A new generation of miniature microwave tuners.
There is increased demand for microwave tuners capable of supporting multiple operational functions. These functions range from communications to radar, as evidenced by requirements for low phase noise, low group delay distortion, wide instantaneous bandwidths (0.5 to 2 GHz are common) and fast tuning speed in one tuner. The tuner is required to be sufficiently versatile to feed diverse specialized downstream equipment that define operational characteristics of the overall receiver system.
In addition to these performance demands, the overwhelming pressure in today's procurement environment is to reduce equipment cost and to deliver to accelerated schedules. This leads manufacturers to attempt to develop core technology that will serve multiple applications.
This article will present microwave tuner designs that address three different sets of operational requirements, yet share modular circuitry. Tuners typically convert RF and microwave frequencies (a few MHz to 40 GHz) to IF frequencies (often in the range of 70 MHz to 5 GHz, depending on application) for further processing. The majority of SIGINT receivers use microwave tuners in the input chain. The complete receiver may include channelizers, IFM subsystems, direct digitizers, and a variety of downstream demodulating and signal processing equipment. The first tuner to be presented maintains high dynamic range while convening a tunable 1,100 MHz instantaneous bandwidth within 2 to 18 GHz to a 5-GHz IF output, and is packaged in a custom SEM-EX configuration (58 [in..sup.3] and 4 lb). The second tuner, a VXI- configuration tuner (180 [in..sub.3] and 7.7 lb), covers 0.01-40 GHz and processes 1-nsec pulses with minimum distortion. A third, more general purpose tuner will also be covered.
The basic function of a microwave tuner is to translate a portion of the microwave spectrum to an output intermediate frequency for signal processing (for example, any 1-GHz section of the spectrum within the limits of 2 to 18 GHz to an output centered at 5 GHz). Small size and light weight are frequently required for mobile applications. Critical specifications that often distinguish tuners are listed in Table 1, including specific parameters representative of three fairly diverse applications.
The top five electrical performance/cost drivers are typically 1) frequency coverage, 2) dynamic range figure-of-merit (input intercept point minus noise figure - see discussion below), 3) phase noise, 4) tuning resolution and 5) tuning speed.
HIGH DYNAMIC RANGE TUNER DESIGN
"Dynamic range" is gaining in importance as the number of signals present has increased for modern environments. A problem with dynamic range is that it is highly dependent on system parameters and the particular definition employed, and it is therefore difficult to compare alternative tuners. In order to evaluate different tuners or variations of the same tuner architecture, a figure of merit was devised. It is defined as:
Dynamic Range Figure of Merit (FOM) = Input IP3 - Noise Figure
where Input IP3 = Input third-order two-tone intercept point (in dBm).
Note: Input IP3 = Output IP3 - Tuner Gain Noise figure is in dB.
The figure of merit improves with a lower noise figure (better sensitivity) or higher third-order two-tone intercept point (reduced spurious levels). It is always referenced to the system input (the only meaningful comparison when complex chains of non-linear elements are involved, such as in receivers). Commercially available microwave receivers have typical figures of merit in the range of -25 to -10. Very high dynamic range tuners require a figure of merit of 0 (or greater). In other words, the noise figure (in decibels) objective is to be lower than the two-tone third-order input intercept point (in decibel/milliwatt).
Dynamic range FOM is a clear differentiator for the three tuner examples given in Table 1. It is also useful for comparing tuners from different manufacturers, since nearly all tuner specifications include noise figure and third-order two-tone intercept point (it may be necessary to convert an output IP3 to input IP3 by subtracting tuner gain).
Three critical factors influenced our choice of tuner architectures: high dynamic range FOM, small size and the need to achieve application versatility. The approach chosen was to upconvert the input RF frequency segment of interest to a first intermediate frequency (IF) above the maximum input frequency, using a high-side local oscillator (LO) frequency. The technical challenge was to develop component and interconnection technology that would allow its practical realization with LO frequencies to 40 GHz and a first IF above 20 GHz.
This general architecture places preselector filters as the first element in the tuner chain, followed by low gain (typically 10-15 dB), high-intercept preamplifiers, followed by an upconverting mixer, amplification and filtering at the first IF (nominally 22.5 GHz), a downconverting mixer and, finally, second IF amplification. This architecture involves the fewest mixers possible for situations requiring coverage of such wide input frequency ranges while simultaneously maintaining large instantaneous bandwidth. Maintaining a small number of mixer conversions, all else being equal, preserves a maximum input two-tone third-order input intercept point (i.e., minimizes two-tone intermodulation).
