EEPW论坛

故事从晶体谈起：当反向偏置（PN结）形成时，半导体的连接点会有一个反向击穿电压。对于敏感的运放输入端晶体管，这种情形是极为不利的。

　　消费者经常将程序控制器返修，因为现场操作员没有正确调整它们。例如，假定一个天然气管道运行在3000 psi。当用校准过的压力传感器测量0到5000psi（对应输出为直流1到5V）时，3000psi由程序变量产生3.4V的输出。工作人员应该在3000psi时将输出调整到3.4V。现在程序变量相当于定点，并且控制环静止不动，换句话说，没有产生输出变化。

　　现场工作人员经常在程序变量和设定都是1V，而不是在它们都是5V时，调节控制器输出为恒值。在这种情况下，输入差分放大器的CMRR（共模抑制比）不令人满意。

　　考虑典型的程序控制回路，包括一个45V电源，从控制室到测量场所的一对导线，一个电流传感器，一个250Ω可变电阻，所有的器件都是串联。测量程序变量的电流传感器在导线中产生4到20mA的电流，代表0到100％的程序变量值。可变电阻上的电流就产生1到5V的控制器输入电压。

　　有时，1到5V的输入信号已经可用，那么可变电阻就是多余的了。那么 
在控制器上就可以不插这个电阻。然而，如果控制器工作在4到20mA信号，而又忘记安装可变电阻，那么控制器的输入大约是45V。危险！

　　程序控制器电路包括输入连接器与第一个运放间的电阻应当大于100kΩ，那么超过运放最大共模抑制比的输入电压也不会导致损坏，不是吗？电流能够好到足以通过一个100kΩ的电阻就可以改变45V电压源的运放特性吗？

　　我曾经问主办摩托罗拉研讨会的工程师，运放±15V供电，用一个100KΩ电阻将45V连到运放的输入端，是否会损坏运放。他回答：“是，运放的输入设备会雪崩击穿，放大功能可能不会失去，但是工作特性会改变”。

　　雪崩？这意味着什么啊？

　　当反向偏置形成时，半导体的连接点会有反向击穿电压。齐纳击穿二极管就是这种工作方式，在电压稳压方面非常有用。但这种情形对于敏感运放的输入端晶体管是极为不利的。像上面描述的电路一样，这种现象通常并不致命，但是会改变运放的表现。也会损坏输入差分信号的平衡。

　　所有我们要保持输入电压不能超出±15V供电电压。我们将二极管放置到控制器的输入端（阴极与＋15V相连，阳极与输入针相连），如果控制器输入电压超出正供电电压，二极管将导通。为了保护反向输入电压不要超出－15V（万一线路错误），我们也在控制器输入端和－15V的电压之间（阳极连接－15V，阴极连接输入针）放二极管。

　　这些二级管一定要好的，在最高操作温度时漏电流要尽量低。在0%(4 mA, 1V)时读取，40 µA为1%的误差。

　　添加了二极管后，顾客再也没有将改进的还控制器返修理过。

　　英文原文：

　　Avalanche: Overwhelming input voltage sets off cascade of process-controller failures

　　Tales From The Cube: Semiconductor junctions have a “reverse breakdown voltage” at which a reverse-biased junction begins to conduct. This situation is unhealthy for sensitive op-amp input transistors.

　　By Walter Lindenbach -- EDN, 7/19/2007

　　Customers often returned process controllers to us for repair, because the field operators could not adjust them correctly. For example, consider a gas pipeline that is to operate at 3000 psi. When measured by a pressure transducer calibrated for 0 to 5000 psi (with an output of 1 to 5V dc), 3000 psi produces an output of 3.4V, which is the process variable. The operator will have adjusted the setpoint to 3.4V also for operation at 3000 psi. Now the process variable is equal to the setpoint, and the control loop should “sit still”—that is, produce no output change.

    Often, the field operator could adjust the controllers for constant output when the process variable and setpoint were both equal to 1V but not when they were both equal to 5V. In this situation, the CMRR (common-mode rejection ratio) of the input differential amplifier was unsatisfactory.

　　Consider a typical process-control loop, which contains a 45V power supply, a wire-line pair (or equivalent) 
from the control room to the measurement site, a current transducer, and a 250Ω “range resistor,” all in series. The current transducer, which measures the process variable, produces a current of 4 to 20 mA in the wire-line pair, representing 0 to 100% of the value of the process variable. This current in the 250Ω-range resistor produces an input for the controller of 1 to 5V.

　　But sometimes a 1 to 5V input signal is already available and the range resistor is unnecessary. Then the installer does not plug it in at the controller. However, if the controller is to work with a 4- to 20-mA signal, and the installer forgets the range resistor, the input to the controller is about 45V. Not good!

　　The process-controller circuits included resistors between the input connector and the inputs of the first operational amplifier (configured as a differential amplifier) that were greater than 100 kΩ, so an input voltage that exceeded the amplifier’s maximum common-mode rating would not cause damage, would it? Could a current great enough to change the characteristics of the op amp come from a 45V source through a 100-kΩ resistor?

　　I asked an engineer leading a Motorola seminar whether applying 45V to a 100-kΩ resistor connected to an op-amp input, operating with a ±15V power supply, could harm the amp.

　　“Yes,” he said. “The op-amp input devices will avalanche. The amp may not fail, but the operating characteristics will change.”

　　Avalanche? What did that mean?

　　Semiconductor junctions have a “reverse breakdown voltage” at which a reverse-biased junction begins to conduct. Zener diodes behave in this way, which makes them useful as voltage regulators. But this situation is unhealthy for sensitive op-amp input transistors. With a circuit like the one described above, this phenomenon is not usually fatal, but it will change the behavior of the op amp. It will spoil the balance of the input differential pair.

　　So, we had to keep the input voltage from exceeding the supply voltages (±15V). We placed diodes at the controller inputs (cathodes to the 15V supply; anodes to the connector-input pins) that would conduct if the controller-input voltage exceeded the positive supply voltage. To protect against input voltages more negative than –15V (in case of a wiring error), we also placed diodes between the controller inputs and the –15V supply (anodes to the –15V supply; cathodes to the connector-input pins).

　　These diodes had to be good; leakage had to be low at the highest operating temperature. At a 0% reading (4 mA, 1V), 40 µA represents an error of 1%.

　　After we added the diodes, customers no longer returned the modified controllers.