A high-speed LSI GaAs 8/spl times/8 bit parallel multiplier

Multiplication is frequently the speed-limiting function in digital signal processing systems. High-speed hardware multiplier ICs can therefore greatly enhance the throughput and bandwidth of many digital systems. In this paper, the design, fabrication, and performance of GaAs parallel multipliers are discussed. The largest of these circuits, an 8/spl times/8 bit multiplier, has 1008 gates, and is by far the most complex GaAs IC demonstrated today. This multiplier forms the 16 bit product of two 8 bit input numbers in 5.25 ns. This corresponds to an equivalent gate propagation delay of 150 ps/gate. The power dissipation ranges between 0.6-2 mW/gate.     

A 70-MHz 8-bit/spl times/8-bit parallel pipelined multiplier in 2.5-/spl mu/m CMOS

A design is presented for an 8-bit/spl times/8-bit parallel pipelined multiplier for high speed digital signal-processing applications. The multiplier is pipelined at the bit level. The first version of this multiplier has been fabricated in 2.5-/spl mu/m CMOS technology. It has been tested at multiplication rates up to 70 MHz with a power dissipation of less than 250 mW. Clock skew, a major problem encountered in high-speed pipelined architectures, is overcome by the use of a balanced clock distribution network all on metal, and by proper use of clock buffers. These issues and the timing simulation of the pipeline design are discussed in detail. Possible extensions and improvements for achieving higher performance levels are discussed. The conversion of the two-phase clocking scheme to an inherently single-phase clock approach is one possible improvement. A design using this approach has been simulated at 75 MHz and is currently being fabricated.     

2.7 ns 8/spl times/8-bit parallel array multiplier user sidewall base contact structure

A high-speed 8/spl times/8-b parallel array multiplier is developed using sidewall base contact structure (SICOS) technology. The two's-complement multiplication algorithm with carry save adder arrays and carry lookahead adders is utilized. A SICOS transistor results in 14-GHz cutoff frequency and 84-ps/gate ECL switching speed. Multiplication is 2.7 ns with a power dissipation of 900 mW.     

A 4-bit/spl times/4-bit multiplier and 3-bit counter in Josephson threshold logic

A fast Josephson circuit using a threshold logic is developed for application to a multiplier and a binary counter. The former is a typical combinational circuit and the latter is a typical sequential circuit. The junction and barrier materials used were Nb-AlO/SUB X/-Nb. An optimized asymmetric two-junction interferometer maximized the operating margin of the threshold gate. A speed-up junction was introduced to decrease the switching delay without sacrificing the operating margin. A dumping resistor, which was inserted parallel to the input signal line of the threshold gate between its two terminals, decreased the reflection of the input signal caused by the gate inductance, thereby ensuring the margin and speed. To demonstrate the high-speed possibility of the Josephson threshold logic, a high-speed experiment for the circuits was performed. The multiplier demonstrated 210-ps operation.     

10 ns 8/spl times/8 multiplier LSI using super self-aligned process technology

Describes a high-speed 8/spl times/8 bit multiplier LSI which uses the newly developed high-speed and low-power bipolar process technology SST-2. SST-2 results in 250 ps delay time and 0.165 pJ power delay product in a low-level current mode logic (LCML) gate. Its multiplication time is about 10 ns, and its power dissipation is about 660 mW. This LSI has a feature called `perfect expandability' for arbitrary scaling of the expanded 8n/spl times/8n bit multiplier without an additional circuit. The results indicate that 32/spl times/32 bit multiplication can be carried out with 55 ns.     

A 16 bit/spl times/16 bit pipelined multiplier macrocell

A 16 bit/spl times/16 bit pipelined multiplier implemented in a two-layer metal 1.5 /spl mu/m CMOS/BULK technology has been developed. The design is based on the well-known modified Booth algorithm and is capable of operating at a 25 MHz clock rate. The multiplier is designed to be used as a macrofunction within larger chip designs. A structured design approach has been utilized so that reconfiguration of the basic array can be performed. The multiplier macrocell measures 1.7 mm/spl times/1.7 mm.     

