How to design smart gas meters to maximize energy efficiency

Electronic water and gas meters are typical representatives of the most challenging low-power designs for embedded control systems that require RF connectivity. These systems are characterized by battery power (for example, wall power is generally not available at the gas and water meter installation points) and require a battery life of more than 20 years. Utility suppliers make this requirement because the cost of expert maintenance alone exceeds the entire cost of the smart meter. Due to the design requirements for ultra-long life, almost all water and gas meters use lithium thionyl chloride (LiSOCl2) chemical batteries, which can have a service life of more than 20 years in the meter due to its very low self-discharge characteristics. However, this battery is expensive (about $1.5/Ah), resulting in a battery BOM cost of up to $10~15 in a single water or gas meter.

Many smart meter vendors decide to differentiate their products by extending their communication range. In their system network topology, a certain number of meters send usage and billing information via sub-GHz networks to repeaters installed on utility poles. Repeaters collect aggregated information and send it to utility service providers via cellular networks or other backhaul channels. Repeaters can support about 1,000 meter nodes. However, repeater costs are often 10 to 100 times the cost of a single meter node. Meter vendors often face pressure from their customers to reduce the number of repeaters in the network. The most practical way to solve this problem is to improve the robustness of the transmitter (TX) link.

There are many ways to improve the TX link budget. One of the most obvious solutions is to increase the transmitter output power using a power amplifier (PA). However, this approach is also the most expensive in terms of battery life. Another solution is to enhance the protocol to minimize message errors and the number of retransmissions that follow. Although this technique saves more power than simply increasing the PA, it still increases the current power budget by about 40%.

Assume that the redesigned smart meter has the following three design requirements:

• More than 40% of the power budget is allocated to TX functions to increase coverage

• Maintain existing LiSOCl2 battery size (A) and capacity (3650mA-hr)

• Maintain the existing battery life for 20 years

The strategy is clear: increase power within the TX budget, but do not increase the overall power budget, which means that power consumption in other functional areas must be reduced, such as RX, active mode, and sleep mode budgets. Figure 1 shows the original power budget and the target budget after the redesign.

Figure 1. Power budget for smart meter applications vs. higher voltage conversion efficiency

To increase the performance of CMOS circuits and reduce their power consumption, chip designers usually use the smallest practical device to build integrated circuits. Generally, embedded processors and RF transceivers are designed using 0.18µm, 0.13µm or even 90nm processes. A key indicator for reducing device power consumption is to reduce the internal operating voltage, thereby reducing CVf switching losses.

Even if a battery -powered device supports a 3.6V supply voltage, the device can often operate at much lower internal voltages.

Almost all devices on the market have an on-chip low dropout linear regulator (LDO) that regulates the output to a very low on-chip voltage, usually 1.8V or lower, when the input voltage is 3.6V. In other words, a linear regulator with an input voltage of 3.6V and an output voltage of 1.8V will produce a conversion efficiency of 50%. Obviously, this efficiency will become worse as the output voltage decreases.

More advanced embedded controllers, such as the C8051F960 MCU in Figure 2, integrate switching regulators that are more efficient than LDO controllers. In most cases, this device can switch with efficiencies as high as 85%, which can reduce the overall current drawn from the battery and extend battery life.

Figure 2: Switching efficiency comparison between traditional MCU and advanced MCU

Using this approach, the current RX power budget can be significantly reduced.

That is, the radio receiver consumes approximately 62.5% of the battery current as would be consumed using a DC-DC buck converter instead of just an LDO. The net result of taking this approach is a reduction in the RX current power budget.

With this change implemented, we are close to meeting the new RX power budget requirement (for example, Figure 3 shows a reduction from 30% to 19%, although the target was 18%). Next, it is necessary to continue optimizing other operating modes in the system.

Figure 3: DC-DC switching converter improves RX power budget and lowers sleep mode power consumption

Typically, battery-powered meters spend 99.9% of their time in low- power sleep mode. Therefore, it is critical to minimize the power consumption of sleep-mode circuits. A few years ago, the best devices were able to consume as little as about 1µA using a 32.768 kHz crystal driving a low-power wake-up clock at 3.6V. With further optimization and improvement, today's devices use only about 700nA for the same functionality at the same voltage. Although the net savings is only 300nA, it is actually a significant saving that can be subtracted directly from the power budget.

