The last few years have seen dramatic improvements in the analog-to-digital converter (ADC) world, changing the landscape of this data-conversion chip beyond recognition. With the migration toward CMOS processes, the number of functions being integrated around the converter are transforming the part into a subsystem solution while maintaining the performance at lower power and smaller die size. As makers adopt finer CMOS geometries, these highly integrated ADCs are beginning to look like complete analog front ends for applications like data acquisition, wireless communications, and audio. They also might find a home in the industrial and instrumentation markets.
While this trend remains popular, efforts continue to increase the level of integration at low supply voltages. Designers are exploring circuits and techniques to bring memory, microcontrollers, and processors like DSPs on board at the digital end. In the RF domain, they're investigating technologies to pack functions like mixers, local oscillators, and amplifiers on chip. Doing so will transform the ADC into a complete radio receiver for mobile communications.
In the last year, several major ADC vendors have demonstrated high-resolution ADCs that include all of the key functions needed to directly link sensors, as well as other such sources of signals, to microprocessors or digital systems. These integrated ADCs have drastically simplified the analog front-end design to enable complete data-acquisition solutions with a minimum number of chips, along with unprecedented levels of accuracy, speed, and resolution. They also keep power consumption within design budgets.
A good example is Cirrus Logic Inc.'s ultra-low-noise data-acquisition IC, the CS553x, which was unveiled late last year (see "Ultra-Low-Noise Data-Acquisition IC Tackles Multiple Sensors," Electronic Design, Oct. 28, 1999, p. 59). Designed for precise measurement of low-level unipolar or bipolar signals, this 24-bit ADC has a lot of on-chip features: a multiplexer, an ultra-low-noise programmable-gain instrumentation amplifier (PGIA), a selectable Sinc filter, calibration, control logic, a clock generator, and a three-wire serial interface for microprocessors and microcontrollers.
In essence, it integrates all of the necessary analog and digital functions to directly and singlehandedly link a variety of sensors to general-purpose microcontrollers. It used to be that two or more chips were required to perform a similar task. A high-performance instrumentation amplifier was needed to achieve ultra-low noise, adding to cost, space, and design time.
To extend the applications range for this line, Cirrus Logic is now in the data-gathering mode. It's exploring a variety of power, speed, and noise combinations to address myriad applications. By readying subgrade members with lower power, the company can meet the requirements of applications like weigh scales, in which battery life takes precedence over high accuracy.
Likewise, higher-speed versions are in preparation for medical equipment and other similar instrumentations. There, conversion speed is an important specification. Meanwhile, Cirrus Logic designers are actively pursuing the challenge of integrating a microcontroller on the same CS553x chip. However, further details were unavailable at the time of this writing.
Also in the microcontroller race is Burr-Brown Corp. This supplier has stepped up its effort to bring a general-purpose microcontroller on board. Toward that end, it has licensed an 8-bit, 8051 synthesizable core from Synopsys Inc. It plans to integrate that core with 20-bit converters, providing data at moderate output rates.
"For the initial introduction, the ubiquitous 8051 processor core was selected for its popularity and its established infrastructure of development and programming tools," states Robert Schreiber, strategic marketing engineer at Burr-Brown. By combining existing development tools with evaluation modules, the company is readying "an integrated development environment that will make it painless for analog engineers to work with these mixed-signal parts," says Schreiber.
The early introductions target smart transmitters, in which size is a constraint and remote processing is desired. "On-chip flash memory will allow it to be programmed in the field or remotely," Schreiber points out. "Plus, it will provide better control of system calibration with added intelligence. In short, it will enable more specialized processing on chip to offload the system microcontroller and relieve it for other tasks."
Under development are two delta-sigma (Δ−Σ) ADC chips for release later on this year. One of them will include an 8051 core (Fig. 1). The other will incorporate an 8-bit RISC core for applications that demand faster on-chip processing. Both devices will offer a variety of memory and I/O configurations.
The 20-bit, 25-kHz ADC with an on-chip 8051 core will provide up to 32 kbytes of program flash memory. In addition, it will have up to 4 kbytes of data flash memory and 1 kbyte of RAM. The chip also is home to a multiplexer with eight single and four differential channels, a low-noise PGA, a clock generator, a Sinc3 filter, serial ports, and an I/O port. It will implement debugging capability.
This integrated converter will be implemented in a 0.6-µm CMOS process to keep power consumption and die size within design budgets. Preliminary specifications indicate that the device will consume less than 10 mW at a 5-V supply and a 4-kHz data rate. Due to the size of the memory on the same die, however, the company says that the user will have to pay a slightly higher premium for an ADC with a microsystem controller on board.
Faster Version Due
Somewhere in the same timeframe, Burr-Brown plans to unwrap a faster 20-bit version that's capable of a 35-kHz output rate. It'll incorporate an 8-bit RISC core. To cut power consumption for the RISC-based ADC solution, the faster 20-bit model won't have the PGA on chip. Although the initial offerings won't support the IEEE 1451.2 common transducer interface standard, Burr-Brown's designers are investigating this option for future releases.
