ‘Aninteresting application of the Spectrum Analyzer ig tts use notas a wide-sweeping device but rather as a CW single frequency receiver for antenna field pat- tern measurement. The general qualifications of a pattern receiver are high sensitivity with low noise, good frequency stability, wide dynamic range and good shielding. As we have’ seen earlier the © analyzer meets these requirements well and has the added fea~ tures of selectable vertical displays for linear, log (av), or square (power) patterns plus X-¥ recorder ‘output signals. The logarithmic display is by far the most useful in this application since i¢ allows simul~ taneous on-scale viewing of the main lobe of an an— tenna’s field pattern along with side lobes that are dawn 60 ab. ‘To make rectilinear pattern plots with the spectrumPage 28 orate PLATFOR \ one woron QD Appl. Note 63 SMALL FED ANTENNA es ssr008¥ af sean swreu Figure 21. analyzer connect the output of the antenna in test to the analyzer input as shown in Figure 27. Connect the X-¥ recorder output jack of the analyzer display unit to the Moseley 2-D or 135 seriesX-Y recorder inputs. ‘The illustration shows the antenna in test mounted on a rotary platform which is common at antenna range installations. This platform allows the antenna In test to be rotated through 360 degrees in azimuth while coupling the antenna coaxial transmission line or waveguide through a rotating joint to the analyzer input. {A smaller directional transmitting antenna capable of being aimed for maximum signal at the receiving an~ tenna is located at a range or distance R, for field measurement, away. This range should satisfy the equation R 2 2p2/ for minimum error in received power due to -muiual coupling effects of the two antennas, where R D X range (distance between antennas) largest aperture dimension of antenna free space wavelength of test frequency (all dimensions in identical units, Le, meters, feet, etc) Ideally, the two antennas would be mounted on towers. above flat unobstructed terrain to minimize multipath reception due to ground waves and reflections from buildings and trees. Set up the equipment as follows: At the Transmitter Location: 1) Set CW frequency and power output of transmitter as desired. 2) Stabilize frequency with phase lock or AFC system. 3) Aim transmitting antenna toward receiving antenna, fos | AL mame i, Spectrum Analyzer Connected for Rectilinear Antenna Pattern Plots At the Receiving Location: 41) Tune spectrum analyzer to approximate transmitter frequency with SPECTRUM WIDTH at 1 MC/CM, SYNC at INT, and IF BANDWIDTH at AUTO-SELECT. 2) Rotate test antenna for a maximum CW response on CRT. Vertically position test antonna for desired langle with respect to transmitting antenna. 3) Stabilize analyzer local oscillator per Operating and Service Manual. 4) Select log display on analyzer so vertical scale is now calibrated in db. Adjust IF gain and input attenu- ator for full seale indication of test signal 5) Reduce spectrum width to zero koeping the re- sponse centered on the display with fine tuning. The analyzer is now operating at only the test signal fre- quency. Refine analyzer frequency slightly to peak vertical response and touch up IF gain so response is, just T em high 6) Adjust SWEEP TIME to some value slightly longer than the total time required to rotate the test antenna through 360 degrees of azimuth. For example, to X-Y record pattern, a relatively slow sweep time is required; therefore set SWEEP TIME to 3 SEC/CM and rotate antenna at 2-1/3 rpm. This results in a hori- zontal display of 42degrees rotation per centimeter of display, 1) Set X-Yreconder gain and position controls to co- incide with analyzer horizontal sweep and vertical de- flection limits. 