slides - Laura Marie Feeney

How do the dynamics of battery
discharge affects sensor lifetime?
Laura Marie Feeney (SICS)
Christian Rohner, Per Gunningberg (Uppsala Univ)
Anders Lindgren, Lars Andersson (Pricer AB)
Battery discharge behavior
●
complex electro-chemical system
●
modeled as a simple “bucket of mA-h”
●
contribution: characterize discharge behavior
–
WONS 2014
2014/04/03
protocol design and evaluation; state of charge estimation
2
current
Wireless sensor networks
time
●
WSN power consumption profile is complex
●
quantify key macroscopic behaviors
–
synthetic loads, systematically defined load patterns
–
intensity and duration parameters based on typical WSN
WONS 2014
2014/04/03
3
Testbed
●
apply load and measure battery output voltage
●
Li coin cell (CR2032)
–
personal/body area networks, wildlife
WONS 2014
2014/04/03
4
Duty cycle
load(current)
100%
60%
30%
3%
time
●
lifetime estimation
–
based on lifetime at 100% duty cycle
WONS 2014
2014/04/03
5
Duty cycle
100%
load (current)
x
7
.
1
x
3
.
3
60%
30%
x
3
3
3%
time
●
lifetime estimation
–
based on lifetime at 100% duty cycle
WONS 2014
2014/04/03
6
Lifetime estimation
load (current)
100%
●
x %
7
.
1 60
1
+
x
3
.
3 0%
6
2
+
x
33 0%
7
2
time
+
60%
30%
3%
for a 10 mA load
–
under-estimates observed lifetime by as much as a factor of 3!
WONS 2014
2014/04/03
7
Lifetime estimation
load (current)
100%
●
60%
30%
3%
for a 10 mA load
–
●
x
7
.
1 7x
.
2
x
3
.
3 x
5
.
8
x
33 0 x
9
time
under-estimates observed lifetime by as much as a factor of 3!
for a 4 mA load
–
only slight under-estimate (~10-15%)
WONS 2014
2014/04/03
8
Why?
●
charge recovery
–
●
intermittent load utilizes more capacity than continuous load
rate dependent capacity
–
low current utilizes more capacity than a higher one
WONS 2014
2014/04/03
9
load (current)
Load Intensity
time
●
●
model consumed capacity =
∑ i×t
or
∫ i(t )dt
compare load patterns with same time-average current
WONS 2014
2014/04/03
10
load (current)
Load Timing
time
●
●
model consumed capacity =
∑ i×t
or
∫ i (t )dt
compare load patterns with same time-average current
WONS 2014
2014/04/03
11
Consumed capacity
●
compare load patterns with same time-average current
●
expect them to consume same total capacity
WONS 2014
2014/04/03
12
Consumed capacity
5
1
%
s
e
c
n
e
r
e
f
f
i
d
●
compare load patterns with same time-average current
●
expect them to consume same total capacity
WONS 2014
2014/04/03
13
What is a battery? It's complicated.
●
●
primary (non-rechargeable) Li-coin cell
–
Li anode oxidized: Li → Li+ + e-
–
Mn02 cathode reduced: Mn02 + Li+ + e- →LiMn(III)02
discharge behavior depends on chemistry and structure
–
even manufacturer specific
WONS 2014
2014/04/03
14
What is a battery? It's complicated.
●
●
primary (non-rechargeable) Li-coin cell
–
Li anode oxidized: Li → Li+ + e-
–
Mn02 cathode reduced: Mn02 + Li+ + e- →LiMn(III)02
discharge behavior depends on chemistry and structure
–
even manufacturer specific
WONS 2014
2014/04/03
15
What is a battery? It's complicated.
Li+
●
●
primary (non-rechargeable) Li-coin cell
–
Li anode oxidized: Li → Li+ + e-
–
Mn02 cathode reduced: Mn02 + Li+ + e- →LiMn(III)02
discharge behavior depends on chemistry and structure
–
even manufacturer specific
WONS 2014
2014/04/03
16
What is a battery? It's complicated.
Li+
●
●
e-
primary (non-rechargeable) Li-coin cell
–
Li anode oxidized: Li → Li+ + e-
–
Mn02 cathode reduced: Mn02 + Li+ + e- →LiMn(III)02
discharge behavior depends on chemistry and structure
–
even manufacturer specific
WONS 2014
2014/04/03
17
What is a battery? It's complicated.
