The design of power conversion and management circuitry for portable computing products presents unique challenges. Dealing with multiple power sources and output voltages, size and weight constraints, and of course all-important battery life, can consume a disproportionate amount of design time. The problem is often compounded by less than ideal input-voltage ranges dictated by battery chemistries and unregulated supplies, particularly if problems in these areas are uncovered late in a product's design. By identifying and dealing with some of these roadblocks, designers can at least be made wary of approaching problems in their systems. Some circuit suggestions may also prove useful as specific solutions, since all have been applied in working portable designs.
A particularly troublesome problem in portable systems occurs when the input supply voltage (battery, wall adapter) range includes voltages above and below the desired output voltage. This situation occurs quite frequently when very common logic supply voltages (5 V and 3.3 V) meet up with common battery topologies, in particular, when generating 5 V from a four-cell alkaline, nickel-cadmium (NiCd), or NiMH battery (3.6-V to 6-V range), or when making 3.3 V from a three-cell battery or one lithium rechargeable cell (2.7-V to 4.2-V range).
These cases don't allow basic boost or buck designs because a boost converter cannot step down and a buck converter cannot step up. The simplest solution would be to simply change the cell count so that a buck or boost design would work, but this is seldom an option since size and weight limits usually prevent cells from being added, and power needs keep cells from being eliminated. Even if the battery voltage does not cover a wide span, and is always lower than the output, there may still be a problem if an ac adapter is used. For example: 5 V can be generated from a two-cell battery with a conventional boost converter, but the ac adapter, with what is often a poorly regulated output, may sometimes have too high a voltage to allow a boost converter to operate properly.
Combining a boost converter with a linear regulator provides what usually is the lowest cost solution for applications requiring a buck-boost function. It might at first seem odd to find a linear regulator in a design where battery life is critical, and that using one for a step-down function would have a large efficiency and battery-life penalty. However, when actual battery profiles are considered, the impact is less than may be expected.
In a voltage profile for four-cell AA alkaline and NiCd batteries operated with a circuit generating 5 V at 70 mA, in the first 30 minutes of use, the alkaline falls below 6 V (where linear regulator efficiency is 83 percent) and the NiCds fall to 5.2 V. Once the boost converter is activated, conversion efficiency is typically 88 percent.
In another example, a circuit operates as a boost (step-up) dc-dc converter followed by linear regulator. When the input voltage is greater than about 5.4 V, the linear regulator supplies the load. When the input falls below this level, the dc-dc converter takes over. The most common implementation of this scheme is with separate ICs for the boost and linear blocks; however in this case, both are included in one IC.
This provides a few advantages other than higher scale integration. One is that the linear and boost converter share one feedback network to set the output voltage. A second feature of this combination of boost and linear blocks is that the linear regulator acts as a type of soft-start switch by not connecting the load on power-up until the boosted output is ready. A third advantage is that operation can be configured for lowest noise or best efficiency by selecting the forward drop of the linear regulator via the IC's E/N pin.
A second scenario that may justify the boost-linear topology occurs when the input voltage is high only when an ac adapter is present. High efficiency is not normally a need when wall-powered. Nevertheless, power dissipation and waste-heat concerns often rule out this approach. Dissipation problems arise when worst-case high- and low-line dc voltages are considered, and are usually compounded by the loose tolerance of low-cost ac "wall cubes." This situation is frequently the type of problem that is uncovered late in a design. As a practical matter, if heat sinks are to be avoided, a regulated wall source (or no wall-powered source) is typically needed to make a boost-linear design work, except when loads are small. The expense of a regulated wall cube vs. an unregulated one, of course, must be weighed against added cost incurred in the power supply to accommodate a wider input range.
If a regulated wall cube can be used, the design-limit output current for the circuit increases over what could be supplied from an unregulated source, since power dissipation is more tightly controlled. There is a change in available linear-regulated output current. Remember that since the linear regulator dissipates very little power when the boost converter is on, the limits are only a function of the maximum input voltage and do not apply when boosting from batteries, only when regulating down from a higher than 5-V (or other output voltage) dc input.
If power dissipation restrictions preclude a linear regulator for stepping down in buck-boost applications, another useful topology is Sepic (single-ended primary inductance converter). This architecture is switch-mode over its entire input range and, by design, exhibits no transition artifacts when the input passes through the range of the output voltage. The circuit operates in a manner not so different from that of a flyback design (with a 1:1 transformer), except that a coupling capacitor also transfers energy to the output.
Efficiency peaks at around 85 percent and is above 80 percent for most of the load range. Though respectable, these numbers are not as good as those of contemporary buck- and boost-only designs, a side-effect of having more components in the energy transfer path.
