Transcript Battery Monitoring Basics

Battery Monitoring Basics
October 2012 Dallas
1
Section 1 – Basic Concepts
• What does a battery monitor do?
• How to estimate battery capacity?
– Voltage lookup
– Current integration
• Factors affecting capacity estimation
• Other functions
– Safety and protection
– Cell balancing
– Charging support
– Communication and display
– Logging
October 2012 Dallas
2
2
What does a battery monitor do?
Battery Subsystem
VPACK
CHG
DSG
Vbatt
ICHG
comm
Load
Gas
Gauge
Tbatt
Battery
VDSG
Charger
VCHG
IDSG
System
October 2012 Dallas
Rs
Monitor
Ibatt
• Capacity
estimation
• Safety/protection
• Charging support
• Communication
and Display
• Logging
• Authentication
Cell
3
3
How to estimate battery capacity?
• Measure change in capacity
– Voltage lookup
– Coulomb counting
• Develop a cell model
– Circuit model
– Table Lookup
October 2012 Dallas
4
4
Voltage lookup
• One can tell how much water is
in a glass by reading the water
level
– Accurate water level reading
should only be made after the
water settles (no ripple, etc)
• One can tell how much charge
is in a battery by reading wellrested cell voltage
– Accurate voltage should only be
made after the battery is well
rested (stops charging or
discharging)
October 2012 Dallas
mL
marks
I(t)
q(t )
V(t)
5
5
OCV curve
Level
rises
same rate
OCV Curve
Voltage
Level
rises
same rate
Full charge voltage
End of discharge voltage
Capacitor
100%
0%
Fullness
Level rises
slower
Voltage
OCV Curve
Level rises
faster
Full charge voltage
End of discharge voltage
Battery
October 2012 Dallas
0%
100%
Fullness
6
6
OCV voltage table: DOD representation
OCV(DOD)
4300
4100
3900
Voltage_a(DOD)
Vmax
Vmin
3700
Voltage_a
3500
Poly_a(DOD)
3300
3100
2900
0
0.2
0.4
0.6
0.8
1
1.2
DOD
Flat Zone
DOD = Depth of Discharge
SOC = State of Charge
DOD = 100% - SOC
October 2012 Dallas
7
7
Current integration
• One can also measure how
much water goes in and out
• In batteries, battery capacity
changes can be monitored by
tracking the amount of
electrical charges going in/out
mL
marks
q (t ) q 0   I (t )  dt
qk q 0 t  k I k
• But how do you know the
amount of charge, q0 , already
in the battery at the start?
• How do you count charges
accurately?
October 2012 Dallas
I(t)
q(t )
Voltage
8
8
Basic Smart Battery System
CHG
DSG
VPACK
Vbatt
ICHG
comm
Gas
Gauge
Tbatt
Battery Model
VDSG
Load
Charger
VCHG
IDSG
Rs
Ibatt
qk q 0 t  k I k
October 2012 Dallas
9
9
Circuit model
Rint
RS
RL
Vbatt
Ibatt
•
•
•
Voc(SOC)
CS
CL
•
DC model
Rint
RS
•
RL
Vbatt
•
VOC a function of SOC
Rint is internal resistance
Rs and Cs model the short
term transient response
RL and CL model the long term
transient response
Vbatt and Ibatt are the battery
voltage and current
All parameters are function of
temperature and battery age
Ibatt
Voc(SOC)
CS
CL
Transient model
October 2012 Dallas
10
10
Table lookup
• Large, multi-dimensional table relating capacity to
– Voltage
– Current
– Temperature
– Aging
• No cell model
• Apply linear interpolation to make lookup “continuous”
• Memory intensive
October 2012 Dallas
11
11
Factors affecting capacity estimation
•
•
•
•
•
October 2012 Dallas
PCB component accuracy
Instrumentation accuracy
Cell model fidelity
Aging
Temperature
12
12
PCB component accuracy
• Example
– Current sensing resistor
– Trace length (resistance)
Gas
