Challenges with Measuring Current when Developing Power Management Schemes for Battery-Powered Devices (Part 1)
While most power supplies used to power the DUT have built-in current measurements that can measure ampere and milliampere level current quite accurately, they may not be able make the microampere current measurements needed to measure sleep mode or leakage.
Battery powered devices are everywhere, from the nearly ubiquitous cell phone, to personal electronics like MP3 players and digital cameras, to medical equipment, to industrial tools and to a vast array of military electronics. For all of these devices, battery runtime is a key requirement. An enormous amount of engineering goes into developing power management schemes to optimize runtime.
In order to understand if a design has achieved an improvement in runtime, an engineer needs to be able to measure current being consumed by the device and determine if his optimized design has indeed lowered overall current consumption. In this article, we will cover methods used to measure current flowing from the battery into the electronic device (or within sub-circuits of the device) and how modern power management schemes create challenges for measuring these currents.
The cell phone – The “poster child” for needing to extend runtime through power management
In the cell phone market, battery life is the single specification that any customer can easily measure on their handset. Insufficient battery life easily dissatisfies users. Therefore, extending battery life by lowering power consumption is a main driver for all aspects of the design of handsets and their constituent components. The trend, of course is heading in the wrong direction. Increased functionality of cell phones, which today include internet access, audio, video and multi-mode capabilities with voice and data, causes increased demand on the battery and lowers run time.
In order to address the increased power demand, cell phone designers have turned to dynamic power usage as a means to significantly reduce power consumption. With this technique, sub-systems within the handset are turned on and off only as they are needed, thus conserving power.
However, these sub-systems sit on the handset’s internal power rails continuously. Even while they are disabled, their quiescent power (aka leakage) is a drain on the battery. Even though the power drain per sub-system is small when they are disabled, the sub-systems have fallen under scrutiny as a source of excess battery drain. In the R&D lab, engineers work hard to make small changes to handset hardware and software to minimize current draw and optimize battery life. It is therefore critical for engineers to be able to accurately characterize total phone current in the lab, and to understand the impact of their design decisions, by measuring the on and off currents of each sub-system independently.
Resulting current profiles create measurement challenges
In most power management systems, the primary method to conserve battery power is to power down and put to sleep sub-systems that are not actively in use. As functionality rises, it becomes harder to find sub-systems that are unused to power down, put to sleep, and save energy. It becomes necessary to turn off and on sub-systems very quickly in between activities on a sub-millisecond timescale. The result is that the DUT draws from the battery a rapidly changing current waveform with events that can take place in a fraction of a second. For example, a GSM cell phone can have current pulses of 2 amperes that last approximately 500 us while the power amplifier is on and transmitting, and then drop back down to the milliampere level for the remainder of the 4.5 ms GSM cycle.
These rapidly changing current waveforms present two measurement challenges: range and speed. First, the dynamic range of current can be greater than 1000:1 or even 1,000,000:1. With full power active currents on the order of 1 to 3 amperes, and with low sleep mode level currents on the order of tens of microamperes, the range of current to be measured often exceeds a typical single measurement range of any measurement instrument (voltmeter, ammeter, scope, digitizer). Range changing becomes necessary, but the speed of the changing in the waveform can be a problem. It may not be possible to change range and make a measurement fast enough to capture the desired measurement during a rapidly occurring event in the waveform. If range changing isn’t possible, then accuracy may be sacrificed by using a higher single measurement range to capture current during these low current consumption events.
To make these measurements may require building into the DUT some kind of test mode that operates the DUT in each of its power states in a slow or static way, allowing each state to be accurately measured. However, this may not be possible depending on the design of the DUT and it certainly does not allow for measurements of real operational performance of the DUT. A final alternative may be to run multiple passes of measurements with the instrumentation on different ranges and somehow combine the information to create a complete picture of dynamic current consumption. This method can be time consuming and may not be practical, as operating the measurement instrument in an over-range condition can cause saturation and invalidate the measurements. Additionally, the particular event that the engineer is trying to capture may not be repeatable (such as hunting for glitches or edges), so multiple passes won’t work.
Emulating the battery
To measure power consumption, often, the battery of the device under test (DUT) is replaced by a power supply, which acts as a battery that never dies and allows the DUT to be tested at a variety of operating points. The power supply then acts as a battery emulator and with a combination of voltage set-points and programmable output resistance settings, the power supply can emulate a fully charged battery, a partially discharged battery, and a battery that is nearly dead.
If the DUT’s current is changing very rapidly, fast current transitions will occur. These sharp transitions can cause significant problems with the power supply and cause the power supply output to momentarily collapse as the power supply’s regulation system is unable to keep up with the rapidly changing current demand being pulled from its output. The result is a droop in the power supply output voltage. This droop in “battery voltage” can cause several problems. One problem is the DUT might turn off because its low battery detect circuit engages. Another problem could be that measurements are invalid because the test conditions are not correct. For example, if you are measuring the RF power output of a GSM power amplifier during that 500 us GSM transmit pulse, and the power amplifier’s input voltage isn’t constant due to power supply droop, the RF power out won’t be correct either.
Some test equipment manufacturers offer specialized power supplies called battery emulating power supplies. These power supplies have very fast regulation systems that allow them to keep their voltage very constant even in the face of rapidly changing currents. Note that these supplies can be unstable depending on the impedance of the DUT and wiring, thus requiring more careful attention when applying these power supplies.
