Field of the Invention
The present invention generally concerns radio-frequency interference (RFI) suppression techniques for application to communication systems, such as satellite systems. In particular, the invention concerns the protection of satellite transponders subject to jamming, to earth terminals subject to RFI, or in any other communication environment where RFI presents a problem.
Background of the Invention
RFI reduction techniques fall into many different categories. The one for which this invention was intended to provide an improvement is a nonlinear automatic gain control (AGC) device referenced under the general heading of Smart AGC.TM.. The Smart AGC.TM. concept is shown in a device 10 which is illustrated in simplified form in FIG. 1A and has been the subject of U.S. Pat. No. 5,379,445, issued on Jan. 3, 1995. The disclosure of that U.S. Patent is incorporated herein by reference. In the Figure, a wanted signal with interference is input to an adjustable time delay 11. The delayed signal is provided to a null zone envelope I/O unit 12, which has a characteristic response curve as illustrated in FIG. 1B. The input signal also is provided to an envelope detector 13 which could be realized by using a broad band radiometer power meter. The output serves as an input to a filter 14 having adjustable time constants. The output of filter 14 is input to the null zone envelope I/O 12 and acts to control its processing in the manner subsequently described. In brief, the RFI suppression characteristics of Smart AGC.TM. are provided by an active nonlinear device which is controlled by observation and processing of the input envelope.
The device 10 disclosed in the above referenced U.S. Patent Application utilized a biased "hard limiter" as shown in FIG. 1C. The hard limiter employed a "null zone," i.e., a range of input voltages for which the output voltage is zero. There are many different realizations of null zone nonlinearities, but the biased hard limiter is easy to implement since it has only 3 levels and has been analyzed extensively.
The hard limiter device used for active RFI suppression has a sensitivity to threshold error as shown in FIG. 1D for the special case where the input ratio of RFI to wanted signal was 30 dB (a factor of 1000 in signal power). In practice, one wants to operate the device at the point where the curve labeled (I+IM)/C is at a minimum. Any other point would be considered to have "tracking error." Tracking error will sometimes cause placement of the envelope threshold at a higher value than the peak of the composite input envelope and will destroy both the interfering and the wanted signals. FIG. 1D shows the (I+IM)/C curve approaching infinity when the threshold-to-interference ratio reaches about 0.275 dB. Another characteristic feature of the null zone device which is evident in FIG. 1D is that the minimum of the curve (I+IM)/C occurs at about 4.77 dB. This means that the best one can do, even adjusting the null zone for its best level, is to reduce an input I/C to an output level of about 4.77 dB. While this may be very good for some applications, it does not satisfy the need for a nonlinear device where the minimum level of (I+IM)/C is lower than 4.77 dB and which is less sensitive to tracking errors.
Summary of the Invention
It is an object of the present invention to provide a nonlinear device which is less sensitive to threshold errors than the biased hard limiter.
It is a further object of the present invention to provide a nonlinear device that results in a better optimum output (I+IM)/C ratio.
It is yet another object of the present invention to provide a nonlinear device that uses signal inversion rather than a null zone in order to provide a nonlinear response characteristic.
These and other objects are achieved by providing a processor for RFI suppression that uses a biased inverting limiter.
Brief Description of the Drawings
FIG. 1A is an illustration of a voltage-variable null zone device used in a signal processor for RFI suppression.
FIG. 1B is an illustration of the null zone envelope input-output curve which results using a conventional biased hard limiter design shown in FIG. 1A.
FIG. 1C is an illustration of a voltage variable null zone for FIG. 1A.
FIG. 1D is an illustration of the theoretical biased hard limiter sensitivity to threshold error for the case where the input I/C ratio was 30 dB (i.e., a factor of 1000).
FIG. 1E is an illustration of a functional block diagram of a conventional bandpass nonlinear device.
FIG. 2 is a biased inverting limiter used instead of a biased null zone limiter.
FIGS. 3A and 3B are illustrations of the envelope nonlinearity of a biased inverting limiter which can be contrasted to the characteristic shown in FIG. 1B.
FIG. 3C is an illustration of the RF nonlinearity for the biased inverting limiter design.
FIG. 3D is an illustration of the theoretical biased inverting limiter sensitivity to threshold error for the case where the input unwanted-to-wanted signal ratio was 30 dB (i.e., a factor of 1000).
FIGS. 4A-4C illustrate the ideal time-domain response of the biased inverting limiter to a sinusoidal input as a threshold V.sub.t,RF, is varied.