Gauge
V (t )  I (t )  Rs
R+
RI (t )
Rs
I (t ) 
October 2012 Dallas
V (t )
Rs   r
13
13
•
•
•
•
ADC Resolution
Sampling rate
Voltage drift / calibration
Noisy immunity
Voltage
Instrumentation accuracy
ADC count
October 2012 Dallas
14
14
Battery model fidelity
Rint
• Steady-state (DC)
• Transient (AC)
• Capacity degradation
RS
RL
Vbatt
Ibatt
Voc(SOC)
CS
CL
DC model
– Aging
– Overcharge
Rint
RS
RL
Vbatt
Ibatt
Voc(SOC)
CS
CL
Transient model
October 2012 Dallas
15
15
Model parameter extraction
• Extract battery model parameter values using actual
collected battery data
– Open circuit voltage (OCV)
– Transient parameters (RC)
– DC parameters (Ri)
• Least square minimization
• Extraction process can be hard and time consuming
October 2012 Dallas
16
16
Temperature
• Temperature is important for
– Capacity estimation
– Safety
– Charging control
• Temperature impacts model
parameters
–
–
–
–
Resistance
Capacitance
OCV
Max capacity
October 2012 Dallas
17
17
Safety
– Accelerates cell degradation
– Thermal runaway and
explosion
• LiCoO2 – Cathode reacts with
electrolyte at 175°C with 4.3 V
• Cathode coatings help
considerably
Heat Flow (W/g)
• High operating temperature
OCV = 4.3 V
Thermal
Runaway
100 125 150 175 200 225 250
Temperature (°C)
• LiFePO4 shows huge
improvement! Thermal runaway is
> 350°C
October 2012 Dallas
18
18
Cell Safety
Safety Elements
•
Pressure relief valve
•
PTC element
•
Aluminum or steel case
•
Polyolefin separator
– Low melting point
(135 to 165°C)
– Porosity is lost as melting
point is approached
– Stops Li-Ion flow and shuts
down the cell
•
Recent incidents traced to
metal particles that pollutes the
cells and creates microshorts
October 2012 Dallas
19
19
Safety and protection
• Short circuit
• Over/under
(charge/discharge)
current
• Over/under voltage
• Over temperature
• FET failure
• Fuse failure
• Communication failure
• Lock-up
• Flash failure
• ESD
• Cell imbalance
Trip-Over
Trip
Margin
(level)
Alert
Trip
Trip
Trip
Level
time
Trip Margin
(time)
Trip-Under
Trip
Margin
(level)
Alert
Trip
Trip
Trip
Level
time
Trip Margin
(time)
October 2012 Dallas
20
20
Overcurrent Protection Details
Battery
Current
AFE
SCP (CHG and DSG)
Turn Off FETs
AFE
Hardware Protection
Recoverable
Gas-Gauge IC
Software Control
Both CHG and DSG
(1-s Update Interval)
AFE
OCP (DSG Only)
Turn Off FETs
Recoverable
2nd-Level Safety OCP
(Blow Chemical Fuse)
Permanent
Recoverable
Recoverable
1st-Level OCP
(2nd Tier)
1st-Level OCP
Turn Off FETs
(1st Tier)
Turn Off FETs
Time
AFE SCP CHG AFE OCP
DSG Time
/DSG Time
0 to ~915 µs 1 to ~31 ms
October 2012 Dallas
Safety OCP CHG/ OCP (2nd Tier) OCP (1st Tier)
DSG Time
CHG/DSG Time CHG/DSG Time
1 to ~60 s
1 to ~60 s
1 to ~60 s
21
21
Basic Battery-Pack Electronics
Charge MOSFET
Discharge MOSFET
Pack+
SMD
SMC
RT
Q1 Q2
Chemical Fuse
Gas Gauge IC
SMBus Overvoltage
Undervoltage
Temp Sensing
bq20z90
LDO
I 2C
AFE
OCP
Cell
Balancing
Second
Safety
OVP IC
bq29412
bq29330
Voltage ADC
Current ADC
Rs
Sense
Resistor
Pack–
• Measurement: Current, voltage, and temperature
• bq20zxx gas gauge : Remaining capacity, run time, health condition
• Analog front end (AFE)
October 2012 Dallas
22
22
JEITA/BAJ Guidelines for Notebook
• Do not charge if T< 0°C or T> 50°C
• Minimize temperature variation among cells
• How do we collect temperature information?
