WebSeminar: December, 2005 Hello, and welcome to the Microchip Technology Web Seminar overview of signal chains. My name is Kevin Tretter and I am a Product Marketing Engineer within Microchip Technology s Analog division. December 2005 Page 1
Agenda Sensors Thermocouples Load Cells Signal Conditioning Amplifier Requirements Gain, A OL, Noise, etc. In order to understand a signal chain and the resulting requirements, one must have an understanding of the sensor being used. Today we will look at two common sensor types: thermocouples and load cells. We will then review the signal conditioning required for these sensor outputs and how this affects component selection. December 2005 Page 2
Thermocouple Overview What is a thermocouple? Two dissimilar metals joined together on one end Temperature difference between the bead and the other end of the thermocouple will cause a voltage (e o ) Metal 1 + e o Metal 2 - Sensors are used to measure a variety of different conditions, including pressure, humidity, strain, etc. The most commonly measured condition is temperature. Thermocouples are very common sensors used to measure temperature. It consists of two different metals soldered together on one end. A temperature difference between the two ends of the thermocouple will cause a voltage. This voltage will change as a function of temperature. December 2005 Page 3
Thermocouple Advantages/Disadvantages Advantages Wide Temperature Range Very Robust Fast Response Time Cheap Disadvantages Very Non-Linear Very Small Output Signals Needs a Temperature Reference There is a wide variety of applications in which temperature measurement is required. Thermocouples are very popular due to their wide operating temperature range which makes them well suited for extreme conditions such as cryogenics (liquid oxygen or hydrogen used as rocket fuels and industrial coolants) and other industrial applications. Other advantages of thermocouples include fast response time, durability, and they are relatively cheap. The disadvantages include a non-linear transfer function and small output voltages. Due to this non-linearity, thermocouples must be compensated, typically in software, to correct for this non-linear transfer function over temperature. This transfer function and output voltage will vary depending on the types of metals used to form the thermocouple. It is important to note that thermocouples measure the relative temperature, not the absolute temperature. Therefore a temperature reference is necessary. This reference is commonly known as the cold junction temperature reference since historically the reference was kept in an ice bath to ensure that the temperature remained constant at 0 degrees Celsius. December 2005 Page 4
Load Cell Overview What is a load cell? Transducer that converts a force into an electrical signal Used to measure anything from pills to trucks Strain gauge based load cells are the most common Gauge will deform as force is applied Measure the resistive change due to this deformation We will now move on to load cells. A load cell converts a force into an electrical signal, typically through the use of a strain gauge or gauges. Load cells are found in all types of electronic scales, from those used in the supermarket to weigh fruit to those used to weigh large trucks. A strain gauge is nothing more than a piece of wire. However, when this wire is placed under strain, the resistance will change. (See the picture on the next slide.) December 2005 Page 5
Strain Gauge Metal-foil Strain Gauge Omega Engineering Inc. Here is a picture of a metal-foil strain gauge. Note that it is one continuous piece of wire bent into parallel lines. This topology is used to maximize the total change in resistance as the strain gauge is placed under pressure. December 2005 Page 6
Example Lever Arm F Strain Gauges Strain Gauges So how are these strain gauges used within a load cell to measure force? Let s look at an example. Typically, four strain gauges are used to measure the force on a lever arm. In the above example, a force (F) is placed onto the end of the lever arm. December 2005 Page 7
Example Lever Arm (Cont.) Strain Gauge (Tension ) F Strain Gauge (Compression ) Strain Gauge (Compression ) Strain Gauge (Tension ) This force causes the lever arm to bend. Although the above illustration is an exaggeration of this bending, it is a true representation of how the lever arm will indeed bend due to the force placed on it. Due to this bending, two of the strain gauges attached to the lever arm will be squeezed together, or compressed. The other two strain gauges will be stretched, or held in tension. December 2005 Page 8
