By Steve Diaper
Until recently, even the lowest-voltage, lowestpower MCUs required a minimum supply of 1.8V to operate, which requires at least two alkaline batteries in series for battery operation. However, a new MCU family now offers a minimum operating voltage of just 0.9V—the end-of-life voltage of a single alkaline battery.
By running from a single battery, you could, for example, replace two smaller cells with a larger single-cell battery within a similar form factor, all the while increasing product battery life. Alternatively, you could take the two existing batteries and connect them in parallel rather than in series, again increasing product battery life by a worthwhile amount. Such parallel battery connections do require a mechanical means to prevent reverse connection of the two cells, but are otherwise a good way to maximise battery lifetime.
Another possibility would be to remove one battery. You might imagine that removing a battery would reduce the battery life of your product by half. However, as we shall see shortly, that need not be the case
Operating from a single cell It is, of course, possible to create a system using a standalone dc-dc boost converter, which can run a standard MCU from 0.9V by boosting the input voltage to provide a 1.8V or greater supply. However, this standalone approach has a number of limitations within a battery- powered embedded system.
To minimise current consumption, it’s desirable to disable the dc-dc converter when it’s not needed. But if the dc-dc converter is turned off, then the MCU will be without a supply voltage. Thus, it won’t be able to maintain a real-time clock or restart the system without an external input of some kind.
Even worse, the MCU will lose its entire RAM contents whenever the dc-dc converter is disabled. However, without disabling the dc-dc converter, then the standby current consumption of the system, even with the MCU in sleep mode, can remain high—often greater than 20μA.
Aside from this, there’s the active efficiency of the dc-dc converter and MCU to consider. Most standalone dc-dc solutions are designed to be most efficient when delivering at least 150mW (and in most cases considerably more) to the load— and are a lot less efficient at lighter loads. By contrast, a typical MCU device will draw less than 30mW from the supply, which can result in surprisingly low dc-dc efficiency in the range of 50% to 70%.
So, is there another, more efficient solution? What if you were to integrate an optimised, low power dc-dc converter on to the same silicon as the MCU? This could readily reduce system cost and board space. If you also included the ability to retain the RAM contents and run a real-time clock using the low input voltage, down to 0.9V, then the MCU could have control of its own power system.
Moreover, if you optimised the standard microcontroller peripherals and functions (including standby mode, wake up from sleep, and fast code execution) for the lowest possible leakage current and power consumption, then the device could support single-cell operation while still having a battery lifetime comparable with a dual-cell implementation.
THE ADVANTAGE OF AN INTEGRATED SOLUTION
The approach adopted by Silicon Labs for its recently introduced C8051F9xx MCU family integrates a highly optimized, step-up dc-dc converter into the MCU. This can boost the incoming battery voltage, between 0.9 and 1.5V, up to a programmable output voltage of between 1.8 and 3.3V. This boosted voltage is then used for the I/O pins and analog peripherals of the MCU. By using an optimised 65mW dc-dc converter, the converter’s efficiency can remain as high as 80 to 90% (Fig. 1).
Not only that, but since this dc-dc converter can supply a total output of 65mW, the boosted output voltage can also be used to provide a voltage supply for external components. In this way, some of the potential problems with interfacing to other higher-voltage ICs or sensors, driving blue LEDs at 3V, or even providing enough voltage to drive an LCD or OLED display can be avoided.
To further boost system efficiency, the MCU core and digital peripherals of the new family run from an internally regulated 1.7V supply, consuming only 170μA/MHz at speeds of up to 25MIPS. Figure 2 gives a simple overview of the power architecture within this new MCU family.
Providing an efficient, integrated power-supply system is not, of course, the whole story. The different operating modes and switching times, as well as the analog, digital, and communications peripherals, can all affect the overall power consumption of the system.
The most obvious datasheet parameters of a low-power MCU include the standby and activemode power consumption figures. Manufacturers often quote a figure of milliamps per megahertz (mA/ MHz) to account for different clock speeds used with the device.
In relation to this, when looking at active power consumption, though it may seem counter-intuitive, a higher active clock speed is more efficient in terms of average power consumption than running a microcontroller at a much lower speed. A CMOS processor is typically much more efficient at the faster end of its operating capability, and can then spend more time in a low-power standby or shutdown mode.
For the same reason, a welldesigned, fast ADC can also provide efficient system measurements. The ADC’s speed in a given system may, however, be restricted by higher input impedances that ultimately require somewhat longer acquisition times.
In addition, for consistent ADC results in a battery-powered system, it’s common to use a separate reference voltage, which is sometimes integrated into the MCU. However, the system’s efficiency can be compromised if a fast ADC, capable of giving a result in a few microapplications seconds, has to wait several milliseconds for the voltage reference to stabilise.
The ADC and Voltage Reference modules used on the new devices from Silicon Labs feature very short wakeup and processing times. The high-speed internal voltage reference is stable within 1.7μs, and is therefore ready as soon as the MCU wakes up. This allows the 300ksps 10-bit ADC to begin conversion immediately.
It’s common within mixed-signal MCUs for the relatively simple analogue comparators to be interruptdriven, capable of waking the device, and operating somewhat independent of the processor core. However, further power efficiencies also can be realised by adding some degree of “autonomous” operation to the ADC module.
The latest Silicon Labs ADC module supports a burst mode, performing a series of 16 conversions and automatically accumulating the result without MCU intervention. It also supports a window-comparator mode, which entails interrupting the MCU when the results are within a particular “window” of values of interest. Furthermore, it’s able to synchronise to the “quietest” part of the dc-dc converter’s operating cycle.
Alkaline is not the only battery Several single-cell battery chemistries can provide an input voltage between 1.5 and 0.9V for the integrated dc-dc converter of these MCUs. This includes almost all AA and AAA style batteries—alkaline, NiMH, NiCd, and lithium primary among them—as well as zinc-air and silver-oxide “button cells.” However, for other battery types, the nominal battery output is higher, such as lithium “coin cells” with a voltage of between 3.0 and 2.0V. In addition, there may be other reasons for using a higher supply voltage. Such applications can still take advantage of the ultra-low power consumption and efficiency by configuring the device in a “two-cell” mode. Referring again to Figure 2, you can see that the dc-dc converter can be disabled completely, allowing the MCU to support an input voltage of between 1.8 and 3.6V.
Estimating system battery life With Silicon Labs’ “Battery Life Estimator,” designers can estimate the battery life of a new design (the software is freely downloadable from the company’s site). For any system or application, when given the designer’s choice of battery type, and the basic powerconsumption parameters in the “Discharge Profile” (Fig. 3), the software compares single, dual series, and dual-parallel battery configurations in terms of total battery life. It takes into account selfdischarge and shelf-life effects.
The output of the software is a graph showing battery voltage against time, and a numerical estimate of battery life (Fig. 4).
By using and modifying the saved Discharge Profiles with measured or estimated values, designers can evaluate the long-term effect of different system features with any battery configuration. Moreover, they can even compare competitive MCU solutions.
By combining efficient and optimised power components with an MCU device, it’s now possible to have an ultra-low power SoC that can run from a single battery cell all the way down to 0.9V.