As the world becomes increasingly networked, digital, and mobile, there is a corresponding need for sensor signal-conditioning and signal-processing products that are smart, inexpensive, and easy to use, and that can reduce the development time of the devices they go into.
The SSP1492, designed for high-volume, battery-powered, consumer, and commercial applications, is a monolithic IC device that works directly with resistive, capacitive, inductive, voltage, and pulsed sensor elements
Because there has been no standard sensor application development environment and processing circuitry, sensor end users have had to become experts in the specific sensor technology they wish to use and start from scratch in developing solutions for each new application. Needless to say, this situation has been a hindrance. Developing ASICs for sensors is a moving target of cost vs. performance over time. Customers have come to expect increased functionality at lower costs. Sensor manufacturers and product developers have wrestled with these demands in the past, but can no longer afford to dedicate large amounts of time and resources to this non-core activity. Instead, they are turning to companies that make it their sole focus. To better understand what one such company has to offer, let's review the classical signal conditioning approach.
Classical Signal Conditioning
The most widely used sensor signal-conditioning technique is to combine an instrumentation amplifier (in amp) and an A/D converter (ADC), as shown in Figure 1. In contrast to a general-purpose operational amplifier (op amp), the in amp is characterized by low drift, common-mode rejection, and high input impedance. In amps have a balanced differential input, meaning that the output voltage is proportional to the difference between the input voltages. The differential inputs are compatible with sensors that are balanced, such as bridge-based sensors, or unbalanced, where the sensor is single-ended with a ground return.
Figure 1. Instrumentation amplifier with external gain resistor applied to a bridge-based sensor
Now: Universal Signal Conditioning
Sensors work as they do because they act as a variable resistor, capacitor, inductor, voltage, current source, or pulse source when their measurand is converted into an electrical signal. If a sensor's output signal is converted into a frequency-based signal, it can be easily isolated and converted into a digital result by counting its time period over a predetermined interval. To effect this conversion, the sensor element is placed into a resistance-capacitance (RC) or inductance-resistance (LR) oscillator (Figure 2). The oscillator circuit consists of a comparator with hysteresis in the positive feedback loop that converts the time-varying sensor signal into a square wave with a varying time period. This signal can be counted and digitized into a value that is proportional to the oscillator frequency or period, and, in turn, proportional to the sensor's measurand. In the RC oscillator configuration, the sensor can be introduced into the circuit as either a variable resistor (R) or as a variable capacitor (C). If the sensor has a voltage-based or pulse-based output, it too can be introduced into an RC oscillator (with fixed R and C) as an injected voltage to modulate the oscillator frequency. In the LR oscillator configuration, the sensor can be a variable inductor (L) or variable resistor (R).
Figure 2. Universal signal conditioning, based on RC and LR oscillators, converts sensor outputs into frequency-based signals
This principle is the basis for the LR and RC sensor oscillator front end of the SSP1492 sensor signal processor chip. Consider the advantages of a conversion scheme based on frequency-based signal conditioning:
- 1. High noise immunity. Digital signals from a smart sensor are more immune to transmission noise than are analog sensor signals.
- 2. Wide dynamic range. Because the signal is in frequency form, the dynamic range is not limited by the supply voltage and noise. A dynamic range >100 dB may be easily achieved.
- 3. Simple integration and coding. An adding pulse counter enables precise integration in time for a frequency-modulated signal, approaching an ideal integrator with unlimited measurement time. The counter result can be processed by a microcontroller with no additional interface circuitry.
As shown in Figure 3, the sensor oscillator front-end circuit consists of a comparator and a HI/LO DAC that functions like a Schmitt trigger. Five parallel inputs connected to the comparator's negative pin are selectable by analog switches. Resistive, capacitive, voltage, or pulse-based sensors can be interfaced to the sensor oscillator through these connections. An inductive or resistive sensor can be interfaced into one of four switch-selectable, negative feedback loops located in parallel with the comparator. Eight general-purpose I/O (GPIO) pins, also switch-selectable, are provided at the negative pin of the comparator for additional sensor inputs.
Figure 3. The SSP1492 sensor signal processor chip can help sidestep some of the cost and time associated with sensor product development and improve performance as well
This architecture permits the SSP1492 to handle up to 15 separate sensor input channels for highly flexible multisensor configurations, especially where collaborative processing of multiple and mixed-sensor inputs is needed. It is the only monolithic IC solution that works with resistive, capacitive, inductive, voltage, and pulsed sensor elements—whether MEMS or bulk-based devices—on a single chip at the same time.
An onboard general-purpose op amp is available for voltage-based sensor outputs that require amplification before their signal is injected into the sensor oscillator. A period counter unit (Figure 4) demodulates and digitizes the output signal from the sensor oscillator section, and can measure two parameters of time-varying signals: cycle period and pulse width. It can be programmed to measure a single pulse or multiple signal cycles, providing user-selectable conversion resolution. This flexible conversion approach easily accommodates a broad range of sensors and applications. The fact that the period counter is a 100% digital device makes the mixed-signal design much more straightforward.
Figure 4. The period counter can provide scalable resolution and conversion times with virtually infinite resolution
The high-speed, pipelined 8051 microcontroller core is user accessible and pipelined to run at up to 18 MIPS. Sensor-specific code can be uploaded into internal RAM from a host computer, or from an external serial EEPROM, and stored in user-customizable firmware memory space. Chip operation may be controlled by the 8051, the host, or by both where the host calls upon specific 8051 ROM routines as required.
