Fullscreen HTML5Med Res (??MB)HTML4

Wideband Oxygen Sensor Controller Mk.2 Accurately measure air/fuel ratios with an improved oxygen sensor Are you involved in car modifications? Have you improved the inlet air-flow or modified the exhaust line with new mufflers? Has your engine been “re-chipped” to improve the timing and fuel parameters or have you fitted bigger injectors? If you answered yes to any of these, then you need to check that your engine is not running too lean or too rich. To do that you need to fit a wideband oxygen sensor and build our improved Wideband Oxygen Sensor Controller. B ACK IN SEPTEMBER and October 2009 we published the original and very popular Wideband Oxygen Sensor Controller. This was designed for use with the Bosch LSU4.2 wideband oxygen sensor. In this substantially revised design, we use the much-improved Bosch LSU4.9 sensor which supersedes the LSU4.2. This has necessitated an upgraded microcontroller, the addition of a sensor to monitor exhaust pipe pressure and a re-designed LED display module. 32  Silicon Chip Most modern vehicles include a narrowband oxygen sensor so that the engine control unit (ECU) can control the air/fuel ratio. Unfortunately, that sensor is only accurate when the fuel/ air mixture is stoichiometric, ie, when the mixture is exactly right to give complete combustion and with all the oxygen used in the burning process. The engine control unit (ECU) normally adjusts the fuel mix to maintain an oxygen sensor signal that’s close to 450mV, the stoichiometric point. In practice, a narrowband sensor has a very sharp voltage change around the stoichiometric point and so the sensor voltage is continually cycling above and below 450mV as the ECU maintains the fuel mixture. This is referred to as “closed loop” operation. It does not matter to the ECU that the narrowband sensor is inaccurate and non-linear outside closed loop operation. To explain further, Fig.1 shows the typical output from a narrowband siliconchip.com.au Pt.1: By JOHN CLARKE oxygen sensor. It has a very sharp response either side of the stoichiometric point (lambda of 1), ranging from about 300mV up to 600mV; the classic “S” curve. For rich mixtures, it ranges from around 600mV to almost 900mV (lambda up to 0.8), is quite non-linear and varies markedly with temperature. It is similarly non-linear for lean mixtures, ranging from around 300mV down to a few mV (lambda of about 1.15). To learn about lambda, refer to the explanatory panel later in this article. The ECU uses its own factory preset values to set rich mixtures for acceleration or lean for cruise conditions. This is referred to as “open loop” operation because the oxygen sensor is not capable of providing accurate feedback about the actual fuel mixture. Now if you haven’t changed anything on your vehicle, then there is little reason to worry about the actual fuel mixtures at any time; the ECU takes care of it all. But if you have made any changes to the vehicle to improve its performance (eg, inlet air filter, throttle body and plenum, injectors, MAP or MAF sensor, custom ECU chip, supercharger or turbocharger, catalytic converter, exhaust manifold, mufflers and resonators, in short, anything that’s likely to result in significant changes to fuel mixtures and oxygen sensor readings) then you need a wideband oxygen sensor and a companion controller. Bosch LSU4.9 oxygen sensor As stated, our new controller is designed to work with a Bosch LSU4.9 wideband oxygen sensor. This sensor is now used in some late-model cars to measure and control the mixtures over the full range of engine operation. Main Features • Accurate lambda measurements on 3-digit display • • Pre-calibrated sensor • S-curve (narrow band sensor) simulation output for ECU • • Heat/data/error indicator LED • Correct sensor heat-up rate implemented • Heater over and under-current shutdown Pressure and temperature correction of lambda reading Adjustable engine-started battery voltage threshold Fig.2 shows the wideband controller output using the Bosch LSU4.9 sensor over a wide range of air/fuel ratios from 0.7 lambda to 1.84 lambda. Our Wideband Oxygen Sensor Controller is housed in a small plastic case, as shown in the accompanying photo. As well as providing an 8-pin socket (CON5) for the wideband oxygen sensor, it has two jack sockets. One of these (CON3) drives a companion 3-digit LED display unit which shows the lambda value. The other jack (CON4) provides a S-Curve Output vs Lambda 1000 900 OUTPUT (millivolts) 800 RICH 700 600 500 400 300 200 LEAN 100 0 0.8 0.9 1 Lambda () 1.1 1.2 Fig.1: the S-curve output from the Wideband Controller simulates a narrowband sensor output (the response follows the Bosch LSM11 narrowband sensor curve). Note the steep slope in the curve at stoichiometric (ie, lambda = 1). siliconchip.com.au Fig.2: the wideband output from the Wideband Con­ troller is linear with respect to lambda values from 0.7-1.84. The resulting signal is displayed as a lambda value on the Wideband Display Unit to be described in Pt.2 next month. June 2012  33 more slowly if there is a sensor error or if the air/fuel ratio is outside its measurement range. +12V Rcal Rcal Ip Vs/Ip WIDEBAND SENSOR Heater Vs Ip SIMULATED NARROW-BAND SENSOR SIGNAL Rcal Ip Vs/Ip Vs H– H– H+ H+ WIDEBAND CONTROLLER +12V 0–5V OUTPUT GND GND2 GND1 Why do you need it? 8.8.8 WIDEBAND DISPLAY Fig.3: here’s how the Wideband Controller is used with a wideband oxygen sensor and with a Wideband Display Unit (to be described in Pt.2), to provide accurate air/fuel mixture readings. As shown, the Wideband Controller has both a display output and a simulated narrowband (S-curve) output. signal which simulates the output from a narrowband sensor. This enables the vehicle’s existing narrowband sensor to be replaced with the Bosch LSU4.9 and still provide for normal ECU operation. As far as the car’s ECU is concerned, the simulated signal is what it would get from a narrowband sensor and so engine operation is normal. By the way, it’s possible