Why so much of a buzz over the basic electric power utility meter by hardware and software companies in the electronics industry? The answer lies in what drives the business: money.
Companies around the world are sensing huge business opportunities in power grid design because of two things. First, utilities in US and Europe are embarking upon replacing over 45% of the existing utility meters by 2015. Second, there are huge requirements in terms of units being created by the swelling consumer and industrial base in the developing nations like China, India, Brazil, etc.
According to one projection, there is a requirement of $19.5 billion worth of utility meter and related communications between 2010 and 2015 with an estimated shipment of over 200 million smart-meters . The graph in Figure 1, below from a different source, gives almost the same projections.
The purpose of this article is to serve as a guide to determining the system-on-chip architecture for a basic utility meter that will 1) serve the basic purposes and applications of the present day market: measurement of consumption, tamper protection, time keeping, display and transmission of the meter-reading, and 2) can be used as the baseline upon which future designs will be derived.
The features and modules discussed here should serve only as a starting point – not as the end of the road – for metering SoCs and should signal what lies ahead. Peripherals like USB, Ethernet, etc are jostling their way into the next generation of the metering SoCs, better known as Smart Meters.
Nowadays, a meter does not do just the measurement, protection, etc but can do a lot more tasks e.g they can interact with the consumer via a smart touch screen display and signal him/her the present consumption. They can also monitor each and every appliances of the home/office and tell which one is consuming how much of energy.
Using Ethernet, the consumer can operate the appliances of one place while sitting at any other place on this earth, like switching on the AC right from the office so that the user gets cozy colder home when he/she reaches home.
Also, AMR (Automatic Meter Reading) is going through a long revolutionary way. Now the options could be to install pre-paid meters, take the reading using IR/ZigBee receivers or using Ethernet or GPRS. Smart Grid is also going to revolutionize the world of metering in terms of the transmission, fault location, etc.
In a very real sense, today's basic metering designs are going to be the beginning of a new world of applications in Smarter Metering.
|Figure 1. Meter Demand Projection between 2008-2012.(Source ABS Energy Research)|
All the above mentioned functions of a utility meter can be achieved using the building blocks as shown below in Figure 2 below . As shown, it consists of the following elements:
1) Analog Front End to measure the Current, Voltage (for energy meter) or Heat (for heat meter) or maybe the output of the Analog Sensors (flow, gas meter),
2) A Flow/Gas measurement unit which works using the output from the digital/analog sensors. (flow, gas meters),
3) Tamper protection and detection logic,
4) RTC (real time clock) for time keeping purposes,
5) Communication peripherals to communicate with the external world like with ZigBee transceivers, RF transceivers or another SoCs,
6) Display drivers to display data like meter reading, date-day-time, etc.
7) Core to process the data to give the consumption and to do other tasks,
8). Memory to save the data like meter reading and time of the tamper event, etc.
Now, let us explore each block one by one.
|Figure 2. Building Blocks of the Basic Utility Meter.|
Analog Front End – Sensing and measuring the input
For energy meter purposes, we need to measure ” input voltage, current on all the three phases and current on neutral line ” to measure the consumption. These quantities can be measured easily using current transformers and sensors.
All these are fed to the Analog block which comprises of Programmable Gain Amplifiers (PGAs, may reside outside the SoC, optional), filters and ADCs (Figure 3 below ). PGAs are fed with raw input from the current transformers or sensors which, in turn, feed the ADCs after filtering and required muxing has been done.
|Figure 3. Utility meter analog interface|
This muxing may or may not be a part of the ADC. These ADCs measure the above mentioned quantities and give results to the core (as shown in Figure 3 above). A number of times, the outputs of the current transformers and sensors are not in a range as required by the ADCs to match up to the desired accuracy of conversion.
In such cases, the PGAs are used to scale their input to give desired results. Depending upon the use-scenario and the cost of the SoC, these PGAs can be placed inside or outside the SoC as they consume more power and may also generate a lot of on-chip noise. Filters are used to filters out the noise component from the input signal.
