AD7721
REV. A
–9–
Input Circuits
The purpose of antialiasing filters is to attenuate out of band
signals that would otherwise be mixed down into the signal
band. With traditional ADCs, high order filters using expensive
high tolerance passive components are often required to per-
form this function. Using oversampling, as employed on the
AD7721, this problem is considerably alleviated. Figure 4a
shows the digital filter frequency response. Due to the sampling
nature of the digital filter, the passband is repeated about the
operating clock frequency and at multiples of the clock fre-
quency. Out of band signals coincident with any of the filter
images are mixed down into the passband. Figure 4b shows the
frequency response of the antialias filter required to provide a
particular level of attenuation at the first image frequency. Fig-
ure 4c shows the frequency response of the antialias filter re-
quired to achieve the same level of attenuation with a traditional
ADC. The much smaller transition band can only be achieved
with a very high order filter.
f
CLK
2
f
CLK
3
f
CLK
0dB
a. Digital Filter Frequency Response
REQUIRED
ATTENUATION
0dB
OUTPUT DATA RATE
ANTIALIAS FILTER
RESPONSE
f
CLK
b. Frequency Response of Antialias Filter (AD7721)
REQUIRED
ATTENUATION
ANTIALIAS FILTER
RESPONSE
OUTPUT DATA RATE
0dB
c. Frequency Response of Antialias Filter (Traditional ADC)
Figure 4. Frequency Response of Antialiasing Filters
Figure 5 shows a simple antialiasing filter which can be used
with the AD7721. The –3 dB corner frequency (f
3dB
) of the
antialias filter is given by Equation 1, and the attenuation of the
filter is given by Equation 2. Attenuation at the first image
frequency is given by Equation 3.
f
3 dB
= 1/(2 π R
EXT
C
EXT
) Equation 1
Attenuation = 20 log
1/ 1+ f/f
3dB
()
2
Equation 2
Attenuation (First Image)=
20log 1/ 1+ 0.986 f
CLK
/ f
3dB
()
2
Equation 3
The choice of the filter corner frequency will depend on the
amount of rolloff which is acceptable in-band and the attenua-
tion which is required at the first image frequency. For example,
when f
CLK
= 15 MHz, R
EXT
= 50 , C
EXT
= 7.84 nF, the in-
band rolloff is 1 dB and the attenuation at the first image fre-
quency is 31.1 dB. Increasing the size of the external resistor
above 50 can cause increased distortion due to nonlinear
charging currents.
R
EXT
C
EXT
ANALOG
INPUT
R
EXT
C
EXT
AD7721
VIN1
VIN2
Figure 5. Simple RC Antialiasing Filter
Figure 6 shows a simple circuit that can be used to drive the
AD7721 in unipolar mode. The input of the AD7721 is sampled
by a 1.6 pF input capacitor. This creates glitches on the input
of the modulator. By placing the RC filter directly before the
AD7721, rather than before the operational amplifier, these
glitches are prevented from being fed back into the operational
amplifier and creating distortion. The resistor in this diagram,
as well as creating a pole for the antialias filter, also isolates the
storage capacitor from the operational amplifier which may
otherwise be unstable.
COMMON
MODE
VOLTAGE
ANALOG
INPUT
R
EXT
C
EXT
AD7721
VIN1
VIN2
Figure 6. Antialiasing Circuits
A suitable operational amplifier is the AD847 if a ±15 V power
supply is available. If only a +5 V power supply is available, the
AD820 can be used. This operational amplifier can be used with
input bandwidths up to 80 kHz. However, the slew rate of this
operational amplifier limits its performance to 80 kHz. Above
this frequency, the performance of the AD820 degrades.
For both filters, the capacitor C
EXT
should have a low tempera-
ture coefficient and should be linear to avoid distortion.
Polypropylene or polystyrene capacitors are suitable.
Offset and Gain Calibration
A calibration of offset and gain errors can be performed in both
serial and parallel modes by initiating a calibration cycle. During
this cycle, offset and gain registers in the filter are loaded with
values representing the dc offset of the analog modulator and a
modulator gain correction factor. In normal operation, the offset
register is subtracted from the digital filter output and this result
is then multiplied by the gain correction factor to obtain an
offset and gain corrected final result.
During the calibration cycle, in which the offset of the analog
modulator is evaluated, the inputs to the modulator are shorted
together internally. When the modulator and digital filter settle,
the average of 8 output results is calculated and stored in the
offset register. The gain of the modulator is determined by
AD7721
REV. A
–10–
Standby
The part can be put into a low power standby mode by writing
to the configuration register in parallel mode or by taking the
STBY pin high in serial mode. During Standby, the clock to
both the modulator and the digital filter is turned off and bias is
removed from all analog circuits. On coming out of standby
mode, the
DRDY pin remains high in parallel mode and low in
serial mode for 2080 clock cycles. When
DRDY changes state,
valid data is available at the interface. As soon as the part is
taken out of standby mode, a synchronization or calibration
cycle can be initiated.