to use the wideband sensor by temporarily installing it into the end of the exhaust pipe, as will be detailed in Pt.2 next month. You might want to do this for easy monitoring of changes to different vehicles. However, the ideal installation is to substitute the original narrowband sensor with the Bosch LSU4.9. A description of the new Bosch sensor is provided in an accompanying panel. Another feature of our new Wideband Oxygen Sensor Controller is an on-board sensor to measure pressure in the exhaust system. We’ll talk more about this later. A red status LED on the front panel indicates when the controller is heating the sensor to its operating temperature. This occurs each time the controller is switched on and it takes less than 10 seconds for the operating temperature to be reached. Once the sensor is at operating temperature, this LED then flashes rapidly. From that point on, the wideband controller is monitoring the signal from the oxygen sensor and feeding a simulated narrowband signal to the ECU. By contrast, the LED flashes Fig.4: inside a narrowband zirconia oxygen sensor. It consists of a zirconia ceramic sensor element with thin platinum electrodes on both sides. 34  Silicon Chip So why is the Wideband Oxygen Sensor Controller necessary? It’s be­ cause a wideband sensor is very different from a narrowband sensor. In its most basic form, a narrowband sensor has only one wire and this is the sensor output. The other connection is via the metal frame of the unit. However, some narrowband sensors have an internal heater and these units may have three or four wires. By contrast, a wideband sensor has six wires (yeah, we know the socket on our controller has eight pins – be patient). This is because the wideband sensor comprises a narrowband oxygen sensor, a heater and an oxygen ion pump which diffuses oxygen ions into or out of the measurement chamber (of the narrowband sensor). The heater and oxygen ion pump need to be controlled externally from the sensor and this is where the Wideband Oxygen Sensor Controller comes into the picture. But we are getting way ahead of ourselves . . . Fig.3 shows the basic set-up. At left is the wideband sensor with its six leads which are all connected to the wideband controller. As already mentioned, this provides a simulated narrowband sensor signal which feeds the ECU. In addition, there is an output to drive the 3-digit Wideband Display Unit. Before we describe how a wideband sensor and its associated controller work, it’s necessary explain the characteristics of a narrowband sensor. Fig.4 shows a cross-section of a typical narrowband sensor. It’s about the same size as a spark plug and is threaded into the exhaust system so that the sensor is exposed to the exhaust gasses. The assembly is protected using a shield that includes slots so that the exhaust gases can pass through into the sensor. The sensor itself is made from a zirconia ceramic material that has a thin layer of porous platinum on both sides. These platinum coatings form electrodes to monitor the voltage produced by the zirconia sensor as the exhaust gas passes through it. For the chemistry-minded, the sensor is called a “Nernst cell”. The device operates by measuring siliconchip.com.au DIFFUSION GAP EXHAUST 20 A REFERENCE CURRENT PUMP CELL O 2¯ MEASUREMENT CHAMBER ZrO 2 O 2¯ CONTROLLER LOGIC PSEUDO REFERENCE SENSOR CELL HEATER WIDEBAND DISPLAY OUTPUT NARROWBAND OUTPUT (SIMULATED) 450mV REFERENCE Vs Vs SENSE COMPARATOR HEATER ELEMENT H+ 62 DIFFUSION PATH Vs/Ip ZrO 2 Ip SENSE AMPLIFIER Rcal Ip ZrO 2 Ip Rcal H– WIDEBAND SENSOR WIDEBAND CONTROLLER Fig.5: the basic scheme for a wideband oxygen sensor and its associated control circuit (at right). the difference in oxygen content between exhaust gas and outside air. The oxygen content of air (about 20.95%) serves as the reference (reference air). In operation, a voltage is produced between the electrodes because the zirconia sensor has a high conductivity for oxygen ions at high temperatures. Some narrowband sensors include a resistive heating element to ensure that they operate within their correct temperature range. The heater also quickly brings the sensor up to its operating temperature and thereby allows the ECU to provide closed-loop operation earlier than would otherwise be possible. So with that brief description of a narrowband sensor under your belt, take a look now at Fig.5. This shows the internal cross-section of the wideband sensor on the left and the connections to the controller on the right. The wideband sensor includes a clever method to obtain a wider, more linear response from a narrowband sensor. This involves a measurement chamber incorporating a pump cell into which a small sample of exhaust gas enters via a diffusion gap. The pump cell moves oxygen ions into or out of the measurement chamber gap in order to maintain a stoichiometric measurement for the sensor cell. For our purposes, the sensor cell is a narrowband sensor. If the measured mixture is lean, then the sensor cell detects excess oxygen. The pump cell then drives oxygen ions out of the measurement chamber until the sensor cell produces a stoichiometric lambda value. Conversely, if the mixture is rich, oxygen ions are pumped from the surrounding exhaust gas into the meassiliconchip.com.au Fig.6: this graph plots Ip (pump current) versus lambda for the wideband sensor. urement chamber gap until the sensor cell again reaches its stoichiometric lambda value. When the mixtures are lean, there is oxygen available in the measurement chamber for the oxygen ions to be transferred. Conversely, when the mixture is rich, oxygen ions for both the pump cell and the pseudo reference chamber are obtained (reduced) from the available oxygen in the sampled exhaust gas. This available oxygen can be reduced from exhaust gases such as carbon dioxide (CO2) and steam (H2O). As a result of the above, the current applied to the pump cell can be either positive or negative, depending on whether oxygen is pumped into or out of the measurement