Seite 20 - Boards & Solution/ECE Magazine July 2013

July 2013 20
EMBEDDED DESIGN
stronger cell. When discharging the pack,
charge can be moved from the stronger cell to
compensate the weaker cell. Instead of wasting
energy, charge is transferred through a highly
efficient circuit, such as a fly-back converter.
As a result, heat generation is limited, the bal-
ancing current is higher, and the balancing
time is significantly reduced. This allows for
active balancing while the pack is in use, where
it can ensure extraction of the maximum ca-
pacity for each individual cell. New ICs, such
as Linear Technology devices LTC3300 and
LT8584, are enabling active balancing in auto-
motive battery packs.
Ideally, active balancing should be enabled as
the cells reach the ends of the SOC range. For
maximum efficiency, active balancing should
be used when necessary. A system maintained
well within the SOC limits would require
much less active balancing than one operating
near the limits. To illustrate, consider a pack
containing many cells with uniform capacity,
and one weak lower capacity cell. If all cells
are charged to 80% SOC and then discharged,
the SOC of the weak cell will slowly diverge
from the rest of the cells. The BMS must de-
termine a suitable point to enable the balancer
to keep the weak cell operational, while the
other cells are continuing to discharge. Figure
4 shows the SOC divergence during the dis-
charge cycle, for two examples: one with a cell
that has a capacity difference of 2% from the
rest of the pack, and another that differs by
8%. The BMS cell measurement error sets a
limit for determining the relative condition
between cells. With a ±2% SOC measurement
error (±10mV), cells could be out-of-balance
from each other by as much as 4% before the
cell measurements would reliably detect the
situation. Enabling an active balancer at a well-
defined point along this discharge curve would
be virtually impossible without cell measure-
ment accuracy very much better than ±10mV.
The implication for measurement accuracy is
not limited to active balancing. From this ex-
ample, a 4% SOC difference translates to a ca-
pacity variation of more than 6.6% (0.066 =
(0.04/0.60) for an SOC range of 22% to 78%
(guard-banded range)). For an automotive bat-
tery that reaches its end-of-life after 20% re-
duction, this is significant unrecovered capacity.
More importantly, a change in cell capacity is a
key indicator of its health, and an unrecognized
capacity change could be a serious issue.
The importance of cell measurement accuracy
becomes clearer when considering the compli-
cations beyond this simple example. For ex-
ample, most packs will have a continuum of
capacity variation, with more subtle and harder
to detect SOC divergences. Also, cells will not
likely be aligned at 80% SOC when discharging
begins, possibly further masking capacity vari-
ation. It is also important to note that multiple
parameters are required for an SOC calculation.
Measurement inaccuracy of these other pa-
rameters does not lessen the need for accurate
cell voltage measurements. To the contrary,
compromising on cell voltage accuracy will
likely widen the distribution of battery life.
The voltage reference within the battery mon-
itor is the primary determinant of measure-
ment error. Any change in the voltage reference
directly degrades the cell measurement accu-
racy. The current generation of battery moni-
tors relies on bandgap voltage references. In
theory, bandgap references are ideal for inte-
gration into complex integrated circuits, such
as a battery stack monitor, because they require
little die space, low power and low overhead
voltage. However, band-gap references are sen-
sitive to mechanical stress, IR reflow, and hu-
midity, resulting in thermal hysteresis and
long term drift. For precision instrumentation
that is required to maintain very high accuracy
for more than 15 years, there is a better choice.
The newest battery monitors, such as Linear
Technology LTC6804, incorporate a sub-surface
Zener voltage reference. Sub-surface Zener
voltage references provide outstanding long-
term stability and accuracy, over time and op-
erating conditions. Using this approach, the
LTC6804 is able to guarantee a total cell voltage
measurement error of less than 1.2mV.
The accuracy of the battery monitor is not lim-
ited to the accuracy of the measurement itself.
Cell measurements need to be considered in
the context of the automobile, where there is
significant electrical noise and transients from
inverters, actuators, switches, relays, etc. This
noise is imbedded within the cell signal, and if
accuracy is important, this noise has to be re-
moved. A modest level of noise reduction can
be achieved by placing an RC filter on each
cell; using a higher order filter circuit on each
cell is impractical, due to cost and board space.
A modest amount of noise can be removed by
processing many samples from each signal
measurement; but given the large number of
cells, the massive data to be transmitted to a
central processor makes this impractical, as
well. A practical and effective solution is to re-
move noise within the battery monitor. As an
example, Linear Technology LTC6804 uses
delta-sigma ADCs with built-in third order
noise filtering. This is in contrast to a wideband
SAR ADC, where fast acquisition has limited
value for a signal corrupted by noise (SAR con-
verters can implement simple on-chip averaging,
but averaging has poor filter characteristics).
To optimize speed and noise reduction, the
LTC6804 delta sigma ADCs can operate with
different corner frequencies, ranging from 27
kHz to 26Hz. For the automotive environment,
the delta sigma approach is quite effective.
As high-powered battery systems continue to
advance into the mainstream, the demands
on the battery monitoring electronics will not
relent. The automobile offers only abuse, while
demanding the highest possible performance
and reliability. To achieve the driving range,
reliability, and safety, requires careful consid-
eration of every small source of lost perform-
ance. To extract every bit of usable energy re-
quires implementing cutting edge technologies,
such as active balancing. It also requires the
most accurate and stable cell voltage measure-
ment possible. n
Figure 4. Detecting cell capacity differences depends on measurement accuracy.