Besieged by conservation issues, utility metering is undergoing somewhat of a renaissance—particularly electricity metering. That's because electrical distribution systems everywhere are fragile. Metering used to be the nearly exclusive domain of EEs with a power specialty. Today, it's wide open to chip, board, and system designers, as well as software writers.

The reluctance-motor electrical-meter movement is more than 100 years old. Now, it's getting displaced by chips that aren't just cheaper and more precise, but also can separate the active and reactive components of power and provide gateways to remote reading.

The metering business resembles the automotive business. While relatively few players dot the landscape, they're wide open to innovations that will give them positioning with the utilities that comprise their market. Also, like automobiles, once you've sold a chip or technology into a product, it represents high-volume sales over a long end-product life cycle.

One example of the new kind of residential meter hangs in the service entrance to my own house (Fig. 1). When we put in solar panels last summer, the contractor helped us negotiate new electric rates with the power company. We're charged for electricity we use, or credited for electricity we generate, at approximately 28 cents/kWh during peak hours. During non-peak hours, the rate goes down to 7 cents. The meter keeps track of both figures, and we settle up once a year.

The photo shows a port on the meter for rapid reading. With it, our meter reader needn't wait for the LED panel to cycle through five displays, but that's pretty low-tech. The trend is toward remote reading, which opens up many possibilities. RF drive-by reading has been used for some years. But it's becoming more sophisticated, with newer wireless technologies generating faster and more secure uploads.

Every utility customer needs at least one electric meter. But many people don't realize that there's really high volume potential in submetering—meters that users purchase to read their own electricity usage.

Submetering was once mainly the province of apartment buildings. Where allowed, building owners would negotiate an industrial contract with the electric company and bill tenants on their actual energy usage. More recently, submetering has become important to large manufacturing companies.

It shows managers how a site's total energy consumption is distributed among the various departments, tenants, or processes within the building or facility. Also, it helps businesses with peak-shaving, load-shedding, aggregation, and other measures that lead to lower energy bills.

Of course, management has to know what to do with the data. There continues to be an opportunity to refine and simplify data collection and storage, as well as the user interface that enables people to display and manipulate data and manage the systems that use electricity.

One example of a medium-sized player innovating in the sub-metering business is a Swiss company, LEM. Its Wi-LEM system lets users create inexpensive, reconfigurable ZigBee-based submetering systems. Assembly takes little more effort than clamping current probes around power leads and mounting the Wi-LEM hardware. The system consists of energy meter nodes (EMNs), mesh gates (MGs), and mesh nodes (MNs).

Each EMN attaches to the electrical wiring for the system or machine it's monitoring with split-core transformers. The EMNs measure active and reactive energy, maximum current, and minimum voltage at 5- to 30-minute intervals. The MG is a standalone ZigBee gateway that manages its EMN network, which is a wireless mesh configuration. Any MG can manage up to 240 EMNs while storing the latest data from its network.

Because communication between an EMN and MG is usually limited to a 25-m line-of-sight range, LEM offers MNs, simple repeaters that extend the network's range as much as necessary. Between a local mesh of EMNs and more remote EMNs communicating through MGs, users can build systems that cover an entire manufacturing center or residential facility.

Dating back to the 1890s, the electromechanical induction meter most of us grew up with is based on reluctance, or an eddy-current motor. A horizontal metallic disc rotates in a field supplied by a permanent magnet. Induced fields from ac voltage and the supplied current create a torque that will rotate the disk. Eddy currents from the disc's rotation through the permanent magnetic field retard that rotation.

The ac line voltage and the power being drawn both affect the disc's rotational velocity, which is therefore proportional to the power demanded by the circuits connected to the meter. The disc rotations increment a multidial clockwork mechanism that maintains a record of energy consumption. Commonly, one disc revolution represents 7.2 Watt-hours (Wh).

The international standard for electricity meters is IEC 61036. It specifies environmental requirements such as how much power the meter itself can dissipate and how much voltage it must tolerate, along with performance specs such as accuracy and electromagnetic compatibility.

IEC 61036 meters fall into Class 1 or Class 2, based on accuracy. Further subdivisions in these classes are based on the nominal operating voltage (UN), the base current used for most measurements (IB), and the highest current at which accuracy is guaranteed (IMAX) (see the tables 1, 2, 3: "IEC 61036 Electrical Requirements").


A great deal of design help is available from chip vendors that manufacture metering chips (Fig. 2). Plenty of information is online for engineers who want to evaluate their options. Analog Devices, Cirrus Logic, Maxim, Microchip, and Texas Instruments all offer chips expressly for digital metering. But it's also a business for fabless semi companies such as Teridian (Fig. 3).

To understand what's inside a basic metering chip, consider Analog Devices' ADE7752, which reads only active power. ADI calls it a "polyphase energy metering IC with pulsed output," meaning it's strictly a measurement device. Intended for wye- or delta-connected three-phase sources, the chip incorporates a total of six 16bit, second-order, delta-sigma analog-to-digital converters (ADCs). The bandwidth of the active power measurement is 14 kHz with an oversampling rate of 833 kHz.

Externally, a voltage sensor could be a potential transformer or a resistive divider (Fig. 4a and 4b). The transformer, of course, provides isolation from the main voltage. With the resistive divider, the chip input is biased around the neutral wire. The virtue of the divider approach is that adjusting the ratio of RA, RB, and VR offers an easy way to calibrate gain.

