- TO DOPPLER AND MTI RAD. INTRODUCTIO! wired to detect targets in the pres. an receiver noise whet uch a were req to deal with more th choes from the natural environments nce they can “clutter” the raz tude larger than aircraft echoes. Whe radar resolution cell, the aircraft s of clutter and discuss sed in the previous chapter the real world, radars have they can also receive e and, sea, and weather, These echoes are called clu! display. Clutter echoes can be many orders of magni cn eireraft echo and a cluter echo appear in the same might not be detectable. Chapter 7 describes the characteristic: ‘methods for reducing these unwanted echoes in order to detect the desired target echoes oving targets in the midst of lant the most powerful method for detecting m which is the change of frequen) The radars discus ence of noise. In detecting targets since ater Si However, clutter is by taking advantage of the doppler effect, the radar echo signal due to the relative velocity between the radar and the moving the detection of ™ aaa ofthe doppler frequency shift with a pulse radar for 3 a inca ste sabjt of is chap Fe ee eee the United Kingdom and night naval engageme lf of World War II to prevail in the crucial ait bat World War Il, fre ae ie ‘Axis powers. Almost all of the radars wed at Sept AGE Re ee pulse systems i their mission without the Mn nately, these pulse radars were able (0 accom performance military air-defense opp This would not be possible 108%. Ale tion and tracking of aircraft depend and all civil air-traffic control radars £0 ve on the doppler frequency shift to separa? & 104 Scanned with CamScanner31 Introduction to Doppler and MTI Radar 105 clutter echoes from the much smaller echoes from moving targets. Clutter echoes can be greater than the desired target c choes by as much as 60 oF 70 dB, or more, depending on the type of radar and the environment MM Rador and Pulse Doppler Radar A pulse tecting moving targets ts either an MTT (moving target indication) radar! of a pulse doppler radar,’ The MTL radar has a pulse repetition frequency (prt) low enough ea faeces range ambiguities as defined by Eq. (1.2), Ruy = off. It does, however, have many ambi- guities in the doppler domain. The pulse doppler radar, on the other hand, is just the op- posite. As we shall see later in this chapter, it has a prf large enough to avoid doppler am- biguities. but it can have numerous range ambiguities. There is also a medium-prf pulse doppler that accepts both range and doppler ambiguities, as discussed in Sec. 3.9. In addition to detecting moving targets in the midst of large clutter echoes, the doppler frequency shift has other important applications in radar; such as allowing CW (continu- ous wave) radar to detect moving targets and to measure radial velocity, synthetic aper- ture radar and inverse synthetic aperture radar for producing images of targets, and me- teorological radars concerned with measuring wind shear. These other uses of the doppler frequency shift are not discussed in this chapter. adar that employs the doppler shift for de- Doppler Frequency Shift The doppler effect used in radar is the same phenomenon that ‘was introduced in high school physics courses to describe the changing pitch of an audi- ble siren from an emergency vehicle as it travels toward or away from the listener. In this chapter we are interested in the doppler effect that changes the frequency of the electro- magnetic signal that propagates from the radar to a moving target and back to the radar. If the range to the target is R, then the total number of wavelengths A in the two-way path from radar to target and return is 2R/A. Each wavelength corresponds to a phase change of 277 radians. The total phase change in the two-way propagation path is then 2R b= ax a 7 4mRIA tn If the target is in motion relative to the radar, R is changing and so will the phase. Dif- ferentiating Eq. (3.1) with respect to time gives the rate of change of phas » which is the angular frequency abs Said S Azv a an d dt x 2fa (3.2) where v, = dR/dt is the radial velocity (meters/second), or rate of change of range with time. If, as in Fig. 3.1, the angle between the target's velocity vector and the radar line of sight to the target is 6, then v, = v cos 6, where v is the speed, or magnitude of the vec- tor velocity. The rate of change of ¢ with time is the angular frequency wy = 27» where Suis the doppler frequency shift. Thus from Eq. (3.2). Scanned with CamScanner ae106 CHAPreR 9 © MT and Pulse Doppler Radar Figure 3.1 Geometry of radar and target in deriving the doppler frequency shift. Radar, torget, ‘and direction of target travel all ie in the some plane in ths illustration, ‘The radar frequency is f, = c/A, and the velocity of propagation c = 3 X 10° nvs. If the doppler frequency is in hertz, the radial velocity in knots (abbreviated kt), and the radar wavelength in meters, we can write = L03v,(kt) _ v-kt) fa (Ha) = TE ~ amy bal ‘The doppler frequency in hertz can also be approximately expressed as 3.43v;,f,, where f; is the radar frequency in GHz and y, is in knots. A plot of the doppler frequency shiftis shown in Fig. 3.2 as a function of the radial velocity and the various radar frequency bands. Figure 3.2 Doppler 10,000, hrequeney shih from @ moving torget os a function of the torget’s radial velocity and the radar frequency band. 3 10m i E 2 F 1 ive Radi velsiy-knots —————— Scanned with CamScannerFigure 3.3 ‘9h eprom ofa simple pulse radar that exrocs the dope ects stationary cluter echoes. The frequency F 3.1 Introduction to Doppler and MTI Radar 107 Simple CW Doppler Radar Before discussing the use of doppler in pulse radar, itis im structive to begin by considering the doppler frequency shift experienced with a CW rudit The block diagram of a very simple CW radar that utilizes the doppler frequency shift to detect moving targets is shown in Fig. 3.3a, Unlike a pulse radar, a CW radar transmits while it receives. Without the doppler shift produced by the movement of the target, the weak CW echo signal would not be detected in the presence of the much stronger signal from the transmitter. Filtering in the frequency domain is used to separate the weak doppler-shifted echo signal from the strong transmitter signal in a CW rada The transmitter generates a continuous (unmodulated) sinusoidal oscillation at fre- quency f,, which is then radiated by the antenna. On reflection by a moving target, the transmitted signal is shifted by the doppler effect by an amount * f. 1s was given by Eq 3.3). The plus sign applies when the distance between radar and target is decreasing (8 closing target); thus, the echo signal from a closing target has a larger Irequency than that which was transmitted. The minus sign applies when the distance is inercasing (a reced- ing target). To utilize the doppler frequency shift a radar must be able to recognize that the received echo signal has a frequency different from that which was transmitted. This is the function of that portion of the transmitter signal that finds its way (or leaks) into the receiver, as indicated in Fig. 3.3. The transmitter leakage signal acts as a reference to determine that a frequency change has taken place, The detector, or mixer, multiplies the echo signal at a frequency f, fy with the transmitter leakage signal f,. The doppler filter allows the difference frequency from the detector to pass and rejects the higher fre- quencies. The filter characteristic is shown in Fig. 3.3a just below the doppler-filter block. Tt has a lower frequency cutoff to remove from the receiver output the transmitter leak- age signal and clutter echoes, The upper frequency cutoff is determined by the maximum ow . transmitter Pale ami cae Echo | Transmitter cw signal) leakage Duplexer oscilator Detector | _[ Doppler |. y, Li (mixer) filter “ Reference signal ceiver Ls} Poppler | l i g Frequency (b) fo) () vom thot extracts the doppler frequency shit from o moving trgel and Lo} Simple CW radar block diogrom the) ert sia is shown a he lower right. (b) Block nse of ore hrequency shift ofthe echo signal from @ moving forge. Scanned with CamScanner1OB CHAPTER 3 © MTIand Pulse Doppler Radar radial velocity expected of moving targets. The doppler filter passes signals with frequency fy located within its pass band, but the sign of the doppler is lox. gga ™PEt the direction of the target motion, CW radars can be much more complicated thar simple example, but it is adequate as an introduction to a pulse radar that utilize, doppler to detect moving targets in clutter. the Pulse Radar That Extracts the Doppler Frequency-Shified Echo Signal One canny simply convert the CW radar of Fig. 3.3a to a pulse radar by turning the CW osciy on and off to generate pluses. Generating pulses in this manner also removes the refer ence signal at the receiver, which is needed to recognize that a doppler frequency ship has occurred. One way to introduce the reference signal is illustrated in Fig. 3.3b, Thy output of a stable CW oscillator is amplified by a high-power amplifier. The amplifier turned on and off (modulated) to generate a series of high-power pulses. The received echo signal is mixed with the output of the CW oscillator which acts as a coherent ref erence to allow recognition of any change in the received echo-signal frequency. By co. herent is meant that the phase of the transmitted pulse is preserved in the reference sig. nal. The change in frequency is detected (recognized) by the doppler filter. The doppler frequency shift is derived next in a slightly different manner than was done earlier in this section. If the transmitted signal of frequency f, is represented as A, sin(27f,1), the received signal is A, sin{27f,(t — T,)], where A, = amplitude of transmit- ted signal and A, = amplitude of the received echo signal. The round-trip time Ts is equal to 2R/c, where R = range and c = velocity of propagation. If the target is moving toward the radar, the range is changing and is represented as R = Ro — v,1, where v, = radial ve- locity (assumed constant). The geometry is the same as was shown in Fig. 3.1. With the above substitutions, the received signal is Veee = A, sin [pra(t + mh = salto 5) © The received frequency changes by the factor 2fy,/c = 2v,/A, which is the doppler frequency shift f,.