A High-Performance CMOS
Processor for Piezoresistive Sensors

Electronic trimming of sensor errors is an emerging trend in sensor technology. In conventional transducer design, calibration and compensation are performed in the analog domain. To store the sensor-specific data, analog "memory" components such as laser trimmed resistors, potentiometers, and discrete resistors or capacitors (some of them temperature dependent) are used. The major problems with such an approach include restricted accuracy of compensation, resulting from the nonlinearity of the sensor; and high capital and labor cost of the trim equipment (e.g., automated laser trimmers).

Photo 1. There are more than 17,000 transistors in the MCA7707, shown here in a microphotograph. The minimum feature size is 1.2 µm.

The emergence of low-cost digital programmable electronics has made it possible to trim the analog functions in the digital domain and to store individualized correction coefficients in nonvolatile digital memory (e.g., EEPROM). Electronic trimming for sensors has evolved into two generic approaches.

• Analog sensor signal processors (ASSPs) digitally adjust the offset and gain of amplifiers, as well as sensor excitation, to achieve sensor calibration and temperature compensation in an analog domain, without signal quantization.
• Digital sensor signal processors (DSSPs) convert the sensor signal into a digital domain using an A/D converter (#ADC); perform calibration and compensation in a digital domain using microcontroller or custom logic; and, if needed, convert the compensated signal into an analog domain using a D/A converter (DAC).

This article introduces the basics of electronic trimming technology and characterizes the design and performance of the MCA7707 CompProcessor (see Photo 1), based on ASSP technology.

Electronic Trimming

Figure 1. A D/A converter enables multiplication of an analog voltage, Vin, by a digital number N, which permits digital trimming of analog signals without converting them into the digital domain.

Electronic trimming for sensor applications allows digital adjustment of the analog component or function (e.g., resistance, capacitance, offset, gain, or current). Digital adjustment permits the set of values to be stored in digital, instead of analog, memory. The key element of the electronic trimming system for most ASSP configurations is the DAC that enables multiplication of an analog voltage by a digital number (see Figure 1).

A DAC requires an input (or reference) voltage, V_in, and generates an output voltage, V_out, that is a product of V_in, calibration constant , and digital number N:

(1)

N can be stored in memory for customization of each sensor's electronic performance, thus forming a foundation for programmable electronic trimming. A DAC in this application is equivalent to a digital potentiometer.

To trim different sensor errors, V_in and V_out terminals can be connected as follows:

Using such configurations, we developed a unique single-chip CompProcessor, the MCA7707 (see Figure 2 below). For low-cost digital trimming, we devised a new DAC and ADC technology that placed 16-bit converters on a very small area of silicon. The small size permitted cost-efficient implementation of the complex systems-on-a-chip with multiple DACs and ADCs.

Click the circuit diagram to see a larger version.

Figure 2. The single-chip MCA7707 calibrates and compensates sensors by changing the offset and gain of the programmable gain amplifier and sensor supply current using five 16-bit D/A converters. Up to 120 compensation segments can be used for compensation of zero and F.S. over the operating temperature range. The selection of coefficients for each of the temperature segments is accomplished by controlling the address of the EEPROM with the output of the 12-bit A/D converter driven by the temperature-dependent bridge voltage. The ratiometric output configuration provides an output that is proportional to the power supply voltage. When used with ratiometric A/D converters, such an output provides a digital pressure value independent of the supply voltage.

ASSP System Architecture
High-performance compensation of sensor errors requires a high level of complexity because sensor errors come in several functional categories, most of which require nonlinear temperature correction. The compensation processor was optimized to perform the electronic calibration and compensation typical for piezoresistive sensors, as well as to support the compensation of other types of resistive sensors.

One important requirement of the CompProcessor architecture was that it support advanced manufacturing technologies. To lower manufacturing cost, three conventional sensor manufacturing operations were integrated into one automated process:

Pretest. The sensor's performance is tested over the compensated temperature and pressure ranges. To carry this out under the control of a host computer, a Microwire interface and three-state output were included. The configuration allows a parallel connection of multiple transducers under test, and digital communication between a specific sensor and the test system. The selection of a given transducer is via a chip-select pin.

Calibration and Compensation. This operation can be performed immediately after pretest, without removing the transducer from the test socket. Calibration and compensation coefficients (4 Kb) are written into the transducer's EEPROM, calculated by the test computer, and downloaded to the transducer via a Microwire interface.

