The Measurement Of Radiated Emissions From Integrated Circuits

Abstract

Newer, larger integrated circuits are becoming significant sources of EMI, due to their higher clock speeds, greater dynamic current consumption, and increased size and complexity. This paper presents an approach for the measurement of radiated emissions potential at the component level, with the objective of quantitatively specifying the level of device emissions that will ensure compliance at the equipment level. Since many integrated circuit radiated emission effects depend upon near-field coupling, the problem is first examined in the light of magnetic and electric field measurements made in the immediate vicinity of the device. The prospect of using far-field measurement methods is also considered, and the relative merits of using both techniques are weighed. R = Radiated I. INTRODUCI~ON C = Conducted Fig. 1. Electromagnetic emission from integrated circuits. The design of telecommunications equipment for EMI compliance is currently hampered by a lack of relevant data on the emissions potential of commercial and custom electronic components. Newer, more complex devices such as 32-bit microprocessors, digital signal processors, and a variety of ASICs (application-specific integrated circuits) have appeared on the scene and represent an increasingly important source of electromagnetic energy. This is due to their higher clock speeds, greater dynamic current consumption, and increased size and complexity. Although these components will not normally fail the radiated emissions limits by themselves, the energy that they generate may excite resonant structures within electronic equipment, causing an EM1 problem. Figure 1 depicts the flow of both conducted and radiated energy from the integrated circuit (IC) into printed circuit board (PCB) and backplane structures, into enclosure cavities and onto connectors and cables. Each one of these elements contributes to the total energy radiated from the equipment. It is evident from this view that limiting the emissions from the device will control the emissions from the equipment. The problem that remains is to quantitatively specify the level of device emissions that will ensure emission compliance at the product level. The purpose of this paper is to propose that devices be characterized in terms of their equivalent magnetic and electric dipole moments, which will provide a measure of the potential for radiated emissions coupling. This is accomplished by representing an IC by a small loop antenna (magnetic dipole) and by a small dipole antenna (electric dipole). The magnetic dipole moment is simply the product of loop area and loop current, and the electric dipoie moment is the product of dipole length and current [I]. Measurement techniques to determine these parameters will be presented in the following sections. II. REASONS FOR REDUCING INTEGRATED ClRCUIT EMISSIONS Direct device radiated emissions are becoming more important in modem high-density high-speed circuitry since the use of ground planes and stripline minimize trace emission, leaving the elevated (2.5 mm typical) wiring within large IC devices as the dominant radiator. Additional benefits of controlling IC radiated emissions are that the device layout and circuit measures necessary to achieve this also tend to reduce package inductance, ground bounce, and conducted noise in general. It follows that the measurement of radiated emissions also provides an indication of some of the principal factors associated with device conducted noise, potentially by means of a single measurement of electric and magnetic fields. This contrasts with the need for conducted noise measurements on hundreds of pins for some of the newer ASICs, CH3169-0/92/0000-0067 $3.00 01992 IEEE 340 suggesting that effort to determine this correlation will be worthwhile. The subject of limits and techniques for direct measurement of conducted noise is not presented in this paper, as it has been widely treated in the literature [2],[3], and because the primary purpose of this effort is to control the radiated emissions contributions of the IC that cannot be effectively mitigated at the circuit pack level. DI. THE NEED FOR BOTH MAGNETIC AND ELECTRIC DIPOLE MOMENTS The EMC Engineer needs to h o w the near-field coupling potential of the IC if significant system coupling along the paths evident in Figure 1 is to be avoided. Near field refers to close-range coupling at distances less than about one sixth of a wavelength. For example, if an IC is mounted on a PCB inside of a metal enclosure, it is desireable to know both the electric and magnetic dipole moments in order to estimate cavity resonance effects. The new high-speed ASICs can emit significant fields up to and beyond one GHz, so this can apply even to small enclosures and structures such as heat-sink assemblies mounted over ground planes. At lower frequencies such as 30 MHz 300MHz, where ASIC-to-cable interface coupling effects may dominate, impedance conditions at the interface may favour magnetic or electric field coupling effects, and again it is desireable to know both characteristics. This suggests that a near-field measurement technique is necessary, at least in the early phases of data gathering. As the correlation between magnetic and electric near fields becomes better understood for different device structures, then one near-field measurementmay suffice. A far-field measurement could also be considered in this situation, but knowledge of near fields would be lost, and again some correlation between far-field and nearfield behaviour would have to emerge. In summary, it was decided to initially pursue near-field magnetic and electric field measurement techniques, and to periodically correlate these measurements with far-field tests of the IC. The near-field tests have the added advantage of being conducted in situ in actual circuit applications, minimizing the need for special fixturing and circuitry to exercise the IC. The development of the method is outlined in sections 4 and 5. Section VI outlines the use of a wideband TEM cell which could be used to obtain equivalent far-field characteristics as described in [4], or dipole moment characteristics as described in [5]. This method would be suitable for use by an IC manufacturer as the IC is tested in isolation inside of the shielded test cell and all extraneous drive circuitry is outside. Iv. THE DETERMINATION OF MAGNETIC DIFQLE MOMENT It is generally not possible or practical to make a direct measurement of the magnetic dipole moment of an IC. If the device consisted of one simple radiating loop, one could measure loop area and loop current, and multiply these quantities to obtain magnetic dipole moment. ICs are very complex sources of radiated fiields, and their magnetic dipole moments are not easily calculated. In very simple cases, a vector sum calculation of dipole moments could be performed if the magnitude and phase of all currents is known[6]. In initial trials, it was decided to infer the magnetic dipole moment from1 magnetic field measurements. A square 2x2 cm shielded loop was fabricated from 0.045 " dia. semirigid coaxial cable. The general approach was to place the probe over the IC so that the plane of the loop was perpendicular to the plane of tlhe IC, as shown in Figure 2. At each frequency of interest, the spectrum analyzer reading in microvolts was recolrded. A calculation was then performed, which determined thle necessary current that must flow in a hypothetical half-loop substituted in place of the IC to p r e duce the Same specmum analyzer readings. For the purposes of the calculation, the half-loop length was made equal to the average distance between opposing IC pins, and the height was made qua l to the height of the device lead-frame above circuit board ground. The half-loop and the measuring loop are in the same plane. The product of the calculated current and twice the area of the half-loop gave the magnetic dipole moment, which was expressed in units of dB (uA-m2). This process was repeated at various probe orientations. Referring to Figure 2, measurements would typically be made in the x-z and y-z planes, at heights h, and h,.