* If the target had been moving away from the radar, the sign of the doppler frequency would be minus, and the received frequency would be less than that transmitted. The received signal is heterodyned (mixed) with the reference signal Ave sin 29ff and the difference frequency is extracted, which is given as Va = Ay cos (2mfit ~ 4rRo/A) fal where Ay = amplitude, f, = 2v,/A = doppler frequency, and the relation f,A = ¢ was used (The cosine replaces the sine in the trigonometry of the heterodyning process.) For st tionary targets f, = 0 and the output signal is constant. Since the sine takes on valves +1 to 1, the sign of the clutter echo amplitude can be minus as well as plus. 08 other hand, the echo signal from a moving target results in a time-varying output (due the doppler shift) which is the basis for rejecting stationary clutter echoes (With doppler frequency) but allowing moving-target echoes to pass 1 +The teem doppler frequency sik, doppler requoncy, and doppler shit ae vied ierchongeabl in is OM Scanned with CamScanner3.1 Introduction to Doppler and MTI Radar 109 If the radar pulse width is long enough and if the target's doppler frequency is large enough. it may be possible to detect the doppler frequency shift on the basis of the fre~ eee change within a single pulse. If Fig. 3.4a represents the RF (or IF) echo pulse train, Fig. 3.4b is the pulse train when there is a recognizable doppler frequency shift, To detect a doppler shifl on the basis of a single pulse of width generally requires that there | be at least one cycle of the doppler frequency f, within the pulse: or that fyr > 1. This condition, however, is not usually met when detecting aircraft since the doppler frequency Ja is generally much smaller than 1/z. Thus the doppler effect cannot be utilized with a single short pulse in this case. Figure 3.4c is more representative of the doppler frequency for aircraft-detection radars. The doppler is shown sampled at the pulse repetition fre- quency (prf). More than one pulse is needed to recognize a change in the echo frequency due to the doppler effect. (Figure 3.4c is exaggerated in that the pulse width is usually small compared to the pulse repetition period. For example, r might be of the order of 1 : us, and the pulse repetition period might be of the order of 1 ms.) | ‘Sweep-to-Sweep Subtraction and the Delay-line Conceler Figures 3.5a and b represent ‘ (in a very approximate manner) the bipolar video (both positive and negative amplitudes) from two successive sweeps* of an MTI (moving target indication) radar defined at the beginning of this chapter. The fixed clutter echoes in this figure remain the same from sweep to sweep. The output of the MTI radar is called bipolar video, since the signal f hhas negative as well as positive values. (Unipolar video is rectified bipolar video with - beet Nk AN Ah = Figure tation of the echo pulse train at ether the RF or IF portion of the rece ®) “ides poke ra Pier the phate detector when the doppler Frequency f4> 1/; (video pulse train forthe doppler equency fa< 1/7, which is usualy the cose for cirerafsurveillance radar. The ‘doppler requency signal is shown doshed in (), 03 if it wore CW. Note that the pulses in (e) have ‘on exaggeraied widih compared to the period of he doppler frequency. the time between two transmitted pultes, or the pulse repetition interval. what ec in hi td bake theta mae cp The won Toop ing et coma ge oucon bom clade oy tin Sy com he ofan oring the ime ofa pulse ropttion Scanned with CamScannerWO CHAPTER 3. © MTland Pulse Doppler Radar Moving targets Yn A @ a f@ —Time Figure 3.5 Two successive sweeps, (o) and (b), of an MII radar A-scope disploy (omplitude os 0 function of time, or range). Arrows indicate the positions of moving targets. When (b) is subtrocied from (a), the result is (c) and echoes from stationary targets are canceled, leaving only moving targets only positive values.) If one sweep is subtracted from the previous sweep. fixed clutter echoes will cancel and will not be detected or displayed. On the other hand, moving ta gets change in amplitude from sweep to sweep because of their doppler frequency shift If one sweep is subtracted from the other, the result will be an uncancelled residue. as shown in Fig. 3.5c. Subtraction of the echoes from two successive sweeps is accomplished in a delay line canceler, as indicated by the diagram of Fig. 3.6. The output of the MTI receivers digitized and is the input to the delay-line canceler (which performs the role of a dopplet filter). The delay 7 is achieved by storing the radar output from one pulse transmission or sweep, in a digital memory for a time equal to the pulse repetition period s0 T = Ty = Wfy The output obtained after subtraction of two successive sweeps is bipoW” (digital) video since the clutter echoes in the output contain both positive and negative amplitudes [as can be seen from Eq. (3.6) when f, = 0}. It is usually called video. though it is a series of digital words rather than an analog video signal. The absolute Bipolar Unipolar digital Ggital) oe video video fo | AD [Pes | fSutwace] —[Absoine] S|. Analog re —lcomener =, valve Ti ‘display Digital MT! ovtpet to sutomate deetio® and daa Scanned with CamScannerFigure 3.7 Block diagram of on MTI radar thot vies o Power amplifier 23 he tronsmiter oY V1 Tntroduction ty Doppler and MITE Radar 14 ‘eo, Unipolar video is needed if Aignily only. ‘The unipolar digital video bn wlop signal by the dipital-to unalog (D/A) converter iF the processed St on a PPE (plan positon tndicntor), Alternatively, the digital sp ping aa mek the detection dee Eyed for ote A a peo 7 a ‘hing und/or rpet recognition, ‘The name delay line can: celer was originally applied when analog lay Hines ‘omy acoustic) were ued in the carly MTL radars, Even though analog delay Kno have been repluced by digital memo- Fies, the mame delay-tine cunecler i sil ued tw denerthe the operation of Fig, 4.6, an analog display is used then converted to an an Signal is 10 be display nals may be use cessing, ‘MT Radar Block Diagram ‘The Necessary for an MTL radar, bu diagram of an MTL + block diagram of Fig, 3,3 Mustrated the reference signal ut it is oversimplified. A slightly more elaborate block i Har employing a power amplifier as the transmitter ix shown in Fig, 4.7. The local oscillator of an MT iadar’y supetheteredyne receiver must be more Stable than the local oscillator for a radu that does not employ doppler, It the phase of the local oscillator were 10 change signticuntly between pulses, an uncancelled clutter Fesidue can result at the output of the delay-tine canceler which might be mistaken for a moving target even though only clutter were present. To recognize the need for high sta- bility, the local oscillator of an MTT receiver is called the stalo, which stunds for stable ched filter, as is usually the case in radar, local oscillator, The IF stage is designed as a ma Instead of an amplitude detector, there is a phase detector following the IF stage, This is a mixer-like device that combines the received signal (at IF) and the reference signal from the coho so as to produce the difference between the received signal and the reference signal frequencies.’ This difference is the doppler frequency. The name coho stands for Pulse modulator Sith Power Duplexer amplifier Sith Sithth Mix —{ w* | pL [uta F Coho amplifier S Reference signal Phase detector fa To delay-line ceancelet Scanned with CamScannerM2 CHApteR 9 © MTIand Pulse Doppler Radar | coherent oscillator to signify that its the reference signal that has the phase of hp } ‘mutter signal. Coherency with the transmitted signal is obtained by using the sym j Coho and the stalo signals asthe input signal to the power amplifir. Thus the yan frequency is the sum of the stalo frequency f; and the coho frequency f,. Ths i, Plished in the mixer shown on the upper right side of Fig. 3.7. The combination om stalo and coho sometimes is called the receiver-etciter portion of the MTI rader (1 % the receiver stalo and coho to also generate the transmitter signal insures better sy than if the functions were performed with two different sets of oscillators, The output, the phase detector is the input to the delay-line canceler, as in Fig. 3.6. The det = canceler acts as a high-pass filter to separate the doppler-shifted echo signals of Tae targets from the unwanted echoes of stationary clutter. The doppler filter might be a sig gle delay-line canceler as in Fig. 3.6; but it is more likely to be one of several other axe elaborate filters with greater capability, as described later in this chapter. The power amplifier is a good transmitter for MTI radar since it can have high ss. bility and is capable of high power. The pulse modulator tums the amplifier on and off ip generate the radar pulses. The klystron and the traveling wave tube have usually been the preferred type of vacuum-tube amplifier for MTI radar. The crossed-field amplifier has also been used, but it is generally noisier (less stable) than other devices; hence, it might not be capable of canceling large clutter echoes. Triode and tetrode vacuum tubes have also been used with success for radars that operate at UHF and lower frequencies, but they have been largely replaced at these lower radar frequencies by the solid-state transistor. The tan- sistor has the advantage of good stability and it does not need a pulse modulator. Before the development of the high-power klystron amplifier for radar application in the 1950s, the only suitable RF power generator at microwave frequencies was the mag- netron oscillator. In an oscillator, the phase at the start of each pulse is random so that the receiver-exciter concept of Fig. 3.7 cannot be used. To obtain a coherent reference in this case, a sample of each transmitter pulse is used to lock the phase of the coho to that of the transmitted pulse until the next pulse is generated. The phase-locking procedure is t- peated with each pulse. The RF locking-pulse is converted to IF in a mixer that also uses the stalo as the local oscillator. This method of establishing coherence at the receiver some- times is called coherent on receive. Further information on the MTI using an oscillator may be found in previous editions of this text or in the chapter on MTI in the Radar Handbook.' 3.2. DELAY-LINE CANCELERS ‘The simple MTI delay-line canceler (DLC) of Fig. 3.6 is an example of a time-domain {filter that rejects stationary clutter at zero frequency. It has a frequency response funeto® 'H(/) that can be derived from the time-domain representation of the signals. , . Frequency Response of the Single Delay-Line Canceler The signal from a target at Ro at the output of the phase detector can be written A rr Vs = ksin(nfa - 0) Scanned with CamScanner| | 42 Delay-Line Cancelers 199 lo AMR, IA, Ry range where fi = doppler trequeney shift, day = a constant phase nee, the time equal to zer0, A= wavelength, and £ = amplitude of the signal, [For conve cosine of Eq. (3.6) hay been replaced by the sine. The signal from the previous radar trans mission is similar, except itis delayed by a time 7,, = pulse repetition interval, and 1% (3.8) Va =k sin fit ~ Ty) ~ dol The amplitude & is assumed to be the same for both pulses. The delay-line canceler sub- tacts these two signals, Using the trigonometric identity sin A— sin B= 2 sin((A ~ B)/2) cos (A + BY/2], we get eee ie " in (3.9) v Vi = Vo = 2k sin (afiT,,) cos | 2afdt Lr dy " 2 The output from the delay-tine canceler is seen to consist of a cosine wave with the same frequency f as the input, but with an amplitude 2k sin (7f,,). Thus the amplitude of the canceled video output depends on the doppler frequency shift and the pulse repetition pe- riod, The frequency response function of the single delay-line canceler (output amplitude divided by the input amplitude 4) is then Hf) = 2sin (afily) (3.10), Its magnitude | H(f)| is sketched in Fig. 3.8. The single delay-line canceler is a filter that does the job asked of it: it eliminates fixed clutter that is of zero doppler frequency. Unfortunately, it has two other properties that can seriously limit the utility of this simple doppler filter: (1) the frequency response function also has zero response when moving targets have doppler frequencies at the prt and its harmonics, and (2) the clutter spectrum at zero frequency is not a delta function of zero width, but has a finite width so that clutter will appear in the pass band of the delay-line canceler. The result is there