Introduction to power measurements
Evaluation of hardware meters

This section examines the characteristics and suitability of physical measuring devices suitable for energy measurement in information technology. The document is based on the proof of concept section related to the acquisition of measuring devices created for the previous MitViDi project. In the Visiiri project, we have significantly expanded this analysis. Whereas the previous study answered the question "what kind of minimal experimental test setup is possible to build using easily available meter for energy measurement," in the Visiiri project we have taken a broader look at the devices available on the market at a practical price. The review covers a wide variety of device types, so a more detailed analysis of how the meters are used, in addition to the purchase price, is also an essential criterion.

The results of the analysis can be used to support the purchase of ready-made commercial measuring devices. Although the models available for sale and the prices change over time, the analysis identifies various types of devices for which this material can be used in procurement and use. We have also assessed the suitability of the devices for use in a measurement laboratory and with the appropriate software (PowerGoblin).

  1. Requirements analysis
  2. Classes of devices to be measured
  3. Measurement equipment
    1. Measurement instruments
    2. Classification of off-the-shelf power meters
      1. AC meter
      2. Measuring PSU
      3. Meter modules
      4. Component modules
      5. Built-in meters
    3. Summary of off-the-shelf power meters
  4. Measurement component modules

Requirements analysis

The starting point for the design of the measurement laboratory is to define its purpose. The measurement laboratory at University of Turku was built primarily to measure the energy consumption of software under different conditions and for the various devices and their interconnections associated with the operation of the software system. A secondary objective was to develop the processes and practices to organise measurements in a practical way using this equipment.

In the design, we explored different types of equipment, as well as the measurement of equipment at the component level and different forms of energy use (DC, AC, battery, different voltage levels). The meters also differed in the quantities they measured (current (I), voltage (V), power (P), energy (E), integrating and direct measurement). In addition, we studied the suitability of different measurement tools for measurement and the supply of openly available power measurement instruments. However, measuring instruments are specialised tools and ordering them from abroad can lead to additional taxes and import duties.

In practice, the constraints on the construction of the laboratory were the availability, cost, quantity and size of the equipment and the practicality of the instrumentation. It was also essential to consider which different types of equipment and their use in practical applications are perceived to consume significant amounts of energy.

Laboratory conditions typically aim to analyse the properties of individual parts and functions in isolation from each other. For example, in new product development, this is a desirable starting point as the aim is to focus on product optimisation. In contrast, the main objective of our measurement laboratory was to try to observe and measure consumption in a way that mimics normal operating environments. This objective approached the problem from a sort of opposite direction, with the analysis of off-the-shelf solutions and the generalisability of the results. Of course, this does not mean that the laboratory could not possibly be used in other ways, such as those required for product development. However, other aspects were prioritised over properties such as absolute accuracy, bandwidth, dynamic range in the acquisition of the instruments.

It is important to note that one of the starting points of the measurement was to extract the energy consumption characteristics of the software from the results. The software itself is an abstract description (algorithm) and does not consume energy, but the software is used to control the hardware and the impact of the software is observed indirectly through the consumption induced by the hardware. At a very least, it is possible to differentiate between the consumption in idle and active states. The difference between these two states can be attributed to the software.

Classes of devices to be measured

The following classification was used to classify the suitability of measuring instruments and methods for measuring different types of equipment. The categories cover quite a large proportion of typical IT purchases in an office environment.

Device AC DC Battery Ext. meter Int. meter SW meter Effect
Mobile devices
Smartphone <= 5V x DC (#1) (#2) All
Tablet <= 5V x DC (#1) (#2) All
Notebooks
Notebook computer <= 24V x DC (#3) x All
Large computers
Desktop computer <= 240V AC / DC (#4) (#3) x All
Server computer <= 240V AC / DC (#4) (#3) x All
Accessories
Switches (#5) <= 240V <= 24V AC / DC Comm
Routers (#5) <= 240V <= 24V AC / DC Comm
4G/5G modems (#6) <= 240V <= 24V x AC / DC Comm
USB devices (#7) <= 12V AC / DC Comm
DC powered accessories <= 12V AC / DC Comm

Description of categories:

  • AC: typically uses AC (built-in AC power supply). Common voltage levels: 100 - 250V.
  • DC: typically uses DC (battery or DC connector). Common voltage levels: 5V, 9V, 12V, 19V, 24V, typically (3,3 - 24V).
  • Battery: typically powered by battery (DC). Some may also be powered with an external DC power supply.
  • External meter: power consumption can be measured with an external meter (internal battery may have an effect on the results).
  • Internal meter: the device might be equipped with internal meters for analyzing the power consumption.
  • SW meter: software based meters (either utilizing the internal hardware meters or models with calibration data) can be used for measuring / estimating the power consumption.
  • Effect: different forms of software-generated consumption to be measured: 1) processing (CPU, GPU, DSP etc.), 2) memory, 3) disk, 4) communication, 5) all of these.

