Teardown: High-voltage Li-ion battery stack

management - the drive for safe power

Stephen Evanczuk - July 31, 2012

At the heart of the Chevrolet Volt, a

sophisticated battery-stack management
system ensures the safety and reliability of
the multicell lithium-ion battery stack that
delivers power on demand to the Volt drive
system.

Within the management system, battery-

monitoring boards use two key subsystems to
reliably monitor cell health and deliver digital
results to a host processor that orchestrates
system operation. Separating those
subsystems, a signal interface ensures
isolation between high-voltage battery-
sensing circuitry and communications devices
on the boards.

In this teardown, we review the challenges associated with high-voltage Li-ion battery-stack
management in automotive applications and discuss how the overall architecture of the Chevy Volt
battery-stack management system meets those challenges. In particular, we discuss the
requirements for Li-ion cell monitoring and focus on the architecture and components used in the
cell-monitoring subsystem, digital-communications subsystem, and isolation interface. We take an in-
depth look at parts selected for this design, including a custom ASIC, the Freescale S9S08DZ32, the
Avago ACPL-M43T, and the Infineon TLE6250G. Finally, we examine the benefits of this specific
solution for mission-critical battery-stack management and consider the trade-offs with possible
alternatives available for similar design challenges.

For further information about the role of isolation in automotive battery management
systems, we have included a series of three in-depth video interviews.
Part 1 addresses the role of isolation in automotive battery management systems;
Part 2 looks at considerations for parts selection for these applications;
Part 3 examines the use of isolation devices in the Chevy Volt battery management
system.

EV Challenges
The Chevrolet Volt is the first production battery EV (electric vehicle), able to run nearly 40 miles
solely on batteries. When battery charge reaches its lower limits, a gasoline engine engages to
generate additional electricity to extend the vehicle’s range by several hundred miles. At the heart of
the vehicle, a lithium-ion battery pack measuring 1.8m in length and weighing 181 kg generates the
16-kWh power needed to turn drive motors, power passenger features, and supply power to a
sophisticated battery-management system that rivals avionic systems in its complexity.

IBM senior vice president Robert LeBlanc has noted that with its 10 million lines of code, the Volt’s
software content surpasses the 7.5 million lines of code said to fly the US DOD F-35 Lightning II
Joint Strike Fighter—a level of software content that itself more than triples the code content of
current jet fighters, according to the US Government Accountability Office. While LeBlanc could
probably have picked a less controversial system for comparison, the Volt does attract its own share
of controversy. Perhaps no other vehicle has faced the same level of scrutiny as the Volt. Indeed,
when a Volt test vehicle caught fire after sitting for weeks following a test crash, the incident kicked
off a government agency review and a buyback offer from GM—even though no battery-related fires
have occurred after "real-world crashes," according to the National Highway Traffic Safety
Administration.

Ultimately, the Volt’s success hinges on public acceptance—and its ability to perform. To that end, in
designing the Volt, GM worked with IBM to simulate performance of the “system of systems” that
power the Volt. Using detailed models of key systems, IBM software verified behavior and even
generated key elements of the software code used in the Volt systems. That approach to code
generation and systems modeling was crucial for ensuring performance of the Volt battery-
management system because of the complex algorithms required to ensure optimal Li-ion cell
performance and lifetime; indeed, optimizing such cells remains a highly active focus of research in
industry, government, and academia. For the Volt, ensuring battery performance resulted in a final
multiboard design (Figure 1) that orchestrates the operation of multiple embedded systems into a
single system responsible for meeting the range, safety, performance, and extended-life
requirements for the Volt's Li-ion battery pack.

Figure 1. The Chevy Volt battery management system partitions functions across multiple
subsystems implemented in several PCBs. The focus of this teardown is the battery-
interface control module—the red, blue, and green boards shown above in the second
column from the right. (Courtesy of UBM TechInsights)
Li-ion technology

Li-ion cell characteristics

The sophisticated system required to meet Volt performance, safety, and reliability requirements
arises directly from the characteristics of Li-ion cells. In a discharging Li-ion cell, lithium is ionized
in a (typically) graphite anode, and the Li ions move through an electrolyte to pass through a
separator to a cathode, resulting in charge flow. The charging process reverses the flow, bringing Li-
ions from the cathode through the separator to the anode.