will be target speeds, called blind speeds, where the target will not be detected and there will be an uncanceled clutter residue that can in- terfere with the detection of moving targets. These limitations will be discussed next. The response of the single delay-line canceler will be zero whenever the Blind Speeds (aful,) in Eq. (3.10) is zero, which occurs when af,I, = 0, *m, £2, magnitude of sin +37,.... Therefore, 2y, n ates n= 01,2, 1M) laa 7, Ga) 2 S$ zi "0 GevT 2h Mh Frequency response |H(f)| of single delayline conceler a» given 3.8 Mogntide of he requency eran ih oes 13.10), where f, = pulse repetition Scanned with CamScannerTE V4 cHaprer 3 © MTIand Pulse Doppler Radar 10 the zero response at zero frequency, there will also be This states that in addition ( ver the doppler frequency f, = 2v,/4 aes a al response of the delay-line canceler whene tiple of the pulse repetition frequency fj, (The doppler shift can be negative of depending on whether the target is receding or approaching, When considering ch speed and its effects, the sign of the doppler can be ignored—which is what is done The radial velocities that produce blind speeds are found by equating Eqs. 3.11) ang 33, and solving for the radial velocity, which gives nA _ MAfp _ “ae he n= 123 fay nth blind speed. Usually only the first blind speeg 1 multiples of vy. If A is measured in meten, the first blind speed can be written where v, has been replaced by Vm the vy, is considered, since the others are intege! fy in hertz, and the radial velocity in knots, v1 (kt) = 0.97 A (m) fp (Hz) ~ A (m) fy (Hz) faa A plot of the first blind speed as a function of the pulse repetition frequency and the vx- ious radar frequency bands is shown in Fig. 3.9. Blind speeds can be a serious limitation in MTT radar since they cause some desl moving targets to be canceled along with the undesired clutter at zero frequency. Basd fon Eq, (3.13) there are four methods for reducing the detrimental effects of blind speeds Figure 3.9 Plot of the first Unambiguous range, nmi blind speed, Eq, (3.13), a5 0600 490300200 100s0 60, 40.30 20 108 SS4 3 2 Yt function of the pulse repetition frequency for the various rodor frequency bands. ind speed, knots First 100 100 1000) 10,000 Pulse repetition frequency. hertz Scanned with CamScanner12 Delay Fine Canvelers 08 1. Operate the radar at ony wave 2. Operate with a high pulye 4. Operate with mone emathy Chow fhequencion) tepetttion Hequeney Han ane putts petition Hequeney 4. Operate with more than an RE fequeney (wavelength) Combinations of two OF mone effect of blind speeds. Huet of limitations, s0 there application Consider the cave wheve a tow RE frequency fy chosen to avokd blind speeds, far example, the first blind speed iy to be than 640 KECapproxinately Mav 1) and the pal is Selected ay 40 Hy (whieh lve an anauguon tae a2 rl) thet the fdr wavelength from Vi. (1.11) iy 2a, ‘his corresponds tw a fhegueney at LAO Me (the VE region of the spectrum), Although there were inany vashareat VED butt Av he 19308 and curly 1940s and there are sil advantages to opeutiny a sahara there frequencies, VII iy not usually considered w desituble frequency chatee for bone bane Sir-surveillance radar for a number of reasons: C1 resolution ln range anal anple ai pata due 1 narrow bandwidths and lane beamwidth, (2) this pation of the electromagnetic spectrunt is erowded with other than radar services (stich ay broadewst FM. anid TV), und (3) low-altitude coverage generally is pour, ‘Thuy attempting to tive low fequeneles to avoid the blind speed problem is not usually a desitable option for many radar applicat On the other hand, prt to avoid bl result, For e: UF the above are alo parvible to futher alleviate the eve Four mths ho partiouhar advantapen ay well ae {8 OL always a clout choice ay 4 avhbel A use In any partie Ho lowe if we choose (0 operite ala high RE trequeney and increase the nd speeds, we would then have to tolerate the many range anibipuitios that uple, if the first blind speed was again choyen to be O40 KEand the wave: 0,1 (an S-band frequeney of 3000 MHZ), the prt would thave to be 6600 eximumn unambiguous range of 12.4 nmi, which iy small for many radar applications, (Such an approach, however, is used successfully inv pulye doppler radars, as discussed later in this chapter.) When two or more prfs are used in a radar, the blind ypeeds wt one prt generally are different from the blind speeds atthe other prls, Thus targets that are highly attenuated with one prf might be readily seen with another prf, This technique iy widely ued with air-surveillance radurs, especially those for civil air-trattic control. A disadvantage of a multiple-prf waveform is that multiple-time: aroun er echoes (thom regions beyond ir jous range) are not cunceled but the required frequency change is often larger than might be possible within the waa frequency bands allocated for radar use, A limitation of multiple frequencies is the need etree cael night be desirable o tolerate the blind ypeeds rather than ies citations of the above methods, As in many aspects of engleering, there ix limitations can be accepted in any particular application Scanned with CamScanner16 CHAPTER 3 © MTL and Pulse Doppler Radar 1 because of the sampled nature of the pulse radar waveform, Thy he cause of ambiguities, or aliasing, in the measurement of iy: (at the prf) can give rise to ambigy, speeds occu it is sampling that is U doppler frequency, just as sampling in a pulse radar ities in the range measurement. ‘The other limitation of the single delay-line canceler is insufficien, attenuation of clutter that results from the finite width of the clutter spectrum. The single delay-line canceler whose frequency response was shown in Fig. 3.8 does what its up. posed 10 do, which is to cancel stationary clutter with zero doppler shift. In the “ea world,” however, the clutter spectrum has a finite width due to such things as the internal motions of the clutter, instabilities of the stalo and coho oscillators, other imperfections of the radar and its signal processor, and the finite signal duration. (The factors that widen the clutter spectrum will be discussed later in Sec. 3.7.) For present purposes we will as- sume the clutter power spectral density is represented by a gaussian function, and is writ- ten as Clutter Attenuation 22 x HEN) = Wo x0 ( £,)- woes (-F) f20 tau where Wo'= peak value of the clutter power spectral density, at f= 0, o- = standard de- viation of the clutter spectrum in hertz. and g, = standard deviation of the clutter spec- trum in meters/second. The relation between the two forms of the clutter-spectrum stan- dard deviation is based on applying the doppler frequency expression of Eq. (3.3) such that @, = 2o,/A. The advantage of using the standard deviation oj, is that it is often in- dependent of the frequency; whereas a, is in hertz and depends on the radar frequency. Nevertheless, we will generally use a in this chapter for the clutter spectrum, “The consequences of a finite-width clutter spectrum can be seen from Fig. 3.10. The frequency response of the single delay-line canceler shown by the solid curve encom passes a portion of the clutter spectrum; therefore, clutter will appear in the output. The greater the standard deviation the greater the amount of clutter that will be passed by the filter to interfere with moving target detection. The clutter attenuation (CA) produced by a single delay-line canceler is Relative response Frequency 3.10 Relotve frequency response of o single deloyline canceler (solid curve) and Me, double delaysine canceler (dashed curve}, along with the frequency spectrum of the clutter {shoded rea). Nole the clutter spectrum is folded over ot the prf and its harmonics because of the some ature of pulse rodor waveform Scanned with CamScanner 732 DelayLine Cancels 17 Pm Ca = = mus | MUN a where H(f) is the frequency respoase of the delay-tine camveler, Substitaming fe NU) (Eq. (3.10)). the clutter attenuation besos (Wo exp [-Pretler me JWoexl-f2034 sintarT ge Te? If the exponent in the denominator a the right-hand pact of this equation is sana, Whe eX ponential term exp [—x] can be replaced by 1 —.t so that Ru re . ore In this equation the pulse repetition period T, bas beea replaced by Uf. The chatter at tenvation provided by a single delay-hine canceler is noe sufficient fe raost NUTT radar applications. If a second delay-line canceler is placed in cascade, the frequency respoase of the two filters is the square of the single delay-line camcelen, ot Hy) = 4 sa" (2/T,) me 10 by the dashed curve (except that we have pled the rela: absolute respoase). Less of the clutter spectrum is imaged veithin the frequency response of the docble delay-tine canceler Beace, it aeaakes me of the clutter. The clutter attenuatioa for the double detay-tine camveter is fe et CA” Bete Jere Re can be cascaded 1 obsain a frequemcy respomse AUF) ‘Additional delay-line cancelers ii ae canceler given by By. (A1O\ where a is which is the nth power of the single delay-tine the number of delay-line cancelers. ‘The clutter attenuativa is a useful measere of the Peete mance va canceling charter tot it bas 3a inherent weakness if one is mot mance of an Mr aqenuatnn can be made ante By tring Othe radar eine? Ta, aacry emul ot be done Knowingly since it aso ehmiares the desire move Se, eons igual, To avoid the problem of ustesing tamer anenagivn a Ws sons Af TEEE defined* a measure of performance Known as the MTT a. nchdes the signal gain 2s well 2 the chuter ameawatton, Wis de ratio at the comput of the chater Ger divided by te MTI Improvement Factor Scanned with CamScanner— 4 | VO CHAPTER 38 MTV amt Dube Dappter Ratan Stgnaleto-ctotter ratio at the input af the clutter Mtter, averaged UniFOEMTY OVEF all tage, antial velocities a nvterest! IV is expressedt as tvinygtefclutter en iGo Se =he ~ Lay S inyprvement factor =) = ingtocttterhy fa” Ca Sn = CAX average gain tay The vertioal line ont the Fightin the above equation indicates that the average is taken with resqevt to chapter fhequeney fe The innprovement favtor can Be expressed as the clit attenuation CA = (Clog) tines the average filter gain, The aventge gain is determing thom the filter response MUP) anal is usually: small compared! to the clutter attenuation, The average gain fie a single delay-tine canceler is 2 ana for a double delay tine cancer is 6 The impeavemient factors for single and double delay-tine cancelers are Lasingle DLO) = =e fa Iwo 1 x DLO = "a * ime? Hradouble DLO) = Se gae Ne oh “The general expression for the improvement factor for a canceler with 1 delay-tine can t elers in cascade is® 4, (n cascaded DLC) = [AS with the previous expressions, this applies when a7 5 is small, A plot of the im provement factor as a function of & 7, is provided later in Fig. 3.13, The ratio of ist measure of the amount af “doppler space” vocupied by clutter. Equation (3.23) also ag plies far the so-called N-pudse cunceler with N= n+ 1, to be discussed next. NPubse Delay-line Canceler A double delay-line canceler is shown in Fig. 3.113. Aca celer with two delay Tines that has the same frequency response as the double delay-tine ‘canceler, but which is arranged differently, is shown in Fig. 3.11b. This is known 3s # three-pulse canceler since theve pulses are added, with appropriate weights as shown. v ‘obtain a sin*( 2/7,) response, the weights of the three