Final Test. Performance is verified again, without removing the transducer from the pretest socket.

The ASSP configuration provides a true analog signal path for the sensor signal, eliminating quantization noise typical of DSSP systems. It includes a high-performance chopper-stabilized programmable gain amplifier (PGA); a programmable current source to excite the sensor; a 12-bit ADC digitizing the bridge voltage (a measure of temperature); five 16-bit DACs; a Microwire interface to an external EEPROM and test computer; and several instrumentation amplifiers and supporting circuitry (oscillator and bias generators).

The temperature compensation processor is driven by the piezoresistive sensor bridge voltage, which is used as a temperature signal (or an external transistor for other sensors, such as metal strain gauges). Two types of compensation are implemented:

Analog. To compensate the first-order temperature errors, a continuous adjustment of offset and bridge current with temperature is performed by connecting two DACs to the temperature signal (bridge excitation voltage).

Digital. The same temperature signal is used to control the address of the EEPROM via an ADC driven by the temperature signal. This compensation corrects residual higher order errors using a multisegment (up to 120) look-up table that resides in the EEPROM).

As illustrated in Figure 2, MCA7707-based compensation and calibration provide:

• Initial offset calibration
• Initial full scale (F.S.) output calibration
• Offset temperature coefficient slope correction
• Offset temperature coefficient nonlinearity correction
• F.S. temperature coefficient slope correction
• F.S. temperature coefficient nonlinearity correction
• Pressure linearity correction

Initial offset correction is implemented by multiplying a fraction of the supply voltage by a 16-bit word to create a voltage that feeds into a summing junction of the programmable gain amplifier, compensating the sensor offset.

F.S. calibration is achieved using two adjustments. Coarse gain is set by adjusting the gain of the PGA with 3 bits; in this case the sensor signal is multiplied by a digital word. Fine F.S. calibration is performed through the adjustment of bridge current using another 16-bit word.

To compensate the linear component of zero and F.S. temperature coefficients, the reference inputs of two DACs (Offset TC DAC and F.S. TC DAC) are connected to the bridge voltage. For a given digital word, the DAC's output voltage will follow the quasilinear bridge voltage change with temperature. By adjusting the multiplier coefficient, compensation of the temperature slope is achieved.

Multislope Temperature Compensation
Digital multislope temperature compensation allows compensation of the arbitrary error curves, which are restricted only by the available adjustment range of the electronics and the shape of the temperature signal. Compensation is implemented using look-up tables for 120 pairs of offset temperature coefficient and F.S. temperature coefficient correction numbers. The numbers are stored in EEPROM. The address of the EEPROM is selected by the output word of the 12-bit ADC driven by the temperature-dependent bridge voltage.

When the bridge voltage changes, different sets of correction coefficients are read from the memory for each incremental temperature span. The temperature span can be estimated as ~1/120 of the compensated temperature range. If the range is ­20ºC to 80ºC, then a new set of correction coefficients will be generated approximately every 1ºC

Calculation of the correcting coefficients' values is performed during the test procedure using curve-fitting into the test data. A larger number of test points makes for a better curve fit accuracy, but increases the test cost.

Figure 3. The MCA7707 uses temperature-dependent bridge voltage to change the address in memory, from which correction coefficients for the sensor offset and sensitivity (supply current) are read, enabling a 120-segment compensation of sensor residual error (after analog TC slope compensation).

The residual sensor temperature error is dependent on the slope of the temperature errors. For example, correcting a 6% nonlinearity over temperature with 60 segments (half of the temperature range) with a perfect curve fitting would yield the maximum residual error on the order of 0.1% (6%/60). Figure 3 is a graphical representation of the compensation.

Pressure Nonlinearity Correction
Pressure nonlinearity is corrected by means of a feedback loop from the output voltage to the current source. A DAC is used to control the feedback depth (see Figure 4).

Figure 4. Pressure nonlinearity correction feeds back a fraction of the output voltage to the current source. The feedback ratio is adjusted by a D/A converter.

