Corporation
SIGNAL PROCESSING EXCELLENCE
131
2
-
1
(MSB)
0
0
1
1
2
-
2
0
1
0
1
Output
0
1/4 Full-Scale
1/2 Full-Scale
3/4 Full-Scale
Table 1. Contribution of the two MSB's
VREF
VDD
470
DIGITAL
RFEEDBACK
IO1
+
-
IO2
GND
ROS
A
VOUT
SP7516
200
400
Figure 2. Unipolar Operation
The most common technique for building a D/A
converter of n bits is to use n switches to turn n current
or voltage sources on or off. The n switches and n
sources are designed so that each switch or bit contrib-
utes twice as much to the D/A converter’s output as the
preceding bit. This technique is commonly known as
binary weighting and allows an n-bit converter to
generate 2
n
output levels by turning on the proper
combination of bits.
In such a binary-weighted converter, the switch
with the smallest contribution (the LSB) accounts
for only 2
-n
of the converter’s full-scale value.
Similarly, the switch with the largest contribution
(the MSB) accounts for 2
-1
or half of the converter’s
full-scale output. Thus it is easy to see that a given
percent change in the MSB will have a greater
effect on the converter’s output than would a
similar percent change in the LSB. For example, a
1% change in the LSB of a 10 bit converter would
only affect the output by 0.001% of full-scale. A
1% change in the MSB of the same converter
would affect the output by 0.5% of FSR.
In order to overcome the problem which results from
the large weighting of the MSB, the two MSB’s can
be decoded to three equally weighted sources.
Table
1
shows that all combinations of the two MSB’s of a
converter result in four output levels. So by replacing
the two MSB’s with three bits equally weighted at 1/
4 full-scale and decoding the two MSB digital inputs
into three lines which drive the equally weighted bits,
the same functional performance can be obtained.
Thus by replacing the two MSB switches of a conven-
tional converter with three switches properly de-
coded, the contribution of any switch is reduced from
1/2 to 1/4. This reduction in sensitivity also reduces the
accuracy required of any switch for a given overall
converter accuracy.
With the decoded converter described above, a 1%
change in any of the converter’s switches will affect
the output by no more than 0.25% of full-scale as
compared to 0.5% for a conventional converter. In
other words the conventional D/A converter can be
made less sensitive to the quality of its individual bits
by decoding.
In the
SP7516/HS3160
the first four MSB’s are
decoded into 16 levels which drive 15 equally weighted
current sources. The sensitivity of each switch on the
output is reduced by a factor of 8. Each of the 15
sources contributes 6.25% output change rather than
an MSB change of 50% for the common approach.
DIGITAL
RFEEDBACK
IO1
+
-
IO2
GND
ROS1
A
VOUT
1
+
-
A2
ROS2
VOUT1
A1, A2, OP-07
4K
4K
ROS2
R
200
VREF
VDD
470
400
SP7516
Figure 3. Bipolar Operation
TRANSFER FUNCTION (N=16)
BINARY INPUT UNIPOLAR OUTPUT BIPOLAR OUTPUT
111...111
–V
REF
(1 - 2
–N
)
–V
REF
(1/2 + 2
–N
)
–V
REF
/2
–V
REF
(1/2 – 2
–N
)
–V
REF
(2
(N – 1)
)
0
–V
REF
(1 – 2
–(N – 1)
)
–V
REF
(2
–(N – 1)
)
0
V
REF
(2
–(N – 1)
)
V
REF
(1 – 2
–(N – 1)
)
V
REF
100...001
100...000
011...111
000…001
000...000
Table 2. Transfer Function