chamber. The oxygen pump thus maintains a stoichiometric lambda value within the measurement chamber. So while the narrowband sensor (sensor cell) is used to detect the stoichiometric mixture, it is the current applied to the pump cell that provides the necessary information to accurately determine the air/fuel ratio. If this sounds like “black magic” then that’s not far from the truth. Most wideband sensors (including the older Bosch LSU4.2) utilise a narrowband sensor similar to the Fig.4 arrangement June 2012  35 Specifications Power requirement: 11V to 15V. Start-up current when heating is 1.6A (~20W) and typically 0.6A (7.5W) when up to temperature. Reading accuracy: typically 1%. Measurement range: 0.7 (rich) to 1.84 (lean) lambda. Reading error indication: LED flashes at 1Hz rate for <0.7 or >1.84 lambda. Engine started battery voltage threshold: adjustable to 15V; 13V setting typical (TP2 = 4.17V). Sensor heating: preheat begins at an effective 2V for 2s then at an effective 7.2V and ramps up at 73.3mV/187.5ms (equivalent to 0.39V/s). Heater maximum effective voltage (Veff): 12Veff after initial preheat and at 13Veff for <30s. Heat-up period: typically <10s. Heater over current error: 4A. Fuse protection: 5A. Heater open-circuit detection error: if current is less than 390mA at initial preheat. Heater drive frequency: 122Hz. Sensor temperature: controlled at 780°C by maintaining the 300Ω impedance of the sensor cell at that temperature. Can be measured as 684mV DC at the wideband output with JP1 inserted. Temperature correction: Ip corrected for sensor temperature between 698°C and 880°C. Pressure correction: Ip corrected for pressures up to 587hPa above standard atmospheric pressure of 1013hPa. Pressure offset adjustment: between sea level (1013hPa) and 2000m (766hPa) above sea level. VR6 adjusted for 1V/1000m when the sensor is plugged for gauge pressure readings. Sensor cell measurement: AC drive at 1.953kHz and 243µA. Sensor cell DC loading: <4.5µA. Reference Current: 20µA. Wideband output: linear 0-5V output for 0.7-1.84 lambda. S-curve output: simulates a 0.8-1.17 range following the Bosch LSM11 sensor curve. S-curve response: 100ms time constant. Wideband reading response: 100ms to a 5% change in oxygen. Indicator LED: pre-startup and 2Veff warm up = dim; during sensor preheat = fully lit; controlled with data = 16Hz flash; error = 1Hz flash. An overheated sensor is indicated with the dim LED. WHERE TO FIND DATA (1) Data for the LSU4.2 and LSM11 sensors mentioned is available at http:// www.bosch.com.au/content/language1/downloads/Section_D.pdf (2) Data on the Bosch LSU4.9 oxygen sensor is available at http:// www.breitband-lambda.de/media/Dateien%28Lambda%29/ LSU49TechProductInfo.pdf (3) A description of the operation of wideband sensors and the difference between the LSU4.2 and LSU4.9 is found at http://www.ee.kth.se/php/ modules/publications/reports/2006/XR-EE-RT_2006_008.pdf.junk (4) More information on oxygen sensors in general can be found at http:// chemistry.osu.edu/~dutta/index_files/Recent%20Publications_files/ Ramamoorthy_R.pdf 36  Silicon Chip where it has a reference air-chamber. However, the Bosch wideband LSU4.9 sensor does away with the reference air, utilising a “pseudo reference” chamber instead. It is truly a clever device. For the pseudo reference, excess oxygen is maintained in this chamber by applying a small reference current to the sensor. This current transfers oxygen ions from the measurement chamber to the pseudo reference chamber. For this chamber to act as a reference, the driving reference current must be sufficient to maintain excess oxygen in the pseudo reference chamber. As with the pump cell, this oxygen comes from the exhaust gas. The partial gas pressure between the two chambers is equalised by having a diffusion path opening in the pseudo reference chamber. The pseudo reference chamber is a big advance because a reference air-chamber needs to be constantly replenished with oxygen from the outside air and the only way oxygen can enter the sensor is via the sensor leads, ie, between the copper wire and its surrounding insulation, a pretty tortuous route! Any contamination of the sensor leads from oils, tars and fuels can affect the oxygen flow to the sensor. The leads are also susceptible to damage if the sensor lead connections are soldered (instead of crimped), as this will melt the wire insulation sufficiently to seal the wire against oxygen flow. However, for a pseudo reference, oxygen replenishment is not affected by sensor lead contamination since it derives its oxygen via a different pathway. It should be noted that both the reference air-chamber and the pseudo reference chamber, whichever is deployed, will be depleted of oxygen over time unless it is continuously replenished. That is because any oxygen in the reference chamber will ultimately diffuse into the measurement chamber to balance out the oxygen partial pressure that is higher in the reference chamber (for the chemistry minded, this is due to Fick’s First Law). Now have another look at the block diagram of Fig.5. Vs is the output voltage from the oxygen sensor cell while Ip is the current into or out of the pump cell. At the stoichiometric point, Vs is 450mV and this is compared against a siliconchip.com.au BUFFER FILTER siliconchip.com.au 10k PWM1 (IC1) AN6 (IC1) x25.45 100nF TP3 (IC4b) Rcal (IC3b) 62 TP12 Ip 20 A AMPLIFIER Vs TP11 AN10 (IC1) x4.7 + PUMP CELL SENSOR CELL (IC3a) OFFSET BUFFER TP4 TP1 Vs/Ip 3.3V +5V VR4 (IC4a) Fig.7: this diagram shows the general arrangement for the pump control and the sensor cell measurement. Buffer stage IC4b supplies current to the pump cell via trimpot VR5 and the paralleled Rcal and 62Ω resistors. The other side of the pump cell connects to a 3.3V supply (formed using buffer stage IC2b and set by trimpot VR3 – see Fig.12). IC3a monitors and amplifies the sensor cell voltage (Vs) by 4.7. Ip Variation with Pressure 20 Ip/Ip at 1013 hPa (%) lambda > 1 15 10 lambda < 1 5 0 –5 Fig.8: this graph shows how Ip (pump current) varies with pressure. The effect on Ip with pressure is greater for lean mixtures (lambda>1). The wideband controller corrects for pressures up to 587hPa above standard atmospheric pressure of 1013hPa (ie, up to 1600hPa). 