Wheatstone Bridge Circuit - Excitation Voltage + Used to measure resistance, inductance, or capacitance A voltage is applied across the bridge and the voltage across the middle of the bridge is measured Four variable elements is called a full bridge Okay, let s leave our lever arm example for a moment and look at how these strain gauges are used in terms of a circuit. A Wheatstone bridge circuit, shown on the left, consists of four resistor elements (in this case our four strain gauges). A voltage is applied across the bridge and the voltage potential across the center of the bridge is measured. In this example, since we have four strain gauges, this would be considered a full bridge load cell. December 2005 Page 9
Circuit for Strain Measurement Excitation Voltage (Tension) (Compression) R + ΔR R - ΔR - + R - ΔR R + ΔR (Compression) (Tension) Tension will INCREASE resistance Compression will DECREASE resistance The configuration shown will maximize the voltage in the center of the bridge Okay, so let s go back to the lever arm with the four strain gauges. As you may recall, two of the strain gauges were in tension (stretched), and two of the gauges were in compression (squeezed). A strain gauge in tension will increase the absolute resistance of the gauge (relative to the gauge s nominal resistance value). A strain gauge in compression will decrease the absolute resistance of the gauge (again relative to the gage s nominal resistance value). By arranging the two strain gauges in tension opposite of each other in the bridge (and consequently the two gauges in compression are opposite), the voltage across the center of the bridge is maximized. December 2005 Page 10
Circuit Characteristics Excitation Voltage A higher voltage will result in a greater voltage across the center of the bridge More power dissipation causes more self heating (error in measurement) Most applications use +5V or +10V as the excitation voltage A Wheatstone bridge circuit has several characteristics that need to be considered. The excitation voltage is the voltage applied across the bridge. A higher voltage will result in a greater voltage across the center of the bridge. However, a higher excitation voltage also results in more current flowing through the resistors. This means more power being dissipated by the strain gages in the form of heat. This heat will cause the gages to heat up. What happens when metal heats up? It expands! This results in a change in resistance and an error in your measurement. December 2005 Page 11
Circuit Characteristics (Cont.) Strain Gauge Sensitivity Smaller nominal resistance values will produce more signal swing Increased sensitivity to change in strain More current flow will produce more heat (causing measurement errors) Values of 350Ω and 1000Ω are common Load cells have a rated sensitivity, usually 1mV/V or 2mV/V Example: 5V excitation voltage with a 2mV/V load cell will result in a 10mV full-scale signal out of the bridge Another variable is the nominal resistance of the strain gauges. A smaller resistance value will cause more current to flow and increases the change in voltage as a function of strain. Keep in mind that Power = Current 2 * Resistance. So more current means significantly more power and self-heating. Most of the time you don t have to worry about this level of detail, as a given load cell will have a stated sensitivity in mv/v. December 2005 Page 12
Bridge Compensation Excitation Voltage V out An additional resistor can be used to compensate for temperature The resistor can be referenced to the positive supply or to ground (as shown) V comp As mentioned earlier, when metal heats up, it expands. The strain gauges are mounted to the lever arm, so if the actual lever arm was to heat up, it would expand and cause the strain gauges to be stretched (tension). This will cause an error in the measurement. To compensate for this, some designs add a resistor. This compensation resistor is located on the lever arm and measures the self heating of the metal on which the gauges are attached. The voltage across this resistor is measured and the resulting data can be used to cancel out the error. This may require additional signal conditioning circuitry and another channel of analog-to-digital conversion. December 2005 Page 13
Signal Conditioning Now that we have explored two common sensor types, we will move on to the signal conditioning circuitry that is required for these sensors. December 2005 Page 14
Requirements Small signals need to be amplified Filtering of higher frequency noise/mechanical resonance Cannot add significant noise to the measured signal The signals generated by these various sensors are very small, usually on the order of mv or even uv and therefore need to be amplified. There may be specific noise sources that need to be filtered out. Some examples include power line hum (50 Hz or 60 Hz), mechanical scale resonance, etc. Finally, any signal conditioning circuitry cannot add significant noise to the system or else the sensor output signal will be lost. December 2005 Page 15