The performance and reliability of this chemical process are dependent on cell temperature and cell
voltage. At low temperatures, the chemical reaction slows, lowering cell voltage. As temperature
increases, the reaction rate increases until the Li-ion cell components begin to break down. At
temperatures above 100°C, the electrolyte begins to break down, releasing gases that can cause
pressure to build within cells designed without pressure-relief mechanisms. At high enough
temperatures, Li-ion cells can experience thermal runaway as oxides break down, releasing oxygen,
which further accelerates the temperature increase.

Consequently, maintaining optimum operating conditions for Li-ion cells is a critical requirement for
the Volt battery-management system. The problem for Volt engineers was ensuring reliable data
collection and analysis to monitor and control the state of the Li-ion cells in the vehicle—a problem
exacerbated by the nature of Li-ion cells themselves.

One distinguishing feature of Li-ion cell technology is that at a given temperature and output current
level, a Li-ion cell maintains a nearly flat voltage output across the middle of its capacity range
(Figure 2). Although this characteristic heightens Li-ion's advantages as a power source, it
complicates engineers' attempts to use simple cell-voltage measurements to provide users with a
measure of remaining battery capacity, or SOC (state of charge). For Volt drivers, an accurate SOC
measurement is important for estimating vehicle range. Indeed, for a nascent EV market in which
“range anxiety” is a key impediment to more widespread acceptance and sales, an accurate
statement of SOC is essential.

Figure 2. At a given temperature and discharge current level, a Li-ion cell such as this
Panasonic CGR18650CG exhibits nearly flat output voltage in the middle portion of its
discharge range. That's an advantage for a power source but a complication for engineers
who need to measure state of charge accurately. (Courtesy of Panasonic)

Furthermore, maintaining SOC within a specific range is important for ensuring extended battery
life. Batteries kept at too low or too high a state of charge will tend to degrade faster than those kept
at intermediate levels, whose specific range is generally determined empirically. If allowed to
discharge too completely, Li-ion cell components begin to deteriorate, leading to permanent
damage; if allowed to charge beyond the recommended upper voltage range, cells can overheat or
experience permanent changes in structure.

In the Volt, GM engineers established a safe SOC window of 58% to 65%, depending on driving
mode, with a lower limit in normal driving mode set to a 30% SOC and higher “mountain driving”
mode, set to 45%, to ensure available capacity while driving up inclines for extended periods. When
the Volt reaches the appropriate lower SOC limit, the vehicle’s gasoline engine engages for
extended-range operation.

Estimating state of charge

Because of the uncertainty surrounding SOC measurement in Li-ion cells, engineers can only create
SOC estimates, typically performed using current-based or voltage-based methods.

Current-based methods offer the most accurate results. Such methods keep track of changes in
charge, essentially counting coulombs added to a battery during charging or subtracted during
discharge cycles and determining SOC relative to the cell in its fully charged state. Self-discharge
losses or inefficiencies in the battery itself, however, can make the “coulomb counting” method
subject to error. In addition, because continuous monitoring is not practical for many applications,
coulomb-counting methods need to use a sampling approach. In automotive applications, the
approach must be sufficiently fast and adaptive to track the rapid discharges associated with
acceleration and the rapid charging associated with regenerative braking.

Voltage-based methods estimate SOC by using the instantaneous voltage output from a cell as the
basis of further calculations that take into account such variables as cell temperature, age, current
output, and discharge rate. When used with accurate characterization data for individual Li-ion cells
across multiple operating conditions, voltage-based methods can deliver accurate SOC estimates.
For a production vehicle such as the Volt, maintenance procedures account for the need for accurate
cell characterization and provide specific tools and procedures that enable the battery-management
system to learn the capacity of a new battery module—or relearn battery capacity if necessary. Li-
ion chemistry

Li-ion chemistry
Li-ion cells include diverse chemistries, each offering particular characteristics for energy density,
efficiency, durability, and nominal cell voltage. The cells that LG Chem manufactured for the Volt
use that company’s manganese spinel cathode lithium-ion chemistry and its proprietary safety-
reinforced separator—a ceramics-coated, semipermeable membrane. Industrywide, Li-ion cells are
manufactured in a variety of forms, including familiar cylindrical shapes; flatpacks, used in mobile
phones; and rigid-plastic prismatic shapes. For the Volt, LG Chem’s basic cell uses a prismatic
package.

