Notice that the source and body are connected together by the metalized contact at the source. Also notice that there is a parasitic diode due to the P-type material of the body and the N-type material of the drain. This parasitic diode is reverse biased normally because the drain is more positive than the body (and source), so does not need to be considered in all applications. Note that the body diode generally has a long reverse recovery time; in synchronous switching applications where the body diode may conduct, an ultrafast diode is sometimes connected in parallel to prevent or reduce the body diode current.

To create a conducting channel in the body of the MOSFET requires a certain amount of gate potential. MOSFETs are specified with a certain gate threshold voltage, usually at the point where the drain current reaches 1 mA, but this varies between manufacturers. As the gate–body isolation is a dielectric, gate–source and gate–drain capacitance values are usually found in the datasheet.

Typical gate thresholds are in the range 4–7 V, however a number of “logic level” devices are now available. A “logic level” device is defined as one that switches fully on at V_gs equal to 5 V; this means that the gate threshold is typically about 2 V. So-called “standard devices” are defined as being fully switched on at V_gs equal to 10 V. A logic level device can also be operated with V_gs equal to 10 V or higher, in which case the on-resistance is lower. Many logic level devices have a higher gate capacitance compared to standard devices, for a comparable saturation current rating.

Fig. 11.4A shows a simple MOSFET circuit with parasitic capacitance C_gd and C_gs; Fig. 11.4B shows a graph of gate voltage versus gate charge. This helps to understand the switch operation when the MOSFET is driven from a current source because the rate of charge is constant (Q_g = current × time) and thus the horizontal axis represents time. The gate voltage initially rises quickly as C_gs is charged. Once the gate threshold voltage, V_gs(th), is reached, the MOSFET starts to turn on; this causes the drain voltage to fall forcing the gate drive current into capacitor C_gd instead of C_gs, so the gate voltage rises very slowly while the drain voltage falls.

Once the MOSFET is turned on and the drain voltage is close to 0 V, the capacitors C_gd and C_gs are both charged at a steady rate. Note that both C_gd and C_gs are capacitors formed across a reverse bias P–N junction and so will have lower capacitance when the reverse bias voltage is high. Now that the drain voltage is close to 0 V, the capacitances of both C_gs and C_gd have increased because of the small reverse bias voltage, and the charging rate is lower, thus the slope of the curve is now less steep than during the initial charge (below V_gs(th)).

Note that a MOSFET gate is normally driven from a resistive or current-limited source. In fact designers often add a resistor in series with the gate, for two reasons: (1) to reduce EMI, by slowing down the turn-on speed; and (2) to prevent high frequency ringing at the MOSFET drain due to resonance of the drain–gate capacitance and the load inductance. The load inductance may be from the inductor in a buck converter, or a transformer primary in a fly-back converter.

MOSFETs have two current ratings—peak current and continuous current. Continuous current ratings depend on the on-resistance of the MOSFET and are based purely on thermal considerations. Peak current ratings are the maximum current that is able to flow. When designing a switching LED driver circuit, the circuit current is pulsed and so the peak current rating is important. However, note that this current is normally quoted at 25°C; at 100°C the peak current is about half this value. As a rule of thumb, always use a MOSFET that has a peak current rating that is 3 times the value needed in the application.

Matched bipolar transistors can be very useful, particularly in current mirror circuits. A current mirror is one where two or more branches carry identical currents; the current in one branch depends on the current in another, hence the “mirror.” Transistors do not have to be matched to make a current mirror. Transistors of the same type have very similar characteristics, so by adding a low value resistor between the emitter and ground any variation in the base–emitter voltage (V_be) is negligible; see Fig. 11.5.

Schottky diodes have the lowest forward voltage drop and the shortest reverse recovery time, but they are more expensive than standard diodes and generally have a limited reverse breakdown voltage range. Instead of a P- and N-type semiconductor junction, the Schottky diode has an N-type semiconductor and metal junction. Reverse leakage is higher than in most P–N junction diodes. They are used for many applications, including reverse polarity protection and as flywheel diodes in low voltage switching circuits. Note that the forward voltage drop across a Schottky junction tends to increase with diode voltage rating, so use the lowest voltage rating suitable to keep the conduction losses to a minimum.