Notes:

  • #1 Depends on SoC.
  • #2 Varies, e.g. Android power profiler.
  • #3 Usually RAPL, ACPI, WMI, NVML, I2C sensors, and other common interfaces.
  • #4 Low-power desktop/server systems can be powered with PicoPSU class devices + high current DC meters.
  • #5 Typically powered by AC, but also support DC supplies / typically powered by both types of power supplies.
  • #6 E.g. 4G/5G modems, multiple types of power supplies.
  • #7 Large number of USB powered devices can be measured by connecting to a powered hub (with a DC input). On the downside, the USB hub adds its own extra load to the result.

Measurement equipment

This section looks at the equipment used for measurement. The first section looks at the physical principles that are used internally by the different types of meters. The second section provides a classification of the various off-the-shelf measurement products according to their principle of use. The following sections list the complete meters, measurement modules and individual integrated circuits found in the study.

Measurement instruments

In short, (electrical) energy consumption can be measured in two main ways. First, we can measure the current flowing through the system. The power and energy consumption can be deduced from this value by taking into account the (known) voltage level and length of the operation. Since the current flow also generates a magnetic field, other types of instruments either directly measure this field or indirectly the current induced by this field on other conductors within its range. Another approach is to store a known capacity of energy in a battery or capacitor, use the energy from that source to power the system, and finally measure the remaining capacity of the power source.

Technique AC DC Suitability Downsides
Shunt resistor (#1) X X Low current, low-cost designs Additional resistive load, accuracy of the resistor value, thermal drift
Hall sensor X X Low-cost designs Magnetic interference
Fluxgate / zero-flux sensor X X High accuracy, resolution & bandwidth High sec power consumption, noise
Current transformer X High voltage & current Saturation, hysteresis, phase shift errors (power and energy meters)
Rogowski coil X High bandwidth signal Requires integrator circuit
Fiber-optic current sensor X X Power grid Size, price
Magneto-resistive current sensor (#2) X X Non-linear behavior, thermal drift, sensitive to external fields
  • (#1) Maximum continuous current for shunts typically 66% of the peak max.
  • (#2) Multiple magnetoresistance technologies: anisotropic (AMR), giant (GMR), tunnel (TMR), colossal (CMR), extraordinary (EMR)

Examples:

  • Shunt resistor: TME
  • Hall sensor: Mouser
  • Fluxgate: Hioki, Mouser
  • Current transformer: TME
  • Rogowski coil: TME
  • FOCS: ABB
  • Magneto-resistive: Digikey (select sensor type: magnetoresistive)

Classification of off-the-shelf power meters

In the study of measurement equipment, the meters were classified into the following categories:

  1. AC meters
  2. Measuring power supply units (PSU) with integrated measurement
  3. Meter modules requiring an external PSU
    • DC supply connected meter
    • USB power meter
  4. Component modules (circuit boards. ICs) and
  5. Built-in meters

AC meter

An AC meter is the simplest to use - the meter is connected between the power outlet and the device to be measured, either with the electricity passing through the meter or with the meter mounted around the power cable (possible to install without shutting down the device). An AC meter is also suitable, for example, for data centers where complex or custom wiring is not possible. AC metering is very suitable for measuring a wide range of equipment, as most IT devices can be supplied with AC power (even battery powered devices). There are also special AC meters for very high voltage and current environments such as power grids. There is a wide selection of measurement technologies for AC.

AC meters need to support measurement over a relatively wide range of currents, as the consumption of devices and equipment differs significantly. Low-cost AC meters may not be able to measure the highest power devices and may have other sorts of limitations. For example, the power factor of the switching power supplies makes the measuring less straightforward compared to purely resistive loads, the different capacitor stages in the system behave as a buffers, adding latency and smoothness to the results. The measurement focuses on the total consumption of the whole system, which also makes it more difficult to pinpoint smaller subsystems as the sources of consumption. Low-cost AC meters might also have other limitations in their design such as relatively long sampling periods, inaccurate internal clocks and readings, and inaccuracies and other limitations in data export.

