AD9853 Analog Devices, AD9853 Datasheet - Page 22

• Manufacturer Part Number: AD9853
• Description: Programmable Digital OPSK/16-QAM Modulator
• Manufacturer: Analog Devices

AD9853
The goal of interpolation is to up-sample the baseband informa-
tion to the system clock rate and to suppress aliases in the pass-
band. The system clock rate is the sample rate of the sine and
cosine signal carriers generated by the DDS in the quadrature
modulator stage. Alias suppression is accomplished by the CIC
filters as described previously. For timing synchronization, the
overall interpolation rate must be set such that the bit rate of the
baseband signal be an even integer factor of the system clock
rate. The importance of the relationship between the data and
system clock rates can not be overstressed. It is restated here for
clarity:
The SYSTEM CLOCK RATE must be an EVEN INTEGER
MULTIPLE of the DATA BIT RATE.
Following is a design example that demonstrates the principles
outlined above.
System Requirements:
• Baseband Bit Rate
• Carrier Frequency
• Modulation Scheme
• System Power
It should be noted that with a 3.3 V power supply, the maxi-
mum system clock rate of the AD9853 is 126 MHz. This sets an
upper bound on the system clock.
The first consideration is to make sure that the required carrier
frequency is within the AD9853’s output frequency range. The
carrier frequency should be
given carrier frequency requirement of 49 MHz means that a
minimum system clock rate of 122.5 MHz is required; a value
within the range of the AD9853’s 126 MHz capability.
We must next ensure that the system clock rate is an even inte-
ger multiple of the input bit rate. Dividing the system clock rate
(122.5 MHz) by the data rate (1.024 Mbps) yields 119.63.
Obviously this is not an integer, so we must select the nearest
even integer value (in this case, 120) as the data rate multiplier.
Thus, a system clock rate of 122.88 MHz is required (120
1.024 Mbps). With 6 REFCLK engaged, the reference clock
input will be 1/6th of the system clock rate, or 20.48 MHz.
Finally, the two interpolator rates must be determined. Since
the FIR filter and interpolator stages will be operating on 16-QAM
symbols, the data rate must be converted from bits/second to
1
ENCODER
I & Q
Q
I
3
3
SYMBOL
CLOCK
1.024 Mb/s
49 MHz
16-QAM
3.3 V
41 - TAP
41 - TAP
FIR
FIR
40% of the system clock rate. The
Figure 36. Block Diagram of AD9853 Data Path and Clock Stages
4
12
2
2
12
MUX
MUX
M
M = 3...31
13
13
POLATOR
POLATOR
INTER-
INTER-
#1
#1
N = 2...63
28
28
SCALER
SCALER
–22–
N
13
symbols/second (baud). Each 16-QAM symbol is composed
of four serial data bits. Therefore, the baud rate at the input to
the FIR filter is 1.024 Mbps/4 = 256k baud. The FIR pulse
shaping filters up-sample by a factor of 4. This fixes the FIR
sample clock at 256k baud 4, or 1.024 MSPS. With the FIR
sampling at a 1.024 MSPS rate, and a previously determined
system clock rate of 122.88 MHz, the interpolators must up-
sample by a factor of 120 (122.88/1.024 = 120).
Rule of Thumb: divide the interpolating burden as equally as
possible among the two interpolators.
Since the required rate change ratio is 120, select a value of 10
for interpolator #1 and 12 for interpolator #2 (10 12 = 120).
This satisfies the requirements for the two programmable inter-
polator stages.
Thus far we have established the rate change ratios for the inter-
polators. However, there is an additional consideration. By
default, the interpolators have an intrinsic gain (or loss) that is
dependent on the selected interpolation rate. Since there is the
potential to have overall CIC gains of greater than unity, care
must be taken to avoid the occurrence of overflow in the
interpolators.
Interpolator Scaling
Proper signal processing in the AD9853 depends on data propa-
gating through the pulse-shaping filter and interpolator stages
with as flat a baseband response as possible. In addition to the
frequency response issue, it is also necessary to ensure that the
numerical data propagating through the interpolators does not
result in an overflow condition.
As mentioned earlier, the interpolators are implemented using a
CIC filter. In the AD9853, the CIC filter is designed using
fixed-point processing and two cascaded CIC filter sections
(Interpolator #1 and Interpolator #2). It is important to under-
stand that in a CIC filter, the integration portion of the circuit
will require the accumulation of values based on the rate change
factor, R. This means that the size of the data word grows in a
manner dependent on the choice of R. In the case of Interpola-
tor #1, the circuit is designed around a maximum R of 32 and
this results in an output register width of 28 bits. The design of
Interpolator #2 requires an output register width of 25 bits.
13
POLATOR
POLATOR
INTER-
INTER-
#2
#2
SYSTEM
CLOCK
25
25
SCALER
SCALER
COS(
SIN(
10
10
C
C
DDS
)
)
10
10
20
20
10
INVERSE
FILTER
SINC
DAC
REV. C