Optimize Wireless Device Battery Run-time

Optimize Wireless Device Battery Run-time Innovative Measurements for Greater Insights Part 1 of 2 Electronic Measurement Group Ed Brorein Applications Specialist August 22, 2012

Why is Optimizing Battery Run-time so Important Now? Increasingly Capable Devices Higher data demands Larger displays and touch screens Always-connected applications Longer autonomous operation Limited scope for improving battery capacity Complex interaction of applications/software/hardware Inadequate design and analysis methods lead to: Shorter device run time Unanticipated periods of high battery drain Additional design cycles to resolve battery drain issues Disappointed customers due to short battery life 2

A Growing Need for Battery Current Drain Testing for all Phases of Product Development In Hardware Development, optimize energy efficiency: Evaluate and optimize overall device and its sub circuits for battery drain In Software Development, validate new code builds: Run application code regression test suites, assess impact on battery drain In Integration and Validation, run suites of benchmark tests: Validate battery drain for all required operational modes Validate operating time with product s battery (battery run-down test) Validate battery drain for realistic network conditions Benefits: Bring smaller, longer running, more competitive products to market Faster time-to-market and at lower expense by reducing development time 3

Challenges to Battery Drain Characterization under Simulated Real-world Use Traditional Solution: Custom RF Stimulus & Current Drain Logging Setup Challenges: Properly powering the device Making accurate, high resolution current measurements Expertise to generate combinations of simultaneous device activities Expertise to simulate wireless network/environment Massive amounts of data created, how to best process and manage Developing effective visualization & analysis tools can be difficult Software development effort that yields a flexible & capable system Developing takes a large amount of resources and time 4

Challenge: Properly Powering a Mobile Device Actual battery response to a GSM pulse load GP power supply response to a GSM pulse load Battery Voltage GP-PS Voltage Instability Unloading Overshoot Battery Current GP-PS Current Battery voltage drops proportionally with current Battery resistance is 150 mω Voltage response and current drain does not match battery Instability & overshoot 10% higher drain experienced 5

Challenge: Measuring Battery Drain for Power Savings Modes 14 s/div 3 ma/div 500 ms/div 50 ma/div Wireless Temperature-Humidity Sensor Wireless devices operate in short bursts of activity to conserve power: Applicable to a wide variety of devices (handsets, sensors, Bluetooth devices, etc.) Long periods of sleep between bursts of activity GPRS Smart Phone Battery Drain for Standby Resulting current drain is pulsed; extremely high peak, low duty cycle, and low average values spanning up to 4 decades challenging to measure accurately! Problem: Traditional solutions do not have the dynamic measurement range needed to accurately measure current drain of mobile wireless devices. 6

Challenges With Traditional Measurement Solutions DC source or battery + Shunt + DUT current - - DUT Most common solution: Shunt + DAQ Typical Performance: ~12 to16 bits resolution ~ 50K to 1M samples/sec ~ 0.2 to 1.0% gain error (shunt and DAQ) ~ 0.05 to 0.2% offset error (mainly DAQ) + - Data Acquisition Equipment Diff Amp MUX Gain Amp ADC PC to log long-term data Data out Commonly Encountered Challenges: Large effort to configure and program Excessive peak voltage drop on shunt Multiple shunts needed for wider range Greater measurement BW = lower accuracy Offset error and noise limits dynamic range of measurement to about 2 decades for reasonable accuracy for minimum (floor) level of the signal 7

Innovation for Battery Drain Characterization N6781A 2-Quadrant SMU for Battery Drain Analysis Specialized DC power supply module for battery drain testing: Innovation: Seamless ranging spans over 7 decades for accurate measurement of battery drain over wide dynamic range Settable battery emulation characteristics Zero-burden current measurement and auxiliary DVM for battery run-down testing Fast transient response for pulsed loads Up to 200 KSa/sec digitizing rate For use in the N6705 mainframe N6705 DC Power Analyzer Mainframe Integrates multiple instrument functions into a single box: 1 to 4 advanced power supplies; >22 different models available Digital voltmeter and ammeter Arbitrary waveform generator Oscilloscope Long term data logger Full functionality from front panel Gain insights in minutes, not days! 8