Different types of AC meters are widely used and with the proliferation of smart home technology, many AC meters for different smart home ecosystems are available from different home electronics stores. Many IT devices fit within the specifications of these meters, but they often have limited temporal granularity and accuracy of the measurement (e.g. several minutes to hours). Their software ecosystems might also be geared towards instant or periodic monitoring, but not accurate logging of the numeric data for the needs of a systematic measurement process. They even might not contain the functionality for accessing the logged data or exporting the data in a clearly documented format.

Measuring PSU

A measuring power supply unit (PSU) or a laboratory PSU is a power supply with an intermittent current measurement. We identified several laboratory power supplies during the study, but excluded them from the acquisition of lab equipment because the cost of these power supplies can become relatively high, especially as the number of devices to be compared increases. Measuring PSUs are suitable for special applications where high accuracy is required and high manufacturing quality is valued. For general purpose "background" current measurement, their cost may be too high and their use is limited to the measurement of a single or just few DC circuits. Their voltage range usually also covers the majority of digital devices, but the smaller models might not provide enough current to larger IT systems.

Meter modules

Another versatile type of meter, apart from AC current measurement, are meter modules requiring an external PSU. We studied several types of such meters and classified them based on their applicability for the project. The suitability of the meters was assessed on the basis of voltage and current ranges, accuracy, sample rate, price, and the ease of data collection. In practice, some lab power supplies are also sold as separate PSU and meter modules. There are roughly two types of meters. A DC supply connected meter can be set up to provide one or more channels with different voltage levels to the devices to be measured. The meter channels the power from its own power supply. Some of these meters are powered by standard "brick" power supplies, similar to those used for powering e.g. notebook computers. The meters' output voltage range covers a very wide range of devices to be measured, but is typically slightly smaller than that of laboratory power supplies. To simplify the design, these meters often do not perform DC-DC conversion, and the output voltage range extends to just below the input voltage.

Another common type of meter is the USB power meter, which similarly channels the power and data from the USB bus to the device or devices to be measured. The meter can also provide power from an external PSU instead, but the USB input is usually needed for passing through the data (bidirectional). The USB bus itself also provides a crude method for defining power budgets of devices. The USB power meter is limited to devices equipped with a USB bus, but on the other hand, especially with EU legislation and guidelines, USB has become more common as a general purpose power connector, especially for battery-powered devices. The main challenge for USB is that the latest USB charging and power delivery standards have made USB power distribution more complex, which means that not all USB-based meters are able to measure all new USB devices.


Component modules

There is also a category of component modules which includes both integrated circuits (IC, single components) and circuit boards capable of measuring power. We have excluded simple analog devices such as Hall sensors from this category since they require a number of external components (e.g. ADC, MCU, stable power supply, calibration and temperature compensation circuits) for building a meter. These modules offer a modern alternative to building custom power meters from scratch and avoid time-consuming steps such as circuit design. The integrated circuits also have a very small footprint and often provide integrated digital output which simplifies the reading of data. The combination of such chips and high precision components such as shunt resistors can lead to high quality designs comparable to high-end commercial products.

In the study we identified evaluation boards and circuit boards offered as popular DIY kits, and different types of integrated circuits that come factory calibrated with various logic for reducing noise, voltage and temperature compensation, among other features. During the study, we did an initial search for such component modules from the inventories of popular DIY maker manufacturers such as Adafruit, Seeed Studio, and the Raspberry and Arduino related maker companies. We also went through the catalogs of popular component distributors such as Farnell and Digikey. We then used "snowballing" approach by identifying the ICs and techniques used for measurements and searched for other manufacturers offering products based on the same or similar components listed in the chip makers' product catalogs.

Manufacturers often offer different variations of integrated circuits for different applications. Typical ways of classifying circuits are their operating principle, output format (digital, analog), quantity to be measured (voltage, current, etc.), measurement accuracy, voltage and current ranges, need for additional components (separate inductor, resistor, etc.) or price. For example, Texas Instruments provides the following graph for classifying the devices based on their maximum handling capacity of common mode voltage, the measurement accuracy, and type of meter (analog out, digital, integrated comparator, integrated shunt).

Built-in meters

In addition to the above, many digital devices such as computers and computer components have built-in current meters that provide a software-based API for performing voltage, current, power, and energy measurements. Depending on the system, measurements can be done on a component-by-component basis (e.g. processor, graphics card, solid-state disks, USB devices, screen's backlight) or for the whole system (e.g. total instantaneous power consumption of a laptop).

