Passive Components:Magnetic Components

Magnetic Components

Integration Issues

It is well known and recognized that magnetic components are to be avoided when designing integrated circuits (ICs) due to their lack of integrability. New developments in the field of magnetic component fabrication are promising devices that can be integrated and miniaturized using monolithic fabrication techniques as opposed to today’s bulk methods. The driving forces for such developments rest in certain applications that benefit or rely on inductive or magnetically coupled devices using ferromagnetic media. Examples of such applications include tuned RF tanks, matching networks, DC–DC power conversion and regulation, network filters, and line isolators/couplers.

Emerging applications requiring more mobility, lower power dissipation, and smaller component and system sizes have been drivers for the development of highly integrated systems and subsystems. To match these trends, it has become necessary to be able to integrate high-quality magnetic devices (i.e., inductors and transformers) with the systems they operate in as opposed to being stand-alone discrete devices. Not only does their discrete nature prevent further miniaturization, but their very nature also hampers improved performance (e.g., speed).

The main features of a monolithic magnetic device are:

1. High values of inductance compared to air core spirals.

2. Enhanced high-frequency performance.

3. Energy storage, DC bias, and power-handling capabilities.

4. Use of ferromagnetic materials as a magnetic core.

5. Photolithographic fabrication of windings and magnetic core.

6. Multilayer mask fabrication for complete magnetic device design.

7. Standard or semistandard IC-processing techniques.

Owing to the use of standard photolithography, etching, and patterning methods for their fabrication, monolithic magnetic devices may appear compatible with IC processes. However, two main characteristics make these devices more suitably fabricated off-line from a mainstream IC process:

1. Coarser design rules. Usually magnetic device designs do not require submicron geometries as demanded by semiconductor designs. This discrepancy means that an expensive submicron process for these components would unnecessarily raise the device cost.

2. Use of ferromagnetic core materials. The use of iron, cobalt, nickel, and their alloys is at the heart of a high-quality magnetic device. Some of these materials are alien and contaminating to semi- conductor cleanrooms. As a result, processing sequences and logistics for full integration with semiconductors is still in the development phases.

With these two major differences, integration of magnetics and semiconductors may require the use of multichip modules or single package multidie cases. Full integration into a single monolithic die requires separate processing procedures using the same substrate [1,2,3,5].

The construction of a monolithic micromagnetic device fabricated on a substrate such as silicon or glass is shown in Figure 6.1. In this diagram, the magnetic layer is sandwiched between upper and lower conductor layers that are connected together by means of an electrically conducting “via.” This structure is referred to as a toroidal device from its discrete counterpart.

Conversely, a dual structure can be made where two magnetic layers sandwich the conductor layer (or layers). This dual structure can be referred to as an EE device since it is derived from the standard discrete “EE” core type. In either case, as required by the operation of any magnetically coupled device, the magnetic flux path in the magnetic film and the current flow in the coil conductor are orthogonal in accordance with Ampere’s circuital law. Interlayer insulation between conductors and the magnetic layer(s) is necessary, both to reduce capacitive effects and to provide a degree of electrical voltage breakdown. These parameters are affected by the choice of insulator systems used in microfabricated circuits due to the differing values of dielectric constants and breakdown voltages used. Some commonly used insulator systems include silicon dioxide and polyimide, each of which has distinctly different processing methods and physical characteristics. Conductor layers for the coil windings can be fabricated using standard aluminum metallization. In some cases, copper conductors are a better choice due to its higher conductivity and hence lower resistive losses. This is especially important if the device is to handle any significant power. The magnetic film layer is a thin film of a chosen magnetic material typically between 1 and 10 mm in thickness. Such materials can be routinely deposited by standard techniques such as sputtering or electrodeposition. The specific method chosen must yield magnetic films with the desired properties, namely permeability, parallel loss resistance, and maximum flux density. These parameters vary with the deposition conditions and techniques, so significant development and optimization has occurred to produce desirable results.

Since the design may call for energy storage and hence gaps in the core, the fabrication method can be modified to incorporate these features. Figure 6.2 shows the geometry of a planar magnetic core with a gap produced by photolithography. In this case, the gap is formed as a result of the artwork generated

for the core design. Figure 6.3 shows the design of a gap using multilayer magnetic films. The energy storage region exists in the insulation between the two magnetic layers.

The fabrication and construction of conductors for the coil usually involves depositing standard interconnect metals (e.g., aluminum) by sputter deposition. The thicknesses are chosen based on the current-carrying capability and the frequency of operation as well as the desired configuration (inductor or transformer). The DC resistance of the conductors must be minimized to reduce DC losses, but the conductor thickness and arrangement must also result in minimal AC losses. This can be accomplished by reduced resistivity (i.e., copper versus aluminum) and by multiple conductor layers to reduce skin and proximity effects at high frequencies.