pulses are +1, -2. +1 When the input is s(2), the outpat of the three-pulse canceler is then st) ~ 290 + T,) + 50+ 2) which is the same as the output from the double delay-tine canceler MO) — t+ T,) — [0+ T,) — sr + TL “Thus the double delay-tine canceler and the three-pulse canceler have the same frequen? response function. A fourputse canceler with weights +1, ~3, +3, —1 has a frequency response FY portional to sin’ (x/7,). A five-pulse canceler has weights +1, —4, +6, —4 th" Scanned with CamScanner32 Delay-Line Cancelers 119 Input Figure 3.11 + (o) va ay PD — owe ling canceler; (b)threepulse cancel. The hwo configurations have the tame frequency {Paty ine r= ty, |} | Delay tine 7» (b) {0} Double del reiponte, The threepulie canceler of (b) lt an example of a tranaverso filer, Figure 3.12 Transversal, ot nonrecurtive, fer for MT Aignal Proceasing, the number of delay lines, there are n + 1 = N pulses available to produce a trequency Tesponse function proportional to sin" (mf7,,) when the weights are the coefficients of the expansion of the binomial series (1 ~ x)" with alternating signs. The binomial weights with alternating sign are given by al (n-1+ D-H! ‘The N-pulse canceler has the same frequency response as n single delay-line cancelers in cascade, where n= N ~ 1. The greater the value of N, the greater will be the clutter attenuation, we (-1t Pe Lent (3.24) Transversal (Nonrecursive) Filter The three-pulse canceler of Fig. 3.11b is an example of a sransversal filter Its general form with n delay lines is shown in Fig. 3.12. The weights ww, are applied to the N = n + 1 pulses and then combined in the summer, or adder. The transversal filter is a time-domain filter with feed-forward lines and taps with weights w,. Scanned with CamScanner120 CHAPTER 3. © MT and Pulse Doppler Radar “The delays 7; are usually equal, but they need not be (Sec. 3.3). The configuration of 3.12 is also known as a nonrecursive filter, feed-forward filter, finite impulse-res (FIR) filter, or finite memory filter. “Optimum” MTI Filter Transversal filters can have other than a sin"(7fT,) frequ response by choosing other than binomial weights. It is logical to ask what might re optimum MTI filter. If one speaks of an optimum, there must be some criterion on whieh it is based. In the case of the so-called optimum MTT filter® the criterion is that it may. mize the improvement factor, Eq, (3.20). It can also be considered as one which mag. mizes the clutter attenuation. It happens that a close approximation to the “optimum” filter is a transversal filter with binomial weights of alternating sign that has a freq response function proportional to sin" (7fT,), where 1 is the number of delay lines, was discussed above. The optimum weights (1,~1) of a transversal filter with a single de lay line are the same as that of the single delay-line canceler, if the clutter spectrum can be modeled as a gaussian probability density function.’ The difference between the thee- pulse canceler and the “optimum” three-pulse MTI filter is less than 2 dB. ® The diffe- tence is also small for higher values of n. Figure 3.13 is a plot of the improvement factor as‘a function of o:/f,» where 0 = the standard deviation of the clutter spectrum assum- ing it is of gaussian shape, and f, = the pulse repetition frequency. ° The solid curves ap- ply for optimum weights and the dashed curves apply for binomial weights. There are two things to be noted from this plot. First, the differences between the two sets of curves ae small so that we will take the optimum MTI filter to be adequately approximated by the filter with binomial weights whose response is proportional to sin" (mfT,). Second, su ficient improvement factor for many applications might be obtained with no more than Wo Figure 3.13. MT improve- 140 iment factor for on Nepulse delayine canceler with binomial weights (dashed 120 curves) and optimum weights {solid curves) as a function of the clutter spectral width ogg 10 | (her Andrews") 8 g 8 i g 6 = 40] 20 ° 0.001 ‘or (Clutter spectral width)/ radar prf) Scanned with CamScanner22 Delay-Line Cancelers 121 or three delay lines, if the improvement factor is the major criterion for the design of the MTI filter. The term optimum is sometimes mistaken for the best that can be achieved. Opti- mum. however, is defined as the best under some implied or specified conditions. A SO- called optimum solution might not be desired if the conditions for which it is defined are ‘not suitable. This happens to be the case for the “optimum” MTI filter mentioned above. {tis optimum if one wishes to have a filter that maximizes the improvement factor oF the clutter attenuation. This may seem to be a suitable criterion, but it is not necessarily what one wants to achieve in a MTI filter. As the number n of delay lines increases in a filter with H(S) ~ sin” (mfT,), the response of H(f) narrows and more and more clutter is re- jected. The narrower bandwidth of the filter also means that fewer moving targets will be detected. If, for example, the ~10 dB width of H(f) is taken as the threshold for detec tion, and if all targets are uniformly distributed across the doppler frequency band, the following reductions in performance occur: 20 percent of all targets will be rejected by a two-pulse canceler 38 percent of all targets will be rejected by a three-pulse canceler 48 percent of all targets will be rejected by a four-pulse canceler ‘Thus if the “optimum” clutter filter is used, the loss of desired target detections is another reason it should not employ a large value of n. “Rectangular-like” Transversal Filter Response If one examines the clutter spectrum, such as the shaded region in Fig. 3.10, it can be seen that a desirable filter should ap- proximate a rectangular passband that attenuates the clutter but has uniform response over ‘as much of the doppler space as practical. It would not have as much clutter attenuation as an “optimum” filter of the same number of delay lines; but as we have scen from Fig. 3.13, the clutter attenuation of the “optimum” generally is far greater than can be used in practice when the number of delay lines n is large. A transversal filter, as in Fig. 3.12, ‘can be designed to approximate a rectangular passband if it contains a sufficient number of delay lines and if the weights w, are chosen appropriately. ‘Some examples of transversal, or nonrecursive, filters for MTI applications that have appeared in the literature have been summarized by Y. H. Mao.'® Procedures for nonre- cursive filter design can be found in classical text books on digital filters. An early ex- ample due to Houts and Burlage,'! based on a Chebyshev filter response that employs 15 pulses, is in Fig. 3.14. Also shown for comparison is the response of a three-pulse can- celer with binomial weights aud the response of a five-pulse canceler with “optimum” weights. Generally, the goal in such filter design is to achieve the necessary improvement factor by choosing the attenuation in the stopband of a bandpass filter, the extent of the stopband, the extent of the passband, and the ripple that can be tolerated in the passband. The large improvement factor that results with the “optimum” MTI filter when n is large can be traded for increased doppler-frequency passband. For example, when fy = 0.001, Eq. (3.22) indicates that the three-pulse-canceler, or double delay-line canceler, (which is close to the “optimum”) has a theoretical improvement factor of 91 dB. This is a large improvement factor and is usually more than is required for Scanned with CamScannert f 122 CHAPTER 3 © MTland Pulse Doppler Radar & Br HAS 0 tp Frequency Amplitude response IH Figure 3.14 Amplitude response for three MT! deloyine concelers. (1) Classicol threepulse canceler, (2) fiveppulse delayline canceler with optimum weights, and (3) 1S:pulse Chebyshev design 1 (Aer Houts ond Burloge."") routine MTI radar applications. Furthermore, it might be difficult to achieve such high values in practice considering the problems of equipment instability and other factors tha can limit the improvement factor. The five-pulse “optimum” of Fig. 3.14 is indicated by Houts and Barlage'' as having an improvement factor of 85 dB for this clutter spread and the 15-pulse Chebyshev design has an improvement factor of 52 dB. If there are many pulses available for MTI processing, an approximately rectangular filter response may be Preferred over the “optimum” since increased doppler-frequency passband is more im- Portant than extremely large theoretical values of improvement factor which are not needed ‘or cannot be achieved in practice. It has been said'’ that even with only five pulses avail- able, a five-pulse Chebyshev design provides significantly wider doppler space than the five-pulse “optimum” design. ‘When only a few pulses are available for processing, there is probably little that can be done to control the shape of the nonrecursive filter characteristic, and there might not be much gained by using other than a filter with binomial weights that has a sin" (n/T) response. Recursive Filters The N-pulse nonrecursive canceler discussed above allows the designer N zeros for synthesizing the frequency response using the classical z-plane procedure for filter design. Each feedforward line and its weight w, correspond to a zero in classical fi. ter design on the z-plane. Filter design using only zeros does not have the flexibility filter design based on poles as well as zeros. Poles can be obtained with delay-line 3 caded delay lines, within the limit oF infinite impulse response (HR) filters. Significantly fewet delay lines (and fewer PU) are needed to achieve desirable frequency-response functions than with nonrecursive ters that only have zeros available for design in the z-plane, Scanned with CamScanner32 Delay-Line Cancelers 123 Vio, (ben = Fi 3. nonical ¢¢ Fue as Canonical configuration of a recursive delayline fier with both feedforward ond | (Aber White ond Ruvin, IRE Nat. Cony, Re. vl. 5, pf. 2, 1957.) ‘The canonical configuration of a time-domain filter with both feedback and feedfor- ward is shown in Fig. 3.15. More usually, the canonical configuration is broken into sec- Hons with feedback and feedforward around individual delay lines. An example is the three- Pole Chebyshev filter of Fig. 3.16a. The frequency response of this recursive filter is shown in Fig. 3.16b, with 0.5 dB ripple in the passband.'? The width of the passband can be changed with different sets of weights. Figure 3.17, due to J. S. Shreve,'* compares the re- Sponse of a nonrecursive and recursive filter. It is seen that the recursive filter provides a frequency response that better resembles the rectangular shape than the nonrecursive, and does so with only two delay lines rather than the four of the nonrecursive filter. The recursive filter provides more flexibility in shaping the frequency response to Produce better response than the nonrecursive filter, and it does it with fewer delay lines. Unfortunately, its utility is limited by poor transient response because of the feedback. Large discrete clutter echoes, interference pulses from a nearby radar, deliberate pulse jamming, or the beginning of a dwell by a step-scan radar (a phased array radar, for ex- ample) can all produce transient ringing that can mask the target signal until the transient dies out. A frequency response with steep sides might allow 15 to 30 or more pulses to be generated at the filter output from a single input pulse because of the feedback." It Relative response 0 vr co) Figure 3.16 Shown in (a} is @ recursive delaysine filter whose frequency response in (b} is based on a threepole Chebyshev design with 0.5-dB ripple in the passband "(ter w. 0. White.) Scanned with CamScanner24 CHaprer 3 © MTI and Pulse Doppler Radar » ; Nowa er eo 4 8 2 of Tame fer 4 I En h ‘Frequency, Figo 3.17 Exomple of the frequency response of a recursive filter with two delay lines and 2 onrecursive filter with fovr delay lines. For the recursive fier the feedforward weights ore 1, ~2, i Grd 1 and its feedbock weights ore 1, 1.561, ond —0.641. The feedforward weights for the nonrecursive fiter are 1, 1.25, -4.75, 1.75, and 0.75. 