OFDM WAVEFORM FEATURE SUPPRESSION Ronald R. Meyer Michael N. Newhouse Abstract Traditional multi-carrier systems developed for commercial applications use features of the waveform to improve the performance of receiver processing. These features, added in transmit processing, include, for example, a cyclic extension to mitigate inter-block interference, pilot tones inserted to assess the channel state, and specific FIT bin patterns for synchronization. An adversary can exploit these features to glean information about the transmitted waveform. This paper discusses a low probability of intercept (LPI) analysis of traditional and militarized OFDM waveforms. Militarized efforts include the elimination of the cyclic extension, pilot tones, and fixed synchronization patterns. We present an overview of the algorithms used to provide additional LPI relative to the commercial systems. In addition, traditional LPI quantitative analyses of the OFDM waveform are presented. Without adding these militarized features, the analysis shows that the OFDM wavefoim is rather easily detected. These analyses show the detectable features in OFDM without cyclic extension decrease as the number of subcarriers increase. Introduction Future Combat Systems Communications (FCS-C) is a DARPA program with the goal of developing and demonstrating communications technologies needed for future combat systems acquisition programs. The Rockwell Collins program sponsored by DARPA uses an orthogonal frequency division multiplexing (OFDM) waveform for multi-path mitigation. OFDM uses multiple, orthogonal carriers with minimal subcarrier spacing to convey information. OFDM supports similar data rates compared to single carrier systems, with simplified receiver structures in channels exhibiting frequency selectivity and narrowband interference. One primary reason to use OFDM is its ability to perform well in a multipath environment where significant time delay spreads may be present. Large time delay spreads Communications & Navigation Systems Department Advanced Technology Center - Rockwell Collins Cedar Rapids, Iowa 52498 induce severe frequency selective fading. OFDM with proper coding provides increased robustness against this type of channel compared to single tone modulation schemes. Commercial off-the-shelf (COB) based OFDM standards and systems unfortunately calntain many attributes that make them vulnerable to detlection. The users of these systems do not have the same security concerns as the military. Encryption of data is vital in both arenas, but covert operation is a unique requirement for military users. OFDM signal features, such as cyclic extension to combat inter-symbol interference, produce easily detectable features that can be generated by non-linear surveillance receivers. For the FCS-C program, Rockwell Collins developed an OFDM waveform that utilizes hybrid techniques with frequency hopping (FH) and direct sequence spread spectrum modulation to provide a system that is both AJ/LPD capable and resistant ito severe channel distortion from multi-path propagation. We will discuss the shortfalls of the commercial standards and present our solutions to overcome these limitations. OFDM Features and Their Generation COTS OFDM systems contain synchronization symbols, pilot carriers, and guard intervals that are vulnerable to detection by non-linear receivers. These cyclostationary features are not suppressed even when using frequency hopping or direct sequence spreading techniques. For example, GPS signals offer protection against traditional linear intercept receivers because the signal is below thermal noise. But, a simple squaring receiver can remove the BPSK spreading code and modulation, and generate a spectral line at double the carrier frequency. This line can be easily detected when the SNR is much less than 1. An OFDM signal with its multitude of carriers seems particularly vulnerable to tlhis type of exploitation. Fortunately, these feature detectors can be defeated, or at 0-7803-7625-0/02/$17.00 02002 IEEE.
