ARTM CHANNEL SOUNDING RESULTS AN INVESTIGATION OF FREQUENCY SELECTIVE FADING ON AERONAUTICAL TELEMETRY CHANNELS

Transcription

1 ARTM CHANNEL SOUNDING RESULTS AN INVESTIGATION OF FREQUENCY SELECTIVE FADING ON AERONAUTICAL TELEMETRY CHANNELS Item Type text; Proceeings Authors Rice, Michael; e Gaston, Davi; Davis, Aam; German, Gus; Bettwieser, Christian Publisher International Founation for Telemetering Journal International Telemetering Conference Proceeings Rights Copyright International Founation for Telemetering Downloa ate 5/6/8 9:34:3 Link to Item

2 ARTM CHANNEL SOUNDING RESULTS AN INVESTIGATION OF FREQUENCY SELECTIVE FADING ON AERONAUTICAL TELEMETRY CHANNELS Michael Rice, Davi e Gaston, Aam Davis, Gus German, Christian Bettwieser Department of Electrical an Computer Engineering Brigham Young University ABSTRACT Initial results of wieban channel souning experiments sponsore by the Avance Range Telemetry (ARTM) program are presente. Data collecte at Ewars EAFB uring the Winter , are analyze in the frequency omain to estimate the number, strength, an elays of the significant multipath reflections observe uring the experiments. We observe that the channel is aequately moele using two or three multipath reflections. The multipath fae events are correlate with recore bit error rates an transmitter location to provie a comprehensive overview of the channel characteristics. Summaries from two test flights are inclue where it is seen that the - an 3-ray channel moels provie excellent moels for the ata. In general, the 3-ray moel captures the essential features of the multipath interference. In this moel the first multipath is a strong specular reflection with relative amplitue greater than.5 an relative elay in the 3 to 7 ns range. The secon multipath is a much weaker reflection with relative amplitue less than.5 an relative elay in the 75 to 35 ns range. KEY WORDS Channel Souning, Channel Characterization, Multipath Faing, ARTM. INTRODUCTION Multipath interference occurs when reflecte replicas of the transmitte wave front arrive within the main lobe of the receive antenna gain pattern []. The ranom phasing of these reflections causes ranom perios of constructive an estructive interference. When the estructive interference reuces the carrier-to-noise ratio at the receiver front en to an unacceptably low level, a multipath fae event occurs which prouces what is often calle a ata outage or signal outage []. Mitigation techniques for faing inclue space iversity, antenna iversity [], interleave error control coing [3], an equalization [3]. The effectiveness of these

3 techniques epens on certain faing characteristics such as the number, strength an time elays of these reflections. One of the tasks of the Avance Range Telemetry (ARTM) Program is to investigate the properties of multipath faing in common aeronautical telemetry environments. This paper reports on the initial results from two ARTM channel souning flights at Ewars AFB in December 998 an February 999. x(t) h(t) y(t) Y(f) = H(f)X(f) Figure : Linear time-invariant system representation of the RF multipath environment. CHANNEL MODELS AND SOUNDING TECHNIQUES We begin by noting that real channels are time-varying. The channel impulse response is a function of the physical geometry involving the airborne transmitter, the receiver, an the reflection points. Since this geometry varies uring the mission, the channel impulse response varies uring the mission. The characterization of these changes or channel ynamics, is treate in [4]. We assume that over a short enough time interval, the channel oes not change an is time-invariant. Thus, uring a sufficiently short interval of time, we moel the aeronautical telemetry channel as a linear, time-invariant system with impulse response h(t) an transfer function H(f) as illustrate in Figure. When a signal x(t) is input to the system, the channel output y(t) is given by the convolution of x(t) with h(t). In the frequency omain, the input-output relationship is Y(f) = H(f)X(f). Observations of receive signal properties at Point Mugu NAWC [5,6], White Sans Missile Range [7], Ewars AFB [8,9], an China Lake NAWC [] suggest that a goo moel for the aeronautical telemetry channel is one consisting of a line-of-sight signal accompanie by a single strong specular reflection whose elay relative to the line-ofsight signal is some fraction of the bit time. Such a channel is moele mathematically as j ( πf τ + θ ) h( t) = δ ( t) + Γe δ ( t τ ) ()