This architecture has particular advantages for the lower half of the frequency coverage, where the greatest signal [TABULAR DATA FOR TABLE 1 OMITTED] density is expected. The dominant spurious in that range is predicted to be
[F.sub.LO] - N [center dot] [F.sub.RF] = [F.sub.IF], with N = 2, 3, 4
where [F.sub.LO] = LO frequency; [F.sub.RF] = RF frequency; and [F.sub.IF] = IF frequency.
These are easily attenuated by choosing the preselector filters to be suboctave (for example 2.0-3.0 GHz, 2.5-4.0 GHz, etc.). The combination of an upconverting first mixer preceded by suboctave preselection can produce extremely high single-signal spurious-free dynamic range. In this case, for frequencies below 10 GHz, input single-signal second-order intercept points exceed +60 dBm! Of course, at all frequencies, the rejection of the images is very high (the first conversion images range from 27-58 GHz).
Internal construction is illustrated in Figure 1. The entire tuner was first partitioned into eleven hermetically sealed modules (two input preselector modules, upconverter, first IF bandpass filter, output downconverter and six synthesizer modules). They are interconnected by subminiature coaxial connectors that preserve both low VSWR and good shielding. The particular coaxial interconnect technology was introduced by Gilbert Engineering Co. (the GPO connector series) and has been widely applied in defense applications. These connectors are forgiving of small misalignments during module installation, provide excellent shielding and operate to quite high frequencies (standard versions operate to 26.5 GHz).
In addition to the use of the GPO connectors, new circuit transitions were developed within the MIC (microwave integrated circuit) modules that allowed low VSWR interface through the module floor. This allowed stacking of modules as illustrated in Figure 2. The construction approach illustrated in Figure 2 has been applied for frequencies to 20 GHz, and it has been successfully tested for environments of extreme shock and vibration. This module integration approach avoids the numerous coaxial cables normally found in conventional microwave tuners and supports easy access for repair and assembly. It has earned the nickname of "Microwave Lego," for the obvious manner in which microwave assemblies can be constructed.
SHORT PULSE TUNER DESIGN
There has been an increasing need for the capability to survey the electromagnetic environment in search for very short pulse microwave emitters. These emitters are currently surfacing for a wide range of industrial applications which take advantage of both the short duration and wide instantaneous bandwidth of the pulses. The pulses from these emitters produce both time domain and frequency domain challenges for system designers. Typical receivers have limited bandwidth and do not consider time domain implications during design. These two deficiencies will corrupt the signal and prohibit accurate reception of the transmitted pulse. Nonstandard approaches to tuner design are required.
Currently, only short pulses generated at baseband frequencies, typically to 5 GHz, can be recorded. This tuner design was required to provide the capability of translating short pulses from frequencies above 5 GHz to an intermediate frequency where they can be recorded. The critical specification for translating these pulses is the preservation of the time domain characteristics of the signal through the tuner. It is intended that this capability will allow a whole new class of signals to be recorded and will open the door for many future developments.
A tuner architecture similar to that of the initial high dynamic range tuner was adopted to this requirement, with several subtle changes. First, the input preselection filters were eliminated, and in their place were substituted a 1- to 19-GHz preamplifier and lowpass filter. This special amplifier was developed to maintain a group delay flatness of less than +25 psec over any 2-GHz portion of the 1- to 19-GHz input range. Second, the first IF was changed to 21.5-23.5 GHz (2-GHz width), and third, the output IF was changed to 0.5-2.5 GHz. Essentially the identical first LO synthesizer as was used in the previous tuner was employed here, and it also served to downconvert signals in the 18- to 40-GHz range. Overall, this is a significant expansion of the original architecture.
The priority for this tuner (Tuner 2) is to minimize distortion of incoming pulses, and to translate the microwave spectrum about the tuned input frequency to an output IF suitable for very wideband direct digital signal processing. In this case, the IF was chosen as 500 MHz to 2,500 MHz for interface with a digitizer that directly sampled the relatively high frequency IF. The IF amplifier covers 0.01-4.0 GHz in order to support direct pass through of ultrashort baseband pulses to the digitizer.