Meanwhile, the company is in the process of releasing these 20-bit versions without the microprocessor cores. Based on 0.5-µm CMOS technology, the 20-bit ADS1216 and ADS1218 will include all of the requisite functions and peripherals needed to deliver a monolithic data-acquisition subsystem on a single chip. The only difference is that the 1218 will come with 4 kbytes of flash memory to store calibration coefficients and configuration data.
Even though merging very precise ADCs with microcontrollers has just begun, some have demonstrated the viability of doing so by relaxing the conversion specifications and going to a larger package. As a result, Analog Devices has achieved that convergence in its Microconverter line launched last year. From initial 12-bit resolution, the Microconverter has been upgraded to 24 bits with the addition of the ADuC824.
This programmable Δ−Σ ADC, which was unveiled last October, is a complete smart-transducer front end. It incorporates two independent (24- and 16-bit) ADCs, a temperature sensor, a PGA, and an 8-bit microcontroller. It also boasts 8 kbytes of program flash memory, 640 bytes of data flash memory, 256 bytes of data RAM, a timer/counter, a 12-bit voltage-output DAC, and serial ports (Fig. 2).
The microcontroller core is 8051-instruction-set-compatible. It operates from a 32-kHz crystal with an on-board phase-locked loop (PLL) generating the required internal operating frequency. Consequently, the output data rate is programmable in software, as is the microcontroller operating frequency for this industrial-class converter.
"Depending on the application, these parameters can be changed on-the-fly," explains Grayson King, ADI's applications engineer. "Also, the resolution varies with the output data rates." For example, the effective number of bits for the Microconverter is 19.2 at a 5-Hz data rate. It goes down to 16 bits at 105 Hz. An internal voltage generator provides the appropriate voltage for programming the chip, allowing the part to operate from a single 3- or 5-V supply. In short, it offers all of the hooks required to design a self-calibrating, user-programmable data-acquisition subsystem.
Sensor Interface Support
Interestingly, the company has added another new feature to this part. To ease communication in a network connection, the ADuC824 supports the IEEE 1451.2 sensor interface standard. Essentially, that standard specifies a digital interface to access the sensor's calibration, correction, or configuration data stored in its nonvolatile memory. By combining this part with a sensor, a manufacturer can produce a smart sensor module that provides signal conditioning and decision making in the same package. The transducer electronic data sheet (TEDS) can then be stored in the nonvolatile memory to be updated in a network environment.
Keeping the analog designers in mind, ADI has readied an integrated development kit that includes both an evaluation board and development software that simplifies their job. "The sample codes available with this tool allow an engineer to play with the Microconverter ADC without writing any code," states King.
The development software includes an assembler, a Keil C compiler, a serial downloader, a simulator, a debugger, and an example code library. It comes on a CD-ROM and can be accessed on ADI's web site. Encased in a 52-pin PQFP, the primary targets for this system-level converter are applications like weigh scales, thermocouples, temperature/pressure measurements, and other slow-changing systems.
Concurrently, ADI is preparing a lower-cost version with 16-bit resolution in the main channel. Much of the cost is due to testing time in high-resolution converters. "Going to 16 bits will save us time on test, which will be passed on to the users," notes King. Smaller packages also are on the drawing board to further cut the price. While the current Microconverter members are based on 0.5-µm CMOS, plans are to migrate toward 0.25 µm for the digital portion of the die.
The 8051 core appears to be the preferred choice for several reasons. Among them are widespread proliferation of the architecture, third-party support, and documentation. "The 8051 is the best-documented 8-bit microcontroller out there," says Tremont Miao, ADI's product line director. The architecture is open and plenty of development software is available through third parties.
With its recent introduction of the MAX1460, Maxim Integrated Products has joined the fray. This 16-bit low-power ADC provides on-chip digital correction of the output over the specified temperature range. For correcting sensor offset and errors, it incorporates a rudimentary digital signal processor (DSP) for performing multiply-accumulate (MAC) operations. The MAX1460 also includes 128-bit EEPROM to store the user-programmed compensation and calibration coefficients.
The MAC instructions and code are actually embedded within the chip, eliminating the trouble of programming the DSP for signal processing. An uncommitted op amp is available to filter the analog output or implement a 2-wire, 4- to 20-mA transmitter (Fig. 3). The frontal section of this 16-bit ADC offers a 2-bit PGA and a 3-bit coarse-offset DAC to condition the sensor's output.
The device has built-in testability that integrates three traditional sensor-manufacturing operations into one automated process. These include pretest, calibration and compensation, and final test. Maxim claims that by eliminating manual calibration and allowing verification of transducer calibration and compensation in the pretest socket, the MAX1460 cuts sensor manufacturing costs by a substantial amount. Operating from a single 5-V supply, this part is designed to resolve 1 µV of differential input signal. It consumes only 400 µA of supply current, making it suitable for low-power applications.
Speaking of low power, the available MAX1462 version will run from battery power at 2.35 V. According to Maxim, that device can operate down to 2.2 V.