Seamless Range Changes Amperes N6781A Seamless Measurement Ranging 200 khz, 18-bit digitizer acts likes single range of 28-bits Range Measurement Accuracy = Seamless range change 3 A ±(0.03% + 250 A) TRANSMIT ACTIVITY 100 ma ±(0.025% + 10 A) RECEIVE & IDLE ACTIVITIES 1 ma ±(0.025% + 100 na) SLEEP STATE 10 µa ±(0.025% + 8 na) FIXED RANGE 9

Comparing a Battery to a Battery Emulator Source Measure Unit (SMU) Powering a Mobile Device Actual battery response to a GSM pulse load N6781A Battery emulator SMU response to a GSM pulse load Battery Voltage Battery Emulator SMU Voltage DUT Current DUT Current Battery voltage drops proportionally with current Battery resistance is 150 mω Voltage response and current drain comparable to the battery Battery emulator SMU set to 150 mω 10

Power-savings Current Drain Measurement Wireless Sensor Example: Test Set Up Wireless weather station: Base unit and wireless temperature sensor Laptop or PC running Agilent 14585A software N6705B DC Power Analyzer with N6781A Source Measure Module LAN cable PM 11:15 78 95 DC in DC power cable 11

Power-savings Current Drain Measurement Wireless Temperature Sensor Example: Measurement Results The need to measure minimum, maximum, and everything in between: Pulse peak current: 21.8 ma Pulse duration: 13.6 msec Pulse avg. current: 9.22 ma Pulse period: 4 sec Average current: 54 µa Pulse contribution: 30.8 µa (57%) Sleep current: 8.7 µa (16%) Other sleep activity: 14.5 µa (27%) 5 ma/div 0.5 sec/div Wireless Temperature Sensor Current Drain 12

Power-savings Current Drain Measurement Wireless Temperature Sensor Example: Measurement Results 21 ma pulses (off scale) 21 ma pulses (off scale) Sleep current base, 20 µa/div 0.5 sec/div Sleep current base, 20 µa/div 0.5 sec/div 100 ma Fixed Range Measurement Seamless Ranging Measurement Range 3 A 100 ma 1 ma 10 µa Measurement Accuracy ±(0.03% + 250 µa) ±(0.025% + 10µA) ±(0.025% + 100 na) ±(0.025% + 8 na) Seamless measurement between these 3 ranges Parameter Fixed Range Seamless Improvement Overall DC accuracy (54 µa avg) 18.9% 0.245% 77 X Sleep current DC accuracy (8.7 µa avg) 115% 1.18% 97.5 X Sleep current AC noise floor ~47 µa p-p ~10 µa p-p 4.7 X 13

Auto-ranging vs. Seamless Ranging Measuring Pulsed Signals Traditional DMMs unable to return desired readings for pulsed signals when set to auto-ranging: Designed to select and use a single range only Takes up to tens of ms to range change. Signal level changes in meantime. Measurement is over-ranged. 200 µs/div Traditional Source-Measure Units output current is also limited by measurement range: Like a DMM, takes up to tens of ms to range change. Signal level changes in meantime. Current limit linked to measurement range. Constantly current limiting with load pulse with subsequent voltage drop-outs and overshoots Seamless Ranging is optimized for pulsed signals: Never over-ranged. Provide a uninterrupted stream of valid readings from multiple ranges As an SMU, measurement and sourcing are totally decoupled assuring glitchless output performance Output Voltage 1V/Div Load Current 500 ma/div Traditional SMU set to auto-ranging measurement, powering a pulsed load (Voltage drops out when load pulses on) 14

Greater Time Resolution Yields Detailed Insights Temperature-Humidity Sensor Example 3 ma/div 91 ms/div 11.7 s/div 3 ma/div Dual Transmit Burst 1-temp, 1-humidity 0.19 sec-long each 8.6 ma average Wireless Temperature-Humidity Sensor 38 sec-long, 7.54 µa sleep periods during burst 200 µsec resolution provides insights. 600 K-points for 120 sec capture 3 ma/div 11 ms/div 99% of the time is sleep; 7.46 µa average (7.54 µa * 99%) 1% of the time is transmit; 86 µa average (8.6 ma * 1%) 12,900 Hrs (1.5 yrs) with 1.2A-Hr battery (93.5 µa total drain) 14.4 ma pulsed, for determining RF PA Power Added Efficiency (PAE) Transmit Burst Pulses 14.4 ma pulse plateau 2 ms pulse duration 2.01 ma idle current 15

