LTC6803-2/LTC6803-4
28
680324fa
APPLICATIONS INFORMATION
DIFFERENCE BETWEEN THE LTC6803-2 AND LTC6803-4
The only difference between the LTC6803-2 and the
LTC6803-4 is the bonding of the V
–
and C0 pins. The
V
–
and C0 are separate signals on every LTC6803 die.
In the LTC6803-2 package, the V
–
and C0 signals are
shorted together by bonding these signals to the same
pin. In the LTC6803-4 package, V
–
and C0 are separate
pins. Therefore, the LTC6803-2 is pin compatible with the
LTC6802-2. For new designs the LTC6803-4 pinout allows
a Kelvin connection to C0 (Figure 22).
CELL VOLTAGE FILTERING
The LTC6803 employs a sampling system to perform its
analog-to-digital conversions and provides a conversion
result that is essentially an average over the 0.5ms con-
version window, provided there isn’t noise aliasing with
respect to the delta-sigma modulator rate of 512kHz. This
indicates that a lowpass filter with 30dB attenuation at
500kHz may be beneficial. Since the delta-sigma integra-
tion bandwidth is about 1kHz, the filter corner need not
be lower than this to assure accurate conversions.
Series resistors of 100Ω may be inserted in the input
paths without introducing meaningful measurement er-
ror. Shunt capacitors may be added from the cell inputs
to V
–
, creating RC filtering as shown in Figure 8. The cell
balancing MOSFET in Figure 11 can cause a small transient
when it switches on and off. Keeping the cutoff frequency
of the RC filter relatively high will allow adequate settling
prior to the actual conversion. A delay of about 500µs is
provided in the ADC timing, so a 16kHz LPF is optimal
(100Ω, 0.1µF) and offers about 30dB of noise rejection.
Larger series resistors and shunt capacitors can be used
to lower the filter bandwidth. The measurement error due
to the larger component values is a complex function of
the component values. The error also depends on how
often measurements are made. Table 17 is an example. In
each example a 3.6V cell is being measured and the error
is displayed in millivolts. There is a RC filter in series with
inputs C1 through C12. There is no filter in series with
C0. There is an interaction between cells. This is why the
errors for C1 and C12 differ from C2 through C11.
Table 17. Cell Measurement Errors vs Input RC Values
R = 100Ω,
C = 0.1µF
R = 1k,
C = 0.1µF
R = 1k,
C = 1µF
R = 10k,
C = 3.3µF
Cell 1 Error
(mV, LTC6803-2)
0.1 4.5 1.5 1.5
Cell 2 to Cell 12 (mV) 1 9 3 0.5
For the LTC6803-2, no resistor should be placed in series
with the V
–
pin. Because the supply current flows from
the V
–
pin, any resistance on this pin could generate a
significant conversion error for cell 1, and the error of
cell 1 caused by the RC filter differs from errors of cell 2
to cell 2.
OPEN-CONNECTION DETECTION
When a cell input (C pin) is open, it affects two cell mea-
surements. Figure 9 shows an open connection to C3,
in an application without external filtering between the C
pins and the cells. During normal ADC conversions (that
is, using the STCVAD command), the LTC6803 will give
near zero readings for B3 and B4 when C3 is open. The
zero reading for B3 occurs because during the measure-
ment of B3, the ADC input resistance will pull C3 to the
C2 potential. Similarly, during the measurement of B4, the
ADC input resistance pulls C3 to the C4 potential.
Figure 10 shows an open connection at the same point in
the cell stack as Figure 9, but this time there is an external
filtering network still connected to C3. Depending on the
value of the capacitor remaining on C3, a normal measure-
ment of B3 and B4 may not give near-zero readings, since
the C3 pin is not truly open. In fact, with a large external
capacitance on C3, the C3 voltage will be charged midway
Figure 8. Adding RC Filtering to the Cell Inputs
(One Cell Connection Shown)
+
100nF
100nF
680324 F08
7.5V
Cn
C(n – 1)
100Ω
100Ω
