REAPer Adaptive Micro-Source Energy-Harvester for Wireless Sensor - PowerPoint PPT Presentation

REAPer Adaptive Micro-Source Energy-Harvester for Wireless Sensor Nodes SenseApp 2017 Ulf Kulau, Daniel Bräckelmann, Felix Büsching, Sebastian Schildt and Lars Wolf, 09.10.2017 Technische Universität Braunschweig, IBR

Introduction and Motivation Undervolting in WSNs REAPer HW Design REAPer Evaluation Software Implementation Field Test Summary Towards more adaptive WSNs WSNs in real environmental conditions Various parameters (especially temperatures) affect the characteristics of WSNs Dependability : Efficiency of transceivers, HW faults, ... Efficiency : Power dissipation, ... Energy budget : Energy Harvesting, Energy storage, ... 09.10.2017 Ulf Kulau REAPer Adaptive Micro-Source Energy-Harvester for Wireless Sensor Nodes Page 2

Introduction and Motivation Undervolting in WSNs REAPer HW Design REAPer Evaluation Software Implementation Field Test Summary Towards more adaptive WSNs Project goal: Robust but efficient WSNs by adapting operation parameters Efficiency Dependability Energy Budget 09.10.2017 Ulf Kulau REAPer Adaptive Micro-Source Energy-Harvester for Wireless Sensor Nodes Page 3

Introduction and Motivation Undervolting in WSNs REAPer HW Design REAPer Evaluation Software Implementation Field Test Summary Towards more adaptive WSNs Project goal: Robust but efficient WSNs by adapting operation parameters Efficiency Adaptive energy harvesting platform REAPer Energy harvesting : → Varying energy budget Voltage scaling (undervolting) : → Adaptive energy efficiency Dependability Energy Budget 09.10.2017 Ulf Kulau REAPer Adaptive Micro-Source Energy-Harvester for Wireless Sensor Nodes Page 3

Introduction and Motivation Undervolting in WSNs REAPer HW Design REAPer Evaluation Software Implementation Field Test Summary Towards more adaptive WSNs Project goal: Robust but efficient WSNs by adapting operation parameters Efficiency Adaptive energy harvesting platform REAPer Energy harvesting : → Varying energy budget Voltage scaling (undervolting) : → Adaptive energy efficiency Dependability Energy Budget 09.10.2017 Ulf Kulau REAPer Adaptive Micro-Source Energy-Harvester for Wireless Sensor Nodes Page 3

Introduction and Motivation Undervolting in WSNs REAPer HW Design REAPer Evaluation Software Implementation Field Test Summary Undervolting in WSNs – Background Voltage Scaling increases energy efficiency significantly Dynamic power dissipation of CMOS p dyn = C L · f cpu · V 2 09.10.2017 Ulf Kulau REAPer Adaptive Micro-Source Energy-Harvester for Wireless Sensor Nodes Page 4

Introduction and Motivation Undervolting in WSNs REAPer HW Design REAPer Evaluation Software Implementation Field Test Summary Undervolting in WSNs – Background Voltage Scaling increases energy efficiency significantly Dynamic power dissipation of CMOS p dyn = C L · f cpu · V 2 DVS: Adapting f cpu to current workload and scale V ( f cpu ) f cpu = 8MHz f cpu = 8MHz f cpu = 4MHz f cpu = 6MHz 100% 100% voltage voltage Task 1 Task 2 Task 2 Task 1 0% 0% T T T T DPM DVS 09.10.2017 Ulf Kulau REAPer Adaptive Micro-Source Energy-Harvester for Wireless Sensor Nodes Page 4

Introduction and Motivation Undervolting in WSNs REAPer HW Design REAPer Evaluation Software Implementation Field Test Summary Undervolting in WSNs – Background Voltage Scaling increases energy efficiency significantly Dynamic power dissipation of CMOS p dyn = C L · f cpu · V 2 DVS: Adapting f cpu to current workload and scale V ( f cpu ) Undervolting: Violate specifications V ( f cpu ) → V ( f cpu ) − ∆ V f cpu = 8MHz f cpu = 8MHz f cpu = 4MHz f cpu = 6MHz 100% 100% voltage voltage Task 1 Task 2 Task 1 Task 2 Task 2 Task 2 Task 1 0% 0% T T T T Undervolting 09.10.2017 Ulf Kulau REAPer Adaptive Micro-Source Energy-Harvester for Wireless Sensor Nodes Page 4

Introduction and Motivation Undervolting in WSNs REAPer HW Design REAPer Evaluation Software Implementation Field Test Summary Undervolting in WSNs – Background Legitimation to use undervolting Threshold Voltage V th of CMOS is temperature-dependent V th ( T ) = V th 0 + α · ( T − T 0 ) MCUs cover a widespread temperature range with a fixed V ( f cpu ) Room Temperature (19°C < T < 24°C) -55°C -35°C -15°C 5°C 25°C 45°C 65°C 85°C 105°C 125°C → MCUs must be able to run below V ( f cpu ) (under normal conditions) 09.10.2017 Ulf Kulau REAPer Adaptive Micro-Source Energy-Harvester for Wireless Sensor Nodes Page 4