FIGS. 5A-5C illustrate the first-harmonic envelope nonlinearity for the biased inverting limiter.
Specifically, FIGS. 5A and 5B show a phenomenon called an "amplitude trap", which occurs when the input signal amplitude is slightly larger by a factor 2/.sqroot.3 than the inversion point in the biased inverting limiter (V.sub.t,RF in FIG. 3C), the output envelope goes to zero (G(A)=0).
FIG. 6 illustrates the unfiltered time-domain response of the biased inverting limiter to a sinusoidal amplitude input at the amplitude trap.
FIGS. 7A and 7B illustrate the fundamental frequency interference power as a function of input amplitude for the biased inverting limiter.
FIGS. 7C and 7D illustrate the third harmonic power output from the biased inverting limiter as a function of interference signal amplitude.
FIG. 8 is a block diagram arrangement illustrating a variation of FIG. 2 showing the application of the biased inverting limiter to an interference reduction device.
FIG. 9 is a block diagram of the biased inverting limiter circuit.
FIGS. 10A-10D are waveforms applicable to the circuit of FIG. 9.
Description of the Preferred Embodiment
To be of practical use, any signal processing scheme that exploits dynamic (time-varying) properties of signals must adapt to those properties. Such a scheme is termed adaptive.
Using x(t) to represent the input signal as a function of time and y(t) to represent the output signal, a nonlinearity obeys the following relationship for some time period:
for arbitrary constants b and c and distinct input signals x.sub.1 (t) and x.sub.2 (t). A nonlinear signal transformation is referred to as being "memoryless" (or as having zero memory) when the output at an arbitrary instant in time depends on the input at the same instant and not on previous values of the input.
The band-limited input to a voltage nonlinearity, x(t), is modeled in complex envelope notation as having a center frequency f.sub.o, a time-varying amplitude A(t), and a time-varying phase shift .phi.(t). To make notation simpler, the time variable may be suppressed.
A general voltage nonlinearity results in harmonic components at DC and an integer multiples of f.sub.o. There are amplitude and phase transfer functions associated with each harmonic component. These transfer functions may be denoted by G.sub.m (A) and F.sub.m (A), respectively, and each set is referred to as the "mth-harmonic envelope nonlinearity". The subscript "1" is suppressed in references to the first harmonic; if a harmonic is not specified, the first harmonic is assumed.
The unfiltered output of a voltage nonlinearity is called v(x), which is v(x(t)) or v(t) for short. Using "RE{}" to denote the real portion of a number, the signals associated with a voltage nonlinearity are expressed in terms of their complex envelope representations as: ##EQU1## This representation is based on well known theory of random signals and noise, or as otherwise explained herein.
The DC component and the mth harmonic components of v(t) for 2.ltoreq.m.ltoreq..infin. (the mth zones) are normally not desired; they are removed by a first zone bandpass filter (BPF). The output of this BPF is expressed as:
When the output amplitude is a negative number, this disclosure uses the convention of inverting its polarity and setting F(A) to 180.degree. otherwise, G(A) is left alone and F(A) is zero. Any physical device may have additional phase shift characteristic contained in the F(A) function, e.g. "AM/PM conversion" of a traveling wave tube amplifier.
The combination of the phase shift F(A) with the cascade of a voltage nonlinearity and a first zone BPF is called a "bandpass nonlinear device". FIG. 1E shows an equivalent bandpass nonlinear device model. In the figure, the bandpass nonlinear device 15 receives an input signal x(t), which is passed to an RF nonlinearity 16. The nonlinearity output v(x) is applied to a bandpass filter 17 (first zone) and is output from the bandpass nonlinear device 15. The input signal x(t) also is sent to a phase shifter 18 which converts the envelope A into a phase variable function F(A). At the output of device 15, the signal comprises the bandpass filtered first harmonic output phase shifted by F(A).
A block diagram of the Smart AGC.TM. which embodies a "biased inverting limiter" is illustrated in FIG. 2. As seen from a comparison with FIG. 1A, the arrangement and composition of system elements are basically the same, but for a substitution of one component. Specifically, the new nonlinear signal processor 10A contains an adaptive nonlinear amplifier 12A that has a biased inverting limiter and has an operational characteristic that does not contain a null zone.
