• 
  Advanced

• EMC Scan
• Notebook
• Tom's Blog
• Events
• 

EMC NOTEBOOK

Spectrum Analyzers as EMI Diagnostic Tools

As EMI consultants, troubleshooting equipment for excessive RF emissions is a common task. As do the test houses, we find the spectrum analyzer a valuable tool. In practice, the spectrum analyzer can yield a lot more information than that needed by the test houses, providing valuable assistance to the troubleshooter.

The test people always test to a specific receiver bandwidth, as specified in the governing standard, whether commercial or military. The frequency span is selected to ensure accurate resolution of the emission at each problem frequency.

When troubleshooting, we find the need to juggle bandwidth and frequency span to get the best feel for the nature of the problem. There are reasons for setting both the bandwidth and frequency span higher or lower than the test people are accustomed to. Below are some examples of how we exercise the spectrum analyzer to assist the diagnosis.

Setting the Span

There are occasions where we want to set the frequency span very wide or very narrow, depending on what we are looking for.

Typically, emissions from clock harmonics will fall off more or less steadily with increasing frequency, but may pop up at some higher frequency. When this occurs, you have almost certainly encountered a resonance, either a cable resonance or an LC (inductive-capacitive) resonance. Generally, you can't eliminate resonances ‚ they will occur, and they will probably occur in your frequency range of interest. As a general rule, the higher you can push the resonant frequency, the less likely it is to cause trouble.

While troubleshooting a resonance, you should set the span to a fairly wide range, making sure you can still see the problem frequency in addition to frequencies well above and below your target frequency. When working on resonances, you want to make sure that you are resolving the problem, not just moving it. If your span is too narrow, you might just move the resonance up or down enough to go off-screen ‚ you might just be moving the problem left or right without solving the real problem. We call this the "water bed effect."

This method works well when you have a fairly low frequency clock source, as the harmonics will be displayed in a row, like crows on a fence ‚ you can easily observe the problem areas. It is not quite so obvious with high frequency clocks, where the harmonics are few and far between. But don't give up hope ‚ you might still catch the resonance by juggling the bandwidth.

Resonances work with broadband as well as narrowband, so look at the ground clutter as well at the narrowband. If you see a bulge in the immediate vicinity of the harmonic, you can suspect a resonance, suitable for further evaluation. If you don't see anything, try stepping up the receiver bandwidth a bit ‚ this will push up the noise floor as well.

By the way, setting a wide span may help you to distinguish between cable and LC resonances. Cable resonances are cyclical in terms of frequency, occuring at half wave intervals. With a wide span, you can identify resonances that repeat, even if they don't all come up to the limit line. LC resonances are individual. While there may be a number of LC resonances, they won't be frequency related.

If you have trouble resolving the narrowband emission from the broadband, you can try stepping down the receiver bandwidth ‚ broadband emissions are proportional to receiver bandwidth, while narrowband emissions are nearly independent of receiver bandwidth. Stepping the bandwidth down by a factor of 3 will drop the broadband about 10dB.

Resolving the Problem Frequency

When observing emissions, you might note the emission amplitude bobbing up and down. The test requirement is to look at the max, but for diagnostic purposes, you need to investigate a little farther. There are two common conditions that can cause this.

The first is that you have two or more harmonic frequencies lying very close together in frequency, of roughly equal magnitude. What happens is that two contributors are nearly equal in frequency but not synchronized ‚ they drift in and out of phase. When in phase, they add to present a maximum. When out of phase, they subtract. The result is a measurement bobbing up and down.

We had a case where we had four clocks adding up ‚ there was a 20MHz, 25MHz, 40MHz and a 50MHz clock, all producing harmonics at 200MHz. In this case, the amplitudes were approximately, so we had a situation where all four harmonics were drifting in and out of phase.

You can get a better look by setting the receiver bandwidth to its minimum and selecting a very narrow frequency span. Depending on your analyzer capability, you may be able to resolve the contributors to individual frequencies. Locking those contributors into place, you can use a second live trace to isolate the dominant contributors, say, with a sniffer probe.

Another condition is when the source is stepping up and down, usually by a considerable amount, maybe 6dB or more. If stepping the bandwidth and span as mentioned above doesn't work, you may have a single source that is switching on and off. This might occur with a serial data bus or channel that is bursting.

In this case, you can set the analyzer to zero frequency span (some analyzers have a specific button for this purpose). With this setting, the analyzer becomes a demodulator. You can set the sweep rate to determine the cycle time ‚ often this will lead you to the source. You may not be able to switch off the source, but at least you know where it is coming from. With the programmable chips, we have encountered cases where the programmer had something running that didn't have to be running. That makes a particularly palatable fix. Just turn off the data stream.

Summary

Spectrum analyzers are very useful for troubleshooting emissions, as well as for calibrated emissions measurements. But don't confine yourself to the same settings as would be used for measuring emissions ‚ you can glean a lot more information if you make use of bandwidth and frequency span settings.

• TOP