PIERS ONLINE, VOL. 7, NO. 4, 2011 380 The Minimum Phase Nature of the Transfer Function of the Impulse Radiating Antenna J. S. McLean1 , R. Sutton1 , and H. Foltz2 1 2 TDK R&D Corp., USA University of Texas — Pan American, USA Abstract— The Impulse Radiating Antenna, a reflector antenna employing a TEM feed, has been shown to provide excellent time-domain pulse reproduction on its principal axis. We examine the on-axis response of a representative, well-designed Impulse Radiating Antenna in the frequency- and time-domains and show that it is the existence of the pre-pulse in the time-domain response that causes the frequency-domain transfer function to be non-minimum phase in nature. Moreover, the hypothetical time-domain response corresponding to the minimum phase frequency domain transfer function derived in turn from the magnitude of the on-axis response of the impulse radiating antenna is essentially identical to the actual time domain response except that the pre-pulse occurs after the main pulse. That is, the time domain response associated with the minimum phase transfer function appears almost as a mirror image of the actual time domain response with the symmetry occurring around the center of the main pulse. 1. INTRODUCTION The Impulse Radiating Antenna (IRA), a reflector antenna employing a TEM feed structure [1], has been shown to provide excellent time-domain pulse reproduction on its principal axis; more specifically as stated in [1], “a step-like signal into the antenna gives an approximate delta-function response in the far field.” While this statement succinctly describes the ideal time domain behavior, the frequency domain counterpart is that perfect time domain pulse reproduction requires satisfaction of the distortionless transfer function criterion in the frequency domain. Distortionless transfer functions, in turn, are a small subset of a more general group, minimum-phase transfer functions. When a minimum-phase network exhibits a transfer function magnitude which is nearly constant with frequency, its associated phase function necessarily satisfies the distortionless transfer function criterion; however, many broadband antennas and other physical systems exhibiting flat or nearly flat transfer functions are not minimum-phase. An example of such a broadband, non-minimum phase antenna is the Log-Periodic Dipole Antenna (LPDA), the transfer function of which exhibits a nominally flat magnitude, but deviates greatly from minimum phase behavior [2]. It is well known that this antenna exhibits poor time domain pulse fidelity. 2. FREQUENCY DOMAIN TRANSFER FUNCTION ~ gives the far field radiated The frequency domain complex vector antenna transfer function H ~ electric field E in terms of the power-normalized incident voltage a at the antenna input port [3]: −jβR ~ (R, θ, φ, ω) E ~ (θ, φ, ω) e = −jω H a (ω) . √ η0 2πRc0 (1) This definition explicitly includes a differentiation of the transmitted signal through the jω factor1 . Thus, a hypothetical antenna with a unity transfer function for all frequency would still differentiate the incident signal in the sense that the radiated electric field would resemble the time derivative of the input voltage. Thus, the inverse Fourier transform of the antenna transfer function may be called antenna impulse response, but it actually gives the far field electric field due to a step incident voltage. Note that the definition for antenna transfer function given in Eq. (1) has units of meters and is identical to the frequency domain counterpart of the normalized impulse response defined in [4, 5]. 1 Strictly speaking, the antenna transfer function does not directly yield the electric field due to the incident voltage; therefore the “impulse response” derived from it does not give the electric field due to a voltage impulse at the input. The definition was arrived at by several different groups of researchers and is a compromise that provides some symmetry between forms for reception and transmission [3]. However, the authors note that there are other conventions that have been proposed. PIERS ONLINE, VOL. 7, NO. 4, 2011 381 However, the differentiation in Eq. (1) is distinct from the intrinsic high-pass nature of the impedance matching and power transfer characteristics of any finite-sized antenna. A finite size antenna driven by a step input cannot sustain a DC electric field that decays as 1/R in the far-field; ~ must have at least one zero at ω = 0. This can also be therefore, one can conclude from (1) that H seen in the relationship between the complex vector effective length ~heff and the antenna transfer function: √ η0 Z0 ~ ~ (θ, φ, ω) = H heff (ω) (2) Z0 + ZA (ω) where Z0 is the normalizing impedance for the antenna input port and ZA is the antenna input impedance. The frequency domain transfer function must exhibit at least one zero at DC. Some antennas (for example, a short monopole) have an effective height that is constant and finite in the limit as ω → 0; however, such antennas have an input impedance ZA that becomes infinite at DC. The authors are unaware of any antenna that has both nonzero effective height and finite impedance at DC. The conclusion is that the relationship between electric field and input voltage must always have at