For current sensing, ADI recommends a current transformer for each channel (Fig. 4c). The common-mode voltage for the current channel can be derived by center-tapping the burden resistor (RB) to the metering chip's analog ground. The transformer turns ratio and the value of RB are selected to provide a peak differential voltage of 500 mV at maximum load.

After the voltages and currents are digitized, high-pass filters remove the dc component from the current signals, eliminating any offset effects. Instantaneous power is obtained by multiplying the current and voltage signals of each phase. To extract the active power component, the instantaneous power signal on each phase is low-pass filtered. The results are added to obtain the total active power.

The metering chip's output frequency, which is proportional to the average active power, is obtained by accumulating the total active power information. The average active power information then can be accumulated to obtain active energy information.

The ADE7752's low-pass filtering approach only records active power and ignores reactive and harmonic currents. Figure 5 shows the unity power factor on top and a condition with a purely reactive displacement power factor (DPF) = 0.5 (current lags voltage by 60 ) on the bottom. In the simple case where the voltage and current waveforms are sinusoidal, the active power component of the instantaneous power signal (the dc term) is one-half the voltage times current times the cosine of phase displacement.

Extending that to the case of any non-sinusoidal voltage and current waveforms and separating those waveforms into their Fourier components: where (see equation):

v(t) is the instantaneous voltage
VO is the average value
VN is the root mean square (rms) value of voltage harmonic N
αN is the phase angle of the voltage harmonic.

Similarly (see equation):

That is, the harmonic active power can be obtained by summing. For more information about the ADE7752, see ADI's application note AN-641 at

Metering gets more interesting when it's necessary to consider power factor (PF), as it is in the European Union as well as in China and parts of India. Europe's power-factor-correction (PFC) standard is EN 61000-3-2, with Amendment A14.

Before switching power supplies became common, power factor was associated with reactive loads. Power companies dealt with it by installing big capacitors at switching yards and substations. Switching supplies changed that situation by introducing currents on the power line that are out of phase with the voltage and consist of multiple harmonics of the power-supply switching rate.

Electrical utilities care about power factor for two reasons. First, the out-of-phase power component represents system capacity that isn't available to do real work. Second, utility regulations prevent the supplier from charging for the out-of-phase "imaginary" component.

EN 61000 PFC's objective in switching supplies is to limit the magnitude of the individual harmonic currents up to the 39th harmonic. Amendment A14 relaxes some requirements, but not for personal computers, monitors, and television sets.

For a look at how energy-metering chips separate reactive power from active power and report it, consider Cirrus Logic's CS5467, which comprises a four-channel ADC and a computation engine. The chip has two current channels and two voltage channels.

As with ADI's ADE7752, external voltage- and current-sensing elements generate signals that are amplified and applied to delta-sigma modulators (second order in the case of voltages, fourth order in the case of currents) inside the Cirrus chip. A cascade of Sinc3 and third-order infinite impulse-response (IIR) filters lies on the outputs of the data converters. The Sinc3 decimates, and the IIR compensates for the 5-magnitude roll-off of the low-pass decimation filter.

The Cirrus chip adds a calculation of apparent power, based on rms calculations on multiple instantaneous voltage and current samples. Apparent power is the combination of active and reactive power. Power factor is the active power divided by the apparent power. The active power determines the sign of the power factor.

Also, the CS5467 calculates the reactive power as the square root of the difference between the square of the apparent power minus the square root of the apparent power. Active, apparent, reactive, and fundamental power get updated every computation cycle.

To achieve average reactive power, the chip averages the voltage and multiplies that value by the current measurement with a 90 phase difference between the two. The 90 phase shift is created by another IIR digital filter in the voltage channel. The filter provides exactly 90 phase shift across all frequencies and uses the ratio of the input line frequency to the sample frequency to achieve unity gain at the line frequency.

Subsequently, the instantaneous quadrature voltage and current samples are multiplied to obtain the instantaneous quadrature power. The product is then averaged over N conversions. For details, see the CS5467 data sheet at

What's new in meters? According to Microchip, which makes standalone meter chips as well as chips that work with its PIC microcontrollers, utilities are primarily driven by the need to eliminate the costs associated with periodic meter readings by employees who record data meter by meter.

In environments that invite power theft, such as apartment buildings, the meters no longer are located on the outside wall of the building. Instead, they're in a locked closet inside the building. Human "meter readers" plug their apparatus into a box on the side of the building to capture data from all of the meters.

Plug-in remote reading isn't just for large residential structures. Easily accessible but tamper-proof interface boxes are appearing in single-home construction as well. They make the human meter reader's job somewhat more efficient and eliminate much of the potential for interactions with guard animals, overgrown rosebushes, and paranoid homeowners. The next step is to make the process wireless and keep the meter-reading human on the street or sidewalk.

Microchip says that as theft of services becomes more of a concern, some meters are including a separate ADC to continually monitor the line voltage to detect incidents. For instance, it could detect if the meter is temporarily disconnected or has its terminals swapped (to make it run backward).

Meters that use transformers, rather than shunts, for current measurement are susceptible to core saturation using external magnets. If that's perceived to be a potential problem, the meter may be given a shunt and a transformer, and the two can be checked against each other.

Microchip's meter-reading products aren't just adaptations of PICs with standard microcontroller peripherals. The company's latest replaces the customary RS-485 interface with a serial peripheral interface. Figure 6 shows a three-phase reference design based on three of these MCP3909 ICs, plus a PIC18F2520 and a PIC18F4550 microcontroller.

The PIC18F2520 performs the power calculations, while the 4550 provides a USB interface to desktop software. A software package that comes with the reference design enables meter calibration.