Feb 1999 16-Bit Parallel DAC Has 1LSB Linearity, Ultralow Glitch and Accurate 4-Quadrant Resistors

DESIGN FEATURES
16-Bit Parallel DAC Has 1LSB Linearity,
Ultralow Glitch and Accurate
by Patrick Copley
4-Quadrant Resistors
Today’s fast paced marketplace has
developed a major appetite for high
resolution, high accuracy, fast digital-to-analog converters. System
requirements in instrumentation,
automatic test equipment, communications, waveform generation, data
acquisition and feedback control systems, among many other applications,
have fueled the need for 16-bit digitalto-analog converters. Not only does
the converter need to meet the stringent speed and accuracy requirements
of the system, it needs to do so in both
unipolar (0V to 10V) and bipolar (±10V)
modes of operation without degradation. To meet and exceed these
requirements, Linear Technology
introduces its LTC1597 16-bit parallel, current output, low glitch,
multiplying DAC with 4-quadrant
resistors. Key features of the new
DAC include:
❏ ±1LSB maximum INL and DNL
over the industrial temperature
range
❏ On-chip 4-quadrant resistors
allow precise 0V to 10V, 0V to
–10V or ±10V outputs
❏ Ultralow, < 1nV-s midscale glitch
impulse
❏ Small 28-pin SSOP package
❏ Low supply power consumption:
10µW typical
❏ Pin-compatible with the LTC1591
14-bit parallel, current output,
low glitch, multiplying DAC with
4-quadrant resistors.
Unique Features
of the LTC1597
The LTC1597 operates from a single
5V supply and provides both unipolar
48k
REF
0V to –10V or 0V to 10V and bipolar
±10V output ranges from a 10V or
–10V reference input using a single or
dual external op amp. The device
achieves bipolar operation using three
additional on-chip precision resistors.
The DAC consists of a precision thinfilm R/2R ladder for the thirteen LSBs.
The three MSBs are decoded into
seven segments of resistor value R, as
shown in Figure 1. R is nominally
48k. Each of these segments and the
R/2R ladder carry an equally weighted
current of one-eight of full-scale. The
feedback resistor, RFB, and 4-quadrant resistor, ROFS, have a value of R/
4. 4-quadrant resistors R1 and R2
have a magnitude of R/4.
The reference pin presents a constant
input impedance of R/8 in unipolar
mode and R/12 in bipolar mode. The
output impedance of the current output pin, IOUT1, varies with DAC code.
48k
1
R2
12k
RCOM 2
48k
48k
48k
48k
48k
48k
48k
96k
96k
96k
96k
ROFS
12k
RFB
12k
4 ROFS
5 RFB
R1
12k
R1 3
6 IOUT1
VCC 23
7 AGND
DECODER
LD 8
WR 9
LOAD
22 DGND
D15
(MSB)
D14
D12
D13
D11
•••
DAC REGISTER
D0
(LSB) RST
INPUT REGISTER
WR
28 CLR
RST
1597 BD
10
11
D15
D14
••••
21
24
25
26
27
D4
D3
D2
D1
D0
Figure 1. The LTC1597 16-bit CMOS DAC uses a precision thin-film modified R/2R architecture to provide unsurpassed accuracy and stability.
Accurate 4-quadrant multiplication applications are now possible with on-chip resistors R1, R2 and ROFS. A built-in deglitcher reduces glitch
impulse to 1nV-s.
18
Linear Technology Magazine • February 1999
DESIGN FEATURES
1.0
16-Bit Accuracy
Over Temperature
INTEGRAL NONLINEARITY (LSB)
0.8
0.6
0.4
0.2
0
– 0.2
– 0.4
– 0.6
– 0.8
–1.0
0
49152
32768
16384
DIGITAL INPUT CODE
2a.
65535
1597 G01
DIFFERENTIAL NONLINEARITY (LSB)
1.0
0.8
0.6
0.4
0.2
0
– 0.2
– 0.4
– 0.6
– 0.8
–1.0
0
49152
32768
16384
DIGITAL INPUT CODE
65535
1597 G02
2b.