least two zeroes at ω = 0, and therefore the response to a step input must be zero average in the time domain. The Impulse Radiating Antenna is primarily a reflector antenna. However, the TEM feed structure itself also radiates and behaves essentially as a P × M antenna. It can be shown that one type of canonical P × M antenna sometimes referred as a balanced transmission line wave (BTW) sensor [6], a backward-radiating, terminated uniform transmission line, has an asymptotic slope of 12 dB/octave in its gain, and thus has an asymptotic slope of 6 dB/octave in its transfer function, indicating a single zero at ω = 0. The Impulse Radiating Antenna can also be shown to exhibit a single zero at DC in its effective length and transfer function. Both the balanced transmission line wave sensor and the IRA are very well matched due to the internal loads and the complex voltage division term in (2) is essentially frequency independent. Thus, when relating the incident input voltage to the far field radiated electric field the high-pass differentiation appears twice: once in the transfer function (and thus the effective height), and again due to the jω factor in Eq. (1) for the electric field. An antenna with a frequency-domain transfer function (as defined in [3]) of unity for all frequency would generate an electric field impulse in the far field in response to a step input, and is said to satisfy the distortionless transfer function criterion. The IRA nearly satisfies the distortionless transfer function criterion and thus is nearly minimum phase in the frequency domain. However, the small departure of its transfer function from minimum phase is quite interesting. Close examination of the time domain response of the IRA reveals a pre-pulse as well as a long shallow tail following the main pulse [1]. The pre-pulse has been shown to be step-like in the time domain. It radiates from the TEM feed structure and necessarily precedes the main pulse from the Figure 1: Schematic drawing of an impulse radiating antenna. The reflector employed by the Farr IRA-3 has a diameter of D = 46 cm and focal length of F = 23 cm and thus F/D = 0.5. PIERS ONLINE, VOL. 7, NO. 4, 2011 382 reflector. It can be shown that having two radiation mechanisms (in this case direct feed radiation and the reflector radiation) with different delay times leads to non-minimum phase behavior if the smaller signal is the first to arrive in the far-field. The lack of minimum phase behavior in turn implies that the distortionless criterion cannot be satisfied, and furthermore, that it cannot be corrected through passive equalization. As discussed above, Eq. (1) relating the electric field to the incident voltage must contain at least two zeroes at DC, and the entire time domain response must average to zero. In [1] some discussion is given concerning how if the area of the pre-pulse could be made to equal that under the main impulse (thus giving zero DC average) then the tail of the response would be small. The area of the pre-pulse is tailored by adjusting the characteristic impedance of the TEM feed of the IRA. Thus, the canonical or ideal response of the Impulse Radiating Antenna is a step followed immediately by an impulse of equal area. The particular IRA characterized here, the Farr Research IRA-3, exhibits a very good on-axis, time-domain response with a very minimal tail. 3. MEASUREMENT The port-to-port forward transfer scattering parameter of combination of a Farr Research IRA-3 impulse radiating antenna and a Farr Research TEM-1 TEM horn was measured in a fully anechoic chamber using an automatic vector network analyzer. The analyzer was calibrated using a so-called transmission calibration; that is, a full two-port calibration was not used due to the very long coaxial cables connecting the network analyzer, which was located outside the chamber, to the antennas. The magnitude of the measured transfer scattering parameter is shown in Fig. 3. Figure 2: Solid model of impulse radiating antenna. A bifurcating ground plane located in the x-z plane has been omitted for clarity. -20 S21 Magnitude (dB) -30 -40 -50 -60 -70 -80 -90 -100 0 5 10 Frequency (GHz) 15 20 Figure 3: The magnitude of the measured transfer scattering parameter S21 of the 2-port network comprised by a Farr Research IRA-3 impulse radiating antenna and a Farr Research TEM-1 TEM horn situated in an anechoic chamber with principal axes aligned and with 3.5 meters separation. PIERS ONLINE, VOL. 7, NO. 4, 2011 383 Transfer Function Magnitude (dB re 1m) 0 -10 -20 -30 -40 -50 0 5 10 Frequency (GHz) 15 20 Figure 4: Magnitude of the frequency domain transfer function of the Farr IRA-3 impulse radiating antennas as derived measured port-to-port insertion loss data. 0 Farr IRA-3 Canonical PxM -10 20*log10|H| -20 -30 -40 -50 -60 1 10 2 10 Frequency (MHz) 3 10 Figure 5: Magnitude on-axis transfer function of a canonical P × M antenna (length 23 cm) and a numerical model of the Farr Research IRA-3 impulse radiating antenna. As expected the slope of the transfer function is asymptotically = 6 dB/octave at the low end of the frequency range. The TEM horn had been previously characterized as described in [8]. From the transfer scattering parameter measurement, the transfer function of IRA was then determined. The magnitude of the frequency domain transfer function is shown in Fig. 4. The transfer function data given here agrees reasonably well with the data provided by the manufacturer. 