1 (hoe JS. Shreve") rs has been suggested’ that the undesirable transient in a stepped-scan radar can be mit ; gated by using the first return from each new beam position to apply initial conditions . the MTI recursive filter to cancel or reduce the transient effects. The clutter retums Cat be approximated by a step-input equal in magnitude to the first return for each beam Pe sition. The steady-state values that would normally appear in the (digital) filter are co culated and loaded into the filter to suppress the transient response. Because of the poor transient response of recursive filters (infinite impulse resp" they are usually avoided in military radars that might be subject to deliberate electro ‘countermeasures. They have been widely used, however, to eliminate clutter in ci doppler weather radars. The Nexrad radar, for example, uses a five-pole (fifth-orde? lipti filer with its notch width under software control, and the Terminal Dopplet West! Radar uses a four-pole elliptic filter!” Recursive filters are used in this application "= cause of the requirements for sharp filter notch widths around zero doppler and # Ma sponse within the passband so thal accurate estimates can be made of weather ee ity and precipitation regartless of the doppler frequency. The elliptic filter is sad '° the advantage of narrower notch widths than any other filter of the same OPUCr es ‘Although the recursive filter seems attractive for MTI, its transient fae it of limited application. There are several other approaches to achieving a M™ dl Scanned with CamScanner2. Staggered Pulse Repetition Frequencies 125 The use of multiple staggered prfs will be discussed next, which increases the first blind ‘speed as well as provide limited filter shaping. This will be followed by the doppler fil- ter bank that provides capabilities not available withthe filters discussed previously. The multiple delay-line canceler, the N-pulse canceler, the transversal (nonrecursive) fil- ter, and the recursive filter are all examples of time-domain filters rather than frequency-do- main filters, The doppler filter bank in Sec. 3.4 is also considered as a time-domain filter. 3.3 STAGGERED PULSE REPETITION FREQUENCIES The use of multiple waveforms with different pulse repetition frequencies allows the de- tection of moving targets that would otherwise be eliminated with a constant-prf wave- form if their radial velocities were at, or in the vicinity of, a blind speed [as defined by Eq. (3.12)]. A simple illustration is shown in Fig. 3.18 which graphs the frequency re- sponse of a single delay-line canceler with two different pulse repetition frequencies (prfs). At prf fi, blind speeds (nulls) occur when the doppler frequency is f, or 2f, (and other in- teger multiples, which are not shown). With prf f = 2f,/3, blind speeds occur when the doppler frequency equals fs, 2fs, or 3f,. It can be seen in Fig. 3.18 that targets not de- tectable because of a blind speed in the frequency response of one prf will be detectable with the other prf. A target is lost on both pris, however, when the blind speeds occur si- multaneously, as when 2f, = 3f2. Thus the first blind speed at prf f, has been doubled in this simple example. The above illustrates the benefit of using more than one prf to re~ duce the effects of blind speeds; but it might be cautioned that it is not usual to use pris with the relatively large ratio of 3/2. There are several methods for employing multiple prfs to avoid losing target echoes due to blind speeds. The pris can be changed (1) scan to scan, (2) dwell to dwell, or (3) pulse to pulse (usually called a staggered prf). A dwell is the time on target, and is First blind speed ‘Amplitude fi 2h ° ra 2h 3h Frequency onse of two prfs fy (dash curve) and fr (solid curve), where fy = oe 38 ind ota of he fst pr re shown otf and 2h, and the Bind speeds ofthe second id 3h, A target ot the blind speeds of one prf are “covered” by the ie sae a when Fh, = 3h, where both prfs hove the same blind speed. Scanned with CamScannerNRO NPE EE 8 ULL ALR be Ug yibor Raat vty ay oanayeat Ueto ME ato Be a wea tt A vateam teat that HOF HY In havent (HSN, ALANS tthe poh apa e AAPA Rem HAE AL A eva tL wl) one We ak at Ae MR ERG be vata a voumdliate done Ue LOppler HHCT Hank as que Paws, HS: A NAN acannon ae availabe: Wl AboEay Hite CAMEOS, aN Way ANNE Ae pn AE AN. SOE BS WRGTOE MERE PROERSNING Nasa he haus Re HIAT AYN AE MELT tttor losign were fotiutated whew the de Jay Dao WAN AMMNRYS nAAtHE ABNEY, SoyiniowtAl MEEL Mopper tlters were diticlt ‘ayatoaeans a ana anatnbs, ao at vay tat fa MEPL lA to employ ote than te awh ANA HARON A AAMT tite Sateen the eqpal aloyetopyment oF dsital technology NSH FW AE ONE LPR, Howweneys alhowwovt the totays te abained By: storing dy eal WORMS AAW HAL AUtAFOANE Heap ame wets eoyitedl This prolly increased A RH! ANAT HOARD HAE HHA eONortN CUBED, Digital doppler filters with many ASA DWN ARE AHH AE HHAT oy A sahistioatos tltory eat he readily: obtained when a ase HARRY OE AKRON at ava HOE puywessinys, Thats the theoretical aspects Of MTL ARE ROE. Wee WeeHE AND, at ayvtatouttte CUIEIISHLY when analog delay lines were all A as av NRATN, WINE GAH DS apoE dis alta MMottOAES, Le aition to aking remeron the ANGE OE ATH soyatiowtet tters, the advantages oltered by digital MTL NONE NAN Cangas BRINE paves” which eative a Loss due to the difference in phase Davo the oth AAT ANNE the NETL rotenone sigma, This is achieved by we of saat Q jaowooanyg (A Nase ant qutteatare, someting that wars always KoA 10 be O8 valve BE MEPL pasoaaingy NE Which Was Mot caNvenient to implenent with an ep avettty Coyote tonne ange cat bee obtained than way possible with, acoustic ay lines © Vvwvaineat tangs athe tol times af aooustic delay’ tines: due to temperate tvages ae ctiommatet fy the aovurate timing at digital methods, 8 Thee ose pe UL aking the delay time in the digital memory synehrone® ' ‘aa thes unk pat soning lttioult to do with acoustic delay lines, # Tre Hovituty aatiorest BY aigital methods allows sagial processors to be really OF ‘atest wath ana ttoventtitter characteristics, Digital processors can be mule Jeanna » Daal NTL As vou stable ant retiable than analog MITA, and requires tess ab acnts aig ayer it the thet Aout agital prosesaing ches not have the serious weaknesses associater with anal atotay Twos HU Nas other charactorstics that have to De understood! if Cull ely ny Be akon af 1 capatalition foe sugaittoantly improving doppler processing!" at antr ___ _ Scanned with CamScanner35 Digital MTI Processing 137 Blind Phases, 1 Fig. 3.7 hada ae ‘The block diagram of an MTI radar that was shown in single processing shar etetor and filter channel. With a single phase detector and at the peak ponte = iere is a loss when the doppler-shifted signal is not sampled doppler signal and he negative values of the sinewave, When the phase between the blind phase is Gifeia pocae at the prf results in a loss, it is called a blind phase. A a blind speed occurs Neate’ blind speed discussed previously. It will be recalled that cycle at each sampling tinue sampling Pulse appears at the same point in the doppler ling time. as in Fig. 3.28a, Figure 3.28b illustrates the loss due to a Amplitude Figure 3:28 (0) Exomple ofo bind speed in on MM rodor. The lxgers dapper hequency ‘equal to the prf. (b) Exomple of the effect of o blind phase in the | channel, ond one a. channel. (d) The I channel of ancther speciol exomple of a blind speed. The pf is twice the Frequency and the phose ofthe sampling it such tht thre is no response ot all since te seek othe zeros ofthe dope equency. Nohing it deed (The @ chone! or the example of in which the sompling is at the positive and negotive peaks of the doppler fr complete recovery of the signal. Pee iinet Scanned with CamScanner— ‘Nant phavo, ‘he sampled signals in this partioular example ae iu ne sume aunt wit at spot auth tt when pe ay aubtenete tra pub LINEN ‘Whoo pute fs subg tea o tavover there (8 LHe PAE Dt When ayn Aeuetoa thom asthe rot 6 gots, anal goon Thy fi Ahlspattewlar cHse, Ha atte Wal enemy is lost The ather hall af the signal emery ean be recovered IEW second Wen, Aeal procorsing channel ts ued and there by 090" phave change of the coho, (eletencey pal that by appli 4 tty phase detector, Uhls xeconal channel iy culled the Q, ar gugy Fane, channel, ‘The original chanel ty culled the For fe pluase chute 1 the cay 9 tral. the Fohanel ty si 2 mfg te the colo tw the Q-chunnel by eo 2 ft The result a he YO" phave change tn the Q channel ty shown in Hp, 2K, Those pulse pairs that hud tery ‘wutput in the Fehannel now have a finite rosie in the Q channel, Likewise thase rates whet hal a thnite residue tn the J chanel now have sere output In the Q ¢ Whrat way toxt tn the J channel iy weoverod in the Q channel, and view versa, Th bination of the Zand Q channels results ina uniforn output with no loss, The example of Fg. INH andl e iy a spectal ewse, Another very special care of g i Mint phase is when the prt iv twiee the doppler frequency (Ubis iy Hot w lind speed) an the phase ot the prt is stich that the samples ovcur whenever the doppler signal in the 1 schanioe! passes thnagty agora erussing, Ply. 1.28, There ts no output in this ease, Inthe Q chanel, Hig, L2Ne, where the phase of the prt is shifted 90°, the doppler signal is sun. bled whenever it iy at w positive oF a negative peak, ‘The muxinun signal is obtained, ' ‘Again what way lost in one channel is recovered in the other channel, NOR CHAPTER 8 MITE and Pulse Doppler Radar Cons Dihe nel com. pen Block Diagram The iock diagram of digital MTL signal processor with J and Q chan- i trols is shown in Fig. 4.29, The signal trom the IF amplitier is split inte two channels, The } Phase detectors in each channel extract the dopplesshitted signal, In the channel the Hi doppler signal iy represented as sly cos (2 fit + dy) and inthe Q channel it is the » Pi except that sine replaces the eosine, The signals are then digitized by the analog-4e r digital (A/D) converter, A sample-and-hold eireuit usually ix needed ahead of the AD converter for mony effective digitizing, Sample and hold is often on the sane chip asthe AMD converter, (Some A/D converters, such isthe Mash type, do not require a sample and bok hold.) The digital words are stored in digital memory for the required delay time(s) ; and are processed with a suitable algorithm to provide the desired dappler filtering. The i ti magnitude of the doppler signal is obtained by taking the square root af 12+ Q*. Some: ha! Uimes, (or simplicity, the sum of the magnitudes of the two channels, [7] + lof is