Power supplies: The best design may not be the smallest

What makes a power-management IC easy to design with? I often hear that integration and component count are the dominant factors. ("This IC requires fewer components, so it's easy to design with.") While that is sometimes true, it's not as true as it used to be. Engineers should not universally apply this rule. Instead, they should consider the entire process of designing the power supply, implementing the design and bringing it to production.

If you consider two factors associated with almost every electronic system, the power supply is actually easier to design, in many cases, when the right components are left outside of the IC.

Start with the fact that powering a digital IC is the most common use of a dc/dc power-management subsystem. As a result of technology developments in the digital domain, the power supply is getting more complex and thus more difficult to design and get into production. For example, the error window for the regulated voltage is getting tighter, and it's often not firmly set until late in the development of the end system.

Therefore, adjustable output voltages (which require external resistors) are often a saving factor for system designers when, at the last moment, they realize that they can save power by running the IC at the low end of the supply voltage range. While external feedback resistors allow for this, let's look deeper. If the system designer chooses to run the digital IC close to the low end of the supply voltage range, and the digital IC is a dynamically changing load (as most digital ICs are), then external compensation allows the designer to tweak the loop response. This ensures that the processor's supply voltage never drops below the allowable range.

Consider the size constraints designers face. As electronic content continues to increase, but there is no corresponding increase in product form factor, size becomes a more difficult issue to resolve. Simplistic thinking suggests that fewer components in the design yield a smaller solution, but often this is not the case. For example, the system designer may want to select the highest-frequency, highly integrated power-management IC possible. And it should deliver the right amount of output current and operate within the required input-voltage range for each power rail. Then once the devices are selected, perhaps the designer might have an easier time of producing a small power supply.

Of course, the designer will have trouble finding a selection of ICs that meet those requirements, but the more difficult part is learning the unadvertised nuances of each one. A better solution is to use a product family that offers a series of voltage ranges and output currents, combined with adjustable switching frequencies. With this approach, each power supply can be designed with the right trade-off between size and efficiency. When this family also includes external loop compensation, it can also solve the problem of transient response.

There are obvious arguments against this approach. One is that each of those component values has to be calculated carefully-something many engineers aren't experienced enough to do with confidence. Another argument for integration is reduced size, Figure 1a and Figure 1b.

(Click to enlarge image)

(Click to enlarge image)
Figure 1: The size benefits of a widely adjustable switching frequency combined with adjustable loop compensation. Both the smaller design (Figure 1a, upper) and the larger design (Figure 1b, lower) provide the same voltage and current ratings, but with different efficiency, frequency, and performance attributes.
While external components do consume space, external loop compensation and other adjustable parameters are set using tiny ceramic capacitors and resistors, with a case size no larger than 0805 (2012 metric; 2.0 x 1.25 mm). In reality, components this small can often be sprinkled in spaces too small for anything else, so there is no size penalty in practice.

Of course, there are footprint (size) savings as a result of running the switching regulator at as high a frequency as possible, for each particular size/efficiency trade-off decision. Perhaps this is the fault of the IC manufacturer, which has not drawn the power supply components to scale in the product data sheet. The output capacitor is much larger than the tiny loop compensation capacitor. Engineers should understand that as they run a switching regulator at a higher frequency to reduce the space requirements, its efficiency is reduced, Figure 2.

(Click to enlarge image)
Figure 2: The trade-offs of running switching regulators as different frequencies; in this case, lowering the frequency from 800 kHz to 300 kHz yields an efficiency improvement of four to five percentage points.

This is a key trade-off: size vs. efficiency, each as a function of switching frequency.

Reliability must also be considered, since each component affects system reliability. We know that passive components are the most reliable parts, and that today's "chip shooters," which place the components on the pc board, are quite accurate. Supplies that cannot easily be set to the precise performance that a given application needs are the ones that pose more reliability issues than does using a few extra passive components. A nonoptimized power supply that does not have appropriate margining creates a greater reliability risk than a few added components.

If system designers had to be experts in every power supply they designed, my argument would be harder to support. That's why engineers have to use advancements in design tools to handle these calculations, based on the specific requirement priorities identified by the system designer. For example, National Semiconductor's SimpleSwitcher regulator IC family, combined with the WeBench online design tools, offer system designers a relatively easy way to configure power supplies.

We also see more interest in power supply modules. This is the extreme case of designing-in a power supply, yet with zero external components. If everything goes as planned in the design, or as hoped, then modules may be an acceptable option. For example, a module may make engineering sense, assuming the space is available, if a design engineer is on the second- or third-generation enhancement of a product design, and thus the likelihood of a surprise is small. It likely will not make commercial sense, however. In contrast, if the engineer is creating a new power supply for an entirely new application, then it's advisable to design a supply that is flexible.

The demands on the power supply increase with each technology advance. When your time is limited and you have to complete one or more power supply designs, don't look for the one with the fewest components.

While this may seem like the easiest route, it often will cause problems later because you won't have any flexibility to overcome the inevitable changes in requirements. Look for switching regulators which, when combined with online design tools, allow you to quickly and easily make important trade-off decisions. These parts will give you some parameters to adjust once you get your prototypes back and you realize that with a little tweaking, your entire system will operate better.

About the author
Mark Davidson (mark.davidson@nsc.com) currently holds the position of marketing director in the Power Management Division of National Semiconductor Corp. During his tenure, he has held various technical marketing roles, including director of marketing for the portable power systems product line. Prior to joining National Semiconductor, he worked in the automotive electronics industry, where he gained valuable experience from product design, manufacturing and applications engineering positions. He graduated from Pennsylvania State University in 1995 with a BSEE.