Figure 2. The outstanding INL and DNL
(typically less than 0.25LSB) and very low
linearity drift allow a maximum 1LSB spec to
be guaranteed over the industrial temperature range.
An added feature of the LTC1597 is
a proprietary deglitcher that reduces
the glitch energy to below 1nV-s over
the DAC’s output voltage range.
The LTC1597 has a 16-bit parallel
input data bus and is double buffered
with two 16-bit registers. The double
buffered feature permits the updating of several DACs simultaneously.
The WR signal updates the input register and the LD signal loads the DAC
register. The deglitcher is activated
on the rising edge of the LD signal.
The versatility of the interface also
allows the use of the input and DAC
registers in a master/slave or edgetriggered configuration. This mode of
operation occurs when WR and LD
are tied together to act as a clock
signal.
The asynchronous clear pin (CLR)
resets the LTC1597 to zero scale and
the LTC1597-1 to midscale. CLR resets both the input and DAC registers.
The LTC1597 also features a poweron reset.
Linear Technology Magazine • February 1999
The LTC1597 has ultralow linearity
drift of well below ±0.2LSB from
–45°C to 85°C. This allows the
LTC1597 to hold its accuracy of 1LSB
integral nonlinearity (INL) and differential nonlinearity (DNL) over time
and temperature. In the past, the
only DACs that approached this
accuracy over temperature were of
the autocalibrated type. These DACs
were very large, very expensive and
therefore not very practical for most
applications.
Figures 2a and 2b show the typical
INL and DNL curves of the LTC1597.
The outstanding 0.25LSB INL, 0.15
LSB DNL (typical) and very low drift
allow a maximum 1LSB specification
over the extended industrial temperature range. For optimum performance,
the REF pin of the LTC1597 should be
driven by a source impedance of less
than 1kΩ. However, the DAC has been
designed to minimize source
impedance effects. An 8kΩ source
impedance degrades both INL and
DNL by a mere 0.2LSB.
5V
LD PULSE
5V/DIV
GATED
SETTLING
WAVEFORM
500µV/DIV
500ns/DIV
Figure 4. When used with the LT1468 and a
20pF feedback capacitor (see Figure 3), the
LTC1597 can settle in an amazing 1.7µs to
within 0.0015%. The top trace shows the LD
pulse; the bottom trace shows the gated
settling waveform settling to 1LSB in 1.7µs.
Fast Settling:
Less than 2µs to within
0.0015% of Full-Scale
Now system designers no longer have
to make tough decisions in the tradeoff between accuracy and speed. The
solution is here. The combination of
the LTC1597 DAC and the LT1468 op
amp provides an industry first: superb
16-bit settling of less than 2µs for a
10V step while maintaining 1LSB DC
accuracy.
Figure 3 shows the application circuit for unipolar mode. Figure 4 shows
the resulting full-scale 10V step settling time of the LTC1597/LT1468
combination. With a 20pF feedback
capacitor, the optimized settling time
to 0.0015% is an amazing ≈1.7µs. A
0.1µF
VREF
2
3
R1
1
REF
RCOM
R1
23
VCC
4
5
ROFS
RFB
ROFS
R2
20pF
RFB
IOUT1
16
DATA
INPUTS
LTC1597
–
6
16-BIT DAC
AGND
10 TO 21,
24 TO 27
DGND
7
+
LT1468
VOUT =
0V TO
–VREF
22
WR LD CLR
WR
LD
CLR
9
8
28
Unipolar Binary Code Table
DIGITAL INPUT
BINARY NUMBER
IN DAC REGISTER
LSB
MSB
1111
1000
0000
0000
ANALOG OUTPUT
VOUT
1111
0000
0000
0000
1111
0000
0000
0000
1111
0000
0001
0000
–VREF (65,535/65,536)
–VREF (32,768/65,536) = –VREF/2
–VREF (1/65,536)
0V
1591/97 F01b
Figure 3. With a single external op amp, the LTC1597 performs 2-quadrant multiplication with
±10V input and 0V to –VREF output. With a fixed –10V reference, it provides a precision 0V to
10V unipolar output.