least, the detection made much more difficult, when the signal is randomized with a PN pattern that is known only to authorized users. With the appropriate randomization, a signal passed through the nonlinear feature generator will have the feature spectral lines smeared out in the frequency domain by the random dither. If the dithering is done correctly, the spectral lines can be spread out over a wide enough bandwidth to make them essentially undetectable. As we shall show in this paper, non-linear detection of the subcarrier spectral lines is somewhat thwarted by the intermodulation bet ween subcarriers. Our methods presented here offer solutions to suppress features so the interceptor is forced to use a wide-band radiomet er. Synchronization Traditional COTS OFDM systems use either a fixed subset of bins or a single, wideband carrier for frequency and timing synchronization. Those that use a fixed subset of frequency bins, such as IEEE 802.11a, sparsely populate one or more OFDM symbols in frequency using a unique pattern. This generates a time domain series containing unique patterns that can be used in the receiver to provide timing and frequency information about the transmitted burst. These correlative time domain patterns are undesirable in a system designed for LPD. A ' wideband, single carrier correlation preamble can be implemented to provide good correlation and LPD properties. However, when employing transmit excision with OFDM, a single carrier correlation sequence cannot be utilized. Due to these LPD and coexistence issues, we desire a synchronization pattern that uses only the frequency bins used in the OFDM symbol without time domain features. Our method generates pseudorandom frequency domain samples on the carriers used in each OFDM symbol as shown in Figure 1. One or more EFTS produce a time domain series that is transmitted at the beginning of each burst. The same pseudorandom frequency domain samples produce an identical series using an IFFT in the receiver. The serial to parallel converter fils only the frequency bins used in the data portion of the symbol. The correlation filters use the correlation vectors extracted from the vector-split function shown in the figure. The receiver uses the IFFT output as a set of correlators. This set of correlators provides precise time synchronization of the received OFDM burst. The frequency offset is evaluated by estimating the drift in the phase of the pseudorandomly generated OFDM synchronization pattern. Figure 1 - Example correlation pattern generator This method of correlation enhances the LPD nature of the overall waveform while preserving the ability to support multiple users in the frequency domain or prevent interference with fiied communications services within the OFDM transmission band. This method could use any of the available FFT bins with receiver and transmitter synchronized PN modulated data. Generation of the synchronization sequence in this manner allows for improved correlation properties. In addition, this method generates a featureless synchronization pattern suitable for OFDM with multiple access capability. OFDM Feature Suppression Commercial OFDM signals are vulnerable to detection if a guard interval for inter-symbol interference (ISI) suppression uses cyclic extension. This threat potentially exploits an OFDM signal that is hopped unless measures are taken to suppress the cyclic guard interval features. Commercial OFDM systems commonly use a cyclic extension of the IFFT outputs at the modulator to generate the guard interval. This periodic extension to each OFDM symbol period can be detected by a cross correlation as follows: x(n) = s(n) s* (n- ToFDM ) where TS is the duration of the entire OFDM symbol (including the guard time, Tgrm,.d), * indicates conjugate operation. Figure 2 below illustrates the alignment of the received signal and a delayed version that will generate a detectable feature by cross correlation. 583
cyclic extension remove the modulation and generate a subcarrier line depends on the modulation. For example, if the subcarriers use BPSK modulation, a second order non-linearity will generate lines at frequencies twice the sub-carriers. Squaring an M sub-carrier complex OFDM symbol produces: M y-1 for n fin The summation in the expansion represents the spectral lines generated from the sub-carriers with the modulation removed from squaring. Each modulated subcarrier will produce M spectral lines at mice the subcarrier frequencies equal to 201, 40, 60,... 2Mw The other terms in the expansion are the intermodulation (IM) products between all of the OFDM sub-carriers. The IM products produce sum and difference products that can fall on top of the other OFDM subcarrier frequencies. In. general, the nr and mfh subcarriers will have an IM product of an an, e j ( m t + h ) nlm + h, ) ej( for n#rn i Since the data symbols are all _+1 and independent, the IM products are indistinguishable from BPSK modulated subcarriers except the IM product is folded to a new subcarrier frequency of (n+m) a. This feature extraction for a 32 subcarrier OFDM waveform is illustrated in Figures 4 and 5 below. Figure 4 shows the OFDM spectrum for this simulation. The signal contains 32 sub-carriers with BPSK modulation on each subcarrier is in Figure 4. MO 900 Figure 3 - Cyclic guard detection with cross correlation guard interval. This eliminates a dead time that effectively corresponds to a periodic odoff modulation at the OFDM symbol rate. An odoff AM component on the OFDM signal will also generate strong rate lines and should be avoided. Spectral lines will be generated by passing the OFDM signal through a non-linearity. The order necessary to SPFDM.db i 10 Figure 4 - OFDM signal spectrum. 584