We plan to publish separate documents on the software-based use of built-in meters at a later stage.

Summary of off-the-shelf power meters

The purpose of the study was to identify available solutions for measuring software power usage in general purpose applications, as well as available commercially available metering hardware that can be utilized without extensive prior knowledge of power measurements. As the cost of precision instruments can be relatively high, especially for small purchases, we have limited our focus to products with a unit price below €100 per channel. The idea of channel-specific refinement is to compensate for possible savings when measuring systems with multiple devices.

After mapping the different types of meters, we found out what different off-the-shelf meters and measurement modules are openly available on the market that can be acquired for in-house use. We used the following criteria:

  • The meter is suitable for measuring the consumption of one or more systems selected for the study.
  • The meter measures current, power and/or energy, or its output values can be used to derive a value for energy consumption over some period of time.
  • The total consumption can be accurately broken down between two time points.
  • no PCB design, soldering or other similar equipment manufacturing process is required to implement the meter.
  • Meter output values are readable by software and standard protocols (software & hardware, e.g. SCPI), digital output.
  • To ensure the scope of the meter mapping work, the price per device is limited to €100 per channel.
Name Price Volt Amp Chan Sample rate Proto Target Accuracy Other
Hardkernel
ODROID SmartPower 3 <50€ 3 - 18 3A 2 200 Hz USB TTY MA documented ascii protocol
ODROID SmartPower 2 - - - 1 - - Discontinued
ODROID SmartPower - - - 1 - - Discontinued
Smaller vendors
Joy-IT JT-UM25C 100€ 24V 5A 1 500 Hz BT M protocol not documented
Powerwerx PWRcheck+ 250€ 60V 40A 1 ? USB TTY MNAL protocol not documented, out of stock
VectorFlux ZS-1100-A <750€ 6V (!) 1,5A (!) 1 1 MHz USB - out of stock
VectorFlux ZS-2102-A <750€ 6V (!) 1A (!) 1 1 MHz USB - out of stock
Qoitech Otii Arc Pro 850€ 5V (!) 5A 1 4 kHz USB M protocol not documented
Qoitech Otii Ace Pro 1500€ 25V 5A 1 50 ksps USB MnA protocol not documented
Joulescope 1300€ 15V 3A 1 250 ksps USB MA protocol not documented
Sistemi P1125 <2300€ 1.8 - 8.2 (!) 3.2A 1 10 kHz USB M Discontinued
Sistemi P1150 1000€ 17V 3.2A 1 125 ksps USB MA Not available yet
Lab power supplies
Owon
OWON SPS series 61/31V 5.1/8.1A 1 USB MnA SCPI protocol
OWON SPM series 60/30V 10/5A 1 USB MNAl SCPI protocol
OWON SPE series <200€ 60/30V 10A 1 USB MNAl SCPI protocol
OWON SP series 60/30V 10A 1 TTY MNAl SCPI protocol
OWON P4000 series 60/30V 3/5A 1 TTY MnA SCPI protocol
OWON ODP3032 30V 3A 2 USB,TTY MA SCPI protocol
OWON ODP3063 30+30+6V 6+6+3A 3 USB,LAN,TTY MA SCPI protocol
OWON ODP6033 60+60+6V 3+3+3A 3 USB,LAN,TTY MA SCPI protocol
OWON ODP3122 30+6V 12+3A 2 USB,LAN,TTY MA SCPI protocol
OWON ODP6062 60+6V 6+3A 2 USB,LAN,TTY MA SCPI protocol
Peaktech
PeakTech P 1565 450€ 16V 40A 1 USB MAl protocol not documented
PeakTech P 1570 650€ 16V 60A 1 USB MAL protocol not documented
PeakTech P 1575 450€ 32V 20A 1 USB MNAl protocol not documented
PeakTech P 1575 450€ 32V 20A 1 USB MNAl protocol not documented
