A portable oscilloscope and keen troubleshooting instincts help solve an EMI problem.
Let's take a look at a conducted Electromagnetic Interference (EMI) problem that was recently identified and resolved at an industrial process control plant. Here, the victim circuit was a signal transported on a coaxial cable. The signal was a low frequency square-wave whose period varied as a function of the sending end transducer's input.
Details of problem
The corrupted signal could be seen on the input to the cable end's receiver plug-in card by using a portable oscilloscope and a 10x voltage probe on Channel-B. Fig. 1 shows the badly corrupted square wave signal, which at the time of measurement was about 1.8V peak-to-peak, with a typical period of about 2ms (500 Hz).
The victim coaxial cable was observed to be 52 ohm, RG-58/U cable approximately 50 ft long. This cable ran from a transducer location outside of a control room to within the control room via an overhead PVC conduit. The transducer end was provided with a 2-wire plus ground, 120VAC branch circuit and a Switch-Mode Power Supply (SMPS) to convert the AC to the required logic level, regulated DC voltages needed to drive the electronics at that end. The control room end of the cable simply went into the back of a proprietary interface card equipped with a BNC coaxial fitting on the back of what appeared to be a PC-style workstation/controller.
Other cables interfaced this workstation/controller to other equipment in the control room. The control room was powered from its own main panel and transformer, (e.g., a solidly grounded separately derived system) which also supplied all the power to the site's other electronics, including the SMPS at the transducer's remote location.
An initial measurement was made by removing the coaxial cable from the PC end interface and attaching a BNC style TEE fining in its place. As shown in Fig. 2 (on page 20), this allowed a "tap" to be made across a noninductive, 50 ohm resistor that was previously soldered in place across the signal pin of one port on the TEE to its metal shield/enclosure. As such, the signal could be viewed with some (but not total) assurance.
The next step was to see if any significant common-mode (CM) current, which could be correlated in some way to the corruption on the signal, existed on the victim coaxial cable itself. This was done by connecting a current transformer (CT) to the portable oscilloscope and then clamping it around the victim coaxial cable. Fig. 3 (on page 20) shows the result of this test. The CM current's waveform somehow looked familiar; but what was it?
Finally the answer came: It looked just like the typical signature current for the AC input to an SMPS power supply (or any full-wave rectifier and large filter capacitor connected directly across the AC line). But, the most interesting question was this: Did this CM current have a timing relationship to the EMI on the victim cable's signal?
To get the answer to the last question, the next step involved the voltage probe and the TEE connection on the portable oscilloscope. This time, however, a two-channel setup was in order so that both signals (CM current and the data signal's voltage) could be viewed simultaneously. This was done and the result is shown in Fig. 4 (on page 20), where a clear relationship between the EMI and the CM current on the victim cable can be seen to exist.
Next, the TEE connection was disconnected from the rear of the PC, but left attached to the B-Channel voltage input on the portable oscilloscope. We could see what the result would be with the CM current's path on the victim coaxial cable's path completely opened at the receive end. The result: A "clean" square-wave signal as shown in Fig. 5 (on page 20). Now, how was the CM current getting onto the victim coaxial cable to begin with?
Another trip out to the transducer end of the victim cable was in order so as to check out the conditions at that end (it now looked as though the origin of the EMI problem might exist there). Upon arriving, we disconnected the SMPS from its branch circuit. Then, with the TEE fitting reconnected to the PC in the control room, another CM current measurement was made on the victim coaxial cable. The result: No appreciable CM current on the cable with the SMPS disconnected.
One more test was made by placing the CT around the AC power supply cord for the SMPS to see if the CM current could be measured there. It was, and from this a tentative conclusion could be drawn: The SMPS was probably the aggressor source of the EMI.
The time had now come to pull the SMPS out and to check it on the bench to see what could be found. The SMPS was powered up on the test bench using a homemade test attachment, as shown in Fig. 6 (on page 22). The shown arrangement allowed the placement of a CT around any one or combination of the three AC line conductors used in the power cord. The first choice: The equipment (safety) grounding conductor. The CT was placed onto the "green wire" and the SMPS was switched on. Except for amplitude, the portable oscilloscope clearly showed a signature current waveform of the same shape and timing as had been observed on the victim coaxial cable. This was clearly not a typical small leakage current as permitted by U.L. under the SMPS's product safety listing. Something was wrong inside the power supply, but what?
A simple continuity tester was used to check from each of the SMPS' line cord pins (unplugged) to the metal frame/enclosure of the SMPS. This showed good continuity for the green wire, an open on the ungrounded side of the line (e.g., the "black wire"), and a short on the grounded side of the line (e.g., the "white wire" or neutral).
Since the input to the SMPS should be virtually symmetrical to "ground" from either line pin, this was definitely not right. A closer look inside of the SMPS revealed a shorted AC line filter capacitor that was connected between "ground" and the neutral pin, as shown in Fig. 7 (on page 26). Thus, some portion of the SMPS' neutral current was dividing between the neutral and equipment grounding paths, which consisted of the green wire and, unfortunately, the grounded copper braid of the victim coaxial cable (frame/enclosure terminated at the PC end in the control room). The control room PC's input power cord then supplied the remaining required connection back into the electrical system's neutral on the transformer.
The SMPS was replaced rather than repaired. When power was reapplied to the equipment, no problems were seen to exist and the system again ran normally. Why the line filter capacitor failed in the SMPS is anyone's guess at this point, since no one at the facility could recall any recent AC system or other ground faults. In addition, no electrical storms had occurred for several months prior to the problem being first observed. Thus, to this date, we are unsure as to what to do to prevent a reoccurrence of the problem, or whether we should be further concerned or unconcerned about it.
The present attitude of the client is that if it fails again, then a more thorough examination will need to be made to uncover the cause and to create a remedy for it.
As a side note, we did note that the entire associated AC system was not equipped at any point with a transient voltage surge suppressor (TVSS). As such, we suggested that adding TVSS protection (e.g., at least at branch circuit level) might be worthwhile at this site, since transient high-voltages can cause dielectric breakdown in the AC line filter capacitors (and other parts) in an electronic system.
Some of you may be wondering how the SMPS' ground-fault current, which was flowing on the coaxial cable's braid, could combine with the desired signal and corrupt it, since this should not normally happen when coaxial cable is used for signal transport. The simple answer is in the opening paragraph of this article: the desired signal is of a low-frequency nature. In fact, it's such a low-frequency signal that it flows on the braid and in the center conductor with very little skin-effect. That's the reason the externally applied EMI CM current (e.g., from the ground fault) could do what it did. Without skin-effect, both the desired and undesired currents were free to flow in the entire available cross-sectional area of the braid, nearly like DC. Thus, the two currents shared the same conductor volume and were algebraically added together to produce a longitudinal (e.g., lengthwise) voltage drop on the cable due to the braid's impedance. This action then produced the combined signal voltage in the differential mode (DM) at the receive end of the cable.
However, it should be noted that if the desired signal had been of a high frequency nature, it would have flowed entirely on the inner surface of the braid, nearest the center conductor; this would have left a negligible amount of shared conductor volume for the two currents to "mix" together.
Next month, we'll discuss relative signal levels and the all important decibel (dB), which is a really useful way to express differences in signal levels for all sorts of reasons in the electrical field. In a future column, we'll discuss how signals are transported and interfered with on coaxial cables in greater detail than above.