 Hello and welcome to this presentation on Op-Amp Precision, Op-Amp's VIO input offset voltage. Today's sensors are everywhere. In our modern world we need to constantly measure everything, weight, UV, temperature, speed, current, even though we live in an increasingly digital world. Many are still purely analog and generally most of them provide an extremely small signal. The question is how to deal with this slow signal in order to transfer it to the digital domain without adding any error that could compromise the information. The solution is to use an Op-Amp to amplify the signal, but this Op-Amp needs to be ideal not to introduce any offset. Unfortunately this kind of device does not exist. Op-Amp's will always impact measurements, however by choosing the right Op-Amp can limit the error on the output as much as possible. In the example shown in this slide, let's consider a battery delivering a maximum current of 140 mA through a shunt of 10 mA. The resulting voltage will be 1.4 mV maximum as the LSB of the ADC is 880 microvolts. We clearly need to amplify the signal to increase the accuracy. The main goal of this video is to identify which Op-Amp parameters are the most relevant with regard to precision. The main limitations to precision in analog integrated circuits are noise and mismatch. Before we start, let's define a few terms. What is offset? When a zero differential input voltage is applied, the Op-Amp output voltage should be zero in practice, which is not the case. The offset voltage by definition is the differential input voltage that is required to make the output voltage zero. The input offset voltage parameter, generally called VIO, is defined as a DC voltage between the non-inverting and the inverting input. It is always specified in the electrical characteristics of a datasheet and can be positive or negative. The VIO value will be different for each Op-Amp and to ensure good precision, it is important to choose the Op-Amp with the lowest VIO because the VIO will be amplified by the gain and added to the total output as an error. Some Op-Amp's with a chopper architecture like TSZ-121 exhibit a VIO of 5 microvolts. They are especially good for precision DC measurements. The cause of the input offset voltage is well known. It is due to the inherent mismatch of the input transistors and components during the fabrication of the silicon die and the stress placed on the die during the packaging process a minor contribution. These effects collectively produce a mismatch of the bias of the input circuit resulting in a differential voltage at the input terminals of the Op-Amp. For CMOS technologies, this equation shows the root cause of the input offset voltage. We can see that the size of the MOS width and length and the doping substrate play a key role in the VIO error. The resulting expression contains four terms since all of them can be positive or negative, they never all add up. Thankfully, this never occurs in practice. They never cancel each other out either. In any case, regardless of the design or process effort, there always remains a small input offset voltage. This also explains why for a dual amplifier where two Op-Amp's are in the same package, the input offset voltage of both channels is different. Let's now look at how the input offset voltage can impact the theoretical measurement of an application. Shunt current sensors are used in precision current sources for feedback control systems. They are also used in a variety of other applications including battery fuel gauging and torque feedback controls in electric power steering and precision power metering. In this slide, we will look at a motor control application. Thanks to the H-bridge, the motor can be driven in both directions. The current measurement helps to give the motor speed and its rotation direction. In such applications, it is desirable to use a shunt with a very low resistance to minimize the series voltage drop. This wastes less power and allows the measurement of high currents without a significant voltage drop. A typical shunt might be 0.1 ohm. The lower the current, the lower the resulting voltage through the shunt resistor. So at high current, the VIO of the Op-Amp has little impact. For low current, it may become critical. When the current is only few amps, the shunt's output signal is only a few hundred of millivolts. For example, at 1 milliamp, the voltage through the shunt is 100 microvolts to which the VIO has to be added. So the Op-Amp demands a very low offset voltage and drift to maintain absolute accuracy. The Op-Amp is used as a differential Op-Amp in order to amplify the voltage drop appearing through the shunt resistor. In this example, the signal is amplified by 1000. With an ideal Op-Amp VIO equals 0 millivolts. So only the first term of the equation is valid. Normally we can read an output voltage of 100 