New radix-(2/spl times/2/spl times/2)/(4/spl times/4/spl times/4) and radix-(2/spl times/2/spl times/2)/(8/spl times/8/spl times/8) DIF FFT algorithms for 3-D DFT

In this paper, new three-dimensional (3-D) radix-(2/spl times/2/spl times/2)/(4/spl times/4/spl times/4) and radix-(2/spl times/2/spl times/2)/(8/spl times/8/spl times/8) decimation-in-frequency (DIF) fast Fourier transform (FFT) algorithms are developed and their implementation schemes discussed. The algorithms are developed by introducing the radix-2/4 and radix-2/8 approaches in the computation of the 3-D DFT using the Kronecker product and appropriate index mappings. The butterflies of the proposed algorithms are characterized by simple closed-form expressions facilitating easy software or hardware implementations of the algorithms. Comparisons between the proposed algorithms and the existing 3-D radix-(2/spl times/2/spl times/2) FFT algorithm are carried out showing that significant savings in terms of the number of arithmetic operations, data transfers, and twiddle factor evaluations or accesses to the lookup table can be achieved using the radix-(2/spl times/2/spl times/2)/(4/spl times/4/spl times/4) DIF FFT algorithm over the radix-(2/spl times/2/spl times/2) FFT algorithm. It is also established that further savings can be achieved by using the radix-(2/spl times/2/spl times/2)/(8/spl times/8/spl times/8) DIF FFT algorithm.     

An ultralow power 8K/spl times/8-bit full CMOS RAM with a six-transistor cell

A fully static 8K word by 8 bit CMOS RAM, with a six-transistor CMOS cell structure to achieve an extremely low standby power of less than 50 nW has been developed. A 2 /spl mu/m, double polysilicon CMOS process was utilized to realize a 19/spl times/22 /spl mu/m cell size. Redundance technology with polysilicon laser fuses was also developed for improving fabrication yield with relatively large chip size, i.e. 5.92/spl times/7.49 mm. In addition, for reducing operational power dissipation while maintaining fully static operation from outside on the chip, an internally clocked low-power circuit technology using row address transition detectors was employed, which results in only 15 mW operational power at 1 MHz by cutting off all DC current paths. The RAM offers an 80 ns address access time.     

A 2K/spl times/8-bit static MOS RAM with a new memory cell structure

A 2K/spl times/8-bit static MOS RAM with a new memory cell structure has been developed. The memory cell consists of six devices including four MOSFETs and two memory load resistors. Two load resistors are fabricated in the second-level polysilicon films over the polysilicon gate MOSFET used as the driver. Thus the memory cell area is determined only by the area of four MOSFETs. By applying the new cell structure and photolithography technology of 3 /spl mu/m dimensions, the cell area of 23/spl times/27 /spl mu/m and the chip area of 3.75/spl times/4.19 mm have been realized. The RAM is nonclocked and single 5 V operation. Access time of about 150 ns is obtained at a supply current of 120 mA.     

A 1000 FPS at 128/spl times/128 vision processor with 8-bit digitized I/O

This paper presents a mixed-signal programmable chip for high-speed vision applications. It consists of an array of processing elements, arranged to operate in accordance with the principles of single instruction multiple data (SIMD) computing architectures. This chip, implemented in a 0.35-/spl mu/m fully digital CMOS technology, contains /spl sim/ 3.75 M transistors and exhibits peak performance figures of 330 GOPS (8-bit equivalent giga-operations per second), 3.6 GOPS/mm/sup 2/ and 82.5 GOPS/W. It includes structures for image acquisition and for image processing, meaning that it does not require a separate imager for operation. At the sensory side, integration and log-compression sensing circuits are embedded, thus allowing the chip to handle a large variety of illumination conditions. At the processing plane, analog and digital circuits are employed whose parameters can be programmed and their architecture reconfigured for the realization of software-coded processing algorithms. The chip provides, and accepts, 8-bit digitized data through a 32-bit bidirectional data bus which operates at 120 MB/s. Experimental results show that frame rates of 1000 frames per second (FPS) can be achieved under room illumination conditions; applications using exposures of about 50 /spl mu/s have been recently reached by using special illumination setups. The chip can capture an image, run approximately 150 two-dimensional linear convolutions, and download the result in 8-bit digital format, in less than 1 ms. This feature, together with the possibility of executing sequences of user-definable instructions (stored on a full-custom 32-kb on-chip memory), and storing intermediate results (up to 8 grayscale images) makes the chip a true general-purpose sensory/processing device.     

An 8-bit high-speed CMOS A/D converter

An 8-bit high-speed A/D converter has been developed in a 1.5-/spl mu/m bulk CMOS double-polysilicon process technology. The design, process technology, and performance of the converter are described. In order to achieve high speed and low power, a fine-pattern process technology and a novel capacitor structure have been introduced and the transistor sizes of a chopper-type comparator have been optimized. High speed (30 MS/s) and low power consumption (60 mW) have been obtained. Computerized evaluations such as the histogram test and the fast Fourier transform test have been used to measure dynamic performance. The linearity error in dynamic operation is less than /spl plusmn/1 LSB. Signal-to-peak-noise ratio is 40 dB at a sampling rate of 14.32 MS/s and an input frequency of 1.42 MHz.     

