Fullscreen HTML5Med Res (??MB)HTML4

Electronic Noses Smell a Big Future By PETER HOLTHAM 8  Silicon Chip Of our five senses – sight, sound, touch, smell and taste, the first three are physical in nature. They also have readily available electronic equivalents. You can buy cameras, microphones and pressure sensors off the shelf to convert light, sound and pressure into electrical signals. Soon, smell sensors will be readily available too. www.siliconchip.com.au S mell and taste are chemical senses, so-called because they detect the presence of different chemicals as molecules in the air (smell) or dissolved in liquids (taste). At present, electronic sensors for both are in their infancy. Smells are simply chemical molecules small enough and light enough to vaporise into the air. A smell may be just one type of molecule or a mixture of many different types. Over 600 different molecules wafting into your nose make up the delicious aroma of fresh coffee, for example. Smell is a vital part of our daily lives and it uses more of the brain than any of the other senses. Smell lets us sample our surroundings and check for danger. Think of the smell of smoke, for example. Molecules of smoke can travel long distances on the wind, showing that smell can act as an early warning system. Even though the human sense of smell is poor compared with many animals, we can easily detect just parts per billion of the toxic gas hydrogen sulphide – the smell of rotten eggs. With training and experience, human noses can check products such as wine, cheese, fish and many other foodstuffs, for quality and freshness. Doctors can diagnose certain diseases from their smell alone. Human noses are sensitive and self-repairing but they are not suited to boring or repetitive tasks. They are also subjective, prone to catching colds and cannot be used to check situations that may be hazardous. Humans cannot smell the fatal presence of carbon monoxide, for example. What we need is an electronic or E-nose, to give an objective readout of the smell-scape that surrounds us. Scientists have been working on E-nose development since the 1980s, their first step being to understand how our biological sense of smell works. How do volatile odour molecules reaching your nose trigger recognition of a smell in your brain? Smell molecules swirl past the turbinate bones to reach the human smell sensors. the eyes, lies the nasal epithelium containing about 5 million smell sensor cells. By comparison, the super-sensitive noses of dogs contain over 100 million sensors. At one end of each sensor cell there are 10 to 20 hair-like smell receptors, bathed in watery mucus. Smell molecules attach to the receptor proteins in the hairs, triggering a cascade of chemical reactions inside the cell. The reactions result in the transfer of sodium ions across the cell membrane in a form of biological amplification. At the other end of the sensor cell there is a connecting nerve or ‘wire’ called an axon. The sodium ions pour into the axon, triggering it to fire with an electrical impulse. Chemical information is now an elec- trical signal on its way to the brain for identification. Bundles of axons from groups of sensors thread their way through holes in the base of the skull. The bundles terminate in two olfactory bulbs, one in each nasal cavity. Inside the bulbs, a cluster of neural networks called glomeruli carry out some signal pre-processing. They function much like Internet routers, sending the electrical impulses for specific smells via mitral cells to the brain. The architecture of the olfactory bulbs results in a 1000 to 1 convergence between individual sensors and the mitral cells. A lot of information about individual sensors gets thrown away but sensitivity increases since contributions from many sensors are The Biological Nose Sniffing sucks a sample of air carrying a smell into your nostrils. A mucus layer on their inner surfaces together with a forest of sticky hairs cleans the air of any stray dust particles. The filtered air swirls past the turbinate bones to the roof of each nostril. Here, just below and behind www.siliconchip.com.au Simplified diagram of the biological smell sensing system. November 2003  9 to appear everywhere smell detection is important. Conductivity Sensors There are two types of conductivity sensor: metal oxide and polymer. Both show a change in resistance when exposed to odour molecules. Thick film metal oxide gas sensors (TGS) have been around since the late 1960s; you can buy them off the shelf from component retailers. They are sintered n-type bulk semiconductor devices made of tin dioxide. The sensor changes in resistance in the presence In a conductivity sensor the resistance of the sensing layer changes when a molecule of gases such as hydrogen, reacts on the surface. carbon monoxide, methane, propane etc. added together. Just 0.1% propane by volFinal signal processing occurs deep ume is enough to decrease the resistin ancient parts of the brain concerned ance of a TGS gas sensor up to 20 times. with motivation, emotion and certain This concentration is well below the types of memory. Actual identification explosive limit for propane. of the smell occurs in the