FMCW waveform generator requirements for ionospheric

Transcription

1 Radio Science, Volume 33 Number 4, Pages , July-August, 1998 FMCW waveform generator requirements for ionospheric over-the-horizon G. F: Earl radar Electronics and Surveillance Research Laboratory, Defence Science and Technology Organisation Salisbury, South Australia, Australia Abstract. The purpose of this paper is to discuss the significance of FMCW waveform generator spectral purity with regard to its propensity to limit over-the-horizon (OTH) radar performance. The fundamental principle of operation of FMCW OTH radar is first described and then further developed in order to yield results in terms of several parameters of a synoptic database of OTH radar propagation data acquired with the Australian project Jindalee OTH radar frequency management system. The method of data analysis is addressed in detail and followed by the presentation of the results of the analysis of 5 years of data. Finally, the high demands of waveform generator spectral purity are emphasized in the sense of avoiding radar performance degradation due to this potential form of internal noise limitation. 1. Introduction The performance of any radar system is ultimately limited by the magnitude of the noise floor obscuring potential target detections. In the case of over-the- horizon (OTH) radars operating in the HF bands, the relatively high external noise fields determine that an appropriate radar design will result in the system being externally noise limited. However, as higher- powered radars are employed in missions involving smaller targets, the radar dynamic range requirements escalate, and a number of phenomena related to instrumental effects can determine that the radar will become undesirably internally noise limited. One such instrumental effect relates to the spectral purity of the waveform generators required at the transmitter and receiver sites. In a sense, the waveform generators may be regarded as the heart of the RF section of the radar, and lack of adequate spectral quality will determine the radar to be internally noise limited. Essentially, the design of the waveform generator requires an appreciation of relevant propagation factors, which are shown to be ionospherically propagated ground backscatter clutter levels and atmospheric noise spectral densities. In this paper the principle of operation of FMCW OTH radar is discussed. The manner in which the dynamic range of the waveform generators can influ- Copyright 1998 by the American Geophysical Union. Paper number 98RS /98/98RS ence the radar detection capability by controlling the noise floor is next discussed, and as a consequence a quantitative relationship linking the radar performance to the waveform generator performance is derived. The relationship establishes the required waveform generator spectral purity in terms of several environmental factors. These factors may be derived from data logged by the Australian project Jindalee OTH radar frequency management system. The methodology of the analysis undertaken in order to establish the waveform generator requirements is next described in detail. Finally, the results of the analysis of 5 years of data are presented and interpreted in the context of OTH radar requirements. 2. Principle of FMCW OTH Radar Operation The principle of FMCW HF radar operation is described in detail elsewhere [e.g., Barrick, 1973]. Essentially, a repetitive waveform is swept between lower and upper frequency limits at a constant repetition frequency. The waveform is radiated at high power (typically several hundred kilowatts) from the transmitter site and propagates via the ionosphere to the region under surveillance. Within that region, energy is scattered back to the receiver site both from the Earth's surface and from targets. The return path is again via the ionosphere. At the receiver site a replica of the transmitted waveform is used as a local oscillator signal that is mixed with the backscattered signals in a process known as "aleramping." Following