As described by analysts at UBM TechInsights and Munro & Associates, the full Chevy Volt battery
pack comprises 288 prismatic Li-ion cells, which are in turn packaged into 96 battery-cell groups to
deliver a system voltage that the analysts measured at 386.6V dc. Those battery-cell groups, in turn,
are combined with temperature sensors and cooling elements into four main battery modules.
Voltage-sense lines attached to each cell group terminate in a connector on top of each battery
module, and a voltage-sense harness joins the connector to a battery-interface module that sits on
top of each battery module. Here, four color-coded battery interface modules operate at different
positions in the battery stack, corresponding to low, medium, and high voltage ranges of dc voltage
offset for the set of four modules.

Data from the battery-interface modules moves upstream to the battery-energy-control module. That
module in turn passes fault conditions, status, and diagnostic information to the hybrid-powertrai-
-control module, which serves as the host controller for vehicle level diagnostics. At any time, the
overall system runs more than 500 diagnostics every tenth of a second; 85% of those diagnostics
focus on battery-pack safety, and the remainder target battery performance and life.

Multilayer board
The downstream analysis of battery performance starts with the focus of this teardown: the battery-
interface control module (Figure 3). This module is a four-layer PCB with the majority of components
mounted on the top, along with orange battery connectors and a black data-communications
connector. The uppermost layer includes a ground plane and signal traces with multiple vias
providing connections to lower layers. In the second layer, power and ground planes spread under
the high-voltage areas of the PCB; the third layer contains signal traces passing under those areas.
The other side of the PCB, the fourth layer, is used for the ground plane and signal traces, and
contains some additional components.

Figure 3. Each of the four battery-interface control-module PCBs in the Chevy Volt
combines multiple sensing circuits with CAN communications circuitry isolated by
optocouplers sitting at the edge of the communications subsystem.
(Courtesy of UBM TechInsights)

The black ATLPB-21-2AK PCB-mounted connector carries 5V reference, low-reference, signal-

ground, CAN bus high-serial-data, CAN bus low-serial-data, and high-voltage-fault signals. The
orange battery connector carries the battery-module-temperature signal, low-reference, and voltage-
sense lines from the battery-cell groups. Sensing subsystem
Sensing subsystem
The heart of the battery-interface control system is a sophisticated sensing subsystem—a complete
embedded system circuit responsible for monitoring the voltage output from each Li-ion cell group
and monitoring the temperature of the battery pack. Cell voltage passes through the battery
connector to the L9763, an ASIC developed by STMicroelectronics and LG Chem.

The L9763 ASIC can monitor up to 10 individual Li-ion cell groups, using on-chip current-sense
amplifiers for cell-load-current monitoring and an on-chip analog multiplexer and sample-and-hold
circuit for cell-voltage monitoring (Figure 4). The device's differential inputs address the need to
ensure millivolts-accurate measurements despite large offset voltages, depending on the cell’s
location in the battery pack. Furthermore, PCB designers used a combination of trace-layout
technique, isolation techniques, and the ground planes mentioned earlier to help ensure signal
integrity in this challenging environment.

Figure 4. The L9763 ASIC includes on-chip circuitry for measuring voltage and current
from Volt cell groups as well as for balancing charge in those cells through passive
resistive cell-balancing techniques. (Courtesy of STMicroelectronics)

Based on those measurements, the device’s on-chip circuitry switches individual cell groups into
external resistor networks to discharge those cells selectively to reduce stress arising from large
voltage differences. This simple, passive technique provides a reliable, low-cost solution for cell
balancing but loses efficiency because energy is lost as heat in discharge resistors (Figure 5).
Alternative cell-balancing approaches use active methods, wherein charge from the highest voltage
cells is stored and redistributed to the lowest cells; the techniques use capacitors, inductors, or
transformers to store or redistribute the charge by switching sequentially between each cell to
accumulate or discharge charge as appropriate. What active methods might gain in energy
conservation over passive methods, however, they trade off in increased system cost and complexity.
 Figure 5. Passive cell balancing (left) switches high cells to a discharge resistor; active
cell-balancing techniques sequentially accumulate charge on a capacitor (shown on the
right) or inductor or use a transformer to distribute charge to low cells.
(Courtesy of STMicroelectronics)

For charging or discharging multicell Li-ion battery packs, designs typically use a constant-current
and constant-voltage method, in which the charging system uses a pair of MOSFETs to reduce
charge current when the desired charge voltage has been reached or increase current during
discharge operations. The L9763 provides a charge pump to drive the power MOSFET devices. The
ASIC passes measurement data for its monitored Li-ion cells through its SPI interface to the
Freescale S9S08DZ32 MCU. The L9763 also provides a 5V LDO output for the MCU. For overall
battery-management functions, individual L9763 devices are linked through on-chip interfaces and
individually addressed through vertical daisy-chain communication by the primary control unit.

