dac16fs Analog Devices, Inc., dac16fs Datasheet - Page 8

• Manufacturer Part Number: dac16fs
• Description: 16-bit High Speed Current-output Dac
• Manufacturer: Analog Devices, Inc.

DAC16
reference amplifier. Completely removing the compensation
network would introduce large linearity errors, reference amplifier
instability, wideband reference amplifier noise, and poor settling
time.
Because the DAC exhibits an internal current scaling factor of
eight times (8 ), the reference amplifier requires only 500 A
input current from the user-supplied precision reference for a
4 mA full-scale output current. In applications that do not re-
quire such high output currents, good accuracy can be achieved
with input reference currents in the range of 350 A
achieved with a 625 A reference current which yields a maxi-
mum 5 mA output current. Figure 22 illustrates how to form
the reference input current with a REF02 and a 10 k precision
resistor.
Reducing Voltage Reference Noise
In data converters of 16-bit and greater resolution, noise is of
critical importance. Surprisingly, the integrated voltage refer-
ence circuit used may contribute the dominant share of a
system’s noise floor, thereby degrading system dynamic range
and signal-to-noise ratio. To maximize system dynamic range
and SNR, all external noise contributions should be effectively
much less than 1/2 LSB. For example, in a 5 V DAC16 applica-
tion, one LSB is equivalent to 76 V. This means that the total
wideband noise contribution due to a voltage reference and all
other sources should be less than 38 V rms. These noise levels
are not easy targets to hit with standard off-the-shelf reference
devices. For example, commercially available references might
exhibit 5 V rms noise from 0.1 Hz to 10 Hz: but, over a 100 kHz
bandwidth, its 300 V rms of noise can easily swamp out a
16-bit system. Such noisy behavior can degrade a DAC’s effec-
tive resolution by increasing its differential nonlinearity which,
in turn, can lead to nonmonotonic behavior or analog errors.
The easiest way to reduce noise in the reference circuit is to
band-limit its noise before feeding it to the converter. In the
case of the DAC16, the reference is not a voltage, but a current.
Illustrated in Figure 23 is a simple way of hand-limiting
voltage reference noise by splitting R
and bypassing the common node with a capacitor. To minimize
thermally induced errors, R1 and R2 must be electrically and
Figure 22. Generating the DAC16’s Reference Input Current
625 A. The best signal-to-noise ratios, of course, will be
Figure 23. Filtering a Reference’s Wideband Noise
0.1 F
0.1 F
REF02
+15V
REF02
+15V
10k
R
REF
22 F
5k
R1
I
REF
C1
=
I
REF GND
REF
V
R
5k
R2
REF
REF
REF
DAC16
into two equal resistors
REF GND
AGND
I
REF
DAC16
I
I
REF
OUT
–8–
thermally well-matched. Thin-film resistor networks work well
here. In this circuit, the parallel combination of R1 and R2
forms a 3 Hz low-pass filter with C1. The only noise source that
remains is the thermal noise of R2 which can be a significantly
lower noise generator than the voltage reference.
Input Coding
The unipolar digital input coding of the DAC16 employs nega-
tive logic to control the output current; that is, an all zero input
code (0000
Conversely, an all 1s input code (FFFF
current output. An expression for the DAC16’s transfer equa-
tion can be expressed by:
Table II provides the relationship between the digital input
codes and the output current of the DAC16.
Digital Input
Word (Hex)
0000
7FFE
7FFF
8000
FFFF
Since the DAC16 exhibits a small output voltage compliance on
the order of a few millivolts, a high accuracy operational ampli-
fier must be used to convert the DAC’s output current to a volt-
age. Refer to the section on selecting operation amplifiers for the
DAC16. The circuit shown in Figure 24 illustrates a unipolar
output configuration. In symbolic form, the transfer equation
for this circuit can be expressed by:
In this example, the reference input current was set to 500 A
which produces a full-scale output current of 4 mA – 1 LSB.
The DAC’s output current was scaled by R3, a 1.25 k resistor,
to produce a 5 V full-scale output voltage. Bear in mind that to
ensure the highest possible accuracy, matched thin-film resistor
networks are almost a necessity, not an option. The resistors
used in the circuit must have close tolerance and tight thermal
tracking. Table III illustrates the relationship between the input
digital code and the circuit’s output voltage for the component
values shown.
Digital Input Word
(Hex)
0000
7FFE
7FFF
8000
FFFF
Table III. Unipolar Output Voltage vs. Digital Input Code
V
I
O
OUT
H
) yields an output current 1 LSB below full scale.
R3 8 I
Table II. Unipolar Code Table
DAC16 Output
Current I
8
8 (2
8 (2
8 (2
0
8 I
(2
REF
16
15
15
15
Decimal Number in
in DAC Decoder
65,535
32,769
32,768
32,767
0
/2
– 1)/2
+ 1)/2
REF
– 1)/2
16
OUT
)
65,535 – Digital Code
65,535 – Digital Code
16
16
16
I
REF
I
I
REF
I
REF
REF
65,536
65,536
H
) yields a zero analog
Comment
Full Scale
Midscale + 1 LSB
Midscale
Midscale – 1 LSB
Zero Scale
Analog Output
Voltage (V)
4.999924
2.500076
2.500000
2.499924
0
REV. B