millivolts. But as we said previously, the world is not perfect and the ideal Op-Amp doesn't exist. If we use a TS507 Op-Amp with a maximum positive VIO of 100 microvolts at ambient temperature, we can see that the output voltage is minus 100 microvolts. This is far from the expected value 100 millivolts, meaning that the speed information is completely incorrect. Moreover, this negative output will indicate to the MCU that the motor is rotating in the reverse way, which is totally incorrect. Here we can see how the VIO of the Op-Amp can have an impact on the entire application if the Op-Amp is not chosen carefully. If we use a precision amplifier with a Charper architecture like the ST's TS-Z121 with a VIO of 5 microvolts at ambient temperature, we can see that we will make an error of 5% on the motor speed, which is really acceptable when we use a gain of 1000. If the motor rotation direction is changed, the theoretical output value considering an ideal Op-Amp will be minus 100 millivolts. We can see by using the TS-Z121 that even in a reverse rotation the measurement remains valid, 5% error on the output. An Op-Amp's VIO can be either positive or negative. This table summarizes the real impact of the VIO with two different amplifiers, TS-507 and TS-Z121, with the TS-507 and a positive VIO of plus 100 microvolts. The result on the output is completely incorrect, even in regards to the motor's speed and rotation direction. If the VIO is negative, the rotation direction of the motor is fine, but we make an error of 100% on the motor speed. Unfortunately, we cannot predict the polarity of the VIO from one part to another. So when we make a design, it is important to take into consideration the fact that the VIO might be positive or negative. However, by using the TS-Z121 Op-Amp, which is a precision amplifier with an extremely low VIO, the impact on the output result will be limited. The fact that the VIO can be positive or negative will not impact the expected results. Another point we need to understand is the real impact of the VIO on an application in terms of effective cost. The VIO also reduces the dynamic range of an ADC. The loss of the dynamic range affects the resolution of ADC circuits because the maximum dynamic range is required for maximum resolution. This table shows the equivalent resolution of the ADC for various input offset voltage ranges. Usually, an Op-Amp can be chosen with a VIO low enough to meet the desired accuracy. It is easy to find an Op-Amp that meets the VIO specification for an 8-bit or 10-bit converter, but it becomes increasingly difficult as the resolution increases. We can see that if the Op-Amp is not chosen carefully, all the money spent on a 12-bit ADC is lost. Indeed, in this example, if the TS-512 Op-Amp is chosen to amplify the signal, only 7 bits of the 12-bit ADC are effective. Whereas by choosing the TS-Z121, which has an extremely low VIO of 5 microvolts, no LSB will be lost when using a 12-bit ADC and the entire resolution of the ADC can be used. Therefore, it is not necessary to spend money on a precision ADC if the Op-Amp is not correctly chosen. As the VIO is principally due to the mismatching of the input components, this parameter will also vary depending on the ambient temperature, so it is important to take into consideration in a precision environment the offset drift in temperature, generally called DVIO DT. The VIO is always multiplied by the non-inverting gain of the Op-Amp and added to the signal amplified by the circuit, which is the minus 100 in this example, the transfer function is determined by equation 1, where a VIO is the maximum value written in the data sheet, adding the effects of the temperature to equation 1 gives equation 2, this allows a fairly accurate calculation, the worst case change in output due to VIO neglecting the effect of the resistors. However, the resistor values also change with temperature and will also affect the gain. Typical drift values for general-purpose precision Op-Amp's lie in the range of 1 to 10 microvolts per degree Celsius. In this schematic, then equals zero as the inputs are grounded, so the output of the Op-Amp reveals VIO multiply minus 101. This slide compares two amplifiers, the TSV611, a standard CMOS Op-Amp, with a maximum VIO of 4 millivolts, and the TSZ-121, a chopper Op-Amp with a maximum VIO of 5 microvolts. At 25 degrees Celsius, we can see that the TSV611 shows an output offset of minus 125 millivolts, whereas TSZ-121 is close to zero microvolts. This shows that if the TSV611 is used in an application requiring accuracy, a calibration must be made at manufacturing level to eliminate the offset introduced by the Op-Amp. But when the temperature changes, we can clearly see on the oscilloscope that the output offset voltage of the TSV611 changes from minus 125 millivolts to minus 90 millivolts. This means that