Other key suppliers simplifying the link to DSPs include Intersil Corp., National Semiconductor Corp., Telcom Inc., and Texas Instruments Inc. These manufacturers are adding all the required bells and whistles to make the high-performance ADC connection seamless to a variety of DSP processors.
Interestingly, manufacturers supplying microcontrollers for data-acquisition systems have expanded their role to compete head on with ADC producers. Producers like Microchip Technology and Atmel Corp. have acquired enough analog expertise to bundle the front-end analog peripherals and converters with their 8-bit controllers. They can now deliver a complete solution to the builders of data-acquisition systems.
Last September, for example, Microchip debuted its 8-bit microcontroller line, PIC16C7xx. It packs a 10- or 12-bit successive-approximation-register (SAR) ADC, enhanced capture/compare/PWM, voltage reference, OTP programmable memory, RAM, oscillator, and timers on the same silicon. The goal is to provide a one-stop shop for embedded-systems designers. Until now, they had to rely on analog houses for the peripherals of the microcontroller-based solution.
Now the company provides both integrated solutions, as well as standalone multichannel 10- and 12-bit ADCs. These include op amps for the designers that prefer to segregate the analog and digital portions of their embedded solution for better accuracy control and noise performance. Like Microchip, Atmel has embedded nonvolatile flash memory, a 6-channel 10-bit ADC, and a host of other peripherals around an 8-bit RISC core. The company hopes to furnish a cost-effective solution to many control applications.
Meanwhile, those flaunting analog strength continue to concentrate on precision, low-noise, micropower, and wide-dynamic-range parts from a tiny low-cost package. Their goal is to make these very-high-accuracy parts extremely simple to use. The calibration is transparent to the user.
Driven by this philosophy, Linear Technology Corp. crafted a 24-bit, third-order Δ−Σ ADC, the LTC2400, nearly 18 months ago (see "High-Resolution ADC Targets Precision Data-Acquisition Applications," Electronic Design, Sept. 1, 1998, p. 68). Since then, the supplier has expanded the line with key enhancements like higher speed, lower noise, and the addition of a multiplexer.
"Our focus has been to enhance the analog part and keep the digital interface simpler," says senior design engineer Mike Mayes. "Unlike others, it does not require any writing into the registers. Nor does it ask the user to go through a complex sequence of steps to calibrate the LTC2400. Its self-calibration is transparent to the user. As it offers no latency, it is easy to multiplex. Latency and redundant data are normally associated with Δ−Σ ADCs."
The family has been expanded with multichannel versions.The latest revision has a fully differential input and differential reference. The micropower 24-bit LTC2410 is designed to maintain the effective number of bits (ENOB) and display extreme stability over the input range of −0.5 to +0.5 V. Its offset error is negligible over the input common-mode voltage of 0 to +5.5 V. This no-latency architecture is crafted to ensure that the linearity doesn't drift over the industrial temperature range of −45° to +90°C while the offset is zero and the drift is minimal.
The sum of this performance is that the total adjusted offset error is within 3 ppm over the differential input, as well as over the temperature range of −40° to +125°C (Fig. 4). The noise should only be 800 nV (rms) over the entire input range. This part has been tailored for simplicity. Sampling now, the LTC2410 is slated for production in a month or two. A multichannel version also is in the works.
On the wideband front, applications demand much higher sampling rates. So mixed-signal IC supplier DataPath Systems has extended its capabilities to design a 5-Msample/s, 16-bit pipelined ADC with an on-chip PGA, a sample-and-hold (S/H) circuit, and a voltage reference. Implemented in 0.5-µm CMOS, the DPS9245 provides user-programmable power dissipation that's dependent on the sampling rate. At 5 Msamples/s, the unit dissipates 400 mW. It drops to 200 mW at 2.5 Msamples/s.
The PGA offers up to 20 dB of gain in seven settings. According to the company, the amplifier's high-resistive input impedance (greater than 1 kΩ) eliminates the need for an external buffer amplifier and complex filtering to handle transients. This integrated converter also boasts a noise density of 8 nV/(check)Hz ±1.0-LSB integral nonlinearity. It has ±0.5-LSB differential nonlinearity. A spurious-free dynamic range (SFDR) of 95 dB can be obtained from this pipeline multistage architecture at a 900-kHz sinusoidal input. The differential input range is 5-V p-p. These specifications are guaranteed over the industrial temperature range.
This part targets ATE, subscriber-line test equipment, scanners, and imaging applications that need the data-acquisition subsystem. According to DataPath's director of marketing, Phil Welsh, "This level of integration at such a high sampling rate is unprecedented."
|Suppliers Mentioned In This Report|
Analog Devices Inc.
Ray Stata Technology Center
Cirrus Logic, Crystal Division
DataPath Systems Inc.
Linear Technolog Corp.
Maxim Integrated Products
Microchip Technology Inc.
National Semiconductor Corp.
Telcom Semiconductor Inc.
Texas Instruments Inc.
Semiconductor Group SC-98085
Literature Response Center
(800) 477-8924, ext. 4500