Upper-Limit Charge Current
Upper-Limit Voltage: 4.25 V
Safe Region
T2 T5
(100C)
T6
T3
(450C)
No Charge
No Charge
4.15 V
T1
October 2012 Dallas
4.20 V
T4
23
23
Why Are Battery Packs Still Failing?
→ Heat Imbalance
• Space-limited design
causes local heat imbalance
• Cell degradation
accelerated
• Leads to cell imbalance
>10ºC
Variation
Between
Cells
Temperature Profile
along Section Line
• Single/insufficient thermal
sensor(s) compromise
safety
October 2012 Dallas
24
24
Cell Balancing
Battery cells voltages can get out of balance, which
could lead to over charge at a cell even though the
overall pack voltage is acceptable.
Cell balance can be achieved through current
bypass or cross-cell charge pumping
October 2012 Dallas
25
25
Passive Cell Balancing: Simplest Form
Rext1
• Simple, voltage based
• Stops charging when
any cell hits VOV
threshold
• Resistive bypassing
turns on
• Charge resumes when
cell A voltage drops to
safe threshold
+
VDiff_End
Rext2
VOV
VOV – VOVH
Cell A
VDiff_Start
bq77PL900, 5 to 10 series-cell Li-Ion
battery-pack protector for power tools
October 2012 Dallas
Ibalance
Battery
Cell
Cell B
ta
tb
tc
td
te
tf
26
26
Fast Passive Cell Balancing
PACK +
1 kW
R4
Cell 2
R1
Q2
1 kW
R4
Cell 1
R2
bq2084/
bq20zxx
Q2
1 kW
R
• Needed for high-power
packs, where cell selfdischarge overpowers
internal balancing
• Fast cell balancing
strength is 10x ~ 20x
higher
RDS(on)
Internal CB
R3
Fast CB
VCell
ICB 
R DS(on)
ICB
VCell

R4
Where R4 << RDS(on)
October 2012 Dallas
27
27
Charging support
• Inform battery charger proper charging voltage and
current
• Conform to specification (e.g., JEITA)
• Reduce charge time
• Extend battery life by:
– Avoid overcharging
– Precharging depleted and deeply discharged cells
October 2012 Dallas
28
28
Communication and Display
• Communication
– To the System or Charger
– Industry specification
• Display
– LED, LCD
– Capacity indication
– Fault indication
October 2012 Dallas
29
29
Logging
• Works like an airplane “blackbox
recorder”
• Record important lifetime
information
– Max/min voltage
– Max/min current
– Max/min temperature
• Record important data for failure
analysis
– Reset count
– Cycle count
– Excessive flash wear
October 2012 Dallas
30
30
Section 2
Battery Fuel Gauging:
CEDV & Z-track
October 2012 Dallas
31
Basic Vocabulary Review
• Capacity
– Design Capacity [mAh]
– Qmax, Chemical Capacity [mAh]
– FCC, Usable Capacity [mAh]
– RM, Remaining Capacity [mAh]
– RSOC [%]
– DOD [%]
– DOD0, DOD1 [%]
• Voltages
– OCV [mV]
– OCV(DOD) [mV]
– EDV [mV]
– EDV 2 [mV]
– EDV 0 [mV]
– CEDV [mV]
October 2012 Dallas
• Current
– C-rate [mA]
– Coulomb Counting
q (t ) q 0   I (t )  dt
32
32
How Much Capacity is Really Available?
Voltage, V
4.5
Open circuit voltage (OCV)
4.0
I • RBAT
3.5
EDV
3.0
0
1
2
3
4
Capacity, Ah
6
Usable capacity : FCC
Full chemical capacity: Qmax
• External battery voltage (blue curve) V = V0CV – I • RBAT
• Higher C-rate EDV is reached earlier (higher I • RBAT)
October 2012 Dallas
33
33
What Does A Fuel Gauge Do?
Which route is the battery taking?
4.2V
3V
October 2012 Dallas
Suppose we
are here
• What is the remaining
capacity at current load?
• What is the State of charge
(SOC)?
• How long can the battery
run?