4. LOW FREQUENCY EXTRAPOLATION OF MEASURED DATA The IRA transfer function magnitude shown in Fig. 4 appears to reach an approximately constant value at low frequency. However, as discussed in Section 2, it is known that there is at least one zero at DC, and therefore one can conclude that there is also a pole at a low but finite frequency. S21 in a two-antenna measurement has at least three zeroes at DC (one in each of the transfer functions plus the jω factor in Eq. (1)), and thus the measured signal decreases very rapidly at low frequency. In the data presented here, unavoidable noise overcomes the measured signal before the pole in the IRA transfer function can be seen. To study the minimum phase behavior of the IRA, a Hilbert transform is performed to derive the minimum phase function from the transfer function magnitude, which requires in principle integration over all frequency. Explicit or implicit (simple truncation) extrapolation of the finite frequency range is necessary. The value of the minimum phase function as ω → 0 is determined by the limiting slope at low frequency. Due to the noise problem described above, the low frequency behavior must be extrapolated. PIERS ONLINE, VOL. 7, NO. 4, 2011 384 100 80 Phase of H (degrees) 60 40 20 0 -20 -40 -60 -80 -100 1 10 2 10 Frequency (MHz) 3 10 Figure 6: Simulated phase of far-field on-axis electric field of Farr IRA-3. A factor of exp(−jkR) has been removed from the data. As can be seen the phase of the transmitted electric field with respect to the source asymptotically approaches 90◦ at low frequencies. This is consistent with there being one zero at DC in the transfer function. The operation of the IRA in the lowest registers of its operating frequency range is similar to that of the so-called P × M antenna or Balanced Transmission Line Wave (BTW) sensor [6]. The antenna exhibits a pattern which is a cardioid of revolution about the principal axis and has an effective length and transfer function the magnitudes of which increase 6 dB/octave asymptotically. The power gain increases 12 dB per octave asymptotically. Fig. 5 shows data for the Farr IRA3 generated using a commercial FEM simulator, Ansoft’s HFSS software, with comparison to a canonical P ×M antenna. In this numerical simulation the equal-delay balun was not modeled. The canonical P × M antenna is a terminated two-wire transmission line with characteristic impedance of 450 Ohms and terminated in a matched load. The length of the line is 246 mm corresponding to the focal length and hence feed dimension of the Farr IRA-3. The data was obtained using a numerical model implemented using the Numerical Electromagnetics Code (NEC). Note that both antennas show 6 dB/decade slope in the magnitude of their transfer functions in low portion of the frequency range. Since both antennas are very nearly perfectly matched at the low ends of their respective operating frequency ranges, it can be shown that the antenna transfer function as well as the traditional effective height rolls off with 6 dB/octave. Thus, the extrapolation of the transfer function to DC requires that there be exactly one zero at the origin. 5. MINIMUM PHASE TRANSFER FUNCTION The minimum-phase quality of the transfer function of an antenna is associated with the propagation of energy through the system. Having a single path through the network or system is a sufficient condition to have minimum-phase behavior. It was shown in [7] that the broadband, double-ridged horn is very nearly minimum phase on its principal axis, but deviates from this condition off-axis. It was surmised that this was due primarily to interference between the direct radiation from the horn’s aperture and fields diffracted by the edge of the horn. In [8], it was shown that an asymmetric or half TEM horn such as the Farr Research TEM-3 exhibits a minimum phase response on its principal axis as well as off axis in the E-plane for angles below the ground plane, but is not minimum phase off-axis above the ground plane. This is because below the ground plane the radiation is essentially entirely due to a single mechanism, diffraction from the edge of the ground plane. The question in this case is whether a similar multiple path effect applies to the IRA. In Fig. 7 we show the measured transfer function phase for the IRA in comparison with the minimum phase function for the antenna. The minimum phase function was computed using the Bode-Hilbert transform of the measured magnitude data, but with the data truncated at 300 MHz and replaced with extrapolated points computed using the 6 dB/octave slope as discussed in Section 4. It can be seen that there is good agreement above 5 GHz, but increasing non-minimum phase behavior below 5 GHz. This is evidence that there are multiple radiation mechanisms. PIERS ONLINE, VOL. 7, NO. 4, 2011 385 Transfer Function Phase (degrees) 90 Black- Measured Blue - Hilbert 0 -90 -180 0 5 10 Frequency (GHz) 15 20 Figure 7: Measured phase of transfer function and minimum phase function computed from the magnitude of the measured transfer function. 