4 where Γ is the strength of the reflection relative to the line-of-sight signal, θ is the phase shift cause by the reflective meium, τ is the elay of the specular reflection, an πf τ is the phase shift cause by the elay ( f is the carrier frequency). Γ an θ are a function of the permeability, permittivity, an roughness of the reflecting surface, the incience angle of the arriving electromagnetic wavefront [], an the gain of the receive antenna in the irection of the arriving specular reflection []. The relative elay τ is a function of the geometry: slant range, elevation angle, etc. The overall phase shift of the specular reflection is γ = πfoτ + θ. For our analysis, we are intereste only in γ an not in its constituent parts so that j γ h( t) = δ ( t) + Γe δ ( t τ ). () The magnitue of the corresponing channel transfer function is H ( f ) = + Γ + Γcos f ( π τ γ ). (3) Some of the ata to be presente later is well moele by a channel consisting of two specular reflections. In this case we have j γ jγ δ ( t τ ) + Γ e δ ( ) (4) h( t) = δ ( t) + Γe t τ where the subscripts are use to enote the magnitue, phase shift, an elay of the first an secon reflections. In general, a multipath channel with L reflections may be moele as an LTI system with impulse response jγ i h( t) = δ ( t) + Γe δ τ L i= i ( t ). (5) In many applications, the accuracy of the moel improves with increasing L. But moels with large L are of limite practical value in simulation an analysis. Thus we prefer channel moels with as small an L as possible. Our ata moeling shows excellent results using L = or. The ieal channel souner is one that transmits an impulse an recors the channel output. When x(t) = δ(t), y(t) = h(t). Since transmitting an ieal impulse is not possible, the alternative metho outline in Figure is often use. Here x(t) is a bipolar NRZ pseuo-ranom (PN) sequence of length M. The PN sequence is use because it has the esirable correlation property [3]: i R XX ( λ) = x( t) x( t + λ) t λ + = T c M M T c < λ T λ < ( M ) T c c (6)

5 moulate PN source x(t) h(t) y(t) replica of x(t) correlator Figure : PN sequence metho for channel souning. The receive signal is correlate with a locally generate replica of the transmitte moulate PN sequence. The resulting correlation peaks ientify the multipath reflections. λ M T c MT c Figure 3: Perioic correlation function for length-m PN sequence moulate with bipolar NRZ pulse shape with a bit perio of T c secons. where T c is the bit perio. A plot of the correlation function (6) is shown in Figure 4 where it is seen that it is perioic with perio T c. To see how channel souning is accomplishe using a signal with this property, consier what happens when x(t) is passe through a two ray channel with impulse response given j γ by equation (). In this case, y( t) = x( t) + Γe x( t τ ). The channel souning proceure computes the cross correlation between the receive signal y(t) an the transmitte PN sequence x(t): R YX ( λ) = = R y( t) x( t + λ) t XX ( λ) + Γe jγ R XX ( λ τ ) The cross correlation function is plotte in Figure 4 for two important cases. In the first case, τ > T c so that the correlation peaks o not overlap. From the cross correlation function, we are able to etermine the magnitue, phase, an elay of the reflection. In the secon case, τ < T c so that the two correlation peaks overlap thereby making the etermination of the reflection properties problematic. (7)

6 Γe jγ M e j Γ γ M λ + Γe jγ M τ istortion cause by correlation peak overlap λ τ? Figure 4: Examples of the cross correlation between the transmitte length-m PN sequence an the output of a -ray channel with impulse response given by Equation (3). The top figure illustrates the case where the elay τ is greater than T c. Two istinct peaks are obvious an are separate by the time elay τ. The bottom figure illustrates the case where τ is less than T c. The two correlation peaks are smeare together making the ientification of the channel parameters problematic. When τ < T c it is still possible to perform some channel moeling using frequency omain techniques. This technique uses the power spectral ensities of the input an output signals which we enote by S X (f) an S Y (f), respectively, an exploits the relationship SY ( f ) = H ( f ) S X ( f ). (8) Using this technique, the power spectral ensity of the receive signal is compute then ivie by the power spectral ensity of the known transmitte signal (the PN sequence in our case) to obtain an estimate of the magnitue of the channel transfer function (which we call H ˆ ( f ) ): Hˆ ( f ) SY ( f ) =. (9) S ( f ) X As an illustration of the technique, assume a -ray channel moel h(t) given by () with corresponing transfer function given by (3). The channel transfer function H(f) is completely specifie by the three parameters Γ, γ, an τ. The parameters that most closely match the ata may be etermine by choosing the parameters to minimize the square error between the assume transfer function H ( f ) an the measure transfer function H ˆ ( f ). Usually, some weighting is use so that small values of H ( f ) an