0%
7/16/2015
34
34
Current Integration Based Fuel-gauging
• Battery is fully charged
• During discharge capacity
is integrated
• State of charge (SOC) at
each moment is RM/FCC
• FCC is updated every
time full discharge occurs
4.2V
Q
0%
3V
FCC
October 2012 Dallas
RM = FCC - Q
SOC = RM/FCC
7/16/2015
35
35
Learning Before Fully Discharged
– fixed voltage thresholds
• It is too late to learn
when 0% capacity is
reached  Learning
FCC before 0%
4.2V
7%
3%
EDV2
EDV1
• We can set voltage
threshold that
correspond to given
percentage of
remaining capacity
0% •
EDV0
FCC
October 2012 Dallas
7/16/2015
However, true voltage
corresponding to 7%
depends on current
and temperature
36
36
Learning before fully discharged with
current and temperature compensation
CEDV
OCV
4.2V
EDV2 (I1)
EDV2 (I2)
October 2012 Dallas
CEDV Model:
Predict V(SOC) under any
current and temperature
• Modeling last part of
discharge allows to
calculate function
V(SOC, I, T)
• Substituting SOC=7%
allows to calculate in
real time CEDV2
threshold that
corresponds to 7%
capacity at any current
and temperature
7/16/2015
37
37
CEDV Model Visualization
OCV curve defined
by EMF, C0
Voltage
OCV corrected by
I*R (R is defined by
R0, R1, T0)
I*R
Further
correction by low
temperature (TC)
Actual battery
voltage curve
Reserve Cap: C1
shifts fit curve
laterally
Battery Low
3%
October 2012 Dallas
4%
5%
6%
7%
8%
9%
38
38
CEDV Formula
CEDV = CV - I*[EDVR0/4096]*[1 + EDVR1*Cact/16384]*
[1 – EDVT0*(10T - 10Tadj)/(256*65536)]*[1+(CC*EDVA0)/(4*65536)] * age
Where:
CV = EMF*[1 – EDVC0*(10T)*log(Cact)/(256*65536)]
Cact = 256/(2.56*RSOC + EDVC1) – 1 for (2.56*RSOC + EDVC1) > 0
Cact = 255 for (2.56*RSOC + EDVC1) = 0
EDVC1 = 2.56 * Residual Capacity (%) + “Curve Fit” factor
Tadj = EDVTC*(296-T) for T< 296oK and Tadj < T
Tadj = 0 for T > 296 oK and Tadj max value = T
age = 1 + 8 * CycleCount * A0 / 65536.
October 2012 Dallas
39
39
39
Impedance Track Fuel Gauging
• Combine advantages of voltage correlation and coulomb
counting methods
• State of charge (SOC) update:
– Read fully relaxed voltage to determine initial SOC and capacity
decay due to self-discharge
– Use current integration when under load
• Parameters learning on-the-fly:
– Learn impedance during discharge
– Learn total capacity Qmax without full charge or discharge
– Adapt to spiky loads (delta voltage)
• Usable capacity learning:
– Calculate remaining run-time at typical load by simulating voltage
profile  do not have to pass 7% knee point
October 2012 Dallas
40
40
Current Direction Thresholds and Delays
8
1
2
3
7
6
5
1.
2.
3.
4.
5.
6.
7.
8.
CHG relaxation timed
Enter RELAX mode
Start discharging
Enter DSG mode
DSG relaxation timed
Enter RELAX mode
Start charging
Enter CHG mode
4
Example of the Algorithm Operation Mode Changes With Varying SBS.Current( )
October 2012 Dallas
41
41
What is Impedance Track?