when using the TSV611, a calibration must also be made on the temperature at the manufacturing level, which is very costly. By using the TSZ-121, which is a zero-drift amplifier, we don't necessarily need this calibration phase as the VIO stays very low even with a large temperature variation, because it exhibits a DVIO DT of 30 nanovolts per degree Celsius. So when the temperature increases by 100 degrees Celsius, the VIO will vary within a range of 3 microvolts. When we speak about precision, another important point must be considered, the CMRR. The common mode rejection ratio is defined as the ratio of the differential voltage amplification to the common mode voltage amplification. This is measured by determining the ratio of a change in input common mode voltage to the resulting change in the input offset voltage. The common mode input voltage affects the bias point of the input differential pair. Because of the inherent mismatches in the circuitry, changing the bias point changes the offset voltage, which in turn changes the output voltage. In general, a rail-to-rail op-amp has parallel input stages made of a p-pair, which work on the low input common mode voltage and at an end-pair, which work on the high input common mode voltage. As seen previously, the mismatch between two NMOS or two PMOS is responsible for the VIO, but there is no link between the mismatch of the NMOS and the mismatch of the PMOS. This means that each pair will generate its own VIO. So depending on the common mode voltage used in the application, the VIO might be different. In a precision environment, the main goal is to achieve the lowest VIO jump when the signal switches from one pair to another. We can have an idea of the impact of CMRR on an op-amp used in differential mode to sense a current through a shunt. Highside current sensing is typically selected in applications where ground disturbance is not tolerated and short-circuit detection is required, such as battery current monitoring. The application shows a single op-amp used in the differential amplifier made of the TSV711 op-amp and four external resistors. It amplifies a small voltage drop across the sensing resistor of 8 mOhms by the gain RF over RG or 1000 while rejecting the common mode input voltage. First, the battery is fully charged and its voltage is at 4.2 volts. The R-load source is 100 mA, so through the 8 mOhms shunt, a differential voltage of 800 mV appears on the input of the amplification stage. If we consider a perfect world, the input should be amplified by the gain, so on the output we should have 800 mV. But despite a lot of effort, we are still not in a perfect world and some errors must be considered. As seen previously, it is important to take the VIO into consideration. The TSV711, which is already a precision amplifier, shows a maximum VIO of 200 mV at 25 degrees Celsius. So, on the output, we will not see 800 mV but 600 mV. So, there is a 25% error due to the VIO. The CMRR due to the mismatch of the resistance must be considered and depending on the precision of the resistance, the CMRR res might be predominant in the total output error. The CMRR of the differential amplifier is given by this equation and if we consider that the precision of the four resistors is 0.1% with a gain of 1000, we can obtain a CMRR of 108 decibels. So, when the battery is fully charged at 4.2 volts, the CMRR due to the mismatch of the resistance will add an error on the output of roughly 17 mV. The TSV711 has its own CMRR specified in the datasheet, which is 74 decibels. In this case, the TSV711's CMRR will be the predominant one and it will cause an error of 340 mV on the output, so an error of more than 42% compared to the theoretical value. When the battery discharges, the input common mode voltage will change as the schematic shows high side current sensing. In the datasheet, the VIO is defined as VCC divided by 2, so in this case, when the battery discharges, VCC is equal to 2.5. This means that the TSV711's CMRR will not have any impact on the output. Only the CMRR due to the mismatch of the resistance will play a role by adding an error of more than 1%. This table summarizes the impact of the CMRR at different battery voltages when using the TSV711. We can see that when the battery is fully charged, the R-Pamp's CMRR is the main contributor to the accuracy of the current measurement. We can also clearly understand that the precision of the measurements will change depending on the voltage level of the battery due to the CMRR of the TSV711 amplifier. Let's keep this application in mind, but replace the TSV711 by the TSZ-121 R-Pamp. The theoretical value of the output is still 800 millivolts without any error introduced by the R-Pamp itself. The TSZ-121 has a maximum VIO of 5 microvolts at 25 degrees Celsius, so by considering this error and a gain of 1000, the output will be 