2 1070 EARL: WAVEFORM GENERATOR REQUIREMENTS Figure 1. DOPPLER SHIFT Spectral content of radar range cell. NOISE the deramping process, the clutter, being characterized by only small Doppler shifts due to either ionospheric motion or scatter from moving ocean waves, is concentrated in a series of spectral lines separated by the waveform repetition frequency. Moving targets are resolved from the clutter lines by frequency offsets corresponding to the Doppler effect as depicted in Figure 1. A commonly used measure of relative radar sensitivity is radar subclutter visibility (SCV), defined as the clutter-to-noise ratio following coherent integration over the radar dwell time, and is shown in Figure 1. As a consequence of the relatively high external noise levels characteristic of the HF bands, the noise floor shown in Figure 1 should be external, i.e., of either atmospheric or galactic origin. In particular, the noise floor should not be due to noise attributable to waveform generators of inadequate spectral purity. In order to test the performance of waveform generators within the laboratory, a configuration as shown in Figure 2 is commonly employed. The low-pass filtered output from the test mixer is subjected to spectral analysis, the signal consisting of a single wanted peak due to the beat frequency between the two signal generators, together with a broadband noise floor, the magnitude of which is a measure of the spectral purity of the waveform generators. The requirements placed on the spurious free dynamic range (SFDR) of the test equipment are very demanding. The results derived in the present study determine that the SFDR of the test equipment is required to be approximately 30 db above the maxi- mum SCV. 3. Specification of Waveform Generator Performance Suppose that as that a consequence a radar employs of performing waveform generat a deramp- such ing test of the type described above, the ratio of the wanted signal peak to unwanted waveform generator induced noise floor is W db, the test being performed with all relevant parameters set at the same values as employed foradar operations. The radar is required to achieve a subclutter visibility S db when used operationally, and we require a relationship determining W in terms of S. For simplicity we will assume that the deramping process yields a noise floor of uniform spectral density. Figure 3 is a frequency domain depiction of the relevant clutter and noise variables at the receiver output. As discussed in the previous section, the clutter from successive range cells is constrained to appear at multiples of the waveform repetition frequency and is shown at levels Cj, Cj+ 1, Cj+2, etc. The waveform generator induced noise levels corresponding to the same range cells are shown at levels Gj, Gj+i, Gj+2, etc. By definition, the waveform generator induced noise associated with each range cell is W db below the corresponding clutter level. Suppose that T is the total received clutter from the entire radar footprint (dbw), that Cma x = clutter originating from range cell yielding peak subclutter visibility (dbw), and that we define F = T- Cmax. Clearly, the total waveform generator induced noise Gtot is the sum of the noise contributions from all range cells, and so G tot = T- W = C ma x q-f- W. The level of internal waveform generator induced noise is proportional to the total clutter level and thus to the transmitted power level. Increasing transmitter WAVEFORM GENERATOR #1 WAVEFORM GENERATOR #2 SPECTRUM ANALYSER Figure 2. Waveform generator test configuration.

3 EARL: WAVEFORM GENERATOR REQUIREMENTS 1071 c j Cj+2 c Noxt G J+t G J+3 Ga RECEIVER BASEBAND FREQUENCY (Hz) Figure 3. Clutter and waveform generator induced noise levels from successive range cells shown at the receiver output, together with the consequentotal waveform generator induced noise and external noise. power to realize greater system sensitivity will fail when this form of instrumental noise is dominant. Waveform generators of adequate spectral purity are mandatory. If we insist on the total unwanted waveform generator induced noise G to t being suppressed Q db below the external noise level Next, in order to maintain an externally noise limited radar, G tot = Next - Q. Rearrangement of the equalities yields By definition, and hence we determine W = Cmax - Next + F + Q. S = Crnax -- Next W-S+F+Q. The degree of spectral purity required of the waveform generator may thus be derived in terms of variables obtainable from the database of environ- mental factors. In particular, the above expression emphasizes the explicit nature of F in enhancing the waveform generator requirement beyond that of supporting merely the maximum subclutter visibility. 4. Available Database Estimates of both peak SCV and total clutter to peak clutter ratio, F, are obtainable from the database collected using the Jindalee frequency management system, which has been described by Earl and Ward [1987]. Of the various subsystems described by Earl and Ward, only the backscatter sounder and background noise component of the surveillance sub- systems are relevant to the present study. The backscatter sounder data consist of eight backscatter ionograms collected from each of the eight antenna beams deployed at the receiver site. Each backscatter ionogram consists of a matrix of points where the 120 columns correspond to the 120 frequencies employed between 6 and 30 MHz at 200-kHz resolution. The 256 rows of the matrix correspond to 256 range cells between the 0- and 12,500-km group delay at a resolution of 49 km. The information stored at each point of the matrix is the measured backscatter clutter

