IllustrationPower the device through a meter with low burden voltage that switches current ranges automatically and fast, sample quickly enough to see radio bursts, and average over whole wake cycles. A multimeter held on a microamp range drops too much voltage or overloads when the radio wakes, which can leave the device unstable or reset it.
A battery device spends most of its life asleep, drawing microamps, and wakes for short radio bursts that draw far more. One instrument has to read both without disturbing the device. This guide covers burden voltage, why a multimeter range switch can upset a sleeping device, how fast an autoranging meter has to switch, what sample rate radio bursts need, and how long to average. It ends with an instrument table and a bench checklist.
This is an engineering summary based on vendor documentation and application notes. Instrument specifications change between versions, so check the linked pages before you buy or test. All sources were read on November 15, 2026.
Why is sleep current hard to measure?
The problem is range. Espressif's ESP32-C3 datasheet lists 5 µA in Deep-sleep mode with the RTC timer and RTC memory on. The same datasheet lists a peak of 335 mA for 802.11b transmission at 1 Mbps and 21 dBm, measured with a 3.3 V supply (ESP32-C3 datasheet, as of November 2026). One meter has to read both ends of that span on the same device.
Espressif's own measurement guide says a module draws several microamps in Deep-sleep and milliamps in active mode, and that this high dynamic current range makes accurate measurement difficult. It adds that ordinary ammeters cannot switch the measurement range fast enough (ESP-IDF current measurement guide, as of November 2026).
Sleep current is often not flat either. TI explains that Standby current on its CC13xx and CC26xx wireless MCUs has two parts: periodic recharge pulses that top up an internal capacitor, and around 70 nA between them. The Standby figure of approximately 1 µA in the data manuals is the average including the pulses (TI SWRA478D, as of November 2026). Engineering reasoning: a meter that only reads the flat part between pulses reports a number that is too low.
The sleep number you measure ends up in a battery-life claim. For how to state that claim so a customer can reproduce it, see how to test and state battery life honestly.
What is burden voltage, and why does a multimeter upset the device?
Most current meters put a small resistor, called a shunt, in series with the supply and measure the voltage across it. TI's Bluetooth LE measurement note describes the simplest version: a resistor in line with the supply and an oscilloscope voltage probe across it. It suggests 10 Ω as small enough not to affect the circuit and large enough to give a voltage you can read with decent precision (TI AN092, as of November 2026).
The voltage across the shunt is the burden voltage. The device sees the supply minus that drop. Espressif warns that ordinary ammeters have a relatively high internal resistance, and that the resulting voltage drop can leave a module below its minimum supply voltage and in an unstable state (ESP-IDF current measurement guide, as of November 2026).
Bench multimeters publish this figure for each range. The Siglent SDM3065X datasheet lists a burden voltage below 0.03V on its 200 μA range and below 0.25V on the next range up (SDM3065X datasheet, as of November 2026). Engineering reasoning: the drop grows with current, so a burden that is harmless during sleep becomes a large drop if the radio wakes while the meter is still on a sensitive range.
The range switch problem
A single fixed range cannot cover both states. TI's note on the CC2541 says the meter must read in the μA range and also handle mA current while the device powers up. Some ammeters limit the maximum current in their low-current modes. With a meter that can switch modes while connected, TI suggests powering up in mA mode, switching to μA mode once the device sleeps, and switching back before any event draws more current, because that may overload the meter and cause unexpected results (TI AN092, as of November 2026).
Engineering reasoning: that method gives a quick sleep reading but misses the bursts, and a scheduled wake can land while the meter is on the small range. If a meter opens the circuit while it changes range, the device runs from its decoupling capacitors until contact returns. If that gap or the extra drop pulls the rail below the brown-out level, the device resets, and you end up measuring a device that has just booted.
How fast does an autoranging meter have to switch?
An autoranging power analyzer changes shunts in hardware as the current changes. Jetperch's theory page for the Joulescope JS220 describes six current ranges and says the instrument picks the shunt that keeps the maximum voltage across it to 20 mV (JS220 theory of operation, as of November 2026). Its specifications list shunts from 0.01 Ω on the 10 A range to 1111 Ω on the 18 µA range (JS220 specifications, as of November 2026). Espressif's guide gives its total voltage drop as 25 mV at 1 A (ESP-IDF current measurement guide, as of November 2026).
Switching speed matters as much as the drop. Jetperch gives a linearized estimate: switching time equals the device capacitance times the voltage glitch it can tolerate, divided by the current step. For a 3.3V target with 10 µF that tolerates a 3% glitch on a 1 A step, the result is about 1 µs, and the JS220 was designed for that speed. Range changes are always make-before-break, so the device stays connected (JS220 theory of operation, as of November 2026).