Diodes are sometimes labeled by their reverse recovery time. When the voltage across a diode is suddenly reversed, an initial current flow will occur in the reverse direction. Reverse recovery time (T_rr) is the time taken to stop conducting when the diode is reverse biased. The labels, fast, ultrafast, and hyperfast are sometimes given. A standard rectifier diode like 1N4007 has a typical reverse recovery time of 30 μs, but an ultrafast version UF4007 has T_rr = 75 ns, which is about 500 times faster. More recent devices are much faster, for example, the STTH1R06 is a 600 V, 1 A rectifier with T_rr ∼ 30 ns.

There are now high voltage Schottky diodes available from a few suppliers, notably Cree and ST Microelectronics. These are known as silicon carbide (SiC) diodes and are rated from 600 V to above 1 kV. These type of diodes are very useful in circuits like power factor correction boost circuits and high voltage, high current, buck circuits, where the short reverse recovery time prevent high switching losses. They tend to be more expensive than ultrafast junction diodes.

Shorter reverse recovery times reduce switching losses. This is because the reverse current often flows through the MOSFET switch when the voltage across the MOSFET is high, so a shorter time gives lower losses. However, a “snappy” diode can sometimes generate radio interference (EMI), due to the fast turn-off causing high frequency ringing. In some applications a “soft-recovery” diode should be used, where the turn off speed in the reverse-biased condition is fast but at a controlled rate of change. If slower diodes must be used, slowing the MOSFET switching speed by adding a resistor in series with the gate may be necessary to prevent the diode from overheating.

In fly-back power supplies, a series connected RC snubber circuit is placed across the primary winding to prevent very high voltages when the MOSFET switch turns off. Some snubber circuits use a medium speed diode in series with a resistor, so that the diode blocks current during switch turn-on, but is conducting in both directions for a period of about 250 ns after switch turn-off. This allows any ringing current to flow through the resistor and thus decay quickly. Alternatively, a fast diode connected in series with a high voltage Zener diode (the two anodes connected together) can be used to clamp ringing.

Zener diodes behave like regular diodes in the forward conducting direction, but breakdown and conduct at a defined voltage in the reverse direction. Low voltage Zener diodes rated below 6 V have a soft knee in their current versus voltage graph; the conduction increases gradually. High voltage Zener diodes (avalanche diodes) rated above about 6 V have a sharp knee and conduction increases very rapidly. Zener diodes can exhibit some noise when breaking down and are often used with a small capacitor in parallel to reduce this effect.

Transorb suppressors are like Zener diodes but are designed to handle high current peaks. Transorbs can be unidirectional or bidirectional and rated from low voltage ∼5 V up to several hundred volts. A Transorb designed for 275 V AC operation will limit the peak surge voltage to below 600 V, even at high transient current levels.

A VDR has high resistance at low voltage and low resistance at high voltage. Thus conduction increases gradually as the voltage across it increases. A VDR can absorb high surge energy; the devices are often rated in Joules rather than Watts because the surge energy is short lived. A VDR rated at 275 V AC will breakdown and limit the voltage to about 710 V at high transient current levels.

A capacitor’s behavior is not ideal. Every capacitor will have some series inductance; this is due to the plate conductors and the lead wires attached to them. This self-inductance can be a problem at frequencies close to the self-resonant frequency and above. Each capacitor will also have series resistance, due to both the conductors and the dielectric of the insulator, this is known as equivalent series resistance (ESR). The ESR will create losses. An equivalent circuit for a capacitor is shown in Fig. 11.6.

In an LED driver, the key function of a capacitor (symbol C) is energy storage. There are two types of storage, slow storage and fast storage. Slow storage is required across the DC supply, to hold up the voltage when powered from a low frequency AC supply. Sometime direct connection is possible (for very low power lamps), but more often a power factor correction stage produces a high voltage DC energy that must be stored. The purpose of this energy storage is to supply power to the LED driver between the peaks of the AC voltage, which is twice every cycle. The AC frequency is typically 50–60 Hz, although 400 Hz is used in some aircraft, so the capacitor must supply energy and hold up the supply for as long as 10 ms.

For slow storage, an aluminum electrolytic capacitor is often used because it has a high energy storage density (they take up less space for an equivalent amount of storage, compared to other dielectric types). These capacitors are made using aluminum foil with a wet dielectric material. Due to this construction, they cannot be used in a high temperature environment for long periods; the dielectric dries out and the capacitor eventually fails.

