The growth of distributed power architecture (DPA) has increased the demand for dc-dc power conversion. As a result, a number of "off-the-shelf" packaged dc-dc converters have become available. Careful design with these converters can help engineers to save space, reduce component count, and simplify the layout of power conversion stages by eliminating the need for discrete solutions.
Not surprisingly, a growing number of engineers are choosing the packaged approach. Designers, though, should be aware of certain issues that arise from using these components. Because isolation is often necessary for either safety or performance reasons, it's important that engineers understand the impact of using isolated dc-dc converters in their designs.
The first question that any engineer must address is whether to use an isolated or a nonisolated dc-dc converter. For instance, when designing safety-critical applications such as products for the telecommunications or medical sector, isolated converters will almost always be required. Very often these devices will have to be certified to the rigorous UL1950 safety standard. In such cases, only UL1950-approved converters that have been independently assessed by a third party for both isolation strength and physical separation of the windings within the component, should be used.
Safety, however, isn't the sole reason for isolation. In most modern noise-sensitive circuitry, it will be necessary to isolate the load and noise presented to the local power-supply rails from the main supply rails of the entire system. To achieve this goal, dc-dc converters with a high galvanic isolation should be the automatic choice. The basic input-to-output isolation can then be used to provide a simple, isolated-output power source. Or, it could be used to generate different voltage rails, dual-polarity rails, and/or nonstandard voltages.
When using isolated converters, it's vital that the designer fully understands the difference between the converter's rated isolation voltage and its rated working voltage. Isolation is defined as the voltage that the device can withstand between input and output, for a fixed time period. An example is one second. This parameter is a measure of the electrical strength of the insulating materials used.
Rated working voltage, on the other hand, is defined as the maximum continuous voltage that can be sustained across the component's isolation barrier. Typically, it's lower than the rated isolation voltage. Determining a converter's working voltage from its specified isolation voltage is a controversial matter. Their relationship depends on the construction of the individual device and the materials used. That aside, the IEC-950 specification for the safety of information technology equipment provides guidelines for calculating working voltage from isolation voltage (Fig. 1).
Despite the many benefits offered by packaged dc-dc converters, many engineers decide to design and implement their own discrete power-conversion stages. This is often because they have a particular custom requirement that cannot, at first glance, be solved by using a standard part. Fortunately, many custom requirements can be addressed through the creative use of packaged, isolated converters.
At the most basic level, the input-to-output isolation of a converter can be used to generate different voltage rails and/or dual-polarity rails (Fig. 2). Furthermore, a negative supply can be generated by connecting the isolated positive output to the input ground rail. Alternatively, it's possible to produce a voltage above that of the main supply by referencing the output to a voltage other than ground (Fig. 3).
The output is isolated from the input, so the choice of reference for the output side can be relatively arbitrary. Where a floating output isn't required, the 0-V output can be connected to the input ground. Doing this converts a single-ended voltage to a dual rail, although the converter's inherent isolation is then lost.
It should be emphasized that, when mixing reference levels, regulated converters will need more careful consideration than their unregulated counterparts. The former usually only have a series regulator in the positive output rail. Therefore, referencing the isolated ground will work only if all the current returns via the converter rather than through external components, such as diodes and transistors.
The galvanic isolation of the output from a dc-dc converter also allows nonstandard voltage rails to be generated by connecting multiple converters in series. This is accomplished by simply attaching the positive output of one converter to the negative output of another. This configuration allows a wide variety of voltage variations to be produced provided that one rule is obeyed: the output current for the whole cannot exceed the output current rating of the dc-dc converter with the highest output voltage.
When connecting converters in series, additional filtering is strongly recommended because the converters' switching circuits won't be synchronized. An unfiltered output produces a ripple voltage that's the sum of the ripple voltages from the individual converters. It can also generate relatively large beat frequencies. (Filtering will be covered in detail later.)
Even in applications with varying input voltage levels or power requirements beyond that of a single packaged solution, it isn't always necessary to revert to a discrete design. For example, many of today's miniature converters are available with wide 2:1 or even 4:1 input voltage ranges. This makes them ideal for applications in which power has to be supplied over a long distance, or where the application is likely to demand varying input voltage levels.
Multiple packaged converters connected in parallel often provide a suitable solution for cases where a single off-the-shelf dc-dc converter is unable to deliver the required output power. Where parallel connection is used, though, it's always good design practice to parallel converters of the same type. This is because, generally, output voltages of different types of converters won't be sufficiently well matched.
