
29
LTC2480
2480f
Reference Current
In a similar fashion, the LTC2480 samples the differential
reference pins V
REF+
and GND transferring small amount
of charge to and from the external driving circuits thus
producing a dynamic reference current. This current does
not change the converter offset, but it may degrade the
gain and INL performance. The effect of this current can be
analyzed in two distinct situations.
For relatively small values of the external reference capaci-
tors (C
REF
< 1nF), the voltage on the sampling capacitor
settles almost completely and relatively large values for
the source impedance result in only small errors. Such
values for C
REF
will deteriorate the converter offset and
gain performance without significant benefits of reference
filtering and the user is advised to avoid them.
Larger values of reference capacitors (C
REF
> 1nF) may be
required as reference filters in certain configurations.
Such capacitors will average the reference sampling charge
and the external source resistance will see a quasi con-
stant reference differential impedance.
In the following discussion, it is assumed the input and
reference common mode are the same. Using internal
oscillator for 60Hz mode, the typical differential refer-
ence resistance is 1M
which generates a full-scale
(V
REF
/2) gain error of 0.51ppm for each ohm of source
resistance driving the V
REF
pin. For 50Hz/60Hz mode, the
related difference resistance is 1.1M
and the resulting
full-scale error is 0.46ppm for each ohm of source
resistance driving the V
REF
pin. For 50Hz mode, the
related difference resistance is 1.2M
and the resulting
full-scale error is 0.42ppm for each ohm of source
resistance driving the V
REF
pin. When F
O
is driven by an
external oscillator with a frequency f
EOSC
(external con-
version clock operation), the typical differential reference
resistance is 0.30 10
12
/f
EOSC
and each ohm of source
resistance driving the V
REF
pin will result in 1.67 10
–6
f
EOSC
ppm gain error. The typical +FS and –FS errors for
various combinations of source resistance seen by the
V
REF
pin and external capacitance connected to that pin
are shown in Figures 15-18.
In addition to this gain error, the converter INL perfor-
mance is degraded by the reference source impedance.
The INL is caused by the input dependent terms
–V
IN2
/(V
REF
R
EQ
) – (0.5 V
REF
D
T
)/R
EQ
in the reference
pin current as expressed in Figure 11. When using internal
oscillator and 60Hz mode, every 100
of reference source
resistance translates into about 0.67ppm additional INL
error. When using internal oscillator and 50Hz/60Hz mode,
every 100
of reference source resistance translates into
about 0.61ppm additional INL error. When using internal
oscillator and 50Hz mode, every 100
of reference source
resistance translates into about 0.56ppm additional INL
error. When F
O
is driven by an external oscillator with a
frequency f
EOSC
, every 100
of source resistance driving
V
REF
translates into about 2.18 10
–6
f
EOSC
ppm addi-
tional INL error. Figure 19 shows the typical INL error due
to the source resistance driving the V
REF
pin when large
C
REF
values are used. The user is advised to minimize the
source impedance driving the V
REF
pin.
In applications where the reference and input common
mode voltages are different, extra errors are introduced.
For every 1V of the reference and input common mode
voltage difference (V
REFCM
– V
INCM
) and a 5V reference,
each Ohm of reference source resistance introduces an
extra (V
REFCM
– V
INCM
)/(V
REF
R
EQ
) full-scale gain error,
which is 0.074ppm when using internal oscillator and
60Hz mode. When using internal oscillator and 50Hz/60Hz
mode, the extra full-scale gain error is 0.067ppm. When
using internal oscillator and 50Hz mode, the extra gain
error is 0.061ppm. If an external clock is used, the corre-
sponding extra gain error is 0.24 10
–6
f
EOSC
ppm.
The magnitude of the dynamic reference current depends
upon the size of the very stable internal sampling capaci-
tors and upon the accuracy of the converter sampling
clock. The accuracy of the internal clock over the entire
temperature and power supply range is typically better
than 0.5%. Such a specification can also be easily achieved
by an external clock. When relatively stable resistors
(50ppm/
°
C) are used for the external source impedance
seen by V
REF+
and GND, the expected drift of the dynamic
current gain error will be insignificant (about 1% of its
value over the entire temperature and voltage range). Even
for the most stringent applications a one-time calibration
operation may be sufficient.
In addition to the reference sampling charge, the reference
pins ESD protection diodes have a temperature dependent
APPLICATIU
W
U
U