Good quality electrolytic capacitors may have a lifetime of 10,000 h at 105°C, for example. But electrolytic capacitors do not suddenly fail, they decay slowly. Lifetime is defined by the manufacturer: some use the value of ESR, where lifetime is when the ESR has doubled from its initial value. Doubling of the ESR means that power dissipation will be doubled for a certain level of ripple current, so some manufacturers measure the maximum ripple current for a given temperature rise in the capacitor. An important “rule of thumb” is that lifetime doubles for every 10°C drop in ambient temperature, so operation at 85°C, giving a 20°C drop, would give 4 times the lifetime (40,000 h for a 10,000 h 105°C rated capacitor). This means 5 years lifetime.

Fast energy storage is required in switching driver circuits, where the switching frequency is often in the range 50–500 kHz. The energy only has to be stored for a short time, as short as a few microseconds, so the main characteristic of the capacitor for this function is to have the ability to store and discharge energy quickly. This means low self-inductance (high self-resonant frequency). Surface-mount components generally have lower self-inductance because they have no added lead inductance. Typically, the capacitors used for fast energy storage are ceramic or plastic film types.

Ceramic and mica capacitors are made using flat dielectric sheets; the simplest construction uses just one insulating layer with a conducting plate on either side. Mica capacitors are very rarely used, but ceramic are fairly common. Higher valued devices use several insulating layers with interleaving layers of metal film. The metal film layers are bonded alternatively to side A, side B, side A, side B, etc.

Plastic film capacitors, such as polyester, polypropylene, polycarbonate, and so on, use two layers of metalized plastic film. One form of construction is identical to that of ceramic capacitors, where flat sheets of metalized film are used. This type of construction is often found in surface-mount polyester capacitors.

Another form of construction for plastic film capacitors uses rolled films. Two metalized layers are placed one above the other and then rolled, so that the two conductors spiral around each other with insulating layers in between. The films are laterally offset from one another so that the conductor of “side A” protrudes from one side, and the conductor of “side B” protrudes from the other side (this technique is sometimes known as extended foil). It is then relatively easy to bond lead-wires to the ends of the resulting cylindrical body. The rolled form of construction provides a metal film around the body of the capacitor; this can be connected to earth or the “earthy” side of a circuit to reduce external electric field pickup. Connecting the outer foil to earth also helps to reduce EMI. The outer foil is marked on the outer case of some film capacitors.

Generally, ESR and self-inductance is more of a problem with aluminum or tantalum electrolytic capacitors. These types of capacitors are normally used to decouple power supplies. Digital circuit designers have long since become accustomed to connecting 10 nF ceramic capacitors across tantalum devices used for power supply decoupling. This is because the higher value tantalum capacitor absorbs low frequency transient currents, while the ceramic absorbs the high frequency transient currents.

Dissipation factor (DF) and loss tangent are terms used to describe the effect of ESR. The value of DF is given by the equation:

where X_c is the capacitor’s reactance at some specific frequency. This is the tangent of the angle between the reactance vector X_c, and the impedance vector (X_c + ESR), where the ESR vector is at right angles to the reactance vector.

One of the most notable problems with capacitors is self-resonance. Self-resonance occurs due to the parasitic inductance mentioned earlier. Consider the self-resonant frequency of capacitors, of various dielectrics, having a lead length of 2.5 mm (or 0.1 in.): a 10 nF disc ceramic has a self-resonance of about 20 MHz; the same value of polyester or polycarbonate capacitor also has a self-resonance of about 20 MHz.

A rough idea of the self-resonant frequency can be found by calculating the inductance of a component lead. For example, a 0.5-mm diameter lead that is 5-mm long (2.5 mm for each end of the component) has an inductance of 2.94 nH in free space. When combined with a 1-nF capacitor, the self-resonant frequency is calculated to be about 93 MHz. Replacing the 1-nF capacitor in previous calculations with a 10-nF capacitor, results in the self-resonant frequency falling to 29 MHz.

Earlier, I wrote that the self-resonant frequency of a 10-nF capacitor with 2.5-mm leads was about 20 MHz, not 29 MHz. The reason for the discrepancy between the calculated frequency and the actual frequency is that inductance in the plates was not taken into account. Adding the inductance of the plates gives a lower self-resonant frequency. As the value of the capacitor increases, the inductance of its plates also increases and so does the discrepancy between the calculated and the actual self-resonant frequency.