In addition, when connecting the converter outputs, it must be remembered that the switching won't be synchronous. Therefore, some form of coupling must be employed. In cases of models with 12- or 15-V outputs, one possible remedy is to use a diode feed arrangement (Fig. 4). Here, the diode voltage drop won't significantly affect circuit functionality. But with 5- and 9-V supplies, the diode drop will usually be too large to make this a suitable connection scheme.
Ideally, the method for connecting converters in parallel would be via series inductors on the output (Fig. 5). This scheme offers lower voltage losses than the diode-based connection method. At the same time, through careful selection of inductors and an additional external capacitor, it also helps to reduce beat frequencies and the ripple from each converter.
In general, parallel connection of converters should be used when a single higher-power converter is unavailable. Additionally, there should always be a correction factor for the maximum power rating that will allow for any mismatch between converters.
Typically, up to three converters can be paralleled, and designers may still maintain a high level of confidence in the circuit's overall performance. It should be noted, though, that converters with regulated outputs shouldn't be connected in parallel. This is because it cannot be guaranteed that the output voltage match between devices would be accurate enough for truly even loading—that's within the tolerance of the internal linear regulator. Consequently, one of the converters could become overloaded. For this reason, when a high-power regulated supply is required, it's normally preferable to parallel unregulated converters, and then add an external regulator.
Because packaged dc-dc converters are noise-generating devices, designers must take filtering requirements into account when selecting a device. First, an engineer must consider choosing a converter that uses pulse-skipping or fixed-frequency technology. In a pulse-skipping scheme, a large range of frequencies are encountered as the device adjusts the pulse interval for loading conditions. This may lead to complex filtering requirements. In contrast, the fixed characteristic of a fixed-frequency converter can significantly simplify filter design.
Certain guidelines apply to the design of filter circuits for fixed frequency dc-dc converters. To reduce ripple on the input or output of such a converter, a passive LC network is often all that's needed. Normally, this approach is preferable to a passive RC network that could suffer from significant power loss through the resistor.
When designing the LC network, the self-resonant frequency of the inductor should be significantly higher than the characteristic frequency of the dc-dc converter. The dc resistance of the inductor should be taken into account as well, to ascertain the expected dc power loss. A good rule-of-thumb for selecting an inductor is that the inductor's current rating should be approximately twice that of its supply current.
One issue that designers need to carefully consider is limiting the current seen at switch-on, so that it cannot damage the packaged part. Fortunately, using a series inductor at the dc-dc converter's input not only helps to filter the noise from the device's switching circuitry, but it also limits the inrush current.
To realize the importance of limiting this current, consider the circuit without the inductor. Input current to the converter may be calculated from the equation:
where V is input voltage, t is a time interval, R is the input resistance associated with pc-board traces and wiring, and C is the converter's input capacitance. When the device is initially switched on (t = 0) this equation simplifies to the familiar I = V/R. If we bear in mind the example of a low-power dc-dc converter with a 5-V input and, say, a 50-mΩ track-and-wire resistance, the inrush current could be as high as 100 A, which could cause problems for the device.
When using a packaged dc-dc converter, designers must also consider how the circuit reacts the input voltage drops out. Input voltage drops have various causes, such as when other circuitry undergoes instantaneous current demand or when devices are plugged in or removed from the supply rail while "hot." When the input voltage drops or is momentarily removed, the output circuit of a dc-dc converter suffers a similar voltage drop.
In these cases, a simple diode-capacitor arrangement can prevent the output circuit from being affected (Fig. 6). The circuit uses a diode feed to a large reservoir capacitor that provides a short-term reserve-current source for the converter. The diode blocks other circuits from draining the capacitor via the supply rail. As it was in paralleling devices, diode voltage drop is also important in this arrangement. For this reason, it's generally recommended that a Schottky diode be used.
Although the use of filtering will prevent excessive current at power-on, under normal conditions there's no protection against an output circuit taking excessive power or even going short circuit. When this happens, the dc-dc converter will try to supply the output by taking a large input current. If this situation isn't rectified, the device will overheat and, eventually, destroy itself.
The easiest way of preventing overload at the output is to use a fuse with sufficient tolerance to the inrush current, so that it won't blow at power-on. Alternatively, a simple circuit breaker can be used. There exists, though, the potential to add some intelligence to the overload scheme by detecting either the input current or the output voltage. The simplest implementation for overload protection at the input is to have the devices supplied by a linear regulator with an internal thermal shutdown facility. But, this has a significant, negative impact on overall efficiency.
If there's an intelligent power management system at the converter's input, replacing the series inductor with a series resistor allows the voltage drop across the device to be used to signal the management system. A similar scheme can be applied at the output in order to determine the device's output voltage. On the other hand, if the management system is on the input side, then the signal will need to be isolated from the power management controller to preserve the system isolation barrier (Fig. 7).
