290 likes | 301 Views
Energy Metering for Free: Augmenting Switching Regulators for Real-Time Monitoring. Prabal Dutta † , Mark Feldmeier ‡ , Joseph Paradiso ‡ , and David Culler †. Computer Science Division † University of California, Berkeley {prabal,culler}@cs.berkeley.edu. The Media Laboratory ‡
E N D
Energy Metering for Free: Augmenting Switching Regulators for Real-Time Monitoring Prabal Dutta†, Mark Feldmeier‡, Joseph Paradiso‡, and David Culler† Computer Science Division† University of California, Berkeley {prabal,culler}@cs.berkeley.edu The Media Laboratory‡ Massachusetts Institute of Technology {carboxyl,joep}@mit.edu
Energy is a critical resource in this domain… So, why don’t more publications provide empirical evidence of a change in energy usage in situ or at scale?
Current energy metering techniques are inadequate cumbersome, expensive, not distributed, not scalable, not embedded cumbersome, expensive, not distributed, not scalable, not embedded, DS2438 ADM1191 BQ2019 BQ27500 [Jiang07] low resolution, low responsiveness, high quiescent power low responsiveness, high cost, high quiescent power
How simply can energy metering be performed? If your platform has a PFM switching regulator…(increasingly, many do) very simply: iCountenergymeterdesign • The network-wide cost of the CSMA overhearing problem • Energy division between route-through and local traffic • Energy benefits of batching or compressing data
This simple design works surprisingly well MAX1724 Our implementation
Outline • Introduction • How does it work? • How well does it work? • How much does it cost? • What are its limitations? • How could it be used?
How does it work? E=½Li2 Lx PFM Regulator Vout Vin VLX S2 iLX Vin Cin S1 Rload Cout Energize Transfer Monitor Source: Maxim Semiconductor
The key insight: each regulator cycle transfers a fixed amount of energy to the load ΔE=½Li2 P=ΔE/Δt
Outline • Introduction • How does it work? • How well does it work? • Range • Accuracy • Resolution • Responsiveness • Precision • Stability • How much does it cost? • What are its limitations? • How could it be used?
A typical mote-class system exhibits a 10000:1dynamic range in current draw (5 µA to 50 mA) iCount offers a dynamic range exceeding 100000:1
iCount exhibits less than ±20% errorover five decades of current draw Common Operating Points iCount exhibits lower error over mote operating range
A Telos mote uses about 20 µJ per second when sleeping iCount resolves less than 1 µJ
A mote’s energy-consuming eventscan occur in as little as 100 µs [Jiang07] iCount responds in less than 125 µsto sudden changes in current draw
iCount is precise over short periods (2 sec) so one or two samples is enough to estimate the instantaneous current All samples fall within ±2% of the median
iCount is stable over long periods (1 week) All samples fall within ±1% of the median
Outline • Introduction • How does it work? • How well does it work? • How much does it cost? • Hardware • Software • Energy • What are its limitations? • How could it be used?
Hardware costs include a wireand a microcontroller counter “wire” Counter HydroSolar Node (v2)
Software costs include initializing hardwareand handling load-dependent counter overflows Initialization Overflow Control Access (15 µs)
Energy costs include switching gate capacitors and handling load-dependent counter overflows 0.01% 1%
Outline • Introduction • How does it work? • How well does it work? • How much does it cost? • What are its limitations? • Efficiency • Voltage dependence • Calibration • How could it be used?
Regulator inefficiency can makebattery gas gauging challenging
Input voltage dependence requires calibration(not fundamental, but an artifact of the MAX1724)
Calibration is required eitherat manufacturing or at run-time Reg Calibration
Estimating per-component current draws from the aggregate Log Regression X = [ones(size(R)) R G B]; p = dE ./ dt; i = p / 3; a = X\i; y = [dt transpose(a)];
Conclusion iCount - simple, functional, research-enabling research
Future directions and enabled research • Hardware profiling – estimating per-subsystem power draw • Model validation – do theory and practice agree in practice and at scale? • Real-time current metering – measuring the instantaneous current draw • Software energy profiling – where have all the joules gone? • Runtime adaptation – equal-energy scheduling by the operating system • Gas gauging – estimating remaining battery energy • Voltage independence – ensuring a cycle delivers the same energy independent of input voltage
Performance summary * Frequency averaged over 1 second