From EDN Europe: Semiconductor sensors bridge the analogue-to-digital divide

A casual observer of the electronics industry could be forgiven for believing that every innovation today is digital. The reality is different, as is clearly seen at the interface between physical properties and electronic circuitry. Today

By David Marsh, Contributing Technical Editor -- EDN, August 3, 2000

From audio to video and including everything in between, today's electronic products appear almost exclusively digital in nature. But as any analogue design engineer will tell you, under the surface, continuous-state electronics are alive and well. Nowhere is this divide more apparent than at the point where physical quantities meet electronic circuitry. As electronics take over from mechanical control systems, the market for sensors continues to grow at a phenomenal rate. One of the best examples is the automotive industry: New cars include just about every type of sensor you can think of and probably several you can't. Domestic appliances are another example; smart washing machines now include—in addition to temperature and flow meters—accelerometers to guarantee smooth load handling. And the manufacturing industry continues to consume huge volumes of sensors for applications that range from process control to robotic manufacture.

To cope with new application challenges, sensor manufacturers develop familiar and unfamiliar technologies. The commonest physical measurements are temperature, pressure, and acceleration; position sensing is another key sensor application (see sidebar "Magnetic fields measure position"). Temperature is still the favourite measurement by far, with thermistors, thermocouples, and resistance-temperature detectors (RTDs) in use throughout the industry. You can find extensive discussions of such devices at process-control vendors' Web sites, such as Omega Engineering's site, www.omega.com). But silicon sensors dominate today's consumer electronics and other applications that require measurements in the typical IC's –55 to 150°C operational range. A mobile phone, for example, may integrate as many as four temperature-sensing elements—one to sense overloads in the power amplifier, another to monitor battery-charge rate, and two in the frequency-synthesiser subsystem to compensate for temperature effects.

Because temperature measurements are so familiar, designers often select discrete sensors without too much thought, tending to use parts they know best (Reference 1). The automotive industry is a case in point: Despite difficulties with linearisation and manufacturing-process dependencies, most temperature sensors in vehicles are still thermistors. The simplest method of measuring temperature with a silicon device relies on the negative temperature characteristics of a pn-diode junction operating at constant current (approximately –2.2 mV/C). But this simple approach requires calibration to compensate for the junction's absolute-voltage value, which varies from device to device. Temperature-sensor ICs, such as Analog Devices' AD590, use two identical transistors that operate at different collector currents, generating a base-emitter difference voltage that's proportional to absolute temperature (the PTAT voltage). In Figure 1, if "r" is the ratio between the collector currents in Q8 and Q11, the differential base-emitter voltage is (kT/q) (ln r), where k is Boltzmann's constant and q is the charge on an electron. Q10 tracks the collector current in Q9 and Q11, forcing device current to equal the PTAT value. Laser-trimmed resistors R5 and R6 calibrate the circuit at 25C and convert device current to voltage, allowing a range of connection options for the two-terminal device, such as to the virtual earth of an operational amplifier. The AD590 costs $3.10/10,000 for the ±3C-accurate J-grade device in a TO-52 package.

A less familiar temperature-sensing method dispenses with pn-diode junctions and uses the spreading-resistance principle, which derives from the one-point method of measuring the specific resistivity within silicon wafers. In the one-point method, a probe makes measurements between a contact point on the wafer's top side and its substrate (Figure 2a). The wafer's resistance is equal to the bulk silicon's specific resistivity divided by p times the diameter of the top-side contact point, so—providing that the top-side contact is negligibly small compared with the wafer's diameter—the measurement is independent of wafer thickness or diameter. For sensor applications, the temperature-detection mechanism relies on the rise in lattice energy in silicon crystal with increasing temperature, so these devices exhibit a positive temperature-coefficient (TC) characteristic. Such devices typically operate from –50 to 150°C, showing a regressive parabolic response that you can easily correct with linearisation tables. (For high-temperature operation, see the Philips KTY84 series that measures up to 300C.)

Infineon uses the spreading-resistance technique in its KTY-series sensors, using a contact hole in the oxide mask to form the measuring point. Using n-conducting silicon with a specific resistance of 7 Wcm and a 22-µm contact hole optimises device characteristics, producing a TC of approximately 0.75%/°C and 1 kWresistance at 25°C. The asymmetric construction and nonohmic contact between the metalisation and silicon in Figure 2a produces a resistor that's dependent on the direction of current flow. To overcome this dependency, KTY-series sensors use a symmetrical construction based on two spreading-resistance sensors in series (Figure 2b). Current then flows through two identical contact holes, and the composite device has a nominal resistance of 2 kW. Excitation current is typically 1 mA, but you must protect the low-capacitance device structure from voltages above 25V peak. Initial device tolerances are specified into 1 and 3% bands. Alternative devices that use the same silicon and a double-sized contact hole are available with a nominal 1-kWvalue at 25°C.

