You Can Troubleshoot EFT Problems
The European EMC Directive dictates that products sold in Europe must comply with several EMC immunity standards, one of which is for electrical fast transient (EFT) pulses. EFT pulses can enter your equipment through I/O cables and can upset the equipment's circuits. So, what can you do to increase your confidence that your equipment will pass compliance testing before you bring it to a test lab? And how can you troubleshoot your equipment if it fails?
The key to troubleshooting EFT (also called burst) problems is to find where current induced by an EFT event enters and exits your equipment. Using a few simple techniques, you can locate EFT current paths and minimize interference with your equipment.
EFT pulses are generated by sparks on AC mains and occur whenever an electrical switch opens or closes or when a motor commutator is running, powering up, or powering down. An EFT event is composed of bursts of pulses, and pulses can occur at intervals of 1 ms or less. Thousands of pulses can reach an EUT as a result of one EFT event, causing stress on the EUT that is more severe than would occur with just a single pulse.
Because it contains many high-frequency components, EFT current in an AC mains cable can couple into your EUT's I/O cables and create current in those cables. If that current enters the EUT, it can cause voltage drops of tens of volts per centimeter in a PCB's ground traces. Such voltage drops can cause bit errors in digital circuits and cause poor signal integrity in analog circuits. The challenge is to find out where and how these voltages interfere with an EUT's circuits and to fix the resulting problems. I've found that once I've measured the effects that EFT voltages have on an EUT, I can correct the problems.
To calculate an EFT-induced voltage drop, you need to measure the EFT current in a cable. That requires an oscilloscope, two matched current probes, and an EFT simulator. You'll need a matched pair of current probes because you need to measure both incoming and outgoing EFT current relative to your EUT. (See "Check Those Current Probes," at the bottom of the article.)
The EFT test waveform described in IEC 1000-4-4 (See Footnote 1 at the bottom of this article) is similar to the human-body-model ESD waveform, but it does not begin with a high-frequency spike. The EFT test pulse has a 5-ns risetime and a 50-ns pulse width.
To calculate EFT voltages, you have to measure the current in the incoming and outgoing cables of the EUT and calculate voltage per centimeter. Once you've made the calculation, you'll know which voltage drops on PCB traces will cause circuit errors. Assume that a cable's inductance is 10 nH/cm; the induced voltage (E) is:
E = L(di/dt)
where L is in H/cm and E is in V/cm.
In Figure 1, the initial change in current is about 1 A in 5 ns. If the current probe has a 5-V transfer impedance, then the 0.5-A/div scale shown in Figure 1 represents a 2.5-V/div oscilloscope setting. During those 5 ns, E is about 2 V/cm. A current that generates more than 1 V/cm can cause problems in many circuits, whereas 50 mV/cm is unlikely to cause a problem in most digital circuits.
To minimize the voltage drops in PCB traces, you must keep EFT currents away from sensitive circuits. Assume that your EUT has two I/O cables. Measure the current entering the EUT with one probe around the EFT input cable. Use the other probe to measure the exiting current in the other cable. If the two amplitudes are equal, you can conclude that all of the entering EFT current exits the EUT through that cable. You can fix this problem by keeping the ground connections of the two cables as close together as possible. The short connection will cause any incoming EFT current flowing in the ground wire to bypass the EUT's circuits and go directly to the exiting cable. That should minimize performance problems.
But if your EUT has more than two cables, the incoming EFT current could divide in the EUT and exit through two or more of them. You'll need to measure the exiting current on as many cables as possible. With multiple cables, you may not be able to keep the ground connections of all the cables close to the cable carrying the entering EFT current. In such a case, filtering the EFT current is an option.