Given that precautions are taken to minimize amplifier and mixer group delay and gain or loss variation as a function of frequency, the time domain response of the above system is limited by the passband filter characteristics. In this case the 2- to 18-GHz band was required to have a nominal 2-GHz bandwidth, whereas the 18- to 26-GHz and 26- to 40-GHz bands supported 800-MHz instantaneous bandwidth. Bandwidths were "nominal" because the actual specification was minimum pulse risetime (with a 10% overshoot goal) of [less than] 0.5 nsec for Band 2 and [less than] 1.5 nsec for Bands 3 and 4. Modeling of the time domain response demonstrated that standard microwave filters were unsuitable for avoiding serious distortion of pulsed response. In particular, the overshoot and ringing with Chebyshev filters was determined to be severe when the incoming pulse spectrum was not centered within the filter bandwidth. It was necessary to optimize filter design specifically for time domain response.
This tuner converts all incoming spectra centered at an input frequency 22.5 GHz below the first LO frequency to an output spectra centered at 1.5 GHz. The risetimes of interest were so short compared to a cycle at the IF frequency (0.33 nsec for one half cycle at 1.5 GHz) that envelope detection offered insufficient accuracy to measure the expected pulse fidelity of the tuner. Rather, a method was devised to test the tuner by phase locking the test source, the tuner LO and the digitizing oscilloscope to an external 10-MHz reference, such that cycle-by-cycle measurements could be made. A difficulty encountered was imperfect modulation of the input test source (the input test pulse had a rise time on the order of 0.4 nsec), and it is therefore probable that the tuner response time is shorter than that implied by the test data. Note that the input carrier frequency is translated down to 1.5 GHz in the conversion process [ILLUSTRATION FOR FIGURE 3 OMITTED].
Sensitivity was also a priority for this short pulse tuner design. The input preamplifiers had typically a 2- to 3-dB noise figure to 26 GHz; and 5 dB for 26-40 GHz. Input protective limiters also added to noise figure. Thus the noise figure of the overall tuner was determined primarily by the amount of preamplifier gain ahead of the first mixers. The higher the gain, the lower the noise figure, until the dynamic range was compromised. The tuner gain in these cases was limited to about 25 dB (to preserve dynamic range) and the resultant tuner noise figures are plotted in Figure 4.
Two additional major constraints on this tuner included minimization of DC power consumption and weight. The actual power consumption achieved was a remarkable 12 W, and weight was only 7.7 lb. The tuner layout is shown in Figure 5. The flat modules in the foreground are synthesizer modules (adapted from Tuner 1). The vertical card mounted modules in the rear of the chassis are individual microwave converters (input upconverter, filter and downconverter, for example) that interface on their lower edge (not visible) with LO distribution modules through blind mate connectors. Thus the microwave converter cards plug into the chassis much like standard PC cards.
Finally, an additional requirement of this development was to comply with VXI mechanical and electrical interface standards. It is expected that the VXI configuration will add important application versatility.
GENERAL PURPOSE SIGINT (TUNER 3) DESIGN
A problem with the two tuners presented (Tuners 1 and 2) is that they output a rather custom IF, and their frequency step size of 500 MHz is relatively coarse. These turners have obvious advantages, including rapid tuning speed and wide instantaneous bandwidths, and they interface with very high performance downstream equipment.
More normal SIGINT applications require a moderate instantaneous bandwidth (500 MHz is common) primarily to enhance probability of intercept in search scenarios (rather than for short pulse response), much finer tuning resolution (usually [less than] 50 MHz) and a conventional output IF center frequency 1,000 MHz is perhaps the most common. A tuner with a 1,000-MHz IF and tuning resolution of [+ or -]2.5 MHz such as that listed as Tuner 3 of Table 1 can interface with a variety of existing receiver equipment (IFMs and converter/demodulators for example), and can offer very good overall performance at a more affordable price (assuming sufficient versatility to accommodate multiple applications, thereby achieving manufacturing efficiencies). One such tuner architecture is shown in Figure 6.
This tuner incorporates much of the up and down converter circuitry of Tuners 1 and 2, but differs significantly in the instantaneous bandwidth (which simplifies filter design) and most notably in the use of YIG oscillator based synthesizers. These were chosen because of the ability to achieve finer frequency resolution (200 MHz on the first LO and 5 MHz on the second LO) at moderate cost. Note also that considerable effort was expended to develop an architecture that could support high dynamic range conversions (with only two oscillators) that covered the full 500-MHz to 26.5-GHz input frequency range. The physical volume of this design (78 [in..sup.3]) requires high-density packaging similar to that used in the high dynamic range tuner presented earlier.