The envelope nonlinearity of the biased inverting limiter is pictured in FIGS. 3A and 3B. The AM/AM transfer function, G(A), of the biased inverting limiter is illustrated in FIG. 3A and has a zero at A=V.sub.t. The nonlinearity effectively traps and eliminates the interference from the first zone while preserving the wanted signal. The AM/PM transfer function, F(A), is illustrated in FIG. 3B and causes the RF phase to reverse by 180.degree. when the input envelope crosses the threshold value of 2/.sqroot.3 V.sub.t,RF.
The biased inverting limiter differs from the biased hard limiter only in that it contains an inverted hard limiter in the location of the null zone, between input voltages of -V.sub.t,RF and V.sub.t,RF (the "RF threshold"), as seen in FIG. 3C.
The biased inverting limiter voltage nonlinearity is expressed by:
The biased inverting limiter can readily be implemented in real-time hardware, since its output consists of only two discrete voltages which can be obtained from high-speed digital logic outputs. Also, the biased inverting limiter has odd symmetry, meaning that v(x)=-v(-x) and that DC and even harmonic outputs are zero under ideal conditions.
On the assumption that input additive Gaussian noise is negligible, the unfiltered output of the biased inverting limiter, v(t), when a single CW signal, x(t)=A cos (2.pi.f.sub.o t+.phi.) is input, is given in FIG. 4A when the input amplitude is less than V.sub.t,RF. As shown in FIG. 4A, the biased inverting limiter reduces to an inverted hard limiter--the output is -1 when the input is positive and +1 when the input is negative. When the input signal amplitude is larger than V.sub.t,RF, as seen in FIGS. 4B and 4C, the biased inverting limiter goes through six transitions in the output during each period of the input signal. Hence, during one period of the input, the output is positive in three distinct time intervals and negative in three distinct time intervals. Because there are three on/off output cycles (not necessarily equal in length) for every input cycle, the third harmonic component of the output is significant.
The ratio of A to V.sub.t,RF dictates the length of each output cycle. When A is much larger than V.sub.t,RF, the voltage of the input signal is above V.sub.t,RF or below -V.sub.t,RF for an amount of time much greater than the amount of time that it is between -V.sub.t,RF and V.sub.t,RF. Hence, the device reverts to a hard limiter (output a positive constant when input is positive, output a negative constant when input is negative) when the RF threshold approaches zero.
Quantitative analysis of bandpass nonlinear devices reduces to finding the first-harmonic envelope functions G(A) and F(A) and their analogs in the higher harmonics. These functions are determined by expanding the response of the voltage nonlinearity to a CW signal of peak amplitude A and frequency f.sub.o as a Fourier series. The coefficients of the Fourier series represent the envelope nonlinearities corresponding to the given frequencies. This technique may be applied to a general voltage nonlinearity v(x). The result is the well-known Chebyshev transform: ##EQU2## where v.sub.m (A) corresponds to the envelope transfer characteristic at the mth harmonic component of the output.
Here, the amplitude transfer functions G.sub.m (A) are equal to the absolute values of v.sub.m (A). If v.sub.m (A) (for a fixed value of m) is positive, then the mth-harmonic phase shift function F.sub.m (A) is zero; if v.sub.m (A) is negative, then F.sub.m (A) is 180.degree.. Typically, the subscripts in G.sub.1 (A) and F.sub.1 (A) are .dropped. Sometimes, G(A) is called the "AM/AM transfer function" and F(A) is called the "AM/PM transfer function."
Solving equation (5), we find that the first-harmonic envelope nonlinearity for the biased inverting limiter is:
Sketches of G(A) and F(A) for the biased inverting limiter are given in FIGS. 5A-5C.
The quantity V.sub.t, is called the "envelope threshold". This point on the G(A) curve is also referred to as an "amplitude trap", since the nonlinearity traps and eliminates signals with envelopes at that point.
For the biased inverting limiter, V.sub.t is calculated by letting G(A)=0 in Equation (6b), then solving for A. This process yields V.sub.t =(2/.sqroot.3)V.sub.t,RF ; therefore, an adaptively obtained value of V.sub.t, must be attenuated by approximately 1.25 dB in order to obtain a suitable RF threshold.
The behavior of G(A) and F(A) in the vicinity of V.sub.t, provides the RFI suppression properties of the biased inverting limiter. The wanted signal causes the composite input envelope to fluctuate above and below the interference envelope; these fluctuations are retained by the nonlinearity, since the amplitude trap acts only on the larger, interfering waveform. Due to the 180.degree. phase difference between the two sides of the trap, the envelope fluctuations due to the wanted signal on both sides of the interferer are preserved-thus overcoming one of the limitations of the biased hard limiter (and all other null zone devices), which removes portions of the wanted signal.