For small value capacitors of less than 1 nF the self-resonant frequency can be approximately calculated by the following equations.

where L is the lead inductance. For a wire in free space, $L=0.0002b\left\{\left[\ln \left(\frac{2b}{a}\right)\right]-0.75\right\}\mu \text{H}$, where “a” equals the lead radius, b equals the lead length. All dimensions are in millimeters (mm) and the inductance is in H.

Using the formulae, if a = 0.25 mm (0.5-mm diameter) and b = 5 mm (2.5 mm each leg), the inductance is 2.94 × 10−3 μH. This is 2.94 nH. When substituted into the frequency equation, with a 1-nF capacitor, the self-resonant frequency is calculated to be 92.8 MHz.

Surface-mount capacitors are in common use now because of their small size. In the past they were often used for high frequency circuits because there is no lead inductance to worry about. This reduction in inductance has benefits for switching power supplies too; where fast pulse rise and fall times are needed. The most popular type of surface-mount capacitor is the multilayer ceramic; it has conducting plates that are planar and interleaved; they have very little parasitic inductance. Some conventional leaded ceramic capacitors are surface-mount devices with wire leads attached. They are usually dipped in epoxy resin or similar before having their capacitance value and voltage rating marked on the outside.

Ceramic capacitors generally have a temperature coefficient that is zero or negative. The terms NPO or COG are used to describe ceramic capacitors with a zero temperature coefficient (NPO = negative positive zero). Other ceramic dielectrics are described by the temperature coefficient; N750 describes a dielectric that has a negative temperature coefficient of −750 ppm/C. More exotic dielectrics are X7R, X5R, and Y5U, which have a higher dielectric coefficient and are used to make capacitors with high capacitance values.

The X7R and Y5U capacitors have a wide tolerance on the component value. The first character defines the minimum temperature, so X = −55°C and Y = −30°C, for example. The second character defines the upper temperature limit, so 5 = 85°C and 7 = 125°C. The third character defines the capacitance change over temperature, so R = ±15% and U = +22%, −56%. Perhaps Y5U should be avoided.

Even if we keep to X7R and X5R capacitors, we have another characteristic to consider. The capacitance of a ceramic capacitor changes with applied voltage. Also, the smaller the package size, the worst this effect becomes. For example, a 0603 capacitor may have its original (0 V bias) capacitance value reduced to less than half with just 6 V bias, so a 2.2-μF capacitor looks like 1 μF. This is illustrated in Fig. 11.7.

This reduction in capacitance with applied voltage is important in many designs. Consider a decoupling capacitor used for the internal power supply of an LED driver, it must supply high current pulses at high frequency to drive the gate of the power switching MOSFET. Calculations may have shown that a 1-μF capacitor would be sufficient, except that we discover that the effective capacitance has been reduced to 400 nF when a DC bias was applied. A capacitor of the same value, but in a much larger 1210 package, would show perhaps 10% drop in capacitance with 6 V bias. Unfortunately, size and cost can be critical in some applications and a 1210 size package may not be an option.

Apart from NPO/COG capacitors, ceramic capacitors exhibit a piezoelectric effect. A high voltage AC signal can generate acoustic noise. The acoustic output increases with physical size, so a surface-mount 1206 size capacitor will generate more noise than a 0805 capacitor in the same circuit. The piezoelectric effect will also cause the capacitance value to change with applied voltage, as described earlier. In applications like alarm clocks and theater lighting, any noise is a big problem because of the very quiet environment they operate in, so alternative capacitor types are recommended.

Polystyrene and polypropylene capacitors have a negative temperature coefficient that fortunately closely matches the positive temperature coefficient of a ferrite-cored inductor. They are thus ideal for making LC filters. Unfortunately, with these dielectrics, capacitors tend to be physically large for a given capacitance value.

Polyester and polycarbonate capacitors are commonly used in LED driver circuits. Polyester capacitors have the worst performance in that they have a poor power factor (high ESR) and a poor (and positive) temperature coefficient. But polyester capacitors are popular because they have a high capacitance density (high capacitance value devices are small) and lower cost. Polycarbonate capacitors have a better power factor and a slightly positive temperature coefficient. Another useful feature of polycarbonate capacitors is that they are “self-healing”; in the event of an insulation breakdown due to overvoltage stress, the device will return to its nonconducting state, rather than become short circuit.