8) Now rotate antenna 180° away from the maximum reception point just set up. This will position the an- tenna in the correct relationship with the start of theAppl. Note 63 sweep in the analyzer to cause the main lobe of the pattern to appear inthe center of the display when the plot is made. 9) Set analyzer SYNCto single sweep and turn onX-¥ recorder servo and pen down switches 10)Simultaneously push SINGLE SWEEP button on ana- lyzer and start antenna scan drive motor. The trace on the CRT and the X-Y recorder will now move hori- zontally at the $-sec/em rate set up with analyzer ‘sweep time, and be deflected vertically by strength of the received signalas the antenna rotates through 360 degrees, ‘Thus a rectilinear plot of the antema field pattern is obtained. The analyzer sweep may also be triggered from an external +3 to +16 volts which ean bbe synchronized with the antenna scan motor switch eliminating the need to manually operate the single ‘sweep button. Exact synchronism is not particularly important since the sweep time of the analyzer is longer than one rotational cycle of the antenna in test. This allows the operator to locate the identical con- tours on the left and right edges of the plot denoting the 360° boundaries. Figure 28. Antenna pattern obtained from Spectrum Analyzer. Vertical and horizontal out- ‘pats on rear panel also allow X-Y recording of display. Figure 28 is a time - exposed photo of an antenna pat~ tern plot on the @ analyzer (the alternate method of trace recording). Notice the deep minima in the plot, some extending nearly 60db below the main lobe. This illustrates the Importance of the logarithmic display and 60 db dynamic range, Previous spectrum ana- Iyzers commonly have had only 40-dbdynamie range which would obscure the actual minima of such a plot and therefore not be a true representation of the an- tenna characteristics. FREQUENCY MEASUREMENTS WITH THE SPECTRUM ANALYZER ‘The basic frequency dial accuracy of the HP 8551 is 1G whieh is adequaiefor most applications of the ana~ lyzer. However, there are occasions when much higher accuracy ‘will enhance the instrument's use- fulness. Fast, accurate frequency measurements are Page 29 possible using the @ Spectrum Analyzer and a good reference marker generator. The marker generator provides a number of accurately known frequencies which are displayed on the spectrum analyzer. Afre- quency to be measured is then simultaneously fed to the analyzer input and comparedto the reference fre- quency markers. When the unknown frequency coin- cides with one of the markers on the display both are at the same frequeney and the unknown signal is de- termined. If the markers are at intervals of 1 to 10 ‘Me, accurate interpolation of unknown frequencies may be made using the CRT graticule. This technique may be varied slightly to measure passive frequency meters also. Astep-recovery diode impulse generator whose repe- tition rate is determined by an accurate crystal-con- trolled oscillator makes an ideal reference mark generator for this application. A generator of this type which produces useful harmonics up to about 6 Ge, is ‘the HP 406A. Three repetition rates controlled by a 01% crystal give accurate reference markers at 1, 10, or 100 Me intervals. This allows quick first approximations with a few widely-spaced markers at 100 Me, or high resolution measurements using the 1-Me markers, Impulse generators using mechanical Switching (mercury-wetted relays) will not work inthis application because of their instability and inherently low repetition rates Figure 28a shows an active frequency measurement setup using a step-recovery diode impulse generator and the spectrum analyzer. The impulse generator is connected to the analyzer input through a coaxial tee providing the reference marks on the display. A small pickup loop 1s connected to the other arm of the tee which brings ina sample of the frequency to be measured. In this example a VHF transmitter is to bbe adjusted