7 Hˆ ( f ) are not ignore by the curve fitting process. For example, Rummler [4] ivie each point by the square of its absolute value. Here we use a ifferent weighting by converting the ata to ecibels (B) to convert equal magnitue ifferences to linear ifferences. In this way, the structure of to 3 B nulls in curve fit. In mathematical terms, the best-fit parameters are expresse as ( Γ, τ ) ( log H ( f ) log Hˆ ( ) H ˆ ( f ) are preserve in the, γ opt = argmin f ) () ( Γ, γ, τ ) where the conversion to B has been mae explicit. The extension of this technique to the 3-ray moel is straight forwar. Since many of the elays encountere uring the ARTM channel souning experiments were less than T c, the frequency omain metho was use to process the ata. EXPERIMENTAL CONFIGURATION The channel souning experiments were conucte as outline in Figure 5. A length-7 PN sequence was transmitte using a Mbit/secon BPSK transmitter in a T-39 aircraft. A -Watt linear power amplifier was connecte to a linearly polarize antenna mounte uner the fuselage. A GPS receiver logge the time an position for post flight ata correlation. The receiving station was locate at Builing 579, the main telemetry receiving site at the Ewars AFB complex. The receiver was equippe with a circularly polarize tracking antenna, a wieban telemetry receiver with a linear 7 MHz IF output, an ata acquisition equipment. The receiver AGC signal was sample at a rate of 5 ksamples/secon an recore with GPS erive time stamps. The bit error counts from the bit error rate analyzer were also logge. The IF output was sample at a rate of Msamples/secon using a high-spee igital oscilloscope. The oscilloscope sampling-trigger was riven by a BPSK emoulator couple with a bit error rate analyzer. Whenever the bit error count exceee a certain level, the sample trigger was asserte. During each trigger event, the igital oscilloscope recore evenly space ata segments. Each segment consiste of, samples, or µsecons of ata. At the en of each trigger event, the samples were ownloae from the igital oscilloscope to a PC an logge with a time stamp. To apply the frequency omain curve fitting techniques of Equations (9) an () to the sample ata, careful associations must be mae between the recore iscrete-time ata an the continuous-time ata assume by the technique. This association is outline in Figure 6. Here we see that we have samples of the output signal y(t) which we enote y(nt s ) where t s = ns is the sample perio. Computing the N-point DFT of the sequence Y e j e jω Y is the iscrete-time Fourier Ω y(nt s ) prouces ( ) πk for k =,,,N- where ( ) Ω= N

8 transform (DTFT) of y(nt s ) an π < Ω π is the iscrete-time frequency. Using samples jω of the input signal x(nt s ) we can compute ( e ) H jω ( e ) πk Ω= N Y jω ( e ) jω ( e ) X an form the ratio = for k =,, N- () X πk Ω= N 7-PN source LPF BPSK mo linear PA bipolar Mbits/sec aircraft fuselage hemispherically omni-irectional antenna tracking antenna wieban telemetry receiver AGC 7 MHz IF A/D sampling trigger high spee igital oscilloscope PC BPSK emo BER analyzer trigger circuit GPS Figure 5: System configuration for the ARTM channel souning flights.

