Design of a Wideband, High Dynamic Range Downconverter Used in Smart Antenna Applications

Design of a Wideband, High Dynamic Range Downconverter Used in Smart Antenna Applications

This article details the design of a wideband, high dynamic range downconverter from concept through production. The downconverter is a part of the Smart Adaptive Appliqué (SAA) used in cellular base stations to increase channel utilization and reduce the impact of multipath in congested urban cellular systems. Because the system employs a wideband approach that digitizes the entire cellular spectrum in a single converter, maximizing the spurious-free dynamic range (SFDR) is a primary design requirement. Innovative packaging techniques enable the downconverter to realize the full potential of the individual components, resulting in better than 89 dB SFDR. The initial design phase involves determining the requirements, frequency plan and gain-noise-intercept (GNI) strategy. Following the initial design, critical circuits are breadboarded and tested to confirm computer predictions. Then the most important part of the design begins: the layout of the printed wiring board (PWB). The best paper designs can fail easily due to poor isolation between circuit card components, resulting in needless hours of debugging to correct layout problems. High isolation layout techniques, including grounding, shielding and signal routing, are discussed at length. Finally, the circuit card assembly (CCA) is tested and the card’s performance is presented.

David A. Hairfield, Phillip D. Harvey, Yvonne L. Krayer and Troy S. Voytko
Raytheon E-Systems Falls Church
Falls Church, VA

The SAA system replaces a cellular provider’s existing passive antenna array with a system that maximizes each channel’s signal-to-noise ratio, reducing signal fading and dropped calls. The equipment rack houses digital signal processors, RF conditioning modules, a transmit switch matrix and linear power amplifiers. Instead of using two antenna elements and choosing the one with the strongest signal (space/selection diversity), the SAA digitizes the entire cellular band from each of four receive array elements and applies a continually varying set of complex weights such that the contributions from each antenna element add constructively. Each cellular channel is weighted independently, effectively creating a high directivity antenna that focuses its beam optimally on each mobile source. The net effect is a reduction in interference, signal fading and dropped calls. Figure 1 shows the SAA system. Figure 2 shows the system’s block diagram. Since the processed output port has a higher signal-to-noise ratio than the bypass (nonprocessed) port, it is always selected by a base station utilizing space/selection diversity.

The RF-to-digital conversion block consists of three parts: the low noise amplifier (LNA), RF-to-baseband downconverter and A/D converter. The downconverter portion provides a baseband image of the entire cellular spectrum suitable for digitizing. Since narrowband gain control is not feasible in a wideband converter, the converter must be capable of accurately reproducing both weak and strong signals simultaneously. As a result, high SFDR is a primary design requirement. Figure 3 shows a production downconverter in its versa module Eurocard extension for instrumentation (VXI) housing. The RF signal flow in the downconverter is counterclockwise from top right to bottom right. Shielding fences are shown with the covers removed.

Initial Design

Frequency Plan

The Federal Communications Commission (FCC) divided the Advanced Mobile Phone System (AMPS) cellular spectrum into two groups of bands to allow competition in any given market. Initially, the allocations consisted of only the A and B bands. As usage increased, additional bandwidth was added for each provider. Because of the way the new bands were allocated, the contiguous nature of the original spectrum was lost, as shown in Figure 4 . Existing cellular providers (who used narrowband receivers) could simply tune their receivers to the new frequencies. However, in this system, which digitizes the entire spectrum, the 22.5 MHz of total A-client bandwidth posed a serious problem: The A/D converters available at the start of the project were not capable of digitizing that much instantaneous bandwidth at the desired dynamic range. Therefore, a dual IF architecture was developed for the A-client provider that reduced the final bandwidth to be digitized. The B-client bandwidth of 14 MHz was handled easily with a single IF conversion. To reduce production costs, the downconverter was required to use a common PWB design that, depending on how it was populated, could service either provider.

