Xcell Journal: The authoritative journal for programmable ... - Xilinx
Xcell Journal: The authoritative journal for programmable ... - Xilinx
Xcell Journal: The authoritative journal for programmable ... - Xilinx
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Figure 4 – J.83 Annex A/C modulator with scatter plot<br />
IP Simulation in Simulink<br />
It takes a lot of time to simulate and test<br />
the functionality of a complex system. You<br />
can use the same J.83 circuit built in<br />
System Generator <strong>for</strong> simulation and verification,<br />
as well as the FPGA implementation.<br />
Within the same environment, using<br />
Simulink <strong>for</strong> simulation, the design is stimulated<br />
with MPEG transport packets and<br />
the appropriate QAM, reset, synchronization,<br />
and other control inputs.<br />
As shown in Figure 4, this stimulus is<br />
shown in the block labeled “Stimuli.” <strong>The</strong><br />
source, inter-packet gap, and burst nature<br />
of the MPEG transport packet may be<br />
chosen at random at the top level, allowing<br />
a test of the full suite of possibilities.<br />
Figure 4 also shows a discrete time scatter<br />
plot of the output of the baseband section<br />
of the modulator.<br />
Simulation of this complex system in an<br />
HDL simulator <strong>for</strong> a meaningful number of<br />
clock cycles (such that several frames of data<br />
may be processed) imposes a huge penalty in<br />
the time taken to complete a simulation.<br />
This makes it an impractical choice, but<br />
sometimes it is the only option when the<br />
design source is in an HDL <strong>for</strong>mat.<br />
This simulation time is drastically<br />
reduced when simulating the model in<br />
Simulink. What might take days to simulate<br />
in a gate-level simulator could be<br />
accomplished in a matter of hours. This<br />
savings in time is highly valuable – not only<br />
do you benefit from superior simulation<br />
speed in Simulink but you<br />
also reap the benefits of a shortened<br />
design cycle, allowing <strong>for</strong><br />
overall rapid IP delivery.<br />
Single and Multi-Channel Designs<br />
<strong>The</strong> modulator is constructed out<br />
of two primary footprints or granularity:<br />
a single-channel implementation<br />
and a four-channel<br />
implementation. A block diagram<br />
of the four-channel granularity<br />
Annex B and A/C are shown in<br />
Figure 5 and Figure 6, respectively.<br />
Each instance of the singlechannel<br />
footprint provides <strong>for</strong><br />
exactly one independent channel;<br />
the four-channel footprint, however,<br />
is optimized to efficiently support<br />
four channels at a time, using resourcesharing<br />
techniques. You select the granularity,<br />
and with that selection, make a<br />
trade-off between resource utilization and<br />
individual channel control.<br />
1<br />
1<br />
1<br />
1<br />
8<br />
8<br />
MPEG<br />
Framer<br />
MPEG<br />
Framer<br />
MPEG<br />
Framer<br />
1<br />
1<br />
Physical Interface<br />
Clock<br />
Mgmt<br />
4x1b<br />
to<br />
4x7b<br />
&<br />
FIFO<br />
Ext. Memory<br />
Memory<br />
Controller<br />
28<br />
1 ch<br />
RS 28 Inter- 28 Rando- 28<br />
1 ch<br />
RS<br />
1 ch<br />
RS<br />
leavermizer DIGITAL SIGNAL PROCESSING<br />
<strong>The</strong> trade-off is essentially in the area<br />
(resource) utilization; the optimized fourchannel<br />
group solution results in a very<br />
efficient and compact design requiring<br />
fewer FPGA resources. However, it imposes<br />
the restriction that the four channels<br />
must share the same controls. <strong>The</strong> singlechannel<br />
solution imposes no such restriction;<br />
the trade-off here is the linearly<br />
increasing FPGA resources used, which is<br />
directly proportional to the number of<br />
channels required.<br />
Multi-channel modulators are automatically<br />
constructed through the use of multiple<br />
copies (also referred to as groups) of<br />
the single- or four-channel implementation.<br />
For example, a four-channel modulator<br />
may be constructed with four copies of<br />
single-channel granularity or a single copy<br />
of the optimized 4-channel granularity<br />
design. Similarly, a 12-channel modulator<br />
may comprise 12 copies of the single-channel<br />
granularity design or 3 copies of the<br />
optimized 4-channel design.<br />
<strong>The</strong> ease of use is evident in that the<br />
Winter 2004 <strong>Xcell</strong> <strong>Journal</strong> 53<br />
1 ch<br />
RS<br />
Controls<br />
(Frame, RST, Configuration &<br />
Channel Controls)<br />
Byte<br />
m<br />
Bits to 4 ch Conv.<br />
to<br />
4-ch Byte RS Inter-<br />
m<br />
4 4 32 32 32 Symbol<br />
Rando- & Enc. leaver<br />
1 m<br />
mizer Sync (204, I=12,<br />
2<br />
FIFO 188) J=17<br />
3 m<br />
4<br />
Clock<br />
Mgmt<br />
4-ch<br />
Diff.<br />
Encode<br />
7<br />
7<br />
TCM<br />
TCM<br />
Figure 5 – J.83 Annex B four-channel granularity design<br />
Controls (Frame, RST,<br />
Configuration, Channel)<br />
TDM<br />
Bus<br />
Symbol<br />
Mapper<br />
m = 4, 5, 6, 7, 8<br />
<strong>for</strong> 16, 32, 64,<br />
128, 256-QAM<br />
Figure 6 – J.83 Annex A/C four-channel granularity design<br />
I<br />
Q<br />
I1<br />
Q1<br />
I4<br />
Q4<br />
Prog. Controls<br />
Interleaver Select<br />
QAM Select<br />
Level Select<br />
TDM<br />
Bus<br />
I<br />
Q<br />
Async<br />
FIFO<br />
Async<br />
FIFO &<br />
8 ch<br />
RRC<br />
� = 0.18<br />
� = 0.12<br />
TDM<br />
Bus<br />
I<br />
Q<br />
I1<br />
Q1<br />
I4<br />
Q4<br />
2 ch<br />
RRC<br />
Filter<br />
(a=0.15<br />
or<br />
a=0.13)<br />
TDM<br />
Bus<br />
Q<br />
Prog. Controls<br />
QAM Select<br />
Annex A or C<br />
I