taken, [AD Digital | Digital converter "| memory | proves Figure 3.29 Block diagram of o digital MTI | doppler signal processor From W amplifier Scanned with CamScanner45 Digital MTI Processing 139 or the “greater oF” of the two channels might be used instead. ‘The Land Q processor of Fig. 3.29 hay « square-taw detector characteristic. A lineurslaw detector can be approxt= mated by the greater ot [1] 4 Ll/2 or Lal + 17172, W required, the combined unipolar output can be converted to an analog signal by a digital-to-analog (D/A) converter for display. Otherwise the digital output might be sub- ject to further processing, There are several methods for impl 1g an A/D converter, depending ‘on the speed nd number of bits required, Since there are (wo channels (the / and the Q), sampling in each can be at one-half the Nyquist rate, which generally makes the implementation of the A/D converters simpler, (The Nyquist rate is twice the signal bandwidth.) The num- ber of quantization levels in the A/D is generally given as 2". D/A converters are usually easier to achieve than A/D converters of the same resolution, ‘The available technology of A/D converters has sometimes been a serious limitation to MTI performance, but there have been continual advances in this technology. A 16-bit A/D, for example, might have a sampling rate of several MHz; a 14-bit A/D, a 50-MHz sampling rate; a 12-bit AD, a 100-MHz sampling rate; and an 8-bit A/D, a 1-GHs sam- pling rate ‘The output of the IF amplifier is usually made to limit at a level consistent with the MTI improvement factor (as in Sec. 3.7) and the full-scale range of the A/D converter. The IF portion of the receiver is then a linear-limiting amplifier. The limiting action means that the peak excursion of the signal that the A/D converter must handle is known, and the A/D can be designed to cover this excursion. The signal should not be allowed to ex- ceed the full-scale range of the A/D converter, since the output would then be degraded and severe harmonics generated. ** Limitation on the Improvement Factor Due to the A/D Converter All analog signals that lie within the same quantization step of the A/D converter are represented by the same digital word. Since the rms value of the noise accompanying the signal is usually greater than the quantization step of the A/D converter, the digital word can change slightly from pulse to pulse in a noiselike manner. Thus the quantization of the analog signal results in noise or uncertainty, called quantization noise, which can limit the MTI improvement factor. Shrader and Gregers-Hansen’® give the limitation to the improvement factor due to quantization noise as = 20 log (2% — 1)V0.75] (dB) (3.35) where N = number of bits. This is approximately 6 dB per bit. (Each bit represents a fac- tor of two in amplitude resolution.) Thus a 10-bit A/D theoretically limits the improve- ment factor to about 60 dB. In practice, the A/D converter generally requires one or more additional bits to achieve the desired performance. The limitation on the improvement factor based on Eq. (3.35) assumed that the quan- tization error was independent from pulse to pulse. This happens when the rms noise is greater than a quantization step. It was suggested by Glen Preston that there will be a qui- ting effect that reduces the quantization noise if the pulse-to-pulse clutter samples are correlated rather than independent, Brennan and Reed” derived an expression for the Scanned with CamScannerAO CHAPTER 3 © MTVand Pulse Doppler Raclar quieting that is a thnetion of the correlation, and they found quieting tbe smut itoccurs the quantization sep is lante compared 10 the rms Noise, a vontion thar preferred tor best performance in a practical nidar system, Dynamic Range The dynamic range is the maximums signal-to Molse rai Hat Cn be hy, Aled by an AMD converter without saturation, The noise level relative tthe gant Sep affects the dynamic range, The available dynamic range (power ratio) is given ge Vk where N= number of bits in the A/D converter (the sign bit is included), and & noise level dived by the quantization fevel, The larger Ais, the Less the aya range A value of & fess than one means that the receiver nolse at the input t9 the AMD convener Ws Tess than the quantization noise, which results in a Toss af detectability. Generally 4 i equals 1 or 2, With A = 2 (4 value recommended by Shradk dynamic range = 2 ty) my and Gregers-Hlansen), the dynamic range for a 10 bit A/D according to Ey, (3.46) is 48.2 dB, which is considerably tess than the 60 dB improvement factor based on at value of 6 dB per bit IF the linitation Fa) ‘on the improvement factor is taken as 2°* (6 dB per bid), the dynamic range given by Fy, BPs (3.20) when k= 2 is seen to be about 1S dB less than the improvenient factor, This the limitation on the improvement factor might be determined by the dynamic range of Ea (3.40) rather than by the quantization noise lintitation ustually given as 6 Db per bit (Thi could be the reason why, in pructice, one oF more bits ane added to the AD convener when the improvement factor limitation is taken as 6 dB per bit) Other Limitations The above huts, for the most part, assumed ideal A/D. converters and ideal and Q detection. There are several practical conditions that need to be considered when good MTL performance is required, Errors and redticed performance ean be die to:'*"* (1) other than a 90" phase difference between the Hand Q reference signals,” 2) gain and phase imbatinee in the wo channels, (3) tinting jitter in the sample-and:hold circuits, (4) nonlinearity in the A/D." and (5) range straddling toss due to the sumpling. ot being at the peak of the output of the matched filter, The degradation in pertormanee caused by the inabil @ channels in phase, amplitude, and frequeney spectrum ean be significantly rede Oy employing a single channel whose signal is down converted (fram TE ar RE. by mixin) to obtain the in-phase and quadrature channels, There have been several methods pi Posed in the past to perfor this form of findpass sampling, also sonnetitnes c ital down conversion, \ technique described by Rader" for a baandfimited signal of ban width & down-converts (mixes) the IF oF RE signal to a center tiequeney of B. The siz! is sampled at a rate of 4A Two Mlters are realized ox a pair of A" phase spliving ae works with several symmetries which are used to reduce the computations, The in-phet and quadrature samples are obtained at an OuLpUt rate of BL Thus been suid’ that Ih oversampling by a factor of 2 trom the Nyquist rate (28) results in a simple and effort implementation of the filters, Only one mixer and one A/D converter are ned NM sampling iy four times faster, J another method for baseband sampling.!* sar the Nya rate are interpolated, based on a number of store pase a y to precisely mateh the Fand UF samples: ken ne “samples, to provide the Scanned with CamScanner3.6 Moving Target Detector 141 quadrature channels. Experimentally it was found that the phuse errors could he reduced by an order of magnitude compared with conventional busebund /,Q processing. Fast Fourier Transform (FFT) Digital filtering involves the use of the Fourier transform. The fast Fourier transform requires less computational effort, and it has been popular for many applications. It has some limitations, however, compared to the classic Fourier trans- form. The number of samples used has to be expressed as 2%, If a filter bank is being xen- erated, all filters have identical responses, they will be uniformly spaced in frequency, and the weighting coefficients are not optimum since they cannot be chosen independently for each filter, The filters possible with a non-FFT filter bank also cun achieve greater atten- uation of moving clutter (such as rain or chaff) because of the greater flexibility available in their design, There are times, therefore, when the classical Fourier transform may be more advantageous than the FFT even though the FFT might be quicker and require less complexity. Loss Due to Blind Phases in Using @ Single Channet? With only a single channel, the ef- fect of a blind phase can be anything from complete loss of the signal to no loss at all, as was indicated by the special cases mentioned earlier in this section in the discussion of the /,@ processor. With a nonfluctuating target and only a single channel there is a loss. of 2.8 dB and 13.7 dB for a probability of detection of 0.5 and 0.9, respectively, and a probability of false alarm of 10-*.* When the target fluctuates pulse to pulse with a Rayleigh probability density function (Swerling Case 2) and with postdetection integra- tion of from 2 to 8 pulses, the loss in using only a single channel is about 2.4 dB when 0.5 < Py <0.9 and 107° < Py, < 107". For a very large number of pulses integrated (greater than 1000), the loss approaches 1.5 dB. When the target fluctuations are described by Swerling Case 1, there is a loss of 3.2 dB in using a single channel with postdetection integration of only two pulses.*® The loss is about 2 dB when more than 20 pulses are integrated. It might be concluded that these losses are moderate, so that a single channel might be acceptable when it is important to reduce hardware complexity. This was the conclu- sion reached in the early days of MTI radar (the 1950s and 1960s) when the delay lines were analog. Hardware complexity, however, is seldom a consideration with digital MTI processing, so that both the / and Q channels are always employed. A block diagram of a digital processor, therefore, almost always shows / and Q processing. 3.6 MOVING TARGET DETECTOR The Moving Target Detector (MTD) is an example of an MTI processing system that takes advantage of the various capabilities offered by digital techniques to produce improved detection of moving targets in clutter. It was originally developed by the MIT Lincoln Laboratory for the airport-surveillance radar (ASR), a 60-nmi radar found at major air- ports for control of local air traffic. The introduction of the MTD represented an innova- tive and significant advance in radar detection of aircraft in the presence of clutter. Scanned with CamScanner142 CHAPTER 3 @ MTI and Pulse Doppler Radar Figure 3.30 Block diagram of the original Moving Target Detector (MTD) signal processor. Original MTD‘”~“? The original MTD concept was designed for a radar simitar up FAA's ASR-8, which operated at § band (2.7-2.9 GHz) with a pulse width of 06 1.35° azimuth beamwidth, antenna rotation rate of 12.8 rpm, average prf of 1040 Hp, a an average power of 875 W. The ASR-8 employed four staggered prfs; but staggering og the pris was not used with the MTD. The original MTD included the following: + Aneight-pulse FFT digital filter bank with eight filters, preceded by a three-puly delay-line canceler. The three-pulse canceler reduced the dynamic range of the sg. nals which the doppler filter bank had to handle, and it compensated for the lack of adequate cancellation of stationary clutter in the doppler filters. The doppler fier bank separated moving targets from moving weather clutter if they appeared in git ferent doppler filters. + Frequency-domain weighting to reduce the doppler-filter sidelobes for better cluter attenuation, © Alternate prfs to eliminate blind speeds and to unmask aircraft echoes from weather clutter. + Adaptive thresholds to take advantage of the nonuniform nature of