June 5, 1997

Linear Supplies feel the (lack of) heat from switchers in low-wattage applications

BILL SCHWEBER, TECHNICAL EDITOR

When batteries are drained--or are unsuitable for your application--your design must get its power from the ac line. ICs that implement advanced supply techniques are challenging linear supplies by lowering costs, cutting size, and increasing efficiency.

Sometimes, batteries simply can't provide what you need in voltage, current, power, or energy value. Even when they can, they'll eventually need a recharge, unless you're using disposable cells. Either way, you have to turn to the power line as the operating or recharging source.

An offline supply takes the ac mains as its source and provides the dc rails you need. As such, nearly any ac-to-dc supply qualifies as an offline supply. Those that are less than approximately 20W find use in basic battery-recharging and -elimination applications, as well as keep-alive, standby, or operational supplies in TVs, games, or cellular and wired telephones. As more high-performance, lower power components become available, even desktop units--which aren't inherently power-supplied-constrained--fit under this wattage level.

The basic linear-regulator-based step-down design offers many advantages and has been the mainstay of low-power designs, despite step-down design's relatively low efficiency. The alternative switching-based designs are just too complex, costly, or idiosyncratic to justify in many of these applications. However, the latest switching-mode ICs have simultaneously solved several of the problems: They reduce component count, minimize operational difficulties, because these ICs include critical auxiliary circuits and functions, and give you a choice among several basic topologies. The result is that the crossover threshold between a linear regulator and a more advanced switching design has dropped to single-digit wattage de-signs.

The term "offline" is somewhat of a misnomer, because it implies that a supply is related to an "online" supply and that you use it only when your system is offline. However, that situation is not so. You could better describe an offline supply as "off-the-line," but the industry has settled on the shorter term.

It's easy to think of an offline supply, such as the ubiquitous wall adapter, as a fairly simple device. But, as is often the case, what looks simple isn't. Users may regard wall adapters, the most common examples of offline supplies, as nearly interchangeable commodities, yet they represent a complicated set of trade-offs regarding the placement of the power-conversion elements as well as fundamental architecture, efficiency, size, and cost.

There's no shortage of available dc/dc power-supply ICs, especially for using a battery as the dc input. You could, for example, rectify and filter the ac line and then have power go through one of these battery-optimized devices. This approach would certainly work, but it might not be best in the size, cost, weight, or perform-ance areas. In fact, it may have more functions and features than you need.

An offline supply has different constraints from those of a battery-operated dc/dc supply. First, although a dc/dc supply may have to boost, buck, or boost and buck the source voltage--depending on the battery chemistry and required output range--an offline supply normally works from an input voltage that is higher than the required output. This approach simplifies design and provides greater design efficiency if you want it. It also lets you consider using a linear regulator, in which the input voltage must be greater than the output value.

However, you may choose to make a different trade-off. Because the ac source is virtually "unlimited," absolute efficiency is not necessarily the overriding goal for the offline supply. You may elect to lose a few percentage points in efficiency if doing so lets you reduce total size, weight, or cost. Such inefficiency causes increased heat dissipation but not reduced operating time, as this inefficiency would in a battery-operated system. A little less operating efficiency and a few more degrees in operating temperature (or reduction in high-end operating range) may be an acceptable trade-off.

An offline supply also gives you more design flexibility. For example, you can feed a signal from the secondary side of the power transformer back to the primary to close a loop, thus allowing the load to exert control over the primary circuit and operating cycle.

Start with supply topologies

Your design may have a mains ac line coming into the box, so your task is to design the right high-level ac to low-level dc supply. However, many system designers prefer to keep the high-level ac outside their systems, which greatly simplifies meeting mandatory safety standards and reduces the internal thermal load. This wall-adapter ap-proach works well, as the millions of wall adapters in use confirm. You still need to decide how to physically partition your low-level ac, ac/dc conversion, and single-dc- vs multiple-dc-output development sections. (See box, "No matter what you do, someone will be unhappy.")

For safety reasons, a transformer isolates most offline supplies from the ac mains. By adding windings or feedback, you can also use the transformer as part of the supply-regulation function in addition to its role as a simple isolation and step-down element. Note, though, that some applications, such as powering a small circuit board or fan in a nonisolated home appliance, require no isolation.

A linear supply is relatively simple to design, has few components, and is reliable and inexpensive at its lowest power levels of a few watts (Figure 1). It generates no EMI/RFI, has no start-up or transient modes, and is easy to troubleshoot, with its simple low-frequency waveforms throughout. Unfortunately, it becomes the largest and heaviest supply when power levels exceed 10 to 20W, has only 20 to 50% efficiency, and may be difficult to design when the input voltage range deviates more than ±10 to ±20% from nominal.

Generally, a linear design needs one secondary winding, rectifier, and regulator per dc output, and the outputs are independent of each other. Don't rule out the fact that a low-dropout regulator can be the core of a simple and low-cost design, especially if you can operate it from a well-regulated supply rail that is just above its dropout voltage. However, this scenario is rare in most offline situations.