For energy-meter purposes, a BP Filter with central frequency of 50-60 Hz is used to accept the desired band of results. A single phase energy meter would have two ADCs on the SoC, one each for the current and voltage. The number of ADCs increases by one for each additional phase.
The choice of ADC is particularly difficult in this case (metering). The accuracy, power consumption and speed are the primary factors governing the choice of the ADC. The most common choice for ADC is between SAR (Successive Approximation) and Sigma- Delta (SD) ADCs. Both of them have their own advantages and disadvantages.
Therefore, the choice depends largely on application and the budget for the SoC. SAR ADCs use sample and hold technique. They capture the input at a particular time instant and then constantly compare it against internal DAC output, tweaking it further to approximate to the captured input value.
And on each conversion, the corresponding value of the DAC output is digitized and stored in SAR register. SAR ADCs have good resolutions, small first-conversion latency and quite wide range. They are also very sensitive to the changes in the input channel value. They also have a very wide bandwidth for the input channel. However, these types of ADCs suffer from linearity error due to this repetitive subtraction and comparison.
In SD ADCs, the input signal is over-sampled over a period of time and then the desired band of signal is filtered, which is then averaged and digitized. SD ADCs are a type of Feedback Flash ADCs.
They make use of the fast conversion speed of the flash ADCs, which can have conversion time of as low as a few nanoseconds for 8-bit operation. The result of the Flash ADC has a large amount of error in it. This output is then fed-back and subtracted from the input. This results in shaping of the noise and as a result, the noise is reduced .
Therefore, SD ADCs have better noise response than their SAR counterparts as SAR ADCs sample only at a single point. However, the input response of SARs is better than that of SD ADCs. Therefore, SAR is preferred when the application requires fast response and low latency and multi-channel data-acquisition.
However, if high precision and high resolution in noisy environment is required, then SD ADCs are the choice . Usually, in metering applications, the lower-end solutions have SAR ADCs whereas, the high-end SoCs have SD ADCs to give more noise-proof conversion result.
Calculation and Storing the Consumption
The intensive calculations involving the consumption are usually done by the core. The measurements involve measurement of active power, reactive power, load factor and actual power ( Figure 4. below ).
|Figure 4. Voltage, current, power and average power relationship.[Source : Wikipedia]|
Active power is the component of power in which the voltages and currents are in phase. And when the two are not in phase, it contributes to reactive power. Inductive or capacitive loads have their current lagging and leading the voltage, respectively.
Although, the reactive power does not contribute to any power transfer and purely reactive loads have zero net energy flow, this reactive power generates a lot of heat and thus arises the need to have better quality transformers and thicker wires to carry more current and sustain more heat. This increases the cost of the energy distribution.
Therefore, the energy providers levy a penalty on consumers who use reactive loads by dividing the actual power consumption by the power factor. As the power factor (load factor) is always less than one, this division effectively increases the consumer's power consumption units.
That is why the load factor is calculated which, further, is used to calculate the Apparent power. Apparent power is equal to (Real power / Load Factor). And real power is the product of root mean square values of the currents and voltages.
As it is evident from the discussion above, for metering applications the core has got a lot of processing to be done like calculation of the root mean square values of the currents and voltage, their product, and then averaging, etc.
Therefore, the core of choice in metering applications is DSP core as it can do the mathematical operation in a very short span of time. Sometimes, this load of the core can also be off-loaded by incorporating an additional MAC (multiply and accumulate) block in the platform which would do most of the calculations and in the meanwhile, core would be free to do rest of the operations like communications, display, monitoring, etc (primarily in smart metering).
The size of the on-chip memories also plays a crucial role in the pricing. The memory size can go from as low as 256 Byte of RAM and 8kB flash to anywhere like 26kB of RAM and 264kB Flash, depending upon the application. In somewhat more advanced applications, we may require upto 2MB of flash and 512kB of RAM. 
Flow/Gas Measurement Logic
This block may comprise of modules like programmable counters, comparator, pulse width modulator, etc. The rate of flow of the gas/fluid can be measured either using digital rotator sensors (with digital output) or analog rotator sensors (with analog output).
In case of digital sensors, the pulse train output of the sensors feeds the counter module in the block to increment the counter. This counter count corresponds to the consumption of the consumer. The counters can be read in periodic fashion to calculate the consumption of the consumer.