The physical manifestation of the amplitude trap at A=V.sub.t can be understood by examining the unfiltered time-domain response v(t) of the RF nonlinearity to the CW input x(t)=V.sub.t cos (2.pi.f.sub.o t+.theta.). This is illustrated in FIG. 6.
Note that v(t) is a bipolar (i.e. no DC component) square wave that repeats every T/3 seconds, where T=1/f.sub.o is the period of the input signal. The fundamental frequency of v(t) is 1/(T/3) or 3 f.sub.o. Because the output is bipolar, there will be no even harmonic components; the harmonics occur at 3 f.sub.o, 9f.sub.o, 15f.sub.o, 21f.sub.o, and so forth. Hence, G.sub.m (A) evaluated at A=V.sub.t is nonzero for m=6k-3 (with k representing all positive integers) and zero for all other values of m, including for m=1 (the fundamental frequency of the input).
This qualitative approach is validated by evaluation of the biased inverting limiter third-harmonic envelope nonlinearity, which has been by solving equation 5.
The solution is expressed as:
Sketches of the first- and third-harmonic envelope nonlinearities are shown in FIGS. 7A-7D.
The peak of the G.sub.3 (A) curve in FIG. 7C is located at the same value of A as the first-harmonic amplitude trap. The maximum value of G.sub.3 (A) is equal to the maximum value of G(A). Therefore, the effect of the biased inverting limiter when the RF threshold is properly set is to transfer the interference energy out of the first zone so that a first zone filter can attenuate it.
The biased inverting limiter performance is shown in FIG. 3D. The best possible interference (I)-plus- intermodulation (IM)-to-carrier (C) ratio (I+IM)/C performance of the biased hard limiter is approximately 5 dB; the best (I+IM)/C provided by the biased inverting limiter is 0 dB. 0 dB is the best attainable effective (I+IM)/C with an optimum amplitude nonlinearity when the input interference is stronger than the input wanted signal and the noise is negligible; therefore, the ideal biased inverting limiter performs at the theoretical limit for an amplitude-based nonlinear device. Note, however, that 0 dB is still not the optimum level attainable by any means, since an ideal subtraction circuit (one which subtracts the CW signal) will reduce interference to zero.
When the envelope threshold is equal to the input amplitude of the interference (V.sub.t /I.sub.in =0 dB), the biased hard limiter leaves interference approximately 3 dB stronger than the remaining wanted signal and intermodulation 0.5 dB stronger than the remaining wanted signal. On the other hand, the biased inverting limiter all but eliminates the interference (output I/C=-29 dB when V.sub.t /I.sub.in =0 dB), leaving intermodulation equal in power to the wanted signal.
If the biased hard limiter is used, a tracking error that results in placement of the envelope threshold at a higher value than the peak of the composite input envelope (V.sub.t /I.sub.in .gtoreq.0.27 dB for an input I/C of 30 dB) will destroy both interfering and wanted signals. The biased inverting limiter does not share the problem; even when the error in estimating the envelope threshold is as much as 0.5 dB in either direction, the biased inverting limiter improves (I+IM)/C by 18 dB (as opposed to 13 dB of improvement when the null zone threshold is set 0.5 dB too low).
The ability of the biased inverting limiter to suppress strong interference has motivated its use in a way which improves upon the previous Smart AGC.TM. approach.
In FIG. 8, an input is provided to two parallel paths leading to a single output. The first path comprises a delay 21 and a linear booster amplifier 22. The second path comprises an envelope detector 23, whose output is provided in parallel to a tracker 24 and a delay 25. The tracker 24 outputs to an envelope to RF conversion circuit 26 which itself outputs to an RF threshold control circuit 28. The delay 25 outputs to a comparator 27, which compares the envelope to a voltage called "max wanted signal". The purpose of the comparator is to bypass the use of a voltage window (V.sub.t,RF) whenever the absolute input envelope is so small that the presence of RFI becomes ambiguous. Thus, the RF threshold control circuit 28 output will be either the filtered envelope or zero. A zero output means the estimated RF is too small to provide a correction. The circuit 28 will input in common to a biased inverting limiter 29 together with the output of the linear booster amplifier 22. A first zone filter 29 is provided before the output of the stage to limit the band of operation to the vicinity of the input band.