Sensing-circuit MCU
As described earlier, SOC estimation for Li-ion cells is a complex task, requiring plenty of processing
horsepower. For this design, each sensing subsystem pairs an L9763 ASIC with a Freescale
S9S08DZ32 40-MHz HCS08 MCU packing internal 32-kbyte flash, 2-kbyte RAM, and 1-kbyte
E2PROM on-chip. An external 4-MHz oscillator provides a reference frequency for MCU clock
operations.

In the GM-LG Chem design, the MCU would be expected to perform the required calculations
needed to estimate SOC based on the voltage and current measurements the L9763 provides.
Although the SOC algorithm is proprietary, the hardware-configuration and -maintenance
procedures suggest these estimation algorithms combine voltage-driven estimates using stored cell-
characterization data, with more direct charge measurement during charge periods for interim
recalibration. The use of the detailed systems-modeling environment described by IBM would
provide an ideal platform for finding appropriate data sets for optimizing SOC calculations, as well
as for validating the approach across a widely sampled set of operating conditions.

HCS08 safety features such its computer-operating-properly watchdog timer help ensure reliable
operation and cause an automatic reset in the event of unrecoverable application-software faults. Of
particular importance in this application, the S9S08DZ32 includes a sophisticated on-chip CAN
controller that can be selectively powered down or placed into sleep mode when not in active use
(Figure 6). To help ensure predictable real-time performance, the on-chip controller includes five
receive buffers, organized as a FIFO buffer, and three transmit buffers that allow prioritization of
outbound messages.

Figure 6. An on-chip CAN controller is a key feature driving the selection of the Freescale
S9S08DZ32 MCU for the battery-interface control-module sensing subsystem.
(Courtesy of Fairchild Semiconductor)
Signal isolation

Signal isolation
In the Chevy Volt system of systems, communications and control are the cornerstone of vehicle
operations, and the Volt provides multiple networks that isolate and protect individual subsystems.
Complex algorithms manage individual Li-ion cell groups and monitor battery packs within each
sensing subsystem on a particular battery-interface control module. Ultimately, however, the critical
data needed for overall battery management is contained in the CAN bus signal interface and a high-
voltage fault signal. At the same time, system safety and reliability depend on safe isolation of the
CAN bus network from the high-voltage sensing circuits. Although isolation can be engineered using
a variety of methods and components, the harsh environment and multiple safety regulations make
optocouplers a preferred solution in this kind of application.

Video interview: Andy Poh, Avago product marketing engineer, discusses the role of
isolation in automotive battery-management systems with Stephen Evanczuk, EDN
contributing editor.

Optocouplers offer high common-mode noise immunity and are essentially immune to the EMC and
EMI associated with electronically noisy environments such as automobiles. In addition, these types
of devices offer thick multilayer insulation, which is vital in the face of long-term dc voltage stress
from the battery pack as well as the fast high-voltage transients that can occur during testing,
charger connection/disconnection, and dc/dc conversion.

In selecting this critical component, key requirements for automotive applications include suitable
package and working-voltage specifications. Although performance specifications such as speed,
data rate, and power consumption remain important, concerns over EMI from fast switching times
and high current transients tend to limit the need for very high-speed devices and instead increase
the requirement for greater flexibility in adjusting slew rate and performance to limit EMI further.
 Video interview: Andy Poh discusses considerations for selection of isolation devices for
use in automotive battery-management systems.

Automotive-grade optocoupler
Avago Technologies’ ACPL-M43T optocoupler provides isolation in the Volt’s battery-interface
control-module PCBs. A member of the Avago R2Coupler family, the M43T is an automotive-grade,
single-channel digital optocoupler in a five-lead SO-5 Jedec package designed for surface mounting.
Along with reinforced insulation, Avago R2Coupler devices such as the M43T use double wire bonds
to reinforce critical functional pads (Figure 7). In addition, the use of hermetically sealed
automotive-grade LEDs has demonstrated extended reliability and a wide temperature range, well
beyond those available with optocouplers based on consumer-grade LEDs. Intended for automotive
applications, the Avago part is manufactured in conformance with the ISO/TS16949 quality system
and is qualified to AEC-Q100 specifications.