Capacitors used across the AC mains supply must be rated X2. For universal AC input, 275 V AC X2 rating is normally used. These capacitors are available in polyester and polypropylene, and 100 nF is a typical value found across the supply connections. This capacitor reduces EMI emissions and absorbs fast transient surges from the mains supply. In a typical application, a VDR (or varistor) is connected in parallel.

In some applications, a standby power supply is created by having a series capacitor (usually X2 rated) and a shunt Zener diode (or an active shunt like the Microchip SR10). One problem with series capacitors is that the capacitance value falls over time. Corona discharge during overvoltage conditions (line borne surges) causes the metalization layers to be burnt away, thus reducing the capacitance value. The current available for the power supply will reduce over time. An example that I have seen is an electric shower control unit that failed because the series-connected 100-nF capacitor, used in its power supply, had its capacitance value reduced to 25 nF after about 3 years of service. It is possible to obtain “low corona” capacitors, which have a longer lifetime in such circuits, but they tend to cost more.

Capacitors from each AC power line to ground (earth) are sometimes used and must be 250 V AC Y2 rated. These capacitors typically have a dielectric of ceramic, polyester, or polypropylene. Capacitance values are readily available in the range 1–47 nF. A value of 2.2 nF is commonly found in power supply designs.

Inductors (symbol L) are used to store energy in switching LED driver circuits. A length of wire creates inductance, but winding insulated wire into a coil can magnify this effect because the magnetic field produced by a wire then couples to adjacent wire. With intercoil coupling, the total inductance is proportional to the number of turns squared. The wire used in multiturn coils is normally soft copper covered with a thin plastic film for insulation; this is sometimes referred to as enameled copper wire (ECW).

Although a simple coil creates inductance, if a magnetic material is placed within the coil the inductance increases considerably. The coil can be wound around a short ferrite or iron-dust rod to increase their inductance, but with this type of core a magnetic field will radiate and may cause interference (EMI). The advantage of this type of inductor is that the saturation current level is very high, considering the inductor size. A typical application for this type of inductor is in the power filter at the input of an LED driver. Many low value inductors have the same appearance as wire-ended resistors, with colored bands marking their inductance value.

Alternatively, the coil can be wound around a toroidal (doughnut) shaped ferrite or iron-dust core. The toroidal shape keeps the magnetic field contained. Some toroidal materials have a distributed air gap, in which case the saturation current is very high. Toroidal inductor winding is not easy, requiring specialized winding equipment, hence these types of inductors can be more expensive than their bobbin-wound counterparts.

Shielded bobbin cores are popular. The coil is wound on a bobbin, within a closed ferrite material. These are of low cost and small physical size. Many can be surface mounted, which make printed circuit board (PCB) assembly easier. The central ferrite core inside the coil often has an air gap to increase the saturation current rating, although this reduces the inductance value.

An inductor’s behavior is to oppose any change in the current flowing through it. This is because the energy stored in an inductor is given by $E=\frac{1}{2}L{I}^2$. To change the current instantly through an inductor would take infinite power. If we ignore physical imperfections due to the construction of an inductor, when a voltage is applied across it the current will increase linearly. If a load is then applied across the inductor, the current falls linearly. If we alternately switch the voltage source and the load across the inductor, the current will rise and fall, but remain fairly constant.

Inductors can be used to filter the power supply lines in switching LED drivers. Due to their energy storage characteristics, they tend to oppose any change in current, so they present high impedance to unwanted interference. Combined with capacitors that are low impedance to unwanted interference, the resulting “T” or “PI” filter considerably reduces the amplitude of high frequency signals.

Inductors can be a source of many problems. High value inductors are bulky. This is because they are usually made up from tens or hundreds of turns of ECW that is wound on a ferrite core. The windings capacitively couple to each other, which effectively introduces a parallel capacitor across the coil. This capacitance causes switching losses in power supplies, or poor filtering in supply input filters. Above the self-resonant frequency, the impedance of the inductor falls due to the dominating capacitive reactance.

Inductors also possess some series resistance due to the intrinsic resistance of the copper wire used. This resistance will cause losses in the power supply and thus limit the efficiency. Heating effects due to this resistance can cause problems. Choosing an inductor for the correct inductance value, without considering the ESR and self-resonant frequency will give poor results.