I I I I I 0 I 304 I 102 310' 410' 510' 610' 0, P P - Figure 5 - ODFM spectrum after squaring operation. After the squaring operation, the signal spectrum is show in Figure 5. The parameters for the simulation were: OFDM symbol duration 0.8 sec, guard interval 0.2 sec, and 256K point FTT for the spectrum of Figures 4 and 5. The spectral lines in Figure 5 are the telltale signature of a multicarrier waveform. For intercept receivers with minimal or no oversampling before the non-linear processing, there will be a considerable amount of aliasing of the carrier lines and IM products as seen in Figure 5. As the number of subcarriers in an OFDM signal increase, IM products will also increase and mask the spectral lines generated by squaring all the subcarriers. In equation 1, the IMD terms generate sum and difference terms that fold onto the other subcarrier bins. For an M subcarrier OFDM symbol total number of IM generated BPSK products is: CM (sum products) + C; (dlfserence products) = M(M-1) = M' Each of these will fold to add a total number of M' IM products at each subcarrier bin. This effect is illustrated in Figure 6. Here the spectrum at the output of the squaring operation is shown for an OFDM signal with 256 subcarriers. The symbol and guard times were kept the same for this simulation. 20 201 I I I I I I Figure 6 - Spectrum at output of squaring operation, 256K point FFT. When squaring an OFDM symbol, it can be shown that the ratio of one subcarrier line power to the power spectral density (at the line frequency) has the relationship: power in subcarrier line 1 oc- IM PSD M' So for every doubling of the number of subcarriers, the regenerated carrier line (with BPSK niodulatioii) is masked 6 db by the IMD between the subcarriers. In this example, there is no advantage for an intercept receiver to use non-linear methods to detect low-level signals. Under these conditions, a better intercept strategy would be to use a linear receiver. However, in waveforms with fewer numbers of subcarriers, a random frequency jitter can be introduced to mask the multi-carrier signature in a nonlinear receiver. One strategy for implementing this algorithm is shown in Figure 7. EP PN Freq PN Freq Figure 7 - Random frequency jitter for suppression of subcarrier features. Hop and symbol rate detectors Non-linear methods exist to exploit other attributes of an OFDM waveform. In general, hop rate and symbol rate detectors perform similarly to square-law and fourth-law receivers. With these receivers a nonlinearity generates a coherent spectral line at the hop rate, or symbol rate. Even though the OFDM carrier frequency is psuedo-randomly varied, the precise time that the carrier changes will be periodic if a constant hop rate is used. Time jittering should be used in an FH-OFDM system to confuse the intercept receiver and remove strong spectral lines at the hop and symbol rate. Jittered signals tend to shorten the intercept receiver's integration time, and broaden discrete spectral lines. Summary In summary, strategies to avoid detection of the FH-OFDM dwells would incorporate dither at several levels into the waveform. Time hopping for each FH- OFDM dwell will mitigate the threat from a channelized radiometer by disrupting the intercept receiver integration 585
period. Time dither can be used also mitigate the threat from a hop and symbol rate detector. For an OFDM waveform with a small number of subcarriers, a non-linear receiver will be able to detect spectral lines regenerate from the subcarriers. In this case, frequency dither can be used to mask this signature. For systems with 256 or more subcarriers, IM products mask these tones and dither is not necessary. The degree of dithering is a function of how covert the system needs to be, so the intercept receiver is forced to use a wideband radiometer, instead of a channelized radiometer or non-linear feature detector. The effectiveness of a covert OFDM waveform can also be improved by the use of directional antennas, power control, and robust coding. All of these methods ensure that the absolute minimum power is required for communication, while minimizing the signal power at the intercept receiver. References Nicholson, David L., Spread Spectrum Signal Design - LPE & AJ Systems, Computer Science Press, 1988. Reed, D. E. and Wickert, M. A., Minimization of Detection of Symbol-rate Spectral Lines by Delay and Multiply Receivers, IEEE Trans. on Comm., COM- 36, pp. 118-120, January, 1988. Rockwell Collins, OFDM for FCS-C Base Period Final Report, Feb 2002, Contract: DAAD19-01-9-0002. Imbeaux, J. Pe$ormance of the delay-line multiplier circuit for clock and carrier synchronization in digital satellite communications, IEEE J. Select. Areas Comm., vol. SAC-1, pp. 82-95, January 1983. Mills, Robert F., Prescott, Glenn E., Waveform Design and Analysis of Frequency Hopping Networks, IEEE trans. on Aerospace and Electronic Systems, Vol. 36, July 2000, p848-858. 586