Designs for Integrated Circuits

Unlike discrete magnetic components, monolithic micromagnetic devices are designed to operate at substantially higher frequencies. Owing to their integrated nature and smaller physical size interconnection and coupling parasitics are lower thus enabling high-frequency response. However, the smaller physical size also places upper limits on characteristics such as inductance, current levels, power levels, and dissipation. With these limits the maximum energy storage, E, is lower. In any inductor, the energy stored due to the current flowing (I) is related to the magnetic fields in the volume of the device by

where L is the transformer or inductor’s inductance in Henries and urI (A) the maximum cuururent carried by the corresponding winding. This is related to the magnetic flux, B , and magnetic field, H, present in the volume of the device. So with a small physical volume one can see from Eq. (6.1) that the energy stored is also small. This limited energy storage capability limits these devices to operate in low power circuits. In order to obtain a high B–H product for more energy storage, a combination of high- and low-permeability regions should be fabricated, i.e., a gap in the high-permeability path is introduced. This gap region helps to maintain a high flux density as well as an appreciable field. The highly permeable region however, while being able to maintain high flux density does not support large magnetic fields due to the fundamental relationship between magnetic field and flux:

In Eq. (6.2), m0 is the permeability of vacuum (4p ´ 10–7 H/m) and mr the relative permeability of the medium in which the magnetic field produces the corresponding magnetic flux density. The size of this gap both determines the energy storage levels and the inductance attainable (lower than the inductance attainable without a gap). In micromagnetic fabrication two approaches may be taken to create this “air gap” region. One is to introduce a planar lithographical feature into the core structure (Figure 6.2) and the other is to rely on multiple magnetic core layers separated by insulating layers (Figure 6.3). The drawback of lithographical gap is the limits imposed by the design rules. In this case the gap may not be any smaller than the minimum design rule, which as was mentioned could be quite coarse. Excessive gap sizes result in very low inductance requiring an increase in number of turns to compensate for this drop. Consequently, electrical losses in these windings increase and also the fabrication becomes more complicated. The drawback of the multiple magnetic core layers is the need to add another level of processing to obtain at least a second (or more) magnetic layer(s). The stack up of these layers and the edge terminations determine the amount of energy storage possible in the device. Unlike the lithographically produced gap, the energy storage in this case is much more difficult to estimate due to the two-dimensional nature of the edge- termination fields in the gap region surrounding the multilayer magnetic cores. In uniform field cases, the energy stored in the volume of the gap can be obtained from Eqs. (6.1) and (6.2) due to the continuity and uniformity of the flux density vector in both the core and gap regions giving

where Vgap is the volume of the gap region in m3. The approximation is valid as long as the gap region carries a uniform flux density and is “magnetically long” compared to the length of the highly permeable core region (i.e., gap length/mr gap � core length/mr mag). Usually this condition can be satisfied with most ferromagnetic materials of choice, but some ferromagnetic materials may have low enough permeabilities to render this approximation invalid. In this event, some energy is stored within the ferromagnetic material and Eq. (6.3) should be modified. Eq. (6.3) is very useful in determining the size of gap needed to support the desired inductance and current levels for the device. For example, if a 250 nH inductor operating at 250 mA of current bias were needed, the gap volume necessary to support these specifications would be about 2 ´ 10–5 mm3, assuming a material with a maximum flux density of 1.0 T. In the planar device of Figure 6.2 with nominal magnetic film dimensions of 2 mm in the normal direction and 200 mm in the planar direction, the required gap width would be about 5 mm. Since the gap in this case is obtained by photolithography, the minimum feature size for this process would need to be 5 mm. If a different material of lower maximum flux density capability of 0.5 T were used instead, the rated current level of 250 mA would have to be downgraded to 62 mA to prevent saturation of the magnetic material. Conversely, the gap length of 5 mm could be increased to 20 mm while maintaining the same current level assuming that adjustments are made to the turns to maintain the desired inductance. Such trade-offs are common, but are more involved because of the interaction of gap size with inductance level and number of turns.