to 108.0 Me: ‘The spectrum analyzer is tuned to 100 Me and SPEC- TRUM WIDTH set to 10 MC/CM The centimeter marks on the graticule now represent 10-Mc intervals from 50 t0150Me. The impulse generator is switched to the 100-Mc repetition rate resulting in a single 100- Me reference mark appearing inthe center ofthe ana- lyzer display. Now the transmitter is keyed momen- tarily and its approximate frequency noted from its position on the display Ifthe transmitter is reason- ably clase to 108 Me, the center marker representing 100 Me is moved to the 1-cm graticule line on the left side of the display by fine tuning the analyzer fre~ quency. Next the SPECTRUM WIDTH is sot at 1 MC/ CM and the analyzer's local oscillator is. stabilized. ‘The analyzer is now sweeping from 99 to 109 Me and the impulse generator is switched to the 1-Me repe- tition rate producing a marker at each centimeter of the display graticule. The spectrum width VERNIER may be used to place the markers exactly over the centimeter marks if necessary. Now the transmitter is keyed and Its frequency compared to the calibrated screen. When the transmitter is correctly adjusted, its output response onthe analyzer display will appear eight centimeters above the 100 Me markers, The photo in Figure 29a shows the 100-Me marker and the 108-Me transmitter signal,Page 30 Appl. Note 63 rome MARKER ae ses TRANSMITTER ) @ esir0ss1 rectum ANALYSER co) Be = gen frees oe Cle ricter A . : * VAF TRANSMITTER Beas conta IN TEST FREQUENCY COMB n SEG 2.860 waRKeR WAVEMETER DIP AT 2.04 Ge @ os1r2s1 epccrRuM ANALY EER @ coo ae Gasca seo PRESENESATOR WAVEMETER LP FILTER IN Test aac Figure 29. (a) Spectrum Analyzer and high accuracy Marker Generator setup for tuning transmitter frequency. (b) Spectrim Analyzer and Marker Generator setup for measuring frequency of passive device. A comb of frequencies may also be anpliedto an RF filter to display its bandpass characteristicsAppl. Note 63 Figure 2b shows how to connect an absorption wave~ meter whose frequency accuracy is to be checked with the spectrum analyzer and impulse generator. ‘This measurement requires a series connection of the im- pulse generator through the wavemeter tothe analyzer Input since the measured device is passive. The gen- erator and analyzer are set for the appropriate fre- quencies to give a display of reference marks on the Seren within the range of the wavemeter. When the wavemeter is tuned t0 one of the marker frequencies it will absorb that particular frequency causing its amplitude to decrease, When minimum amplitude of the marker response occurs, the wavemeter is pre- cisely at the marker frequency andthe wavemeter dial calibration is verified. Since the impulse generator Fepetition rate is accurate to within .01% it is en- Page 31 tirely adequate to verify the accuracy of the wavemeter in test. The repetition rate of 1 Me makes calibration possible every 1 Mc throughout the wavemeter range if desired. ‘The © G592A Wavemeter may also be ‘checked in this manner as can any passive wavemeter up to about 6 Ge, ‘Many other possibilities exist with this technique in- cluding measurement of bandpass and low -pass fil- ters, transmission wavemeters, and tuned circuits above 10 Me. Broadband impulse generators which generate frequencies from 1 to 10 Ge are available. Using one of these as a source feeding a bandpass filter in test, the spectrum analyzer would show a complete display of the filter's response.Appl. Note 63 Page 33 APPENDIX A SECTION I a. WAVEFORM ANALYSIS. ‘The Fourier integral permits the representation of a function f(t) by the expressions: 19-B [Plel* aw o where * Fw) = fste TF at @ is the Fourier integral or Fourier transform of f(t). Extensive tables of Fourier transforms are available, and a short table of some of the more important ones 4s included with this discussion as Appendix A-IL