brain’s more The trouble with metal oxide senmodern frontal cortex. sors is that they are not particularly selective and are easily poisoned, esE-Noses pecially by sulphur compounds. They Electronic nose designers are fol- also need a continuous power supply lowing Nature’s plan. They use a of over 500mW to heat up the sensor. sampling device to act as nostrils and Nevertheless, they have found wide an array of chemical sensors to mimic use as gas leak detectors. the olfactory epithelium. Signal proThin film metal oxide sensors using cessing hardware and software takes silicon micro machining methods are the place of the olfactory bulbs and now starting to appear. They use oxthe brain. ides of tin, zinc, titanium and iridium, The difficulty lies in the sensor doped with catalysts such as platinum stage. Until recently the only way and palladium. A micro hotplate to analyse a sample of air was by structure reduces heater power by a using complex and expensive labo- factor of 10, compared with thick film ratory-based instruments such as gas devices. Because thin film sensors chromatographs. Routine analysis of are now being made in high volumes smells with this technology is out of (1000-2000 per silicon wafer) the cost the question. But now new smell sen- per sensor is falling rapidly. sor technologies based on conductivity A second type of conductivity or resonance are beginning to appear. If they can be integrated into low cost chips or modules, E-noses will start Conductivity sensors manufactured by AppliedSensor (www.AppliedSen-sor. com) – micro sensor (left) and thick film sensor (right) . 10  Silicon Chip The AppliedSensor quartz crystal microbalance sensor. The diameter of the crystal is 6 mm. sensor is based on polymers. Cyrano Sciences uses this technology in its “Cyranose 320 handheld electronic nose”. Conductive carbon black is blended homogeneously with different non-conducting polymers. The different blends are deposited between pairs of electrodes as thin films on an alumina substrate. The result is an array of typically up to 32 chemiresistors. When odour molecules come into contact with the resistors, the polymers act like a sponge and ‘swell up’. Swelling progressively breaks carbon black pathways and the resistances increase. Once the smell goes away, the polymers ‘dry out’ and shrink, the conductive pathways rejoin and the resistances decrease. The ratio of the smell-on to smell-off resistances becomes the output of the sensor array. Any individual sensor responds to a variety of odour molecules. By varying the amount of carbon black in the polymer or the polymer itself, an array of sensors can be built to yield a distinct pattern of resistances for different odours. The cost of polymers and carbon Internal details of the AppliedSensor micro conductivity sensor (left) and thick film conductivity sensor (right). www.siliconchip.com.au The principle of the QCM sensor. black is low and the electronic interface is simple, making this ideal portable E-nose technology. An array of 32 sensors per chip is a long way short of human sensing capability but still allows reliable smell recognition with suitable software. gram. That amount of methane in a one-litre container gives a concentration of just 1.4 parts per billion. QCMs can be made to respond to different smells simply by changing the polymer coating but they are most sensitive to volatile organic compounds. The Surface Acoustic Wave (SAW) sensor is a cousin of the QCM, operating at a much higher frequency. An AC signal applied to the input creates an Piezoelectric Sensors This family of sensors also has two members: quartz crystal microbalance (QCM) and surface acoustic wave Polymer sensor principle. (SAW) devices. QCM types consist of a quartz crystal disk a few millimetres acoustic wave that ‘surfs’ over the in diameter with metal electrodes on surface of the sensor to the output. each face. The QCM resonates at a fre- Although the AC signal is recreated quency in the range 10-30MHz when at the output, it is shifted in phase. The phase shift depends on the mass excited by an oscillator. During manufacture, a thin polymer of the sensing polymer layer covering coating is applied to one face to act as the sensor substrate. This in turns the sensing material. Odour molecules depends on the odour molecules adsorb onto the polymer, increasing absorbed. A typical SAW sensor operates at the mass of the QCM and reducing its BITSCOPE AD 9/10/03 1:38 PM Page 1 resonant frequency. QCMs can detect 400MHz but its sensitivity is similar mass changes of as little as one pico- to the QCM. Because SAW devices The Electronic Sensor Technology zNose® using fast gas chromatography with a SAW sensor. can be made using standard semiconductor technology, they are cheaper than QCMs. An American company called Electronic Sensor Technology has already developed the zNose, which combines fast (10 seconds) gas chromatography with a SAW sensor. The main disadvantage of this family is that more complex electronics are needed compared with conductivity sensors. Mosfet Sensors Metal oxide silicon field effect transistors (Mosfets) can be also used as odour detectors. The gate electrode is coated with a catalyst