Another approach to preventing voltage droop is to use a standard power supply and just put a large capacitor near the DUT during test. The power supply will no longer experience the high current transients because the current pulses will actually come out of the capacitor. However, this approach has issues as well. First, the capacitor may need to be physically quite large and located right at the DUT, where space may by constrained. Also, any current measurement built into the power supply will no longer give the correct values because the power supply will be measuring the current into the capacitor, while the real DUT current is the current coming out of the capacitor.
Measuring current using a DMM
While most power supplies used to power the DUT have built-in current measurements that can measure ampere and milliampere level current quite accurately, they may not be able make the microampere current measurements needed to measure sleep mode or leakage. Furthermore, if the engineer is trying to measure current flowing between two sub-systems, such as the current out of a power management unit (PMU) into a sub-system, then the engineer may turn to a DMM. However, to use the DMM as an ammeter, the power path needs to be routed from the power supply through the DMM to the DUT. This adds complexity to wiring and sources for noise pickup.
Another issue with using a DMM as an ammeter is burden voltage. When a DMM is configured as an ammeter, the current to be measured is flowed through a calibrated current sense resistor, or current shunt, inside the DMM. The DMM measures the voltage drop across the shunt and calculates the current. The voltage drop inside the DMM reduces the available voltage at the DUT and hence places a burden on the circuit. This burden voltage can be hundreds of millivolts, so using the DMM as an ammeter can have negative effects on the test conditions (Figure 1).
The final issue with using a DMM as an ammeter is speed. DMM’s, to achieve high accuracy, make integrated measurements, sometimes over multiple power line cycles to reduce noise. This means that each accurate reading can take hundreds of milliseconds. With such a slow measurement rate, an engineer cannot capture a rapidly changing waveform.
DMMs can give very accurate current measurements for static conditions and can give accurate measurement of overall (i.e., integrated over time) current consumption. However, their burden voltage and inability to measure dynamic waveforms means they have limitations for use in gaining insights into power optimization. DMMs can’t help an engineer to see the operational impact of their power management schemes in order to answer the question “Did my design change save power by successfully turning on/off this sub-system for the required number of milliseconds or microseconds?”
Measuring current using an oscilloscope
When trying to measure a rapidly changing waveform, an oscilloscope seems like the best choice. Scopes have sufficiently wide bandwidth, which means rapidly changing waveforms due to power management should pose no problem. However, scopes don’t measure current, so in order to use a scope, the engineer may turn to a current probe. These are readily available, but current probes are not known for accuracy below 1 mA. So, while current probes may be the available choice and one of the simpler ways to measure current, there are tradeoffs on overall accuracy.
A current sense resistor, or current shunt, could be used in conjunction with a scope (Figure 2). Current shunts are low cost and simple to use. However, selecting the appropriate sized current shunt can be a big problem. If the dynamic range of current to be measured is 1000:1, what size resistor should be selected? If the shunt is sized to measure the lowest current accurately, there will be a large voltage drop across the shunt during the high current events, and this will place an unbearable burden voltage on the circuit. If the shunt is sized for the high current, there will most likely not be enough voltage signal for the scope to measure, as most scopes have only 8-bits to 10-bits of vertical resolution and limited vertical accuracy. By going with several shunts for different sized measurements, engineers can solve the signal level issue, but then switching shunts means interrupting the measurement, so this solution suffers from the same shortfalls as using a DMM and changing ranges. So, while shunts are simple transducers that are easy to use, they suffer from limitations on measurement range.
Measuring current with specialized instrumentation
There are also some specialized instruments that can aid the engineer who is trying to optimize battery runtime. Along with battery emulating power supplies (available from a number of manufacturers) mentioned earlier in this article, Agilent Technologies offers the N6705 DC Power Analyzer (see figure 3).
Figure 3: The Agilent N6705 DC Power Analyzer enables R&D
engineers to visualize current vs time into the DUT
The DC Power Analyzer is a tool specifically designed for an R&D engineer doing investigative work with low power DUTs. It combines up to 4 power supplies, voltmeter, ammeter, scope, datalogger, and arbitrary waveform generator. As an integrated instrument, the DC Power Analyzer can measure current directly (no need for current shunts or current probes) and show that current on a scope-like display or datalogger display to allow engineers to visualize current vs time or power vs time. Unlike an oscilloscope, the DC Power Analyzer has an 18-bit digitizer and can measure currents down to 1 microampere. This measurement resolution gives a wide measurement range that means fewer or no range changes when measuring dynamic currents. It is a useful tool for engineers as it gives insight into power consumption of the DUT.
With battery life being a critical specification on portable devices, sophisticated power management schemes are being employed. The result is that the currents being drawn by these devices are time-varying, with sub-millisecond events and wide dynamic ranges. Engineers will need to characterize these currents in order to determine if they have achieved improved optimization of available battery power. There are a number of solutions that can be used to make dynamic current measurements. Each carries with it a tradeoff of cost, simplicity, speed, and accuracy. Understanding the limitations of each solution can help an engineer to determine what tool and methodology to employ to get the best measurements they can.
This article covered the measurement challenges associated with measuring dynamic currents on low-power, battery-powered devices. Next month, look for part two of this subject, where we will describe a new solution being introduced that will uniquely address and solve these measurement challenges.