You can filter the incoming EFT current by installing a ferrite core around the cable carrying the incoming current. You can verify the effectiveness of the ferrite core by measuring the outgoing currents and calculating the values of E that appear in the PCB traces. Remember, it's voltage per centimeter of inductive drop that can upset your circuits.
|Figure 1. During the initial 5-ns rise, EFT induced current can reach 1 A.|
Another option is to use shielded cables. Shielded cables can limit the EFT current that enters your EUT because EFT-induced shield currents can be diverted to the chassis. A shielded cable works best when you can roll its shield braid back over the insulator before attaching the cable to a connector. Rolling the braid gives you a 3608 connection to the EUT's metal chassis at the cable's entry point.
You can check the effectiveness of the shield by placing one current probe on the cable outside the EUT and the other probe on the cable just inside the EUT. When making this two-probe measurement, you should connect the shield of the current probe that is inside the EUT to the EUT's chassis. This will prevent EFT-induced current in the probe's cable from entering the EUT's circuits.
If you don't form a 3608 connection around your I/O cable, you'll reduce the shield's effectiveness. Figure 2 shows the output of the current probe outside the enclosure (bottom trace) and the output of the probe inside the enclosure (top trace). In this case, the cable's shield isn't rolled back; instead, it is twisted into an 8-in. pigtail that connects to ground. The shielded cable reduced the EFT current by about 6 dB at the initial peak.
|Figure 2. Using a shielded cable that has an 8-in. connection to ground reduces the incoming EFT current (lower trace) by 6 dB.|
Figure 3 shows that reducing the pigtail's length to 1 in. reduces the EFT current by 20 dB. Finally, Figure 4 shows the result for a good 3608 connection of the shield to the EUT. The performance in Figure 4 is excellent, for it shows a 40-dB reduction in the EFT current. Note the 0.05 A/div vertical scales for the upper traces of Figures 3 and 4.
|Figure 3. Reducing the shield's ground connection to 1 in. cuts incoming EFT current by 20 dB. Note the difference in vertical scales between the two traces.|
|Figure 4. Creating a 3608 shield connection to ground at a cable's entry point cuts incoming EFT current by 40 dB.|
Follow the Trail
Sometimes you have to trace the EFT currents as they travel through a PCB to the exiting cables. To do that, measure the PCB currents with a shielded magnetic probe on your scope's channel 2 while measuring and triggering on the incoming EFT current with channel 1. Once you know the paths of the EFT current, you can redesign the PCB runs to divert the EFT current away from sensitive circuits. T&MW
1. IEC 1000-4-4 (formerly IEC 801-4), Electromagnetic Compatibility (EMC) Part 4: Testing and Measurement Techniques--Section 4: Fast Transients/Burst Immunity Test, 1995. International Electrotechnical Commission, Geneva, Switzerland.
Doug Smithis the manager of EMC development and test at Auspex Systems in Santa Clara, CA. Before joining Auspex early in 1996, Doug was a distinguished member of the technical staff at AT&T Bell Labs. He can be reached by phone at 408-986-2000. E-mail: email@example.com.
Check Those Current Probes
Like any sensor, a current probe isn't perfect. Fortunately, you can correct for a probe's inaccuracy. You can check a probe's error and, when you need two probes, you can measure the difference between each probe's transfer impedance.
To check a probe's error, take a wire and fold it in half. Place the folded end of the wire into the current probe so the folded end is fully inserted but does not project through the other side of the probe (see the figure). Pass a current through the wire. Because the wire is folded, the net current through the wire is zero, and the probe should sense no current. Any current you measure will be an error signal.
A typical probe will have a transfer impedance tolerance of ±2 dB. A probe is usually specified with a transfer impedance in ohms. Therefore, a probe with a transfer impedance of 5 V will have a 5-V output when 1 A of current passes through it. The ±2-dB tolerance translates into a difference of as much as 40% between two probes. Therefore, you need matched probes to perform the measurements in this article.
Check for differences between the probes by inserting the same current-carrying wire through both probes. But make sure the probes are oriented in opposite directions relative to the current flow. Set up an oscilloscope to measure channel 1 plus channel 2. The sum of the two channels will be the difference between the probes. --Doug Smith