795 millivolts and therefore results in an error of 0.5%. As the resistors are still precise at 0.1%, the error on the output is still the same and add 2.1% of error. In this case, the TSZ-121 has a CMRR of 115 decibels, so its own impact on the total output error is not predominant. When VBAT equals 4.2 volts, it represents only 3 millivolts on the output, so just 0.4% of the error. As already seen with the TSV711, the main error was introduced by the CMRR of the R-Pamp. In this case, by choosing a more precise R-Pamp, such as the TSZ-121, which exhibits a very high CMRR of 115 decibels, the main error is now introduced by the external resistors. With a VCC equals 4.2 volts, the total error has been divided by 10 by using the TSZ-121. Moreover, we can also see that with a very precise R-Pamp, the error on the output will be roughly the same, even with a different VBAT, which was not the case when using the TSV711. When the battery discharges, the input common mode voltage will change, in this case, the resistors are impacted the most, so the error is the same as previously with the TSV711. The power supply rejection ratio PSRR and the differential voltage amplification AVD are also important parameters when making precision measurements, but the Op-Amp's PSRR will have a low impact if the power supplies are well decoupled. It is the same thing for the AVD if the gain of the amplifier is not so high, less than 1000, this parameter will not create any issues. In order to take into account all the parameters likely to have an impact on precision, we can use this equation. Where the first parameter expresses the input offset voltage, the second parameter, the AVD, the third parameter expresses the CMRR, the fourth expresses the PSRR, and the last parameter expresses the input voltage drift with temperature. In addition to errors in the voltage domain, that is, voltage offset, VIO and input voltage noise density, EN, current domain errors, such as the input current, IN, are also important sources of error, especially for high source impedances, above 100 kilo ohm. The technology used for the Op-Amp can significantly impact the whole precision of a system. The input bias current parameter, IIB, is defined as the average of the current into the two input terminals with the output at a specified level. The input circuitry of all Op-Amp's requires a certain amount of bias current to operate properly. The input bias is defined by the following formula. CMOS and JFET have a much lower input current than the standard bipolar. Indeed, for a bipolar architecture, part of the current coming from the current source will flow in the input with a ratio of 1 per square beta. The CMOS transistor is driven by a gate, and there is an insignificant current inside. The small input current that can appear in a CMOS technology is mainly due to leakage of the ESD diode. For sensors with small source impedances, current domain errors dominate. While for higher source impedances, current domain errors dominate, especially for bipolar Op-Amp's. The input bias current, even if it represents a very small current, might affect the precision of a measurement, especially when we need to measure a low current using an Op-Amp. In this application, we want to measure UV radiation using a UV sensor. The UV sensor delivers a small current depending on the intensity of the UV source. A transimpedance circuit is used to convert the current delivered by the UV sensor thanks to the feedback resistance 10 mega ohm. The capacitance in the feedback helps to stabilize the system. The UV source is set with an index of 4, and for this level of radiation, the UV sensor will generate a current of 104 nanoamps. This very small current is amplified by the 10 mega ohm resistor, resulting theoretically in an output voltage of 1.04 volts. First, let's use a bipolar Op-Amp, such as the LM2904. This kind of Op-Amp might have an input bias current up to 200 nanoamps, or twice the current we want to measure, so it will completely affect the output results. The voltmeter will display minus 0.96 volts, which corresponds to index 1 in the UV sensor conversion table. Now, we can keep the configuration as it is, and just replace the LM2904 with a CMOS Op-Amp, such as the TSV611. In this case, we can see that the output is close to the theoretical value, and the level of output voltage corresponds to UV index 4. In case of an application where the current to be measured is extremely low, or where the input impedance is very high, it is mandatory to use a CMOS Op-Amp, so as not to affect the measurement. In the precision domain, the TSZ-121 is generally the best candidate, as it offers extremely good parameters due to its chopper architecture. But the input stage of chopper stabilized amplifiers does not behave like conventional amplifier input stages. The TSZ-121 uses switches on the