Feedback increases (with a positive feedback) or decreases (with a negative feedback) bridge current Ib when output voltage is increasing, creating a nonlinear sensor response to the applied pressure. This nonlinearity in the current source can compensate sensor nonlinearity, often by an order of magnitude. The compensation is governed by the following function:

(2)

where:

Ratiometric Voltage Output Configuration
The voltage output configuration is shown in Figure 2. The transducer configuration includes several decoupling capacitors, several bias resistors, and an EEPROM. Despite its simplicity, this configuration's performance is excellent, limited only by the stability and hysteresis of the sensor used. The range of sensors that can be compensated covers most of the resistive devices, and is summarized as follows:

   Offset . . . . . . . . . . . . . . . . ±100% F.S.
   F.S.  . . . . . . . . . . . . . . . . . 5–30 mV/V
   Offset TC  . . . . . . . . . . . . . . . 20% F.S.
   Offset TC nonlinearity  . . . . . . . . . 4% F.S.
   F.S. TC . . . . . . . . . . . . . . . . -20% F.S.
   F.S. TC nonlinearity  . . . . . . . . . . 5% F.S.
   Temperature range  . . . . . . . . -40ºC to 125ºC
   

The MCA7707 delivers a very tight transducer ratiometric output calibration and compensation (@ 5 V):

   Offset  . . . . . . . . . . . . . 0.500 V ±200 µV
   F.S.  . . . . . . . . . . . . . . 4.000 V ±200 µV
   Offset accuracy over temp.  . . ±4 mV (0.1% F.S.)
   F.S. accuracy over temp.  . . . ±4 mV (0.1% F.S.)

The initial offset of 16.4 mV and F.S. of 55.8 mV of a repeatable piezoresistive sensor were converted into 0.5000 V and 4.5000 V transducer outputs, respectively. The nonlinear sensor offset and F.S. temperature errors, which were on the order of 20% to 30% F.S., were reduced to under ±0.1% F.S. The actual test data of the uncompensated sensor and the compensated transducer output are shown in Figure 5 .

Figure 5. Large raw sensor errors (A) are calibrated (B) and temperature compensated (D) to levels not previously achievable with off-the-shelf single-chip components. Sensor temperature nonlinearities are plotted in (C).

4-20 mA Transducer Configuration
The basic MCA7707 ASIC was designed in a ratiometric configuration, but also supports the popular fixed-output 4-20 mA 2-wire transducer configuration. All the basic components of the voltage output mode appearing in Figure 2 are incorporated. To create a 2-wire current loop, an uncommitted on-chip op amp is used, which, along with an external transistor, creates a programmable current source (see Figure 6). The loop current is set by the voltage across resistor RA, and by the feedback loop formed by resistor RC.

Click the circuit diagram to see a larger version.

Figure 6. In a two-wire 4­20 mA circuit configuration, a 4 mA current is used to power a transducer, and an incremental 0-16 mA current transmitted over the same pair of wires is proportional to the measured pressure. Current output permits transmission over a long distance without loss of accuracy due to the cable resistance. The external transistor forms the controllable current loop. The MCA7707 controls the voltage across resistor RA, and if a 50 value is selected, a 0.2­1.0 V range would have to be set during a calibration procedure. If necessary, the programmable gain amplifier output can be divided using resistors RB and RC.

A voltage regulator (e.g., REF02) provides independence from supply voltage changes. The regulator supplies a stable 5 V to MCA7707, used as reference voltage. An additional function of this regulator is to increase the operating voltage. Because the MCA7707 is implemented in a dense CMOS technology, it has a limited maximum voltage to 7 V. Typical 4­20 mA current loops supply a range of 20 V to 40 V, which the voltage regulator reduces to 5 V for the ASIC. The additional components shown in the circuit enhance the circuit's functionality:

• A diode in series with the positive power supply terminal protects against reversed polarity in the field.
• Voltage spike protection is accomplished by a special diode (transzorb) connected across the power terminals.
• A reduction of the power dissipation in the output transistor is provided by optional resistor RD.

Summary
The CompProcessor (A7707) simplifies the manufacturing of transducers (e.g., pressure, acceleration, force). A single automated process handles pretest, calibration, compensation, and final test. MCA7707 is specifically designed to interface with resistive sensors. It can compensate a sensor to better than 0.1% of the sensor's inherent repeatability over a temperature range of ­40ºC to 125ºC. The base technology has exceeded by an order of magnitude the price-performance capability of other competing technologies such as microprocessor-based (digital signal sensor processing) systems by processing the signals in the analog domain.

Home | About Us | Site Index | Privacy Policy © 2006 Questex Media Group, Inc.. All rights reserved. Reproduction in whole or in part is prohibited. Please send any technical comments or questions to our webmaster.