Christian Burrer, director of semiconductor-sensor marketing at Infineon, notes, "Compared with thermistors, KTY temperature sensors demonstrate superior linearity and long-term stability. Tolerances as low as 1% and SOT-23 packages that allow standard pick-and-place assembly provide further advantages in mass-production applications, such as the automotive and HVAC (heating, ventilation and air-conditioning) industries." Burrer observes that spreading-resistance devices aren't intended to compete with pn-junction devices because the process doesn't suit further integration. The typical cost of a KTY sensor in an SMD package is around 15 cents (10,000).

Integration suits temperature sensors

Unlike many semiconductor-sensor processes, pn-diode devices immediately suit IC fabrication. You can select from a range of general-purpose and application-specific temperature-measurement ICs from vendors such as Analog Devices, Dallas, Maxim, National Semiconductor, ON Semiconductor, and TelCom. Targeting general-purpose use, Analog Devices' AD7414 adds a 10-bit ADC and SMBus/I² C-bus-compatible port to a PTAT-based temperature sensor. The AD7414 uses a chopper-stabilised amplifier to provide differential drive to the 0.25C-resolution ADC, and provides new measurement results every 400 µsec. Filter circuits that you can disable attenuate noise on the serial clock lines, which operate at frequencies as high as 400 kHz. An address-selection pin allows as many as eight I² C-bus addresses (see Reference 2 for the latest information on SMBus and I² C interfaces). Tiny SOT-23 surface-mount versions are available with the midscale 0x00 value centred at 0 or 25C. The AD7414 is sampling now, with full production in the third quarter of 2000. Expect to pay around 90 cents in greater-than-10,000 quantities.

Dallas Semiconductor's temperature-measurement ICs include an unusual selection of serial digital interfaces and packaging options. For example, the DS1615 temperature recorder measures –40 to +85°C in 0.5°C increments, with a basic accuracy of 2°C. Communications use a three-wire synchronous serial port or UART, and packaging options include a 16-pin DIP, SOIC, and flip-chip SMD; prices start at $5.43 (1000). The DS1921 shares similar circuitry but adds a lithium battery and packages the stand-alone subsystem within a stainless-steel "iButton" can. Common circuitry includes a real-time clock and paged nonvolatile-memory areas, together with a unique 64-bit serial number that's useful for device tracking in applications such as good-in-transit monitoring (Figure 3). The clock triggers measurements at intervals from 1 to 255 minutes and stores as many as 2048 results in nonvolatile memory. You can programme the IC to stop taking measurements when the datalogging memory is full, or you can set a rollover mode that starts overwriting new results from the first memory location. Result timestamps let you reassemble the reading sequence. In histogram mode, the IC sorts measurement results into 2°C steps and increments the count value in one of 63 associated data bins, together with a timestamp. There's no rollover facility in histogram mode—the IC stops storing results when memory is full. But in the worst-case scenario of each reading being an identical value, there's sufficient memory to store data from 45 days at the maximum 1-minute sample rate.

The iButton package uses a one-wire serial interface in which power and data share one line, requiring only a ground to complete the connection. A dedicated probe makes contact with the can to exchange data with a PC or microcontroller host. The probe powers the datalogger during data exchanges, preserving battery life and letting you retrieve data even if the 3V lithium supply is exhausted. Physically, the iButton has a 16-mm diameter and is 6 mm deep; it's rugged enough to withstand outdoor use, even on vehicles. The DS1921 is available now at $9 (1000), and you can order a $25 evaluation kit that includes the probe from the company's dedicated iButton Web site (www.ibutton.com), where you will also find free driver software.

Other vendors that use one-wire interfaces include Maxim, whose MAX6575 sensor includes a facility for microcontroller-controlled triggering and an interface that accommodates as many as eight devices on one serial line. The MAX6575 costs approximately 79 cents (1000) and is available now in a six-pin SOT-23 package. Because temperature measurements are so common, there's a host of devices that we can't accommodate here, so don't overlook alternative products (see sidebar "For more information..."). For example, National Semiconductor has a range of analogue and digital temperature sensors, as well as excellent design and application information at www.national.com/appinfo/tempsensors. The company's recently introduced LM92 distinguishes itself with a worst-case accuracy of ±1.5°C from –25 to +150°C and uncorrected error of just ±0.33°C at room temperature. The device integrates a 12-bit-plus-sign ADC, two window comparators, and an I² C-bus interface in an eight-pin narrow SOIC package. Available now, the guide price is $2.05 (1000).