6.0E9 Black - Measured Blue - Hilbert Impulse Response (m/s) 5.0E9 4.0E9 3.0E9 2.0E9 1.0E9 0.0E0 -1.0E9 -5 0 5 10 15 20 Time (ns) Figure 8: The impulse response of the IRA as derived from frequency-domain data measured in an anechoic chamber (black). For comparison, the impulse response derived from a hypothetical transfer function with the same magnitude but Hilbert minimum phase (blue). 2.0E9 Black - Measured Impulse Response (m/s) 1.5E9 Blue - Hilbert 1.0E9 5.0E8 0.0E0 -5.0E8 -1.0E9 -4 -2 0 Time (ns) 2 4 Figure 9: Same data as Figure 8, on an expanded scale. Note that the rectangular pre-pulse that precedes the main pulse in the measured data. If the transfer function is modified to be Hilbert minimum phase, the pre-pulse is transposed such that it follows the main pulse in the response derived from the minimum phase transfer function. Figures 8 and 9 show, for the curves labeled “measured” data, the inverse Fourier transform of the measured transfer function. For these curves no extrapolation or windowing was used. The shapes correspond to the electric field that would be produced by a step input, which is the intended PIERS ONLINE, VOL. 7, NO. 4, 2011 386 mode of operation for the IRA. The negative going pre-pulse is clearly evident starting 1.6 ns prior to the main pulse. The step-like appearance of the pre-pulse leads one to believe that it is associated with a pole at low frequency. Since the deviation from minimum phase behavior is at low frequency, one might guess that the pre-pulse is the most important cause of the deviation. As further confirmation, one can carry out a “thought experiment”, in which a new hypothetical transfer function is generated, consisting of the actual magnitude from measurements, with the minimum phase function derived from the Hilbert transform. This hypothetical transfer function is then used to generate a time response via an inverse Fourier transform. The same procedure was used in [8] to analyze a TEM horn, and it was seen that in directions where the measured transfer function was not minimum phase, the new hypothetical time-domain response differed from the actual response in that the precursors to the main pulse in the actual data were transposed to the opposite (later) side of the main pulse, thus providing a clean onset to the main pulse. Thus the precursors could be seen to be directly responsible for the deviations from minimum phase. A similar computation was carried out for the IRA. The results are the blue curves labeled as “Hilbert” in Figs. 8 and 9 It can be seen that in the time-domain response computed from the minimum phase function, the pre-pulse is transposed about the main impulse such that it later. This behavior can be qualitatively explained in terms of a simple model based on rays representing major sources of radiation. In general, the frequency-domain minimum-phase criterion will be satisfied only if the time-domain field from the strongest radiation source is the first to arrive at the observation point. Enforcing the minimum phase condition in the frequency domain re-orders the time-domain response such that the main impulse is first. 6. CONCLUSIONS The impulse radiating antenna exhibits a transfer function that is essentially minimum phase at high frequency, but has a small but significant deviation from minimum phase at low frequency. By comparing the actual impulse response to a hypothetical impulse response for a transfer function that is modified to be minimum phase, it can be seen that the pre-pulse due to direct radiation from the feed is the most prominent cause of the non-minimum phase behavior. ACKNOWLEDGMENT This work was supported in part by the Army Research Office through grant W911NF-06-1-0420. REFERENCES 1. Baum, C. E., E. G. Farr, and D. V. Giri, “Review of impulse-radiating antennas,” Review of Radio Science (W. S. Stone, ed.), Chapter 12, 1996–1999, Oxford University Press, 1999; Also presented at the 1999 URSI General Assembly, Toronto, Canada, Aug. 1999. 2. McLean, J. S. and H. Foltz, “Minimum-phase / all-pass decomposition of LPDA transfer functions,” Proc. of the 2009 IEEE International Conference on Ultra-Wideband, (ICUWB 2008), Vancouver, Canada, Sept. 2009. 3. McLean, J. S., R. Sutton, A. Medina, H. Foltz, and J. Li, “The experimental characterization of UWB antennas,” IEEE Antennas and Prop. Magazine, Vol. 49, No. 6, 20–30, Dec. 2007. 4. Farr, E. G. and C. E. Baum, “Extending the definitions of antenna gain and radiation pattern into the time domain,” Sensor and Simulation Notes, Note SSN 350, Nov. 1992. 5. Farr, E. G. and C. E. Baum, “Time domain characterization of antennas with TEM feeds,” Sensor and Simulation Note, 426, Oct. 1998. 6. Farr, E. G. and J. S. Hofstra, “An incident field sensor for EMP measurements,” Sensor and Simulation Note, 319, SSN 319, Nov. 6, 1989. 7. McLean, J. S., H. Foltz, and R. Sutton, “The directional dependence of the minimum-phase property of the antenna transfer function,” Proc. Loughborough Antenna Conference, Loughborough, UK, Nov. 2009. 8. McLean, J. S., A. Medina, R. Sutton, and H. Foltz, “Directional dependence of the minimum phase property of the TEM horn transfer function,” 2010 AMEREM Conference, Toronto, Canada, Jul. 2010; Also submitted to UWB-SP-2010.