• archives
• previous
  issue

Analog Inputs and RF Immunity

• by William D. Kimmel, PE
• and Daryl D. Gerke, PE
• Kimmel Gerke Associates, Ltd.

When dealing with RF (radio frequency) immunity, we find that the sensitive analog input circuits are usually the first to be affected specifically the low level analog inputs that are attached to a remote sensing device. The low bandwidth of analog circuits doesn’t protect inputs from high level RF energy a sensitive amplifier will be overwhelmed by an RF source of sufficient amplitude, even at higher radio frequencies. Let’s take a look at the problem and then look at some solutions.

The Problem

Figure 1 shows RF at the input. The source may be common mode (CM), differential mode (DM), or a combination. Common mode rejection is useful for audio frequencies, but is ineffective at radio frequencies. Hence, both CM and DM need to be addressed.

In any event, the effectiveness is limited as the RF input amplitude is high enough to drive the amplifier into non linearity this is a pronounced effect when the input amplitude is driven past the power rails, but the effects will be significant at much lower levels.

Sound improbable? What if your amplifier is looking for 10 millivolts of signal, and you hit it with, say, 10 volts of RF? Somewhere along the line, non linearity will occur. In the radio world, that is known as a detector or demodulator. If the offending input is AM, the output will be the demodulated audio signal. If the input is FM, the output will be simply a DC offset (Frequency deviation in FM is too small to be discriminated by such a crude detector.

Unless you have control over the non linearity of the device itself, you will need to intercept the RF before it reaches the nonlinear element. If you can shield the cable, your problem is solved. Unfortunately, it is usually not feasible to adequately shield the analog sensors you may have other conflicting factors, such as single point grounding to eliminate the 50/50Hz hum. So you are usually left with filtering.

The Fix

You will need to filter both CM and DM. Often the signal will be differential, but even if you have a common line to circuit ground, you will likely to still need to filter the common line. Figure 2 shows a filter that will block both CM and DM. Starting from the source, the first step is to suppress the common mode. This can be a common mode choke, followed by capacitors to ground. Preferably, the capacitors would go to enclosure ground, but circuit ground or an isolated ground can be used, albeit with less effectiveness.

This is followed by a differential mode filter, shown as an LC filter, the L being individual inductors, not common mode chokes, and with the capacitors going from line to line. This will suppress the DM that may arrive on the cable, and also any CM to DM conversion that may have occurred due to loop areas on the cable or source, and any impedance imbalances on the cable or CM filter. So tolerances on the CM filter elements is not critical.

We have indicated series common mode chokes and inductors, but if you can tolerate the DC drop, resistors can be used instead it’s easier to get a good filter with resistors than inductors, and input impedances of an op-amp are usually high enough that the series resistors are not a problem.

What About Source Non linearity?

Having protected the input amplifier, we now have to turn to the source electronics protecting the input amplifier won’t do a think if rectification occurs at the source. If you have active electronics, the issue is the same as with the input amplifier. More often, you will have passive sensors. Figure 3 shows the filter for a simple nonlinear source the key is, you need to block RF from both the high side and the low side of the signal path. Further, you need to block DM, so you can’t get by with just CM chokes. The amount of filtering needed will need to be determined on a case by case basis, but you need very good filtering here.

Figure 4 shows the treatment for active electronics. Here, there would typically be three wires, one common, one for voltage supply and one for signal all of which are at risk. So you need some series impedance in each leg, followed by capacitance to common in this case, common mode will probably dominate, so you can consider using a three wire common mode choke. You might find it desirable to have some series impedance immediately at the output driver, as well.

Non linearity is not limited to PN junctions, even if they get all the press. There are many nonlinear elements in the world, including rusty junctions (nemesis of Ham radio operators), space charge devices (remember the vacuum tube diode?). But we find rectification occurring at junctions of metals with organic elements and other electrolytes. These are not well documented, but they do exist, and may catch you by surprise. We note with interest that polymer rectifiers are being researched. So if all else fails, take a look at the possibility of nontraditional rectifying sources.

Summary

When dealing with RFI, look first to the analog input devices if you can’t block the RF by shielding the cable, you must turn to filtering, both CM and DM. Fortunately, there is usually a significant frequency gap between the audio signal and the RF, so aggressive filtering is usually feasible.

Don’t forget about non linearities in the sensor source. Active electronics and diodes are the most common situation, but nontraditional rectifiers do exist, and you need to be alert for the possibility.