In contrast, switching supplies use some form of PWM and thus are 80 to 90% efficient yet more complex than are linear supplies. Switching supplies generate EMI/RFI that you need to suppress in most applications and have some start-up and duty-cycle characteristics that you must prepare for, both to ensure reliable operation and to maximize efficiency. You also need an isolated feedback path in most cases, either using an optoisolator or a transformer winding.

The buck technique steps down a higher voltage and then follows with a separate switching regulator for each output (Figure 2). It can handle a wider range of ac-line inputs and, depending on the required dc output values, may be able to use one secondary winding and rectifier that is common to all. As a result of the buck technique's higher efficiency, you can use a smaller transformer than with the linear design. If you don't need isolation, you may be able to run the supply directly from the rectified line without any intervening transformer.

In the flyback design, the mandatory transformer serves isolation, voltage-transformer, and energy-storage functions (Figure 3). A bias or bootstrapped winding provides isolated feedback for secondary-side regulation via the primary side of the transformer, along with a source of power for the regulator circuit once it is running. You can get multiple outputs at low cost, although in most designs one priority output has the tightest regulation, and the others are more loosely regulated; there is some cross-regulation between the outputs. A flyback design lets you more easily produce a negative supply output in addition to positive ones.

Further variations are a secondary-side post regulator that uses a primary-side regulator as well as one or more for the nonmain outputs of the secondary side (Figure 4). This approach provides good output regulation and output independence (Reference 1).

You're not restricted to using only one approach in a multiple-output offline supply. Depending on the voltage/current levels and combinations, you may decide to mix them to achieve the cost, efficiency, and independence you need.

One big virtue of linear supplies is that they have few transient-operating-mode difficulties. In contrast, the various switcher topologies have many. Devising the circuitry to control these modes has traditionally inhibited using switchers in place of linear supplies at the lower power levels. This approach requires considerable OEM-design expertise and debugging, as well as many discrete components. This situation is changing, however, as switching-IC vendors incorporate much of the "good-behavior" circuitry into their devices in addition to obvious functions, such as output protection against shorts or excessive loads.

For example, many of the new ICs include "soft-start" circuitry that controls the output voltage on power-up so that the voltage rises gradually rather than abruptly. This approach minimizes output overshoot and possible damage to sensitive loads. The Cherry CS-5102x series, for example, lets you provide a soft-start capacitor to control duty cycle during power-up and thus match the start-up time to the load.

The frequency of switching operation is another issue. (See box, "Choosing the frequency isn't easy.") You may want to synchronize operation to a specific system frequency to minimize interference with adjacent signals or harmonics, such as from a video-display refresh source. Synchronization works the other way, too: A sync output lets other switching controllers "slave" their frequency to yours, thus reducing the potential for mixing and beat frequencies among various fundamental frequencies or harmonics.

All current-mode PWM controllers can suffer from instability when their duty cycle becomes greater than 50%. To prevent this problem, controllers such as those in the Cherry series provide slope-compensation circuitry, which adds a small voltage to the current-sense voltage. At the other end of the load curve, most supplies must work under no-, low-, and full-load conditions; a few exceptions occur when the output load is both constant and always present. The output characteristics, response time, and regulation of a lightly loaded supply differ widely from those of a heavier load. Consider how the vendor accounts for this variation.

Undervoltage lockout (UVLO) is another subtlety. The supply should shut itself down when the input ac falls below a threshold. However, it's insufficient to simply have a fixed on/off threshold. Once the supply shuts off due to low line, the unloaded transformer-output voltage will probably rise and may go above the trip point unless the undervoltage is severe. The supply tries to start, the transformer output is again loaded down, the supply shuts off, and the process repeats itself. The UVLO circuit must have sufficient hysteresis under all normal line variations.

The apparent external simplicity of a switching-regulator IC hides the complexity within. For example, the TOPSwitch-II three-terminal (drain, source, control) offline PWM switch from Power Integrations results in a typical first-pass application schematic (Figure 5a), and the device itself incorporates many of the diverse functions for the overall design (Figure 5b).

Decisions, decisions

The obvious offline supply specifications include the nominal output voltage and current, along with the desired accuracy and line/load regulation. However, you should consider many more parameters when seeking the best offline topology and then try to optimize the resulting design. Among these parameters are:

• Do you need a universal supply that can handle 85 to 265V ac, or will a supply for nominal 110, 115, or 230V-ac ±15% operation satisfy you? With the restricted input range, your design is easier, smaller, and cooler than the universal range requires.

• Is there one output, or multiple outputs?

• Do the outputs have a common side, or are they isolated from each other?

• Is there a negative output?

• If you need feedback, do you prefer optoisolation, which requires an-other discrete component, or a winding on the transformer, which possibly defines a custom design?

• What is the dynamic load range: Is the supply, for example, going to just charge batteries, or will it also charge batteries while the user is running a laptop computer?

• Do frequency sensitivities exist in the circuit that the switcher operating frequency might affect?