A microprocessor-compatible high-speed 8-bit DAC

The increasing use of microprocessors in systems which receive or generate analog signals has created a need for data converters which interface to those processors. A D/A converter which includes all registers and logic required for 8-bit microprocessor interface, and can be fabricated with a standard bipolar linear process is described. The system interface timing is specified such that the converter appears as a memory location to the microprocessor. It can be programmed to operate in a wide variety of modes and can interface with the fastest MOS and TTL microprocessors. The converter offers high-speed multiplying operation and an output current mode multiplexer. Status latches are provided to store multiplexer and code select commands. Nonsaturating multilevel logic operating nearly in the linear region provides gate delays of less than 5 ns when fabricated on the same chip with precision linear functions.     

A Hi-CMOSII 8K/spl times/8 bit static RAM

A Hi-CMOSII static RAM with 8K word by 8 bit organization has been developed. The RAM is fabricated using double polysilicon technology and p- and n-channel transistors having a typical gate polysilicon length of 2 /spl mu/m. The device was realized using low-power high-speed-oriented circuit design and a new redundancy circuit that utilizes laser diffusion programmable devices. The new RAM has an address access time of 65 ns, operating power dissipation of 200 mW, and standby dissipation of 10 /spl mu/W.     

A 50-/spl mu/A standby 1M/spl times/1/256K/spl times/4 CMOS DRAM with high-speed sense amplifier

A 1M word/spl times/1-bit/256K word/spl times/4-bit CMOS DRAM with a test mode is described. The use of an improved sense amplifier for the half-V/SUB CC/ sensing scheme and a novel half-V/SUB CC/ voltage generator have yielded a 56-ns row access time and a 50-/spl mu/A standby current at typical conditions. High /spl alpha/-particle immunity has been achieved by optimizing the impurity profile under the bit line, based on a triple-layer polysilicon n-well CMOS technology. The RAM, measuring 4.4/spl times/12.32 mm/SUP 2/, is fit to standard 300-mil plastic packages.     

A sub-10-ns 16/spl times/16 multiplier using 0.6-/spl mu/m CMOS technology

A 16/spl times/16-b parallel multiplier fabricated in a 0.6-/spl mu/m CMOS technology is described. The chip uses a modified array scheme incorporated with a Booth's algorithm to reduce the number of adding stages of partial products. The combination of scaled 0.6-/spl mu/m CMOS technology and advanced arithmetic architecture achieves a multiplication time of 7.4 ns while dissipating only 400 mW. This multiplication time is shorter than other MOS high-speed multipliers previously reported and is comparable to those for advanced bipolar and GaAs multipliers.     

A 14-bit delta-sigma ADC with 8/spl times/ OSR and 4-MHz conversion bandwidth in a 0.18-/spl mu/m CMOS process

A 14-bit 8/spl times/ oversampling delta-sigma (/spl Delta//spl Sigma/) analog-to-digital converter (ADC) for wide-band communication applications has been developed. By using a novel architecture, a high maximum out-of-band quantization noise gain (Q/sub max/) is realized, which greatly improves the SNR and tonal behavior. The ADC employs a fifth-order single-stage structure with a 4-bit quantizer. It achieves 82-dB SNDR and 103-dB SFDR at 4-MHz conversion bandwidth with a single 1.8-V power supply.     

A 32-element 8-bit photonic true-time-delay system based on a 288 /spl times/ 288 3-D MEMS optical switch

A large-scale three-dimensional microelectromechanical-system optical switch is used for the first time to realize a true-time-delay (TTD) beamformer for phased-array radar applications, with a capacity of 32 antenna elements and eight bits of delay. The 288 /spl times/ 288 optical switch has a median loss of 1.4 dB and all measured 82 944 paths exhibit less than 2.3 dB loss at 1310 nm. The TTD beamformer exhibits a loss variation of 1.5 dB, which is equalized using a mirror-offset technique.     

A 4K/spl times/8 dynamic RAM with self-refresh

A 4K/spl times/8 MOS dynamic RAM using a single transistor cell with on-chip self-refresh is described. The device uses a multiplexed address/data bus. Control of the reconfigurable data bus allows the RAM to operate on either an 8-bit or a 16-bit data bus. The memory cell is fabricated using a double polysilicon n-channel HMOS technology using polysilicon word lines and metal bit lines. Self-refresh is implemented with an on-chip timer, arbiter, counter and multiplexer. A high-speed arbiter resolves simultaneous memory and refresh requests. Redundant rows are used for increased manufacturing yields. Polysilicon fuses are electrically programmed to select redundant rows.