Fast switching has side effects. The same page says a range change can briefly push real current into or out of the target's bypass capacitors, and that this current cancels out over time (JS220 theory of operation, as of November 2026). For how the regulator shapes quiescent current and load steps, see LDO vs buck converter for battery devices.
What sample rate do Bluetooth LE and Wi-Fi bursts need?
Radio bursts are short. TI's CC26xx note says connectable advertising has a base transmit time of 144 µs, with 1 µs added for every additional bit. In its connection event example, the device is awake for 2425 µs (TI SWRA478D, as of November 2026).
TI says a simple multimeter is not sufficient here because consumption varies over time, and that an oscilloscope can be used as long as its sampling rate and bandwidth are good enough (TI SWRA478D, as of November 2026).
Instruments differ widely. The JS220 digitizes 2 million samples per second and keeps a 300 kHz measurement bandwidth (JS220 theory of operation, as of November 2026). The Otii Arc Pro lists 4 ksps in the ±19 mA range, 1 ksps in the ±2.7A range and an analog bandwidth of 400 Hz (Otii Arc Pro specifications, as of November 2026). The SDM3065X lists DC current at up to 1000 readings per second (SDM3065X datasheet, as of November 2026).
Engineering reasoning: a slow logger can still get the average right if it captures all the charge, but it cannot show the shape of a burst. You need the shape to find a wake that runs long, a packet retry, or a radio that stays on. For how advertising and connection intervals set that timing, see what Bluetooth Low Energy is.
How long should you average?
Average current is what goes into battery math, and it depends on the window. TI's CC2541 note says a single average current cannot be given in a datasheet, because it depends heavily on connection parameters, so every average needs its use case stated (TI AN092, as of November 2026).
TI's CC26xx note shows the arithmetic. A connection event of 2425 µs averages 3.619 mA. Spread over the connection interval with Standby in between, the average current while connected is 9.74 µA (TI SWRA478D, as of November 2026). Engineering reasoning: the radio is awake for a small share of each interval in that example, yet the event supplies most of the average. An error in the burst measurement moves the battery estimate far more than an error in the sleep floor.
Sleep current can also change with the wake pattern. TI reports that a CC26x0 advertising every 100 ms settles at a Standby current of 1.57 μA, higher than the data manual figure, because of how its recharge interval adapts (TI SWRA478D, as of November 2026).
Qoitech specifies its Otii Arc Pro accuracy for averages over more than 1s (Otii Arc Pro specifications, as of November 2026). Engineering reasoning: set the window to a whole number of wake cycles, and make it long enough to include rare events such as sensor reads, log writes, connection updates and recharge pulses.
Which instrument fits which measurement?
Sourced cells link to the page they came from, as of November 2026. Wording without a link is engineering reasoning.
| Instrument | How it measures | Good for | Limits |
|---|---|---|---|
| Multimeter on a fixed range | An internal shunt for each range. You pick the range by hand. | A steady sleep floor on a device that stays asleep for the whole reading. | The SDM3065X lists a burden voltage below 0.03V on its 200 μA range and DC current at up to 1000 readings per second (SDM3065X datasheet). TI says a basic multimeter is not sufficient for average current over time (TI AN092). |
| Shunt resistor and oscilloscope | A resistor in the supply line, with a voltage probe across it. | The shape of each burst, with equipment most labs already own. | One resistor sets one range. TI suggests 10 Ω for measuring BLE bursts this way (TI AN092). Sleep current then shows up as a very small voltage. |
| DC power analyzer or source-measure unit | Supplies the device and measures its current over time. | Supply and meter in one. TI calls a DC power analyzer the most accurate way to measure power consumption (TI SWRA478D). | A bench instrument. TI used an Agilent N6705B with an N6781A module and set its ranges to Auto (TI SWRA478D). |
| Fast autoranging analyzer (Joulescope JS220) | Switches among shunts in hardware, make-before-break. | Bursts and sleep in one capture: 2 million samples per second and a 300 kHz measurement bandwidth (JS220 theory). | Needs a host computer (JS220 theory). In an immunity test at 3 V RMS/m, the JS220 measured up to 10 µA of current induced by RF, observed and not guaranteed (JS220 specifications). |
| Autoranging supply and logger (Otii Arc Pro) | Supplies the device and switches current ranges automatically. | Long battery-style runs. Accuracy is ±(0.1 % + 50 nA) up to 19 mA in auto range (Otii Arc Pro). | 4 ksps in the ±19 mA range and an analog bandwidth of 400 Hz (Otii Arc Pro). Short states are smoothed. |
| Nordic Power Profiler Kit II | An analog measurement unit with a high dynamic range. | Espressif describes it as covering several microamps to 1 A (ESP-IDF guide). | Resolution varies between 100 nA and 1 mA, depending on the measurement range (ESP-IDF guide). |
How do you set up the bench so the number holds up?