Figure 5 shows the dual IF architecture that solved the A-client problem without requiring the generation of a third independent LO signal. The output of the first mixer is split into two separate IF bands. The two IFs then are filtered separately and mixed against the second LO or a divided second LO and recombined at baseband, effectively compressing the 22.5 MHz of RF bandwidth into only 15.75 MHz of baseband bandwidth, which is digitized easily. Digital signal processing after the downconverter corrects the inversion of the A' band caused by the use of low side LO injection.

Figure 6 shows how B-client downconversion is performed. Since the entire B-client spectrum can bedigitized using a single IF, a large portion of the PWB is not populated (dotted-line sections).

GNI

The A/D converter’s full-scale input power is –3 dBm, which is the downconverter output level used for single-tone spurious measurements. Two-tone measurements are made 6 dB below full scale to avoid overdriving the A/D. In both cases, the design suppresses the undesired harmonic and intermodulation products well below the noise floor to minimize their impact on system performance. The bandwidth used for noise floor measurement is 30 kHz (the cellular channel bandwidth).

The total gain from the antenna to the downconverter output is 35 dB. With a system noise figure of 5 dB, this noise figure places the output noise floor at –89 dBm, or –86 dBFS (86 dB below full scale). This level supports the design requirement of 80 dB SFDR easily.

A/D Dithering

The dynamic range of a multistage A/D converter can be limited by its own internally generated spurious products, which are caused primarily by the converter’s subranging architecture that reuses digitizing circuitry dozens of times over its full-scale input voltage range. With a pure sine wave input, the crossover distortion caused by subranging occurs at regular intervals, creating spurious responses. However, by adding uncorrelated noise to the input signal (dithering), this regularity is disturbed, spreading the spurious power over a larger bandwidth and reducing the level of A/D-generated spurs. The nondithered dynamic range of the A/D converter is 80 dB, which is the desired minimum system SFDR. Dithering the A/D converter increases its dynamic range to better than 90 dB, eliminating this source of nonlinearity from the overall SFDR.

To generate the dither noise, the broadband white noise from a noise diode is amplified, then band limited and injected into the baseband RF path through a diplexer. Band limiting prevents the dither noise from adding to the designed noise floor in the desired baseband spectrum and increasing the system noise figure.

Prototype Breadboarding

Initially, each amplifier, filter and mixer was mounted on its own connectorized breadboard, as shown in Figure 7 . As the design matured and the cascade was changed to accommodate various features and requirements, these individual pieces were moved easily from one location to another in the chain, providing nearly instantaneous confirmation of the computer predictions. By the time the layout of the PWB began, the design had already been proven. All that remained was packaging the design such that the isolation between components on the final CCA approached the inherent isolation between the connectorized breadboard components. (In addition to giving the design team added confidence prior to the PWB layout, the connectorized components have proven extremely valuable as lab resources in subsequent designs.)

The PWB Design

While careful GNI analysis shows the potential SFDR that can be attained, it takes into account only predicted intermodulation performance of active components and spurious responses from desired paths. Poor matching can seriously degrade a device’s spurious performance while uncontrolled ground currents, sneak paths and waveguide moding can cause high level internal and external spurious responses. It is the job of the engineer to design a PWB that can realize the potential of GNI analysis. In general, that means designing a PWB that maintains its RF integrity well beyond the highest frequency of interest. In this case, even though all the frequencies were below 1 GHz, a microwave approach to isolation, grounding and shielding was used.

Multilayer PWBs maximize the available ground coverage and allow the use of stripline interconnections to maximize isolation. Dual lamination leaves the solder side of the board almost exclusively ground, obviating the need for shields on both sides of the PWB. The only nonground pads on the solder side, aside from through-hole connector pins, are test points (none of which carry RF).

The Toledo rule was used in designing the signal routing: "When you want to keep two parts of a circuit from talking to each other, put one here and put the other in Toledo..." Obviously, separation is a key component of isolation. When combined with shielding and the use of stripline interconnections, high isolations are possible.