Magnetizing (core) losses are also present and are due to the energy required to make the magnetic fields in the core to align with each other. In a switching circuit these losses are continuous and can cause core heating. These losses increase rapidly if the magnetization is forced to operate outside its linear region. The presence of an air gap in inductor and transformer cores makes them suitable for high magnetic saturation levels. Transformer cores that have no air gap are prone to saturate easily.

The saturation current (I_sat) quoted by manufacturers is usually at the point where the inductance drops by 10%. If the inductor is expected to handle discontinuous switching current, where the current drops to or near zero each switching cycle, the peak current should be kept well below the saturation level (I suggest I_max = 0.5 ⋅ I_sat, but preferably 0.25 ⋅ I_sat). This is because magnetization losses will cause core heating. Take care, because sometimes the current rating given by manufacturers is the DC current that causes a certain amount of heating, due to the winding resistance; the saturation current could be a lower current value. Some manufacturers quote a saturation current at the point where the inductance has fallen to 60% of the zero current value!

Sometimes an inductor data sheet will give a “Q” value at a certain frequency. This is the voltage or current magnification value in a tuned circuit. It indicates the ESR of an inductor, $Q=\frac{\omega L}{R}$, which is more accurate than the DC resistance measurement. This is because of the “skin effect.”

The “skin effect” raises the resistance of wire at high frequencies. This effect occurs when AC current flows in the wire, producing a magnetic field concentrated at the center of the wire. This magnetic field forces the electrons to travel down the outside surface (hence “skin” effect). The effect gets stronger as the AC frequency increases. This can be a serious problem for inductors working at a few hundred kilohertz, but can be alleviated by the use of multiple stands of insulated copper wire, twisted together.

Litz wire is made from a number of ECW strands, with an overall covering of cotton braid. This is the type of wire used to make ferrite rod antennas for radios working in the low and medium frequency range (LF and MF). This wire has a lower skin effect because the current is shared down each of the strands; the surface area of all the strands combined is considerably larger than the equivalent diameter of solid copper wire.

More recently, copper wire suppliers have been producing flat wire, which is a thin strip rather than a round wire. Flat wire has a much larger surface area than a round wire of similar cross-section and thus exhibits less skin effect at high frequency. Flat wires are suitable for use in inductors and transformers, in high frequency switching power supplies

Off-the-shelf transformers are available with double or multiple windings, with or without an air gap in the magnetic core. Fly-back power supplies, including isolated LED driver circuits, use gapped cores; the air gap allows high magnetic flux density within the core—the energy is stored and then released. A forward converter is a popular power supply topology, which uses ungapped cores because magnetic energy is not stored in the core—it is immediately transferred to the secondary winding. Forward converter power supplies are rarely used in LED driving.

Transformers with multiple windings are used to create a step-up or step-down primary-to-secondary turns ratio. This allows the duty cycle of the switching circuit to set within a certain range. Very small duty cycles less than 5% should be avoided because of the difficulty in controlling the switching (due to delays in the system). Duty cycles greater than 50% can cause instability unless external compensation circuits or slope compensation are added. In some cases, such as where the input voltage range is very wide, a wide range of duty cycle may be unavoidable.

Another reason for an additional winding is to create a “bootstrap.” A bootstrap circuit creates a power supply for the switching circuit, typically in the range 8–15 V. The switching circuit will be powered from the main power source initially, but this can be inefficient if the power source is of high voltage. Once switching starts, the voltage developed on the bootstrap winding can be used to self-power the switching circuit. Suppose the device needs 2 mA to operate, when powered from a 300 V DC supply it will dissipate 600 mW, but when powered from a bootstrap winding at, say, 10 V it only dissipates 20 mW.

Carbon film resistors are low noise devices with a negative temperature coefficient. Component tolerances of 1 and 5% are standard, although 0.1% are available albeit more expensive. Through-hole resistors are constructed by applying a carbon film onto a ceramic rod, and then cutting a spiral gap in the film to increase the resistance. The spiral conductor is actually a lossy inductor. Surface-mount devices have a carbon film applied to one side of a ceramic sheet and a laser is used to cut part way across the film to alter the resistance. The short length of carbon film has very little inductance.

Metal film resistors have a lower noise than carbon film types, and a lower temperature coefficient. Component tolerances of 1% are standard, although precision devices in an E96 range of values with 0.1% tolerance and 15 ppm temperature coefficient are available at a higher cost. These resistors are constructed by applying a number of metal film layers, of different metals, to a ceramic former to achieve the correct resistance and a low temperature coefficient. In through-hole resistors, a spiral gap is sometime cut around the metal film to increase the resistance value and this increases the inductance slightly.