Another aspect of the design is the conductor for coil windings for an inductor or for primary and secondary windings in the case of a transformer. The number of turns is usually selected based on the desired inductance and turns ratio (for a transformer), which are typically circuit design parameters. As is well known, the number of turns around a magnetic core gives rise to an inductance, L, given by

In this relation, N is the number of turns around a magnetic core of cross-sectional area, A (m2) and magnetic path length, l (m). The inductance is reduced by the presence of a gap since this will serve to increase the path length. The choice of conductor thickness is always made in light of the AC losses occurring when conductors carry high-frequency currents. The conductors will experience various cur- rent redistribution effects due to the presence of eddy currents induced by the high-frequency magnetic fields surrounding the conductors. The well-known skin effect is one of such effects. Current will crowd toward the surface of the conductor and flow mainly in a thickness related to the skin depth, d,

hand, a thinner metallization increases the DC resistance, while providing better AC conductor utilization. The optimum situation is somewhere in between these two extreme cases as can be seen from Figure 6.5. The principles presented in this section regarding design issues are at the core of every magnetic component design for ICs. However, many design details especially at elevated frequencies are beyond the scope of this text. It is important to note that many of the limitations on high-frequency designs (100 MHz and higher) are imposed by the properties of the magnetic materials used in the cores of these devices.

Magnetic Core Materials

The most common magnetic materials used for discrete magnetic components operating at higher frequencies are ferrites. This is mainly due to their high resistivity (1–10 W m). Despite a low saturation flux density of ~0.3 T, such a high resistivity makes ferrites suitable for applications up to 1 MHz where hysteresis core losses are still limited. When the frequency is raised over 1 MHz the core losses become excessive thus degrading the quality factor and efficiency of the circuit. Moreover, the permeability of all magnetic materials experiences a roll-off beyond a maximum upper frequency. Commonly used ferrites operate up to 1–2 MHz before permeability roll-off occurs. Higher roll-off frequencies are available, but with higher loss factors. Ferrites, however, are not amenable to IC fabrication since they are produced by a high-temperature sintering process. In addition, their low flux saturation levels would not result in the smallest possible device per unit area. A set of more suitable materials for IC fabrication is the magnetic metal alloys usually derived from iron, cobalt, or nickel. These alloys can be deposited as thin films using IC fabrication techniques such as sputtering or electrodeposition and possess saturation flux levels of 0.8 to as high as 2.0 T. Their main drawback due to their metallic nature is much lower resistivity. Permalloy, a common magnetic alloy (~80% nickel and 20% iron) has a resistivity of 20 ´ 10–8 W m, with a saturation flux density of 1.1 T. Other materials such as sendust (iron–aluminum–silicon) have improved resistivity of 120 ´ 10–8 W m and saturation flux density of 0.95 T.

To overcome the problem of low resistivity, the magnetic layers must be deposited in thin films with limited thickness. Since eddy currents flow in the metallic films at high frequencies, their effect can be greatly reduced by making the film thickness less than a skin depth. In this case, the skin depth in the magnetic film, dm, is given by

In a thin film of permalloy (mr = 2000), the skin depth at 10 MHz is 3 mm. To limit eddy current losses in the film its thickness must be chosen to be <3 mm. This limitation will conflict with the inductance requirement, since a larger inductance requires a thicker magnetic film (see Eq. (6.4)). Such difficulties can be overcome by depositing the magnetic film in multiple layers insulated from one another to restrict eddy current circulation. Such a structure would still provide the overall thickness needed to achieve the specified inductance while limiting the eddy current loss factor. The use of multilayers also allows the reduction of die size due to the buildup of magnetic core cross section (A in Eq. 6.4) in vertical layers rather than by increasing the planar dimensions. As a result, it can be seen that a trade-off exists between number of layers and die size to yield the most economical die cost.

In addition to eddy current losses due to the low magnetic metal resistivity, hysteresis losses occur as in any magnetic material due to the traversing of the nonlinear B–H loop at the frequency of operation. This is due to the loss of energy needed to rotate magnetic domains within the material. This loss is given by

which is the area enclosed by the particular B–H loop demanded by the circuit operation and f the frequency of operation. This loss is expressed in many forms depending on the application. In many cases it is given in the form of a “parallel” or “shunt resistance” and it therefore presents a reduction in impedance to the source as well as a reduction in the overall quality factor of the inductor or transformer. It also represents a finite power loss since this loss is simply V2/Rp (W), where V is the applied voltage (Fig. 6.6).

Notice that Rp is a nonlinear resistance with both frequency and flux level dependencies. It can be specified at a given frequency and flux level and is usually experimentally measured. It can also be extracted from core loss data usually available in the form

In this relation, k, a, and b are constants for the material on hand. This model is useful for circuit simulation purposes thereby avoiding the nonlinear properties of the magnetic material. However, care should be exercised in using such models since with a large enough excitation, the value of the shunt resistor changes.