Battery life improvement, if any, will depend on the shape of the battery discharge curve. If the majority of operating time is spent with the battery significantly higher than VOUT/0.8 (0.8 represents a conservative estimate of Sepic efficiency), then a Sepic should provide more "mileage." If, however, the battery hovers slightly above VOUT for most of its life, then linear-regulation losses will most likely be less than those of a Sepic. Even in this case, high linear regulator power dissipation at high (wall cube) input voltages may still point to a Sepic design even with no efficiency advantage.
In multiple supply systems, usually only one of the regulated output voltages is inconvenient to generate because of the step-up, step-down requirements mentioned earlier. In such a case, an effective solution might be to cascade the difficult-to-generate supply on the output of the easier-to-generate one.
For a low-power design (100 mA), the connection provides an elegant solution to many switchover problems. A more typical solution for this might use steering diodes to select the higher of the two input sources. This approach is fine when the input voltages are high and the diode forward voltage drop is small compared with the battery voltage. But when the battery is a two-cell alkaline stack, a 0.3-V Schottky voltage drop wastes more than 10 percent of the available energy-a stiff penalty. The loss can be reduced by replacing steering diodes with power MOSFETs-this is frequently done with success, but not without added cost and complexity.
A battery-to-ac wall adapter to back up battery switchover circuitry is a common source of circuit bugs that are too often uncovered late in a design. Problems that can arise include: excess power loss in steering diodes, unforeseen reverse current through FET switch body diodes, Vcc dropout, and system reset during handoff from one power source to the next.
In another type of buck-boost solution, the connection works well when boosting from a low-voltage battery power (two AAs), or stepping down from a somewhat regulated wall adapter: It counts on the wall-powered supply to always be greater than the battery, but not more than 6.5 V to stay within the power and voltage limits of the SOT23 regulator IC. Since heat sinks are usually not appropriate for portable designs, an all-switch-mode approach would be more appropriate for higher power levels, or if the wall source is unregulated.
In a complete power supply design for a medium-power portable device such as a full-featured organizer or handheld inventory or point-of-sale (POS) computer, the peak output is about 2 W. Portable operation is from two AA batteries (typically NiCd) which are boosted to 3.3 V at up to 600 mA (during load peaks). Since the input from the ac or car adapter covers too wide a range, linear regulation is not an option. This also allows dc input voltages to be as low as 4 V. Most power steering diodes or FETs are eliminated by combining power sources at the output of the regulators rather than at the input. This is made possible by ICs that have milliamp-level shutdown modes as well as very low operating currents.
The supply also uses a lithium coin cell to back up RAM when no other power source is available. Since the cell voltage may be as low as 2.7 V, a small (micro-8) charge-pump boost converter generates a regulated 3.3 V when commanded by the BATT NG (battery no good) line. An IC generates the high-voltage supply required by the LCD panel. Its operation is also gated by "OK" signals from the dc supply or the battery.
A complete power-supply circuit often requires a bit of additional voltage detection, logic, diodes, etc. to work out all of the operating kinks.
This is typical of most system-level power-supply designs after pragmatic decisions trading performance and cost are considered. For this circuit, such decisions or compromises are made in a number of locations:
- A Sepic design was not chosen. The boost-only topology provides as much as a 10 percentage points of efficiency improvement over Sepic in this case, particularly when considering the minimum input voltage-as low as 1.8 V from the two AA batteries.
- Dc power source detection is done with a logic gate rather than a comparator circuit. It was decided that the need to accurately detect, and reject, a marginally low dc supply was not critical to the design.
- A MOSFET switch and a diode are needed to steer battery or dc power to the LCD supply dc-dc boost converter.
- LCD bias (24-V LCD) must fall to 0 V when shut down, but a boost dc-dc converter's output only falls to its input voltage when the IC is shut down. This is so because an input-to-output current path still remains. This limitation is often fixed by breaking the current path with a pnp transistor switch at the VLCD output, but since a MOSFET switch is already present, it takes care of this by switching off power at the input to the inductor.
- A MOSFET switch is needed to disconnect the majority of the 3.3-V logic load from the backup supply when the lithium battery is powering RAM, real-time clock and other backed-up circuits. It may not be needed if these other circuits can be "clocked down" or disabled with their own logic shutdown pins.
- Care was taken to insert no MOSFET switch, diode or sense resistor in series with the battery input. Since the battery voltage is low and input current at this pin is high, additional resistance or voltage drops here can have a devastating impact on efficiency. Even when the cost of a design must cut to the bone, consider that 10 cents spent in the battery current path does more to extend battery life than 10 cents spent or not cut anywhere else in the circuit.
See related chart