While an internal, rechargeable lithium battery is usually the best solution for portable kit nowadays, there are still times when using replaceable cells with an external power option, probably from a USB source, is more appropriate. This DI shows ways of optimizing this.
The usual way of combining power sources is to parallel them, with a series diode for each. That is fine if the voltages match and some loss of effective battery capacity, owing to a diode’s voltage drop, can be tolerated. Let’s assume the kit in question is something small and hand-held or pocketable, probably using a microcontroller like a PIC, with a battery comprising two AA cells, the option of an external 5 V supply, and a step-up converter producing a 3.3 V internal power rail. Simple steering diodes used here would give a voltage mismatch for the external power while wasting 10 or 20% of the battery’s capacity.
Figure 1 shows a much better way of implementing things. The external power is pre-regulated to avoid the mismatch, while active switching minimizes battery losses. I have used this scheme in both one-offs and production units, and always to good effect.
 |
|
| Figure 1. | Pre-regulation of an external supply is combined with an almost lossless switch in series with the battery, which maximizes its life. |
The battery feed is controlled by Q1, which is a reversed p-MOSFET. U1 drops any incoming voltage down to 3.3 V. Without external power, Q1’s gate is more negative than its source, so it is firmly on, and (almost) the full battery voltage appears across C3 to feed the boost converter. Q2’s emitter-base diode stops any current flowing back into U1. Apart from the internal drain-source or body diode, MOSFETs are almost symmetrical in their main characteristics, which allows this reversed operation.
When external power is present, Q1.G will be biased to 3.3 V, switching it off and effectively disconnecting the battery. Q2 is now driven into saturation connecting U1’s 3.3 V output, less Q2’s saturated forward voltage of 100–200 mV, to the boost converter. (The 2N2222, as shown, has a lower VSAT than many other types.) Note that Q2’s base current isn’t wasted, but just adds to the boost converter’s power feed. Using a diode to isolate U1 would incur a greater voltage drop, which could cause problems: new, top-quality AA manganese alkaline (MnAlk) cells can have an off-load voltage well over 1.6 V, and if the voltage across C3 were much less than 3 V, they could discharge through the MOSFET’s inherent drain–source or body diode. This arrangement avoids any such problems.
Reversed MOSFETs have been used to give battery-reversal protection for many years, and of course such protection is inherent in these circuits. The body diode also provides a secondary path for current from the battery if Q1 is not fully on, as in the few microseconds after external power is disconnected.
Figure 1 shows U1 as an LM1117-3.3 or similar type, but many more modern regulators allow a better solution because their outputs appear as open circuits when they are unpowered, rather than allowing reverse current to flow from their outputs to ground. Figure 2 shows this implementation.
 |
|
| Figure 2. | Using more recent designs of regulator means that Q2 is no longer necessary. |
Now the regulator’s output can be connected directly to C3 and the boost converter. Some devices also have an internal switch which completely isolates the output, and D1 can then be omitted. Regulators like these could in principle feed into the final 3.3 V rail directly, but this can actually complicate matters because the boost converter would then also need to be reverse-proof and might itself need to be turned off. R2 is now used to bias Q1 off when external power is present.
If we assume that the kit uses a microcontroller, we can easily monitor the PSU’s operation. R5 – included purely for safety’s sake – lets the microcontroller check for the presence of external power, while R3 and R4 allow it to measure the battery voltage accurately. Their values, calculated on the assumption that we use an 8-bit A–D conversion with a 3.3 V reference, give a resolution of 10 mV/count, or 5 mV per cell. Placing them directly across the battery loads it with ~5–6 µA, which would drain typical cells in about 50 years; we can live with that. The resistor ratio chosen is close to 1%-accurate.
Many components have no values assigned because they will depend on your choice of regulator and boost converter. With its LM1117-3.3, the circuit of Figure 1 can handle inputs of up to 15 V, though a TO-220 version then gets rather warm with load currents approaching 80 mA (~1 W, its practical power limit without heatsinking).
I have also used Figure 2 with Microchip’s MCP1824T-3302 feeding a Maxim MAX1674 step-up converter, with an IRLML6402 for Q1, which must have a low on-resistance. Many other, and more recent, devices will be suitable, and you probably have your own favorites.
While the external power input is shown as being naked, you may want to clothe it with some filtering and protection such as a poly-fuse and a suitable Zener or TVS. Similarly, no connector is specified, but USBs and barrel jacks both have their places.
While this is shown for nominal 3 V/5 V supplies, it can be used at higher voltages subject to gate-source voltage limitations owing to the MOSFET’s input protection diodes, the breakdown voltages of which can range from 6 V to 20 V, so check your device’s data sheet.