19
DESIGN FEATURES
– 50
– 60
– 70
– 80
500kHz FILTER
– 90
80kHz FILTER
–100
30kHz FILTER
–110
10
100
1k
10k
FREQUENCY (Hz)
– 50
– 40
VCC = 5V USING TWO LT1468s
CFEEDBACK = 15pF
REFERENCE = 6VRMS
– 60
– 70
– 80
500kHz FILTER
– 90
–100
80kHz FILTER
10
100
– 60
– 70
– 80
500kHz FILTER
– 90
80kHz FILTER
–100
–110
1k
10k
FREQUENCY (Hz)
1591/97 G03
VCC = 5V USING TWO LT1468s
CFEEDBACK = 15pF
REFERENCE = 6VRMS
– 50
30kHz FILTER
30kHz
FILTER
–110
100k
6a. Unipolar-mode full-scale: the noise and
distortion (N + D) is less than –96dB for signal
frequencies up to 30kHz. Out to 100kHz, the
N + D is less than –78dB.
SIGNAL/(NOISE + DISTORTION) (dB)
– 40
VCC = 5V USING AN LT1468
CFEEDBACK = 30pF
REFERENCE = 6VRMS
SIGNAL/(NOISE + DISTORTION) (dB)
SIGNAL/(NOISE + DISTORTION) (dB)
– 40
10
100k
100
1k
10k
FREQUENCY (Hz)
100k
1591/97 G05
1591/97 G04
6b. Bipolar-mode zero-scale: the N + D is less
than –96dB for signal frequencies up to
30kHz. Out to 100kHz, the N + D is less than
–82dB.
6c. Bipolar-mode full-scale: the (N + D) is less
than –96dB for signal frequencies up to
30kHz. Out to 100kHz, the N + D is less than
–79dB.
Figure 6. LTC1597 multiplying-mode signal-to-noise vs frequency
detailed discussion of 16-bit settling
time can be found in Linear Technology Application Note 74, “Component
and Measurement Advances Ensure
16-Bit DAC Settling Time.”
The ability to minimize settling time
is limited by the need to null the DAC
output capacitance, which varies from
70pF to 115pF, depending on code.
This capacitance at the amplifier input
combines with the feedback resistor
to form a zero in the closed-loop frequency response in the vicinity of
200kHz–400kHz. Without a feedback
capacitor, the circuit will oscillate.
The choice of 20pF stabilizes the circuit by adding a pole at 1.3MHz to
limit the frequency peaking and also
optimizes settling time. The settling
time to 16-bit accuracy is theoretically bounded by 11.1 time constants
set by the feedback resistance and
capacitance.
OUTPUT VOLTAGE (mV)
+10
Glitches in a DAC’s output when it
updates can be a big problem in precision applications. Usually, the
worst-case glitch occurs when the
DAC output crosses midscale. The
LTC1597’s new proprietary deglitcher
reduces the output glitch impulse to
1nV-s, which is at least ten times
lower than any of the competition’s
16-bit voltage output DACs. In addition, the deglitcher makes the glitch
impulse uniform for any code. Figure
5 shows the output glitch for a midscale transition with a 0V to 10V
output range.
Unipolar 0V to 10V Outputs
with a Single Op Amp
Figure 3 shows the circuit for a 0V to
10V output range. The DAC uses an
external reference and a single op
amp in this configuration. This circuit can also perform 2-quadrant
multiplication where the REF pin is
driven by a ±10V AC input signal and
VOUT swings from 0V to –VREF.
VREF
+
5V
0.1µF
1/2 LT1112
–
2
3
R1
1
REF
RCOM
R1
5
23 4
VCC ROFS
ROFS
R2
RFB
33pF
RFB
IOUT1
16
DATA
INPUTS
LTC1597-1
–
6
1/2 LT1112
16-BIT DAC
AGND
10 TO 21,
24 TO 27
COMPETITOR’S DAC
DGND
7
+
VOUT =
–VREF
TO VREF
22
WR LD CLR
WR
LD
CLR
0
1nV-s TYP LTC1597
–10
9
8
28
Bipolar Offset Binary Code Table
DIGITAL INPUT
BINARY NUMBER
IN DAC REGISTER
1
2
TIME (µs)
3
4
03 .eps
Figure 5. The proprietary deglitcher 1595
reduces
the output glitch to less than 1nV-s, which is
ten times less than any other 16-bit, voltageoutput DAC. Further, the deglitcher makes
the glitch uniform, independent of code.