ICs take the strain out of pressure

Pressure measurements address applications as diverse as monitoring central-heating pressure in tall buildings, manifold absolute pressure in turbocharged engines, robotic-arm pneumatics, and the human respiratory system. Traditionally the preserve of resistance-strain gauges and quartz-crystal devices, silicon pressure sensors now measure values from millibar to hundreds of bar, relying on the piezoresistive effect to transform pressure into electrical output. Compared with resistance-strain gauges, silicon pressure sensors are more sensitive, are highly linear, and exhibit very low temperature and pressure hysteresis effects. The fatigue-free nature of silicon planar construction also provides high load-cycle stability and is cheaper to manufacture than quartz crystal. And unlike quartz, silicon pressure sensors suit both static and dynamic pressure measurements and offer the potential for integration with active circuitry.

At the core of a typical silicon pressure sensor from Infineon, a diaphragm with four ion-implanted piezoresistors forms a Wheatstone-bridge circuit. Pressure causes the diaphragm's geometry to change, varying each resistance in proportion with mechanical stresses in the crystal. Under compressive tension, resistance reduces as the number of charge carriers increases in the direction of the compressive force; conversely, dilatory tension increases resistance. For maximum sensitivity when the diaphragm deflects, Infineon uses a device layout that increases the value of two oppositely sited resistors while another pair decreases (Figure 4a). If each resistor is initially equal in value and responds equally under pressure, the output voltage is Vin X DR/R. The circuit's response is then linearly proportional to pressure up to the point where the diaphragm deflects sufficiently to cause all common expansion in all four resistors (the "balloon effect").

The sensor's operating range depends on the diaphragm's thickness and its surface area, together with the geometry of the piezoresistors. Infineon classifies its KP and KPY-series pressure sensors into low-pressure (50 - 100 mbar), medium-pressure (250 mbar - 25 bar), and high-pressure (60 - 400 bar) ranges. Alternative construction geometries balance output signal level against balloon-effect non-linearities. The low-pressure sensors use four independent bridge systems in a ring configuration to produce output voltages of around 35 mV from a 5 V supply, two to three times greater than a single bridge at this pressure level (Figure 4b). Each diaphragm is an annular shape and about 20 µm thick. The medium- and high-pressure sensors use single circular diaphragms of between 20 - 300 µm, generating full-scale output voltages that range from 25 mV for a 250-mbar device to 550 mV for a 400-bar device.

One potential problem with all-silicon construction is the material's temperature dependency, but a simple positive TC circuit provides compensation over the normal –40 to 125C operating range. Expect to pay around $5.00/10k for a mid-range device such as Infineon's KP 203, which includes a silicon temperature sensor. Vendors including Fujikura, Honeywell, Kulite, Melexis, Motorola, and SenSym also offer pressure sensor ICs complete with integral or external temperature compensation circuitry. Fujikura's PSM series includes a digitally-programmable ASIC signal conditioning circuit that's mounted underneath the silicon planar pressure sensor in piggyback style (Figure 5); the PSM series is available now starting at around $11.00/10k.

Packaging details determine the sensor's application and contribute to long-term stability. Packaging options range from plastic and metal custom enclosures to conventional DIPs and SMDs, with various pressure port arrangements. A pressure sensor with one side open to the air makes relative-to-atmospheric-pressure (or "gauge pressure") measurements; an absolute pressure sensor measures the unknown pressure relative to a vacuum. Connecting both sides of the diaphragm to unknown pressures simply results in relative pressure measurements. The most rugged construction method uses a Kovar baseplate for hard-mounting the diaphragm within a stainless steel device housing. Kovar is a predominantly iron-nickel-cobalt alloy that has the closest thermal match to silicon, so you'll also find the material in many temperature sensors.