1900 1800 1700 1600 1500 1400 1300 1200 1100 1000 900 –15 800 –10 700 450mV reference. If Vs is higher than the 450mV reference, the mixture is detected as “rich” and the Vs sense comparator output goes high. This “informs” the controller logic that Ip needs to change, to pump oxygen ions into the diffusion gap in order to regain a stoichiometric measurement. Similarly, if Vs is lower than the 450mV reference, the exhaust mixture is detected as “lean” and the comparator output goes low. As a result, the controller adjusts Ip to pump oxygen out of the diffusion gap. Note that if there is no Ip control, the sensor cell behaves like a standard narrowband sensor with an output voltage above 450mV for rich mixtures and below 450mV for lean mixtures. However, with current control, the pump current is adjusted to maintain a 450mV reading from the sensor cell. Variations in the sensor cell voltage indicate the change in mixture in either the rich or lean direction, while Ip (the pump current) shows whether the mixture is actually rich or lean. A negative Ip indicates a rich mixture and a positive current indicates a lean mixture. The Ip level indicates the lambda value. Fig.6 shows a graph of Ip versus lambda for the wideband sensor. The lean region curve (lambda from 1-1.84) was developed from a graph of Ip versus oxygen concentration provided in the Bosch LSU4.9 data and the equation: Lambda = [(Oxygen% + 3] +1] ÷ [1 - 4.77 x Oxygen %]. For the rich region, a 4-step graph provided in the LSU4.9 Bosch data sheet is used. Another calculation is made to convert the lambda value to the voltage required at the wideband output as shown in Fig.2. Similarly, the lambda value is converted to an S-curve response for the simulated narrowband (S-curve) output as shown in Fig.1. Ip is sensed by measuring the voltage across a 62Ω 1% resistor (in parallel with Rcal). However, during the calibration of each sensor, the actual resistor used by Bosch is 61.9Ω (a 0.1% tolerance value from the E96 range). Rcal is trimmed so that the voltage across this resistor, measured against lambda, is the same for each sensor. In fact, Rcal can be a value ranging between 30Ω and 300Ω, depending on the characteristics of the individual sensor. The value for Ip shown on the vertical axis of Fig.6 is therefore not the total pump current. AMPLIFIER Rcal VR5 Pressure in hectoPascals (hPa) In the graph, Ip only relates to the voltage across the 62Ω resistor. So while Fig.6 shows Ip varying between -1.85mA and 1.07mA, the actual total current range could vary from -2.23mA to 1.29mA if Rcal is 300Ω or -5.67mA to 3.28mA if Rcal is 30Ω. This total current needs to be supplied by the wideband controller circuit. Pump sensor control Fig.7 shows the general arrangement for the pump sensor control. As can be seen, a filtered pulse width modulated (PWM) signal from a microcontroller (IC1) is applied to buffer stage IC4b. This in turn supplies current to one side of the pump cell via trimpot VR5 and the paralleled Rcal (located inside the wideband sensor) and 62Ω resistors. The other side of the pump cell con- nects to a 3.3V supply. When the output of IC4b is at 3.3V, there is no current through the pump cell. For positive current through the pump cell, IC4b’s output goes above 3.3V. Conversely, when IC4b’s output is below 3.3V, the pump cell current is negative. In practice, IC4b’s output can swing between 5V and 0V to allow for the current range required for the lambda extremes of measurement (0.7 to 1.84). The pump cell current (Ip) is monitored using op amp IC3b which operates with a gain of 25.45. Its output is in turn monitored using the AN6 input of microcontroller IC1. Op amp IC3a monitors and amplifies the sensor cell voltage (Vs) by 4.7. The 20µA reference current is also applied to the sensor cell at this point. Note that while this is called a reference June 2012  37 Advantages Of The LSU4.9 Oxygen Sensor In September and October 2009, we published a wideband controller based on the LSU4.2 wideband sensor from Bosch. While this sensor is similar in many respects to the LSU4.9, the latter has some distinct advantages. Perhaps the most important advantage is that the sensor now has a pseudo reference for oxygen that replaces the atmospheric air reference of the LSU4.2. For the LSU4.2, this reference air had to pass through the leads of the sensor and this made it prone to problems due to contamination with oils, tar and fuels preventing the flow of the required oxygen. The LSU4.9 is not subject to any contamination that can cause measurement inaccuracies. Other advantages of the LSU4.9 over the LSU4.2 are a faster response to mixture changes, a faster heat-up period and a revised higher resistance for the sensor cell. So while the sensor cell of the LSU4.2 has a resistance of 80Ω at its 750°C operating temperature, the LSU4.9’s sensor cell is at 300Ω at its operating temperature of 780°C. This higher resistance results in a more accurate measurement of the sensor temperature. The way in which the LSU4.9 is used with a Wideband Controller also differs from the LSU4.2. While the LSU4.2 heater could be driven from a PWM (pulse width modulated) voltage at 2Hz or more, the recommended heater-driver PWM frequency for the LSU4.9 is greater than 100Hz. Additionally, the ramping up of heating applied to the sensor has been revised to include a preheat at low voltage. These measures ensure that the sensor is not damaged due to thermal shock or from moisture during sensor heating. Air/fuel ratio & lambda Lambda is simply the ratio of the actual air/fuel ratio to the stoichiometric air/fuel ratio. For petrol, the stoichiometric air/fuel ratio (the mass of air required to completely burn a unit mass of fuel) is 14.7:1. However, this can drop to 