Measure the board you will ship. Espressif does not recommend measuring module current on a development board, because some circuits on the board still draw power during sleep. It suggests powering the bare module through the meter instead (ESP-IDF current measurement guide, as of November 2026).
TI recommends a regulated DC supply over a real battery during measurement, to remove variables caused by a defective or low battery. On its CC2541 keyfob, TI found that the accelerometer, buzzer, LEDs and buttons still drew residual current even with the radio chip's pins set to an optimized state (TI AN092, as of November 2026). Engineering reasoning: every part on the rail counts in sleep, including the charger and fuel gauge. For how those parts behave, see battery management systems for small devices.
Watch what else touches the circuit. Jetperch's example is a 3.3V rail feeding a typical multimeter or oscilloscope input, which draws 330 nA (JS220 theory of operation, as of November 2026). Engineering reasoning: a debug probe or UART adapter left connected can feed current into the device through its pins or draw current from them, so disconnect it or power it from the measured rail.
For how the electronics, firmware and test work fit together in a project, see how the electronics work is organized.
What should you check before you trust a sleep current number?
Run through this list before a sleep or average current number goes into a spec, a battery estimate or a customer document.
- Is the current measured on the product board, with development-board circuits, debug probes and UART adapters removed or accounted for?
- What is the burden voltage on each range the meter uses, and does the supply stay above the brown-out level during the largest burst?
- Does the meter change range without opening the circuit, and how fast does it switch?
- Are the sample rate and bandwidth high enough to show the shortest radio state you need to see?
- Does the averaging window cover whole wake cycles, including rare events such as sensor reads, log writes and connection updates?
- Was the device powered from a regulated source at the battery voltage you plan to ship, and is that voltage written next to the result?
- Are the radio settings, such as advertising or connection interval and transmit power, recorded with the number?
- Was the measurement repeated on more than one board, and at the temperatures the product will see?
Questions about measuring sleep current
Can I measure sleep current with a regular multimeter?
For a steady sleep floor, often yes, if the device stays asleep for the whole reading and the burden voltage on that range is low. TI describes powering up with the meter in mA mode, switching to μA mode once the device sleeps, and switching back before any event draws more current (TI AN092, as of November 2026). It will not show bursts or a true average.
What shunt resistor value should I use?
It depends on the current you want to read. TI suggests 10 Ω for viewing BLE bursts on an oscilloscope (TI AN092, as of November 2026). Autoranging analyzers switch among several shunts instead. The JS220 spans 0.01 Ω on its 10 A range to 1111 Ω on its 18 µA range (JS220 specifications, as of November 2026).
Why does my device reset when I put a meter in series?
A common cause is the voltage the meter drops when current rises. Espressif notes that ordinary ammeters have a relatively high internal resistance, and that the drop can leave a module below its minimum supply voltage and in an unstable state (ESP-IDF guide, as of November 2026). Use a meter with low burden voltage that changes range quickly, and check the supply at the device pins during a burst.
How can Duobay help with power measurement on a new device?
Duobay develops custom devices with a Shenzhen engineering team that covers electronics, firmware, enclosure, app and test, and builds samples and small batches. Duobay is also developing an AI hardware debug box for this kind of bench work; it is in development, and details may change. Firmware source and the SBOM are delivered to you, and source files are assigned to you at every paid gate. Initial calls are free, and we do not publish prices. Tell us about your device.
If you are developing a custom battery-powered device, tell us about your device.
Sources. All pages accessed November 15, 2026. Instrument specifications and app note revisions change, so confirm the current version before you rely on a figure.
- Espressif, ESP-IDF Programming Guide, Current Consumption Measurement of Modules, ESP32-C3 (link)
- Espressif, ESP32-C3 Series Datasheet, version 2.5 (link)
- Texas Instruments, Application Note AN092, Measuring Bluetooth Low Energy Power Consumption (SWRA347) (link)
- Texas Instruments, Measuring CC13xx and CC26xx Current Consumption (SWRA478D) (link)
- Siglent, SDM3065X Series Digital Multimeter datasheet (link)
- Jetperch, Joulescope JS220 User's Guide, Specifications (link)
- Jetperch, Joulescope JS220 User's Guide, Theory of operation (link)
- Qoitech, Otii user manual, Otii Arc Pro hardware and tech spec (link)
Planning a device like this? Tell us what it needs to do. A Duobay engineer replies within 1 business day. Send a project brief →