All conductors have some series inductance, simply due to having a certain length. This is typically 6 nH/cm. In fact some high frequency circuits just use a thin wire bond to form an inductor. Resistors are conductors and therefore have inductance too. Some types have more inductance than others. Even a thick-film surface-mount resistor has inductance, although of considerably lower value than other types. Wire-wound resistors have a significant inductance because of their construction; as we have seen with inductors, when a wire is wound into a coil its inductance increases in proportion to the number of turns squared. Carbon or metal film resistors that have had a spiral gap cut through their surface will have more inductance than a carbon composition type. All through-hole components will also have some inductance due to the wire leads connected at either end.

Resistors also have capacitance. The two ends have a certain cross-sectional area and are spaced a certain distance apart, separated by a ceramic dielectric. This capacitance is small, typically 0.2 pF, so has little effect in an LED driver circuit operating up to 1 MHz. However, at high frequencies this parasitic capacitance can be significant in parts of the circuit that are of high impedance.

An earth plane on high speed PCBs may cause problems when an inductor is placed on the board because of the capacitive coupling between the ends of the inductor and the earth plane. This capacitance forms a parallel tuned circuit with the inductance and may cause the filter to be detuned. One solution is to remove the earth plane from the area below the inductor. An alternative solution is to mount the inductor on spacers above the board, so reducing the capacitance.

Ceramic capacitors should be protected with a moisture-resistant coating. If moisture is absorbed into the ceramic material, the capacitance value will change. Moisture can also be absorbed into plastic packages, so a conformal coating over the whole board is preferred.

Some consideration should be given to storage of components before boards are assembled; metalized sealed bags should be used, perhaps with desiccant material. This will prevent moisture being trapped into an assembled board and avert the risk of damage during soldering (as the moisture boils off). Each component is given a moisture sensitivity level (MSL) by the manufacturer. This gives board assemblers an idea how to use the components; if MSL = 1, no real care needs to be taken. However, if MSL = 3 (say for a large integrated circuit), the IC has to be put in an oven to remove moisture before assembly. If this is not done the IC package could be damaged by any moisture inside turning to vapor during the heating process in PCB assembly.

Via holes are used to connect tracks on opposite sides of the board or to act as heat conduits below thermal pads. Through-hole PCBs have plated through holes that are 1 mm or larger in diameter. Surface-mount boards do not need holes large enough for component leads; hence they tend to be smaller in diameter. Metalized “via” holes 0.3 mm in diameter are common (used to connect two tracks rather than for component leads).

A problem with via holes arises when the board is heated. Glass and epoxy board, for example, FR4 type, has a high coefficient of expansion at temperatures above 125C. Above 125C the board goes through its glass transition temperature and its coefficient of expansion is greater than normal; Z-axis expansion increases the thickness of the board, and can cause fractures between the tracks and the via hole pads.

Soldering causes a problem due to the heat applied to the board; in wave soldering the board is heated to about 300C, which is way above the glass transition temperature. To reduce the problem of “via-hole” damage, all plated through holes should have a wall thickness of 35 μm or more. Temperature cycling of completed boards also causes problems.

On the surface, there is a temperature coefficient mismatch between components and the board. Leadless chip carrier (LCC) devices have an expansion coefficient of 6 ppm/C, but for the board it is 14 ppm/C (below the glass transition temperature) in the XY plane. Above the glass transition temperature the PCB has a coefficient of expansion of 50 ppm/C. Again, temperature cycling strains the solder joints and can lead to failure. A small gull-wing IC does not have a problem in this respect because the leads can flex a little.

PCBs built on aluminum sheet are often used with power LEDs. They can also be used for higher power LED driver circuits. Traditionally, copper-clad invar has been used within some PCBs to restrain expansion and to distribute heat. This should be used with polyamide boards, rather than glass and epoxy types.

A lacquer over the PCB called “solder resist” is commonly used to restrain solder within a solder pad area. However, surface-mount ICs use smaller packages than conventional leaded devices, so thin tracks of solder resist placed between the pads are not practical.

PCBs that have a fine track pitch tend to have 0.05 μm gold plating. If the gold is thicker it causes embrittlement. Gold or nickel plating gives a flat surface and makes surface-mount component placing easier.