4 1072 EARL: WAVEFORM GENERATOR REQUIREMENTS (dbw) at the given operating frequency and group delay. Likewise, the background noise information to-noise ratio established in step 3 and so corresponds to a radar delivering 1 kw of transmitter power to a consists of eight sets of background noise spectral single log-periodic transmitting antenna. Throughout density data (dbw/hz) defined at the same 120 frequencies as employed for the backscatter data. Both sets of data were recorded synoptically at the Jindalee Experimental Facility at Alice Springs, Australia, over the 5-year period inclusive and the period August 1990 through June 1991 inclusive. The data logging systems were designed to run unattended and were scheduled to collect data every 15 min. Once a day the data were transferred to tape and the study, Q was set equal to 10 db, thereby determining that the internal noise floor due to the waveform generators was maintained 10 db below the measured external noise floor, resulting in a system degradation of 0.4 db. 6. On the basis of the above, for each 1-MHz band of an ionogram, there are 40 estimates of each of the variables being investigated (eight beams each recorded at 200-kHz resolution). For each of the varisubsequently sent to Salisbury, where all statistical ables S, F, and W, the 40 available estimates were analyses were conducted. scanned in order to yield the maximum values, and only the maximum values were retained for further 5. Ionogram Data Reduction analysis. Thus, from each ionogram run, each of S, F, and I4 / were retained at 1-MHz resolution, together The data for a single set of eight ionograms and background noise data were analyzed in the following with the group delay of the range cell yielding the peak subclutter visibility. manner: 1. Using values of transmitter power measured during the backscatter sounder run and recorded along with the ionogram data, all clutter estimates were adjusted to correspond to 1 kw of sounder radiated power. 2. A radar bandwidth (specifying the radar spatial resolution) was nominated and used to adjust the sounder clutter estimates to those corresponding to a radar of the specified bandwidth. All data reported in this paper employed a radar bandwidth of 10 khz as typically used for aircraft detection. 3. At each of the 120 frequencies defining the ionogram, the largest single-cell clutter level was combined with the corresponding noise level to calculate the peak clutter-to-noise ratio. This estimate corresponds to the subclutter visibility attainable at that frequency with a radar characterized by the same effective radiated power as the sounder, namely, 1 kw of transmitter power and a single log-periodic antenna. The radar was characterized by a coherent integration or "dwell" time of 1 s. The group delay corresponding to the peak clutter signal was also noted. 4. At the same 120 frequencies the total clutter was summed (linearly) in order to derive the quantity T defined above and was then used in conjunction with the peak clutter found in step 3 to derive F, also defined above. 5. The data from steps 3 and 4 were then used in order to establish W for each of the 120 frequencies. For this purpose, S was set equal to the peak clutter- 6. Statistical Analysis A statistical analysis was conducted on the data resulting from the above analysis of individual ionograms. The analysis could be applied to any nominated month and consisted of the following procedure: 1. A matrix was organized such that the columns corresponded to the hours of the day, and the rows corresponded to the 1-MHz bands between 6 and 30 MHz. 2. Within each element of the matrix a histogram was established, allowing the statistics relevant to a given variable to be accumulated. 3. A month's data was then analyzed using the procedure detailed above, and the resulting data were entered into the appropriate histograms. 4. The histograms were then interrogated to provide data on the median and 10 and 90 percentiles of the variables under investigation. 5. The data relevant to a specific variable and statistic were then fed to a contouring routine for hard copy output and retained as data files for further analysis. 7. Presentation of Results Representative results of the statistical analysis are presented in Figures 4-7 and Plate 1. Figures 4-6 consist of contour plots of data pertaining to October Median estimates of each of S (peak clutter-tonoise ratio), F (ratio of total clutter power to peak

5 EARL: WAVEFORM GENERATOR REQUIREMENTS 'R' Z TIME (UT) Figure 4. Peak clutter-to-noise ratio (in decibels): October 1987 median data. i i! [ i i I uj uj 6,.. [,.. [., ' ] ' ' ß TIME (UT) Figure 5. Total clutter power to maximum single-cell clutter poweratio (in decibels): October 1987 median data.

6 1074 EARL: WAVEFORM GENERATOR REQUIREMENTS u i ' ' I TIME (UT) Figure 6. Waveform generator dynamic range (in decibels)' October 1987 median data. 3O u I I I ' ' ' I ' ' ' I ' ' ' TIME (UT) Figure 7. Group range of maximum single-cell clutter (in kilometers): October 1987 median data.