Li+
e-
i
●
●
primary (non-rechargeable) Li-coin cell
–
Li anode oxidized: Li → Li+ + e-
–
Mn02 cathode reduced: Mn02 + Li+ + e- →LiMn(III)02
discharge behavior depends on chemistry and structure
–
even manufacturer specific
WONS 2014
2014/04/03
18
~10mA
current
Output voltage under load
6.2ms
time
●
apply a load to a battery
●
e.g. lower bit-rate, sub-GHz transceiver
WONS 2014
2014/04/03
19
Output voltage under load
~10mA
6.2ms
●
battery output voltage drops
–
internal resistance
–
efficiency of electro-chemical reactions
WONS 2014
2014/04/03
20
Output voltage under load
~10mA
6.2ms
time
●
battery output voltage drops
–
internal resistance
–
efficiency of electro-chemical reactions
WONS 2014
2014/04/03
21
Battery output voltage under load
85% soc
20% soc
1mA load, 11h
●
load response also depends on state-of-charge
WONS 2014
2014/04/03
22
Battery output voltage under load
2.78V
2.96V
2.53V
2.82V
85% soc
20% soc
1mA load, 11h
●
load response also depends on state-of-charge
.
WONS 2014
2014/04/03
23
Load
current
22mA x 10s
consumes
~5% capacity
1mA x 11h rest x 11h
time
●
alternate short intense loads and low current loads
●
drain ~5% capacity each period
–
based on standard battery test sequence
WONS 2014
2014/04/03
24
Device lifetime
ETC Battery and Fuel Cells Sweden, AB
●
rate dependent capacity
●
charge recovery
●
state of charge dependence
WONS 2014
2014/04/03
25
Device lifetime
device fails
●
●
ETC Battery and Fuel Cells Sweden, AB
2.0V cut-off
device failure
–
battery cannot maintain output voltage under load
–
depends on cut-off voltage for device electronics
not determined by the charge consumed
WONS 2014
2014/04/03
26
Device lifetime
device fails
●
●
ETC Battery and Fuel Cells Sweden, AB
2.0V cut-off
device failure
–
battery cannot maintain output voltage under load
–
depends on cut-off voltage for device electronics
not determined by the charge capacity consumed
WONS 2014
2014/04/03
27
Testbed
●
WONS 2014
2014/04/03
macroscopic behaviors are
well understood
–
rate-dependent capacity
–
charge recovery
–
soc dependence
–
temperature (future)
28
Testbed
●
very little data for cheap, non-rechargeable batteries
–
Panasonic Li-coin cell (CR2032), ~225mA-h
WONS 2014
2014/04/03
29
Testbed
●
large-scale measurement program – cost efficient
●
simple, controlled resistive loads
–
sensor behavior is messy, simplified loads with realistic
parameters
–
variation between batteries (cheap manufacture)
WONS 2014
2014/04/03
30
WSN-typical parameters
●
WONS 2014
2014/04/03
Experiments
•
1 - 25 mA
•
2 – 150 ms load
•
50 ms - 2.0s rest
•
300 μA mean
•
20-22ºC (ambient)
31
WSN-typical parameters
●
CC2420 (250kb/s)
–
–
●
Experiments
tx: 17mA
•
1 - 25 mA
rx: 18mA
•
2 – 150 ms load
•
●
–
●
50 ms - 2.0s rest
•
300 μA mean
•
20-22C (ambient)
contikiMAC
125ms (2-16Hz)
pTunes
–
6/100ms
–
11/350ms
q
●
cpu (active)
–
–
●
1.2-2.4mA (tmotesky)
4.3-5.7mA
(cc2480)
adc
–
●
1.2mA
WONS 2014
2014/04/03
●
TR1001 (2-115kb/s)
●
ANT+
–
tx: 12mA
–
search: 2.8mA
–
rx: 1.3-3.8 mA
–
tx: 11-15mA
–
rx: 17mA
–
avg: 20-200μA
SX1212 (25kb/s)
–
rx: 3mA
–
tx: 16mA
32
Experiments
●
simple, synthetic loads, systematic investigation
–
●
●
over 50 experiment configurations
loads with same time-average current
–
focus on intensity (load current) and timing
–
reduce rate dependent capacity effects
metric: capacity consumed (until 2.0V cut-off)
WONS 2014
2014/04/03
33
Results
●
rate-dependent capacity
–
as current decreases, utilizable capacity increases
WONS 2014
2014/04/03
34
Results
●
load patterns with 1mA load current
WONS 2014
2014/04/03
35
Results
●
load patterns with 4mA peak current
WONS 2014
2014/04/03
36
Results
●
load patterns with 10mA load current
WONS 2014
2014/04/03
37
Results
●
load patterns with 25mA load current
WONS 2014
2014/04/03
38
Applications
●
design
–
●
evaluation
–
●
data suggests high currents with low duty cycle is OK
parameterization and validation of analytic battery models for use
in e.g. simulation
real time state of charge estimation
–
state of charge on deployed device
–
needed for management, also lifetime balancing
–
lack of on-device resources
–
off-line modeling? on-device voltage tracking?
WONS 2014
2014/04/03
39
Conclusions
●
●
in-depth characterization of Li-coin cell (CR2032)
–
large scale testbed for systematic measurement
–
not a bucket of mA-h
applications
–
design, evaluation, state of charge estimation
WONS 2014
2014/04/03
40