9 which are frequency omain samples of H ( f ) over those frequencies where X(f) has support [5]. Thus we have k H k =,,..., N / Nt jω s H ( e ) π k = Ω= () N k N H k = N / +,..., N Nts x(nt s ) X jω ( ) e πk Ω= N x(t) h(t) H ( f ) =? y(t) sample y(nt s ) DFT Y jω ( ) e πk Ω= N H jω ( e ) πk Ω= N = Y X jω ( e ) jω ( e ) k H Nt s πk Ω= N Figure 6: Diagram illustrating the relationship between iscrete-time processing an the continuous-time channel impulse response. Thus while it is impossible to sample h(t) given by (), (4), or (5) irectly (they are not banlimite signals), it is possible to moel samples of H(f) over a limite frequency range [5]. Using the DFT s of the sample input an output signals an Equation () in place of H ( f ) in Equation (), we are able to apply the frequency omain technique to etermine the least squares parameters for our channel moel using the sample ata. RESULTS This paper reports on two of the ARTM channel souning flights outline in Table. The basic iea of these flights was to gather information from ifferent locations at ifferent frequencies using ifferent receive antennas. This ata represents a course sampling of the multipath environments common to lan-base aeronautical telemetry an gives us some iea of the number, strength, an elays of the multipath reflections to be expecte. Table : Technical specification summaries for ARTM Flights an. ARTM flight ientifier Flight Flight carrier frequency 5.5 MHz MHz receive antenna 8-foot parabolic reflector -foot parabolic reflector segment spacing msec 5 msec flight path Black Mountain Cors Roa

10 Some representative curve fits from Flight are illustrate in Figures 7 an 8. These figures also illustrate the ata moeling technique using the DFT s of the sample transmitte an receive signals. Figure 7 emonstrates an example where the -ray channel moel is an excellent match to the ata an suggests the existence of a strong specular reflection with a relative amplitue of.97 an a ifferential elay (from the line-of-sight path) of 47 ns. It is clear from the plot in the lower right han corner that the ae complexity of the 3-ray moel is unnecessary in moeling this particular ata segment. In this case the thir ray has a relative amplitue of only.4 an coul be ignore. Figure 8 is an example where the 3-ray moel is require. These results suggest the presence of a strong multipath reflection elaye 74 ns from the line-of-sight signal an a weaker multipath reflection elaye 3 ns from the line-of-sight signal. Both of these examples illustrate an important relationship between the - an 3-ray moels. The first reflection in the 3-ray moel is usually the strong reflection associate with a short elay an is consistent with the reflection moele by the -ray channel moel. The secon reflection of the 3-ray moel is usually a weaker reflection with a long elay. The inclusion of this aitional multipath reflection improves the accuracy of the moel. Generalizing to L multipaths (see Equation (5)), it is evient that improve moeling accuracy can be realize at the expense of computational complexity. Our experience with these moels an the available ata suggest that a 3-ray moel aequately captures all the essential features of the channel istortions cause by multipath propagation. The ata use in Figure 8 also illustrates an important caveat of channel characterization using the PN cross correlation metho. The PN cross correlation for the ata of Figure 8 is plotte in Figure 9. The PN cross-correlation ata show three peaks at ns (the line-ofsight path), ns, an 37 ns. The ns spacing between the first an secon correlation peaks is well below the ns spacing require by the PN correlation metho to resolve the iniviual multipath reflections. The correlation peak at ns is a false correlation peak resulting from the out-of-phase interference between the line-of-sight path an the multipath reflection elaye 74 ns. When the elay between the multipaths is greater than ns, the PN cross correlation metho oes a nice job of ientifying the elay. This is illustrate by the thir correlation peak at 37 ns in Figure 9 an is consistent with the frequency omain moeling results in Figure 8.

11 Spectrum (B) frequency (MHz) Spectrum (B) frequency (MHz) H(e jω ) (B) Γ =.97 γ =.6 ra τ = 47.3 ns H(e jω ) (B) Γ =.93 γ =.3 ra τ = 46. ns Γ =.4 γ = -.4 ra τ = 8.6 ns frequency (MHz) frequency (MHz) Figure 7: Representative ata from Flight. This ata was recore uring the first segment at time stamp 7:3:. Starting in the upper left an moving clockwise: () the DFT of the transmitte signal; () the DFT of the receive signal; (3) the resulting sample channel transfer function together with the least squares curve fit for the 3-ray moel; (4) the resulting sample channel transfer function together with the least squares curve fit for the -ray moel.