clutter. + Clutter map to detect crossing targets with zero radial velocity that would otherwise be canceled by an ordinary MTI. + Centroiding of multiple reports from the same target for more accurate location measurements. The range coverage of this processor totaled 47.5 nmi. The MTI processor was preceded by a large dynamic range receiver to avoid the re duction in improvement factor caused by limiting (Sec. 3.7). The output of the receiver IF amplifier was fed to / and Q phase detectors. From there the A/D converters changed the analog signals to 10-bit digital words. Figure 3.30 is a block diagram of the MTD. but further explanation is probably required before the block diagram becomes clear Coherent Processing Interval (CPI) The range was quantized into +-nmi intervals, which were approximately equal to the range resolution of the pulse, The azimuth angle 10 3-pule S-pulse doppler ‘Weighting canceler filter bank and magnitude jE revo : ‘Clutter ma Zero velocity filter A Memory (clutter map) a ~ Scanned with CamScanner16 Moving Tiryet Detector 149 Wa tutte Lute degree tntervals, which were abt one-hulf-benmwidth, Thus there Weve 1o81 6 469,000 cange:azinuth vexolutian cells within the 47.4-tmni range tres Ii evel F dogroe uctiiuth cell there were 10 buen (rmmnitied nt we constant prt, On: revel, thers ten pulws were called w cilicvend processing interval (C1), Vbe ten pulses ‘ofthe CPL were prow wt by an eight pulse filter bank thit wine preceded by a three pulve delay line eaneeter, The titer bunk prontaccd eight conta, filters, I here were then 2,920,000, Humgecazinith doppler resolution cells, Hach of these cells had its own adap tWe threshold (clerertho! ler} whlch wae controlled by the amount tf elulter echo seen An the vieinity of the target, ti the next {degree azimuth cell, the prt of the 10 pulse CPL ‘Was changed to eliminate blind specu and to anmusk mov Wy largety hidden by moving weather clutter (us will be described later), Changing the prl every 10 pulses (every hall beamwidth) eliminated second-time-around clutter echoes that would normally degrade a conventional MTT using pulye-to-pulse stagger of the prf wer Filter Bank ‘The filter bank was implemented by a HET Although there wer in the CPL, there are only eight doppler filters. Since the three-pulse cunceler requires all three pulses before it can cancel clutter, the first two output pulses were discarded, ‘The discarded pulses are culled fill pulsey since they are needed 10 “fill” the canceler before cancellation of the elutter ean begin, ‘The frequency responye of euch filter of the FFT filter bunk hud a (sin xy/x shape with the first sidelobes 13.2 dB down from the peuk, These sidelobes were high and generally did not have sutficient rejection of nearby targets or clutter, The nidelobes were reduced in this MTD by subtracting from the output of exch filer | the sum of the two adjacent filters.’ [If A, B, und C represent the unweighted outputs of three contiguous filters, the weighted output applied to filter B is B = (A/4 + C74) Ideally the three-pulse cunceler is not needed if the filter bank can be designed to ob- tain the necessury clutter attenuation for both stationary und moving clutter. It was not Possible, however, to eliminate the three-pulse canceler inthe original MTD. Inthe block diagram of Fig, 3.30, the magnitude is the operation (2? + Q)'”, or an approximation thereto, 10 pulses Clutter Map & conventional MT! processor eliminates stationary clutter, but it also climinates aircraft moving on a crossing trajectory (one perpendicular to the radar line of sight) which causes the aircraft's radial velocity to be zero, This is unfortunate since the radar cross section of an aircraft is relatively large when viewed at the broadside aspect presented by u crossing trajectory. The MTD took advantage of this large cross section to detect targets that normally would be lost to a simple MT! radar. It did this with the aid of a clutter map that stored the magnitude of the clutter echoes in a digital memory, The clutter map established the thresholds used for detecting those aircraft targets which pro- a ea arete canceler removed all echoes with zero velocity, the zero- velocity filter had to be re-established in order to produce the clutter map, as indicated in Fig. 3.30, Each of the 365,000 range-uzimuth resolution cells of the clutter map stored the average value of the output of the zero-velocity filter received during the last eight soars (2. ‘On each scan, one-eighth of the output of the zero-velocity filter was added Scanned with CamScannerWARE 8) NEP ant Pulao Hayter Nala {ws ovo ott atte gata ate te Ae anys Ne nt HET We DHE AUST TOOTS Awnuhor avt xvawe wats ave ovnmy ate that aBenATE He HIR WONT A Het Ha taal AVY AteLafay ib ee Mth one Rng FA ate enon FH He eee OE Py Ag AUtiony vA, At ett apy cfhmneAT ocenltpy Hae valve HT Tey ng wy \wistout 19. ah ayanypt tte siatnt tee esta He Hiwatt wets allen the i AHO AE OMA Lito ANHHTY gore datlaT whiny, Phe MEEEY jvtaoved Hibhily: xtc aneful jp Wty this toctionrquie tee foie ceaoentg tuyere thiat welll five ww repec bol Wy a almpley AEE wont Avkptive Povahohis Ax nventioned, each af the 3,020,000 hinge: a¢tnautte doppler re what cots hat as ataptive dhivabokE fo altos ogee tanh AvP MnVTiN! Gntgete er ae ARO OF MONK CHAO, Cometdor thes aight toppter tere ciageraminieel at the tap at Pig AAV dintowtat wy ptt LY (Hie th anion cvmieeittate sunty ane is ane gi Aiton andl anne the twKE OL the Aan whiet WHE he alactiaasel fi Hie neat aubasetion) Vow aw thio atitownt vritonin fae wetocting the thealiolds, depending on dhe location oF the titer Witty WApwEE HY Aer vetowlty” The mataptive threstokt wetting for filter no. | Gwhieh wire the Alter cwntered al an twaltal vetoetty) way fotornntid By what wie staeed the etter naps Fl ave tno hts far Mion now Eiheogh 7 were aot hy wetter CVARC stunt te whit [a deweribed AW See 8 Panel vhowa We Ny 7 Mile tly, the clation CVARC in the MED eatablialod & threshokt Based on the mean level ot the wignals meetved Hi the 16 range colle centered annanal the tango cell of tntoreat The Lo range wells cateapeane ava elbataaned ane fall ale obit ind one halt nities ahead at the unio celal titereat, Phe mien value ob the ap Wale tn these 16 cells, when multiplied by ai appraprlate veamgtat, eatabllebont thie kL at edt at the monger veloety Alters thiengh 7. The the thavelokd vont awtaptodt tothe oval en feommient a the tala ptt the roving: two tMlters, 2 aint X, avttel He aljawent to the ger velinetty tillog he Heth abl was selected as te tanger a hat given by the eater aap al tig elittor CHAR Lage cecties Hut nov tigt chatter, whieh mab be eontived a real target veace fy a che Moma MIL with a single Alter covering the dyplen apne, ate tuat detwetedd tn the MUD tuevolead tn tlie thranyhs apie Mt Unmwesking Moving Uurgets in Moving Clutter — Aw-satvantaye ot w titer tank te tal allows mowing tantets te be veparated thon moving etutter (acl ay raed av that el Hon can take pee, Since the sloppter shifted radar eet foam a novi. target all {Hout not always) hae a tallal vetoetty gevater than the magia anatiinet ¥ell (or Bist blind speed tte toppler Reueney can fol aver int an of the ight titre of the titer wank, This bs shown Happening with peE-2 OF Fy. NL Te te tango! vob {e represented by the solld vertteat at Lying. 16 the rtp at the ms tiint uc five velit, In this ease I fohly aver tte AMEE HH, we nllnteat ny te st 0% ‘Tho eal ects ty abe in Milter Hs ence, Hatake thie vets ere teat tt € wih pet 2. On pet ts however, the alroratt ech folds inter a lifer tite, ht Uhl Alter ws. 7, The ein ecths comalnn in Mer we, H altco lin tatial vohociny tm toos af @ axtnum unambigwns velocity. tn this manner the change tn prt amd the ue a 10 Alter bank allow moving atrera echoon ta unseat in leust one CH that bE Scanned with CamScanner3.6 Moving Target Detector 48 Fiternumber S647 8 1 2 a 4 5 irra aliases Tre 7 aircraft —| Seer es ee = Radia velocity Figure 3.31 Detection of an cirerat in rain by use of two pris and a doppler filer bank, thawing tha yrmasing of he cicras echo on pr de to doppler ntdover when I masked rain on prh2. 1 (From Mucho,*® Copytigh 1974 IEEE) rain echoes. A change in prf of ubout 20 percent was used in the MTD to allow this un- masking of the aircraft echo. (The echo at S band from precipitation might typically have a spectral width of about 25 to 30 kt; and might be centered anywhere from —60 to +60 kt, depending on the wind conditions and the antenna pointing direction.*!) Interference and Saturation Pulses from a nearby radar might be received by the radar and appear as a large signal among the 10 pulses normally processed as a CPI. In the MTD an interference eliminator compares the magnitude of exch of the 10 pulses against the average of the 10. If any pulse is greater than five times the average, all information from that range-azimuth cell is discarded. A saturation detector detects whether any of the 10 pulses within a processing interval saturates the A/D converter and, if it does, the entire 10 pulses of that particular CPI are discarded. Post-processor Centroiding The output of the MTD processor was a hit report which contained the azimuth, range, and amplitude of the target echo as well as the filter num- ber and prf. On a particular scan, a large aircraft echo might be reported from more than ‘one doppler filter, from several coherent processing intervals, and from adjacent range cells. As many as 20 hit reports might be generated by a single large target.*? A post- processor grouped together all reports that appear to originate from the same target and interpolated to estimate the target's azimuth, range, amplitude, and radial velocity, The target amplitude and doppler were used as the basis to eliminate small cross-section and low-speed echoes (such as birds and insects) before the target reports were delivered to the automatic tracking computer. The automatic tracker also eliminated false reports that did not form logical tracks. The output of the automatic tracker is what is displayed to the operator. Since the MTD processor eliminated a large amount of the clutter and had a low false-detection rate, its output could be reliably transmitted to a different location, if desired, by narrow-bandwidth telephone lines. Scanned with CamScanner46 CHAPTER 3 © MTI and Pulse Doppler Radar Performance The improvement factor of the original MTD measured om an i surveillance radar (ASR) was about 45 dB, which was said to be a 20-d8 increase g the ASR's conventional three-pulse MTI processor with a limiting IF amplifier. n agg. tion, the MTD achieved a narrower clutter rejection notch at zero velocity and at the bling speeds, The MTD has proven to be an important mode! for the design of radars for de. tecting moving targets in stationary as well as moving clutter. Second Generation MID®? The original MTD, as does any electronic development, useq the technology that was available when it was first developed. Advances in digital hard. ware occurred rapidly, however, and an improved version soon appeared. The origina hard-wired processor was replaced by a parallel microprogrammed processor (PMP) which ‘operated out to the full instrumented 60 nmi range of the radar. The characteristics ofthe second generation MTD are summarized as follows: * The filter bank was implemented as a generalized transversal (FIR) filter