Several useful properties of Fourier transforms are applicable to spectrum analyzer problems. The first property is that of amplitude modulation. If a carrier represented by e+J®ct is multiplied by a function £(t) using @) F, lw) Sir(ey Het elotar @ fe(tye TOreedtar F(a) = Flo-) ‘This simply states that the spectrum of a function times a carrier is the spectrum of the function with the origin shifted to the frequency we Another useful property is that of filtering and convo~ Iution (see Appendix A-II}. We can use this property to go from the Fourier transform of 2 non-recurring transient to the transform of a series of the same transient spaced by a constant time interval. This periodle funtion can be represented by 40) f(t-nT) O) ‘which can, in turn, be represented by £50) [ Eac-on] een 6 where * represents convolution. 6 (t) is the delta or singularity function in time characterized by infinite height, zero width, and unit area. Since convolution in time is equivalent to multiplication in frequency, using pair 68 of Appendix A-Il () HE e(-)] Fl) © ‘This states that the spectrum of a periodic function is a frequency comb with teeth separated infrequency by ‘the repetition frequeney and having areas equal to 1/T times the amplitude of the single transient spectrum at the frequeney of the impulse. We will illustrate the use of these first properties by two simple examples: Example (1) Single Tone AM elt) = By (14 Mcosu,t) cos w.t Cy) here [(t) = 1 + Mcosw,.t @ cos wt = eet elect @ igteere: Wer Using transform pair 58 of Appendix AI Fo) =2740(u) + M(B (w+ Wey) + 8 (w-w,,)]b (10) pore ly using the modulation property Ride li ites) catbsos eS eSlancn SF Bedarra uPense ogee tg t + Lolorw,)+ Mlolorayr a, )eB(ureg “alt any = at a A eit tn Figure A-1. Single Tone AM Spectrum Example (2) Rectangular Pulse Train 1 (8 [t] <3 Feriodia with period T 0, [el>y i using pair 7 of Appendix A-I. Fw)= Bt sin (yr/2) as) “wrt Flos Bye sin tus/2 y ‘This function is illustrated in Figure A-2. If the pulse train had been used to modulate a carrier cos wet, the resulting spectrum would have been identi€al in form to that of Figure A-2 except divided into two pasts; one centered about the positive carrier Irequeney, we, and the other about the negative carrier frequency, -w,.eo wo Figure A-2. Rectangular Pulse ‘Train and Spectrum Single Tone Frequency Modulation ‘The frequency of a single tone FM signal consists of a constant term plus 2 sinusoidal torm al =o + Be sim oy * agi as) ~ Carrier Frequency Wm ~ Modulation Frequency Aw ~ Peak frequency deviation OW = uot + Aw C05 wnt ae, {A single tone FM signal is then of the form Vit) = cosg (*) = C08 (wot + Su COS went) (17) w= Jagt gIMEC0S tnt, glut g ME C08 Ut lot IME 008 in na diy = $2 nde: of modulation Using the identity! ixaine (8) + 25049) sin 9 + 2 Jats) cos 29 9) + 2jdg() sin 30+. an the modulation theorem ¥(w) = = . vo (1) 8 (we a) + En og) [6 (wi, 28a) + CaP o(or4, Nag)! 0) where Jy (Mi) is the Bessel function of the first kind, nth order and with argument My. Appl. Note 63 ‘The spectrum is made up of a train of impulse fre~ quency functions having amplitudes given by Bessel Functions and centered about the carrier. ‘The odd terms (n odd) have odd symmetry about the ‘carrier while the even terms (n even) have even symmetry about the carrier. ‘A useful assymptotic expression for J,, (My) is” In (My) —e_C_ cos (Mp = a) 1) Pe Mr ae C, X= constants In the limit, the various components (including the carrier) oscillate harmonically with amplitudes de- creasing with the square root of modulation index ‘Two representative FM spectrums are shown in Fig- ure 3, Ibis interesting to compare Figure 3 with the single tone AM spectrum of Figure 1. Le co) oni ast Figure A-3.Single Tone FM (a) My= 1; (0) M;= 10 ‘Spectrum of a Time Product (Of special interest is a function of the form y(t £() g(t) where f(t) may be amplitude modulati a(t) frequency modulation. Using the produet-in-timesconvolution-in-frequency theorem yv@ Yl © 8 Flu) * Glo) T Bowman, “Introduction to Bessel Functions", Dover, 1958, p. 89. 