DVAL
The DVAL pin or the DVAL/SYNC pin, when programmed as
a DVAL pin, is used to indicate that an overrange input signal
has resulted in invalid data at the ADC output. Small overloads
will result in DVAL going low and the output being clipped to
positive or negative full scale, depending on the sign of the
overload. As with all single bit DAC high order sigma-delta
modulators, large overloads on the inputs can cause the modula-
tor to go unstable. The modulator is designed to be stable with
signals within the input bandwidth that exceed full scale by
20%. When instability is detected by internal circuits, the
modulator is reset to a stable state and DVAL is held low for
2080 clock cycles. During this period, the output registers are
set to negative full scale. Whenever DVAL goes low,
DRDY will
continue to indicate that there is data to be read.
Varying the Master Clock Frequency
The AD7721 can be operated with clock frequencies less than
10 MHz. The sample rate, output word rate and cutoff fre-
quency of the FIR filters are directly proportional to the master
clock frequency. The analog input is sampled at a frequency of
2f
CLK
while the output word rate equals f
CLK
/32. For example,
reducing the clock frequency to 5 MHz leads to a sample fre-
quency of 10 MHz, an output word rate of 156.25 kHz and a
corner frequency of 76.4 kHz. The AD7721 can be operated
with clock frequencies down to 100 kHz.
Power Supply Sequencing
If separate analog and digital supplies are used, care must be
taken to ensure that both supplies remain within ±0.3 V of each
other both during normal operation and during power-up and
power-down to completely eliminate the possibility of latch-up.
If this cannot be assured, then the protection circuit shown in
Figure 7 is recommended. The 10 resistors may be required to
limit the current through the diodes if particularly fast edges are
expected on the supplies during power-up and power-down.
If only one supply is available, then DV
DD
must be connected to
the analog supply. Supply decoupling capacitors are still re-
quired as close as possible to both supply pins.
10nF
1mF
10nF
1mF
10V10V
IN4148
IN4148
AV
DD
DV
DD
AD7721
Figure 7. Powering-Up Protection Scheme
switching the positive input of the modulator to the reference
voltage and the negative input to AGND. Again, when the
modulator and digital filter settle, a gain correction factor is
calculated from the average of 8 output results and stored in the
gain register. After the calibration registers have been loaded
with new values, the inputs of the modulator are switched back
to the input pins. However, correct data is available at the inter-
face only after the modulator and filter have settled to the new
input values.
The whole calibration cycle is controlled by internal logic, and the
controller need only initiate the cycle. The calibration values
loaded into the registers only apply for the particular analog input
mode (bipolar/unipolar) selected when initiating the calibration
cycle. On changing to a different analog input mode, a new calibra-
tion must be performed. The duration of the calibration cycle is up
to 6720 clock cycles for the unipolar mode and up to 9024 clock
cycles for the bipolar mode. Until valid data is available at the
interface, the
DRDY pin remains high in parallel mode and low in
serial mode. Should the part see a rising edge on the
SYNC pin in
serial mode or on the DVAL/
SYNC pin (if programmed as a
SYNC pin), then the calibration cycle is discontinued and a syn-
chronization operation will be performed. Similarly, putting the
part into standby mode during the cycle will discontinue the cali-
bration cycle.
The calibration registers are static and retain their contents even
during standby. They need to be updated only if unacceptable
drifts in analog offsets or gain are expected. On power-up in
parallel mode, the offset and gain errors may contain incorrect
values and therefore a calibration must be performed at least
once after power-up. In serial mode, a calibration on power-up
is not mandatory if the CAL pin is grounded prior to power-up
as the calibration register will be reset to zero. Before initiating a
calibration routine, ensure that the supplies have settled and that
the voltage on the analog input pins is between the supply voltages.
Calibration does not affect the synchronization of the part.
Synchronization
Data is presented at the interface at 1/32 the CLK frequency. In
order that this data is presented to the interface at a known
point in time or to ensure that the data from more than one
device is a filtered and decimated result derived from the same
input samples, a synchronizing function has been provided. In
parallel mode, the DVAL/
SYNC pin must first be configured as
a
SYNC pin by writing to the control register. In serial mode,
there is a dedicated
SYNC pin. On the rising edge of the SYNC
pulse or the DVAL/
SYNC pulse, the digital filter is reset to a
known state. For 2080 clock cycles,
DRDY remains high in
parallel mode and low in serial mode. When
DRDY changes
state at the end of this period, valid data is available at the inter-
face. Synchronizing the part has no affect on the values in the
calibration register.
