双DAC保存系统偏置电压，在电源校准程序中决定了该电压。
设计电阻桥式传感器与5V单电源供电的ADC接口是一个新的挑战。有些应用需要输出电压在0V到满量程电压（如4.096V）之间以高精度波动。用大多数单电源仪表放大器，当输出信号接近0V，接近单电源最低输出摆幅限制时，会出现问题。一个好的单电源仪表放大器可在接近于单电源地的范围内摆动, 即使有真正的轨对轨输出，也不能达到地。
在这个应用中，传感器是一个精密的测压元件，其额定负载5kg，即约11磅。在铝盘上测重大约150g的物体，即大约5盎司。由于铝盘自重，即使没有任何物体称重，仪表放大器的输出信号也不能低到0V。现在，问题是如何补偿仪表放大器的输出偏置电压和铝盘本身产生的电压值。
软件弥补系统偏置是最简单的方法。电源启动期间，铝盘上没有称重物体，系统可以获取偏移电压，并将数据记录在单片机内存中。随后，当有物体称重时，从获得的数据中减去它即可。但是，这种做法不能达到5kg满量程，仅能达到5-0.15kg或4.85kg。
本设计方案说明如何利用单片机实现硬件补偿。当电源启动后，运行软件程序复位系统偏移。解
决方案如图1所示，基于四个来自于Linear公司IC的简单电路。精密参考电压源IC1，有高达50mA的最小输出电流。它提供4.096V输出电压驱动测压元件，并设置12位ADC（IC3）的满量程范围。高精确仪表放大器LT1789-1（IC2）的特点是在0到70°C温度范围内，最大输入失调电压为150 µV，轨对轨输出电压相对地110mV范围内摆动时，最大输入失调偏置电压是 0.5µV/°C。通过精密电阻R2（阻值为500Ω）设定增益，当称重是5kg时，输出范围为4.096V，其最大输入信号是VCC×S=4.096V×2 mV/V=8.192 mV，这里S是该传感器的灵敏度。
双通道DAC（IC4）的DAC_A输出在仪表放大器参考引脚处，提供200mV的参考电压，避免放大器本身近地饱和，但传输特性不是线性关系。放大器最坏情况下输出偏移是：VREF+VPAN±VOFFSET=200 mV+125 mV±500×150 µV=325 mV±75 mV="250" mV/400 mV，这里VPAN=125 mV，是铝盘自重产生的电压值。
电路补偿基于测压元件平衡的系统偏置图示" border="0" height="220" hspace="0" src="http://files.chinaaet.com/images/20100811/a778bc16-8f86-4100-b136-bf7f7f43fbc8.jpg" vspace="0" width="500" />
因此系统输出偏移是250到400mV。电源启动，微控制器运行程序设置DAC_A输出为200mV，同时，增加双通道DAC（IC4）的DAC_B输出直到等于ADC（IC3）管脚2的系统偏置，转换结果就是000h。这一结果是可能的，因为IC4包含两个12位2.5V满量程电压的DAC，最低有效位（LSB）等于0.61mV，小于IC3为1mV的分辨率。这个数字相当于该天平的分辨率：5000g/4096=1.22g。当最大负载5kg时，仪表放大器的最大输出电压是4.096V+VOUT_TOTAL_OFFSET_INA=4.346V/4.496V，低于4.62V高饱和温度的最坏情况。
IC3有一个单极差分输入，所以可以从+IN输入电压中减去一个恒定电压值等于IC4的DAC_B提供的系统偏置。在第一个半时钟周期内，ADC采样和保持正向输入电压。这阶段结束后，或在获取时间内，输入电容切换到负输入并开始转换。在IC3输入处的RC输入滤波器的时间常数为0.5µs，允许在正负输入电压利用最高为200kHz时钟频率在转换时间的第一时钟周期内达到12位精度。如果想增加时间常数，必须降低时钟频率。
此外，DAC和ADC有三线串行接口，可方便地将数据传输到最高采样率为12.5kS/s的普通微控制器。当ADC处于没有转换的时候，它会自动把功率降至1nA的电源电流，而且如果单片机通过其引脚3来关闭IC1，电路限制电源电流在最坏情况下仅为1mA，因为所有的IC集成电路都是微功耗的。
英文原文：
Circuit compensates system offset of a load-cell-based balance
A dual DAC stores the system-offset voltage, which gets determined during a power-on calibration sequence.
