FROM EDN EUROPE: Energy consumption demands electronic metering

A spin-off from Galileo Ferraris' discovery of the induction motor back in 1885, the electromagnetic wattmeter is one of the few remaining electromechanical devices of the Victorian age. Its operating principle relies on the interaction between two magnetic fields that are proportional to voltage and current, causing a metal disk to rotate at a velocity that represents circuit power. Its longevity owes much to traditionally static electricity demand patterns from applications such as heating and incandescent-bulb lighting—about one-third of America's domestic hot water still takes its heat from resistance elements—together with low installation cost and long service life. Increasingly however, today's consumers present electricity suppliers with a massive burden of reactive and nonlinear loads that electromagnetic meters can't adequately measure. For example, electric motors account for a staggering 40% of today's global electricity consumption. The power flowing through such loads comprises the active component that performs work in terms of heat, light, or mechanical energy conversions, as well as the reactive power element that's necessary for device operation—for example, to magnetise a motor's stator.

The implications that reactive loads have on supply-network stability can be profound. To visualise this, consider a purely resistive circuit, where voltage and current are exactly in phase and the power waveform is a positive-value sinewave of twice line frequency, with the resistor dissipating all the energy that the supply delivers (Figure 1A). By contrast, reactive power is the "phantom power" that flows in capacitive and inductive circuits. In a purely inductive circuit, the voltage and current waveforms are 90° out of phase, and the power waveform is a sinewave of twice line frequency that centres on zero—so the inductor is alternately absorbing and returning power to the line (Figure 1B). Neglecting shunt capacitances, loads such as motors comprise a resistance in series with an inductance, so some percentage of the supply-line energy dissipates in the resistor while the balance alternates between absorption and return, moving the current waveform away from zero.

Everyday nonlinear loads from fluorescent lights to home electronics worsen the situation by drawing current peaks and injecting harmonic interference into the supply—so much so that the European Union, for instance, classifies and regulates equipment within its EMC legislation that often compels manufacturers to include active power-factor correction circuitry (Reference 1). Crucially, electricity suppliers must provide more current for power factors below unity, creating greater losses in the transmission system and requiring additional capacity to maintain system stability. This fact is driving some suppliers to consider tariff structures that account for the consumer's power-factor environment. Utility operators accordingly require more comprehensive metering capabilities than electromagnetic meters can deliver—such as the ability to measure harmonic currents and to record and store peak load profiles. These capabilities demand electronics.

The market potential for energy-meter (or e-meter) ICs is vast. Developing countries such as China and India are rolling out new networks, while utilities across the industrialised nations continually replace meters that have met their lifetime limits. According to STMicroelectronics (ST), manufacturers will produce around 53 million e-meters in 2005, 70% of which will be electronic meters. Predicting an 8% compound annual growth 2005 to 2007, the company says that the most important markets are China (16 million units/year), India (11 million), and Europe (8 million), but that new markets are also emerging in Latin America and the Middle East. The European market requires medium- to high-end meters that include automated meter reading, LCD displays, and multi-tariff billing, while the developing regions mainly demand low-end meters that only perform energy calculation. Other vendors targeting e-metering include Analog Devices, austriamicrosystems, Cirrus Logic, TDK Semiconductor, SCL India, and Texas Instruments. Of these, Analog Devices is today's leader with around 100 million e-meter IC shipments to its credit.

Measurements of interest include instantaneous active power, which is the product of the current and voltage waveforms at any time. The systems that follow exclusively use ADCs, sampling and integrating to compute results; for instance, one method to derive rms values squares the instantaneous sample values, averages a number of them, and square-roots the average. Multiplying these rms current and voltage values yields apparent power in VA (Volt-Ampere) units and represents the maximum mount of real power that the load consumes. Integrating the instantaneous active power sample values for an integer number of line-cycle periods for one second yields active average power—or the real power flow rate in Watts (that is, Joules/sec). Energy is the time-integral of power, which e-meters accumulate in kW hours to reflect the consumer's energy consumption. Power factor (PF) is the active-to-apparent power ratio, which varies from unity for resistive loads to zero for a purely reactive load. Neglecting any harmonics that invalidate the relationship, the angle between active power and apparent power in a single-phase circuit represents the phase shifts that reactive loads cause (Figure 2). Sampling systems can calculate reactive power by phase-shifting either the current or the voltage signal by 90° and then multiplying the instantaneous values. Performing this calculation for the same sample group as the active power calculation yields reactive power in VA reactive (VAR) units, permitting direct comparison between the two quantities.