Additionally, the thermal dissipation in a series resistor on the output can be used to determine overloading, while preserving isolation. A thermistor, or another thermally sensitive device mounted close to the sense resistor, can be used to indicate an overload condition. In such instances, the system temperature must be known in order to provide a suitable offset for different operating environments.
When using packaged dc-dc converters in a DPA application, the designer might find it necessary to calculate the leakage current of the device. In the case of galvanically isolated dc-dc converters, the isolation barrier within the device has a capacitance. This is a measure of the coupling between the input and output circuits. Provided that it's the largest coupling source, a calculation of the leakage current (IL) between input and output circuits can be estimated.
Assuming that there's a known isolation capacitance (CIS) and a known frequency (F) for either the noise or the test signal, then the expected leakage current can be calculated from the isolation impedance. The general equation for isolation impedance at a given frequency is given by:
The leakage current can then be calculated by dividing the test voltage by the value of Z F.
As the above demonstrates, leakage currents are larger for higher test or noise voltages, and lower for lower isolation capacitances. Therefore, to obtain low leakage current and high noise immunity, designers should select high-isolation dc-dc converters with an appropriately low-isolation capacitance.
Because dc-dc converters are heat-generating devices, temperature specifications are the key to successful DPA design. It's important to distinguish between the terms "operating range" and "specification range," as well. The specification range of a device is the range of temperatures in which it will perform key datasheet specifications, like rated power and noise output. Still, the operating range, which is generally wider than the specification range, is the complete temperature range that the converter will work in, albeit with performance degradation.
This is a critical issue, as some converters may claim operating ranges of −40° to 85°C, and yet only meet specifications between 0° and 70°C. The noise figure, for example, could be more than double outside of the specification range. As a result, performance specifications need to be analyzed carefully.
Keep the following instance in mind. If a datasheet states that a converter with an operating range of −25° to 75°C derates at 3% per degree above 60°C then, at the maximum operating temperature of 75°C, the power will be just above half the datasheet specification. Indeed, there are instances of 15-W converters that will only deliver 7 W at their maximum operating temperature. Because of this, it's vital for the engineer to ensure that the chosen converters offer the requisite performance across the application's full temperature range.
Within the last five years, a large number of surface-mount, low-power dc-dc converters have been launched. But, while most of these devices offer some advantages in power conversion applications, only now are technologies emerging which allow designers to capitalize on the true benefits of surface-mount technology (SMT). Great care should be exercised when selecting SMT converters.
One of the main problems in this area is that many existing surface mount dc-dc converters are, in fact, modified through-hole components. These converters are based upon standard hybrid circuitry, with pins bent to form level contacts. There are a number of disadvantages to using such devices.
First, the physical bending of the pins can introduce an extra level of mechanical stress into the component. And, because they weren't originally designed as true surface-mount components, these devices are rarely able to withstand the vapor phase and infrared reflow soldering processes of the modern automated manufacturing line.
For example, on most modern high-volume assembly lines, an IC must withstand lead temperatures of up to 280°C to achieve compliance with the standard CECC00802 reflow profile. Often, this means a temperature at the component surface in excess of 300°C. Yet, with many hybrid SMT dc-dc converters constructed by using thermoplastic casings and 220°C solders, the maximum reflow temperature is significantly lower than that of an IC.
As a result, designers incorporating these converters into their applications must allow for two separate mounting and soldering processes—standard components and converter hybrids. Naturally, this two-step process adds to assembly time and increases manufacturing costs, thus eroding many of the fundamental benefits associated with the use of surface-mount components.
Strict Coplanarity Requirements
Another problem with hybrid converters relates to what's called pin coplanarity, or alignment. International standard IEC 191-6:1990 defines strict requirements for coplanarity in surface-mount devices, making the positioning and size of the converter leads absolutely vital. Compliance to the standard, however, can be difficult to achieve consistently with the traditional methods used in hybrid manufacture. This can mean time-consuming and costly problems because even a slightly misaligned pin can lead to inaccurate placement on the board, poor soldering performance, and part rejection.
To address this issue, designers should be certain that their dc-dc converter manufacturer has systems in place which will allow them to verify the coplanarity of each and every component that they supply. Newport Components, for instance, has developed its own customized, automated system for testing coplanarity. This system uses a technique that builds a three-dimensional model of the device and calculates x, y, and z coordinates of a lead to within 10 µm. As a result, the company is able to offer surface-mount converters that guarantee 0.1-mm pin coplanarity.