such as platinum and exposed to the air through a window. Smell molecules react with the gate, altering the gate charge and thereby varying the conductivity of the device. The gate and drain of the transistor are connected together to form a 2-terminal device. The voltage (around 2V) at constant current (100µA) is recorded as the sensor response to Digital Oscilloscope Logic Analyzer + from 5 $59 ANALOG = DIGITAL Convert your PC into a powerful Scope and Logic Analyzer! Now you can analyze electronic circuits in the analog and digital domains at the same time. BitScope lets you see both analog AND digital logic signals to find those elusive bugs. USB and Ethernet connectivity means you can take BitScope anywhere there is a PC or Network. BitScope Hardware • 100MHz Input BW • 40MS/s Sample Rate • Dual 32K Buffers • 4 Analog Inputs • 8 Digital Inputs • Waveform Generator • SMART POD Probes www.siliconchip.com.au BitScope Software • Windows or Linux • TCP/IP Networking • Advanced DSP • Digital Scope • Analog Scope • Logic Analyzer • Spectrum Analyzer Applications • Electronics Labs • Remote data logging • Engineering students • Scientific research • Robotics and control www.bitscope.com USB or Network connection to Windows and Linux PCs! November 2003  11 The Cyranose® 320 portable E-nose manufactured by Cyrano Sciences (www. cyranosciences.com), photo courtesy of Cyrano Sciences. The AppliedSensor MOSFET sensor construction. An AppliedSensor 1.5mm x 1.5mm MOSFET sensor chip on a TO8 header. the smell. These sensors respond to gases like hydrogen, hydrocarbons, ammonia and carbon monoxide. With a silicon carbide substrate instead of plain silicon, Mosfets can operate at temperatures up to 600°C, as in car exhausts, for example. Processing the Signals Sensors are just part of the E-nose story. Adding the electronic equivalent of olfactory bulbs and the brain turns the raw sensor data into a recognised smell. Two stages are normally required: signal pro-cessing and pattern recognition. Signal processing compensates for baseline drift and reduces sample-to-sample variation. The signals from an array are often also scaled or normalised to cover a similar range. Pattern recognition is the crucial step in identifying a smell from the processed data. Firstly, extracting some features from the data reduces the dimensions of the measurement space. Consider the 32 outputs of a conductive polymer sensor chip. The measurement space will have 32 dimensions. This can cause problems in analysis of the responses, not the least of which for humans is trying to visualise a 32-dimensional hyper-space. Often the sensor responses will overlap, so there is a lot of redundan12  Silicon Chip cy in the 32 dimensions. Complex mathematics are used to project the 32 onto a smaller space, preferably in two or three dimensions which can be visualised by humans. Once in a lower dimensional space, the odour pattern can be classified by comparison with known smell responses stored in a database. Here again, complex mathematical techniques such as artificial neural networks are used. These ensure that an unknown smell is matched to the most likely known smell in the database, even if the match is less than perfect. Applications With new chip level sensors becoming available and abundant computing power to process the responses, where are the E-nose applications? The answer is almost everywhere, your car could soon have several, your home several more. A silicon carbide Mosfet exhaust gas sensor can respond fast enough to monitor the air-fuel ratio of individual cylinders in a car. Thin-film conductivity sensors will soon be monitoring cabin air quality, opening and closing fresh air vents as required. In the home, sensors will also monitor air quality, sniffing out carbon monoxide, an early indicator of a fire. One day soon they might find their way into your coffee machine to check that your morning cup is just the way you want it. E-noses are finding widespread use in the food and drink industry. Customers rely on aroma as an indicator of the quality of the food they buy. E-noses are already monitoring the exact ripeness of fruit and vegetables and the quality of fish, cheese, meat and many other foods. Doctors have used smell as a diagnostic tool for centuries. Commercial E-noses are already being tested for rapid diagnosis of lung cancer. They are also being used to screen bacterial cultures for early detection of lethal bugs. Recent events have made everyone aware of terrorism. A major force behind E-nose development in the USA is the need to replace sniffer dogs checking for explosives. Smell sensing technology is still in its infancy but the hardware and software are now starting to appear. More research and development is required but the day of low cost electronic noses all around us is fast approaching. SC Acknowledgement The assistance of Olivia Deffenderfer, Applications Scientist at Cyrano Sciences and Jan Mitrovics, Executive Director Germany, at AppliedSensor GmbH with the preparation of this feature is gratefully acknowledged. The response of an AppliedSensor MOSFET sensor to exhaust gas composition, showing gas from individual cylinders. www.siliconchip.com.au