Mark Strzegowski, marketing manager of Analog Devices' energy metering product line, observes: "With today's resistance to building more power stations, operators are looking for smarter ways to ensure the stability of their supplies other than simply adding another generation facility." Strzegowski affirms that comprehensive measurements and better communications between the e-meter and utility are essential to optimise network performance. Communications techniques range from an industry-standard infrared link to low-power RF readers that allow the meter reader remote access, while Italian utilities are using the power line to implement remote metering schemes. Strzegowski says that electricity suppliers are also becoming more security-conscious, so anti-tamper techniques have become another essential feature of contemporary e-meter ICs. Further considerations include the range of current sensor interfaces that an e-meter IC supports (see sidebar "E-meters drive current sensor evolution").

While a single IC can measure the key energy values using conventional signal processing techniques, the analogue interface that extracts voltage and current information has strict accuracy requirements over a wide signal range. In Europe, the recent IEC-62052/62053 metering standards that supersede the IEC-61036 family refer to a meter's class. The US has its own set of standards in ANSI-C12, while much of the rest of the world echoes IEC-61036—for example, Saudi Arabia (Reference 2). Within the IEC scheme, a Class 1 meter is accurate to within 1% of reading in unity power-factor, 45-65 Hz environments. It maintains this accuracy for 0.5 PF (capacitive) and 0.8 PF (inductive) conditions. The specifications distinguish between directly-connected meters that use current shunts and meters that use current transformers, and make additional accuracy allowances for temperature range, magnetic fields, and harmonic content in waveforms such as those that half-wave rectification, burst-control, and phase-control produce. The limit of error is 3.0% for Class 1 meters in the last three of these tests, and the meter must also maintain its linearity within 0.8% for a 5^th harmonic waveform with an energy content equal to half its maximum measurement range.

End-users find some of these quantities hard to interpret, complicating comparisons in meter performance between alternative suppliers. This difficulty—along with the fact that the IEC standards only apply to the Watt-hour tariff meters that utility companies supply, and not to the secondary meters that many factories and businesses install to manage their energy consumption—forms the rationale for another specification. BS-DD8431, where "DD" signifies draft-for-development, is under review at the British Standards Institution (BSI) to complement IEC-62052/62053 and harmonise meter specifications. It proposes several crucial changes, such as Classes referring to an e-meter's absolute accuracy and removing references to frequency. The intention is that these proposals will gain international acceptance and become a global standard. And standards are key to industry—according to a joint study by BSI and the UK's Department of Trade & Industry that reveals statistics for the first time, standards contribute £2.5bn annually to the UK's economy (Reference 3).

Reflecting the limitations of first-generation electronic e-meters, the established standards distinguish between meters that integrate readings over an arbitrary period and those that make continuous measurements. The most straightforward approach to maintaining signal bandwidth is to sample relatively fast, and this is the approach that ADI takes with its chips. Strzegowski says that with no decimation, the fixed-function signal processing architecture that appears in products such as the ADE7753 preserves the chip's 14-kHz input bandwidth. He notes that some other techniques won't record harmonics—or worse, can become unstable in the presence of signals with large harmonic content. Etienne Moulin, product manager of ADI's energy metering product line, adds: "One of the biggest challenges is to minimise the crosstalk between the relatively large signals that appear on voltage channels and the small ones that current transducers yield." He points out that with shunts of around 200 µΩ, sense voltages lie in the 10s of µV at low currents, but asserts that careful layout can constrain crosstalk below 120 dB within a 60-dB dynamic range. Moulin says that ADI's high-precision ADCs and fixed signal-processing architecture processes as much information as possible in the digital domain, thereby improving accuracy and minimising long-term drift, as the minimalist ADE7757 demonstrates (Figure 3): "Our implementation only requires circuit designers to adjust their systems for gain and offset errors."