2 Tid p. 108.Appl. Note 68 ‘This 48 easily interpreted for the case of periodic signals with £(t) a slowly varying function compared tog() (oF vice versa). Since the spectrum of fh), periodic function is a train of impalses, the combina tion spectrum is a train of spectra of f()) having an envelope of G(w) the spectram of git). This is illus trated in Figure 4 for the case of a pulse train with amplitude modulation, v0 yew Spectrum of an Amplitude ‘Modulated Pulse Train Figure A Page 35 b. INTERPRETATION OF THE ANALYZER RESPONSE Previously, the Fourier transform was defined as Fle)= J serHotae @) In general, F(w) is complex having a real and an imaginary part Fo) © Frole) + § Fim(e) 2) ‘which may also be expressed in polar coordinates as [PI] = CFM) + Filo)? @3) (0) = tan! Fim(o) (ea) Free) A spectrum analyzer of the type described displays ‘only (29), the amplitude spectrum. ‘The phase infor- mation 18 lost. The general time function f(t) then cannot he uniquely determined by the amplitude spec tum alone, An interesting example of this is illus trated by comparing tho spectrum of single tone AM with single tone FM of small deviation. The amplitude spectra are almost identical (see Figures A-1andA-3). ‘As a result, a spectrum analyzer would give virtually the samo display for those two radically different funetions. Even though the amplitude spectrum will not give a unique description of the function (0), it Is still very useful for signal and waveform analysis.Page 36 Appl. Note 63 APPENDIX A SECTION II TABLE OF IMPORTANT TRANSFORMS EXPLANATION OF THE TABLE The time functions ond corresponding frequency functions in thie table ere rolated by the following expressions (Direct ronsform) averse wansform) The 1/2 multipin the inverse transform erises merely bbocause the integration is wctten with respect fo o, rather than eyclc frequency. Otherwise the expressions ore ident- «al excopt for the difference of signin the exponent. AS a resul, Rnctions nd their wansforms can be interchanged ‘with only slight modification Thus, Fl) i the dee! trons form off) 18 also ve that 2rf(—w) is the direct ransform cof Ft). For example, the spectrum of a-88% pulie i rec!= ‘engular (poi 6) while the spectrum of « rectangular pulse of he form 3 (par 7). kis poi 1S i the countere port ofthe welknown fact that he spaciom ofa constant {G0 sc spike ot ero frequency. The freevency functions in th table ore in mony cases Isted both os fonctions of u and ako of p. Thi b done sorely for convenionce.Fip) in all cos i found by subst- futng for on F(a). (Not simpy p for othe notation would erdinaly indicat, That i in the vsual mathematical Convention one would wie Fo) = FUE) = Gip) where the change in letter indicates the resulting change in func tional form. The rotion used above hos grown through sage ond cous no confusion, ones understood) Ths, n the notation tioh = faerie ay = frtpt ea siackner esas eget wollen vs sstomoe in ig ape The fequeney foncione hove been plotied on lest empleo requeney scl end where convein abo Gn fogeire cle The late cls eon bing out chor ces not evident nthe Inecr ple. Thos, many of he spec re ehympoi tof orsacond depos hypebols con @llnsar plot. On @ tg plt these exymtotes become Stoight lines of slope tof “2 Les 8 or 12 docove) The tine foncion I the tele hove ell been normalized to conventnt pel amplitudes orcs or slopes For anyother sropmere) ipl ok ds Efi preps ore Thuy he spearom of tecengular pole 10 vols in empl tue and 2 seconds fog i rom pair 7) 20 seconds. ‘Agcy upon mulipticton by «conto. hong epprs- crsiom, the equency fonctions become. Ber slo Th pal mulled by tho eqwency funeon represents cinple