1. Chemistry table in Data Flash:
OCV = f (dod)
10,000 foot View
dod = g (OCV)
2. Impedance learning during discharge:
R = OCV – V
I
3. Update Max Chemical Capacity for each cell
Qmax = PassedCharge / (SOC1 – SOC2)
4. Temperature modeling allows for temperature-compensated
impedance to be used in calculating remaining capacity and
FCC
5. Run periodic simulation to predict Remaining and Full
Capacity
October 2012 Dallas
42
42
Close OCV profile for the Same BaseElectrode Chemistry
Voltage, V
4.2
•
OCV profiles close for all
tested manufacturers
•
Most voltage deviations from
average are below 5mV
•
Average DOD prediction
error based on average
voltage/DOD dependence is
below 1.5%
•
Same OCV database can be
used with batteries produced
by different manufacturers as
long as base chemistry is
same
•
Generic database allows
significant simplification of
fuel-gauge implementation at
user side
3.93
3.67
3.4
0
0.1
0.2
0.3
0.4
0.5
0.6
DOD, fraction
0.7
Manufacturer A
B
C
D
E
0.8
4
0.9
1
October 2012 Dallas
r, %
2.67
1.33
43
43
Resistance Update
400
Ra
300
200
100
Before Update
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
dod
Discharge direction
October 2012 Dallas
44
44
Ra Table: Interpolation and Scaling Operation
k: Present grid
m: Last visited grid
Grid 14
Ra_new
Ra_old
Grid 0
• R = (OCV – V) / Avg Current.
Averaging method is selectable
• Resistance updates require
updating 15 values for each cell
• A new resistance measurement
represents the resistance at an
exact grid point. Exact value
found by interpolation
• All 15 grid points are
ratiometrically updated from any
valid gridpoint measurement.
Changes are weighted
according to confidence in
accuracy
Step 1
Interpolation
Step 3
Step 2
Scale “After”
Scale “Before”
October 2012 Dallas
45
45
Timing of Qmax Update
October 2012 Dallas
46
46
FCC Learning
1.3 10
4
8000
1.2 10
4
1.1 10
4
7600
1 10
7400
9000
7200
4
0
0.2
0.4
0.6
0.8
V, mV
FCC, mAh
7800
8000
DOD
SMB FCC
true FCC
Ra grids
Voltage
October 2012 Dallas
47
47
Modeling temperature
R
R
i
• Based on a heating / cooling model **
• Heating is from the internal resistance
• Cooling is from heat transfer to the
environment, i.e., T  Ta 
• How many thermistors?
Ibatt
Voc(Vsoc)
C
Vbatt
hc := heat transfer coef
A := cell surface area
m := cell mass
cp := specific heat
m cp
Ta := ambient temp
dT
1
2
2
 I batt
Ri  Vbatt  Voc  I batt Ri   hc AT  Ta 
dt
R
Heating
Cooling
** “Dynamic Lithium-Ion Battery Model for System Simulation”, L. Gao, S. Liu and R. A. Dougal, IEEE Transaction on Components and Packaging
Technologies, vol. 25, no. 3, September 2002.
October 2012 Dallas
48
48
RemCap Simulation (concept)
Start of discharge
V
I*R
(loaded)
OCV
Δ V > 250mV
EDV
Vterm
Time
ΔQ/2
I
ΔQ/4
Qstart
ΔQ
ΔQ
. . . . .
ΔQ
RsvCap
Time
RemCap
Constant Load Example
October 2012 Dallas
49
49
Z-track Accuracy in Battery Cycling Test
•
Error is shown at
10%, 5% and 3%
points of discharge
curve
•
For all 3 cases, error
stays below 1%
during entire 250
cycles
•
It can be seen that
error somewhat
decreases from 10 to
3% due to adaptive
nature of IT algorithm
Remaining Capacity Error, %
1
0.5
0
0.5
1
1.5
0
50
100
150
Cycle Number
error at 10%
error at 5%
error at 3%
October 2012 Dallas
200
250
300
50
50
CEDV, Impedance Track Comparison
Property
CEDV
Impedance Track
Worst error new, learned
+/-2%
+/-1%
Worst error aged, learned
+30% (+/- 15% with age
data)
+/-2%
Data collection
3 temperatures, 2 rates,
Fitting to obtain
parameters.
2 weeks
Chemistry selection test,
Optimization cycle
1 week
Instruction flash
small
large
Voltage accuracy
requirement
20mV/pack
3mV/pack
State of charge initialization
(host side requirement)
No
Yes
FCC temperature
compensation
No (with rare exceptions)
Yes
FCC rate compensation
No (with rare exceptions)
Yes
Learning cycle in production
required
Not required
October 2012 Dallas
51
51
October 2012 Dallas
52