Greater Time Resolution Yields Detailed Insights GPRS Smart Phone Discontinuous Receive (DRX) Example Receive current pulses Receive activity / RSSI Wake up / idle pedestal Sleep base 50 ma/div 500 ms/div Sleep current base 5 ms/div 50 ma/div Baseband activities GPRS Smart Phone Battery Drain for DRX Standby 1.22 ma sleep current during 1.25 s paging interval Able to correlate specific activities to current drain duration and level: A large portion of energy is used during the short burst of activities 100 µsec or better resolution provides detailed insights on RX bursts Validate efficiency of individual activities both in current draw and duration DRX Burst Current pulse details 23 ms pulse: 239 ma peak & 92 ma ave. Current drain waveform is often the easiest (or sometimes only) way to observe activity duration 16

Analyzing & Optimizing Power Savings GPRS Mobile Phone Standby Example: Set up DRX (Discontinuous Receive) is a power-savings mode for mobile phone standby operation PC running 14585A SW controlling N6705B to log & display DUT current drain Interface RF antenna cable DUT battery in SMU DC out 8960 Wireless Communications Test Set emulates base station, maintains standby with DUT for DRX-off and DRX-on modes Device Under Test N6705B with N6781A SMU set to emulate battery to directly power DUT. High speed digitizing measurement system captures DUT current drain

Analyzing & Optimizing Power Savings GPRS Mobile Phone Standby Example: DRX-off Results DRX slot rate set to 0.48 sec High level of activity for standby Found average current was 19 ma and peak was 136 ma with DRX-off 26 hours standby time with 500 mah battery (~ 1 day) Data log of current drain of a GSM mobile phone for DRX-off 18

Analyzing & Optimizing Power Savings GPRS Mobile Phone Standby Example: DRX-off Results Further zoom-in with DRX-off shows 80 msec long RX-related current pulses every 235 msec. Each 80 msec long pulse is actually a series of numerous RX-level pulses Expanded data log of a GSM mobile phone for DRX-off 19

Analyzing & Optimizing Power Savings GPRS Mobile Phone Standby Example: DRX-on Results Data log display for DRXon setting shows substantially less activity than for DRX-off Observing 15 to 20 msec pulses every 470 msec Found average current dropped to 2.8 ma with DRX-on, an impressive 85% reduction from DRXoff Standby time increases from 1 day to 1 week for a 500 mah battery Data log of current drain for DRX-on of a GSM mobile phone 20

Analyzing & Optimizing Power Savings GPRS Mobile Phone Standby Example: Analyzing Improvements Standby current drain, continuous RX I avg = 19 ma Standby current drain, discontinuous RX Power savings mode of operation I avg = 2.85 ma (85.8% reduction!) What things contributed to the power savings and by how much? Difficult to precisely ascertain from a data log display, random activity over time Using statistical distribution profiles to analyze provides additional insights 21

Analyzing & Optimizing Power Savings Complementary Cumulative Distribution Function (CCDF) % PDF % CDF = PDF 100 % CCDF = 1 CDF 100 0 value 0 value 0 What is a CCDF? value Time-related changes Continuous RX Amplitude changes Using CCDF profiles is effective for analyzing complex current drain signals Vertical axis is % occurrence Vertical shifts are time-related changes Horizontal axis is amplitude Horizontal shifts are amplituderelated changes Using a log scale the end details are expanded CCDF profile of standby current drain, continuous RX 22

Analyzing & Optimizing Power Savings GPRS Mobile Phone Standby Example: CCDF Profile Analysis 2.4 ma 14.3 ma Discontinuous RX 3.0% 0.16% Continuous RX Comparing CCDF profiles of standby current drain, continuous RX vs. discontinuous RX A (vertical) change of 2.8% of RX activity at 128 ma contributed 18% of the power savings A (horizontal) change of 11.9 ma of the idle current activity (~ 90%) contributed 55% of the power savings The remainder of the power savings is from reduced baseband activity The more random the signal is the more helpful it is to have alternate ways to analyze results 23