Data EEPROM is available onboard the chip to provide 64 bytes of nonvolatile storage for factory calibration coefficients and user settings. It can also be used to create an IEEE 1451 transducer electronic data sheet (TEDS) to store critical sensor parameters that enable plug-and-play operation. A typical TEDS for a single-axis accelerometer is shown in Figure 5.
Figure 5. Typical Accelerometer TEDS
Two powerful hardware-based, fixed-point math engines and a software floating-point engine enable the chip to perform high-order functions for linearization, calibration, and temperature compensation of the sensor outputs. This feature includes a built-in high-order polynomial fitting function (up to 11th order) for calibration and linearization:
y(x) = Ax11 + Bx10 . . . + Jx + K
where the A, B . . . J, and K coefficients may vary from sensor to sensor, and can be easily uploaded for each sensor into nonvolatile memory from a host computer during production testing.
A CORDIC (COordinate Rotation DIgital Computer) math engine performs the vector and scalar calculations for sensor data processing (Figure 6). These functions include trigonometric, inverse trigonometric, geometric, absolute magnitude, long integer, and scaled fractional multiply, divide, add, and subtract.
Figure 6. Supported Mathematical Functions
During normal operation, the user loads onchip register pairs with values and constants from memory. A MODE register is set to the appropriate mode for the type of calculation to be performed, and a control and status bit is set to a high state to initiate the calculation process. The CORDIC engine resets this bit when the calculation is complete and is very efficient at reducing code space and time required (<1 ms) for the microprocessor to make these calculations. It also reduces the amount of SRAM needed to store values during a calculation.
The high-speed clock oscillator is based on an RC oscillator and is used as the time base for all the digital circuitry including the microprocessor and period counter reference. Through firmware, the user can command the oscillator into power-down and wake-up modes. There is also an external clock option for higher accuracy applications.
The band-gap voltage regulator is stable across different power sources and supplies voltage to the sensor oscillator and high-speed clock oscillator sections on the chip. Constant voltage maintains oscillator accuracy to ensure sensor performance. This is especially important for battery-powered applications where battery voltage may drop off over time.
The SSP1492 is designed for 3 V operation but can operate down to 2.2 V. It can communicate with a host system using either SPI or I2C serial digital protocols. The SPI interface can be either master or slave, but is typically in slave mode; the I2C is slave mode only. With the SSP1492 in slave mode, the host computer has direct access to all of the special function registers, allowing full access to all peripherals and ROM routines.
Devices are available as 4.3 X 4.3 mm bare die or in a 64-pin quad-flat package, but planned revisions will shrink die size to ~2.0 X 2.0 mm. Universal sensor evaluation kits include a USB interface, user programmable EEPROM, and development/system software analysis tools. The user needs only to supply the sensor element and a few passive components to create a smart sensor solution within a few hours for quick and easy evaluation.
The piezoresistive pressure sensor in Figure 7 is interfaced to the SSP1492 using a 4RC dual differential configuration. Each of the four piezoresistors of the sensor bridge is individually selected and connected with an external capacitor to the oscillator section using the onchip configurable analog switches and drivers. A macro program controls the sequencing of the analog switches and measurement of each bridge resistor with a fixed capacitor using the sensor oscillator in RC mode. Though each resistor is measured separately, all are dependent on the other series/parallel resistors in the bridge network. If the four differential measurements are algebraically combined, the cross-terms will cancel out and produce a result proportional to the sensor signal output, result = (R3 – R4) – (R2 – R1), which can be converted into the appropriate engineering units for pressure.
Figure 7. Bridge-based piezoresistive pressure sensor using the SSP1492 for signal conditioning
A capacitive, 3-axis accelerometer (Figure 8) incorporates a micromachined 3D seismic mass. Acceleration on three axes is detected by eight capacitor plates integral to the seismic mass. Acceleration measurement is based on variations between the gap and the overlapping plate area between planar surface pairs for each capacitor.
Figure 8. Triaxial capacitive accelerometer using the SSP1492 for signal conditioning
The capacitors are connected in parallel to the CAP1 pin of the sensor oscillator section. The output drivers of the general-purpose I/O section on the SSP1492 selectively ground each capacitor so that each can be switched into the sensor oscillator section that is operating in RC mode. Four capacitors are taken as two pairs of differential-based measurements for acceleration along each axis of motion.
The mathematical difference of two differential pairs provides acceleration measurement along the ±X and ±Y axes; the sum of the four pairs provides the ±Z axis measurement.
In a Nutshell
The SSP1492 is a low-cost, highly flexible universal sensor drive and signal processing platform that uses innovative technologies to allow direct and fast interfacing to nearly every type of sensor. It is the only such monolithic IC device that works with resistive, capacitive, inductive, voltage, and pulsed sensor elements—both MEMS and bulk-based—all on the same chip. It is well-suited for high-volume, battery-powered, consumer and commercial applications, where cost, size, power consumption, and time-to-market are the critical factors for product success.
The authors wish to thank John Garay from Sensor Platforms, Inc., for his support, review, and constructive input.
This article was adapted from a paper presented at Sensors Expo 2005, in the session, "Nanotechnology and MEMS/MST/Micromachines (M3): A Global Perspective on Technology, Applications and Commercialization."
George Hsu, BSEE, and Joseph Miller can be reached at Sensor Platforms, Inc., Santa Rosa, CA; 707-543-8540, [email protected], [email protected], www.sensorplatforms.com. Edward Zdankiewicz, BSME, MSME, PE, can be reached at Primedyne Systems Inc., Cleveland, OH; 216-374-3377, [email protected].