EXAMPLES OF SELF-INTERFERENCE

The interference sources are often within the receiving or transmitting device itself. This is likely to be more common in the future, as miniaturization of electronics and the design of systems-on-a-chip (SoCs) become prevalent. These designs will necessarily require that sources of digital and other switching waveforms be physically close to the antennas and other sensitive points of radio receivers and transmitters, exacerbating the EMC problem within the device itself. Because wireless sensor network nodes are designed to be small, highly integrated, and will of necessity incorporate both wireless transceivers and microcontrollers (and likely switching power converters), understanding self-interference is an important ability for the network node designer.

While the fundamental physics involved is often straightforward, self-interference EMC problems can present themselves in many different ways. The following is a list taken from actual experience, of a variety of self-interference EMC problems. As one can see, the affected systems, the interference sources, the symptoms, and the solutions can be quite varied:

Ostensibly identical 2 micron geometry microcomputers used in a radio pager were found to have different desensitization effects depending on which fabrication facility made them. Chips from some facilities (but not others) desensitized VHF pagers (i.e., pagers receiving in the 132–174 MHz band); this was cured by separating the microcomputer ground return to the battery from the voltage multiplier integrated circuit (IC) ground return on the pager's controller circuit board. Discovering this fix took two passes of the circuit board layout and two months' development effort, but it stopped the slow and expensive factory practice of sorting microprocessor ICs (with identical part numbers) by the fabrication plant code stamped on each part. As a cost reduction, these microcomputers were then shrunk from 2 to 1.5-micron geometry due to an improvement in lithography available from the chip vendor. This eliminated desensitization in the VHF band entirely (even for the original board design), but now caused desensitization of pagers in the UHF band (420–512 MHz). These microcomputers employed an integrated phase-locked loop (PLL) synthesizer to generate their system clock; the fix for the UHF desensitization was an inductor in series with the supply pin for the chip's PLL oscillator. This fix, while successful for UHF pagers, introduced desensitization on VHF pagers. Because production would be shut down, different controller circuit boards were then put into production for otherwise identical VHF and UHF pagers, greatly complicating factory flow. An additional 6 months were needed to develop and convert to production a controller board that could be used on all bands.

A radio pager design was within six weeks of its scheduled product introduction date when a desensitization problem involving the liquid crystal display driver was discovered. It was found that, if two pages were sent to the same pager, sensitivity to the second page was about 6 dB less than to the first. This was quickly correlated to display activity: If the display were active (as it was immediately after receiving a page), desensitization occurred. After much effort, it was concluded that the root cause of the problem was the pad drivers on the display driver IC itself — a new chip just entering production. The IC design team (which was not co-located with the pager design team) had not considered desensitization in the IC design, nor was it specified in the IC specification given to them. A new revision of the chip was made, resulting in a program dealy of three months.

A capacitive switching voltage converter caused a significant desensitization problem in the receiver of a consumer electronic product. In addition to the desensitization, signals from the converter entered the voltage-controlled oscillator (VCO) of the receiver's synthesizer, broadening the spectrum of the receiver's local oscillator and resulting in a loss of adjacent channel selectivity. After much experimentation, the problems were solved by improved grounding in areas around the converter, and by the placement of the switching converter external capacitors in a particular orientation so that circuit nodes with fast, but inverse, voltage changes were placed close together.

A service provider received field complaints of poor reception in a particular geographical area. The desired signal was strong; interfering signals present on nearby channels, while also strong, did not have the mathematical relationship necessary to produce receiver intermodulation, nor were they strong enough to cause blocking of the receiver. After much investigation, it was found that the 83-kHz switching frequency of an inductive switching converter was coupling to the receiver's first VCO. This produced a spur on the VCO 83 kHz away from the desired signal that, when mixed with an undesired received signal 75 kHz away, produced an interfering signal in the intermediate-frequency (IF) passband of the receiver that led to the desensitization. The root cause of this desensitization was found to be the coupling between the switching converter inductor and an inductor in the VCO circuit; the two were on separate circuit boards, but when the product was assembled, the two circuit boards were placed next to each other, separating the two inductors by only 2 mm. A new circuit board layout that avoided this proximity effect was required.

Shortly after the market introduction of a pager, it was noticed that the pager's alert (its "beep") was distorted and, in some cases, too short. Because there was a large backlog of orders, this was a crisis at the worst possible time. After much investigation, the source of the problem was found to be the microcomputer clock crystal oscillator circuit. Traces on the multilayer printed circuit board leading to the liquid crystal display (LCD) were laid out underneath the oscillator circuit. The LCD lines were active when a page was received, and the switching energy of these lines was coupled to the clock oscillator, causing clock jitter. Because the microcomputer used this oscillator as the timing reference to synthesize the alert frequency, the tone of the alert was affected. A new layout of the circuit board, with the LCD lines moved away from the sensitive clock oscillator, corrected the problem. This failure resulted in a 3-week delay after shipments were announced. This was an example of a microcomputer desensitizing itself.

A two-way portable radio in the 900-MHz band suffered from a 3-dB desensitization problem on a particular channel. The problem was finally traced to the 968th harmonic of a 972-kHz, 5-V square wave on a digital controller, desensitizing the receiver at 940.896 MHz. One of the advantages one has when dealing with such high harmonic orders is that the problem usually can be cleared by a small capacitor (in this case, 27 pF) placed on the offending source. Small value capacitors may have a minimal loading effect on the signal driver, but may significantly reduce high harmonics.

The preceding "rogue's gallery" of EMC problems is, of course, far from exhaustive. Many other types are possible; nearly all occurrences in the course of routine engineering practice seem to be unique. It is only by understanding the physics involved in the EMC problem, specifically, how interference (noise) is created and coupled into sensitive victim circuits, that trends become apparent and techniques to prevent EMC problems, and cure them when they do arise, can be identified.

EXAMPLES OF SELF-INTERFERENCE

EXAMPLES OF SELF-INTERFERENCE

RhiFaT