|
Searching for Emission Source
| by William D. Kimmel, P.E. |
| and Daryl D. Gerke, P.E. |
|
If we expect to kill the emissions at or near the source, we first need to identify the source. Usually that is straightforward, but sometimes it is elusive.
The first thing to note about emissions is that it takes a periodic or nearly periodic source in order for the receiving device to capture the emission. Transients or any occasional switching action will not show up in the test.
The primary emission sources are the clocks and the switching power supply or regulators and, occasionally, bus noise. Even there, you may have to do some digging, but there are some subtle sources that can catch you by surprise. Let"s start with the common sources, then proceed to some of the sleepers.
Clocks
Higher frequency emissions (say, above 30MHz) are mostly caused by clocks. Typical industrial controls run at 20MHz. Computer and telecom clocks run into the GHz range. Since the frequency is quite high, the problems are in the radiated frequency range.
Often, the clock activity is confined to a single chip, say, a microprocessor, but even there, you may be surprised at what pops up. In a typical microprocessor application, the clock input will be stepped up to a higher frequency for processor timing, or subdivided to a lower frequency, usually for I/O timing, so you will have multiple frequencies generated within the chip. Unless you are synthesizing frequencies (discussed below), you can derive all the relevant frequencies by taking integral multiples and integral fractions of the input clock frequency. Frequency division and multiplication are often generated in powers of two, but that is not a certainty.
Each generated frequency will generate harmonic frequencies. As the clocks are already harmonically related and synchronized, there will be numerous cases of multiple contributors to a harmonic, but the highest frequency generated tends to be the dominant problem. Subdivided frequency emissions go down 6dB every time you divide by two, and typically don"t show up as a problem unless multiplied again with, say, bus activity.
If you are distributing the clocks around the board or even from board to board, note that every time the clock gets to a new chip, you create a new emission source. This situation is common with telecom, where high speed clocks go all over the board. Risetime control, a very effective means of controlling emissions, needs to be applied to each chip that regenerates a clock frequency.
One interesting phenomena occurs when you have multiple asynchronous clocks with common harmonic frequencies. We had one case where the equipment had a 20, 25, 40 and 50MHz clock, all contributing harmonics at 200MHz. Looking at the emission at ordinary receiver settings, the emission level bounces up and down, as the harmonics drift in and out of phase. If you set the instrument to a very narrow span and bandwidth (less than 1kHz), you can usually resolve them into individual contributors--this is sometimes helpful to isolate the dominant culprit.
One other point, there is a general belief that clocks generate mostly odd harmonics--not true. A fully symmetrical wave will generate only odd harmonics, but that only occurs in the textbooks--even minor asymmetries quickly generate even harmonics. In addition, CMOS always generates even harmonics on the power rails. So, with the possible exception of the second and fourth harmonic, you can expect even and odd harmonics in approximately equal amounts.
Power Switching Devices
The most common power switching devices are the switching regulators and converters and, increasingly, motor drives. Power switching devices generate much lower frequencies than clocks, typically 10kHz up to 100kHz, but we have seen them as high as 500kHz. As these have a slower rise time than clocks, the harmonic content is also lower but since the switching voltage and current amplitudes are much higher than logic devices, the harmonic content can get quite high. Mostly, conducted emissions from switchers occur up 30MHz and, hence, are primarily conducted emissions, but we have seen radiated emission problems in excess of 300MHz.
The low frequency problems are almost always due to inadequate filtering immediately at the switching device. Unfortunately, effective filtering usually requires inductive elements, which tend to be bulky and costly.
Frequency Synthesizers
When multiple frequencies are needed, one technique is to synthesize them from a single clock source--chips are readily available for this purpose. As handy as they may be, they do create a problem for the EMI engineer who has to run them down. We never know what frequencies they are generating. Ordinarily, you can look at a schematic or even a populated board to see what the crystal frequencies are. With a chip, you can"t find the frequencies anywhere, except in the designer"s log book. Of course, you eventually find them while emission testing.
As we mentioned, the first step in containing emissions is to identify the problem frequencies, then trace them down to the source. But you may be surprised to find a frequency that is not used by the system. We don"t quite understand why, but we often encounter a frequency generated within the chip that does nothing except to create problems for the EMI engineer.
Parasitic Oscillations
Here again, when looking at an emission profile, you might encounter a frequency that you can"t trace to any clock source. Well, it may be from the synthesizer mentioned above, but it may also be parasitic in nature.
The usual source is a linear circuit, especially op-amps or regulators (both linear and switching). Regulators are basically negative resistance devices. Employing filter elements for EMI may create an oscillator, commonly in the tens of MHZ. The good news is that these oscillations are usually obvious and corrected early in the design stage.
Op-amps, with their high gain, will oscillate with even a small amount of positive feedback, often oscillating in the hundreds of MHZ, and if they are of low amplitude (say the low micro volt range), the condition will be undetected by the designer. Oscillations are usually caused by inadequate decoupling or trace routing.
Energy Efficient Devices
As we know, switching devices burn up most of their power during switching transitions, so there is a strong impetus to switch as fast as possible. The biggest benefit is lower power dissipated in the device, allowing smaller devices to be used. In battery-driven applications, the overall power consumption may be important.
We have already talked about the benefits in power supplies and motor drives, but one that really blind sided us was emissions from a relay. We started with an emission profile and quickly identified the candidate emitters. There was an emitter oscillating at typical SMPS frequencies, but all switchers and regulators had been accounted for.
Ultimately, the emitter was traced to a relay. Now what is more mundane than a relay, you say? Well, if you haven"t looked the relay technology lately, you are in for a surprise--gone are the days when relays consisted of an energizing coil and contacts. Modern relays can do much more and, amazingly, are often plug-replaceable with the original electromechanical relay.
In this case, the relay was designed for low power consumption, an important factor in the application. Now, the minimum current necessary to hold the relay contacts is not conducive to reliable switching. In fact, you need to energize the coil with a generous over-current to make sure the relay switches quickly and reliably. The problem is that the energizing current is way above that needed to sustain the relay. This is usually not a problem--relays are generally used in power applications, and the energy burned up in the relay coils is lost in the round off.
But if you need high efficiency, there is an alternate, the high efficiency relay. It works by applying the entire current to the coil in order to switch quickly. Once the contacts have been closed, the current is reduced to a sustain level. The most energy-efficient way to do this is by applying a lower average current, and this is best accomplished by a low-duty cycle-switching current into the coil. Hence, the emission.
Summary
When chasing down an emission problem, the first step is to identify the emission source, almost inevitably a periodic waveform. The usual suspects are clocks and power switching devices, but there are some sources that are not quite so obvious, including parasitic oscillations, frequency synthesizers and high efficiency control devices. We think the quest for high efficiency is going to create new emission sources, so be alert for new technologies.
|