1111
1000
1000
0111
0000
ANALOG OUTPUT
VOUT
LSB
MSB
0
20
Ultralow 1nV-s Glitch
1111
0000
0000
1111
0000
1111
0000
0000
1111
0000
1111
0001
0000
1111
0000
VREF (32,767/32,768)
VREF (1/32,768)
0V
–VREF (1/32,768)
–VREF
1591/97 F02b
Figure 7. With a dual op amp, the LTC1597 performs 4-quadrant multiplication. With a fixed
10V reference, it provides a ±10V bipolar output. For fast bipolar settling applications, an
LT1468 can be used for the output amplifier.
Linear Technology Magazine • February 1999
DESIGN FEATURES
Bipolar ±10V Output
with Two Op Amps
0.92LSB (140µV) at 17 bits for room
temperature. The circuit uses the
LTC1597 in its unipolar mode with
the reference input inverted (–VREF,
by means of R1 and R2 and an external op amp) for the output voltage
range 0V to VREF. When the sign bit
changes, the analog switch changes
the reference input polarity to noninverting (VREF) for the output range 0V
to –VREF.
modes of operation. For AC signals less
than 40kHz, the THD+noise is superb
(better than 90dB) and is still very good
out to 100kHz (78dB). Filtering at the
output of the LT1468 is necessary to
reduce the noise bandwidth to acceptable levels. The wider the bandwidth,
the higher the noise floor.
The LTC1597 contains all the 4-quadrant resistors necessary for bipolar
operation. For a fixed 10V reference,
the circuit shown in Figure 7 gives a
precision –10V to 10V output swing,
with a minimum of external components: a feedback capacitor and a
dual op amp. The bipolar zero error is
8LSB maximum over temperature. If
two LT1468 op amps are used instead
of the LT1112, the circuit can perform wider bandwidth 4-quadrant
multiplication, where the reference
input is driven by a ±10V AC input
signal and VOUT swings ±10V .
Figure 6 shows a graph of the multiplying mode total harmonic distortion and
noise of the LTC1597/LT1468 combination in both unipolar and bipolar
17-Bit Sign Magnitude DAC
Gives Perfect Bipolar Zero
Figure 8 shows a novel application of
the LTC1597, a 17-bit sign magnitude DAC, and the resulting output
coding. This circuit has an extremely
accurate bipolar zero error, which is
the offset voltage of the current-tovoltage op amp plus the bias current
times the DAC feedback resistor. For
the LT1468, this corresponds to a
maximum bipolar zero error of
16
94dB SFDR
Digital Sine Wave Generator
Figure 9 shows the circuit diagram
for a variable frequency digital waveform generator. The circuit shows the
bipolar configuration for the LTC1597
but the unipolar configuration will
work just as well. For a sampling
frequency of 50kHz and an output
sine wave frequency of 1kHz, the second harmonic distortion is –94dB and
the third harmonic is –101dB. The
on-chip deglitcher circuit minimizes
the code-dependent glitch (which
14
15
Bipolar Sign Magnitude Code Table
LTC203AC
DIGITAL INPUT
BINARY NUMBER
IN DAC REGISTER
LSB
SIGN MSB
1
2
15V
LTC1236A-10
2
1111
0000
0000
0000
0000
1111
1
1
1
0
0
0
3
6 VREF
ANALOG OUTPUT
VOUT
1111
0000
0000
0000
0000
1111
1111
0000
0000
0000
0000
1111
1111
0001
0000
0000
0001
1111
VREF (65,535/65,536)
VREF (1/65,536)
0V
0V
–VREF (1/65,536)
–VREF (65,535/65,536)
+
4
5V
LT1468
–
0.1µF
15pF
3
R1
2
RCOM
1
REF
5
RFB
23 4
ROFS
SIGN BIT
20pF
R2
R1
ROFS
RFB
LTC1597
10 T0 21
24 TO 27
–
6
I0UT
16 DATA
INPUTS
16-BIT DAC
AGND
7
+
LT1468
VOUT
22
DGND
WR LD CLR
9
8
28
WR
LD
CLR
Figure 8. This 17-bit sign-magnitude DAC uses the LTC1597 in its unipolar mode with the reference bit inverted (–VREF) for the output range 0V
to VREF. When the sign bit changes, the analog switch changes the reference input polarity to noninverting (VREF) for the output range 0V to
–VREF. The resulting circuit produces an impressive bipolar zero error of 140µV (0.92LSB) max at room temperature—less than 1LSB at 17 bits.