13.8:1 when 10% ethanol is added. A lambda of 0.7 for petrol is equivalent to an air/fuel ratio of 0.7 x 14.7 = 10.29:1. Similarly, a lambda of 1.84 is equivalent to an air/fuel ratio of 27.05:1. The stoichiometric air/fuel ratio is typically 15.5:1 for LPG and 14.5:1 for diesel. These values can differ depending on the actual fuel composition and for diesel it varies between winter and summer. Lambda is probably the best measure of air/fuel mixtures since it is a universal value and not dependent on the specific fuel. current, it is not a critical value and the word “reference” indicates that the current is for the pseudo oxygen “reference”. Note also that any variation in the reference current does not affect the calibration of the wideband sensor when it comes to accurately measuring the oxygen content in the measurement chamber. Instead, that calibration depends on the Rcal adjustment. The reference current must be sufficient to constantly maintain excess oxygen in the pseudo reference. The recommended current to do this is 20µA. Trimpot VR4 is used to provide an offset voltage which is buffered by IC4a and is set so that IC3a’s output is 2.5V when the sensor cell voltage is 450mV. The microcontroller monitors IC3a’s output at its AN10 input and varies the pump current to maintain a 2.5V reading. This effectively maintains the sensor cell for monitoring stoichiometric mixtures. The measured Ip value when the 38  Silicon Chip sensor cell is measuring stoichiometric is used to determine the lambda value. One complication with Ip is that it is dependent on exhaust pressure which is always above atmospheric pressure. Fig.8 shows the change in Ip versus pressure. As a result, our Wideband Oxygen Sensor Controller provides pressure correction up to 587hPa above standard atmospheric pressure (1013hPa). At this pressure, Ip is reduced by 12% for lean mixtures and 9% for rich mixtures. This correction requires an air-hose connection from the exhaust manifold to the Wideband Controller. It is optional though. If you don’t utilise pressure correction, then the readings can be manually corrected using the graphs of Fig.6 and Fig.8. Note that the exhaust pressure does not have any effect on stoichiometric readings because Ip at stoichiometric is zero. Another complication is that Ip is also dependent on temperature. As a result, any variation in the sensor cell temperature will affect the Ip readings, resulting in inaccurate lambda values. Fig.9 shows how the sensor cell resistance varies with temperature. The change in Ip with temperature is some 4% per 100°C. There are two ways to ensure the lambda readings remain accurate. One way is to correct for the effect of temperature using the graph and the 4% change per 100°C. We actually do this in the Wideband Controller itself but it is only useful for small temperature changes when variations in exhaust gas flow across the sensor can cause a momentary temperature variation. The main method to ensure accurate readings is to maintain a constant temperature for the sensor. That’s done by using the sensor’s heater. Heater element control In this case, the Wideband Controller maintains the sensor’s temperature at 780°C. In operation, its temperature is measured by monitoring the impedance of the sensor cell. This has high impedance at room temperature, falling to 300Ω at 780°C. The impedance of the sensor cell is measured by applying an AC signal to it. Fig.10 shows the circuit arrangement. A 5Vp-p (peak-to-peak) AC signal is applied to the sensor cell via a 220nF capacitor and 10kΩ resistor. The capacitor blocks DC and the resistor forms a voltage divider with the impedance of the sensor cell. When the sensor cell has an impedance of 300Ω, the voltage swing across it is 145.6mV peak-peak. IC3a has a gain of 4.7 so its output is 684.5mV peak-peak. The microcontroller measures this 684.5mV signal at its AN10 input and maintains the 300Ω sensor impedance by controlling the heater current. Fig.11 shows the heater control circuit. Mosfet Q1 is connected in series with the heater element across the 12V supply and is driven by a PWM signal from IC1 (RB7). The heater current is monitored via a 0.1Ω resistor in series with Q1’s source and the resulting voltage across this resistor is filtered using a 22kΩ resistor and 100µF capacitor and fed to input AN4 of the microcontroller. If the heater is disconnected or goes open circuit, the lack of current will be detected and this will switch off the Wideband Controller. Similarly, if the heater current besiliconchip.com.au Sensor Cell Resistance versus Temperature 10000 Ip = 4%/100°C 1000 Sensor Cell Resistance () comes excessive, the controller will switch off Q1 and the heater. Note that there is a strict “ramp-up” of power that must be applied in order for the sensor to be heated gradually. This is to prevent thermal-shock damage to the ceramic sensor. It works like this: initially, the sensor is not heated until the engine starts and this allows any condensation to be blown out of the sensor. Then there is a sensor preheat period that begins with an effective 2V being applied to the heating element for two seconds. The heater voltage then increases to an effective 7.2V which then ramps up by 73.3mV every 187.5ms. This is equivalent to 0.39V/s and just under the maximum ramp-up rate of 0.4V/s specified by Bosch. In order to set the effective heater voltage, we also have to monitor the battery voltage to calculate the required duty cycle of the PWM waveform. In addition, the battery voltage is monitored to detect when the engine has started. Basically, the circuit detects when the battery voltage rises above its normal resting voltage with the engine is off. This rise occurs when the engine is started and the alternator begins charging the battery. In practice, the battery voltage can vary from around 12.5V with the engine off to more than 14V when the battery is charged. As shown in Fig.11, the battery voltage is measured using a voltage divider comprising 20kΩ and 10kΩ resistors, together with a 100nF capacitor to filter out voltage spikes. In operation, the impedance of the sensor cell is constantly monitored and as soon as it reaches 300Ω the preheat is complete and power to the heater is controlled to maintain this value. Once the sensor has reached operating temperature (780°C), the pump control circuit begins to operate. 