Gain Stage Gain 5 mv 10 mv 50 mv 100 mv 500x (54dB) 250x (48dB) 50x (34dB) 25x (28dB) Full-scale Input = 2.5V This slide shows a very simplified block diagram of a gain stage. Depending on the output swing of the sensor and the full-scale input to the A/D converter, the gain stage may need to amplify a signal by as much as 500x or more. Please note that you would typically want to leave some headroom in a real world application so that the signal does not saturate (either in the amplifier or in the A/D converter). Laws govern the amount of headroom required for weigh scales that are used for commercial purposes. December 2005 Page 16
Gain Stage (Cont.) Amplifier must have enough open loop gain to maintain linearity Must be low noise to not dilute the measured signal For thermocouple applications, low bias current is required Several considerations must be made when selecting an amplifier or amplifiers to use in a gain stage. The amplifier must have enough open loop gain to provide signal gain while still maintaining performance. As noted earlier, low noise becomes critical when working with small signal sensors. Bias current may also be a factor in some circuit designs, as we will see a little later on. December 2005 Page 17
Open Loop Gain (A OL ) OL ) 125 db 110 db 100 db 75 db 54 db 50 db 25 db Open Loop Gain Available to Ensure Linearity Open Loop Gain Used For Amplification 10 100 1k 10k 100k 1M 10M 100M This slide shows a typical open loop gain plot for an amplifier. The gain remains flat over a certain frequency range and then begins to roll off at 20dB per decade. In this example, the amplifier has 110dB of open loop gain. Let s also assume that we are using this amplifier in a gain of 500x (or 54dB). Such gains are not unreasonable, considering the low amplitude of the signals coming out of most sensors. December 2005 Page 18
Open Loop Gain (A OL ) (Cont.) In the example, the amplifier has an open loop gain of 110dB A gain of 500x is 54dB In this case, the amplifier has only 56dB (110dB 54dB) of linearity Linearity = 56dB = 1 part in 631 < 10 bits In this example, 54dB of the open loop gain is used to provide the required gain. The remaining 56dB is what s left to ensure linearity. 56dB is equivalent to 1 part in 631, which is less than 10 bits accuracy. Although this amplifier can handle a gain of 500x, it will probably limit the overall system performance in terms of linearity due to a lack of open loop gain. There are several possible solutions for this: 1) Get an amplifier with more open loop gain. These are available, but for a premium price. 2) Use multiple gain stages to achieve the overall desired gain. For example, a gain of 25x followed by a gain of 20x will achieve an overall gain of 500x. A designer must be careful though not to introduce too much noise, offset errors, etc. when implementing a multiple stage design. December 2005 Page 19
Amplifier Noise Keep in mind that the voltage noise of an amplifier is INPUT REFERED Any gain in the amplifier circuit will also gain up the amplifier noise Therefore, the noise of the amplifier must be compared relative to the input signal, not the output signal It is important to note that an amplifiers voltage noise is input referred, so any gain or attenuation in the circuit will also affect the amplifier noise. Also, keep in mind that typically amplifier noise is not flat across the entire frequency range. An amplifier may have really low noise at 1 khz where it is specified in the datasheet, but may have terrible 1/f noise that results in a much higher noise floor at lower frequencies. This becomes critical in low frequency applications. December 2005 Page 20
Amplifier Noise (Cont.) N amp N adc N tot = (N 2 amp + N 2 adc ) Noise sources add as the square root of the sum of the squares When considering overall system noise, it is imperative to understand how the various noise sources within the circuit interact. Contrary to popular belief, noise sources do not add in a linear fashion! The above example illustrates that noncorrelated noise sources add as the root of the sum of the squares. December 2005 Page 21