Suiting single-phase measurements, the similar but more capable ADE7753 comprises two input channels, each of which has a programmable-gain-amplifier (PGA) with a maximum gain of 16. Two second-order sigma-delta modulators follow that oversample at 894 kHz, allowing digital filters to push the noise spectrum far beyond the 40 Hz–2 kHz bandwidth of main interest. Digital averaging transforms the raw bitstream into signed twos-complement 24-bit data, which is optionally available to the host microcontroller. A selectable digital integrator in the device's current path allows the use of Rogowski coils, as well as current shunts and transformers. The signal-processing hardware continuously computes the rms values of the voltage and current channels, and a phase-shift adjustment is available to ensure that the values align correctly. The hardware computes the instantaneous active power from these samples, from which it can derive all other transformations. To compute instantaneous reactive power, a phase-shifter displaces the current channel's signal by 90° to the voltage signal, again using line-cycle integration to accumulate average values. An SPI connection allows a host microcontroller to retrieve and display measurement values. Like most of its peers, ADI offers a range of evaluation boards that implement reference designs and encourage experimentation (see sidebar "Dev kits speed time-to-market").

Frank Forster, business development manager for catalogue microcontrollers at Texas Instruments, emphasises that adding value is the key: "Utilities don't just want to replace Ferraris meters, they want to maximise the opportunities that electronics allow to reduce the total cost of ownership." By adding a dedicated e-meter front-end to its MSP430 RISC core, TI built a single-chip e-meter that frees designers to concentrate on adding facilities such as bespoke displays and communications links. Forster comments that communications abilities not only reduce or dispense with human meter readers but also allow new capabilities, such as load management. Bidirectional communications permit the utility to actively manage the consumer's loading, for example by only connecting washing-machine and dishwasher circuits at off-peak times. Requiring a powerline modem to form the gateway into the household, such abilities may span automatically switching tariff structures to encourage off-peak use to disconnecting consumers altogether if the load level compromises network stability—a scenario that may especially apply within developing nations. Forster says, "Because the dedicated front-end leaves the MSP430's CPU with 0% loading, designers can implement the complex algorithms that applications such as powerline modems require." He adds that with meter tampering running deeply into double-digit percentages in some territories, such communications also provide additional security for the utilities.

The analogue-front-end (AFE) of the MSP430FE42x family—the ESP430CE1—includes three 24-bit second-order sigma-delta ADCs that each return a 16-bit result for computation (Figure 4). These ADCs sample in parallel at up to 4096 samples/sec, taking 1 second at this rate to return new measurement data. Forster explains that while only two ADCs are necessary to compute the energy flow in a single-phase application, a third ADC can help implement anti-tampering schemes in hardware: "We can detect mains voltage disconnections and reversed meter connections with just two ADCs, but you also need to measure the neutral return current to detect meter frauds such as earthing the load." Fraudsters often attempt to disconnect the mains voltage to the IC to stop it from recording, which stratagem designers can counter by adding a coil around the current path to provide a secondary supply voltage. By reversing either the voltage or current connections, fraudsters attempt to create negative energy to "rewind" the meter; the MSP430 detects this condition, allowing designers to take appropriate action in software. By substituting the earth connection for neutral or by selectively earthing some loads, fraudsters divert current from the metering circuit to lower readings, hence it's necessary to compare live and return-path currents (this ability also detects partial earth faults, thereby improving consumer safety). If the meter uses a current transformer, another possible fraud involves placing a large magnet close to the current transformer to quash its ability to sense current. One solution places a current transformer in the live side and a shunt in the neutral phase, again comparing each phase's flow.

An internal temperature sensor allows the ESP430CE1 to support temperature compensation algorithms, which with characterisation can minimise temperature dependencies within the transducers. To interface with current shunts and current transformers, the AFE also includes programmable-gain-amplifiers (PGAs) that operate over a times-one to times-32 range. Alternatively, designers may prefer Rogowski coils. Communications between the AFE and the CPU employ mailbox registers, and the AFE has its own parameter registers and hardware multiplier that enable the subsystem to handle the data processing. Some 28 parameter registers allow the CPU to set the AFE's response to cater for corrections factors such gain, phase, and offset errors, and to set thresholds such as anti-tamper imbalances between live and neutral phases. The 16-bit MSP430 RISC core is available with 8, 16, or 32 kbytes of Flash, and has a 128-segment LCD driver peripheral and dedicated memory that support static and multiplexed displays. Essential on-chip elements include a crystal oscillator module, watchdog timer, brownout detector and supply supervisor. There's also a bootstrap loader, USART, basic timer and timer-counter/capture-compare block, as well as two 8-bit I/O ports. Software can configure the chip between full operating mode and five low-power modes that successively disable on-chip clocks to reduce the 500µA maximum active current to 2.5µA (0.1µA typical at room temperature).