300  at 780°C 100 10 600 700 siliconchip.com.au 1100 1000 1200 Fig.9: this graph shows how the sensor cell impedance varies with temperature. The change in Ip with temperature is about 4% per 100°C. 20 A REFERENCE CURRENT 5Vp-p (1.953kHz) RB6 (IC1) 220nF 10k AN10 (IC1) TP11 (IC3a) 300 3.3V 684.5mVp-p x4.7 + SENSOR CELL AMPLIFIER 145.6mVp-p Vs Vs/Ip SENSOR CELL IMPEDANCE MEASUREMENT Fig.10: the temperature of the sensor cell is monitored by measuring its impedance using the circuit configuration shown here. +12V +12V H+ HEATER ELEMENT BATTERY VOLTS AN3 (IC1) 20k H– Circuit description Refer now to Fig.12 for the complete circuit details. It’s based on a PIC16F1507-I/P microcontroller (IC1) and we have used nine of its 10-bit analog-to-digital (A/D) converters and three of its PWM outputs. It runs with an internal 16MHz clock oscillator. The remainder of the circuit consists of a pressure sensor, Mosfet Q1 (to control the oxygen sensor heater), some op amps and a few other components. The op amps are rail-to-rail types 900 800 Temperature of Sensor Cell (°C) RB7 (IC1) 10 22k AN4 (IC1) D G S 10 F EARTH1 Q1 MOSFET 10k 100nF 0.1 EARTH2 HEATER CONTROL Fig.11: the heater element is connected in series with a Mosfet (Q1) that switches the power on and off at 120Hz. Temperature control is achieved by driving the Mosfet with a PWM signal to vary its duty cycle. June 2012  39 CON1 +12V D1 1N4004 F1 5A A REG1 LM317T 10 K 100 F 16V VR1 500 A GND1 4 1 MCLR Vdd AN0 PWM4 10k PWM1 MPX2010 PRESSURE SENSOR IC2: LMC6484AIN TP5V 19 100nF 15 1M 17 5 100nF TP2 PRESSURE PORT VR2 10k 8 +5V Q3 BC337 B TP10 VR6 10k PRESSURE OFFSET CON2 3 MPX2010 PRESSURE SENSOR 12 TP9 9 4 7 2 1 INSTALL ONLY FOR TESTING JP1 AN1 RA1 IC1 PIC16F1507 –I/P AN6 TP8 RB6 2 6 RA5 AN10 RC4 D Q1 IRF540N 10 G S 0.1  5W B 22k 10 16 RB7 K A D2 C D4 100 F K 14 220nF 11 62k 10k 20 A 13 20 6 TP11 100k 3 1 IC3a K 470k 22pF * CHANGES REQUIRED FOR O 2 ¯ IN AIR MEASUREMENTS (SEE TEXT IN PT.2) WIDEBAND OXYGEN SENSOR CONTROLLER 2 470k  LED1 A SC 510 * Vs A Vss A ZD2 15V 1W K D2-D4: 1N4148 2012 100nF A 470 AN4 10 F TPV– D3 K A E Q2 BC327 AN9 AN7 100nF C E 100 F 18 10k 100 F AN8 THRESHOLD VOLTAGE 1 2 3 4 Vs/Ip 11 10k PWM3 VACUUM PORT 10 F 150 7 IC2b 100nF 1k 3 2 AN3 4 5 6 +5V H– 100nF VR3 10k 10 F 20k GND2 100nF 120 ADJ K ZD1 16V 1W H+ +5V OUT IN K 1N4004 A K ZD1, ZD2 A K Fig.12: the full circuit uses microcontroller IC1, several CMOS op amps (IC2-IC4), a Mosfet (Q1) to control the heater in the oxygen sensor and a pressure sensor. The microcontroller and op amps monitor & control the wideband oxygen sensor and drive the Wideband Display Unit. IC1 also provides a simulated narrowband output (via IC2c). and comprise an LMC6484AIN quad op amp (IC2) and two LMC6284AIN dual op amps (IC3 & IC4). These have a typical input offset of 110µV, a high input impedance of more than 10 Teraohms (>10TΩ), a 4pA input bias current, an output that can swing to within 10mV of the supply rails with a 100kΩ load, and a wide common mode input voltage range that includes the supply rails. 40  Silicon Chip Power for the circuit comes from an external 12V supply, ie, the car battery. The +12V rail is fed in via fuse F1 and applied directly to the heater circuit (via H+ at CON1). It’s also fed in via reverse polarity protection diode D1 and applied to an LM317T adjustable regulator (REG1) and to 12V regulator REG2 (LM2940CT-12). Fuse F1 will blow if the sensor is connected when the supply polarity is reversed. That’s because, in this situation, there’s a low-resistance current path through the heater element and the body diode in Q1. Trimpot VR1 allows REG1’s output to be set to exactly 5.00V. This rail supplies microcontroller IC1 and op amps IC2 and IC3. The +12V rail from REG2 supplies IC4. The battery voltage is measured at the AN3 input of IC1 via a 20kΩ and siliconchip.com.au REG2 LM2940CT-12 IN +12V OUT GND 3 1 IC2a 2 12 13 RING TIP SLEEVE 10 WIDEBAND DISPLAY OUTPUT CON3 RING TIP 150 8 IC2c 9 H– 150 14 IC2d H+ 10 F (NOT USED) SLEEVE CON4 SIMULATED NARROWBAND OUTPUT TO OXYGEN SENSOR CON5 3 Vs/Ip 2 1 +12V 4 8 Rcal 5 6 Vs 7 Ip 8 5 VR5 1k 7 IC4b 6 Rcal 4 62 TP3 Ip 22k +5V TP5 100nF 7 IC3b 4 TP7 560k* 5 8 22k 6 IC3, IC4: LMC6482AIN 3.3nF +5V VR4 10k 3 100k 1 Vs/Ip B E IC4a G C D D S LM317T LM2940CT-12 IRF540N 10kΩ voltage divider connected between the +12V input rail and 0V. This divider reduces the applied voltage by two thirds and results in a maximum of +5V at the AN3 input for a battery voltage of 15V (note: 5V is the upper limit for analog-to-digital conversion by IC1 for a maximum 10-bit digital value of 1023). Trimpot VR3 across the 5V rail provides the 3.3V reference voltage siliconchip.com.au 2 TP4 BC327, BC337 LED Additional supply rails TP6 560k K A TP1 GND IN GND OUT This is necessary because zero pump current is required during the sensor heat-up period. It’s also necessary when there is a fault in the sensor’s heater element or the connection to it. IC4b is driven from the PWM1 output of IC1 via a 10kΩ resistor and 100nF capacitor. These RC components filter the PWM output to produce a steady DC voltage. The PWM signal is output at 15.625kHz and its duty cycle can be varied from 0-100% to produce an effective DC voltage ranging from 0-5V. IC1’s PWM4 and PWM3 ports (pins 15 & 17) provide the wideband and narrowband signal outputs respectively, again using PWM control. As shown, the PWM4 output is filtered via a 10kΩ resistor and 100nF capacitor and buffered with IC2d. The wideband display output is then fed to CON3 via a 150Ω resistor. By contrast, the PWM3 output is filtered using a 1MΩ resistor and 100nF capacitor to give a slower, smoothed response that’s similar to