inputs that continually chops the input signal at 100 kHz to reduce input offset voltage down to 5 microvolts. The dynamic behavior of these switches induces a charge injection current on the input terminals of the amplifier. The charge injection current has a DC path to ground through the resistors seen at the input terminals of the amplifier. Higher input impedance causes an apparent shift in the input bias current of the amplifier, resulting in a higher input bias current than conventional CMOS Op-Amp's. It is hard to find Op-Amp's that can be used across a wide range of source impedances, for example, 10 ohms to 10 mega ohms, and still achieve DC precision. A comparison of the state-of-the-art DC specifications of CMOS and precision chopper Op-Amp's is shown in the following graph. If these Op-Amp's are used to interface a sensor with a certain source impedance, RS, the resulting offset is given by equation 1. Although the offset performance of a chopper Op-Amp is better than the rest of the competition across a wide range of impedances, it can be seen that its offset performance starts to degrade rapidly when the source impedance, RS, exceeds a threshold given by equation 2. We can clearly see that if the sensor used has an impedance higher than 1 mega ohm, it is better to choose the TSV-711 rather than the TSZ-121. Noise is also a key parameter in a precision environment. It is part of life, and we have to deal with it. When an electronics component even passive is added to a system, it will add noise that will impact the signal-to-noise ratio. Noise is not easy to understand, as it is non-periodic, and it must be considered using statistics. The easiest approach is to think of it being in the frequency domain, even if engineers generally prefer the time domain. All internal sources of noise contribute to the overall noise generated by the operational amplifier. The Op-Amp noise is modeled with three noise sources. One source for the input noise voltage, EN, and two sources for the input noise current, IN. A current issued from a current noise source and flowing into a resistor generates voltage noise according to Ohm's law. All sources are physically independent and therefore uncorrelated. Two other sources of noise can be added due to the gain resistances, RG and RF. This noise source can be expressed. As a spectral density in nanovolts per square root and hertz for voltage sources or picouamps per square root and hertz for current sources, which can be seen as the noise energy at a given frequency. As an RMS value for a given bandwidth, let's have a look at each noise source. The noise voltage source in the classical Op-Amp architecture showing a combination of two different noise types. At lower frequencies, it is done 500 hertz, flicker noise, also called 1 over F noise or pink noise appears. 1 over F noise is caused by defects at atomic level in semiconductor devices. This noise is the main contributor at low frequency. It is generally expressed in nanovolts per square root and hertz. After the F corner frequency, the voltage noise source becomes white noise. It is a result of thermal agitation of the charges in an electric conductor and it is also expressed in nanovolts per square root and hertz. It is the main contributor of noise at higher frequencies. This is why, generally in the datasheet, the noise spectral density is provided at different frequencies. The current noise source also adds its contribution to the overall noise, especially if the Op-Amp is surrounded by high impedances. But the input noise current for CMOS input Op-Amp is extremely small and generally does not affect the design as it is roughly 5 femto-amps per square root and hertz. But the input current noise for bipolar Op-Amp or chopper architectures is in the range of 100 pico-amps per square root and hertz. Resistors will add also white noise according to the equation square root of 4KTR. We can see that the greater the resistance, the higher the noise. This table summarizes the different noise sources from a mathematical point of view. The second column represents the noise of each source expressed in spectral noise density, nanovolts per square root and hertz. As it is a noise referred to the output, it is generally multiplied by the gain of the circuit. The last column expresses the RMS value of the same noise. It is the integration of the spectral noise over the bandwidth of interest. It is important to consider that the output noise is added in a quadratic sum. This equation expresses the overall RMS noise on the output. In order to significantly reduce the noise level in an application, it is important to minimize the value of the resistors and reduce the bandwidth as the wider the bandwidth, the higher the RMS value. This is done by inserting a capacitor in parallel with an RF resistor. The cutoff frequency can be calculated