Linear Technology Magazine • February 1999
21
DESIGN FEATURES
2
LTC1236A-10
15V
10V
6
+
4
5V
LT1001
0.1µF
–
3
R1
n = 24–32 BITS
SERIAL OR
PARALLEL
DATA INPUT
n
1
REF
5
RFB
23 4
ROFS
15pF
n
SERIAL
OR BYTE
LOAD
REGISTER
2
RCOM
R2
R1
ROFS
RFB
PHASE ACCUMULATOR
PARALLEL
n
DELTA
PHASE
REGISTER
M
n
Σ
n
PHASE
REGISTER
CLOCK
FREQUENCY CONTROL
PHASE
TRUNCATION
16 BITS
LTC1597
16-BIT DAC
AGND
10 T0 21
24 TO 27
–
6
I0UT
n SINE ROM
16 DATA
LOOKUP
INPUTS
TABLE
+
7
LOWPASS
FILTER
LT1468
22
fO =
DGND
M • fC
2n
WR LD CLR
9 8 28
fC
CLR
Figure 9. This digital waveform generator produces a 1kHz sine wave with a second harmonic distortion of –94dB. The sampling frequency is 50kHz.
BIPOLAR MODE
DAC OUTPUT ERROR
DAC OUTPUT ERROR
UNIPOLAR MODE
DAC TRANSFER CURVE WITH VOS IN CURRENT-TO-VOLTAGE OP AMP
DAC TRANSFER CURVE WITH VOS IN CURRENT-TO-VOLTAGE
OP AMP AND REF INVERTING OP AMP
GAIN ERROR =
2VOSI-to-V + 4VOSINV
GAIN ERROR =
VOSI-to-V
DAC TRANSFER CURVE
WITH IDEAL OP AMP
OFFSET ERROR =
VOSI-to-V
0
NEGATIVE FULL-SCALE
ERROR = 2VOSI-to-V
65,535
32,768
CODE
BIPOLAR ZERO ERROR =
3VOSI-to-V + 2VOSINV
DAC TRANSFER CURVE
WITH IDEAL OP AMP
32,768
CODE
0
65,535
1720 G01
1720 G01
Figure 10. The effect of op amp offset on the LTC1597 gain and offset errors in unipolar mode (left) and bipolar mode (right); op amp offset
has virtually no effect on DAC linearity; it merely shifts the end points.
Table 1. Amplifiers recommended for use with the LTC1597, with relevant specifications
Amplifier Specifications
Amplifier
VOS
µV
IB
nA
AOL
V/mV
Voltage
Noise
nV/ Hz
LT1001
25
2
800
10
0.12
LT1097
50
0.35
1000
14
LT1112 (dual)
60
0.25
1500
LT1124 (dual)
70
20
LT1468
75
10
22
Gain Bandwidth
Product
MHz
Power
Dissipation
mW
0.25
0.8
46
0.008
0.2
0.7
11
14
0.008
0.16
0.75
10.5/op amp
4000
2.7
0.3
4.5
12.5
69/op amp
5000
5
0.6
22
90
117
Current Noise Slew Rate
pA/ Hz
V/µs
Linear Technology Magazine • February 1999
DESIGN FEATURES
causes distortion) by making the glitch
impulse both ultralow and uniform
with code.
Op Amp Selection
Considerations
A significant advantage of the
LTC1597 is the ability to choose the
I-to-V output op amp to optimize system accuracy, speed, power and cost.