Lowest-cost designs for applications such as monitoring automotive vacuum servos employ modified plastic packages that suit pick-and-place assembly lines. Motorola's latest addition to its pressure sensor family is the MPXV4115V, which comes in an 8-pin SOIC made from a chemical-resistant polyphenyl sulphide plastic that suits harsh environments. The useful measurement range is 1.15 bar with an absolute maximum pressure rating of 4 bar. Internally, the diaphragm and signal conditioning circuitry is soft-mounted to the frame using a room-temperature vulcanisation (RTV) silicone adhesive. A fluorosilicone gel transmits pressure to the diaphragm but isolates the die surface and the bond wires from environmental contaminants. A stainless steel cover provides a flat surface contact area for the user's pressure port and suits O-ring seals. Temperature compensation and scale factor calibration provides an output voltage that's accurate to 1.5% from 0 to 85C at levels; the 4.4 V output span suits direct connection with ADCs. Available now, the guide price is $7.80 in 10k quantities.

The piezoresistor approach is by no means the only way to measure pressure with a silicon device. Infineon uses an integrated surface micromachining technology based on the company's submicron BiCMOS process for its KP120 capacitive sensor IC, sharing a common process technology base with Hall-effect ICs. The KP120 integrates the capacitive polysilicon membranes, a sigma-delta ADC, the signal conditioning circuitry, a digital filter and a DAC for the analogue interface on a single chip. The IC measures absolute pressures from 0.2 - 1.2 bar with an overall accuracy of 1% between 0 and 85C. The KP 120 will be available in full production quantities in Q3/00 in a 8-pin surface mount package, and costs $6.50/10k.

More radically, light measures pressures at temperatures up to 300C in applications such as automotive in-cylinder combustion monitoring. Bookham Technology considers that its application-specific optical circuit (ASOC) technology optimally measures the in-cylinder measurements that improve closed-loop emissions control, also dispensing with devices such as crankshaft position, mass airflow, and knock detection sensors. The technique relies on a passive optical sensor in each cylinder head that connects with a hybrid interface IC through a fibre-optic cable. The hybrid transmits an optical pulse that's modified by cylinder pressure before being reflected and converted to an electrical signal that the engine management system can use. Bookham announced the success of prototype trials at last year's Frankfurt motor show; in the US, Optrand is another company with similarly capable devices (see www.optrand.com for extensive product and application information).

Acceleration propels MEMS

As you drive off in your new car, you may imagine that the airbag inflation sensor in front of you is a state-of-the-art micro-electromechanical system (MEM) device - but there's every chance that you'd be quite wrong. According to Ian Cattermole, sales manager at Hamlin, the company's reed switch-based technology currently has about 80% of the automotive crash sensor market. But Hamlin's associate company VTI produces a range of MEM accelerometers and - together with vendors such as Analog Devices and Motorola - reports tremendous interest from designers who need to accurately measure forces in motion. Today's automotive industry offers accelerometer vendors huge opportunities in occupant restraint and vehicle stability control applications, and this market is sure to grow as vehicles move to drive-by-wire systems (Reference 3). Other fast-growing areas include industrial machine monitoring, robotics, and even white goods (Reference 4).

VTI's capacitive sensor accelerometers are made from single crystal silicon on glass for stability. These devices measure bidirectional acceleration and are internally protected against g-force overloads. The bulk micromachined sensor comprises three silicon layers isolated by thin glass layers, with a cantilevered mass-beam structure at the centre. Bulk micromachining is an etching process, removing silicon along the crystal's axis to form trenches and leave a mass that's suspended by membrane material. Force deflects the resulting silicon beam spring, causing a capacitance change that's detected between metal electrodes within the outer silicon layers. VTI's SCA320 series adds an ASIC onto the ceramic substrate to provide calibrated analogue voltage outputs within a 1.5g range. The hybrid device is available in an 8-pin, 12.7-mm square package; guide price is $23.60/1k+ quantities. The company's SCA600 is a lower-cost range in plastic packages and suits measurements from 61 to 3g; an inclinometer variant measures angles between 20° and provides a ratiometric voltage output.

Motorola also chooses bulk micromachining and hybrid construction for its accelerometer products. Marc Osajda at the company's automotive sensor marketing division observes, "You can combine micromachining with IC fabrication, but device complexity affects yield and isn't necessarily cost effective. [ensp]Another reason for our strategy is that we have different sensing elements with various axis configurations and sensitivities that we can combine with a set of control ASICs, allowing us to quickly develop custom solutions." Off-the-shelf devices include the MMA1201P for z-axis sensing and the MMA2200W for x-axis work; either device has a 40g range and provides a linear ratiometric output centred on Vcc/2 (nominally 2.5 VDC). Measurement bandwidth is typically 400 Hz, and other sensitivities are available on request. On-chip selftest features include g-cell diagnostics, low voltage and clock monitoring circuitry, and calibration EEPROM integrity status checks. The guide price is $7.50/10k.