In case of analog sensors, comparators can be used to generate a train of pulses which would then feed the counter module in this block.
Display and transmission of meter reading
Another important aspect in metering application is display and transmission of meter-reading so that the consumer could be billed. LCD 7-segment glass displays are probably the most common method to display the current meter reading. It indicates the present value of the counter and then the person taking the reading can note it down.
In some SoCs, the LCD driver may not be a part of the SoC and an external LCD driver is used. In such cases, the SoC needs to transmit the data to the driver. This can be performed by the communication peripherals present on the SoC like I2C, SPI, UART, etc.
Reading can also be transmitted wirelessly to a remote LCD driver or data logging sub-station via ZigBee, IR, etc. In such cases, we may need to modulate the transmission data before sending it to the transmitter (e.g infrared communication).
This can be accomplished easily using peripherals like Pulse Width Modulator which can generate the desired modulation. The corresponding modulation carrier is then transmitted to SPI which then puts the data on the carrier and transmits it out of the SoC.
Protection against Tamper and Time keeping
Protection against any theft is very vital for the metering applications as it protects the supplier/distributor against any commercial losses. The consumer might attempt to tamper the meter in order to make meter slow so that it gives a reading lower than the consumption.
Therefore, it becomes very important to protect meter against any such adventure. The capability to detect the tamper can be added to the SoC in several ways. However, integrating it along with the RTC (Real Time Clock) can be very effective for metering applications.
This way, the tamper detection as well as its corresponding day-date-time can be stamped effectively using one module only. Any tamper attempt can be logged in the internal memory, along with the time stamp of the tamper event.
The meter can also indicate the same by blinking a particular LED or by displaying the tamper event along with the time stamp on the LCD display. This enables corrective measures to be taken when the meter-reader comes to take the meter reading.
Also, the time keeping also plays a key role as the meter has to do averaging at a regular interval depending upon the applications. For such purposes, the regular interrupts from the RTC can indicate to the core to do the same. As per the guidelines provided by the Open Metering System Specifications, following in Table 1 below are the intervals over which the core does the averaging,
|Table 1. Averaging duration for different applications.|
RTC can also have the mechanism to allow only the authorized access to its registers so that any attempt to hack the RTC registers to disable the tamper detection or other features can be prevented. Also, the RTC should also have the capability to run in battery-mode in case of main-power failure.
This way, the meter would be able to perform its critical tasks like tamper detection, even when no power is there. However, this arouses the need to have a design architecture that would be running on very low current consumption so that we can have a long battery life with a normal operation.
This is of special emphasis in flow/gas meter applications where the meter operates on battery only. In such cases, a minimum battery life of 10-15 years is expected from the SoC. Thus, there are certain “take cares” that should be considered while designing the SoC including 1) the modules that consume power in the stop modes, 2) the boot-up time of the core from stop modes, and, 3) the current consumption in the stop mode recovery.
Also, the current figures in the stop mode operation are also vital as they contribute to the stop-mode consumption of the SoC. Typical stop mode currents, which are prevalent in the industry are around 0.5 microAmperes and typical normal operating currents are in the range of 4.3 milliAmperes .
Sunil Deep Maheshwari has been a Design Engineer at Freescale Semiconductor for three years, where he has worked on designs ranging from Digital Signal Controllers, Power Train circuitry to Metering. He earned his Bachelor in Engineering in Electronics and Communications from NSIT, Delhi University, India.
 MCF51EM256 Reference Manual & Datasheet (www.freescale.com)
 Analog to Digital Convertors, (http://en.wikipedia.org/wiki/Analog-todigital_converter)
 “Choosing SAR vs High-Speed Sigma-Delta ADCs ” Bonnie Baker, EDN Article,http://www.edn.com/article/CA6313377.html
. A Novel Automatic Utility Data Collection System using IEEE 802.15.4-CompliantWireless Mesh Networks by Jin Zhu and Recayi Pecen
 Metering Solutions from Freescale (www.freescale.com/metering)
 KNX Smart Metering Solutions.