A MIXED-SIGNAL PROGRAMMABLE ARRAY USING ANTIFUSE INTERCONNECT R. Timothy Edwards, Kim Strohbehn, and Stephen E. Jaskulek Johns Hopkins University Applied Physics Laboratory Laurel, Maryland The ability to use radiation-tolerant programmable digital logic parts in hardware intended for use in space missions provides significant cost savings. The savings result partly from shorter design cycles and lower risk, but mostly from reduced parts acquisition and flight qualification costs. Without these parts, many projects would be unable to afford custom gate array implementations, and would be forced to resort to discrete component designs with the accompanying increase of system size, mass, and power consumption. A similar situation exists in the mixed-signal field today. Almost all spacecraft contain numerous moderate-performance analog and digital processing and I/O circuits. These circuits, used for applications such as status monitoring, motor and temperature control, and signal conditioning and processing, are widely distributed throughout the spacecraft. Traditionally, these mixed-signal circuits have been implemented almost exclusively using discrete components, because the cost savings are not significant enough to justify a custom ASIC design. However, the resources used by such circuits, including mass, power, and volume, add up very quickly. The ability to implement such designs using general-purpose, programmable analog or mixed-signal arrays would be advantageous in all respects, analogous to the advantages gained using a Field-Programmable Gate Array (FPGA) for digital circuits: lower parts acquisition and qualification costs, higher levels of integration, lower power consumption, fast turn-around time for design cycles, and improved reliability. Field-Programmable Analog and Mixed-Signal Arrays (FPAAs and FPMAs) address this need for quick turnaround time in analog flight hardware development. Whereas FPGA architectures typically consist of essential basic combinatorial and sequential logic blocks (flip-flops and mux-based logic, for example), FPAAs vary widely in the choice of the core programmable analog module. The granularity of the module tends to be a function of the performance of the interconnect: Slower (high-resistance, on the order of 1 kOhm) interconnect types (such as SRAM- or EPROM-based) can seriously degrade analog perforamance if inserted at arbitrary points in an analog design. Thus, modules for RAM-based analog arrays tend to have a large granularity such as an entire filter section and allow interconnect programmability only at certain critical points, such as between filter sections or in series with the resistor in the amplifier's feedback loop. Faster (lower-resistance) interconnect types allow a finer module granularity, to the limit of gate array architectures, which allow programmability at the level of individual transistors, but does not solve the stated problem of quick turnaround time and flight qualification. In between, antifuse interconnect (with resistance as low as 15 to 20 Ohms) offers a good tradeoff between functionality and performance on one hand and quick development time and pre-flight qualification on the other. In the past year, we have entered into a cooperative agreement with Actel Corporation, known for manufacturing radiation-tolerant and radiation-hardened FPGA parts based on its antifuse technology. We have designed a field-programmable analog array test chip which has been fabricated in the same technology as Actel's RT-SX line of radiation-tolerant chips. This fabrication process is a 0.22 micron, three-metal process optimized for digital circuitry. The metal-to-metal antifuses, when programmed, typically achieve resistances of 20 Ohms or lower. The low resistance of interconnects allows us to reduce the granularity of our basic analog building blocks to individual components: capacitors, resistors, amplifiers, and analog switches. Our test chip contains twelve modules, each containing a differential amplifier, eight programmable capacitors, four programmable resistors, 64 CMOS switches, and internal and inter-module routing resources. We will report on the measured performance of the test chip, particularly as relates to the difficult problem of constructing high-performance analog circuits in a substantially digital fabrication process. Of particular concern are the linearity of the capacitors, component matching, and performance of a test suite of basic circuit configurations including a charge shaping chain, ADC, and filters. Additionally, we will report on the steps we believe are necessary to make the next generation design a commercially viable product.