Meanwhile, Analog Devices uses monolithic construction for its iMEMS (integrated micro-electromechanical system) products (Figure 6). The company's dual-axis ADXL202 accelerometer uses surface machining techniques that are compatible with integrated circuit fabrication, allowing a digital connection to the host microcontroller. The ADXL202 converts force in the range 2g into two pulsetrains, one each for the x- and y-axes, whose duty cycles are proportional to acceleration. You can adjust the duty cycle period in the range 0.5 - 10ms with one resistor, or use the analogue voltage outputs. Bandwidth is also adjustable from 0.01 Hz - 6 kHz using capacitors. The noise floor is low enough to resolve signals under 5 mg in a 60-Hz bandwidth. The device is available in an 8-pin ceramic leadless chip carrier ($12.50/1k) or 14-pin SMD ($15.96/1k) for the commercial grades. To make your job easier, development tools include a reference design, spreadsheet-based interactive design software, and evaluation boards.

Magnetic fields measure position When you think of semiconductor position sensors, you'll probably think of Hall-effect devices in gear-tooth sensing applications. Such devices have proved their worth in tough environments such as camshaft position and anti-lock brake wheel sensors in millions of vehicles. Hall-effect sensors detect the presence of a magnetic field close to a semiconductor slice. A constant current source biases the slice; when no magnetic field is present, the voltage across the slice is negligible. But when a magnetic field appears that's perpendicular to the slice, a voltage appears that's proportional to field strength. With appropriate electrical and magnetic biasing, Hall-effect switches detect ferrous metals at small air-gap distances from DC to about 100 kHz. Hall-effect technology is also useful for less familiar applications that require measurements on relatively high-level magnetic fields, such as indirect AC and DC current sensing (see Allegro's site at www.allegromicro.com for representative details). Magnetoresistors are a less familiar component for detecting magnetic targets and ferrous metals at frequencies from dc to gigahertz. Widely used in hard-disk drive heads, the technology is now finding wider applications in automotive and industrial applications. A basic magnetoresistor (MR) is built from an indium antimonide/nickel antimonide (InSb/NiSb) semiconductor bar that's doped with tellurium. Following a melt process, the nickel-based material solidifies into parallel needles throughout the bar that provide the device's sensitivity to perpendicular magnetic fields. Bars are cut into slices and ground, etched and polished into wafers about 15 - 25µm thick. Conventional photolithography then allows selective etching into various layouts, such as a meandering path that maximises the sensitive semiconductor material content within a device. Arranging two such devices and a biasing magnet in one package constructs a differential sensor with a typical unbiased resistance of around 500W. Adding two external resistors in a Wheatstone bridge circuit generates a cosine voltage output waveform as ferrous gear teeth rotate nearby. By appropriately arranging two differential sensors, you can detect linear angle as well as speed from the resulting quadrature waveform. Like Hall-effect devices, MR sensors require small air gaps (,1mm) but can produce an output signal .1 V peak from a 5 V supply; for a given air gap distance, signal amplitude is independent of the target's rotational frequency. Giant magnetoresistors (GMRs) extend the air gap range to about 25 mm, opening up entirely new applications. A GMR is built from a stack of iron and cobalt layers separated by non-magnetic copper. Layer separation is so small that the cobalt layers couple to from an artificial anti-ferromagnet; Infineon's devices comprise 11 layers with a total thickness of just 25nm. The soft magnetic iron layers that cover the hard magnetic cobalt align with external magnetic fields and produce the maximum resistance change (.4%) to perpendicular fields. Unbiased resistance is typically .700Wand response broadly follows a cosine law; critically, the device responds to magnetic field direction, not strength. Possible configurations include single sensors, a Wheatstone bridge with two passive resistors, and a full-bridge that comprises four GMRs to double the output voltage. A full-bridge device such as Infineon's GMR C6 can also provide sine and cosine outputs for angle sensing applications. Other MR and GMR vendors include Honeywell and Philips.

 

Author Information

You can reach Contributing Editor David Marsh at forncett@ compuserve.com .

REFERENCE

1. Strassberg, Dan, " Stay off the hot seat when choosing temperature sensors ," EDN Sept 1, 1997.

2. See www.smbus.org/specs for SMBus information and www.semiconductors.com/i2c/facts/index.html for I² C data.

3. Marsh, David, " Drive by wire fuels network highway race ," EDN , April 13, 2000.

4. Lemaire, Christophe (Analog Devices), "Direct and indirect out-of-balance detection for future generation washing machines," paper to 1999 Appliance Manufacturer Conference and Expo, Nashville, TN, USA.