12 3 Spectrum (B) 5 5 Spectrum (B) frequency (MHz) frequency (MHz) H(e jω ) (B) Γ =.9 γ = -.8 ra τ = 7. ns H(e jω ) (B) Γ =.86 γ = -.6 ra τ = 73.9 ns Γ =.3 γ =.7 ra τ = 3. ns frequency (MHz) frequency (MHz) Figure 8: Representative ata from Flight. This ata was recore uring the ninth segment at time stamp 7:3:. Starting in the upper left an moving clockwise: () the DFT of the transmitte signal; () the DFT of the receive signal; (3) the resulting sample channel transfer function together with the least squares curve fit for the 3-ray moel; (4) the resulting sample channel transfer function together with the least squares curve fit for the -ray moel.

13 correlation (magnitue) Γ =.33 γ =.3 ra τ = 37. ns elay (ns).5 in phase quarature correlation elay (ns) Figure 9: PN cross correlation results for the ata presente in Figure 8. These plots are a goo example of both the utility an anger of using the PN cross correlation metho for the channel parameter estimation. The tallest peak at (normalize) elay ns represents the line-of-sight signal. The frequency omain least-squares curve fits in Figure 8 show that there are two elays accompanying the line-of-sight path. The first elay is a strong elay with relative amplitue.86 an relative elay 74 ns. The relative elay is much shorter than the minimum ns spacing require for elay estimation using the PN cross correlation metho. In this example, the phase ifference between the line-of-sight path an the short-elay path causes a false peak at ns in the correlation. This false peak is obvious in the top plot. The frequency omain least squares curve fit in Figure 8 ientifies a secon elay with relative amplitue.3 an elay 3 ns. This elay is obvious in the correlation curve in the top figure where a peak with relative amplitue.33 at 37 ns is observe.

14 3 Receive Spectrum (B) frequency (MHz) H(e jω ) (B) Γ =.3 γ =.5 ra τ = 36. ns frequency (MHz) correlation (magnitue) Γ =.3 γ = -.7 ra τ = 36. ns elay (ns) Figure : Sample ata from Flight showing an example of a long-elay multipath reflection. The upper left plot is the DFT of the receive signal (the transmitte signal is ientical to the one shown in Figures 7 an 8). The upper right is a plot of the measure transfer function together with a plot of the least squares fit using the -ray moel. The bottom is a plot of the magnitue of the PN correlation function which shows a correlation peak at 36 ns. Note the close agreement between the frequency omain technique an the PN cross correlation metho when the elay is well in excess of ns. Figure illustrates sample results from Flight. This plot is an example of a long multipath elay of 36 ns. As expecte, a multipath characterize by a elay this long has a relative amplitue which is smaller than the relative amplitues associate with shortelay multipath reflections. The frequency omain an PN cross-correlation methos prouce very similar results. This reinforces our confience in the moeling results obtaine using the frequency omain technique.

15 Figure : Parameter summaries of the -ray moel (left column) an 3-ray moel (right column) for the Black Mountain Run (ARTM Flight ) at Ewars AFB with an aircraft at an altitue of 5, feet. Figures through 4 summarize the - an 3-ray moeling results using the frequencyomain technique. Figures an summarize the results from Flight at L-ban while Figures 3 an 4 summarize the results from Flight at S-ban. We make the following observations: General Form for the Channel Moel: In general, we observe two multipath reflections together with the line-of-sight signal. The first multipath is a strong specular reflection with relative amplitue greater than.5 an relative elay in the 3 to 7 ns range. The secon multipath is a much weaker reflection with relative amplitue less than.5 an relative elay in the 75 to 35 ns range. To emonstrate that these numbers are reasonable, we consier the cross sectional slice of terrain for Flight at time 7:3: (the high-altitue Black Mountain run) shown in Figure 5. The ifferential path elays for a single multipath reflection with reflection points A, B, an C as shown were compute. The ifferential elay assuming a reflection from point A (the ege of the 3- B antenna beamwith) is 8 ns. The ifferential elay assuming a reflection from point