Figure 7. For automotive-grade R2Coupler devices such as the ACPL-M43T optocoupler,

Avago reinforces critical functional pads with double wire bonds (shown in the highlighted
area). (Courtesy of Avago Technologies)

The device is well suited to the Chevy Volt battery-pack requirements, with specifications that
include 567V continuous working voltage, 6000V maximum transient overvoltage, 5-mm creepage,
and 5-mm clearance. The device features common-mode transient immunity of 30 kV/µsec for either
logic high or logic low output at 10-mA forward input current, reducing the likelihood that transients
from other automotive subsystems might enter the CAN transmissions network.

The device’s 1M baud speed is more than sufficient for this kind of design. Furthermore, the device
features an open collector output, letting engineers tune the output slew rate to reduce
electromagnetic emissions that could be fallout from fast switching times in downstream
components—including CAN transceivers, despite the relatively low EMI inherent in the CAN
physical-transmission protocol.

In the battery-interface module PCB, the M43T devices sit on the edge of the communications
section, isolating it from the high-voltage-sensing subsystems that are further shielded by ground
planes in deeper PCB layers. The isolation interface provides three individual M43T optocouplers for
each of the three lines brought out from each sensing circuit—specifically, the Freescale S9S08DZ32
CAN Tx output pin, MCU CAN Rx input pin, and high-voltage fault signal from the MCU. Output
from the MCU CAN Tx pin, for example, passes through a shielded signal layer in the PCB to reach
the pin 1 anode of an M43T device to energize the embedded LED, resulting in a change of state at
pin 5Vo (Figure 8). The isolated signal is then passed to the communications output stage of the
battery-interface module.
 Figure 8. Avago ACPL-M43T optocouplers isolate signals between the Freescale
S9S08DZ32 MCU and Infineon CAN transceiver. (Courtesy of Avago Technologies)

Video interview: Andy Poh describes how engineers used the Avago ACPL-M43T
optocoupler in the Chevy Volt battery-management systems.

CAN physical signaling

CAN physical signaling
Sitting at the end of the communications signal chain, the Infineon TLE6250G CAN transceiver is an
AEC-qualified IC that provides CAN physical-layer signaling between the physical cable and CAN
protocol handler—here, the S9S08DZ32 MCU (via optocoupler isolation). Rated for a 1M baud CAN
transmission rate, the device handles the conversion between the CAN_H and CAN_L signals on the
differential signal wire and the CAN dominant and recessive bits transmitted and received by the
S9S08DZ32.

The eight-pin TLE6250G includes Tx, Rx, Vcc, GND, CAN_H, and CAN_L pins, as well as two mode-
control pins: INH and RM. When the TLE6250G senses a change on the Rx pin from the CAN
recessive state to CAN dominant state, the device switches CAN_H high and CAN_L low (Figure 9).
This symmetrical change in state works to reduce EMI because electromagnetic emissions caused by
a rise in CAN_H are balanced by the transition in the opposite direction by CAN_L.

Figure 9. In the CAN physical layer, the symmetrical change in CAN_H and CAN_L helps
reduce EMI. (Courtesy of Infineon Technologies)

The device supports three operating modes: normal, standby and receive-only. When the RM pin is
set low, the device operates in receive-only mode, which can be useful for diagnostics. When the
INH pin is set high, the device enters a low-power standby mode that disables both transmit and
receive functions.

Next-generation systems
The Chevy Volt is certainly one of the most complex distributed embedded system applications to
enter production in the commercial market, and its design pushes the state of the art in multiple
areas. Among the systems most critical to the Volt’s success and that of the EV market at large, the
car’s Li-ion battery and associated battery-management system demonstrate the growing
importance of software and electronics content in automotive applications. According to a recent
McKinsey market study, by 2025, emerging Li-ion technologies stand to boost battery capacity 80%
to 110%, with an attendant price drop that would make the total EV cost of ownership competitive
with that of conventional vehicles powered by internal combustion engines. For engineers, the
challenges remain in exploiting the full potential of emerging Li-ion battery systems in the face of
even higher dc voltages, battery capacities, data rates, and customer expectations.

For a deeper look inside the Chevrolet Volt, including additional teardowns, please visit
www.driveforinnovation, sponsored by Avnet Express.

Related articles

● Teardown reveals Chevy Volt's electronic secrets

● Battery-stack-monitor ICs scrutinize the cells
● GM invests in wireless power technology for Chevy Volt
● Wireless-power technology, development kit target Chevy Volt
● Automotive electronics: What's hot in 2012
● Electric-car development spurs investment in US battery manufacturing
● GM battery packs to be made in the US – but not the cells
● The Tesla Roadster: Sporty and electric