Table 1 shows a sampling of op amps
and their relevant specifications for
this application.
The LTC1597 is designed to minimize the sensitivity of INL and DNL to
op amp offset; this sensitivity has
been greatly reduced compared to that
of competing multiplying DACs. Figure 10 summarizes the effects of op
amp offset for both modes of operation. Note that the bipolar LSB size is
twice its unipolar counterpart. As Figure 10 shows, op amp offset has a
minimal effect on DAC linearity; it
merely shifts the end points.
LT1611/LT1613, continued from page 13
The amplifier’s input bias current,
which flows through the feedback
resistor, adds to the output offset
voltage. The amplifier’s finite DC openloop gain also degrades accuracy. The
DAC gain error is inversely proportional to the open-loop gain and
feedback factor of the op amp. In
unipolar mode at full-scale the feedback factor is 0.5; for a 0.2LSB of gain
error (REF = 10V) at 16 bits, the openloop amplifier gain should be greater
than 650,000.
The op amp’s input voltage and
current noise also limit DC accuracy.
Noise effects accuracy similarly to
voltage and current offsets and adds
in an RMS fashion. As with any precision application, and with wide
bandwidth amplifiers in particular,
the noise bandwidth should be minimized with a filter on the output of the
op amp to maximize resolution.
VIN
80
+
SHDN
70
Wherever system requirements
demand true 16-bit accuracy over
temperature, the LTC1597 provides
the best solution. The LTC1597 has
outstanding 1LSB linearity over
temperature, ultralow glitch impulse,
on-chip 4-quadrant resistors, low
power consumption, asynchronous
clear and a versatile parallel
interface.Combined with the LT1468
op amp, the LTC1597 provides the
best in its class, 1.7µs settling time to
0.0015%, while maintaining superb
DC linearity specifications.
SW
D1
L1B
15µH
LT1611
68.1k
SHDN
VOUT
–10V/60mA
NFB
GND
10k
65
+
EFFICIENCY (%)
75
C1
22µF
Conclusion:
C2
0.22µF
L1A
15µH
VIN
3.6V–7V
85
Referring to Table 1, the LT1001
provides excellent DC precision, low
noise and low power dissipation. The
LT1468 provides the optimum solution for applications requiring DC
precision, low noise and fast 16-bit
settling.
C3
6.8µF
60
55
50
0
50
100 150 200 250
LOAD CURRENT (mA)
300
350
1611 TA02
Figure 15. 12V supply at L1A increases
efficiency to 81% and output current to
350mA.
85
VIN = 6.5V
EFFICIENCY (%)
75
VIN = 3.6V
(847) 956-0666
(803) 946-0362
1613 • TA01
(800) 441-2447
Figure 16. 4-Cell to –10V inverting converter delivers 75mA from a 4V input.
LT1611 4-Cell to –10V
Inverting Converter
A –10V low noise output can be generated in a similar manner as the –5V
circuit described above. Figure 16’s
circuit can deliver –10V at up to 60mA
from a 3.6V input. Efficiency, graphed
in Figure 17, reaches a high of 78%.
80
70
L1: SUMIDA CL562-150
C1: AVX TAJB226M010
C2: X7R CERAMIC
C3: AVX TAJA685M016
D1: MOTOROLA MBR0520
VIN = 5V
65
60
Conclusion
55
The flexibility of individually controlled
outputs in multiple-supply applications can make several LT1611/
LT1613 converters attractive compared to a multiple-output flyback
50
0
25
50
75
100
LOAD CURRENT (mA)
125
150
1611 TA02
Figure 17. 4-cell to –10V converter efficiency
Linear Technology Magazine • February 1999
design with one large switching regulator and a custom transformer.
Changing an output voltage on a
multiple output flyback requires
changing the transformer turns ratio,
hardly a simple task. Conversely,
individual control of each output, using the multiple LT1611/LT1613
approach, provides for complete control of each output voltage as well as
supply sequencing. The LT1611 and
LT1613 SOT-23 switchers provide
small, low noise solutions to power
generation needs in tight spaces.
23