Its brand-new STPM01 chip marks ST's entry into the e-meter marketplace. This IC features a single-point calibration adjustment via one-time-programmable memory, and the ability to tackle single and three-phase applications from Class 0.2 using an external crystal oscillator to Class 1 using the on-chip RC option. It includes the normal pulse-output interface and supports current shunts and transformers, as well as Rogowski coils. An SPI port allows a host microcontroller to read active, apparent, and reactive energy, line frequency, and rms current and voltage values. The chip also has a zero-crossing detector output that allows external logic to synchronise load switching with ac-line's zero point to constrain interference generation. An on-chip voltage regulator eases power-supply design in lowest-cost implementations, and the part guards against some twenty fraud/tamper strategies.

Perhaps most interestingly, this chip employs a different approach to active power calculation that relies on an algorithm from ST's partner, Slovenian e-meter manufacturer Iskraemeco. The conventional approach multiplies voltage and current samples to yield instantaneous power values. In practice, these comprise a dc term that represents instantaneous active power together with an ac term that low-pass filters endeavour to suppress. Gianfranco Di Marco, application engineer for system architectures at ST's central application laboratory, explains that because these filters don't have perfect brick-wall responses, sinusoidal ripple of typically around –22 dB remains, requiring averaging to smooth in order to accurately calibrate the meter. By contrast, the STPM01 feeds the digitised voltage signal—which is dynamically far more stable than the current signal—through a differentiator (Figure 5). This differentiator isn't required for use with Rogowski-coil current sensors and is therefore user-selectable. Subsequent stages perform multiplication, integration, subtraction, and division, with the last stage intrinsically cancelling out the ac term to leave just the active power term. Di Marco notes that the benefit is an increase in calibration speed due to no requirement for averaging—a significant consideration in high-volume manufacture.

Ironically for a system that by definition sits beside a bountiful energy supply, low power consumption is essential for e-meters. From a formal viewpoint, ADI's Strzegowski points out that the IEC regulations make it undesirable for domestic meters to exceed 2W and 8VA for circuits that connect to the ac-line voltage supply. From another perspective, TI's Forster says that if the utility operators replaced all the Ferraris meters in Germany with e-meters that use his company's MSP430-based solution, the country could decommission one nuclear power plant. Forster reckons that electronic e-meters achieve power savings of around 90% compared with electromechanical devices—and it's the utilities who foot the bill for this power, so this term forms another part of the total cost-of-ownership equation.

From a circuit designer's viewpoint, ever-lower power consumption also makes it far easier to derive transformerless supplies from the ac line. For example, experiments here show that it's an uphill struggle to derive more than about 20mA at 5VDC from 230VAC 50Hz supplies using a traditional capacitor/resistor dropper circuit. This technique—which dates back to valve radios and exploits the fact that voltage and current are 90° out-of-phase to constrain power dissipation in the resistor—often suits circuits of up to about 50mW that don't require galvanic isolation (Reference 4). It's not necessarily the most efficient arrangement, as some current always flows in the shunt Zener that stabilises the supply line and protects the IC from transients. Above this 50mW level, it's back to low-power transformers that typically provide around 40% regulation, or time to consider off-line switch-mode topologies from vendors such as Power Integrations and Supertex. The Power Integrations parts family employs high-frequency isolation transformers to supply loads of several Watts, while the Supertex SR036/037 devices deliver up to 1.5W from an external MOSFET without transformers or inductors.

Given software-selectable power-saving modes, it may be cheaper and more resourceful to consider using a low-power main supply with an ultracapacitor from vendors such as Maxwell Technologies. The company already has design wins for an ultracapacitor-based approach with US e-meter manufacturer SmartSynch, where its components save on costs for lead-acid or lithium cells that many meters use for memory backups, extending the e-meter's lifetime to 10 years and more. One possibility might be to adapt a low-current dropper circuit to tricklecharge the ultracapacitor during power-save mode, and use the stored energy to support burst-mode full-power operation. The availability of electromechanical components from China and India perpetuates mechanical impulse counters as many designers' choice of totaliser, as is obvious from the number of e-meter ICs that still offer pulse-driver outputs. Such counters are super-cheap, easy to drive from simple pulse outputs, and intrinsically retain their readings when the power fails.