PeakTech P 1585 650€ 32V 30A 1 USB MNAl protocol not documented
Rigol
Rigol DP711 350€ 30V 5A 1 TTY MnA SCPI protocol
Rigol DP712 350€ 30V 5A 1 TTY MnA SCPI protocol
Rigol DP811 700€ 40/20V 5/10A 1 USB,LAN MNAl SCPI protocol
Rigol DP811A 850€ 40/20V 5/10A 1 USB,LAN MNAl SCPI protocol
Rigol DP821 700€ 60+8V 1+10A 2 USB,LAN M SCPI protocol
Rigol DP821A 800€ 60+8V 1+10A 2 USB,LAN M SCPI protocol
Rigol DP813 750€ 20/8V 10/20A 2 USB,LAN MNAl SCPI protocol
Rigol DP813A 900€ 20/8V 10/20A 1 USB,LAN MNAl SCPI protocol
Rigol DP822 750€ 20+16V 5+16A 2 USB,LAN MnAl SCPI protocol
Rigol DP822A 900€ 20+5V 5+16A 2 USB,LAN MnA SCPI protocol
Rigol DP831 500€ 30+30+8V 2+2+5A 3 USB,LAN,TTY Ma SCPI protocol
Rigol DP831A 750€ 30+30+8V 2+2+5A 3 USB,LAN Ma SCPI protocol
Rigol DP832 400€ 30+30+5V 3A 3 USB,LAN,TTY MnA SCPI protocol
Rigol DP832A 600€ 30+30+6V 2+2+5A 3 USB,LAN Ma SCPI protocol
Rigol DP932E 550€ 30+30+6V 2+2+5A 3 USB,LAN Ma SCPI protocol
Rigol DP932U 650€ 32+32+6V 3A 3 USB,LAN MnA SCPI protocol
Rigol DP932A 850€ 32+32+6V 3A 3 USB,LAN MnA SCPI protocol
Rigol DP2031 1350€ 32+32+6V 3+3+5A 3 USB,LAN MnA SCPI protocol
Joy-IT
Joy-IT PS360-C 200€ 60V 6A 1 USB MnA protocol not documented
Joy-IT PS1440-C 400€ 60V 24A 1 USB MNAl protocol not documented
Joy-IT RD6012-C 200€ 60V 12A 1 USB MNAl protocol not documented
Joy-IT RD6006-C 300€ 60V 6A 1 USB MnA protocol not documented
Twintex
Twintex PPA100-40A 40.5V 10.2A 1 USB,LAN MNAl
Twintex DSP-1520 15V 20A 1 USB MNAl
Twintex DSP-3210 32V 10A 1 USB MNAl
Twintex PPS-1560 32V 10A 1 TTY MNAl SCPI protocol
Twintex PPS-1560 15.5V 60.5A 1 TTY MNAL SCPI protocol
Twintex PPM-1820 19V 21A 1 TTY MNAl SCPI protocol
Twintex TPM-2010E 20V 10A 1 USB MNAl SCPI protocol
Twintex PPW-2045 20.5V 45.5A 1 TTY MNAL SCPI protocol
Circuit board kits
Adafruit
Adafruit INA260 <10€ 36V 15A 1 I2C MNAl
Adafruit INA228 <15€ 85V 10A 1 I2C MNAl
Adafruit INA3221 <15€ 26V 3.2A 3 I2C MnA
Adafruit INA219 <10€ 26V 3.2A 1 I2C MnA
Adafruit INA169 <10€ 2.7 - 60 5A 1 Analog MnA Analog output only
Adafruit USB Power Gauge <10€ - - - - - Discontinued
Nordic Semi
Nordic Semi Power Profiler Kit II 100€ 5V (!) 1A (!) 1 100 kHz USB -
Nordic Semi NRF6707 - - - 1 - - Discontinued
Smaller vendors
Joy-IT SBC-DVA 48V 8A 1 I2C MnA IC not documented
Curious Electric ISL28022 60V 32A 1 I2C MnA IC not documented
NCD PR3-6 - 5A 8 I2C - Only voltage/current

Description of categories:

  • Price: Retail price for a single unit
  • Type: Type of inputs/outputs. DC = general purpose DC meter
  • Volt: Supported voltage levels. Not all meters support voltage ranges down to 0 Volts.
  • Amp: Maximum sustained current for measurements.
  • Chan: Number of channels for measurements
  • Sample rate: How many samples can be collected per second. Low (< 1 kHz), medium (1 < x < 100 kHz), high (> 100 kHz)
  • Proto: Supported hw/sw protocols for communicates with the meter
  • Target: Suitable target devices (SUT) that can be measured with the meter:
    • Mobile (5V, >3A), Notebook (20V, >10A), Large computer (12V, >50A), Accessories (12V, >3A)
    • m/n/l/a = limited support for this category of devices (5V, >2A) (20V, >3A) (12V, >10A) (12V, >2A)

Measurement component modules

There are also different kinds of power measurement ICs and circuit boards available as building blocks for the Arduino / Raspberry Pi maker community.

Examples of IC vendors: