You are on page 1of 4

The Origins and Modeling of Co-Site Interference in Military and Commercial Radios

Michael B. Steer, Nikhil M. Kriplani, Kevin G. Gard, Jie Hu and Gregory J. Mazzaro
Department of Electrical and Computer Engineering North Carolina State University Raleigh, NC, USA, NC27695-7914 Abstract: Co-site interference in communications system results when a third radio interferes with the link between a transmitter and an intended receiver. Interference results when the interfering signal is in the same band as the intended communication but also results from timefrequency transients, frequency pulling, and multiple mixing, and possibly re-radiation, in the RF front end. New phenomenology related to time-frequency effects is presented here that explains origins of some forms of cosite interference, particularly in frequency hopping communication systems. Keywords: radio; interference; communications. Introduction Co-site interference has been notoriously difficult to conceptualize, predict and model. In part this is because communication devices are designed to mitigate the effects of interference. As a result co-site interference is in part a result of second-order circuit effects. At the same time it is computationally prohibitive to model the receivers at the circuit level. Rarely is a circuit-level model available and even then unintended coupling such as electromagnetic leakage within the RF front end, spurious passband responses of filters, and frequency pulling sensitivities of oscillators are not captured in typical design data. For both speed and model fidelity reasons, behavioral models are required. These models capture the non ideal effects. For example, a suitable behavioral filter model incorporates spurious passbands. Circuit-type models capture bi-lateral effects important in capturing time-frequency responses. In contrast a MATLAB/Simulink/SystemVue type simulation typically captures unidirectional effects only. The paper reviews time-frequency effects that can lead to unexpected co-site interferences; and demonstrate what can be achieved using bilateral behavioral modeling. Radio Architectures Commercial radio architectures support communication over at most a few narrow frequency bands enabling them to use fixed filters. This has advantages in managing dynamic range and suppressing unwanted radiation from the radio. However military radios often use doubleconversion architectures and bi-lateral mixers. When compared to commercial radios, these encounter more significant dynamic range issues, more likely to be susceptible to co-site interference, and are more likely to generate spurious emissions. It is necessary to model these radios with signals at many different frequencies. Military radios are evolving potentially employing tunable filters and replacing diodes in mixers with integrated circuits which enable direct conversion and low-IF conversion. Still the issue of signals over a very broad range of frequencies impacting the radio is present. Consider the bi-lateral double-conversion transceiver circuit shown in Figure 1(a) which is a typical front end used in wide-band military radios. This circuit has fixed bandpass filters but selectable local oscillators of frequency fLO1 and fLO2. The first mixer up-converts the signal so that the much higher Qs of filters at microwave frequencies can be used to realize filters with very steep skirts. The heterodyne stages themselves are narrow band using ringdiode mixer circuit that has the advantage of being bilateral. The filters are designed based on low frequency prototype circuits which have the bandpass forms shown in Figure 1(b). Design proceeds by designing the radio in the steady-state. However, in reality, there are appreciable turn-on transients as can be seen in Figure 2 (the waveform is at the output of the first mixer). A waveform with a turnon transient such as this has additional frequency content including content that comes from the filters in addition to mixing from the nonlinear diodes. Time-Frequency Effects RF engineering proceeds by conceptualizing circuits in the frequency domain. Very little thought is given to the transient response, or what should really be described as the time-frequency response. This has been found to be significant with radios as the high Q filters almost invariably employed have two affects that produce sometimes surprising responses. The first phenomenon is that filters take a long time to reach steady state, and a longer time to discharge. This is not a charging or discharging response with RC-like circuits but instead refers to energy that is stored on a bandpass filters internal resonators. The second phenomenon is that the group delay across the filter bandwidth varies, this is particularly pronounced with Chebyshev filters. These two effects produce what can be viewed as anomalous responses.

(a)

(c)

(d)

Figure 1. Architecture of the RF front end of a typical military radio: (a) Double conversion architecture bandpass filter number 2, BPF#2, has a center frequency considerably above the frequency of the input signal; (b) details of the mixer realized as a ring-diode mixer; (c) realization of BPF#2 as a third-order lumped-element filter; and (d) model of a diode comprising an ideal diode with a series resistance and shunt capacitance.

(a)

Figure 2. Waveform at the output of the first mixer in Figure 1(a) resulting from a pulsed RF signal at the input to the radio. The LO is 1.1 GHz and the RF is 700
MHz for an output IF of 400 MHz. Frequencies are staggered so that the waveform can be viewed.

(b)

As an example consider the bandpass filter shown in Figure 3(a). This is a 7 th order Chebyshev filter which is the typical filter in the front-end of a cellular phone. The Chebyshev filter as the necessary sharp circuits required to reject close-in out-of-band signals. This circuit has been modeled and measured with similar responses. The circuit is driven by two-RF pulses. First one at 988 MHz which switches to another at 998 MHz one microsecond latter.

(c)

Figure 3. Filter response: (a) a 7 th order Chebyshev bandpass filter as is typical in a cellular phone; (b) amplitude and group delay response of the filter; and (c) transient response to sequential RF pulses.

The filter in Figure 3 is terminated in a linear resistor so the output response in Figure 3(c) is entirely due to linear effects. What is happening here is the first pulse applied to the filter has a long group delay relative to the second RF pulse. As a result there is a period of overlap, most noticeably from 1.1 to 1.5 microseconds when the two tones are present simultaneously at the ouput. Co-Site Interference Figure 4 presents a time-moving Fourier analysis of the signal shown in Figure 3. What is clear here is that the frequency content apparently shifts starting at the frequency of the first pulse, 988 MHz, and transitions to the frequency of the second pulse at 998 MHz. During the transient, the frequency goes up to what is considered to be the third-order intermodulation frequency of a mixing process. However, all process are entirely linear. Thirdorder intermodulation response is regarded as the measure of a nonlinearity and indicates the level of spurious tones generated. In radio systems this is interference that can affect the reception by other radios. The waveform and Fourier analysis reported here are for the transmission response but similar results are obtained for the reflected signal. In this case the third-order frequencies are retransmitted and can result in co-site interference.

being discrete as would be expected from the application of two tones. The spurious tones (in addition to the two applied tones) have appreciable levels at frequencies considerably distant from the pulse frequencies. For example, in WCDMA radios the spurious tones in the channels adjacent to the channel being used must be 80 dB down to avoid interference with pother radios. Clearly the level of the spurious tones shown in Figure 5 exceeds this level.

Figure 5 The reflected response from a handheld communicator excited two sequential tones. Co-Site Interference in Frequency Hopping Systems Recently we have used the above analysis to analysis cosite interference in frequency hopping communications systems [1] which are commonly employed in ad-hoc networks such as military radio networks and WiFi networks. The particular property here of concern is the long persistence of the RF energy in a filter. When the RF excitation energy is removed from a bandpass filter, the internal filter resonators hold considerable RF energy. This energy takes a very long time to dissipate, much longer than the inverse of the bandwidth of the filter would indicate. The phenomenology is similar to what has been described earlier in this paper. The effect is that, considering a particular frequency band, the residual energy from one user overlaps the time slot of next user. Again an effect that could not be predicted from steadystate considerations. The explanation for the apparent discrepancy is to consider that a 7 th order bandpass filter is described by a 14 th order differential equation [2]. Solution of such an equation consists of what can be interpreted as a large number of staggered time constants. It is not possible for a human to intuitively understand the transient response of a 14-th order differential equation. The time-domain response is an exponential like transient but with what is known as a

Figure 4. Time moving Fourier analysis of the waveform showing Figure 3 (c). Considering the more realistic case in a radio of an amplifier on the other side of the filter. The nonlinear mixing at the input of the amplifier results in the spectrum of the response becoming richer. The measured reflected response of a commercial handheld radio system with a filter and amplifier (with a nonlinear input is shown in Figure 5. The third-order frequencies can be clearly identified and now the spectrum is continuous rather than

long tail. The response takes a very long time to reach a steady-state value. Instead humans must work with a few parameter approximation. The suitable approximation is fractional calculus [37]. Here, instead of dealing with the Laplace variable, s, and its integer powers (up to s14), a non-integer power of s is considered (say s where could be less than 1). Conclusion The aim of this paper was to illustrate that the steady-state view of radio operation will not lead to a complete understanding of all sources of co-site interference. Co-site interference can result from time-frequency effects that occur when the excitation tone is not steady-state. It is believed that the time-frequency effect outlined here illustrates a possible source of co-site interference that cannot be anticipated from frequency combinations of RF tones and local oscillator mixing. Acknowledgement This material is based upon work supported by the U.S. Army Research laboratory and the U.S. Army Research Office as a Multi-disciplinary University Research Initiative on Standoff Inverse Analysis and Manipulation of Electronic Systems under grant number W911NF-05-10337.

References 1. G. J. Mazzaro, M. B. Steer, K. G. Gard, and A. L. Walker, Response of RF networks to transient waveforms: interference in frequency-hopped communications, IEEE Trans. Microwave Theory and Tech., December 2008. 2. G. J. Mazzaro, M. B. Steer, and K. G. Gard, Filter Characterization Using One-Port Pulsed RF Measurements, Institute of Engineering and Technology Microwaves, Antennas, & Propagation, In Press. J. R. Wilkerson, K. G. Gard, A. G. Schuchinsky, and M. B. Steer, Electro-thermal theory of intermodulation distortion in lossy microwave components, IEEE Trans. Microwave Theory and Tech., December 2008. K. B. Oldham and J. Spanier, The Fractional Calculus. New York: Academic Press, 1974. H. S. Carslaw and J. C. Jaeger, Operational Methods in Applied Mathematics. New York: Dover, 1963. I. Podlubny, Fractional Differential Equations. New York: Academic Press, 1999. K. S. Miller and B. Ross, Fractional Calculus and Fractional Differential Equations. New York: Wiley, 1993.Nadell, R.,

3.

4. 5. 6. 7.

You might also like