Amplifier Noise (Cont.) Amplifier noise can affect overall system performance Increase in System Noise (%) 50% 40% 30% 20% 10% 0% Noise Analysis 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Namp/Nadc Using the previously stated relationship, this graph shows how the noise from an input amplifier can affect the overall system noise level. Note that the x-axis is the ratio of the amplifier noise to the converter noise, so a value of 1 indicates the point at which the amplifier and the converter are equally noisy. At this point, the overall system noise increases by 44% due to the amplifier. If the amplifier is half as noisy as the converter (the point of 0.5 on the graph), the system noise increases by about 12% due to the amplifier. A good rule of thumb is to design the input stage such that the noise budget is at least 20 db (a factor of 10) below the noise of the converter to prevent the input stage from greatly affecting the overall system performance. As A/D converters increase in performance, this task becomes more and more difficult. December 2005 Page 22
Input Bias Current Amplifiers have a specification called input bias current (or I bias ) This is the amount of current flow into the inputs of the amplifier to bias the input transistors Can range from ua down to pa CMOS (JFET) amplifiers generally have less bias current then bipolar (BJT) amplifiers Another important specification for an amplifier used for signal conditioning is input bias current. As shown on the slide, the input bias current is the current that flows (or leaks) into the input pins of the amplifier. The amount of bias current is dependent on the amplifier design, with CMOS amplifiers generally having less bias current than a bipolar amplifier. The next slide will examine one situation in which this characteristic is critical. December 2005 Page 23
Input Bias Current (Cont.) I bias is a very important specification when using thermocouples or other high impedance sensors As the bias current flows through the resistors, a voltage error is created This illustration shows a thermocouple with a simple protection/filtering circuit. Thermocouples can be located in very harsh environments, and are prone to picking up large noise spikes (AC motor starters, shorting out in an electric furnace, etc.). This noise can potentially damage the circuitry connected to the thermocouple. In the above circuit, the resistors and capacitor form a lowpass filter. The resistors also offer some protection to the rest of the circuitry from these large transients. Large resistors are desirable to increase ESD protection and filtering of noise. A larger resistor means that you can maintain the same filter corner with a smaller capacitor, which is always good for size/cost requirements. For the above reasons, a large series resistor is desirable. However, this is where bias current comes into play. As the bias current flows through the series resistor, a voltage drop occurs which causes an error in the measurement. Recall that Voltage = Current * Resistance, so the larger the product is of the bias current and resistor, the larger the error voltage. In this case, less bias current is desirable, as it allows the designer to use a higher impedance sensor and still maintain an acceptable error voltage level. December 2005 Page 24
Summary The circuitry required for a given signal path always starts with the sensor Sensors produce small signals that need to be amplified Signal chain circuitry can also provide filtering and isolation Every data acquisition system is different In summary, all signal chain analysis begins with the sensor. In order to design an effective signal chain circuit, the designer must be aware of the sensor characteristics and requirements. In most cases, the sensor will produce a small output voltage swing that will need to be amplified prior to being converted to a digital signal. The signal chain circuitry can also provide filtering of the input signal as well as isolation for the rest of the circuitry including the A/D converter. December 2005 Page 25
References AN684 Single Supply Temperature Sensing with Thermocouples AN695 Interfacing Pressure Sensors to Microchip s Analog Peripherals AN717 Building a Bridge Sensing Circuit FilterLab Active Filter Design Software Free Downloadable from the Microchip Web Site Lowpass, Highpass, Bandpass Filter Design Microchip Technology has a variety of other application notes relating to data acquisition system design, as well as general application notes on operational amplifier topologies and AC/DC specifications. Finally, the FilterLab software is available from the Microchip Technology web site. This free software will help you design active operational amplifier filter circuitry. December 2005 Page 26
WebSeminar: December, 2005 Thank you for your time as we reviewed several common sensor types and signal conditioning circuitry. Please check back often for more Microchip Technology Web Seminars covering a wide variety of topics. December 2005 Page 27