rather than as an FFT so as to provide more flexibility in design and reduce the doppler-fiter sidelobes. * A two-pulse delay-line canceler (rather than the three-pulse canceler of the original) preceded the filter bank. © The zero radial-velocity filter was also designed as a transversal filter. It utilized the Chebyshev criterion to produce relatively uniform filter gain across the portion of the doppler space not covered by the nonzero-doppler filters. © The clutter map had one cell for each range-azimuth cell rather than one cell for each range-CPI. * The nonzero-doppler filters were based on the method of DeLong and Hoffstetter (mentioned in ref. 53). They provided sidelobe levels 10 dB lower than the original MTD, and thus gave better performance in rain. When the two-pulse delay line pre- ceded the filter bank, the number of bits required in the transversal filter ‘so that the filter weights only had to be 3 or 4 bits plus sign. «The CPI consisted of eight pulses. One CPI had a 900-ys period, and the other bad a period of 1100 bs. The prfs were 1111 Hz and 909 Hz. © Correlation and interpolation algorithms clustered the threshold crossings in fan8* ‘azimuth, and doppler to provide in a single report the best values of the radar 0 servables of a single target, whether it be aircraft, automobile, or bird. Target sigh) strength was one of the observables as well as range, angle, and doppler. The interpolation was one part in 64 across the band of eight doppler cells. «An area CFAR was used after the MTD thresholding to eliminate single CP! (°° is, interference, and weather clutter that were not remo’ signal processing algorithms. The scan-to-scan correlator was an automatic tracker that deleted those tages scan-to-scan behavior was not what would be expected of an aircraft. This low-speed targets as well as echoes that lacked spatial correlation from sca” Wy ‘The scan-to-scan correlator determined when a track was of a quality sufficient Jow the data associated with it to be displayed. Scanned with CamScanner36 Moving Target Detector 147 Experimental demonstration of the second-generation MTD installed on an ASR-7 at MIT Lincoln Laboratory showed that the PMP processor produced approximately 500 to 600 range-azimuth-doppler threshold crossings per scan out of a total of about 3,000,000 cells within the radar coverage. There were typically 5 to 15 threshold cross- ings per aircraft target. The FFT of the original MTD constrained the type of filter characteristic that could be obtained. Performance was sacrificed for reduced cost. By employing FIR filters in the second generation MTD, the frequency response of each filter of the filter bank could be shaped as desired. Less power was wasted by using the eight-pulse FIR filters since only one fill pulse was required rather than the (wo fill pulses of the 10-pulse FFT. The re- duced number of pulses allowed better performance with a shorter observation time. Both the first and second generation MTDs were developed by the MIT Lincoln Lab- oratory. The MTD concept further evolved into a third configuration used in the ASR-9 airport-surveillance radar developed for the FAA by Westinghouse (later known as Northrop-Grumman) at Baltimore, MD. Third Generation MTD for the ASR-9° The first operational air-traffic control imple- mentation of the MTD concept was for the ASR-9, a radar employed at the major airports of the United States and elsewhere in the world. It utilized the basic philosophy of the second generation Lincoln Laboratory MTD and significant extensions by Westinghouse. The pulse periods of the two CPls in the ASR-9 were not equal, but were in the ra- tio of 9/7. The longer-pulse-period CPI processed 8 pulses and the shorter-period CPI processed 10 pulses, for a total of 18 pulses. The different number of pulses allowed the doppler space covered by the two CPIs to be more equal than the unequal coverage of the two CPls of the original MTD (whose filters were sketched in Fig. 3.31). There were ap- proximately 21 pulses received over the radar’s half-power beamwidth, but only 18 of these were used for establishing the clutter map that helped to contro! the false-alarm rate. ‘Twelve-bit filter weights were used rather than the five-bit weights of the Lincoln MTD. This enabled the filter to be designed to achieve 40-B doppler sidelobes for rain rejec- tion and 44-dB ground-clutter rejection (relative to noise) for the heavy clutter filters. In mountain clutter, an alternate filter design was used that provided 52 dB of attenuation at the expense of less rain-clutter rejection. Sensitivity time control (STC) was employed to reduce the large clutter echoes that exceeded the available dynamic range of the system. ‘The range and azimuth resolutions of this radar were improved compared to the orig- inal MTD. In a simple CFAR, the threshold increases when there are two nearby targets, thus reducing the probability of resolving them in range. The ASR.9 CFAR algorithm, however, omitted the strongest cluster of samples from the estimate of interference (that determined the threshold) so that a neighboring target of comparable echo strength didn’t cause the threshold to be raised. Without this cluster editing, the two targets would ‘ot be resolved and only the larger target would be detected. When detections were obtained that extended for more than two beamwidths in az~ imuth, the shape of the response in azimuth was compared to what would be expected froma single target. A significant deviation from the expected response indicated that the extended azimuth signature was caused by two targets, thus increasing the probability of resolution, Scanned with CamScanner; AB CHAPTER 9 © MTT and Pulse Doppler Radar Accuracy with the MTD" ‘The accuracy of the original MTD was 101 1 ood be obtained with a conventional scunning radar, The angle in the original My a termined on the basiv of the two CPhs rather than ws Is unuully done In w scanning nae that uses ull the pulsex received from u target to estimate Its center by bean sping In the Westinghouse, or third xenerution, version a typleul wireralt night yive ring y 35 individual threshold erownings (which were culled primitive reports) in rane, acinyy, doppler, and amplitude, A large ulrcraft might produce 10) much reports, ‘These have y be centroided to produce w single report of range, azimuth, doppler, and ninplitude, ‘thy radar must also provide an accurate measurement of angle and be able W resolve ty ' nearby turgets. Algorithms were developed 10 rerolve unequal size targets separated in yy, imuth by less than two beumvwidihy and separated in range by fess than tw» range cell, which was | nmi.* Since a lurge (urget might affect up to five range cells, range resoly. tion in the MTD hud (o be obtained by using « simplified form of pulve-shape matching For azimuth resolution and estimate of angle, one of four algorithms wan used, depend. ng on the extent of the datu In angle. The most eluborute of the four employed hearmshape matching when target primitive reports were obtained on three or more CPI with the same prf. The typical number of primitive reports with this algorithm was greater than 30. The rms angle accuracy was suid to be ubout 0.04 beamwidth. lt When the réceived signal was so strong that it saturated the A/D converter, the a curacy of centroiding and the ability to resolve two targets were degraded. In the first two ' versions of the MTD, any CPI that experienced saturation by a large target or interfer cence signal was eliminated. {t was found, however, that eliminating the CPI cauved ypli ting of the target due to premature termination of u report, The Westinghouse MT avoided this problem by providing information about its presence (rather than nothing) when the data saturated. | Hit, oF Primitive, Reports In the above discussion of the various evolutions of the MTD, there were various values given for the number of hit reports, of primitive reports, gem crated by a single aircraft target. According to the references cited in this section, these varied from 15 to 100. Apparently the number depends on the type of MTD, the radar with which it was used, and the nuture of the trials that were conducted. Extension to Other Radars The three generations of MTD systems described above Were well suited for use in an S-band medium-range airport-surveillance radar. The MTD has also been adapted for use with a solid-state transmitter, as in the ASR-12 air-traffic com trol radar.’” Although the individual techniques that constitute the MTD have applicabit ity to other types of air-surveillance radars, it is more difficult to apply in its entirely MTD systems as described above to long-range radars or those at other frequencies. Som™ features of the MTD concept described here might not be suitable for military ait surveillance radar applications since they might make the radar more vulnerable to cour termeasures. “itis generally accepted that a good radar can revolve two targets of equol size if ls ore saporoed Saige Si Eocmmrah. nr hw gc tos eee ae oan of Peas 0 Bre laos! 0.8 pulse width opart Scanned with CamScanner27 Limitations to MT Performance 14 — x LIMITATIONS TO MTL PERFORMAN In this section we consider the degradati antenna scanning modulation, tie in performance of MTI radars caused by (1) 2) internal fluctuations of clutter, (4) equipment instabtly nd (4) limiting, ‘The adverse effects of the A/D converter and other aspects of digi tal processing on MT! performance were considered in Sec. 3.5 : The limitations to MTT performance to be discussed cause the clutter spectrum to widen, More clutter energy is then passed by the doppler filter, which lowers the 1m- Provement factor. If the clutter power spectral density can be expressed as a gaussian func tion with a standard deviation 7, in Hz, it can be represented as w= Woerp(—L. ) (371 where Wy is the peak value of the clutter power spectral density at f= 0. This equation was given previously as Eq, (3.14). The clutter standard deviation is sometimes written in terms of the radial velocity, in units of metervsecond, and denoted v,. The two mary be related by v7, = 2.4,/A, where A = wavelength in meters, (The standard deviation €x- pressed in velocity, «%, is sometimes, but not always, independent of frequency for wind- blown clutter.) If the MTI filter has a frequency response H(/), the clutter attenuation is {a repeat of Eq. (3.15)] [wor [mnlanl a CA= (3.38) “The improvement factor. Eq. (3.20), is found by multiplying the cluter attenuation by the average gain of the filter. ; ‘The improvement factors for a single delay-line canceler and a double delay-line can- cl is Wed by a gaussian spec- celer were derived in Sec. 3.2 when the clutter spectrum is represen trum. The general expression forthe improvement factor with m delay ine cancelers in cascade was given by Eq. (3.23) as y where fy isthe pulse repetition frequency. of Pt. We shall use this expression to deter- c sian. i i rement factor when the clutter spectrum is gaus mere Pre N different effects that contribute vo the widening of the clutter spectrum, deviation and if each is gaussian and independent of one another, the overall standard is Beet on" rv) ; cts, such as oscillator stability, quantiza- rhere the subscripts indicate the various effects, wich ms OucNN fey INCE tion of the A/D converter, antenna scanning a, =(oi+ 0340 Scanned with CamScanner