The China rule, a corollary to the Toledo rule, was used for partitioning: "When you can’t place a part in Toledo, build a Great Wall around it..." Each functional circuit is partitioned into rectangular areas that can be covered with removable shields. No RF components are placed on the outside of a potential shield area, which allows the shielding of all RF components if necessary. Prototype testing confirmed that shields were needed in obvious places such as driver amplifiers and high rejection IF filters. Later, during FCC emissions testing, an additional shield was needed to suppress an LO signal unexpectedly radiating from the input LNA.

Nearly 2000 ground vias dot the PWB. The average spacing between vias is 200 mils, which is one-tenth of a wavelength at 3 GHz (well above the highest frequency of interest). For electrical performance, there can never be too many ground vias.

For volume production testing and debugging, every non-RF node connects to a test point on the solder side of the PWB. Both bare PWB and in-circuit testing are possible. (Production shops prefer that all nodes have test points, but it was decided that allowing RF test points on the unshielded solder side of the card posed too great a risk for SFDR and electromagnetic interference.)

Performance

Figure 8 shows gain vs. frequency for a typical A-client downconverter. The A" and A portions of the cellular spectrum (824 to 835 MHz) appear at baseband from 1.625 to 11.625 MHz. The A' portion (845 to 846.5 MHz) appears (inverted) from 15.875 to 17.375 MHz. The plot also shows the dither noise from DC to approximately 1 MHz.

Tabulated RF performance of a production A-client downconverter CCA is listed in Table 1 . Except for the noise floor data, the listed data are for the cascade of the LNA and downconverter. The data show that all performance goals were met.

Conclusion

The downconverter achieved all performance goals. The success of the design was due in large part to the conscientious and conservative layout and packaging of the PWB, which allowed the design to fully realize its potential. This step in circuit card design, which is often overlooked, can cause even the best paper designs to fail and result in countless hours of debugging and rework. High isolation layout techniques, including grounding, shielding and signal routing, are the keys to maintaining the integrity of paper designs.

The downconverter is now in production and plays a critical role in the performance of the SAA system. To date, the SAA system has completed two separate field trials successfully and has proven the value of smart antennas in congested urban networks. Data taken from the field trials show a marked reduction in dropped calls due to signal fading and multipath interference and a corresponding increase in network reliability and call duration.

Acknowledgment

The authors wish to thank Thomas M. Butler, David E. Etienne, Allan H. Kaplan and Martin P. Koefoed for their assistance throughout the project. The Toledo rule is attributed to J. Roger Coleman; the China rule is attributed to Phillip D. Harvey. n

David A. Hairfield received his BSEE and MSEE from Virginia Polytechnic Institute and State University, where he specialized in communications and electromagnetics, in 1982 and 1984, respectively. Prior to completing his MSEE, he worked for McDonnell Douglas in St. Louis, MO. From 1984 to 1988, Hairfield designed microwave integrated circuits for Harris Corp. in Florida. He has been a design engineer for Raytheon E-Systems since 1988, where he has designed numerous CCAs for receiver applications.

Phillip D. Harvey received his BSEE from the University of California, San Diego in 1984, specializing in communications, and is currently pursuing his MSEE at George Washington University. He has worked for Eldyne Inc. designing shipboard RF systems and providing technical assistance to the US Navy for electromagnetic interference and compatibility between various ship systems. Currently, Harvey is the project engineer for Raytheon E-Systems’ SAA and an engineering manager for both commercial and Department of Defense RF development efforts. Since 1990, he has focused on RF system design for wideband, high dynamic range signal intercept and direction-finding systems.

Yvonne L. Krayer received her BSEE in 1988 from Auburn University, where she specialized in electromagnetics. She has been a design engineer with Raytheon E-Systems since 1989 where she has designed RF switch and converter assemblies for numerous programs.

Troy S. Voytko has been a design engineer for Raytheon E-Systems since 1989 and has been responsible for the layout of dozens of RF PWBs. He has also worked for Vega Laboratories and Delta Electronics and was a radar technician in the US Marine Corps.