How to determine the current range of a 4-20mA device?

　　About 30 years ago, someone told me he had a 4-20mA current source. He wanted to determine if the current was out of range, or if the wire was broken! But no one knew how to detect that. If you want a tough challenge, just tell me that this is an analog function, and no one knew how to detect it. So I started thinking, how can you tell if the 4.0mA current has dropped to 3.70mA or less? It would be great if we could determine that, so that we could say it is an illegal state.

　　I decided to use Bob Widlar's new LM10, which combines a voltage reference and a power amplifier . Of course I would build an effective circuit to detect and send an error message when the current is too low. Later, after drawing a circuit diagram and repeatedly modifying it, I built a useful circuit. The customer also thought the circuit was useful.

　　Since we knew that there was an instrumentation and industrial magazine whose readers often used "4~20mA" devices, I also submitted this circuit as a "design idea" to this magazine. So the magazine published this circuit. About three months later, we received a delightful letter, showing unprecedented interest in this little circuit that the magazine had never experienced before. Well, I guess so.

　　Today, the LM10 is still in production and sold, but the price is $2.30, which is obviously a bit high compared to the actual value of such a simple function. When we got the cheaper LM4041-ADJ (37 cents per unit in volume), I thought, "This part should be able to do this." The LM4041-ADJ has a small gain stage and a 1.2V voltage reference, so it can do these functions (see figure).

　　The LM4041-ADJ in the figure can be used to implement the current judgment function.

　　Main Specifications

　　The gain of the 4N28 is fairly modest (0.1 to 0.3), but it produces a small flag that can be detected when it drops close to "ground potential". The LM4041-ADJ can sense 4.0mA of current through the 332Ω resistor and turn on the 4N28. If the current drops below 3.7mA, the LM4041 turns off the optocoupler. Even a simple circuit can achieve very useful functions. You don't have to find a 30-year-old magazine.

　　If you want to check the actual levels this circuit switches and senses, you might want to put a small triangle wave tester through the calibrated current above and below 3.7mA. If the output duty cycle is exactly 50%, you'll know the threshold is correct. You can fine-tune this if you want better than 2% accuracy.

　　I used to work for Teledyne , and if you know Greek, you'll find that Teledyne means "distance and force." Even with isolation from voltages hundreds of volts above or below ground, this circuit can generate a small force. There is no electrical connection. So isolation does not result in lower accuracy or higher cost.

　　When is long-term stability testing necessary?

　　A few years ago, a friend called me about a new NSC amplifier and asked, “Do you have any data on the long-term stability of this op amp?” Without even looking it up, I told him that we didn’t. It was just a fairly new, run-of-the-mill op amp that wasn’t trying to be a technology leader in low bias or low drift. So, we weren’t bragging, there was no risk involved. But the customer wasn’t happy.

　　"Why don't you have complete information on this?" I tried to explain that we don't have the resources to collect this kind of data for every small product, nor can we afford to collect and analyze the data and delay the product launch. Even if we publish this data immediately and then update the data sheet, users will still ask the same question if they look at the original data sheet.

　　The amount of manpower, engineering, and technician work required to conduct a proper study would be very large. Few customers have such an urgent requirement that it is worthwhile for us to collect such data. Generally, if we have good data on previous circuits using this process, the changes are quite small, such as a small change in wiring or current magnitude, or a change in output stage, and here we assume that the new device will be very similar to the old one.

　　Now, let's do a little drift test just to make sure nothing goes wrong. For example, we'll load up three boards, each with 30 devices, to see if the boards are performing as expected. We'll compare the data before and after 1,000 hours of high temperature operation. We could put one of the boards in an oven and collect 2,000 hours of data. But this kind of testing is usually a boring exercise. When we're done with all the operations, there's nothing to brag about. There's nothing to write in the report other than "passed."

　　The caller was still not satisfied. "Okay, I'll look into it and ask your competitors what their long-term drift is," he said. To which I responded, "Go ahead," and started giving him the phone numbers of some of our competitors. But then I paused and told him, "But they'll tell you the same story. They don't have the luxury of doing accurate life testing on every good circuit they ship."

　　When do we need to do extensive testing and data logging? When we have a new process or circuit that promises superior performance. A new low-drift op amp? Of course. When National Semiconductor introduced a new chopper-stabilized amplifier a few years ago, we did all kinds of life testing to make sure we didn't have any unstable products. What does it mean to have a good amplifier when there aren't any unstable amplifiers? We did a lot of finicky testing to analyze enough data to determine a typical drift rate of "0.006μV/month" (e.g., the LMP2012).

　　What drift do you expect to see in one or two years? Our standard rule is "If the time is N × 1000 hours, we expect the drift to be n = &rad ic ;N." This often makes customers happy and leaves them with nothing to say. Because: 1. This rule is often true, or close to it; 2. If the customer wants to get data, it will be a lot of work for him! In this way, he will not ask questions again soon. For example, if the time is two years (that is, 16,000 hours), the drift can be estimated by √16 = 4.

　　LM199AH Voltage Reference

　　When the new LM199AH was released 38 years ago, it was a new circuit designed using a new process to eliminate all possible causes of long-term drift. Its output tolerance was ±3%, but long-term stability was typically 0.0020% per 1000 hours, or 20ppm.

　　We did a lot of preliminary testing to screen out unstable devices and then put them into comparison circuits so that we could use a nice six-digit digital voltmeter (DVM) to compare multiple voltage reference sources (such as a temperature-controlled standard battery, a temperature-controlled bandgap voltage reference, and several other pretty good Zener diodes).

　　By using multiple voltage references, we can avoid the situation where all devices under test (DUTs) appear to drift at the same time. Is the problem due to all DUTs drifting? No, because the other voltage references show the same degradation at the same time, which means the problem is with the DVM's voltage reference. This effect can be "stripped out" or at least ignored.

　　One day, I had an idea to take out a bunch of LM199AHs and solder them together in groups of four, using a small resistor (499Ω) to average their outputs. The output seemed to have less noise and lower drift. So we did another test. Soon I found four groups of four LM199AHs.

　　I compared the average output of eight LM199s to the average output of another eight LM199s, and the results are indeed good! Some tests showed less than 2μVpp noise in a limited bandwidth (4Hz). If I took the average output of 16 LM199s, the output noise would probably be even lower! Most users won't need noise that low, but by averaging several circuits, you can make the noise go down square root until you run out of energy, space, and power .

　　By Bob Pease

　　Analog designer Bob Pease is one of the legends of Silicon Valley. More than 20 years ago, he opened a column in "Electronic Design". Even when he traveled abroad to the Himalayas, he would insist on publishing column articles, which undoubtedly became the most popular content in each issue. The content of the column is quite comprehensive, and you can discuss current sources and ancient tube amplifiers, or his obesity. He has a clear point of view, but he communicates with a sense of humor. Whether you agree with his point of view or not, you will like him as a person. Not long ago, this 71-year-old man was unfortunately killed in a car accident.