7 EARL: WAVEFORM GENERATOR REQUIREMENTS I 75dB 7O J F ivl'"', M J J A BOND JFMAMJ J A SaND JFMAMJ J A SaND Plate la. Waveform generator dynamic range (in decibels)' upper decile data db JFMAMJ J A 8OND JFMAMJ J A SaND JFMAMJ J A 8aND Plate lb. Waveform generator dynamic range (in decibels): upper decile data.

8 1076 EARL: WAVEFORM GENERATOR REQUIREMENTS single-cell clutter power), and W (waveform generator dynamic range) are presented. Figure 7 is a plot of the median group range at which the peak clutter signal occurred and has been included in order to facilitate interpretation of the other data. In particular, there are regions at low frequencies (less than directive transmitter gain has been enhanced by 12 db, and thus the effective radiated power of the radar would be 37 db greater than that used to generate the contour plots. Adopting the 10-dB margin of external to internal noise employed in this study, at around 1300 UT in October 1990 when the upper decile about 10 MHz) during local daytime (approximately (low-powered radar)waveform generator dynamic UT), where the propagation conditions range requirement is 93 db, such a radar would need relate to vertical incidence returns on the backscatter to incorporate a waveform generator characterized by ionograms. Apart from the effect of the strong vertical incidence returns, this domain is also characterized by very low atmospheric noise levels due to ionospheric D-layer absorption, and for both reasons the data are irrelevant to propagation phenomena of interest to OTH radar performance. The apparently high values of peak clutter-to-noise ratio and waveform generator dynamic range generated in this domain were excluded from further analysis by a mask defining regions of no relevance to the present study. The plots clearly depict the diurnal behavior of each of the three variables under investigation as a function of radar operating frequency. From the October 1987 median data the following conclusions may be drawn: 1. The ratio of the total clutter power to peak single-cell clutter power may be as large as 20 db. 2. Radar subclutter visibilities of 53 db can result a dynamic range of 130 db. 8. Conclusion The principle of operation of FMCW OTH radar was discussed and used to establish a quantitative relationship between attainable operational radar dynamic range and waveform generator dynamic range. The manner in which a database of relevant environ- mental factors was used in order to establish a statistical description of the requirements was then described, and quantitative estimates of waveform generator spectral purity were derived. Failure to incorporate waveform generators of adequate spectral purity will determine that the radar will become internally noise limited and thereby unable to achieve the small target detection capability which would otherwise be fundamentally controlled by environmental factors. The study provided a further demonstration of the practical application of the calibrated synoptic database to problems involving the specification of critical elements of OTH radar instrumen- from the operation of a radar with an effective radiated power corresponding to 1 kw of power being fed to a single log-periodic antenna. 3. A similarly powered radar requires waveform tation. generators capable of exhibiting 76 db dynamic range in a standard laboratory deramping test. Acknowledgment. The author acknowledges the efforts The data for Plate 1 were derived from the analysis of his colleagues N. Borgas and A. Udina, who developed of the 5 years of data by software designed to extract the software necessary to pursue the above analysis. the upper decile value of the variable from the relevant contour plots established for the individual References months, on the basis that certain regions of the contour plots were excluded as discussed above. The Barrick, D. E., FMCW radar signals and digital processing, monthly upper decile contour plots were scanned as a NOAA Tech. Rep. ERL 283-WPL26, 22 pp., function of universal time, and the maximum value of Earl, G. F., and B. D. Ward, The frequency management waveform generator dynamic range was extracted system of the Jindalee over-the-horizon backscatter HF independently of radar operational frequency. radar, Radio Sci., 22, , The stringent requirement imposed on the wave- G. F. Earl, DSTO Electronics and Surveillance Research form generators becomes apparent when the present Laboratory, Building 200, Commerical Road, Salisbury 5108, data are scaled to be more representative of an South Australia, Australia. ( fred.earl@dsto. operational OTH radar. Suppose, for instance, that defence.gov.au) the radar transmitting facility incorporated sixteen 20-kW power amplifiers, each feeding a log-periodic antenna. The radar power level is 25 db above the 1 (Received November 4, 1997; revised April 28, 1998; kw employed in the present study. In addition, the accepted May 4, 1998.)