Take a top-down approach to simplify PSU selection

Paul Boughton
Stephen Dodson looks at the latest issues in selecting the right power supply.

For many systems designers, the number and variety of parameters that require consideration can make selecting a power supply a daunting task. This is especially true when environmental and regulatory issues add to the list of specification requirements.

Even board-level dc/dc converter modules can be challenging to select, as factors such as cooling requirements often complicate an apparently simple application. With planning, it is not difficult to find the solution that best suits the application. It is the application that drives this top-down process, starting with understanding the power requirements and how to integrate the power solution.

Most offline ac/dc converters offer universal input voltage capability that typically spans 90-264 VAC at 47-63Hz. This wide range allows configuration-free use anywhere in the world, relieving the system designer from the concern that end-users may incorrectly set their equipment.

This is a desirable feature, but remember that the optimal input-protection fuse ratings are rarely the same for 110V and 230V operation as these voltages mean a 2:1 spread in steady-state input current. The input fuse needs to be sized for the lowest operating voltage.

Some applications use the ac/dc converter to drive load circuits directly rather than via further dc/dc conversion stages. As a result, power-supply vendors offer a wide range of output voltage and current options, such as the 5V and +/-12V that traditionally power systems built from discrete logic and analogue circuitry.

Despite the trend towards ever lower voltage operation in systems that use complex digital logic ICs, many industrial designers continue to favour these levels as they offer better noise margins in electrically hostile environments. Ac/dc supplies are readily available with output voltages from 3.3-48VDC, in single and multiple output combinations. Single output supplies are often used in DPA (Distributed Power Architecture) and IBA (Intermediate Bus Architectures), where POL (Point of Load) converters are used to provide the various necessary voltages locally.

In terms of output current capability, it is tempting to over specify the supply to ensure it runs safely within its ratings under all conditions. Inappropriately over specifying a supply can be a costly mistake both in terms of initial purchase cost, and in running costs.

Most converters operate at their maximum efficiency towards the higher end of their operational range, say 80-85 per cent, and they are not anywhere near as efficient below 50 per cent of full load. Some multiple output supplies also require a minimum load to maintain regulation.

The output power and number of output rails greatly influence a power supply's physical dimensions. A glance at various manufacturers' catalogues reveals that encapsulated or open-frame board-mount ac-dcconverters typically have 1-3 outputs capable of delivering 5-30W. For example, 20W board mounted supply measures just 67 x 40 x 18.6 mm. At the opposite end of the scale, a rack-mounting mainframe may accommodate more than 20 plug-in modules of differing output voltages and currents that deliver in total more than 2kW from a single-phase ac input. For some applications forced-air cooling will be essential, which introduces restrictions such as minimum clearances between the supply's chassis and other hardware for air entry and exhaust paths. By contrast, the desktop supplies that most often power IT & medical equipment are enclosed and self-cooling, which normally restricts their power capabilities to about 150W. Most such units are single output, but multiple output versions are available.

Always check the converter's efficiency curves for the best fit for your application. Energy savings apart, it's worth remembering that as power supply efficiencies approach 90 per cent, a 1 per cent increase in efficiency reduces dissipation by as much as 10 per cent, with implications in not only cooling the power supply but in the heat generated within the end equipment.

Recently legislation has been introduced stipulating the minimum efficiency and no load power consumption of external power supplies. These include EISA (Energy Independence & Security Act 2007), CEC (Californian Energy Commission) and Energy Star. These pieces of legislation are specifically aimed at desktop and plug top power supplies due to the volume in use. It is thought that there are currently 1.5 billion externals in use in the US alone, which account for 6 per cent of the national electricity bill. In Europe it is thought that reducing the no load power consumption of external supplies will save up to 5TWh per year by 2010. When selecting your external power supply, it is essential to ensure that it meets these directives otherwise you could find the market for your end equipment being severely restricted.

Other power supply selection criteria may include methods of fault protection, remote sensing as well as ways to control output rails such as soft-start and, using multiple supplies, sequencing.

Virtually all supplies include some form of current limiting. Two popular techniques are foldback, when the supply progressively reduces its power output as an overcurrent event takes hold, and 'hiccup' mode when the supply successively removes and re-applies power until the fault condition clears. Power supplies will typically have some form of over voltage protection where the unit is shut down by an independent control loop. Internally, many supplies also monitor their own operating temperatures at critical points and shut down if they detect a dangerous situation.

Remote sensing is used in high-current applications where the lead resistance from the supply to the load introduces unacceptable voltage drops. A 4-wire connection measures the voltage at the load and overcomes cable losses to regulate the output voltage at the load.

The simplest alarm/status indication is the power-good signal, which is basically the output of a voltage comparator that reports the supply's output level is within a determined threshold. Others include input power fail, over-temperature warnings and fan fail. More sophisticated supplies may include a digital interface that permits a range of programmable control and status reporting facilities. The industry-standard PMBus (power management bus) protocol that many supplies implement consists of a defined language together with SMBus (system management bus) hardware that is very similar to the two-line I2C bus. These allow the host system to continuously monitor the supply and to control many of its operating characteristics.

These are the issues most users need to consider, but others may apply. For instance, in a high-reliability scenario, it's normal to include additional supplies that can take over if a main supply fails. The level of this redundancy technique is cost and system dependent. When backing up each supply with another (100 per cent redundancy) is too expensive, designers employ 'n + m' redundancy to improve reliability, with maybe one spare supply per four operational units theoretically providing a 20 times availability increase.

Stephen Dodson is Applications Engineering Manager at XP Power in Reading, UK. www.xppower.com.