The microwave modules and oscillators constitute the key enabling technology for the tuner shown in the block diagram of Figure 6. Prototype modules have recently been completed for the coverage to 18 GHz, and the measured converter noise figure, input third-order two-tone intercept point and dynamic range figure-of-merit for 0.5-18 GHz are presented in Figure 7.
A key to this very compact (78 [in..sup.3]) tuner design is the availability of miniature YIG oscillators. YIG oscillators are well-known products with some three decades of use in microwave receivers. Users often associate large size and susceptibility to vibration with YIGs. However, there has recently been a resurgence in new miniature YIG technology, and the construction of these new devices has been remarkably improved. The tuner design incorporates a miniature broadband (6-10 GHz) YIG tuned bipolar oscillator with improved magnetics to enhance tuning speed, and a dramatically new permanent magnet YIG tuned bipolar oscillator (10.9 [+ or -]0.1 GHz) for the second local oscillator [ILLUSTRATION FOR FIGURE 8 OMITTED]. An obvious advantage of the use of a permanent magnet structure is reduction of power consumption.
The feasibility of the "general purpose" SIGINT tuner with fine frequency resolution and low phase noise is critically dependent on the phase noise of the oscillators. In today's market, it is expected that the ability to process high data rate QAM communications signals will be imperative to avoid product obsolescence, and therefore a goal of [less than] 0.4 [degrees] integrated phase noise has been set for this tuner.
The measured prototype oscillator phase noise leads to the predicted overall phase noise of less than 0.3 [degrees] for the complete tuner (with architecture as shown in Figure 6). This clearly supports the specification goals stated in Table 1. Although the developmental activity for this general-purpose SIGINT tuner is still in process, it appears highly likely that the new microwave component and subsystem technology presented will support the aggressive performance objectives of Table 1, Tuner 3.
Microwave tuner solutions for three differing sets of specifications have been presented: 1) a high dynamic range 2- to 18-GHz tuner with 1.1-GHz instantaneous bandwidth and very low phase noise in an extremely compact SEM-EX configuration (58 [in..sup.3]); 2) a 0.01- to 40-GHz tuner in a VXI configuration for short (1 nsec) pulse processing, with excellent sensitivity and very low power consumption; and 3) a general-purpose 0.5- to 26.5-GHz compact (78 [in..sup.3]) SIGINT tuner with good sensitivity, 500-MHz instantaneous bandwidth, fine tuning resolution (2.5 MHz) and low phase noise.
A tuner architecture with a common upconversion of the input (2-18 GHz) spectrum to a 22-GHz first IF before a second conversion to the output IF was applied to the three diverse requirements. This has allowed sharing of a base of microwave circuit and module technology.
The mechanical integration of hermetic microwave modules has been accomplished with novel blind mate connector technology that avoids coaxial cables and minimizes space. Basic internal construction of the MIC modules and key filter and oscillator technology was also presented. In combination, this powerful new technology promises to continue the evolution of SIGINT tuner capability towards smaller size, lighter weight and improved electrical performance.
This work has benefited from the contributions of a team of engineers at the Watkins-Johnson Co. on our general "Miniature-Microwave-Receiver" project. Some 18 design engineers in such diverse fields as mechanical design, microwave filter and amplifier design, synthesizer and oscillator design, tuner subsystem design and manufacturing process design have contributed is essential ways. We have also benefited greatly from comprehensive review and technical contributions of highly participatory sponsors.
We particularly wish to note the contributions to short pulse testing methodology by Joseph Orgnen of QiesTech, Inc., of Manassas, VA.
E. James Crescenzi, Jr., Richard G. Ranson, Roger D. Fildes, and Thomas Spivey are with the Microwave Products Division of Watkins-Johnson Company. Jeffrey A. Cummings is with the US Army CECOM, RIDEC, Intelligence and Electronic Warfare Directorate.
|Printer friendly Cite/link Email Feedback|
|Author:||Crescenzi, E. James, Jr.; Ranson, Richard G.; Fides, roger D.; Spivey, Thomas; Cummings, Jeffrey A.|
|Publication:||Journal of Electronic Defense|
|Date:||May 1, 1995|
|Previous Article:||Simulators, simulators - everywhere.|
|Next Article:||Microwave technology - quo vadis?|