Detailed Battery Run-down Testing GPRS Mobile Phone Example: Test Set Up 8960 Wireless Communications Test Set N6705B DC Power Analyzer with N6781A in measure-only mode RF DUT with battery Battery V and I measure only Control interface Base station emulator, provides stimulus for DUT PC running N5792A Interactive Functional Test (IFT) SW, and 14585A Battery Drain Analysis SW High speed digitizing measurement system captures DUT battery rundown voltage and current 24

Detailed Battery Run-down Testing GPRS Mobile Phone Example: Test Set Up Using actual battery gives most realistic assessment of DUT performance N6781A regulates zero volts at output while measuring current, acting as a zero burden ammeter, eliminating voltage drop issues of using a shunt Auxiliary DVM input simultaneously measures battery voltage DUT battery + + _ + _ V out - _ Battery current drain 0 Volts Zero burden ammeter A Aux in voltage measurement V out + + N6781A source/measure unit _ DUT 25

Detailed Battery Run-down Testing GPRS Mobile Phone Example: Results Logged min, avg & max volts, amps, & watts 24 min/div Markers at start and shutdown determine: Voltage Current Power I avg = 233 ma V avg voltage = 3.82 V Charge = 843 mah Energy = 3.19 Wh Run time = 3 hr 38 min V shutdown = 3.44 V Findings: Delivered charge (843 mah) less than spec d (1000 mah) V shutdown high (target 3V) N6781A SMU and 14585A Software Measuring Battery Run-down on a GSM/GPRS Mobile Phone Often energy (unspec d) is more related to actual runtime than Charge (spec d) 26

Influence of Environment and Activities on Battery Drain Cellular Mobile Device Example Network Power changes User DRX Navigation Handover UE Voice Video Server Fading Download E-mail Upload Messaging 27

Challenge: Easily Exercise Device Capabilities for a Variety of Environmental Conditions Requirements Exercise All Applications (Voice, Video, Messaging, e-mail, Download, Upload etc) Exercise All Network Conditions (Handovers, Power Level Changes, DRX) Flexibility to quickly setup, change, and re-run tests Traditional Solutions Custom software for application Use a general purpose graphical test executive program environment Test mode operation of device Challenges Custom S/W is a huge programming effort Learning curve for test implementation and measurement methods Tests a limited subset of UE features Specialized Approach Simultaneous and multi-threaded exercising of device under defined network conditions 28

Influence of Environment and Activities on Battery Drain Cellular Mobile Device Example: Test Set Up PXT Wireless Communications Test Set DUT N6705B and N6781A SMU in battery emulation mode RF DUT DC Supply Control interface LAN Optional PC or internet connection for running server applications PC running N5972A Interactive Functional Test (IFT) SW and 14585A SW for battery drain analysis 29

Influence of Environment and Activities on Battery Drain Cellular Mobile Device Example: Stress Testing Results N5972A provides simultaneous activities (SMS/MMS, FTP, battery profile etc.) Predictably or randomly changing over short or long periods of time Enables user-experience testing scenarios Stresses phones and finds issues earlier, vs. sequential testing Simultaneously log current drain to correlate to network conditions and device activities FTP SMS Failure due to SMS buffer limitations in DUT Failure Protocol Log Failure caused by unique combination of Activities Cell Power 30

In Summary More than ever optimizing battery life is a top priority Devices operate in short bursts of activity to save power, draw pulsed current Current drain spans 4 decades, challenging for traditional equipment A wide dynamic range on the time scale provides greater insights on activities Analyzing distributions help to quickly quantify power savings from change Difficult to observe directly in a data log display Battery characteristics have a lot of influence on run-time Power source needs to be the actual battery or reasonable emulation A battery run-down test yields insights on actual capacity delivered Need to emulate environment and exercise device activities for accurate realworld battery run-time results and stress testing 31

Don t miss part two, September 19 th : Optimize Wireless Device Battery Run-time Impact of the Battery, its End Use, and its Management Go to: www.agilent.com/find/events Brochure available Hardcopy: order part # 5991-0160EN Online: http://cp.literature.agilent.com/litweb/pdf/ 5991-0160EN.pdf Thank you for attending! Questions? 32