the response from a standard narrowband sensor. This signal is buffered using IC2c and fed to CON4 via a 150Ω isolating resistor. OUT ADJ OUT IN referred to earlier and this is buffered by op amp IC2b. This op amp drives one side of the pump cell, at the Vs/Ip connection, via a 150Ω resistor which isolates the op amp output. In addition, the Vs/Ip voltage is measured at the AN0 input of the microcontroller to ensure that the pump current can be set to zero by applying the same voltage (from the PWM1 output) to pump drive buffer stage IC4b. While IC2 & IC3 are provided with a 5V supply, IC4 is a special case because IC4b’s output is required to swing from 0-5V to drive the pump cell with current. To ensure this, IC4’s positive supply rail needs to be more than +5V and its negative rail needs to be less than 0V. As a result, REG2 is included to provide a nominal 12V supply. This supply is nominally 12V because the regulator cannot deliver 12V unless the input is just over 12V. If the input voltage to REG2 is less than 12V, its output falls accordingly. This isn’t important since we only want more than 5V for IC4 and REG2 is basically used to limit the positive supply to +12V. Transistors Q2 & Q3, diodes D2-D4 and their associated capacitors are used to derive the negative supply rail for IC4. This circuit is driven by the RA1 output of IC1 which generates a 3.906kHz square-wave signal. Q2 & Q3 buffer this signal and drive a diode pump consisting of D2 & D3 and two 100µF capacitors. This produces a negative supply rail of -2.5V. Diode D4 clamps this rail to June 2012  41 Parts List 1 PCB, code 05106121, 149 x 76mm (availble from SILICON CHIP) 1 ABS box, 155 x 90 x 28mm (Altronics H0377) 1 MPX2010DP 10kPa temperature compensated pressure sensor (Sensor1; optional) (Jaycar ZD1094) 2 M205 PCB-mount fuse clips 1 5A M205 fuse (F1) 1 DIL20 IC socket 1 DIL14 IC socket 2 DIL8 IC sockets 2 PCB-mount 3.5mm stereo switched jack sockets 1 4-way SIL socket strip (can be cut from a DIP8 IC socket) 1 2-way PCB-mount screw terminals (5.04mm spacing) 1 3-way PCB-mount screw terminals (5.04mm spacing) 4 M3 x 5mm machine screws 4 M3 x 10mm machine screws 2 M3 x 15mm machine screws 5 M3 nuts 1 3-6.5mm IP65 cable gland 20 PC stakes 1 2-way pin header, 2.54mm pitch 1 jumper plug for pin header 1 100mm cable tie 1 70mm length of yellow medium duty (2A) hookup wire 1 70mm length of red medium duty (2A) hookup wire 1 70mm length of black medium duty (2A) hookup wire 1 120mm length of green medium duty (2A) hookup wire 1 150mm length of light blue heavy duty (7.5A) hookup wire 1 4m length of green heavy duty (7.5A) hookup wire +0.6V when the negative supply generator is not working, ie, when IC1 is not in circuit or if there is a fault in the negative supply generator. Zener diode ZD2 limits the total supply that can be applied to IC4 to 15V. Op amp IC3b is connected as a differential amplifier to monitor the voltage across the paralleled 62Ω and Rcal resistors. It operates with gain of 25.45 as set by the 560kΩ and 22kΩ feedback resistors. The 3.3nF feedback capacitor rolls off high frequencies and prevents amplifier instability. IC3b’s output is referenced to the Vs/Ip voltage (at +3.3V) by the 560kΩ 42  Silicon Chip 1 2.5m length of red heavy duty (7.5A) hookup wire 1 140mm length of 3mm heatshrink tubing (or 20mm yellow, 40mm red, 40mm black, 40mm green) Semiconductors 1 PIC16F1507-I/P microcontroller programmed with 0510612A.hex (IC1) 1 LMC6484AIN quad op amp (IC2) 2 LMC6482AIN dual op amps (IC3,IC4) 1 LM317T adjustable regulator (REG1) 1 LM2940CT-12 12V low-dropout regulator (REG2) 1 IRF540N 100V 33A N-channel Mosfet (Q1) 1 BC327 PNP transistor (Q2) 1 BC337 NPN transistor (Q3) 1 3mm red LED (LED1) 1 16V 1W zener diode (ZD1) 1 15V 1W zener diode (ZD2) 1 1N4004 1A diode (D1) 3 1N4148 switching diodes (D2-D4) Capacitors 4 100µF 16V PC electrolytic 4 10µF 16V PC electrolytic 1 220nF MKT polyester 8 100nF MKT polyester 1 3.3nF MKT polyester 1 22pF ceramic Resistors (0.25W, 1%) 1 1MΩ 1 1kΩ 2 560kΩ 1 510Ω 2 470kΩ 1 470Ω 2 100kΩ 3 150Ω 1 62kΩ 1 120Ω 3 22kΩ 1 62Ω 1 20kΩ 2 10Ω 4 10kΩ 1 0.1Ω 5W Trimpots 1 500Ω multi-turn trimpot (3296W type) (Code 501) (VR1) 4 10kΩ multi-turn trimpot (3296W type) (Code 103) (VR2VR4,VR6) 1 1kΩ multi-turn trimpot (3296W type) (Code 102) (VR5) Sensor Parts 1 Bosch LSU 4.9 Broadband Oxygen sensor (Available from TechEdge http://wbo2.com/lsu/sensors.htm part #17123, Bosch. Part # 0 258 017 123) 1 Bosch connector for LSU 4.9 sensor (Available from TechEdge http:// wbo2.com/cable/connkit.htm part #017025) 1 6-way sheathed and shielded lead with 2x7.5A wires for heater (Available from TechEdge http://wbo2.com/cable/default.htm part #DIY26CBL; includes 1 x 8-pin circular multi-pole line socket part #P8PIN) 1 8-pin circular multipole panel microphone plug connector (Available from TechEdge http://wbo2.com/cable/connkit.htm part #S8PIN) resistor between its pin 5 input and the output of op amp IC2b. As a result, when there is 0V across the 62Ω resistor, IC3b’s output sits at 3.3V. Sensor cell voltage Op amp IC3a monitors the sensor cell voltage (Vs). As already noted, IC3a is set so that when Vs is at 450mV, its output is 2.5V. To do this, trimpot VR4 provides an offset voltage which is buffered using op amp IC4a. A 2.5V setting means that IC3a can swing symmetrically above and below this level to drive IC1’s AN10 input (pin 13). This voltage swing allows an expanded measurement of any variation above or below 450mV from the sensor cell. The reference current applied to the sensor cell is derived via two series resistors (62kΩ and 510Ω) between the +5V supply rail and the Vs terminal of the sensor cell (in the oxygen sensor). When the controller is running and measuring correctly, the Vs terminal is at the Vs/Ip voltage (3.3V) plus the 450mV of the sensor cell. The 62kΩ and 510Ω series resistors deliver the recommended 20µA current to the cell. That current is calculated as (5V - 3.3V - 450mV) ÷ siliconchip.com.au (62kΩ + 510Ω) or 19.99µA. The actual current