Introduction and Motivation Undervolting in WSNs REAPer HW Design REAPer Evaluation Software Implementation Field Test Summary Is this a good idea? Undervolting will lead to a higher unreliability: Operating devices outside their specification Calculation errors, losses, resets, failures may affect the application 09.10.2017 Ulf Kulau REAPer Adaptive Micro-Source Energy-Harvester for Wireless Sensor Nodes Page 5

Introduction and Motivation Undervolting in WSNs REAPer HW Design REAPer Evaluation Software Implementation Field Test Summary Is this a good idea? Undervolting will lead to a higher unreliability: Operating devices outside their specification Calculation errors, losses, resets, failures may affect the application Our Perspective: WSNs need increased energy efficiency and offer fault tolerance (ideal) Fulfill WSN tasks even with limited energy budget! 09.10.2017 Ulf Kulau REAPer Adaptive Micro-Source Energy-Harvester for Wireless Sensor Nodes Page 5

Introduction and Motivation Undervolting in WSNs REAPer HW Design REAPer Evaluation Software Implementation Field Test Summary IdealVolting – Adaptive undervolting scheme IdealVolting implementation on undervolting capable node INGA v1.6.1 Primary MCU SPI Transceiver Secondary MCU I2C ATtiny84 ATmega1284p AT86RF233 1. Control loop to ascertain ideal voltage levels → Find most energy efficient but reliable operating point individually 2. Supervised-Learning approach → Collect and predict ideal operating points Kulau et.al., IdealVolting – Reliable Undervolting on Wireless Sensor Nodes , ACM Transactions on Sensor Networks (TOSN), 2016 09.10.2017 Ulf Kulau REAPer Adaptive Micro-Source Energy-Harvester for Wireless Sensor Nodes Page 6

Introduction and Motivation Undervolting in WSNs REAPer HW Design REAPer Evaluation Software Implementation Field Test Summary Architecture of REAPer Integrate IdealVolting to energy harvesting and vice versa... Energy-Harvester 1 . 71 V ≤ V ≤ 2 . 44 V Sensor Node Primary MCU Wiper-Pos. Potentiometer I2C ATmega1284p I2C Buck-Converter Settings GPIO Harvester Status BQ25570 Secondary MCU ATtiny84 Voltage Energy Storage ADC V STOR 09.10.2017 Ulf Kulau REAPer Adaptive Micro-Source Energy-Harvester for Wireless Sensor Nodes Page 7

Introduction and Motivation Undervolting in WSNs REAPer HW Design REAPer Evaluation Software Implementation Field Test Summary Static Overhead Efficiency Static current consumption of REAPer Quiescent current of the entire REAPer platform Test conditions: Energy Storage initially charged to V STOR = 5 V , no load at buck-converter Mean (nA) Min (nA) Max (nA) Normal 567 . 23 ± 15 . 45 546 . 0 592 . 0 Normal + Buck 708 . 24 ± 13 . 93 672 . 0 742 . 0 09.10.2017 Ulf Kulau REAPer Adaptive Micro-Source Energy-Harvester for Wireless Sensor Nodes Page 8

Introduction and Motivation Undervolting in WSNs REAPer HW Design REAPer Evaluation Software Implementation Field Test Summary Static Overhead Efficiency Static current consumption of REAPer Quiescent current of the entire REAPer platform Test conditions: Energy Storage initially charged to V STOR = 5 V , no load at buck-converter Mean (nA) Min (nA) Max (nA) Normal 567 . 23 ± 15 . 45 546 . 0 592 . 0 Normal + Buck 708 . 24 ± 13 . 93 672 . 0 742 . 0 → Reasonable overhead below 1 µ A 09.10.2017 Ulf Kulau REAPer Adaptive Micro-Source Energy-Harvester for Wireless Sensor Nodes Page 8

Introduction and Motivation Undervolting in WSNs REAPer HW Design REAPer Evaluation Software Implementation Field Test Summary Static Overhead Efficiency Charging characteristics Exemplary charging curve at V in = 1000 mV input voltage (Energy storage: Cap 1 F) t 0 → 1 : Cold start phase for V STOR ≤ 1 . 8 V with integrated charge-pump t 1 → 2 : Boost-Converter and duty-cycled MPPT is active for V STOR > 1 . 8 V I in V STOR 5 10 t 2 4 5 . 35 V V STOR [ V ] I in [ mA ] 3 t 1 1 . 8 V 5 2 t 0 1 0 V 0 0 0 100 200 300 400 500 700 600 800 t [ s ] 09.10.2017 Ulf Kulau REAPer Adaptive Micro-Source Energy-Harvester for Wireless Sensor Nodes Page 9

Introduction and Motivation Undervolting in WSNs REAPer HW Design REAPer Evaluation Software Implementation Field Test Summary Static Overhead Efficiency Efficiency of the charging Considering the energy that is stored by the capacitor ( C = 1 F) E = 1 2 · CV 2 (1) Efficiency η can be derived by comparing stored Energy against input energy: E η = (2) E in t 0 → t 2 Where E in t 0 → t 2 is based on... the time of charge t 0 → t 2 , the input current I in and the input voltage V in 09.10.2017 Ulf Kulau REAPer Adaptive Micro-Source Energy-Harvester for Wireless Sensor Nodes Page 10