The mechanism that sets V.sub.t,RF from estimation of the interference amplitude 23, 24, 26, 27, 28 requires time to operate and as such must be equalized with delays in parallel paths 21, 25. Present-day high-frequency delay devices are lossy; hence, a linear booster amplifier 22 is required in the RF path. To ensure that the net insertion losses encountered by the tracking path and the RF path are as equal as possible, power compensation, in the form of either an amplifier or an attenuator, must be included in the envelope tracker 24.
After the tracker has been power-compensated, the voltage along the tracking path is V.sub.t ; an additional 1.25 dB of attenuation is required to convert this to the RF threshold, since V.sub.t,RF =(.sqroot.3/2)V.sub.t. This factor of 1.25 dB can easily be accounted for in the power compensation at the output of the tracking filter; it has only been separately noted as "envelope to RF conversion" 26 for the purpose of illustration. The detected amplitude of the input signal-plus-interference-and-noise (the output of envelope detector 23) is compared with a fixed maximum amplitude-determined from a priori knowledge of the wanted signal--to ascertain whether the interference is sufficiently high for the biased inverting limiter to have a beneficial effect. If the comparator 27 indicates that the interference does not exist or is not powerful enough to significantly affect the composite envelope, the RF threshold control circuit 28 sets V.sub.t,RF to zero; otherwise, the control circuit 8 passes the value of V.sub.t,RF estimated by the tracking path to the biased inverting limiter circuit 29. The biased inverting limiter circuit processes the RF signal and threshold estimation, then passes the output to the first zone bandpass filter 30.
A functional block diagram of the biased inverting limiter circuit is given in FIG. 9. There the RF input is provided to the emitter-coupled logic (ECL) window comparator 31 and the ECL hard limiter 32 in parallel. The outputs from the two ECL devices are applied to an exclusive NOR 33 and then to an ECL-to-bipolar conversion unit 34 before being output as an RF signal. There are three ECL voltage comparators used in the circuit. Ideally, the comparators turn on (output a "logic 1" which is typically -0.9 volt in ECL) when the voltage at the noninverting (positive) input terminal is greater than the voltage at the inverting (negative) input terminal, and turn off ("logic 0" or typically -1.7 volts) when the reverse is true. In this mode, the comparators function as one-bit analog-to-digital converters.
A combination of wanted and interfering RF signals that are terminated with purely resistive 50-ohm loads to ground (the "RF input") is applied to the noninverting terminals of three voltage comparators (two of these comprise the "ECL window comparator" unit 31 having an output I/O characteristic shown in FIG. 10A; the third is the "ECL hard limiter" unit 32). The inverting terminals are biased as follows.
The inverting terminal of the ECL hard limiter 32 is grounded; hence, it has the voltage transfer characteristic (VTC) as shown in FIG. 10C, behaving as a hard limiter that has been attenuated by approximately 8 dB and has a negative DC offset of approximately 1.3 volts.
One of the comparators in the "ECL window" is biased with a DC voltage, V.sub.t,RF, that is proportional to the interference amplitude. The other comparator in the ECL window is biased with -V.sub.t,RF. The logical complement of the negatively biased comparator output is tied to the output of the positively biased comparator. A third wire that extends from the connection of the two comparator outputs carries the logical "OR" of these outputs (known as a "wired OR" because of the absence of a gate) which is shown above the "ECL window comparator" unit 31 as a function of instantaneous RF input voltage.
The outputs of the window comparator 31 and ECL hard limiter 32 are then passed through an "exclusive NOR" (XNOR) gate 33, which returns a logic 1 when both inputs are the same and a logic 0 when the inputs differ from each other. The resulting VTC is shown in FIG. 10B.
The XNOR output waveform is converted to a bipolar signal by a capacitor that blocks its DC component 34. The design result, shown in FIG. 10D, matches the ideal VTC except that the design result has approximately 8 dB (20 log.sub.10 {1/0.4}) less gain. This reduction in gain was expected to influence the wanted signal and interfere equally and, thus, have no effect on interference suppression performance.
When the biased inverting limiter is integrated into an RFI suppression system, the RF threshold is set by dynamic tracking of the interference envelope. For this circuit, a tracker may be used or the mechanism that sets V.sub.t,RF for the biased inverting limiter prototype may be manually controlled. In the latter case, it is a resistive voltage divider that consists of a "pull-down" resistor connected to the positive DC voltage that powers the integrated circuits (+5.0 v) and, through a potentiometer ("pot"), to ground. V.sub.t,RF, the voltage across the pot, can be varied between zero and 1.7 volts. The additive inverse of V.sub.t,RF is obtained from an operational amplifier connected in an inverting loop with two equal feedback resistances.