I'm working on a circuit that senses AC presence and measures DC output from a switching power supply and a battery. They will stay inside a grounded metal case but out in the field, so the AC mains are already protected by a varistor and the AC presence sensor is opto-isolated, so I guess that is enough for the AC detection part. For the DC measurements though, on my prototype I'm just using a resistive voltage divider to scale down the battery and supply levels to the ADC range, but I'm wondering if any protection would be recommended as well. The measurements are taken on the switching supply and battery charging outputs, so they are already somehow isolated from the AC input. No wiring or sensor inputs are coming from outside the metal case apart from the AC mains, so, do you think any protection would be recommended on the ADC inputs other than clamping diodes to the ADC voltage input rails?
2 Answers
The typical way to protect ADC inputs (and many other high impedance inputs) is to have a current limiting resistor and clamping diodes at the pin. Many chips already have clamping diodes inside the chip, but it is often useful to have some outside as well (especially when the integrated clamping diodes do not have proper documentation on their specs, which is most of the time).
Your resistor-divider is already going to function as a current limiting resistor. As a general rule, you want your resistor values to be as high as is practical-- with something in the neighborhood of 10K to 500K being reasonable.
As the resistors go higher in value you start to get weird effects from the parasitic capacitance (lowers your frequency response) and the leakage current of the ADC and diodes (messes up your accuracy). If you are unsure about these things, then pick something on the low side of 10K to 500K. If you have already figured out the math or are willing to just try it and see if it works then go with something higher.
There would be one diode between GND and the input. And another diode between the input and VCC. If you have a signal that goes + and -, then connect one diode to the V+ rail and the other to the V- rail. Orient the diodes in such a way that it works (turn the diodes around and nothing will work). Put the diodes between your ADC and the resistor divider.
It works like this... If you get a voltage spike coming in, the diodes will redirect it to a power rail and not directly into the chip. The current limiting resistor will, duh, limit the current to something reasonable. You want to current to be limited for several reasons: 1. You don't want so much current that the diodes blow up. 2. You don't want so much current going into the power rails that it effects the voltage regulation of that rail.
With this setup, the maximum voltage on the input pin will be VCC + the forward voltage drop of the diode. And the minimum voltage on the input will be GND - the forward voltage drop.
There are hundreds or thousands of appropriate diodes for this. The BAV99 is just one of many, but is a good one to start looking at.
-
\$\begingroup\$ A good general rule is unless the ADC has protection diodes internally that are explicitly described as being for use under normal conditions, or you have talked to an engineer at the ADC manufacturer and specifically asked about using the integrated clamping diodes, you should design your hardware as if they weren't there at all. You can't make any assumptions about the durability or reliability of the integral protection diodes unless you have specifically talked to the manufacturer about them, or the data-sheet is explicit about their properties. \$\endgroup\$ Commented Aug 7, 2013 at 7:20
-
\$\begingroup\$ Thanks David. Since the ADCs will be measuring slowly varying DC voltages, I don't need to worry about the frequency response, so the resistors used will limit the current very well (they are taking up to 48V and scaling it down to 1.1V), so I guess the recommended diodes will be enough then. I was just wondering if the fact the box will be in an open field there should be any other EMI protection measures besides the intrinsic faraday's cage provided by the grounded metal box.. hehe. \$\endgroup\$– ClaudioCommented Aug 7, 2013 at 15:32
-
\$\begingroup\$ @Claudio EMI protection is always an issue, and there is no such thing as a perfect Faraday cage for any USEFUL electronic device. There are always holes and/or cables going into or out of a device, which messes up any Faraday cage effect that you have going on. Good shielding is never a substitute for proper PCB routing, signal termination, protection, and filtering. \$\endgroup\$– user3624Commented Aug 7, 2013 at 15:37
1) Spark Gaps will clip the highest voltage static electricity, but it still not low enough, but still they are FREE. http://www.carousel-design.com/SparkGap.html
2) The NEW Analog ADG5462F might be useful in some situations. http://www.analog.com/en/products/switches-multiplexers/fault-protected-switches-multiplexers/adg5462f.html