By using a low-power sleep mode device, the sleep mode budget can be reduced from the previous 8% to 5% (as shown in Figure 4), thus achieving the design goal. However, this is only reaching the goal, not exceeding it, and further improvements are still needed to achieve the overall design goal. The last key point is how to reduce the power consumption in the working mode .

Figure 4. Effect of sleep mode improvements on sleep mode power budget

Reduce active mode power consumption

It is important to distinguish the main power consumption tasks in meter applications. In the example of a gas or water meter given in this article, there are two main tasks:

• To calculate flow, the reed switch status is checked 20 times per second.

• Creates a wireless data packet every 15 seconds and transmits this data to the wireless transmitter for broadcast.

In many metering applications, a device called a register encoder is used to record the flow of gas or water. In the metering system, this appears as a series of switching events or pulses. In traditional metering systems, the CPU must wake up and sample the switch state of the I/O pin. If the switch is a physical reed switch, additional CPU bandwidth is required to debounce the switch and control the pull-up resistor to ensure the pulse is valid and minimize leakage current by closing the switch. Performing this function in software can consume more than 1µA even in the most optimized system.

A better approach is to use a dedicated input capture timer that runs automatically even when the device is in sleep mode. This technique has many advantages over software-based approaches. First, the number of switch times can be accumulated in hardware registers, requiring little CPU intervention. In addition, functions such as switch debounce, pull-up resistor management, and automatic calibration can be integrated directly into the hardware. Using two timer inputs, orthogonal decoding functions can be supported to determine the direction of traffic, giving the system backflow detection capabilities and anti-tampering functions. At 3.6V, even at a sampling rate of up to 500Hz, the dedicated low-power input capture timer consumes only 400nA of current, which is a significant improvement over the method of performing this function in software.

When the CPU is running, instructions are usually fetched from nonvolatile memory (such as Flash memory). It is common for 40% of the active mode current to be used for Flash read operations. Therefore, in any case, using dedicated hardware peripherals (rather than the CPU) to move data can save power . When preparing a packet for RF transmission, the data needs to be edited many times. For example, suppose a 20-byte information payload needs to be transmitted from the meter to the concentrator. Initially, these 20 bytes reside in SRAM; however, this data may contain customer private information and must be encrypted; then, a cyclic redundancy check (CRC) is calculated and appended to the encrypted information; finally, the entire information will be encoded (for example: Manchester, 3:6, etc.) before being transmitted to the wireless transceiver through the serial peripheral interface (SPI). All of these functions can be implemented in software by the CPU. However, using dedicated hardware to perform the task will make the system more efficient, such as the dedicated packet processing engine (DPPE) shown in Figure 5.

Figure 5 Processing time and power savings using the DPPE hardware module

Using DPPE not only reduces the time required to execute functions, but also reduces the current consumed during this time because the Flash memory is not accessed. This ultimately reduces the power consumption in the active mode by up to 90%. When the above improvements are completed, we can exceed the energy saving target in the active mode, and the required power consumption only accounts for 6% of the total budget, as shown in Figure 6.

Figure 6 Results of using DPPE to reduce power consumption of smart meters

By using the above three techniques, we were able to successfully increase the TX power budget to 70%, which is entirely the result of saving power in RX mode, sleep mode and active mode. In other words, we can achieve the overall design goal of increasing TX reliability without using a larger battery capacity or reducing battery life.

The examples shown in this article illustrate how energy savings can be achieved in smart meter applications by reallocating the overall budget. However, energy savings can also be valuable in many other ways. An obvious example is the ability to use smaller, lower-cost batteries . Another benefit is the ability to extend battery life for the same battery . Yet another potential benefit is greater design margin and reduced warranty liability. Consider a scenario where a meter manufacturer produces millions of meters per year, each with a 20-year warranty. If a meter fails after 15 years of use due to excessive power consumption, the manufacturer could potentially be liable for tens of millions of meters. Thus, additional design margin gives peace of mind to both engineers and investors.