John MonteithToday we hear a lot of talk about energy harvesting and alternative power sources. To make effective use of these energy sources, everything in the entire system must make efficient use of power. While today’s microcontrollers are very efficient when using the digital sections, when analog measurements must be made, more current is required.
To describe where we may be able to conserve power, I will start with describing a typical application that you may find connected to an energy harvesting device. Consider a remote wireless node consisting of a strain gauge monitoring the structure of a bridge. Such a note is traditionally implemented using Lithium batteries, but there is increasing interest in replacing the batteries with energy harvesting devices that eliminate the need for costly maintenance.

Obviously, the biggest portion of the power budget of a system like this is the wireless link, but there are other less obvious users of the power budget that must be considered. Power management of the microcontroller’s resources is a key area that must be addressed. This is as much a software task as it is a hardware task. Efficient power management on the microcontroller may not be enough to meet the power budget of the overall electronics system, so other parts of the system may have to be examined.

One area that must be examined is the strain gauge itself, which is typically a Wheatstone bridge configuration. In an effort to reduce current, many companies have gone from the traditional 350 ohm strain gauge to 1K ohms or higher. But many still power the bridge all of the time in an effort to stabilize the bridge due to self heating. In simple terms, if you are applying 3 volts to the bridge, you will have 3mA of current being drawn all the time. In an effort to reduce current even further, some engineers have gone to a scheme where the bridge is switched on long enough to stabilize, then the measurement is taken, and then the bridge is switched off. While a better alternative than having the bridge powered all the time, the 3 mA is being drawn for a relatively long time.

A better solution might lie in looking at the system from a totally different angle. Instead of measuring the analog voltage output of the Wheatstone bridge that has been allowed to stabilize, we can look at the measurement in the time domain by combining the resistors of the bridge with a capacitor while only energizing the strain gauge for approximately 100uS, thus saving power and minimizing self-heating effects.

This is where looking at alternative hardware architecture like the Time to Digital Converter may provide a better solution. The Time to Digital Converter can be implemented in a low power, all-CMOS process and can be employed to make ultra low current measurements, while still preserving the precision of the measurement. Utilizing the Time to Digital Converter, traditional real-world parameters that can be resolved to measurements of resistance, capacitance or inductance can be done with high precision.

Back to our real-world example, if you combine the resistors of your bridge with a capacitor, resistance can be measured to very high precision by measuring the change in discharge time when the strain gauge is flexed. This change in resistance can then be resolved in much the same way as you would with a traditional ADC solution, but looking at it as a change in time instead of a change in voltage. The largest gain in power savings comes from the fact that the strain gauge is only powered for a very short time, 100uS. This dramatically lowers the whole system power consumption making it possible to use energy harvesting.

Conceptually, the circuitry would consist of a charge control switch to charge the capacitor to a pre-determined set point, whereby the charging switch opens. Simultaneously the Time to Digital Converter is started and the discharging switch is closed to discharge the capacitor through the resistor being measured. Measurement would run until the capacitors voltage reaches a pre-determined end point and the Time to Digital Converter is stopped. A typical discharge cycle is around 100uS, so very little energy is used to power the resistive circuit during measurement and the Time to Digital circuitry only runs long enough to capture the data, process it and spit it out to the microcontroller.

While this example looks at measuring resistance, other real world parameters that can be resolved using capacitance and inductance can be measured in a similar manner. The Time to Digital Converter is a very flexible measurement method that has a lot of potential applications.