• Do you want a switcher-controller IC that includes the power element, requiring fewer parts, or one that uses a separate power element and so has more flexibility?

• Do you need to develop a family of offline supplies, so commonality of design is desirable even if each member of the family is not optimal by itself?

• What efficiency do you need? Or, how much heat load can your system accept as a result of the supply's operation? How much efficiency are you willing to trade for performance, component count, or cost? (See box, "So talked-about yet so hard to figure.")

• Do you need any sequencing of supplies? How does output rise time depend on topology and load? How can you ensure the proper supply sequencing, especially if the load on each output varies?

• What output behavior do you need on other outputs if one output is shorted or fails?

To answer these diverse needs, you have many IC-based choices to consider. (See box, "Vendors of offline-switching and related ICs.") Although each IC meets the basic performance requirements, each also emphasizes another performance parameter or technique for optimization. For example, the Unitrode UCC3807 PWM controller lets you program the maximum duty cycle, thus preventing core saturation by limiting the volt-seconds driving the transformer primary.

For noise-sensitive applications, the LTC1538-AUX dual synchronous switching regulator operates at a constant, user-defined frequency, allowing easier filtering. Each regulator maintains this constant frequency even at low output currents as it selectively drives dual N-channel MOSFETs.

Motorola's MC33362 high-voltage switching regulator operates directly from the rectified 120V-ac line and includes an on-chip 500V, 2A MOSFET power switch, an active offline start-up MOSFET that reduces the number of external components needed for this phase of operation (Figure 6). The company also offers the MC33363 for 240V-ac line operation with a 700V, 1A switch.

To ease your selection of inductors, the ADP1148 (3.5 to 20V-dc input) and ADP1149 (3.5 to 48V-dc) synchronous switchers from Analog Devices drive external power MOSFETs as fast as 250 kHz using a constant off-time current-mode architecture. This approach yields constant ripple current in the inductor, rather than the larger ripple current characteristic of wide-input-range regulators.

Don't let the high-level integration of these ICs lull you, however. Al-though they greatly simplify an offline supply design, the nature of switching supplies inevitably requires many external passive components and often a few active ones for a final, reliable circuit. You may find that it's easier to use a transformer, rectifier, and basic filter, followed by an encompassing dc/dc package, such as National's LM2825, which provides 3.3 or 5V in a 24-pin DIP without external components (Reference 2).

Passive yet critical transformers

It's ironic that the transformer, an ancient electrical device that English physicist and chemist Michael Faraday and American physicist Joseph Henry developed in the 1830s, plays such a key role in modern switching-supply design 160 years later. Although it's easy to calculate the basic turns ratios you need, it's hard to determine most other transformer parameters. Among the factors affecting basic performance and efficiency are the kind of core, the way the turns are wound with respect to the core and the other windings, the use of a standard vs a custom design, and the kind and size of wire used. You can obtain useful application notes--although specific to one vendor's product line--about transformer design and selection in Reference 3. (Reference 4 is also helpful.)

An added factor in transformer selection is that, in most designs, the transformer also serves as the line-isolation element and so is subject to safety codes. This situation affects winding spacing, bobbin design, and even wire insulation. For example, you can trade off spacing of windings vs use of triple-insulated wire, which provides redundant insulation layers.

Before you develop an offline supply design, investigate what IC vendors already have done. Nearly all have produced application notes and reference designs that range from basic design guidelines and analysis through ready-to-manufacture completed designs (Reference 5). Be careful, though: It's generally unwise to substitute vendor-specified components unless you know what you are doing and are willing to validate your modified design; the same caveat holds true for any layout changes.

Be prepared to relax some of your requirements if you use an available reference design. The converse is also true: Be willing to use a reference design that is better than you need if it more quickly gets you a working design. Finally, recognize that second-source or even similar switching devices from different vendors are uncommon. With few exceptions, each vendor has a proprietary twist on one of the major topologies, provides a different selection of attributes and features, and implements both the standard and extra features in a different way. You have to carefully study the vendor's offerings and ask questions of its applications group to make meaningful comparisons because of the differences among vendor designs.

If you need multiple outputs, consider mixing topologies, such as buck switching and linear regulator, rather than limiting yourself to just one technique. You may find that a linear regulator is still the best choice for a low-wattage supply for a bias circuit, for example, because of the linear supply's simplicity, low noise, or quick and reliable start-up when you apply power.

References

1. Levin, Gedaly, "Secondary Side Post Regulator for Switching Power Supplies with Multiple Outputs," CS-5101 Application Note, Cherry Semiconductor Corp.

2. Schweber, Bill, "DC/DC converter in IC package eases power-supply design-in," EDN, May 23, 1996, pg 17.

3. Data Book and Design Guide, 1996 to 1997, Power Integrations Inc.

4. Leman, Brooks, "Spreadsheet simplifies switch-mode power- supply flyback-transformer design," EDN, March 14, 1997, pg 103.

5. Schweber, Bill, "Reference designs reshape design engineering," EDN, May 9, 1996, pg 73.

| EDN Access | Feedback