CMOS 8-bit A/D converters
Large values of source resistance where an input bypass capacitor
is not used will not cause errors as the input currents settle out prior
to the comparison time. If a low pass filter is required in the system,
use a low valued series resistor (< 1 kΩ) for a passive RC section or
add an op amp active filter (low pass). For applications with source
resistances at or below 1 kΩ, a 0.1 µF bypass capacitor at the inputs
will prevent pickup due to series lead inductance or a long wire. A
100 Ω series resistor can be used to isolate this capacitor (both the
resistor and capacitor should be placed out of the feedback loop)
from the output of the op amp, if used.
Analog Differential Voltage Inputs and
These A/D converters have additional flexibility due to the analog
differential voltage input. The VIN(–) input (Pin 7) can be used to
subtract a fixed voltage from the input reading (tare correction). This
is also useful in a 4/20 mA current loop conversion. Common-mode
noise can also be reduced by the use of the differential input.
The time interval between sampling VIN(+) and VIN(–) is 4.5 clock
periods. The maximum error due to this time difference is given by:
V(max) = (VP) (2fCM) (4.5/fCLK),
V = error voltage due to sampling delay
VP = peak value of common-mode voltage
fCM = common mode frequency
For example, with a 60 Hz common-mode frequency, fcm, and a
1 MHz A/D clock, fCLK, keeping this error to 1/4 LSB (about 5 mV)
would allow a common-mode voltage, VP, which is given by:
(5 x 10*3) (104)
VP + (6.28) (60) (4.5) + 2.95V
The allowed range of analog input voltages usually places more
severe restrictions on input common-mode voltage levels than this,
An analog input span less than the full 5 V capability of the device,
together with a relatively large zero offset, can be easily handled by
use of the differential input. (See Reference Voltage Span Adjust).
Noise and Stray Pickup
The leads of the analog inputs (Pins 6 and 7) should be kept as
short as possible to minimize input noise coupling and stray signal
pick-up. Both EMI and undesired digital signal coupling to these
inputs can cause system errors. The source resistance for these
inputs should generally be below 5 kΩ to help avoid undesired noise
pickup. Input bypass capacitors at the analog inputs can create
errors as described previously. Full scale adjustment with any input
bypass capacitors in place will eliminate these errors.
For application flexibility, these A/D converters have been designed
to accommodate fixed reference voltages of 5V to Pin 20 or 2.5 V to
Pin 9, or an adjusted reference voltage at Pin 9. The reference can
be set by forcing it at VREF/2 input, or can be determined by the
supply voltage (Pin 20). Figure 6 indicates how this is accomplished.
Reference Voltage Span Adjust
Note that the Pin 9 (VREF/2) voltage is either 1/2 the voltage applied
to the VCC supply pin, or is equal to the voltage which is externally
forced at the VREF/2 pin. In addition to allowing for flexible
references and full span voltages, this also allows for a ratiometric
voltage reference. The internal gain of the VREF/2 input is 2, making
the full-scale differential input voltage twice the voltage at Pin 9.
For example, a dynamic voltage range of the analog input voltage
that extends from 0 to 4 V gives a span of 4 V (4–0), so the VREF/2
voltage can be made equal to 2 V (half of the 4 V span) and full
scale output would correspond to 4 V at the input.
On the other hand, if the dynamic input voltage had a range of
0.5 to 3.5 V, the span or dynamic input range is 3 V (3.5–0.5). To
encode this 3 V span with 0.5 V yielding a code of zero, the
minimum expected input (0.5 V, in this case) is applied to the VIN(–)
pin to account for the offset, and the VREF/2 pin is set to 1/2 the 3 V
span, or 1.5 V. The A/D converter will now encode the VIN(+) signal
between 0.5 and 3.5 V with 0.5 V at the input corresponding to a
code of zero and 3.5 V at the input producing a full scale output
code. The full 8 bits of resolution are thus applied over this reduced
input voltage range. The required connections are shown in
These converters can be operated in two modes:
1) absolute mode
2) ratiometric mode
In absolute mode applications, both the initial accuracy and the
temperature stability of the reference voltage are important factors in
the accuracy of the conversion. For VREF/2 voltages of 2.5 V, initial
errors of ±10 mV will cause conversion errors of ±1 LSB due to the
gain of 2 at the VREF/2 input. In reduced span applications, the initial
value and stability of the VREF/2 input voltage become even more
important as the same error is a larger percentage of the VREF/2
nominal value. See Figure 8.
In ratiometric converter applications, the magnitude of the reference
voltage is a factor in both the output of the source transducer and
the output of the A/D converter, and, therefore, cancels out in the
final digital code. See Figure 9.
Generally, the reference voltage will require an initial adjustment.
Errors due to an improper reference voltage value appear as
full-scale errors in the A/D transfer function.
ERRORS AND INPUT SPAN ADJUSTMENTS
There are many sources of error in any data converter, some of
which can be adjusted out. Inherent errors, such as relative
accuracy, cannot be eliminated, but such errors as full-scale and
zero scale offset errors can be eliminated quite easily. See Figure 7.
Zero Scale Error
Zero scale error of an A/D is the difference of potential between the
ideal 1/2 LSB value (9.8 mV for VREF/2=2.500 V) and that input
voltage which just causes an output transition from code 0000 0000
to a code of 0000 0001.
If the minimum input value is not ground potential, a zero offset can
be made. The converter can be made to output a digital code of
0000 0000 for the minimum expected input voltage by biasing the
VIN(–) input to that minimum value expected at the VIN(–) input to
that minimum value expected at the VIN(+) input. This uses the
differential mode of the converter. Any offset adjustment should be
done prior to full scale adjustment.
2002 Oct 17