Input Immunity

By William D. Kimmel, P.E.

and Daryl D. Gerke, P.E.

Kimmel Gerke Associates, Ltd.

Immunity problems, both RFI and transients, mostly enter the circuit board by wire (signal, power and ground), rather than by field coupling to the circuit board. The most common problem is simply jamming the input device, so the first order of business is to protect the input device. Let’s take a look at the problem, why it occurs and the steps necessary to remedy it.

The Problem

RFI effects dominate in the analog devices and transient effects dominate the digital devices. Digital circuits ignore low level RFI, but will respond to fast transients, which have sufficient amplitude. Analog circuits are mostly slow enough to ignore fast transients, but will respond to RFI. It’s not immediately obvious why RF in the MHz range will affect an audio frequency amplifier, but it definitely does, and there are actually several failure modes.

First, let’s take a look at the assumption that the frequency is too high for an audio frequency amplifier to respond. Let’s suppose you have an amplifier with a 10kHz bandwidth and a 20dB/decade rolloff. So the amplifier response is 20dB down at 100kHz, and 80dB down at 100MHz. Thus, an amplifier that is looking for a one millivolt signal in-band will also respond to a 10 volt interference source at 100MHz – and 10 volt noise is easily achievable with a one watt handheld radio. The bottom line is, if you hit the amplifier hard enough, it will respond.

The second is out-of-band response. When you assume the characteristic 20dB/decade rolloff, you are assuming a very simple RC equivalent circuit. But you really need to factor in the parasitic inductance and capacitance that become significant at higher frequencies. Thus, if you hit a resonance at 150MHz, canceling out the input capacitance, your amplifier will respond to even low-level interference. In practice, you can expect a number of adverse resonances appearing at random frequencies.

The third is non-linearity. As long as the amplifier is operating in the linear region, the out of band response is not troublesome – sooner or later, the amplifier chain will filter them out of the picture. But when the excursion is large enough to drive the circuit into non linearity, you have a demodulator. The output from the demodulator depends, of course, on the nature of the input signal. FM or CW will be demodulated as a DC offset out of the first stage.

As we know, all semiconductor devices are nonlinear – we go to great lengths to make them linear, usually achieved by heavy feedback or other compensation, but even the best amplifier becomes nonlinear when you hit the rails and the amplifier stops working. In practice, non- linearity shows up before then.

We may even install nonlinear elements at the front end, for a variety of reasons (e.g., transient protection). A simple silicon diode or the input base-emitter junction of a bipolar transistor has a fairly sharp knee at about 0.7 volts. If the input interference is less than about 0.7 volts, the diode conducts very little, and usually does not play a significant role. But once the threshold is reached, the diode starts to conduct heavily.

The Solution

The solution is simple in principle – reduce the input interference below the level needed to avoid nonlinear response. Once you reach non linearity, the damage is done, and very difficult to cure.

Amplifiers operating in the audio range are relatively easy to deal with – the key assumption is that the offending interference is high frequency enough that you can filter it without killing the desired signal.

We like RC filters wherever possible. Most amplifiers are high input impedance, allowing a sizable resistance without degrading the input signal. Resistors of any value are easy to come by, and they behave predictably. The big problem with inductors is that their self-resonant frequency limits high-frequency performance, not to mention the cross resonances with the capacitor and the input device capacitance and inductances. Resonances tend to pop up unpredictably, and at much lower frequencies than you might guess. Ferrites are more effective at higher frequencies, and are preferred over inductors. Ferrites are quite temperature sensitive, however – beware if you have a wide operating temperature.

Filtering digital signals is much more of a problem. If the signal bandwidth is too close to the interference frequency, differential mode filtering is not an option and alternate measures will be required. But transients tend to dominate the digital device.

The problem with integrated circuits is that they are not fully isolated – they share a common substrate, which forms a common PNP transistor or, looked at another way, an SCR. This is the cause of substrate latch-up – a condition that can only be cleared by powering down and often smoking the chip. Happily, the chip houses have largely overcome this condition, but it still occurs, notably in transistor arrays.

Recently, we had a case of RFI to an ordinary transistor circuit. We tend to think in terms of integrated circuits, with production lifetimes of a few years, but there is a world of electronics, low tech by most standards, that were designed years ago and are still in production, or at least still in use and, thus, still in need of maintenance.

In this case, the circuit had been designed more than 30 years ago and is still in production – why change it? It works well in its application and has years of history pointing to its success. But 30 years ago, we were living mostly with discrete transistors. Well, the manufacturer had the foresight to buy lots of components while they were still available, but eventually he ran out of parts. He elected to re-layout the board and replace a number of discrete transistors with a transistor array.

Unfortunately, the substrate was connected to one of the transistor emitters, which was also signal common.

Common Mode

Once you have solved the differential mode interference issue, you need to turn to common mode interference. In actuality, interference is largely common mode, which often manifests itself as “ground bounce.” If the input circuit is adequately protected, that part is okay, but currents traveling along circuit ground may cause problems downstream. This is especially acute in two sided circuit boards, where ground impedances are inevitably high. Ground impedance in multilayer circuit boards is low (assuming you haven’t chopped it into islands), so you won’t get significant ground bounce on the board itself – it shows up at connectors downstream.

The good news is that high frequency common mode interference can be filtered with common mode chokes and transformers, with little signal degradation. High-speed serial channels make good use of this fact.

Other Problem Areas on Board

Input errors are relatively easy to identify, and the remedies well defined. But interference can attack other parts of the circuit board, too. Output devices can be attacked, too, but the most common symptom is simply output data errors. Interference that gets by the driver will almost always involve the substrate, and these are hard to track down.

Random digital errors (not apparently associated with input or output) usually are a result of ground bounce or interference on the DC supply. Two sided boards have major ground bounce problems, and such errors can show up almost anywhere on the board. Multilayer boards are pretty much immune from interference due to ground bounce (assuming you haven’t chopped up the ground plane) – if a problem occurs, it will usually be ground bounce at a connector.



Summary

When dealing with RFI, you need to protect the input circuits (output circuits may be vulnerable, too, but this is not the first place to look). Low-level analog circuits are the most vulnerable to RFI, especially including low bandwidth amplifiers. Digital circuits are also vulnerable to RFI, but at a higher level. On the other hand, transient effects are mostly observed with digital circuits. All of these need to be solved by blocking the interference before it reaches the input pins.

Once you have the input circuits protected, you need to make sure the common mode currents are kept within reason. Ground bounce is the major outcome of common mode, often affecting circuits internal to the board and especially at connectors downstream. So keep ground impedances as low as possible.