AD9853 Analog Devices, AD9853 Datasheet - Page 16

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

AD9853
receive end. The values actually programmed on the serial con-
trol bus are “K” and “t,” which will define N as shown in the
above code-word structure equation. As can be seen from the
code-word structure equation, two check bytes are required to
correct each byte error. Setting t = 0 and K > 0 will bypass the
Reed-Solomon encoding process.
Since Reed-Solomon works on bytes of information and not
bits, a single byte error can be as small as one inverted bit out of
a byte, or as large as eight inverted bits of one byte; in either
instance the result is one byte error. For example, if the value
“t” is specified as 5, the R-S FEC could be correcting as many
as 40, or as few as 05, erroneous bits, but those errors must be
contained in 5 message bytes. If the errors are spread among
more than five bytes, the message will not be fully error corrected.
When using the R-S encoder, the message data needs to be
partitioned or “gapped” with “don’t care” data for the time
duration of the check bytes as shown in the timing diagram of
Figure 26. During the intervals between message data, the de-
vice ignores data at the input.
The position of the R-S encoder in the coding data path can be
switched with the randomizer by exercising Register 1, Bit D3,
via the serial control bus.
RANDOMIZER FUNCTION
The next stage in the modulation chain is the randomizing or
“scrambling” stage. Randomizing is necessary due to the fact
that impairments in digital transmission can be a function of the
statistics of the digital source. Receiver symbol synchronization
is more easily maintained if the input sequence appears random
or equiprobable. Long strings of 0s or 1s can cause a bit or
symbol synchronizer to lose synchronization. If there are repeti-
tive patterns in the data, discrete spurs can be produced, caus-
ing interchannel interference. In modulation schemes relying on
suppressed carrier transmission, nonrandom data can increase
the carrier feedthrough. Using a randomizer effectively “whitens”
the data.
The technique used in the AD9853 to randomize the data is to
perform a modulo 2 logic addition of the data with a pseudo-
random sequence. The pseudorandom sequence is generated by
a shift register of length m with an exclusive OR combination of
the nth bit and the last (mth) bit of the shift register that is fed
back to the shift register input. By choosing the appropriate
feedback point, a maximal length sequence is generated. The
maximal length sequence will repeat after every 2
but appears effectively “random” at the output. The criterion
for maximal length is that the polynomial 1 + x
ducible and prime over the Galois field. The AD9853 contains
the following two polynomial configurations in hardware:
The seed value is fully programmable for both configurations.
The seed value is reset prior to each burst and is used to calcu-
late the randomizer bit, which is combined in an exclusive XOR
with the first bit of data from each burst. The first bit of data in
a burst is the MSB of the first symbol following the last symbol
of the internally generated preamble.
x
x
15
6
+ x
+ x
5
14
+1
+1 :MCNS (DOCSIS) compatible.
:DAVIC/DVB compatible.
n
+ x
m
clock cycles,
m
be irre-
–16–
PREAMBLE INSERTION BLOCK
As shown in the block diagram of the AD9853, the circuit in-
cludes a programmable preamble insertion register. This register
is 96 bits long and is transmitted upon receiving the T
signal. It is transmitted without being Reed-Solomon encoded
or scrambled. Ramp-up data, to allow for receiver synchroniza-
tion, is included as the first bits in the preamble, followed by
user burst profile or channel equalization information. The first
bit of R-S encoded and scrambled information data is timed to
immediately follow the last bit of preamble data.
For most modulation modes, a minimum preamble is required.
This minimum is one symbol, two bits for DQPSK or four bits
for either 16-QAM or D16-QAM. No preamble is required for
either FSK or QPSK.
In conformance with DAVIC/DVB standards, the preamble is
not differentially coded in DQPSK mode. However the pre-
amble data can be differentially precoded when loaded into the
preamble register. The last symbol of the preamble is used as
the reference point for the first internal differentially coded
symbol so the preamble and data will effectively be coded differ-
entially. In the D16-QAM mode, the preamble is always differ-
entially coded internally.
MODULATION ENCODER
The preamble, followed by the encoded and scrambled data is
then modulation encoded according to the selected modulation
format. The available modulation formats are FSK, QPSK,
DQPSK, 16-QAM and D16-QAM. The corresponding symbol
constellations support the interactive HFC cable specifications
called out by MCNS (DOCSIS), 802.14 and DAVIC/DVB.
The data arrives at the modulation encoder at the input bit rate
and is demultiplexed as modulation encoded symbols into sepa-
rate I and Q paths. For QPSK and DQPSK, the symbol rate is
one-half of the bit rate and each symbol is comprised of two
bits. For 16-QAM and D16-QAM, the symbol rate is one-
fourth the bit rate and each symbol is comprised of four bits. In
the FSK mode, although the 1 and 0 data is entered into the
serial data input, it effectively bypasses the encoding, scrambling
and modulation paths. The FSK data is directly routed to the
direct digital synthesizer (DDS) where it is used to switch the
DDS between two stored tuning words (F0:F1) to achieve FSK
modulation in a phase-continuous manner. By holding the input
at either 1 or 0, a single frequency continuous wave can be
output for system test or CW transmission purposes.
Differential encoding of data is frequently used to overcome
phase ambiguity error or a “false lock” condition that can be
introduced in carrier-recovery circuits used to demodulate the
signal. In straight QPSK and 16-QAM, the phase of the re-
ceived signal is compared to that of a “recovered carrier” of
known phase to demodulate the signal in a coherent manner. If
the phase of the recovered carrier is in error, then demodulation
will be in error. Differential encoding of data at the transmit end
eliminates the need for absolute phase coherency of the recov-
ered carrier at the receive end. If a coherent reference generated
by a phase lock loop experiences a phase inversion while de-
modulating in a differentially coded format, the errors would be
limited to the symbol during which the inversion occurred and
the following symbol. Differential coding uses the phase of the
“previously transmitted symbol” as a reference point to compare
to the current symbol. The change in phase from one symbol to
X
ENABLE
REV. C