does not affect the accuracy of lambda measurement unless the current is reduced down to near zero or is increased above 40µA. Link setting When installed, jumper JP1 ties IC1’s RA5 (pin 2) input low. This selects a test mode for checking that the sensor impedance is correct (ie, 300Ω). In this mode, instead of the wideband output from IC2d providing 0-5V for lambda measurement, it outputs a value that corresponds to the impedance of the sensor cell. Since this impedance depends on the temperature of the sensor, it’s useful for ensuring that part of the control circuit is working and that the sensor is not being overheated by exhaust gas when installed in a vehicle. Trimpot VR2 sets the threshold voltage for “engine-started” detection. This is so that the engine can blow out any condensation in the sensor before any electrical heating of the sensor begins. As stated previously, engine-started detection is achieved by monitoring the battery voltage. Typically, a 12V lead-acid battery is below 12.9V when the engine is off but rises above 12.9V when the engine starts and the alternator begins charging. In operation, the battery voltage is compared with the threshold voltage on TP2 (AN8 of IC1), as set by VR2. This threshold voltage can be set anywhere from 0-5V, corresponding to a battery voltage range of 0-15V. Basically, the TP2 voltage is set to 1/3rd the required engine-started battery voltage. For example, if this voltage is selected to be 13V, TP2 is set at 4.33V. When the wideband controller is used as a portable air/fuel ratio measuring instrument, TP2 can be adjusted to 4V or less. This will ensure that the sensor is heated when power is first applied. However, it also means that the sensor MUST be protected from moisture ingress and from physical shock when not in use. Heater current Mosfet Q1 drives the sensor’s heater with a DC voltage derived from a 122Hz PWM signal delivered from IC1’s RB7 output (pin 10). The heater current (and the Mosfet’s source current) is monitored via the AN4 input siliconchip.com.au at pin 16. That’s done by monitoring the voltage across the 0.1Ω 5W resistor. LED1 is the status LED. It’s driven from the RC4 output of IC1 via a 470Ω current-limiting resistor. As stated previously, it turns on when the sensor is heating and then flashes rapidly once the operating temperature is reached. It flashes more slowly if there is a sensor error or if the air/fuel ratio is outside its measurement range. Pin 4 of IC1 is the MCLR reset input. It’s pulled high via a 1kΩ resistor and ensures that IC1 is reset on power up. Two grounds Note that the circuit uses two grounds. One (GND2) is for the heater, while the other (GND1) is for the rest of the circuit. These two grounds are connected to the car chassis via separate wires. Without this separate earthing, the switching current applied to the heater would cause inaccuracies in the measurements of voltage and current and for the wideband output. Pressure sensing The pressure sensing circuit comprises the pressure sensor (Sensor1) itself and offset trimpot VR6. The specified sensor has differential pressure inputs and differential outputs. These outputs are connected to AN7 & AN9 (pins 7 & 9) of IC1. With a 5V supply, each output sits at 2.5V when there is equal pressure on each input port. Unequal pressures result in a differential output of 1.25mV per kPa, although the resolution of the pressure sensor readings with a 10-bit A/D converter is about 3.9kPa (or 39hPa). This resolution is sufficient to allow Ip to be corrected to within 1%. The highest pressure that we compensate for is 587hPa (58.7kPa) above atmospheric, which gives a differential sensor output of 73.38mV. The resulting correction, as determined by the microcontroller, reduces Ip by 12% for lean values and by 9% for rich values. These corrections are in accordance with the graph shown in Fig.8. The pressure sensor is set up by plugging (blocking) one of its differential air inlets to allow the sensor to work as an absolute pressure (often called “Gauge pressure”) sensor rather than as a differential sensor. This is best done when the sensor is located at sea level, where the standard air- This is the Bosch LSU 4.9 wideband sensor that’s used in conjunction with the Wideband Controller. pressure of 1013hPa is available. That way, the sensor will respond to variations in pressure above and below standard atmospheric pressure, giving a positive output for pressures above atmospheric and a negative output for pressures below atmospheric. If one input is plugged at higher altitudes, the sensor output will be referenced against the lower pressure in the plugged inlet and the actual output will be a positive value when measuring standard atmospheric pressure instead of 0. In other words, the pressure sensor output will be offset according to the amount that the plugged input is below atmospheric pressure. As a result, offset trimpot VR6 has been included to counter this effect. Basically, it allows the lower pressure reading to be offset, not at the sensor itself but in the way the sensor’s output is mathematically manipulated by the software. In practice, VR6 is set to give a 1V output per 1000m above sea level. For a sea level setting, its output (TP10) is set at 0V. At higher voltage settings, IC1 provides compensation for the approximate 11kPa drop in pressure per 1000m in elevation above sea level. Note, however, that this only applies for elevations up to 2000m above sea level, at which point the change in pressure becomes non-linear. As a result, we do not correct for pressure offset above 2000m. If the pressure sensor is not required, then the AN7 and AN9 inputs must be tied to 0V and 5V respectively. That will stop the AN7 and AN9 inputs from floating and will also indicate to IC1 that the sensor is not connected. We show how these AN7 and AN9 inputs are tied to the supply rails in the construction details to be published next month. We’ll also publish the SC details for the display readout. June 2012  43