LTC2758_15 Datasheet, PDF (19/24 Pages) Linear Technology – Dual Serial 18-Bit SoftSpan IOUT DACs

LTC2758
APPLICATIONS INFORMATION
Op amp offset contributes mostly to DAC output offset
and gain error, and has minimal effect on INL and DNL.
For example, consider the LTC2758 in unipolar 5V output
range. (Note that for this example, the LSB size is 19µV.)
An op amp offset of 35µV will cause 1.8LSB of output
offset, and 1.8LSB of gain error; but 0.4LSB of INL, and
just 0.1LSB of DNL.
While not directly addressed by the simple equations in
Tables 4 and 5, temperature effects can be handled just as
easily for unipolar and bipolar applications. First, consult
an op amp’s data sheet to find the worst-case V_OS and I_B
over temperature. Then, plug these numbers in the V_OS
and I_B equations from Table 5 and calculate the tempera-
ture-induced effects.
For applications where fast settling time is important, Ap-
plication Note 120, 1ppm Settling Time Measurement for
a Monolithic 18-Bit DAC, offers a thorough discussion of
18-bit DAC settling time and op amp selection.
Recommendations
For DC or low-frequency applications, the LTC1150 is the
simplest 18-bit accurate output amplifier. An auto-zero
amp, its exceptionally low offset (10µV max) and offset
drift (0.01µV/°C) make nulling unnecessary. For swings
above 8V, use an LT1010 buffer to boost the load current
capability. The settling of auto-zero amps is a special case;
see Application Note 120, 1ppm Settling Time Measurement
for a Monolithic 18-Bit DAC, Appendix E, for details.
The LT1012 and LT1001 are good intermediate output-amp
solutions that achieve moderate speed and good accuracy.
They are also excellent choices for the reference inverting
amplifier in fixed-reference applications.
For high speed applications, the LTC1468 settles in 2.1µs.
Note that the 75µV max offset will degrade the INL at the
DAC output by up to 0.9LSB. For high-speed applications
demanding higher precision, the amplifier offset can be
nulled with a digital potentiometer.
The Typical Application on the last page shows a composite
output amplifier that achieves fast settling (8µs) and very
low offset (3µV max) without offset nulling. This circuit
offers high open-loop gain (1000V/mV min), low input bias
current (0.15nA max), fast slew rate (25V/µs min), and a
high gain-bandwidth product (30MHz typ). The high speed
path consists of an LTC6240, which is an 18MHz ultralow
bias current amplifier, followed by an LT1360, a 50MHz
fast-slewing amplifier which provides additional gain and
the ability to swing to ±10V at the output. Compensation is
taken from the output of the LTC6240, allowing the use of
a much larger compensation capacitor than if taken after
the gain-of-five stage. An LTC2054 auto-zero amplifier
senses the voltage at I_OUT1 and drives the non-inverting
input of the LTC6240 to eliminate the offset of the high
speed path. The 100:1 attenuator and input filter reduce
the low frequency noise in this stage while maintaining
low DC offset.
Precision Voltage Reference Considerations
Much in the same way selecting an operational amplifier
for use with the LTC2758 is critical to the performance of
the system, selecting a precision voltage reference also
requires due diligence. The output voltage of the LTC2758
is directly affected by the voltage reference; thus, any
voltage reference error will appear as a DAC output volt-
age error.
There are three primary error sources to consider
when selecting a precision voltage reference for 18-bit
applications: output voltage initial tolerance, output voltage
temperature coefficient and output voltage noise.
Initial reference output voltage tolerance, if uncorrected,
generates a full-scale error term. Choosing a reference
with low output voltage initial tolerance, like the LT1236
(±0.05%), minimizes the gain error caused by the reference;
however, a calibration sequence that corrects for system
zero- and full-scale error is always recommended.
A reference’s output voltage temperature coefficient affects
not only the full-scale error, but can also affect the circuit’s
INL and DNL performance. If a reference is chosen with
a loose output voltage temperature coefficient, then the
DAC output voltage along its transfer characteristic will
be very dependent on ambient conditions. Minimizing
the error due to reference temperature coefficient can be
achieved by choosing a precision reference with a low
output voltage temperature coefficient and/or tightly con-
trolling the ambient temperature of the circuit to minimize
temperature gradients.
2758fa
For more information www.linear.com/LTC2758
19