to give minus 3 decibels at 5 to 10 minutes to the highest frequency to pass. In this example, we can see the contribution of each source. Here the noise is expressed in spectral density. To have a better understanding regarding the real impact it has on the output, let's transpose it to the VRMS value for a bandwidth of 30 kilohertz. This is the noise voltage contribution to the output of the current noise source. This is the contribution of the RF resistance. This is the RG resistance. This is the equivalent voltage noise source of the op-amp. We can see that the impact of the current noise on the output is negligible compared to the other noise sources. The voltage noise source of the op-amp represents the main part of the noise in this application. We can see that the contribution of the voltage noise of the op-amp is much higher than the other noise sources. However, if no care is taken regarding the value of the external resistor, their impact will become non-negligible. To get the total error generated by the noise sources, you have to calculate the quadratic sum of each noise source. In this example, there will be 5.4 millivolts RMS. Here is an example of a half bridge strain gauge. The variation of the strain resistance will make a small variation of the input voltage. The small voltage variation must be detected with precision to give a correct value of the weight applied on the strain gauge. Previously, we have seen that the VIO is extremely important for this kind of measurement. If the VIO introduces a large error, there is still a possibility to calibrate the system and completely remove the VIO, but the noise is a non-periodic signal and it cannot be calibrated. So, for an application requiring precision, it is also important to take noise into consideration. In this case, the bandwidth is limited to 10 kilohertz thanks to the 1.5 nanofarad capacitor in the feedback. Let's have a look at the impact of the noise with two different op-amps. The TSV731 which is a high accuracy amplifier and the TSZ121 which is a precision chopper amplifier. The calculation shows the noise impact on the output. The VIO and DVIO DT must be taken into consideration regarding the whole error. If the bandwidth of the application is quite large, there is no difference in term of noise error between the two op-amps as they have the same white noise around 35 nanovolts per square root in hertz. The strain gauge is a DC application so it is important to limit the bandwidth of the application in order to reduce the noise amplitude. Here, the bandwidth is limited to 20 hertz thanks to the 795 nanofarad capacitor which was added in the feedback. In this case, we can see a significant difference between the two devices. In fact, the TSV731 shows a 1 over f noise that the TSZ121 does not have due to its chopper architecture. With the TSZ121 the noise impact on the output is reduced by 4. The noise can also be expressed in VPP and in case of the TSZ121 with a bandwidth limited to 20 hertz, the noise will be around 960 micro vpp. It is a first order noise calculation as we have to consider a stop bandwidth with a very sharp edge. This means in the worst case roughly 600 microvolts of error might be added to the output value caused by the noise. This is a summary of the most important parameters to consider when we speak about precision for an op-amp. It represents the input referred error that an op-amp might introduce in the whole measurement and of course this error must be multiplied by the gain of the configuration. If the op-amp has an offset drift over temperature of 10 micro volts per degree celsius it means that if the ambient temperature increases by 70 degrees celsius the input offset will increase by 700 micro volts. If the common mode voltage application can vary from 0 to 3 volts and if the amplifier has a cmrr of 80 decibels you can expect a vio variation of 300 microvolts. And if the power supply used to power the op-amp varies by 10 percent and if the amplifier shows a psrr of 80 decibels you can expect a vio variation of 50 microvolts. Noise is expressed on a 10 hertz bandwidth and in the application it is necessary to add a sharp 10 hertz filter to consider this 10 micro vpp otherwise the noise must be integrated over the entire bandwidth. The TSZ-121 which is a chopper op-amp shows very good characteristics regarding all 5 parameters. In this slide you will find the precision op-amp's portfolio with the main features of each series and the ideal applicability of each one. Here are some tools that will be useful to continue learning about operational amplifiers. Don't miss the opportunity to design and perform simulations with our platform eDesign Suite. You will find the links to access the different tools below in the video description. Thank you for watching this video. For more information visit www.st.com slash op-amp.