16 B is 3 ns while the ifferential elay assuming a reflection from point C is 5855 ns. These values boun the possible elays of multipath reflections from Figure : Parameter summaries for the -ray moel (left column) an 3-ray moel (right column) for the Black Mountain Run (ARTM Flight ) at Ewars AFB with an aircraft at an altitue of, feet. the groun at this point in the test flight. The moeling results summarize in Figure 8 are within this range. Note that for the ata moele in Figure 8, we see that the strong, short-elay reflection occurs off the ry lake be approximately half way between the aircraft an the receiver while the weak, long-elay reflection occurs in the foothills of the Black Mountains. Amplitue-Delay Relationship: In general, strong multipaths are characterize by short elays on the orer of 5 ns. Multipath reflections with longer relative elays on the orer of to 3 ns are much weaker. This is to be expecte since the longer paths that lea to longer elays usually impose greater attenuation. Frequency Depenency: Given the limite ata available at this time, it is ifficult to etermine which ifferences in Figures through 4 are frequency epenent an which are geometry epenent. Figures 3 an 4 emonstrate that there is tremenous

17 variation over a single flight path. Even though the frequencies an flight paths were ifferent, the results from the curve fitting at the two frequencies are quite similar. It is clear that there is significant multipath interference at both the L- an S-bans. Figure 3: Parameter summaries for the -ray moel (left column) an 3-ray moel (right column) for the mile portion of the Cors Roa Run (ARTM Flight ) at Ewars AFB with an aircraft at an altitue of 4, feet. Altitue Depenency: An interesting relationship exists between transmitter altitue an channel characteristics. Figure summarizes the results for the Black Mountain run for an aircraft at an altitue of 5, ft while Figure summarizes the results for the same flight path for an aircraft at an altitue of, ft. The low altitue run exhibite long elays on the orer of to 3 ns which are completely absent in the high altitue run. With a few exceptions, we see that the long-elay multipaths are weak reflections with a relative amplitue of. or less. During the high altitue runs, the increase attenuation ue to the higher altitue an ifferent incience angles resulte in multipaths too weak to trigger a ata acquisition event.

18 Figure 4: Parameter summaries for the -ray moel (left column) an 3-ray moel (right column) for the east-most portion of the Cors Roa Run (ARTM Flight ) at Ewars AFB with an aircraft at an altitue of 4,5 feet. Relationship to Bit Error Rate: Figures 6 an 7 present the bit error count together with the channel souning trigger events on the same time axis. The plot facilitates a correlation between channel conitions an the recore bit errors. The event numbers below the trigger events are a key in cross-referencing the channel conitions summarize in Tables an 3. For the sake of brevity, the tables list average values for the -ray moel least-squares fit parameters. In general, the large amplitue reflections cause the largest increases in bit error rates. This is to be expecte. What these figures suggest is that even relatively small multipath reflections with amplitues.3 or less can cause significant increases in the bit error rate.

19 Figure 5: Cross-section of the aircraft-to-groun station terrain for Flight, time 7:3:. The vertical scale is exaggerate to show the terrain features where the elevations are inicate. The line-of-sight path is shown as the soli line from the aircraft to the groun station. The ifferential elay for a single multipath reflection for 3 ifferent reflection points is compute: point A, the ege of the 3-B beam with; point B, the ege of the low hill overlooking the ry lake be; point C, near the top of the Black Mountains irectly uner the aircraft. The ifferential elays are 8 ns for point A, 3 ns for point B, an 5855 ns for point C. CONCLUSIONS We have summarize results from two ARTM channel souning experiments esigne to provie a course sampling of the terrain an frequencies typical of lan-base aeronautical telemetry. Channel moeling was performe in both the time-omain (PN cross correlation) an frequency omain (least squares curve fit to an assume moel). The two methos provie consistent results when the ifferential path elays exceee ns. We emonstrate that the channel parameters erive from this analysis were within the bouns impose by the transmitter-to-receiver geometry. It is shown that -ray an 3-ray multipath moels provie an excellent match to the ata obtaine thus far. The 3-ray moel is characterize by a strong line-of-sight propagation path; a strong shortelay specular reflection with relative amplitue greater than.5 an elay in the 3 to 7 ns range; an a weaker, long elay specular reflection with relative amplitue less than.5 an elay in the 75 to 35 ns range. Finally, we emonstrate the correlation