Luca Bruno, ITIS Hensemberger, Monza, Italy; Edited by Charles H Small and Fran Granville -- EDN, 8/16/2007
It’s a challenge to interface a resistive bridge sensor with an ADC receiving its power from a 5V single-supply power source. Some applications require output-voltage swings from 0V to a full-scale voltage, such as 4.096V, with excellent accuracy. With most single-supply instrumentation amplifiers, problems arise when the output signal approaches 0V, near the lower output-swing limit of a single-supply instrumentation amp. A good single-supply instrumentation amp may swing close to single-supply ground but does not reach ground even if it has a true rail-to-rail output.
In this application, the sensor is a precision load cell with a nominal load of 5 kg, or about 11 lbs, to weigh objects on an aluminum pan weighing approximately 150g, or approximately 5 oz. Because of the pan’s weight, the instrumentation amplifier’s output signal can never go down to 0V, even if there are no objects to weigh. Now, the problem arises of how to compensate the instrumentation amp’s output-offset voltage and the voltage that the pan itself produces.
A software approach is the simplest way to compensate the system offset. During power-up, there are no objects to weigh on the pan, and the system can thus acquire the offset voltage and hold the data in the microcontroller’s memory, subsequently subtracting it from the data it acquired when there was an object to weigh. This approach, however, does not reach the 5-kg full-scale of the balance, reaching only 5–0.15 kg, or 4.85 kg.
This Design Idea shows how to achieve hardware compensation using a microcontroller that, on power-up, starts a software routine to reset the system offset. The solution is a simple circuit based on four ICs from Linear Technology in Figure 1. A precision voltage reference, IC1, has a high minimum output current of 50 mA. It provides an output voltage of 4.096V to power the load cell and to set the full-scale of the 12-bit ADC, IC3. The highly accurate LT1789-1 instrumentation amplifier, IC2, features maximum input-offset voltage of 150 µV over the temperature range of 0 to 70°C and maximum input-drift-offset voltage of 0.5 µV/°C over the temperature range of 0 to 70°C with rail-to-rail output that swings within 110 mV of ground. You set the gain through precision resistor R2 to a nominal value of 500Ω to give an output span of 4.096V when the load is 5 kg and its maximum input signal is VCC×S=4.096V×2 mV/V=8.192 mV, where S is the sensor’s sensitivity.
The output of DAC_A of dual-DAC IC4 provides a reference voltage of 200 mV at the refer
ence pin of the instrumentation amp to avoid saturation near ground of the amplifier itself, where its transfer characteristic is not quite linear. The amplifier’s total worst-case output offset is: VREF+VPAN±VOFFSET=200 mV+125 mV±500×150 µV=325 mV±75 mV="250" mV/400 mV, where VPAN="125" mV and is the voltage that the pan’s weight produces.
The system-output offset is thus 250 to 400 mV. On power-up, the microcontroller starts a routine that sets the output of the DAC_A equal to 200 mV, while it increases the output of the DAC_B of dual-DAC IC4 until it is equal to the system offset on Pin 2 of ADC IC3, and the result of the conversion is 000h. This result is possible because IC4 contains two 12-bit DACs with the same full-scale voltage of 2.5V, making 1 LSB equal to 0.61 mV, which is smaller than IC3’s resolution of 1 mV. This figure corresponds to the resolution of the balance: 5000g/4096=1.22g. The maximum output voltage of the instrumentation amp with a maximum load of 5 kg is 4.096V+VOUT_TOTAL_OFFSET_INA=4.346V/4.496V, which is less than the minimum worst case over temperature of 4.62V high saturation.
IC3 has a single unipolar differential input, so you can subtract from the +IN input voltage a constant voltage of value equal to the system offset that that DAC_B of IC4 provides. During the first one and a half clock cycles, the ADC samples and holds the positive input. At the end of this phase, or acquisition time, the input capacitor switches to the negative input, and the conversion starts. The RC-input filters on the inputs of IC3 have a time constant of 0.5 µsec to permit the negative and positive input voltages to settle to a 12-bit accuracy during the first clock cycle of the conversion time, using the maximum clock frequency, which is 200 kHz. If you want to increase the time constant, then you must use a lower clock frequency.
Furthermore, the DAC and ADC have a three-wire serial interface that easily permits transferring data to a wide range of microcontrollers with a maximum sampling rate of 12.5k samples/sec. When the ADC performs no conversions, it automatically powers down to 1 nA of supply current, and, if the microcontroller shuts down IC1 through its Pin 3, the circuit draws a worst-case supply current of just 1 mA, because all the ICs are micropower.