SYNC is latched internally on the rising edge of DCLK which is
a delayed version of the clock on the CLK pin. Should
SYNC
go high coincidentally with DCLK, there is a potential uncer-
tainty of one clock cycle in the start of the synchronization cycle.
To avoid this,
SYNC should be taken high after the falling edge
of the clock on the CLK pin and before the rising edge of this
clock.
AD7721
REV. A
–11–
a linear phase response. This is very difficult to achieve with
analog filters.
Analog filters, however, can remove noise superimposed on the
signal before it reaches the ADC. Digital filtering cannot do this
and noise peaks riding on signals, near full-scale, have the po-
tential to overload the analog modulator even though the aver-
age value of the signal is within limits.
0.0
–50.0
0.5f
CLK
–100.0
–150.0
GAIN – dB
0.1f
CLK
0.2f
CLK
0.3f
CLK
0.4f
CLK
FREQUENCY
0.0f
CLK
Figure 9a. 128 Tap FIR Filter Frequency Response
0.0
–50.0
–100.0
–150.0
1.0f
CLK
/320.8f
CLK
/320.6f
CLK
/320.4f
CLK
/320.2f
CLK
/320.0f
CLK
/32
FREQUENCY
GAIN – dB
Figure 9b. 83 Tap FIR Filter Frequency Response
SERIAL INTERFACE
The AD7721’s serial communication port allows easy inter-
facing to industry-standard microprocessors, microcontrollers
and digital signal processors. The AD7721 is operated in self-
clocking mode, the AD7721 providing the serial clock. The
RFS signal is also provided by the AD7721 by tying RFS to
DRDY.
Figure 10 shows the timing diagram for reading from the
AD7721.
DRDY goes high to indicate that a conversion has
been completed.
DRDY remains high for one internal clock
(15 MHz) cycle and then goes low for the next 31 clock cycles.
New data is loaded into the output shift register on the rising
edge of
DRDY. When DRDY goes low, the data is accessed
from the AD7721. Although the AD7721 has a 12-bit digital
output in the parallel mode, sixteen bits of data are available for
transmission in the serial mode, starting with the MSB. Serial
data is clocked out of the device on the rising edge of SCLK
and is valid on the falling edge of SCLK.
CIRCUIT DESCRIPTION
Sigma-Delta ADC
The AD7721 ADC employs a sigma-delta conversion technique
that converts the analog input into a digital pulse train.
Due to the high oversampling rate, which spreads the quantiza-
tion noise from 0 to f
CLK
/2, the noise energy which is contained
in the band of interest is reduced (Figure 8a). To reduce the
quantization noise further, a high order modulator is employed
to shape the noise spectrum, so that most of the noise energy is
shifted out of the band of interest (Figure 8b).
The digital filter that follows the modulator removes the large
out of band quantization noise (Figure 8c), while converting the
digital pulse train into parallel 12 bit wide binary data or serial
16 bit wide binary data.
BAND OF
INTEREST
QUANTIZATION NOISE
f
CLK
/2
a.
BAND OF
INTEREST
NOISE
SHAPING
f
CLK
/2
b.
BAND OF
INTEREST
f
CLK
/2
DIGITAL FILTER CUTOFF FREQUENCY
WHICH EQUALS 152.8kHz (10MHz) OR
229.2kHz (15MHz)
c.
Figure 8. Sigma-Delta ADC
Digital Filter
The digital filter that follows the modulator removes the large
out of band quantization noise, while converting the one bit
digital pulse train into 12-bit or 16-bit wide binary data. The
digital filter also reduces the data rate from f
CLK
at the input of
the filter to f
CLK
/32 at the output of the filter. The output data
rate is a little over twice the signal bandwidth which guarantees
that there is no loss of data in the signal band.
The AD7721 employs 2 FIR filters in series. The first filter is a
128 tap filter that samples the output of the modulator at f
CLK
.
The second filter is an 83 tap half-band filter that samples the
output of the first filter at f
CLK
/16 and decimates by 2. The
frequency response of the 2 filters is shown in Figure 9.
Digital filtering has certain advantages over analog filtering.
First, since digital filtering occurs after the A/D conversion, it
can remove noise injected during the conversion process. Ana-
log filtering cannot do this. Second, the digital filter combines
low passband ripple with a steep roll off, while also maintaining

AD7721ARZ

Mfr. #:
Manufacturer:
Analog Devices Inc.
Description:
Analog to Digital Converters - ADC CMOS 12-/16-Bit 312.5kHz/468.75kHz
Lifecycle:
New from this manufacturer.
Delivery:
DHL FedEx Ups TNT EMS
Payment:
T/T Paypal Visa MoneyGram Western Union

Products related to this Datasheet