20 between channel parameters an bit error rates. This analysis showe that that the magnitue of the reflection is the ominant factor in etermining bit error rate. Table : Summary of average best-fit parameters for the -ray moel using ata from Flight. The event numbers correspon to the numbers marke in Figure 6. The measure BER is the average BER over the interval uring which the ata segments were recore. Event number average Γ average τ (ns) measure BER Table 3: Summary of average best-fit parameters for the -ray moel using ata from Flight. The event numbers correspon to the numbers marke in Figure 7. The measure BER is the average BER over the interval uring which the ata segments were recore. Event number average Γ average τ (ns) measure BER

21 Figure 6: Correlation between recore bit error counts an channel conitions for the first part of the low-altitue Black Mountain Run from Flight. Each point on the top line is a one plus the number of bit errors recore by the bit error rate analyzer uring a ms interval. The lower line plots the trigger events for the igital oscilloscope. The trigger events allow us to perform a time correlation between bit error rate an recore channel outputs. The event numbers below the trigger events are a key in cross-referencing the channel conitions summarize in Table. Note that event numbers, 3, 4, 7, an 8 correspon to channel conitions characterize by multipath reflections with relative amplitues less than.3 an inclue both long (greater than ns) an short (less than ns) elays. Thus we see that even weak multipath can cause significant BER performance egraation.

22 Figure 7: Correlation between recore bit error counts an the channel conitions for the Cors Roa Run from Flight. Each point in the top line is the number of bit errors recore by the bit error rate analyzer uring a 5 ms interval. The lower line plots the trigger events for the igital oscilloscope. The trigger events allow us to perform a time correlation between bit error rate an recore channel outputs. The event numbers below the trigger events are a key in cross-referencing the channel conitions summarize in Table 3. Event numbers -4 occurre when the aircraft was at the mi-point of the Cors Roa run. At this point, the channel is characterize by long elays (longer than the ns bit time) an amplitues less than.5. Event numbers 5 occurre when the aircraft was at the eastern en of the Cors Roa run where the channel is characterize by short elays (less than half the ns bit time) an amplitues greater than.5. This plot illustrates clearly that the shorter, stronger elays egrae BER performance much more than the weaker, longer elays.

23 REFERENCES. Rice, M. an E. Law, Aeronautical Telemetry Faing Sources as Test Ranges, in Proceeings of the International Telemetering Conference, volume 33, Las Vegas, Nevaa, October 997, pages Jakes, W., Microwave Mobile Communications, IEEE Press, Piscataway, New Jersey, Proakis, J., Digital Communications, Thir Eition, McGraw-Hill, New York, Lanon, D. Doppler Banwith Characterization of ARTM Channel Souning Data, in Proceeings of the International Telemetering Conference, volume 35, Las Vegas, Nevaa, October Bishop B. an E. Law, S an L Telemetry Bans: Signal Level Comparison, Technical Report, Point Mugu, California, October, Law, E., private communication, Naval Air Weapons Center WPNS, Point Mugu, California. 7. Law, E., Performance of PCM/FM During Frequency Selective Faing, in Proceeings of the International Telemetering Conference, volume 3, Las Vegas, Nevaa, October 995, pages Reemann, J., Ewars Range Telemetry Evaluation, in Proceeings of the International Test an Evaluation Conference, Lancaster, California, April Frien, D., BER Analysis of an F-6 Test Run at Ewars AFB, in Proceeings of the International Telemetering Conference, volume 33, Las Vegas, Nevaa, October 997, pages Welling, K., Analysis of JDAM Tests at China Lake, in Proceeings of the International Telemetering Conference, volume 33, Las Vegas, Nevaa, October 997, pages Balanis, C., Avance Engineering Electromagnetics, John Wiley an Sons, New York, Rice, M. an D. Frien, Antenna Gain Pattern Effects on Multipath Interference in Aeronautical Telemetering, in Proceeings of the International Telemetering Conference, volume 33, Las Vegas, Nevaa, October 997, pages Peterson, R. an R. Ziemer an D. Borth, Introuction to Sprea Spectrum Communications, Prentice-Hall, Englewoo Cliffs, New Jersey, Rummler, W., A New Selective Faing Moel: Application to Propagation Data, Bell System Technical Journal, volumn 58, July-August, 979, pages Oppenheim, A. an A. Willsky, Signals an Systems, Secon Eition, Prentice-Hall, Upper Sale River, New Jersey, 997.