RC el. A one coulo ine pul (pale 8) applied oh ler would prodvee on output Genpsne regome) with the pcm 5 1 coon, representing the time function go coulomb (which has the ‘Smersions of empores) Ore I vol stp fncion (oir 25) would proce the euput specrom = x vols, which represens the tine funtion (I= vols pai 48) The entries 1S through 65 inthe table ere singular fun tons for whch the nenfoms os defined above exit only ta limi For example, 15 may be thought of os he lini of pale (muliplied by 1) er »—>0. PROPERTIES OF TRANSFORMS There are @ number of important relations which describe whet happens fo the transforms of functions when the func- tions themselves ore added, multiplied, cowolved, ete. These relations state mothematicelly many of the operations countered in communications systems: operations such as linear amplification, mixing, modulation, filtering, sampling, tc, The2e relations are all readily deducible from the defin- ing equations above; but for ready reference some of the mote important ones are listed in tho Tablo of Properties (beck page). ‘Again, because of the similarity of the direct and inverse ‘transforms, a symmetry exists in these properties. Thus, laying 6 function muhiplics its spectrum by a complex expo- rental; while multiplying the function by a complex expo= rental delays ts spectrum. Multiplying any two funetions kePage 37 Appl. Note 63 cat | 6 “0 0 oa “ay by 0s <1 805,52] eat e sain? 20H <1',20" th ye =o Taney 957 = ane oan! (eae SNOWONAS ANaNDSMs ‘ON SNOLLONNS WIL SWYOSSNVUL LNVIOdWI JO aTaVL CAppl. Note 63 (homed sz dos 1 ]fpe-egremsenge sons Page 38 + (saws wan SNOLLONTS AON|NOIYSPage 39 Appl. Note 63 2 uy eu ape — f ieee oo [r= en] =] [ose (ou span (oat) JONYOUINOIS Nouvasde “onu] SWHOJSNVUL JO SHILUEdOUd KPage 40 Appl. Note 63 APPENDIX B IF AMPLIFIER RESPONSE Mention was made in the test of the phenomenon of decreased sensitivity and resolution that results when 8 CW signal Is swept by the IF amplifier at @ high rate compared to the bandwidth squared, Assuming a Gaussian response for the amplifier, the resulting fe itn Figure B-L A sweep frequency signal as illustrated in Figure B-1 can be represented by att) = oC a/Te) 2 ey using pair 10 of Appendix A-I “1/2 tral? Slo) = yar ol? Fra} (B-2) where += VGTs)/@rFs) « Tt we assume a Gaussian response, (2 w/a)? ey = or ¥/2 (72) (ea the product of S(.) H(.) gives Yl) = Sha) He) = nex. ‘The output transient is the inverse transform of this function, again using pair 10 v0 = oft (]] : 5 vg # Substituting back for ¥ and simplifying es w(t) = Tae ° a He Para Tae pevpaa? | ape oo) ‘The envelope of y(t) Is then 82,2 1 “a WO -——atoaa ew |e]: Dy " Pepa] "| te] Note that for low sweep rates ™ 1 Fs 7 32 x(t) = exp cal 4 . @8) ‘This, as was stated earlier, is a plot of the frequency response of the IF amplifier. DISTORTION If the condition on (B-8) is not satisfied, the resulting transient will be altered in beth width (time duration) and amplitude, ‘The reduction in amplitude will be “her (B-9) ‘ Noting that 8 = (n/Vin®) af where Af is the 8 db band- width, (B10) ‘A plot of this fuietion in db versus -Fa/(TyA 2) is included as Figure B-2. we solve for the 3 dh time duration At from equation (B-8) by setting the function to 1/2 and solving for the appropriate At, we get avin? sr ate as en Ina like manner, the $ db bandwidth of the fanction (y18 2Vint 2 aye [: ey [ee (p12) ‘Te Hato of these times 18 1 Pye?) 2 Be [se] @9)Appl. Note 63 ‘This is the ratio of the effective resolving bandwidth of a spectrum analyzer to the bandwidth of the IF amplifier as a function of sweep rate. Rewritten in terms of $ db bandwidth af, Afeft se [os (yt | ‘This function is plotted in Figure B-2. Figure B-2. Sensitivity Loss and Normalized Effective Bandwidth vs Normalized Sweep Rate Fy = Sweep Width T, = Sweep Time Af = 9 db IF Bandwidth Stet = Impulse Bandwidth 1985-1